# **The Role of Saliva Cortisol Measurement in Health and Disease**

# **Edited By**

# **Margareta Kristenson**

*Professor of Social Medicine and Public Health Science Department of Medical and Health Sciences, Linköping University Sweden*

# **Peter Garvin**

*Post-doc at the Department of Medical and Health Sciences, Linköping University Sweden*

# **Ulf Lundberg**

*Professor Emeritus of Biological Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University Sweden*

© 201 by the Editor / Authors. Chapters in this eBook are Open Access and distributed under the Creative Commons Attribution (CC BY 4.0) license, which allows users to download, copy and build upon published chapters, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications. The book taken as a whole is © 201 Bentham Science Publishers under the terms and conditions of the Creative Commons license CC BY-NC-ND.

# **CONTENTS**


# **FOREWORD**

The ability to assess the biomarker cortisol as an index of Hypothalamic Pituitary Adrenal (HPA) axis activity has provided enormous insights about the relations between psychosocial and physical environmental characteristics and the human stress response. Having a reliable and valid index of stress has also proven invaluable in examining the role of stress in mental and physical health. The ability to assess cortisol in saliva samples has opened up vast new areas of scientific exploration, particularly at the borders of social science with public health and medicine. The collection of salivary cortisol is a relatively unobtrusive procedure that can then be analyzed at low cost. It is remarkable that this book on salivary cortisol as a human stress biomarker is authored by a group of scientists from Scandinavia since, along with some prestigious scientists in Germany, much of the pioneering work on neuroendocrine biomarkers of stress has emanated from Sweden and Norway.

This volume arrives at an opportune moment with exponential growth in the use of salivary cortisol as a biomarker of stress coupled with remarkable interdisciplinary research on the borders of the social sciences and health. Thus we can ask two key questions about salivary cortisol and scientific research: What have we learned about the utility of salivary cortisol as a biomarker of stress? How should we use this tool to assess emerging scientific questions? Reading this book provides in depth answers to these questions.

This book provides a balanced, careful, and thorough review of literally hundreds of studies relating salivary cortisol indices to sociodemographic background characteristics of individuals such as socioeconomic status and gender, psychosocial working conditions (*e.g.,* job control), perceived stress, and psychological resources such as social support. Studies of associations between salivary cortisol and biomarkers of cardiovascular and immune function as well as sleep processes are reviewed along with work on the relations between salivary cortisol and major health outcomes (*e.g.* cardiovascular disease, breast cancer) as well as mental health (*e.g.,* depression). The authors identified all potentially relevant articles then applied systematic conceptual and methodological inclusion criteria to filter out irrelevant or sloppily conducted studies. They then systematically analyzed the remaining studies, tabling results in a manner that is easy to read and understand. Each table is organized by methods of saliva collection according to variables of interest (*e.g.,* sociodemographic background, disease outcomes). The results of hundreds of studies are then discussed within each topic area taking into account the patterns of findings and implications for measurement and theory. As the reader will be able to see herein, the quality of data and the clarity of conclusions about salivary cortisol as a stress biomarker vary considerably because of measurement protocols, statistical and methodological controls, and important conceptual issues having to do with static versus dynamic measures and inter versus intra person comparisons.

The authors have done all of us who are interested in the interplay among environment, personal background, stress, and disease, a marvelous favor. They have extensively and accurately reviewed what we know about salivary cortisol as an index of human stress. The authors have provided direction as well for how future research on salivary cortisol as a biomarker of stress should proceed.

> *Gary W. Evans*  Cornell University USA

# **PREFACE**

This book is based on a combination of fascination and frustration; fascination on the wish to use saliva cortisol measurement because of its many advantages but frustrations over opposing results in the literature. Several discussions at different meetings led to the development of a network of researchers from Sweden, funded by the Swedish National Research Council. This network was soon expanded to also include colleagues from Norway and Denmark. Thus, this was a Scandinavian network working on measurement of Cortisol and the name ScanCort was taken.

The main aim of the group was to try to understand the results from different studies on saliva cortisol measurement and thereby better understand how and when saliva cortisol assessment best could be made. A hypothesis was that, seemingly, divergent findings could be effects of differences in the theoretic assumptions made and methods used.

This led over to a decision to perform a literature review focusing on if the many different ways of evaluating the levels and dynamics of salivary cortisol especially with regard to time points of assessment and analyses of data affect the interpretation of cortisol measurement in various contexts.

The literature review was, of course, more work than expected but it was also a very exciting learning experience! We are grateful for the economic support given by Swedish National Research Council. We thank Gary W. Evans for being insightful, constructive and generous by reviewing all chapters and Lorna O'Brien for skillful language control. As editors we thank all colleagues in the ScanCort group for an unforgettable time together, for stimulating discussions and hard work. My specific thanks goes to my two co-editors professor Ulf Lundberg and PhD Peter Garvin for their work, enthusiasm and friendship.

We do hope that this book will be of use for all those who are involved in the challenging but fascinating field of stress research and want to use saliva cortisol measurement. We do believe that this can be a useful biomarker in many settings, if caution is taken in the choice of methods used.

> *Margareta Kristenson*  Linköping University Sweden

# **List of Contributors**

**Anette Harris, Ph.D.** Researcher at Uni Health and Department of Psychology, Bergen University, Bergen, Norway.

**Anne Helene Garde, Ph.D.** Senior researcher at the National Research Centre for the Working Environment, Copenhagen, Denmark.

**Åse Marie Hansen, Ph.D.** Senior researcher at the National Research Centre for the Working Environment, Copenhagen, Denmark.

**Berndt Karlson, M.D., Ph.D.** Doctor in Occupational and Environmental Medicine,at the Department of Public Health and Clinical Medicine, Umeå University, Sweden.

**Björn Karlson, Ph.D.** Professor at the Department of Psychology, Lund University, Lund, Sweden.

**Christin Mellner, Ph.D.** Researcher at the Department of Psychology, Stockholm University, Sweden.

**Christina Halford, M.D., Ph.D.** Researcher at the Department of Public Health and Caring Sciences, Uppsala University, Sweden.

**Frida Eek, Ph.D.** Associate professor at the Division of Occupational and Environmental Medicine, Lund University, Sweden.

**Ingibjörg H. Jonsdottir, Ph.D.** Associate professor at the Institute of Stress Medicine, Gothenburg, Sweden.

**Lars-Gunnar Gunnarsson, M.D., Ph.D.** Medical Doctor in Occupational and Environmental Medicine at Örebro University Hospital, Örebro, Sweden.

**Margareta Kristenson, M.D., Ph.D.** Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, Sweden.

**Nanna Hurwitz Eller, M.D.** Medical Doctor in Occupational and Environmental Medicine at Bispebjerg Hospital, Copenhagen, Denmark.

**Oskar Lundgren, M.D., Ph.D.-student** in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, Sweden.

**Peter Garvin, Ph.D, MPH.** Post-doc in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, Sweden.

**Petra Lindfors, Ph.D.** Associate professor in psychology at the Department of Psychology, Stockholm University, Sweden.

**Roberto Riva, Ph.D.** at the Department of Psychology, Stockholm University, Sweden.

**Roger Persson, Ph.D.** Senior reseacher at the National Research Centre for the Working Environment, Copenhagen, Denmark.

**Torbjörn Åkerstedt, M.D., Ph.D.** Professor at the Stress Research Institute, Stockholm University, Stockholm, Sweden.

**Töres Theorell, M.D., Ph.D.** Professor Emeritus at the Stress Research Institute, Stockholm University, Stockholm, Sweden.

**Ulf Lundberg Ph.D.** Professor Emeritus of Biological Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University, Sweden.

A special thanks to Professor Holger Ursin, Uni Health, Bergen University, Bergen, Norway for participating in discussions and commenting the outline and contents of the chapters.

# **CHAPTER 1**

# **The Role of Saliva Cortisol Measurement in Health and Disease. Introduction - Why This Book?**

#### **Margareta Kristenson1,\*, Peter Garvin2 and Ulf Lundberg3**

*1 Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, 58183 Linköping, Sweden; <sup>2</sup> Post-doc at the Department of Medical and Health Sciences, Linköping University, Sweden and <sup>3</sup> Professor in Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University, Stockholm University, Sweden.* 

**Abstract:** In recent decades, the technique of using ambulatory saliva sampling for measuring cortisol levels has become increasingly popular in field research and clinical studies aimed at investigating bodily responses to psychosocial stress and other psychological and clinical conditions. This interest is paralleled with frustrations on opposing and ambiguous results. To get a deeper understanding of the seemingly contradictory results, the Scandinavian cortisol and stress network (Scancort) was formed, based on 20 researchers from the disciplines of public health, psychology, biology and medicine. This book is based on a critical review of the existing empirical literature on salivary cortisol, aiming to evaluate the usefulness of salivary cortisol as a biomarker in various settings. In particular, this book focuses on how the many different ways of evaluating the levels and dynamics of salivary cortisol (*i.e.*, with regard to time points of assessment and different algorithms used to integrate data from multiple time points) affect the interpretation of cortisol measurements in various contexts. One main question is to find out if it is possible that different results of studies involving cortisol assessments are functions of differences in the theoretic assumptions made and the methods used.

**Keywords:** Salivary cortisol, stress, cognitive activation theory of stress, adults, ambulatory, single time point measures, deviations measures, area under the curve, laboratory test, dexamethasone.

#### **INTRODUCTION**

Cortisol is a stress hormone that can be measured in blood, urine and saliva. In recent decades, the technique of using ambulatory saliva sampling for measuring cortisol levels has become increasingly popular in field research and clinical studies aimed at investigating bodily responses to psychosocial stress and other psychological and clinical conditions. This non-invasive method is easy to administer and therefore can be implemented in large-scale study designs. Additional advantages are that, compared with blood sampling, saliva measurements do not induce discomfort or pain and do not interfere with the participants' normal activities and environment. Furthermore, saliva samples provide practical advantages; the rate of deterioration of cortisol is low in samples stored at room temperature. It has been shown that cortisol concentrations are relatively stable even after storing for a week at room temperature [1, 2]. Moreover, it has been shown that saliva can be stored in a refrigerator for at least 3 months without loss of cortisol, and cortisol concentrations are not affected by freezing of samples to any major extent [2, 3].

This interest in the use of saliva cortisol measurements is paralleled with frustrations on opposing and ambiguous results. Several psychological and physiologic conditions have been associated with increased and decreased cortisol levels.

#### **Context**

To get a deeper understanding of the seemingly contradictory results, the Scandinavian cortisol and stress

**<sup>\*</sup>Address correspondence to Margareta Kristenson:** Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, 58183 Linköping, Sweden; Tel: +46 10 103 5075; Fax: +46 10 103 1865; E-mail: margareta.kristenson@liu.se

network (Scancort) was formed. The network consists of about 20 researchers from the disciplines of public health, psychology, biology and medicine. It has been financed by the Swedish Research Council to gather competence and experience in Scandinavia regarding cortisol measurements.

In particular, the aim of the network has been to evaluate (and further develop):


The aim of the network has not been to investigate or compare performance characteristics and feasibility of different laboratory methods for quantification. For further reading on this topic, we recommend earlier work published elsewhere, for instance by Hansen and colleagues [4, 5].

This book has been written for other researchers who are interested in cortisol research. The sections on cortisol and stress theories are kept to introductory overviews. The main part of the work has involved scanning the existing literature and compiling the results on cortisol and various variables.

#### **AIM**

This book is based on a critical review of the existing empirical literature on salivary cortisol, aiming to evaluate the usefulness of salivary cortisol as a biomarker in various settings and what the results from measurements of cortisol mean in different study designs. In particular, the book focuses on how the many different ways of evaluating the levels and dynamics of salivary cortisol (*i.e.*, with regard to time points of assessment and different algorithms used to integrate data from multiple time points) affect the interpretation of cortisol measurements in various contexts. One main question is to find out if it is possible that different results of studies involving cortisol assessments are functions of differences in the theoretic assumptions made and the methods used.

### **CORTISOL**

Cortisol is one of the main stress hormones, and essentially prepares peripheral organs for action. The release of cortisol is mediated by the Hypothalamus-Pituitary-Adrenal (HPA) axis. The principal stimulus of the HPA axis is Corticotrophin-Releasing Hormone (CRH), which is locally produced in hypothalamus. When facing an acute stressor, CRH increases markedly and induces increased secretion of Adrenocorticotropic Hormone (ACTH) from the pituitary gland. ACTH stimulates the cortex of the adrenal glands to secrete cortisol into the circulation.

The role of cortisol has been discussed in 2 main areas: metabolism and inflammation. An increase in cortisol levels increases the supply of energy and oxygen, which translates to a temporary increase in blood pressure, blood glucose levels and free fatty acids. Therefore, it has been suggested that long-term dysregulation of cortisol leads to metabolic abnormalities [6].

The anti-inflammatory properties of cortisol are well known and have led to the widespread clinical use of exogenous cortisol (hydrocortisone). However, the role of endogenous cortisol is much more complex than regulating metabolism and inflammation. It has been suggested that cortisol modulates expression in approximately 10% of the body's genes [7].

Cortisol acts on the intracellular nuclear receptors found in most cell types throughout the body, regulating the transcription of target genes [8]. Thus, in contrast to stress hormones such as adrenaline and noradrenaline, which rapidly enter the bloodstream after exposure to a stressful situation, the secretion of cortisol is generally slower and peaks typically after 20-30 min [7].

In normal physiology, a typical pattern for most individuals shows a distinct diurnal variation in cortisol levels, peaking approximately 30-45 min after awakening and declining throughout the day, with lowest levels at night, around 04:00 h [9, 10] (Fig. **1**). It is not uncommon for the peak values in the morning to be 10-fold higher or more compared with the levels at night.

**Figure 1:** Diurnal variation of cortisol secretion. Figure modified from Ranjit *et al.* [7] and Kudielka *et al.* [8].

#### **STRESS**

The word stress has several connotations and is sometimes used in different ways. Our work is based on the definition developed by Levine and Ursin in Brown *et al.* [11]. They divide the term stress into 4 different entities, as seen in Table **1**. These 4 meanings of stress can and should, according to Levine and Ursin, be measured separately to further understand the concept of stress and the role of psychoneuroendocrinology in health and disease [11, 12].

**Table 1:** Four formal definitions of stress according to Levine and Ursin [11]


This introduction begins by discussing the psychobiological stress response and how this is, or could be, linked to measures of stressors and self-reported feelings. Hans Selye, often claimed to be the father of the stress concept, conceptualized stress as the General Adaptation Syndrome (GAS). Based on experimental animal studies, Selye postulated that exposure to stressors is followed by a generalized response consisting of 3 stages as shown in Fig. **2** [13].

Stage 1, the alarm stage or acute stage, is triggered by exposure to any kind of potentially harmful stimulus. This stage has been described by Cannon as the well-known catchphrase "fight or flight," when the body quickly mobilizes energy to handle a potential threat [15]. An acute response is characterized by a

#### **6** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson et al.*

hormonal shift towards a catabolic state. The stress hormones cortisol, adrenaline, and noradrenaline increase, whereas anabolic hormones promoting repair and growth such as insulin and sex steroids decrease [16]. After exposure and this rapid mobilization, the physiologic response declines [14]. The second stage is referred to as the stage of resistance, or stage of adaptation, when the physiologic response remains high in order to meet the demands of a prolonged stressor. If the duration of exposure is further prolonged, the body eventually reaches the third stage, the stage of exhaustion. The exposure to stressors has now triggered a dysfunctional state and hormonal imbalance, in which the physiologic response is weak, despite being exposed to a stressor normally triggering a strong response.

**Figure 2:** The three stages of General Adaptation Syndrome. Modified from Selye [13, 14].

When transferred to laboratory stress tests of humans or ambulatory sampling in normal populations in everyday life, individuals who are in the exhaustion stage based on GAS should, according to this model, be less responsive when exposed to new acute stressors.

This is supported by empirical data, comparing groups with clinical or subclinical signs of fatigue and exhaustion with groups with less or no signs of exhaustion, the latter group showing a higher responsiveness [17-19]. However, opposite findings with higher cortisol responsiveness in patients characterized by stress-related fatigue have also been reported [20]. Thus, an important notion in stress theory is that the response to an acute stressor is determined more by individual characteristics and the history of previous and current stressors, than by the actual tested acute stressor itself [12]. These observations have led to the incorporation of cognitive function in stress theory, a contribution not originally included in Selye's early work.

## **THE COGNITIVE ACTIVATION THEORY OF STRESS**

There are several stress theories, somewhat overlapping, which incorporate cognitive function in frameworks to understand stress and how a stimulus is translated into a physiologic response. One of the more widespread is the Cognitive Activation Theory of Stress (CATS), formalized by Ursin and Eriksen [12]. The main component in CATS is the feedback to the brain from the outcome of the response, which alters both the exposure to the stimulus and the perception of the stimulus in similar situations henceforth (Fig. **3**). Whether a stimulus is considered exciting or threatening depends on previous experiences and expectations of the outcome [12]. The process is dynamic; the stressor and outcome are evaluated and reevaluated in similar future situations.

Consonant with this, the feeling of being stressed can be linked with both positive and negative outcome expectancy. Thus, according to CATS, there is no point in trying to measure stress by objectively focusing on the external load of exposure. Attempts to measure stress should be focused on the subjective experience and feelings elicited by the stressors and the stress response [12]. CATS has implications on GAS as it leads to the conclusion that it is not possible to make a general dose-response association between load of exposure and stage of exhaustion.

**Figure 3:** The cognitive activation theory of stress. From Ursin and Eriksen [12].

Folkman *et al.* [21] have summed up the importance of cognitive function in filtering stress load with the statement that "it is not stress *per se* (referring to exposure to a given stimulus), but how people cope with it, that affects health and well-being".

According to CATS [12], the stress response depends on acquired expectancies of the outcomes of stimuli and available responses. Exciting or threatening depends on the individual appraisal of the situation, which is based on previous experience and expectations of the outcome. Thus, the stimulus expectancies may be distorted by psychological defense mechanisms, at least in humans. The response outcome expectancies to the available responses are defined as positive, negative or none. This offers formal definitions of concepts such as coping, hopelessness, and helplessness. For example, helplessness means that individuals cannot see any relation between their actions and the outcome of a threatening situation; hopelessness means that they believe that any action would lead to failure or even a catastrophe. The theory suggests that, if the man/animal is coping successfully, the threat or demands has a short phasic training effect on the body. If the man/animal is in a state of hopelessness or helplessness, it may lead to sustained activation and a catabolic strain effect on the body or lack of adequate response.

### **THE CONCEPT OF ALLOSTASIS**

In 1988, Sterling and Eyer [16] introduced the concept of allostasis in stress theory. The term allostasis literally means "to stand in variability," denoting stability through change. It was introduced as an antonym to the well-known term homeostasis meaning "to stand equally," denoting stability through constancy. Allostasis is based on the observation that most physiologic variables have a diurnal variation, determined by specific behavioral states and environmental events. Sterling and Eyer [16] argued that the term homeostasis may be misleading as it wrongly implies that different systems are kept constant at a "normal level." They claimed that more important for maintaining health is the ability to respond, thereby causing an appropriate arousal when facing an environmental challenge. A more adequate terminology would address the variation rather than the chronic state that homeostasis implies. In their allostatic model, health is defined as a state of responsiveness, including also the ability to restitute. An insufficient restitution leads to a sustained arousal, which in turn inevitably leads to the inability to respond appropriately.

 This concept has been further elaborated by McEwen and Wingfield [22] who also pointed out that although homeostasis applies to a limited number of systems essential for life such as maintenance of an adequate body temperature, blood pH and glucose level, and oxygen tension, allostasis is a necessary process to support homeostasis in these systems.

Both Sterling and Eyer [16] and McEwen and Wingfield [22] acknowledge that regulation of allostasis is multileveled, involving feedback at several levels from several hormones. However, cortisol is still described as a key player in allostatic regulation.

#### **CORTISOL MEASUREMENTS**

#### **Ambulatory Saliva Sampling**

Cortisol in saliva has been shown to reflect concentrations in serum with good precision [23, 24]. In particular, it is suggested that cortisol in saliva reflects the concentration of free cortisol (unbound to carrier proteins), which is believed to be the biologically active form. Convenient methods for ambulatory saliva sampling have been developed. A common, simple, and well-known device consists of a plastic tube and a cotton/polyester swab by which people can collect saliva themselves, for example during a normal day at home or at work, and then send the samples to the laboratory for analysis.

In addition to the marked diurnal variation, the day-to-day variation in cortisol levels within the same individual is also considerable. The correlation between cortisol levels on consecutive days has been reported to be around r=0.5 in several studies [18, 25, 26]. This variation can be explained to a large extent by temporary states in individuals, which may vary from day to day [18, 26, 27]. In order to obtain more reliable values, sampling over several days is often used. The mean levels over 2 or 3 days give more reliable results when testing whether people have a general capacity to respond in a certain way to a new challenge, for example, to working conditions compared with non-work at home.

In addition to the variation within subjects on different days, the variation between subjects is also high [9]. Therefore, in order to compensate for inter- and intra-individual variations in statistical analysis, it is commonly recommended to use relatively large samples of participants for meaningful intergroup comparisons, for example, between patients and healthy controls.

#### **Standardized Laboratory Stress Testing**

Cortisol secretion in response to a defined stressor exposure such as external stimuli (*e.g.*, light) or cognitive and emotional activation is sometimes of interest as an indicator of the regulation of the HPA axis. Several models have been developed, such as the Trier Social Stress Test (TSST) and other standardized stress tests, *e.g.*, the Stroop Color Word Test, Anger Recall, Mental Arithmetic, and the Cold Pressor Test.

The Physiologic effects vary widely across tasks [28], making it hard to compare studies with each other if only one stressor is used and therefore a combination on two ore more stressors are recommended. Testretest stability of laboratory stress responses increases if the laboratory stress includes more than one stressor [17].

Standardized laboratory testing and ambulatory saliva sampling have pros and cons. Using standardized testing in laboratory settings allows full control over the setting, including type of exposure and time for sampling. However, in addition to higher cost, problems are that according to cognitive stress theory, people with different experiences and expectations have different physiologic responses to the same stressor [11, 12]. Also, the stressors chosen may not be valid and easily transferred into everyday life conditions, and generalization to other contexts may be reduced.

Ambulatory saliva sampling under natural conditions, on the other hand, may be representative of the individuals' responses to everyday life, but may lack precision in defining the stressors as well as the time point when the samples were taken [29].

#### **Dexamethasone Suppression Test**

The dexamethasone suppression test is a specific test of the regulation of the HPA axis. Dexamethasone is a synthetic steroid, more potent than cortisol, which exerts a negative feedback to the pituitary to suppress the secretion of ACTH and, consequently, of cortisol. It is used clinically by monitoring cortisol after administration of the drug. In short, in a functional HPA axis, cortisol levels should be suppressed, whereas non-suppression of cortisol after administration indicates dysregulation of the HPA axis.

#### **Standardization and Control of Confounders**

All saliva sampling for cortisol assessment requires careful control in terms of sampling procedure and information on factors potentially influencing cortisol levels. The circadian variation of cortisol puts strong demands on these measurements to be clearly defined in terms of the actual time for sampling and the individual's circadian rhythm, such as time of awakening in the morning. Even small variations in time of sampling, especially in the morning, could have significant effects on the results. Clear instructions and compliance by the participants are needed to control for food intake, beverages, tobacco, alcohol, physical activity, medicine, sleep the night before, and other confounders.

### **Different Cortisol Measurements**

This book focuses on how the many different ways of measuring and evaluating the level and dynamics of salivary cortisol may influence the results. In the literature on saliva cortisol assessments the main types of measurements can be divided into the following groups:


#### **Single Time Points**

Measurements at single time points were more common in earlier studies, and are still common for serum cortisol. For example, a blood sample for cortisol analysis can be taken at a morning visit to the research laboratory.

With regard to saliva cortisol, repeated ambulatory sampling is common, where people collect saliva samples at home and/or at work at different times of the day. Measurements from these single time points are presented in several studies. It is also common that means or sums of 2 or several measurements over one day or several days are used to determine the level of cortisol. For example, Cohen *et al.* [30] measured salivary cortisol for several days in order to investigate the relation between cortisol and socioeconomic status (SES) defined by education and income. The term static can be used here to describe the use of single time points (or means) to determine a certain concentration. These measurements do not provide any information on the dynamics of the system, *i.e.*, whether the individual can respond adequately to a challenge or unwind and relax after stress exposure. Sometimes, one of these single time point measurements is called the basal or rest level, when it represents a rest period or the level observed before exposure to a standardized stress test.

However, to evaluate if the level is high or low relative to other samples or other individuals, it is crucial to know when a sample is taken, because cortisol levels typically follow a diurnal variation.

#### **Deviations/Slopes Between 2 or More Measurements**

When more than one saliva sample is used to assess changes in cortisol levels with time, these measurements can be used to quantify variations in activation of the HPA axis. These dynamic measurements aim to determine a change in secretion and indicate the individual's capacity to respond to or recover after stressful stimulation.

The main types of dynamic measurements seen are differences or slopes between 2 or more measurements. In ambulatory research, several deviations can be identified.

One of the most commonly used dynamic measurements is the difference between awakening and adjacent time points, typically 15 min and 30 min after awakening [31]. This deviation/difference is often called the Cortisol Awakening Response (CAR) or Awakening Cortisol Response (ACR) and can be defined as absolute and relative. For example, Riva *et al.* compared absolute CAR between female patients with fibromyalgia and healthy controls [32] and Olsson *et al.* [20] investigated relative CAR in patients suffering from stress-related fatigue.

Another measurement derived from ambulatory saliva sampling is the difference between morning and evening levels of cortisol, the cortisol decline over a day, which is often called diurnal variation. Although simplified, it is believed to capture patterns of secretion over the day by measuring cortisol at different points in time throughout the day or days. In the literature, this has been done by either deliberately including the peak value in the morning and subtracting evening values (see for instance [33]), or deliberately excluding the peak value in the morning and subtracting evening values (see for instance [34]). The rationale for the former would be to study the general capacity of the dynamics of cortisol secretion. The rationale for the latter would be that the peak value has a large effect on the deviation [34].

In addition, deviations have been presented between midday/afternoon, or late evening. For example, Nater *et al.* [35] compared morning versus evening salivary cortisol between people with Chronic Fatigue Syndrome (CFS) and a non-fatigued group.

Using standardized laboratory stress testing, dynamic measurements can focus on either reactivity during the stress test or recovery after the stressor. Reactivity is commonly measured as the difference between peak/stress level and baseline/pretest level, answering the question "how high was the increase in cortisol?" Recovery is commonly measured as either the difference peak/stress level and poststress level (after a fixed time point) or how long it takes in minutes to return to baseline values after finishing the test. For example, Kristenson *et al.* [17] investigated cortisol reactivity to a standardized stress test (TSST) in Lithuanian (Vilnius) men exposed to long-term psychosocial stress compared with healthy Swedish (Linköping) men (Fig. **4**).

### **Serum Cortisol responses to a laboratory stress test**

#### **Area Under the Curve Calculated from 2 or More Measurements**

The AUC captures an integrated value of cortisol excretion over a period of time. It is believed to combine information from several consecutive time points into one variable to facilitate statistical analysis. This calculation is done in 2 (principally different) ways [37]. AUC with respect to increase calculates the area under the curve using the first value as reference. In other words, it captures a change in secretion (*e.g.*, after a certain stressor) at a certain time. The AUC with respect to ground also includes the area created by the basal level (pre-test level). In other words, it captures the overall concentration at a certain time. It has also been shown empirically that there is a reasonably good correlation between AUC with respect to ground and levels of free cortisol in 24-h urine samples (r=0.4) [38].

As with the deviations mentioned earlier, the values of AUC are highly dependent on the time frame used in calculation. AUCs can be categorized as morning, mid-day, diurnal, and laboratory stress test. Both AUC with respect to ground and AUC with respect to increase may be computed for any time frame.

#### **Effects of the Dexamethasone Suppression Test**

This test typically measures the overnight effect of oral administration of 1 mg dexamethasone. Most healthy participants seem to respond well and their HPA axis is totally suppressed the day after, although in other cases the cortisol levels are not affected. This is commonly described as non-suppression. For example, Lange *et al.* [39] performed the Dexamethasone Suppression Test (DST) in 21 female borderline patients and 23 healthy controls [39].

#### **Relation of Saliva Cortisol Levels to Covariates - Stressors, Buffers and Outcomes**

The main aim of several studies using saliva cortisol measurements was to examine the importance of health determinants in relation to cortisol levels. In addition to the sampling procedure and standardization of the test setting, validity of results is also related to proper use of statistical analyses and control of confounders.

The distribution of saliva cortisol in a population is usually negatively skewed, which can be solved by logarithmic transformations of the data before use of parametric methods. This means that the influence of a few very high values is reduced. As logarithmic measurements are difficult to interpret, measurements are usually transformed back to actual values (*e.g.*, nmol/L) before presentation in tables and figures.

Lack of control for age and sex, time when the cortisol sample is taken, the individual's normal awakening time, various forms of medication, *e.g.*, antidepressant treatment, and many other factors may confound the result and modify the effects on the cortisol responses.

#### **CORTISOL AND HEALTH**

Cortisol is related to several health determinants. To a large extent, these health determinants can be logically organized in a hierarchy, as described in the model from Marmot and Wilkinson [40] (Fig. **5**).

This structure is also relevant according to CATS theory, *i.e.*, in terms of these factors being stressors, buffers, physiologic correlates, or measures of morbidity.

**Figure 5:** Model suggesting a link between external factors and health. Modified from Marmot and Wilkinson [40].

Stressors include life conditions in terms of individual SES and ethnicity and psychosocial work environment. Low SES, immigrant status, a poor work environment in terms of high job strain and poor social support, and poor balance between effort and rewards, are factors known to be associated with a higher risk of premature death and ill-health, most pronounced for mental ill-health, Cardiovascular Disease (CVD), and musculoskeletal disorders.

With regard to buffering resources, according to the CATS model, the stressor, *per se*, might not be informative enough regarding an individual's response in terms of straining or training to exposure. Whether the exposure is stimulating and exciting or threatening depends on the individual's appraisal of the situation, which in its turn is based on previous experiences and expectations of the outcome. The stimulus expectancies may, however, be distorted by psychological defence mechanisms, *e.g.*, denial or wishful thinking. This response outcome expectancy can be measured using psychometric scales of psychological resources, such as mastery, coping, self-esteem, sense of coherence [41], and psychological well-being [42].

Sleep is of particular importance in this context as it is the most important part of the restitution. During sleep, deep slow-wave (according to electroencephalography) sleep in particular, catabolic processes are replaced by anabolic processes and immune functions are enhanced. Sleep is also of importance for adequate metabolic functioning and it has been indicated that memory consolidation is strengthened during sleep.

With regard to physiologic correlates, long-term exposure to stressful conditions, such as work-related stress, is known to lead to increased risk for several disease groups, of which CVD is one. Therefore, the relationship between cortisol levels and established biomarkers for CVD risk are relevant. These include Body Mass Index (BMI), waist hip ratio, lipid levels, blood pressure, and insulin resistance (plasma glucose and insulin levels).

The concept of allostatic load combines these factors with catecholamines (as a measure of the autonomic nervous system activity) and increased coagulation (*e.g.*, fibrinogen level). Another physiologic pathway to observed outcomes of high stress (or stressor load) is immunologic factors; *e.g.*, Inter Leukins (IL), especially IL-6, and factors related to vulnerability (matrix metalloproteinase). Additional physiologic measurements include other hormones, *e.g.*, growth hormones.

With regard to morbidity measures, several mental and somatic health problems have been associated with cortisol secretion. Examples of mental health problems are perceived stress (as defined by Cohen [43]), depression, vital exhaustion, and burnout. Examples of somatic health problems are CVD, cancer, musculoskeletal disorders, and pain [44-47].


**Table 2:** Example of main table as used in the chapters

Information in table:

Reference Last name of first author.

Year Year of publication.

Exp Exposure (or outcome depending on content in chapter).

Des. Study design where C-S stands for cross-sectional, C-C for case-control, Pros. for prospective and Exp. for experimental design.

n cort Number of participants with cortisol measurements.

M/W Indicates number of me and women in the study group.

Arrow up indicates a positive significant association, arrow down a negative significant association and 0 a nonsignificant finding.

#### **OUTLINE OF THIS BOOK**

In the chapters of this book, a literature review of recent empirical studies is presented, relevant for the relations described earlier for health determinants/outcomes in relation to cortisol. This was done by sorting the evidence according to the methods used for measuring cortisol: single time points, deviation/slope, and AUC measurements or dexamethasone test according to the following schemes (Table **2**). All chapters present results in tables following this template*.* Each chapter includes 2 types of tables: 1 with full information on the study design, type of sample, and the methods used for sampling saliva, measurements of covariates and results; according to the three schemes above. The second table summarizes the findings from the first table, using arrows (up or down) and zeros to indicate positive, negative, and nonsignificant findings.

5 different categories of single time points are used for presenting results, 4 for deviations and 4 for AUC. These are defined as follows:

#### **Single Time Points**


#### **Deviation**

Using ambulatory saliva sampling:


Using standardized laboratory stress testing:

b4. Standardized laboratory test; reactivity (difference between peak and baseline/pretest level) and/or recovery (difference between peak and poststress level, *i.e.*, time when return to baseline was expected); when relevant; baseline before exposure to stressor.

#### **AUC**


#### **Dexamethasone Suppression Test**

No stratification on different doses in dexamethasone test has been made.

#### **Book Content**

The book contains 7 chapters on salivary cortisol in relation to a broad spectrum of factors. (Table **3**).

#### **14** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson et al.*


**Table 3:** Evaluated factors in relation to salivary cortisol

In each chapter, and in the summary in Chapter **9**, the findings are discussed in relation to the question: Is it possible that divergent results of studies are related to different theoretic assumptions and methods used?

#### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **Socioeconomic Status, Demographic Variables and Salivary Cortisol**

**Peter Garvin1,\*, Nanna Hurwitz Eller2 and Anette Harris<sup>3</sup>**

*1 Post-doc at the Department of Medical and Health Sciences, Linköping University, 581 83 Linköping, Sweden; <sup>2</sup> Medical Doctor in Occupational and Environmental Medicine at Bispebjerg Hospital, Copenhagen, Denmark and <sup>3</sup> Researcher at Uni Health and Department of Psychology, Bergen University, Bergen, Norway* 

**Abstract:** This chapter evaluates the association between salivary cortisol and socioeconomic variables (level of education, occupational status, income and related composite measures), ethnicity, age and sex. There were many non-significant findings for all variables, indicating that the associations with cortisol levels are relatively small. Regarding the significant results, there were some consistent trends. It is implied that high SES, regardless of how it is measured, is associated with a higher cortisol deviation throughout the day, and a higher capacity to react with increase in cortisol following a laboratory stress test. Regarding ethnicity, results consistently hint at a higher deviation throughout the day amongst Caucasians in comparison to Hispanics and Afro-Americans. Analyses on sex were not fully consistent, possibly due to influences of the menstrual cycle on cortisol levels. In addition, it has been reported that men and women respond differently to different stressors used in laboratory stress tests. For age, the significant findings found may hint at a small but general increase in cortisol levels throughout the day with increasing age.

**Keywords:** Salivary cortisol, socioeconomic status, educational level, occupational status, income, ethnicity, age, sex, single time point measures, deviations measures, area under the curve.

### **INTRODUCTION**

## **Socioeconomic Status**

Low Socioeconomic Status (SES), whether measured as educational level, income, occupational status, or other indicators are consistently associated with increased morbidity and mortality regardless of context [1- 3]. The mechanisms have been discussed for decades, but are as yet not fully elucidated. Suggested mechanisms are typically based on a lower material standard and lower financial resources, a higher exposure for both environmental risk factors and behavioral risk factors, and/or lower psychosocial resources such as coping and higher psychosocial risk factors such as depressive mood and hopelessness [4- 9]. As it has been hypothesized that behavioral factors and psychosocial factors may have an impact on stress hormones (see Chapters **4** and **7** for overviews), it has been suggested that at least part of the detrimental effects of low SES are mediated by stress hormones, in particular with reference to dysregulation of the Hypothalamo-Pituitary-Adrenal (HPA) axis [10].

A recent review of the literature on salivary cortisol and SES has been performed by Dowd *et al.* [11]. The review was based on 14 studies, and suggests that lower SES is related to a blunted pattern of cortisol secretion, although there are many inconsistencies in the results [11]. In the summary of the 14 studies, 4 showed an association between low SES and higher cortisol levels, 2 showed an association between low SES and lower cortisol levels (regardless of when cortisol was measured), 4 showed mixed results and 4 were non-significant [11]. The authors suggest that part of the explanation for the inconsistency was the variation in approaches regarding cortisol measurements and state that a better theory and study design should help clarify the expected and observed relationships between SES and cortisol levels. In this chapter, we go through the papers covered in the review by Dowd *et al.* [11] and add some more papers not included in the review. A brief review of ethnicity is also included in this chapter, as is sex and age.

**<sup>\*</sup>Address correspondence to Peter Garvin:** Post-doc at the Department of Medical and Health Sciences, Linköping University, 581 83 Linköping, Sweden; Tel: +46 10 103 8875; Fax: +46 10 103 1865; E-mail: peter.garvin@liu.se

#### **AIM**

To examine to what extent associations between different measures of SES and cortisol measurements and can be found, and which of the cortisol measurements seem to be of highest relevance. A second aim was to examine to what extent different cortisol measurements were related to ethnicity, age and sex.

### **METHOD**

#### **Search Strategies**

In a first step, an online search of the NCBI PubMed database was conducted (National Library of Medicine, National Institutes of Health, Bethesda, MD, USA. http://www.ncbi.nlm.nih.gov/PubMed). The search covered the period up to October 2009 (allowing e-publications if a full paper was published electronically prior to journal publication). Search terms were selected with reference to relevant PubMed terms and key words (see detailed description for each of the biological markers below), in combination with salivary cortisol in its truncated form ("saliva\*"). The limitations were set only to include studies matching "human" and "English".

In a second step, studies on patient populations were excluded (*e.g.,* cancer, diabetes, and major depressive disorder). Studies on genome variations, pregnant women, and pharmacological interventions were also excluded.

In a third step, all articles retrieved from each search were briefly read. If no direct statistical analysis between salivary cortisol and the evaluated factors were presented in tables, figures, or text, the paper was excluded. Intervention studies were included if associations with the factors of interest were presented as baseline characteristics. However, the effects on salivary cortisol in response to the intervention are not included in this review. Articles were also excluded if another (prior) publication from the same study material was already included in the evaluation.

#### **SES**

The terms "socio", "socioeconomic" and "SES" were used in combination with truncated salivary and cortisol as three searches. These yielded in 62 hits. In combination with truncated salivary and cortisol, the term "educational" yielded 39 hits, "income" yielded 40 hits, and "occupational" yielded 83 hits.

#### **Ethnicity**

The terms "ethnicity" and "race" were used in combination with truncated salivary and cortisol as two searches. These yielded 46 hits.

#### **Sex**

In addition, papers were included if sex-specific analyses were reported in other papers found in searches on socioeconomy. In total, there were 84 hits.

### **Age**

The term "age difference" in combination with truncated salivary and cortisol was used. In addition, papers were included if analyses on age were reported in other papers found in searches on socioeconomy. In total, there were 61 hits.

### **RESULTS**

Several studies control for the effects of SES and/or ethnicity, without presenting associations between cortisol and SES or ethnicity per se. Similarly, most studies control for the effects of sex and age, but do not explicitly present associations with those factors. Therefore, the number of studies is reduced in comparison to the numbers found in the searches. After meeting all exclusion criteria, 21 papers remained, describing at least one measure of salivary cortisol in relation to any of the measures of SES. There were 6 papers on ethnicity, 18 on sex and 12 on age. In total, 210 associations between salivary cortisol and any of the variables evaluated have been studied, comprising 108 associations on SES, 17 on ethnicity, 50 on sex, and 35 on age. The proportions between salivary cortisol and all markers tested were: for single time points, 108 (51%); for deviations, 73 (35%); for Area Under Curve (AUC), 26 (12%); and for dexamethasone tests, 4 (2%).

#### **Educational Level**

#### *Quantitative Analysis on the Studies Evaluated*

In 13 studies [12-24], there were 44 analyses on the relationship with salivary cortisol Table **1a**. Of these 27 were on single time points, 12 on deviations, 5 on AUC, and 1 on dexamethasone suppression test. In total, 12 of the analyses (27%) showed significant associations with salivary cortisol, whereas the other 32 (73%) were non-significant. Of these, 8 of the significant findings were found for single time points (30%), 3 for deviations (25%), and 1 for AUC (20%). The significant findings were mainly clustered in two categories: higher level of education was associated with low evening samples in 3 out of 6 studies (50%); higher level of education was associated with a higher capability to react on laboratory stress tests in 2 out of 3 studies (67%).


**Table 1a:** Summary of main findings of associations between measures salivary cortisol and high educational status sorted by year of publication

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; C-S, cross-sectional.

#### *Consistency of the Material*

Most of the associations evaluated were non-significant findings. None of the studies report any association between the level of education and cortisol levels at awakening [15, 17, 20, 23, 24], or between cortisol levels and Cortisol Awakening Response (CAR) [14, 15, 20, 22, 23]. These reports included small-scale studies and large-scale studies in different contexts. Arrows in opposite directions were found only for single time points of cortisol measurements in the morning; two studies report associations between a high cortisol level and high educational level [12, 14], whereas one study suggest an association between a low cortisol level and a high education level [23].

### *Methodological or Contextual Explanation for Divergent Findings*

The studies reporting inconsistent results for morning values are based on study designs that differ greatly. Based on the sampling protocol in each study, the highest control of when samples were actually taken is found in Bennett *et al*. [14], where all participants received a telephone call at normal awakening time, with a second call 30 min after as a reminder to take the second sample (constituting the morning sample in the study). Brandstädter *et al.* present a large-scale study, in which participants were instructed to leave a morning sample between 07:00 h and 09:00 h [12]. In the study by Garcia *et al.* participants were instructed to leave samples at 07:00 h as the second sample in the study [23].

## **Occupational Status**

## *Quantitative Analysis on the Studies Evaluated*

In 5 studies [12, 13, 25-27], there were 17 analyses on the relationship with salivary cortisol. Of these Table **1b**, 11 were on single time points, 5 on deviations, and 1 on a dexamethasone suppression test. In total, 6 of the analyses (35%) showed significant associations with salivary cortisol, whereas the other 11 (65%) were non-significant. Of the significant findings, 2 were found for single time points (18%), 3 for deviations (60%), and 1 for dexamethasone suppression test (100%).



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; C-S, cross-sectional.

## *Consistency of the Material*

The number of studies is too low to fully evaluate the association of occupational status and salivary cortisol. However, the reported finding of deviations hint at a higher diurnal variation [25] as well as a higher capability to respond to a laboratory stress test [13] amongst subjects with higher occupational status.

## *Methodological or Contextual Explanation on Divergent Findings*

Steptoe *et al.* [26] report non-significant associations between salivary cortisol and occupational status when evaluating the entire study population. They report that there is a difference between men and women; women with higher grades of employment tend to have somewhat higher levels of cortisol throughout the day compared with women with lower grades, whereas men with high grades of employment tend to have somewhat lower cortisol levels throughout the day compared with men with lower grades [26]. The possible sex difference may have a contextual explanation in manual/nonmanual work; the former may be associated with higher levels of stress hormones as a preparation for physical tasks.

#### **Income**

#### *Quantitative Analysis on the Evaluated Studies*

In 7 studies [12, 13, 19, 20, 22, 23, 28], there were 25 analyses on the relationship with salivary cortisol Table **1c**. Of these were 15 on single time points, 5 on deviations, and 5 on AUC. In total, 13 of the analyses (50%) showed significant associations with salivary cortisol, whereas the other 13 (50%) were non-significant. Of the significant findings, 8 were found for single time points (50%), 2 for deviations (40%), and 3 for AUC (60%).



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; C-S, cross-sectional.

#### *Consistency of the Material*

As with educational status, there is a possible contradiction in the findings presented by Brandstädter *et al.* [12] compared with Garcia *et al.* [23] regarding morning values of salivary cortisol (see above). Regarding other significant associations, there seems to be a consistency in that the findings suggesting a higher diurnal variation amongst subjects with higher income [20] are in line with findings of lower evening values [20, 23] and a lower AUC with respect to ground throughout the day [20]. Also, the reported higher capability by Kristenson *et al.* to respond to a laboratory stress test [13] amongst subjects with higher income are in agreement with Kraft *et al.* [28], who suggest that high family income is associated with higher levels of cortisol throughout a laboratory stress test (speech task).

#### *Methodological or Contextual Explanation for Divergent Findings*

There are no apparent systematic differences between the studies showing associations and the studies reporting non-significant findings. Thus, no clear contextual explanations can be found.

#### **Other SES Measures**

Using the search terms described earlier, there were a number of studies that used measures of SES that deviate from the classic three: education, occupational status, and income. Those are combined in this section.

#### **Quantitative Analysis on the Evaluated Studies**

In 6 studies [15, 19, 29-32], there were 22 analyses on the relationship with salivary cortisol Table **1d**. Of these were 10 on single time points, 10 on deviations, and 2 on AUC. Four of the analyses (18%) showed significant associations with salivary cortisol, whereas the other 18 (82%) were non-significant. Of the significant findings, 3 were found for deviations (30%), and 1 for AUC and (50%).


#### **Table 1d:** Summary of main findings of associations between measures salivary cortisol and other measurements of SES

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Pros, Prospective.

### **Consistency of the Material**

Most of the associations evaluated were non-significant findings. None of the cortisol measures based on single time points showed any significant association with SES measures. Cohen *et al.* [19] suggest that, although there are non-significant findings regarding single time points in both studies, a high SES (as composite measure) is associated with a lower AUC with respect to ground throughout the day.

Ranjit *et al.* [31] suggest that low material hardship is associated with a steeper slope following the morning peak (based on a higher peak). This is in line with the results by Steptoe *et al.* [32] where an increased financial strain is associated with a lower awakening response. Although different constructs are used to measure SES, this may be somewhat in contrast to Wright *et al.* [15] who suggest that participants with lower (self-rated) SES have a higher awakening response.

### *Methodological or Contextual Explanation on Divergent Findings*

There are no apparent systematic differences between the studies showing associations and the studies reporting non-significant findings. Thus, no clear contextual explanations can be found.

## **SES (All Measurements)**

As there is a considerable overlap in construct between common measures used to capture SES, the results for education status, occupational status, income level, and other measures of SES are aggregated and presented together as one entity.

### *Quantitative Analysis on the Evaluated Studies*

In 21 studies [12-32], there were 108 analyses on the relationship with salivary cortisol. Of these were 63 on single time points, 31 on deviations, 12 on AUC, and 2 on dexamethasone suppression test. Of these, 34 analyses (31%) showed significant associations with salivary cortisol, whereas the other 64 (69%) were non-significant. Of the significant findings, 17 were found for single time points (27%), 11 for deviations (35%), 5 for AUC and (41%) and 1 for dexamethasone suppression test (50%)

## *Consistency of the Material*

There are more non-significant findings than significant findings. However, there are few studies contradicting a general pattern where subjects with higher SES tend to have a somewhat higher diurnal deviation throughout the day, a lower AUC with respect to ground throughout the day, and a higher capacity to react with an increase in cortisol following a laboratory stress test.

## *Methodological or Contextual Explanation on Divergent Findings*

The studies using a measure of SES that deviates from the classic three (education, occupational status, and income), *i.e.,* using a composite measure, subjective SES, or other measure, do not seem to show significant findings to the same extent as the other three (with the exception of AUC; both studies that included this measure reported a significant association).

## **Ethnicity**

### *Quantitative Analysis on the Evaluated Studies*

In 7 studies [14, 20, 22, 33-36], there were 17 analyses on the relationship with salivary cortisol Table **1e**. Of these were 6 on single time points, 9 on deviations and 2 on AUC. In total, 10 of the analyses (58%) showed significant associations with salivary cortisol, whereas the other 7 (42%) were non-significant. Of the significant findings, 3 were found for single time points (60%) and 6 for deviations. The significant findings on deviations were clustered in category b3 (deviation throughout the day) where 6 out of 6 findings were significant.

**Table 1e:** Summary of main findings of associations between measures salivary cortisol and ethnicity sorted by year of publication.


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Exp, experimental.

### *Consistency of the Material*

There is a high degree of consistency in the presented results. All three studies that compared Caucasians and Afro-Americans and evaluated a deviation throughout the day reported a higher deviation among Caucasians [20, 34, 35]. It is also suggested that Caucasians have higher deviation throughout the day than Hispanics [35, 36], and Hispanics have higher deviation than Afro-Americans [35]. The results on deviations is also supported by difference at single time points by two studies, where it is reported that Caucasians have a higher level in the morning and a lower level in the evening in comparison to African-Americans [14, 20].

### *Methodological or Contextual Explanation for Divergent Findings*

There are no apparent clear contradictions in the results. Wilcox *et al.* [33] examined stress reactivity in caregiving postmenopausal women and found that more African Americans (58%) than Caucasians (14%) showed >50% increase in cortisol during a test interview about negative aspects of being a caregiver. This laboratory task might reflect a more uncontrollable and stressful situation as a caregiver amongst African Americans than amongst Caucasians.

### **Sex**

### *Quantitative Analysis on the Evaluated Studies*

In 18 studies [12, 15-17, 22, 27, 30, 37-47], there were 50 analyses on the relationship with salivary cortisol Table **1f**. Of these were 21 on single time points, 22 on deviations, 6 on AUC, and 1 on dexamethasone test. Of these, 20 of the analyses (40%) showed significant associations with salivary cortisol, whereas the other 30 (60%) were non-significant. Of the significant findings, 5 were found for single time points (23%), 11 for deviations (50%), 3 for AUC (50%) and 1 for dexamethasone test (100%).


**Table 1f:** Summary of main findings of associations between measures salivary cortisol and sex sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional. a

 Differences vary with menstrual cycle. b

 Different results for different stressors. c

Difference for workdays but not for weekends.

#### *Consistency of the Material*

Overall, there were small differences reported between women and men. The highest consistency was found in laboratory stress tests, where men had a higher response than women in 4 studies (50%) [37-39, 46]. The associations in the early hours of the day hint at a possible difference, where women tend to have somewhat higher values in the morning, as assessed by CAR or single values in the morning (41%) [12, 15, 17, 27, 41, 47].

#### *Methodological or Contextual Explanation for Divergent Findings*

It has been reported by Stroud *et al.* [39] that different stress tests have different patterns for women and men. Women appear more physiologically reactive to social rejection challenges (such as being systematically excluded by associates during a conversation), but men react more to achievement challenges (where study participants were told that the investigator studied the relation between intelligence and performance) [39]. This might have implications on choice of stressors depending on whether men or women are to be studied.

Several studies indicate that cortisol levels are influenced by the menstrual cycle, which complicates analyses on sex differences. It is suggested that estradiol induces changes in corticosteroid-binding protein levels [38]. Kirschbaum *et al.* [38] suggest that men in general have a stronger hypothalamic drive in response to stressful stimulation than women. However, although the difference is consistent between men and women in the follicular phase or women using contraceptives, there were no difference between women in the luteal phase [38].

### **Age**

#### *Quantitative Analysis on the Evaluated Studies*

In 12 studies [12, 16, 17, 21, 22, 29, 30, 40, 41, 45, 48, 49], , there were 34 analyses on the relationship with salivary cortisol Table **1g**. Of these, 17 were on single time points, 11 on deviations, 5 on AUC, and 1 on dexamethasone test. Eleven of the analyses (32%) showed significant associations with salivary cortisol, whereas the other 23 (68%) were non-significant. Of the significant findings, 5 were found for single time points (29%), 4 for deviations (36%), 1 for AUC (20%) and 1 for dexamethasone test (100%).

### *Consistency of the Material*

Most of the associations evaluated were non-significant findings. The reported significant associations may hint at a general increase in salivary cortisol levels with increasing age. Four independent studies show that increased age was associated with increased cortisol levels, at different time points throughout the day. Moreover, two out of two studies conclude that increasing age is associated with a higher reactivity in laboratory stress tests.

#### *Methodological or Contextual Explanation for Divergent Findings*

It has been suggested that age differences are more apparent in older ages [41]. There were however no apparent differences in mean age between the studies reporting significant differences and the studies reporting non-significant findings.


**Table 1g:** Summary of main findings of associations between measures salivary cortisol and age sorted by year of publication

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional.

Most studies, with significant and non-significant findings, had a fair sample size. Thus, lack of statistical power is an unlikely explanation to the large number of non-significant findings.

### **DISCUSSION**

#### **General Remarks**

Before interpreting the results further, there are 3 aspects that should be considered.

First, we studied salivary cortisol levels alone, focusing on the feasibility of using this approach with regard to socioeconomic and demographic variables. A number of other studies using concentrations in sera and/or urine that we omitted may be of relevance when determining associations between cortisol and the factors studied.

Second, a large proportion of the articles are based on relatively small study populations. This leads to the possibility of a high number of beta errors in the presented non-significant findings. On the other hand, the results may suffer from publication bias, where non-significant findings are not reported explicitly in some papers, even though analyses were done on cortisol and socioeconomic or demographic variables. It is somewhat of a scientific oddity that there are numerous studies that adjust for age, sex, SES, and ethnicity without reporting if there are associations between cortisol and these factors to begin with. Indeed, several studies report low *R*<sup>2</sup> values for regressions on cortisol levels, also when including SES, ethnicity, age, and sex in the models. For example, O'Donell *et al.* [45] reports an *R*<sup>2</sup> value of about 0.13 when studying various cortisol measures adjusting for income, sex, age, body mass index, depression, smoking, self-rated health, and awakening time.

Third, the search strategies used may be somewhat incomplete. It is likely that associations between any of the variables covered in this investigation and cortisol in saliva have been studied and presented in papers that could not be identified in our search. In particular, this might be the case for age and sex, where analyses are presented but hidden from key words and titles, as those analyses are not a primary aim of the paper.

A description of all papers covered in this section can be found in Table **2**. 

**Table 2:** Included studies on socioeconomic status and demographic variables sorted by appearance in text in this chapter 


#### **28** *The Role of Saliva Cortisol Measurement in Health and Disease Garvin et al.*

#### 


#### *Socioeconomic Status, Demographic Variables The Role of Saliva Cortisol Measurement in Health and Disease* **29**


#### **30** *The Role of Saliva Cortisol Measurement in Health and Disease Garvin et al.*



#### *Socioeconomic Status, Demographic Variables The Role of Saliva Cortisol Measurement in Health and Disease* **31**


#### **32** *The Role of Saliva Cortisol Measurement in Health and Disease Garvin et al.*

#### 


#### *Socioeconomic Status, Demographic Variables The Role of Saliva Cortisol Measurement in Health and Disease* **33**


#### **34** *The Role of Saliva Cortisol Measurement in Health and Disease Garvin et al.*

#### 


#### *Socioeconomic Status, Demographic Variables The Role of Saliva Cortisol Measurement in Health and Disease* **35**


#### **36** *The Role of Saliva Cortisol Measurement in Health and Disease Garvin et al.*

#### 


#### *Socioeconomic Status, Demographic Variables The Role of Saliva Cortisol Measurement in Health and Disease* **37**


*Abbreviations:* ACR, Awakening cortisol response ANCOVA, analysis of covariance; ANOVA, analysis of variance; BMI, body mass index; CAR, cortisol awakening response; C-C, case– control; CESD, Centers for Epidemiological Studies Depression; CPT, cold pressure test; C-S, cross-sectional; DEXA, dexamethasone; DHEA, dehydroepiandrosterone; EIA, enzyme immunoassay; ELISA, enzyme linked ; Excl, exclusion; Exp, Experimental; GLM, general linear model; HPA, hypothalamo-pituitary-adrenal; MDD, major depressive disorder; Pros, Prospective; RIA, radioimmunoassay; SES, socioeconomic status; TSST, Trier Social Stress Test.

### **SES, Similarities and Differences Between Different Measures**

It should be acknowledged that income, education, and occupational status are different entities that are not easily interchangeable [50]. In this context, however, all 3 may be aggregated as proxies for SES, as there is generally a high correlation between SES measures [13], to see if the associations with salivary cortisol follows a general pattern. The number of studies is still low even after such an aggregation.

The general pattern that emerges is hinting at an association between higher SES and a higher deviation throughout the day, a lower AUC with respect to ground throughout the day, and a higher capacity to react with an increase in cortisol following a laboratory stress test.

However, the differences in SES are consistent but small, in particular for diurnal deviation and AUC throughout the day.

In one study by Cohen *et al.* [19], the overall difference in total concentration throughout the day is attributed to small differences (non-significant for each time point) throughout the day that accumulate to a significant difference in total concentration. In a larger study, Cohen *et al.* [20] demonstrate significantly lower levels in the later part of the day but non-significant differences earlier, consequently with significant diurnal variation due to lower evening levels.

It seems that other measures than income, occupational status and education has a lower frequency of significant findings than studies using the more conventional three measures. This may have to do with the validity of other composite or other proxies for SES.

Of the measures tested, income had the highest proportion of significant findings (50%). This might be explained by a supposedly high correlation between actual income and social status, whereas use of ordinal data of occupational grade and education attainment may suffer from higher variability at individual level, thus diluting the associations with social status. Moreover, data on income may be more feasible for linear analyses (regardless on correlation with social status) than occupational grade and educational attainment. Indeed, most studies using income had significant associations with at least one of the cortisol measurements used in the different studies.

There are several factors that add to the complexity when studying SES and cortisol: One factor of plausible explanatory value is the level of physical challenges at work. For example, Steptoe *et al.* [26] report differences between men and women, where women with higher grades of employment tend to have somewhat higher levels of cortisol throughout the day compared with women with lower grades, whereas men with high grades of employment tend to have somewhat lower cortisol levels throughout the day compared with men with lower grades. The higher levels among women with high grade may partly be driven by psychological challenges and high job demands, whereas the higher levels among men with lower grade may be driven more by physiological challenges. In line with psychological challenges at work, Kunz-Ebrecht *et al.* [51] suggest that there is an interaction between SES and job demands, where mean CAR is high in groups with low SES and high job demands, and low in groups with low SES and low job demands.

Another factor may be the commonly used explanation that individuals in low SES more often are exposed for stressful episodes. Several research groups support their findings by a theoretical assumption that stressful episodes triggers a physiological process by which initial exaggerated cortisol stress response may result in lower overall cortisol levels over time. However, in cross-sectional studies, stressful events and its perceptions naturally vary in duration, intensity and temporality amongst different individuals. Thus it might be a simplistic tool to evaluate solely hyper- or solely hypo-secretion of cortisol, or to study cortisol levels in a linear fashion in regards to SES. In a study using an unconventional statistical approach, Li *et al.*  suggests that subjects with low SES are overrepresented in the extremes in both ends, eliciting both high and low cortisol values [52]. If this pattern is correct, it would generally dilute the associations between cortisol levels and SES in statistical analyses assuming linear or ordinal structure in data.

Further, there is an emerging discussion on the biological mechanisms behind the detrimental effects of low SES. In addition to the ongoing discussions on possible epigenetic effects [53], a growing literature suggests that at least part of these associations may be explained by exposure *in utero* or in the first years in life. As examples, Lupien *et al.* has suggested that young children are affected by the parental socioeconomic situation [54, 55]. In concordance, Gustafsson *et al.* have suggested that SES in early life seem to be more important than current SES for cortisol levels [56].

#### **Ethnicity**

When studying ethnicity and salivary cortisol, a complex web of potential covariates unfolds. In the studies that we have reviewed, we cannot know if the results found are a product of genetic differences, behavioral differences, or if ethnicity is a proxy for SES. However, the studies on ethnicity yielded the highest proportion in this chapter (58%), and it may be of relevance that the studies on ethnicity in this overview follow a clear trend: Caucasian study populations have a higher diurnal variation than African American study populations. It is also suggested that Caucasians have higher diurnal variation than Hispanics, who, in turn, have higher diurnal variation than African Americans. This ethnicity ladder is in congruence with a translation to a socioeconomic ladder, where Caucasians in many societal contexts have higher status than African Americans and Hispanics. If so, these results give further support to the findings on SES, emphasizing that a higher status seem to be associated with a higher diurnal variation. Importantly, a recent large population study (n=935) using cortisol samples from three days by Hajat *et al.* [57] confirmed the presence of a flatter diurnal deviation in African-Americans and Hispanics compared to Caucasians, and also confirmed a more flat deviation in groups with low SES compared to groups with high SES. The flatter deviations, mainly an effect of lower cortisol levels in the morning, are coherent with other large scale populations studying ethnicity and or/SES in diverse populations using a well standardized cortisol sampling throughout the day [12, 20, 35, 58]. Thus, it seems likely that the found differences over ethnicity at least in part are explained by environmental factors.

### **Sex**

Based on the studies that we have reviewed, it can be stated that there are no large differences in cortisol levels between men and women. It may however be noted that if anything at all, the five significant findings on single time points all hint at a somewhat higher level amongst women in comparison to men. The absence of significant findings for single points in this investigation does not provide support to any great extent for biological differences between men and women. It has however been reported that menstrual cycle might have a substantial impact on cortisol levels for premenopausal women [38]. Thus, there might be a larger variation comparing women over the menstrual cycle than it is comparing women and men.

The most pronounced finding on sex differences hints at a higher reactivity in laboratory stress test settings for men [37-39, 46]. When explaining differences in laboratory stress tests between men and women, focus should, as pointed out and discussed by Stroud *et al.* [39] and Kirschbaum *et al.* [38], be focused on the anticipation and interpretation of the stressors used. If men and women are likely to differ in their interpretation of the situation, then differences in stress tests are psychological rather than biological phenomena. This is further supported by Kirschbaum *et al.* [38] who reported that pharmacological stimulation yields similar patterns with peaks in the same range and at the same level for both men and women.

#### **Age**

There is some evidence that cortisol levels increase somewhat with increasing age. Even though the number of studies is low, the differences that follow an age gradient are higher in the later part of the day than in the morning. This is further supported by Deuschle *et al.* [59], who suggested that plasma cortisol does not differ in the daytime, but that cortisol levels in the evening increase with age. Thus, they suggest that higher age is associated with a flatter diurnal slope [59]. In concordance, Van Cauter *et al.* [60] have also suggested that the diurnal slope becomes lower as age increases. Still, it should be remembered that the results hint at a relatively small increase in cortisol levels with increasing age.

#### **CONCLUSIONS**

The number of studies with significant findings is relatively low. However, there is a fair degree of consistency in that high SES, regardless of how it is measured, is associated with a higher deviation throughout the day, and a higher capacity to react with an increase in cortisol following a laboratory stress test. Regarding ethnicity, results consistently hint at a higher deviation throughout the day amongst Caucasians in comparison to Hispanics and Afro-Americans.

There are no apparent consistent large differences between men and women. There is a tendency that women in general have somewhat higher levels than men. There is also a tendency that men have a higher cortisol response than women in laboratory stress test settings, but these responses are dependent on the stressors used.

The studies on age are few but provide some support for a small increase in salivary cortisol with increasing age.

### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **Psychosocial Work Stressors and Salivary Cortisol**

#### **Björn Karlson1,\*, Petra Lindfors2 , Roberto Riva<sup>3</sup> , Christin Mellner<sup>4</sup> , Töres Theorell5 and Ulf Lundberg6**

*1 Professor at the Department of Psychology, Lund University, Box 117, 221 00 Lund, Sweden; <sup>2</sup> Associate professor in psychology at the Department of Psychology, Stockholm University, Sweden; <sup>3</sup> Student at the Department of Psychology, Stockholm University, Sweden; <sup>4</sup> Researcher at the Department of Psychology, Stockholm University, Sweden; <sup>5</sup> Professor at the Stress Research Institute, Stockholm University, Stockholm, Sweden and <sup>6</sup> Professor in Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University, Stockholm University, Sweden* 

**Abstract:** This chapter systematically reviews how different measures of salivary cortisol are related to different measures of psychosocial work stress. Divergent findings were scrutinized with respect to study quality and the methods used. Measures of work stress included concepts reflecting those included in the demand-control-support model or the effort-reward-imbalance model. General bibliographic databases (PsychINFO and PubMed) were searched up to September 30, 2009. Two reviewers extracted data on study characteristics and study quality. In total 27 articles fulfilled the inclusion criteria. Cortisol measures were grouped into single time points at different times during the day, deviations at different time periods during the day, reactivity and recovery after a standardized laboratory test, area under the curve from deviations and reactivity measures. A large proportion of the analyses of the associations between cortisol and psychosocial work stressors showed nonsignificant findings. However, of the significant findings, most results showed that a high work stress was associated with high cortisol levels. Significant relationships were evenly distributed across different measures of psychosocial work stress. As regards salivary sampling or statistical analysis, no strategy seemed superior but some strategies have only been used in the past few years. Typically, older studies were of lower quality. Low quality studies tended to have a higher proportion of significant findings which is a reason for concern. The relatively few significant findings may be because many psychosocial work stressors were of mild or moderate intensity and the study groups were rather small and fairly homogeneous, thus variability was too small to reveal any effects. The results indicate a normal, healthy response to work stress in most workers, according to CATS and the Allostatic Load Models.

**Keywords:** Salivary cortisol, working adults, psychosocial work stress, work load, job strain, job demands, job control, effort, reward, social support.

#### **INTRODUCTION**

It has been a popular idea among stress researchers and those organizing work to use salivary cortisol as an index reflecting work stress. This is partly associated with the hope that salivary cortisol would be an easily administered objective measure of work stress. More specifically, its exponents hoped that salivary cortisol would prove useful for employers who may want or require objective evidence in order to accept subjective claims of a work environment being harmfully stressful. Similarly, salivary cortisol could be useful for employees wanting to use high salivary cortisol concentrations as an argument for claiming that they are exposed to a highly stressful environment or when arguing for the beneficial effects of interventions aimed at reducing work stress. However, the interpretation of the results of salivary cortisol concentration analyses in relation to work conditions is more complicated than was initially expected. Working populations differ from patient populations or from populations of retired individuals in the sense that those who are working are healthier and younger than patients and retired persons. Thus, cortisol regulation in working individuals may show a healthier pattern than in other groups. The study of cortisol regulation at work should be more closely linked to the assessment of stressors than cortisol regulation in patient groups.

**Margareta Kristenson, Peter Garvin and Ulf Lundberg (Eds) © 2012 The Author(s). Published by Bentham Science Publishers**

**<sup>\*</sup>Address correspondence to Björn Karlson:** Professor at the Department of Psychology, Lund University, Box 117, 221 00 Lund, Sweden; Tel: +46 46 222 9117; Fax: +46 46 222 4209; E-mail: bjorn.karlson@med.lu.se

However, this is only partly true; working individuals can also be exhausted and depressed although they still manage to carry out their work tasks.

In this context, increased cortisol excretion when mobilizing resources to endure a situation and successfully cope with it plays an important role in the organism's response to increased demands. The expected effects of a high cortisol level include a decrease in the negative psychological reactions to the stressor, such as feeling uneasy in an adverse psychosocial work environment. When this mechanism functions adequately, a working individual feels less uneasy in a poor psychosocial work environment than would be the case if cortisol levels stayed low. At the group level, this effect decreases the association between feelings of uneasiness and high cortisol excretion as reflected in blood and salivary cortisol concentrations.

With regard to psychosocial work stressors, most of these daily stressors are of mild or moderate intensity. Irritations caused by lazy colleagues, a constant lack of time to do satisfactory work or poor leadership are milder stressors than natural disasters, emergency situations, giving birth to children, and exposure to unexpected violence. Even in occupations where violence or natural disasters may be expected to be part of the work (police officers, prison staff, and fire fighters), strongly stressful situations are relatively rare and may not be captured in a few measurements of salivary cortisol during a single work day. In addition, successful coping with work stressors during ordinary work days is according to the CATS model inducing positive expectancies and a successive attenuation of the stress response and rapid return to baseline.

In the systematic analysis of research findings on salivary cortisol in relation to psychosocial work conditions, the bodily responses in terms of cortisol excretion to milder stressors must be considered. Most of the published research within the field has investigated psychosocial work conditions through the widely accepted theoretic models of work stressors such as the Demand-Control-Support (DCS) model [1, 2], or the model of Effort-Reward-Imbalance (ERI) [3]. This means that the stressors investigated have indeed been mild in character. However, long-lasting levels of mild or moderate stress may also be harmful.

### **AIM**

This chapter examines how different measures of salivary cortisol are related to different measures of psychosocial work stress by systematically reviewing the literature. In addition, divergent findings have been scrutinized with respect to study quality and methods used.

### **METHODS**

### **Search Strategies**

### *Procedure*

Literature searches were performed in two general bibliographic databases: PsychInfo (until September 2009) and PubMed (until September 2009). Articles were identified by combining different sets of search terms in separate searches ("cortisol" AND "job demands" OR "work demands" OR "job strain" OR "work strain" OR "job control" OR "work control" OR "social support" OR "job stress" OR "work stress"), ("cortisol" AND "work" OR "job" AND "ERI" OR "effort" OR "reward" OR "effort reward"). The searches were limited to published articles written in English and including adult study participants (age 19+ years). Reference lists from relevant review articles were also scrutinized. The search was completed with articles known by the authors of this chapter but not identified through the search terms. This resulted in 957 hits, including duplicates (*e.g.*, articles found in both databases and across different searches). Further evaluation of articles was performed according to specific inclusion criteria.

### *Inclusion Criteria*

For an article to be included, cortisol had to be analyzed from salivary samples. This means that studies including parallel examination of, for instance, salivary and urinary cortisol were included. The exposure variable had to include psychosocial work stress, mainly including concepts relating to the DCS model or the ERI model. However, use of the original measures of these models was not required because this would have restricted the number of articles unnecessarily. The researched measures had to be conceptually similar to the psychosocial work stressors included in the DCS model or the ERI model. The exposure variable had to be either self-reported by study participants or objectively defined. In the latter case it had to be defined as some kind of continuous measure of workload that could vary over time. This means that psychosocial work stress defined in terms of occupation, job title, or work day in contrast to a day off was considered insufficient. The study participants had to be in gainful employment (*e.g.*, excluding students and charity workers), and the exposure studied had to involve work life (*e.g.*, excluding laboratory studies).

The first selection of articles based on publication titles and abstracts resulted in 103 articles. All of these articles were retrieved and read by at least two of the authors of this chapter. Assessment of study inclusion criteria was made independently by each author. Disputes were settled by consensus. This evaluation resulted in a final selection of 27 articles to be included in the systematic review (Fig. **1**).

**Figure 1:** Brief description of the retrieval of publications included in the review.

#### **Evaluation Process**

In addition to summarizing the articles in line with the matrix shown in Table **1**, the authors of this chapter also rated the quality of the articles.

#### **46** *The Role of Saliva Cortisol Measurement in Health and Disease Karlson et al.*

**Table 1:** Descriptives of the included articles on salivary cortisol parameters and exposure sorted by year of publication 


#### *Psychosocial Work Stressors and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **47**


#### **48** *The Role of Saliva Cortisol Measurement in Health and Disease Karlson et al.*


#### *Psychosocial Work Stressors and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **49**


#### **50** *The Role of Saliva Cortisol Measurement in Health and Disease Karlson et al.*



#### **52** *The Role of Saliva Cortisol Measurement in Health and Disease Karlson et al.*

#### 


**Table 2:** Brief summary of main findings of associations between salivary cortisol parameters and exposure/outcome/biological marker in studies sorted by type of exposure and year of publication (arrows up/down or 0 indicate results)




Exposures are expressed in terms of the assumed stressful pole of the concept, so an upward arrow will always indicate a higher cortisol measure under the conditions of a higher stress exposure. a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground.

#### *Quality Rating*

A separate matrix (Table **3**) was used to rate the quality of each study included with respect to the criteria used by Chida and Steptoe [4]. These criteria include whether or not relevant confounders were addressed or controlled for. Specifically, the following eight confounders were evaluated: age, sex, smoking status, participants' adherence to procedures for salivary sampling (*e.g.*, by electronic monitoring, self-reported time of sampling), steroid medication, time of awakening, sampling days (number of days and type of sampling day; week day or weekend), and clear and standardized instructions to study participants regarding the sampling procedure.

Three levels of judgments for each of the criteria were used giving 0, 1 or 2 points. In general, 2 points involved statistically controlling for a confounder (*e.g.*, as a covariate) or analyzing group differences and taking the results into consideration. One point typically included addressing the confounder in the text and indicating awareness of its potential role as a confounder and/or discussing why a confounder was not measured or statistically controlled for and the potential consequences of not including a confounder among the study variables. As regards standardized instructions, a maximum of 1 point was given when study participants were reported to have received any instructions, irrespective of the levels of detail. When no report of instructions was included, the rating was 0 points. Six additional criteria were rated from 0 to 2 points and one from 1-2 points, giving a possible total score range of 1 to 15.

The principles for evaluating the other 7 confounders are specified below:

*Age* - included in statistical analysis, controlled for or otherwise accounted for (2 points); discussed but not statistically analyzed or accounted for (1 point); not analyzed or discussed (0 points).

*Sex* - sex was taken into account in the analyses, *e.g.*, separate analyses of women and men or included as a between group factor (2 points); if possible sex effects were discussed but not analyzed, even when the study include one sex only (1 point); only one sex was included in a study and/or possible sex effects were not discussed, or women and men were not studied separately in any way (0 points).

*Smoking status* - evaluated in the same way as age.

*Adherence* - adherence was monitored technically with electronic devices, or analyzed based on other data and treated by exclusion of non-adherent participants or otherwise treated in a stringent manner (2 points); only addressed or discussed (1 point); not addressed at all (0 points).

*Steroid medication* - exclusion of study participants taking some kind of medication or of post-menopausal women (2 points); potential effects of steroid medication were discussed (1 point); such medication not mentioned or discussed at all (0 points).

*Waking time* - controlled or in some way accounted for or handled (2 points); discussed but not measured or accounted for (1 point); not mentioned or discussed (0 points).

*Sampling days* - cortisol sampling on more than 1 day analyzed and accounted for (2 points); one sampling day or possible effects or sampling days discussed (1 point); when it was unclear on which days sampling had occurred the article was excluded from further analyses.

### *Global Quality Evaluation*

A quality rating based on scoring a set of variables may give an indication of many quality aspects of a study, but there is also a risk that the pre-defined scoring principles result in an over- or underevaluation of a study. For instance, in an article with a high score it is still possible that poor biochemical analysis or inadequate statistical methods influenced the study findings and result in an overevaluation of its quality. Similarly, a low quality score may following from not statistically controlling or not discussing specific confounders that were of minor importance in the study setting. This means that a well-designed study failing to meet the quality criteria scored lower, which may result in underevaluation of the study quality. To adjust for this bias in quality scores relating to the predefined criteria, a global quality evaluation was performed, in which a study was classified as being of high, moderate, or low quality. This assessment included taking into consideration the sample size in relation to the study design and statistical analyses, clarity in defining the exposure measure(s), specificity of the exposure measure(s), reliability with regard to the biochemical analyses, restrictions with respect to generalizing the results (including details on drop-out rates, selection bias, and heterogeneity of the study participants with respect to sex, age, occupation, and socioeconomic status).

### *Procedure*

Initially, all 6 authors summarized and evaluated one of the articles in accordance with the summarizing matrix, in order to fine tune their inter-rater concordance and the design of the matrix. Each of the 60 articles selected from the abstract, was then read, briefly summarized, and preliminarily evaluated according to the matrix independently by 2 of the authors in varying pair-wise groupings with each author reading about 20 articles.

In the next step, all groups of authors went through the summaries and preliminary evaluations of the articles read, which resulted in another 33 studies being found not to fulfill the inclusion criteria because only urinary or plasma cortisol was used, because they did not fulfill the criteria for exposure (*e.g.*, not having used or sufficiently described a measure for self-reported job stress), resulting in 27 articles remaining for in-depth review. These articles were read once more by all authors with each author independently scoring the articles in accordance with the quality criteria matrix. These second scorings for each article were discussed among the authors and disputes were settled by re-evaluating the article until consensus was reached. Each study was also given a global quality evaluation score by the author groups. The remaining 27 articles [5-31] were re-read once more with each article being read by two authors in varying groupings in order to summarize the study and relevant findings (Tables **1** and **2**).

### **RESULTS**

### **Quality Assessment**

The quantitative summary of the quality score had a possible range from 1 to 15 and the quality scores among the 27 articles ranged from 5 to 14 (median=10) (Table **3**). Fig. **2** shows that the quantitative quality scores increased over time, which means that recent studies were of higher quality than the earlier ones. According to the global quality evaluation summarized in Table **3**, 11 articles were of high quality, another 11 articles were of moderate quality, and 5 articles were of low quality. As shown in Fig. **3**, articles of high or moderate global quality had similar quantitative quality scores. In contrast, articles of a low global quality had lower quantitative quality scores than the others.

**Figure 2:** Quality scores of articles by year of publication.

**Figure 3:** Quality scores of articles by global quality classification.

#### **Design and Study Participants**

Fifteen studies used a cross-sectional design, 2 used a repeated-measures design, 2 used a longitudinal design, and 8 used a combination of these designs. Four of the studies that used a mixed design were found to combine cross-sectional and repeated-measures designs; another four used a longitudinal design combined with either a cross-sectional (*n* = 1), case-control (*n* = 1) or a repeated-measures design (*n* = 2).

With regard to the study participants, the 27 articles included blue-collar and white-collar workers of different socioeconomic status.

#### **Exposure Measures**

Sixteen of the articles investigated associations between salivary cortisol and one exposure measure; 4 articles investigated associations between salivary cortisol and two or three exposure measures; 3 articles investigated associations between salivary cortisol and 4, 5, and 10 exposure measures, respectively. Of the 27 articles, 21 included subjective exposure measures only and 2 included only objective exposure measures. Four articles included both subjective and objective exposure measures. Of the 27 articles, 15 included subjective exposure measures that were related to the DCS model, assessing one or more of the dimensions, usually using some version of the Job Content Questionnaire [32]. Thirteen articles focused exclusively on the dimensions included in the DCS model; 4 articles focused exclusively on dimensions included in the ERI model [3]. However, 2 articles included measures relating to both the DCS model and to the ERI model. Six articles used some other subjective exposure measure. Seven articles included an objective exposure measure, and 4 of these articles combined the objective exposure measure with a subjective measure. Often, the objective exposure measures were based on some kind of administrative assessment of workload, such as time pressure, conflicting tasks or insufficient number of nurses in an emergency ward.

### **Cortisol in Relation to Work Stress Exposure**

In the 27 articles, 185 analyses of associations between a cortisol measure and a psychosocial exposure measure were done. However, this number includes overall analyses only; the total number of analyses increased if, for instance, sex specific analyses were included. Of these 185 analyses, the distribution of statistically significant associations with respect to the 3 main categories of cortisol measures evaluated in Table **2** were as follows: 20/97=21% of the single time point measures, 14/48=29% of the deviation measures, and 8/40=20% of the Area Under the Curve (AUC) measures. The proportions for the different subcategories of cortisol measures included in Table **2** were as follows: single time point measures a1 (4/30=13%), a2 (6/9=67%), a3 (5/22=23%), a4 (1/21=7%), a5 (4/14=29%), deviation measures b1 (6/21=29%), b2 (4/10=40%), b3 (4/17=24%), AUC measures (increase/ground) c1 (4/17=24%; 3/6=50%), c2 (0/1; 1/9=11%), c3 (0/1; 0/6). In total, 42 (23%) of the 185 analyses showed statistically significant findings. Of these significant results, 29 (69%) were positive with a higher cortisol value being associated with a higher stressor exposure, and 13 (31%) were negative. The negative relationships were evenly scattered across the different categories of cortisol measures.

With regard to the different types of cortisol measures, Table **1** makes it clear that some ways of statistically analyzing salivary cortisol are new to the area, focusing on salivary cortisol and exposure to psychosocial work stressors. For instance, research including the deviation in the morning, that is, the Cortisol Awakening Response (CAR), was published in 2004; research investigating the AUC was published in 2007. In addition, studies on other cortisol measures, such as suppression of cortisol after intake of dexamethasone, are limited. Such limitations may be due to practical reasons and restrictions associated with field studies. Of the 27 articles, only 1 article included a dexamethasone test. The findings showed a higher suppression related to high work stress (lower reward, higher burnout and vital exhaustion), which according to the authors was interpreted as a heightened HPA axis negative feedback.

Table **2** was structured to include the main dimensions in the 2 predominant models of work stress (DCS and ERI) that formed the basis for the literature searches. Table **2** also shows the results for different dimensions of psychosocial work stress. The proportion of significant relationships between different salivary cortisol measures and measures of psychosocial exposure were as follows: 10/51 (20%) for imbalance indices, 13/57 (23%) for demand dimensions and 12/61 (20%) for resource dimensions. Some measures of work stressors (*e.g.*, referring to high levels of perceived stress at work) did not obviously fit into 1 of these 3 main dimensions and these were labeled "other stress measures". Of these other stress measures, 7/16 (44%) were significant.

### **Effects of Design**

Study design is a factor that may be related to the number of significant findings. For instance, compared with cross-sectional designs, repeated-measures designs are likely to yield more robust measures and increase statistical power even with smaller sample sizes, which in turn increases the chances of detecting significant effects. Considering this, an initial intention was to investigate the number of statistically significant findings in relation to study design. However, few articles included repeated-measures designs or longitudinal designs; of these, the number of significant findings was limited.

**Table 3:** Quality criteria according to Chida and Steptoe [4] and own additional criteria


#### **60** *The Role of Saliva Cortisol Measurement in Health and Disease Karlson et al.*


W55/M43

Decision latitude and

demands, and job strain score,

dichotomized by cut-off >1

#### *Psychosocial Work Stressors and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **61**


A (1 p) = addressed (*e.g.*, mentioned in text but not included in statistical analyses). 

C (2 p) = controlled statistically (covariate).

E (2 p) = examined group differences (*e.g.*, *<sup>t</sup>*-test differences in smoking between 2 groups but confounder not included as covariate in subsequent analyses). 

O = objective (exposure definition). 

S = subjective (exposure definition).

#### **Effects of Quality**

With regard to the relationships between quality and significant findings, the proportion of significant findings varied somewhat depending on the quality of the study. In articles with a quality score below the median score of 10, 13/48 (34%) of the associations between salivary cortisol measures and any psychosocial exposure measure were significant. The corresponding figure for articles with a quality rating above the median score was 29/147 (20%). Similar analyses with respect to the global quality evaluations showed that of the high quality articles, 15/61 (25%) of the associations were significant. For the moderate and low quality articles, the proportions of significant findings were fairly similar to that of the high quality articles 16/70 (23%) for moderate quality and 11/54 (20%) for low quality.

#### **DISCUSSION**

The findings from this systematic review of 27 articles including 185 analyses investigating linkages between different measures of salivary cortisol and psychosocial work stressors showed that most of the published findings were nonsignificant. This pattern of findings suggests that there are no consistent associations between salivary cortisol and exposure to psychosocial work stressors among working individuals. However, along with the nonsignificant results, a number of significant findings showing an association between salivary cortisol and psychosocial work stressors was identified. Most of these significant relationships were in a positive direction showing that higher cortisol levels were associated with a higher degree of psychosocial work stress. These findings suggest that, in most cases, exposure to psychosocial work stress is far from severe and the increase in cortisol levels represents a normal, healthy activation response. According to CATS and the Allostatic Load Model, this response pattern should not induce any health problems. More specifically, it seems that most working individuals manage to cope with psychosocial work stressors or recover from the strain associated with psychosocial work stressors before the strain has any measureable effects on salivary cortisol levels resulting in changes in salivary cortisol patterns. With regard to different psychosocial work stressors, deviations in salivary cortisol are perhaps more sensitive to some types of psychosocial exposure than others. The present review of psychosocial work stressors focused on a specific and limited number of exposures. As a result of this focus, a limited number of articles was reviewed and the data obtained from these articles did not allow for a fine-tuned analysis to differentiate between different work stressors. The psychosocial work stressors were roughly categorized into three different groups including concepts relating to either demands, resources, or imbalance. At this level of analysis there were no differences with regard to the ratio of significant relationships. In addition, there was a fourth category labeled "other stress measures". In contrast to the other 3 categories, which mainly assessed perceived stressors, this fourth category included measures assessing perceived work stress in terms of responses to perceived stressors. This fourth category including other stress measures had a higher ratio of significant relationships than the other categories.

The overall findings and lack of robust significant associations may also be related to methodological and statistical strategies. For instance, strategies for sampling saliva and methods of statistically analyzing derived cortisol measures, such as CAR and AUC, can be hypothesized to influence the results. Of the different methods of statistically analyzing salivary cortisol measures, single points in time at waking, midday, and evening were the most common measures in the 27 articles reviewed, followed by deviation measures in the morning (*i.e.*, CAR); the other methods of statistically analyzing cortisol were less frequent. There was a fairly high ratio of significant findings for a few of the cortisol measures, including single measures in the morning and AUCground. But these measures were less frequently used and cannot be concluded to be more successful than any other measure. Moreover, as the overall ratio of significant relationship was quite low, the number of analyses for each cortisol measure was too small to reveal any clear pattern with regard to the more successful strategies. In addition, the distribution of significant relationships was comparable across the main categories of cortisol measures applied (*i.e.*, single time point measures, deviation measures, and AUC measures), which indicates that no clear pattern emerged with regard to comparisons between static and dynamic cortisol measures. However, some of the cortisol measures, such as the dynamic CAR and AUC measures, have not been used until recently, which means that there are far fewer published articles including these measures compared with articles including single points in time. The quality of the published articles has increased over the years. Recent publications are of higher quality than older articles. Perhaps some of the findings reported in earlier research are related to inadequate sampling procedures or to not controlling for potential confounders. This reasoning is supported by the quality evaluation in this systematic review, which showed that the proportion of significant associations between psychosocial work stressors and salivary cortisol was higher in low quality articles.

Apart from methodological and statistical strategies, the nonsignificant findings presented in this review may be related to the homogeneity of the groups studied. Many of the field studies focus on individuals working in a specific organization (*e.g.*, health care), at a specific location (*e.g.*, a hospital) and often include only one group of employees (*e.g.*, nurses). This means that the study participants were homogeneous with respect to occupation but also in terms of psychosocial exposure. For instance, study participants working in the same organization and performing similar work tasks can be classified as exposed to high or low job strain. Typically, this classification is empirically based, which means that the contrasts between individuals in homogeneous groups are too small to allow small group differences in cortisol excretion to be detected. This means that for some of the articles, the many nonsignificant findings may be attributable to a lack of statistical power when comparing subgroups of the full sample studied; in other cases, small groups may have produced significant chance findings. Also, with regard to cortisol, the interindividual variation is very high, which further reduces the power to identify significant associations with psychosocial factors such as work stress. In addition, a certain degree of interindividual variation in the self-report measures of psychosocial stressors will contribute to reduce the ability to detect associations. It may be noted that there is usually a lack of temporal match between cortisol and self-report data, the latter mostly being a global rating based on the individual's mental representations of a longer period of time. However, the findings from the present review are in line with research on work stress and urinary cortisol [37].

From a biological perspective, cortisol responses seem to be mobilized during situations characterized by extreme demands and efforts such as child birth [33, 34] and sailing around Cape Horn [35]. However, the psychosocial work stressors included in this systematic review are not of extreme intensity. Instead, psychosocial work stressors can be considered to be of mild to moderate intensity. Ambulance drivers, policemen, and fire fighters do experience intensive work stress during emergency situations. The present review includes studies on intensive care nurses [10, 31], another group that has to deal with intensive bouts of psychosocial work stress. The articles on intensive care nurses examined unexpected emergency situations and these situations did elicit clear cortisol responses. In contrast, the milder psychosocial work stressors experienced daily by most working individuals are unlikely to elicit any clear cortisol responses. The effects are likely to reflect the relatively small and inconsistent effects found in this review. In addition to daily psychosocial work stressors being of mild to moderate intensity, there is also a potential habituation effect. Specifically, working individuals who are repeatedly exposed to mild to moderate psychosocial work stressors often adapt to the stress. From a biological perspective, this adaptation involves a successive reduction of the cortisol response. This line of reasoning draws on findings from experimental stress provocation research [36] showing habituation effects on cortisol secretion. According to CATS this means that most workers have adapted positive expectancies regarding the outcome from exposure to their work demands. Another factor to keep in mind when trying to explain associations between psychosocial work stressors and cortisol is that cortisol, along with other corticosteroids, reduces the negative feelings associated with a stressor. Accordingly, study participants experiencing intensive psychosocial work stress may actually underreport stress levels due to biological mechanisms. This can be illustrated by the study of intensive care nurses in which the objectively recorded emergency situations were associated with increased concentrations of salivary cortisol but not to self-reports of stress [10]. Similar findings have been reported among ambulance service personnel [38]. These findings were based on a repeated measures design, which is often assumed to be more sensitive in detecting effects than crosssectional designs. However, on the basis of this review it is not possible to draw any conclusions on the effects of study design.

In view of previous research showing clear linkages between exposure to psychosocial work stressors and various cardiovascular disorders, it is perhaps surprising that there are no consistent associations between psychosocial stress at work and salivary cortisol. But compared with the research on cardiovascular disorders, which is based on large samples to allow the effects of psychosocial work stressors on cardiovascular events and mortality to be delineated, the research on salivary cortisol is based on smaller samples and often of a different character. For instance, the studies reviewed here included working individuals with no severe health problems. Most articles are based on young and middle-aged working individuals. This means that longitudinal studies of salivary cortisol, psychosocial work stress and cardiovascular events and mortality are needed before discarding salivary cortisol as one of the key bodily mechanisms involved in the development of work-related cardiovascular disorders. The findings from the present systematic review show no consistent support for cortisol as one of the key bodily mechanisms involved in the development of various disorders related to psychosocial work stress. Perhaps this is due to cortisol not only being related to the exposure to psychosocial stressors but also the fact that the acute and long-term responses of the Hypothalamo-Pituitary-Adrenal (HPA) axis and the associated secretion of cortisol differ. Such differences relating to habituation are seldom discussed in field studies of working individuals. Most of this research focuses on examining the linkages between various aspects of psychosocial work stress and different cortisol measures without taking into account habituation effects other than years in employment or similar factors providing secondary measures of physiologic habituation. However, some significant associations were found. Exposure to high work stress was more often associated with high cortisol levels than low cortisol values. This indicates that when found, the linkages between psychosocial work stress and cortisol levels were associated with a physiologic activation rather than with a downregulation of the HPA axis activity. According to CATS and the Allostatic Load Models physiologic activation represents a normal, healthy response if followed by deactivation after work.

The number of studies reviewed was fairly small (*n* = 27), which means that additional research is needed before drawing any firm conclusions on associations between salivary cortisol and psychosocial work stressors. However, future studies would benefit from a careful evaluation of the positive and negative effects of the study design, issues relating to statistical power, relevant confounders, instructions given to study participants, the timing of salivary sampling, and the number of samples a study participant is asked to collect. All these factors are likely to influence the findings and quality of the research.

#### **CONCLUSIONS**

The present systematic review of 27 articles investigating the associations between cortisol and psychosocial work stressors showed that there was a large proportion of nonsignificant findings, with no strategy for sampling saliva or statistically analyzing the data that was superior to others. Some significant associations were found and these showed that exposure to high work stress was more often associated with high cortisol values than low values. This indicates that the linkages between psychosocial work stress and cortisol levels are more likely to be associated with normal physiologic activation than with a downregulation of HPA axis activity. With regard to the nonsignificant findings, it is possible that most of the articles reviewed examined exposure to stress of mild to moderate intensity and that the groups contrasted were too homogeneous, thus giving too little variation to reveal effects.

#### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **CHAPTER 4**

# **Perceived Stress, Psychological Resources and Salivary Cortisol**

**Christina Halford1,\*, Ingibjörg H. Jonsdottir2 and Frida Eek<sup>3</sup>**

*1 Researcher at the Department of Public Health and Caring Sciences, Uppsala University, Box 564, 751 22 Uppsala, Sweden; <sup>2</sup> Associate professor at the Institute of Stress Medicine, Gothenburg, Sweden and <sup>3</sup> Associate professor at the Division of Occupational and Environmental Medicine, Lund University, Sweden.* 

**Abstract:** The aim of this chapter was to analyze associations between measures of cortisol in saliva with measures of perceived stress, using the Perceived Stress Scale (PSS), and of psychological resources in terms of mastery, locus of control, self-esteem and sense of coherence. Only studies on healthy individuals were included and cortisol measures were grouped into single time point measures, deviation measures, Area Under the Curve (AUC), laboratory test responses, and dexamethasone suppression. For both Perceived Stress Scale (PSS) and for psychological resources, most results of associations with saliva cortisol were nonsignificant particularly for single measures and for cortisol awakening response. For PSS the largest proportion of significant findings (38%) was seen for morning AUC, however with conflicting results. For psychological resource constructs, mastery and sense of coherence were related to lower cortisol level at baseline in standardized rest and high mastery was related to steeper diurnal slope in two studies. For self-esteem, no associations showed significant results. Differences in findings may to a large extent be dependent on theoretical assumptions made and methods used.

**Keywords:** Salivary cortisol, perceived stress, mastery, locus of control, sense of coherence, single time point measures, deviations measures, area under the curve, laboratory test, dexamethasone.

#### **INTRODUCTION**

Since the 1960s it has been virtually unanimously acknowledged that psychological stressors are among the most potent stimuli of the Hypothalamic-Pituitary-Adrenal (HPA) axis [1]. Although the importance of psychological factors for HPA axis activity is undisputed, research on associations between HPA axis activity and subjectively perceived stress as well as associations between HPA axis activity and different psychological resource constructs is inconclusive, and knowledge is still very limited.

The Perceived Stress Scale (PSS) [2] is a widely used instrument for measuring the perception of stress. It measures the degree to which situations in one's life were appraised as stressful during the last month. The purpose is to assess how unpredictable, uncontrollable, and overloaded respondents find their lives, by asking how often the respondent has experienced certain feelings and thoughts during the last month. Some items also cover queries about current levels of stress experienced. The questions are of a general nature and hence not specific to any subgroup population.

The concept of psychological resources refers to psychological factors perceived as potentially protective of well-being and health in the face of stressor exposure. The present literature study takes its departure point from four well-established psychological resource constructs: self-esteem, mastery, locus of control, and sense of coherence [3]. Self-esteem refers to a relatively stable sense of overall self-worth; a sense of being a person of value, and an acceptance of personal strengths and weaknesses [4]. The Concept of Locus of Control (LoC) refers to an individual's general beliefs regarding their ability to influence events [5]. Mastery refers to generalised beliefs of control in terms of "the extent to which people see themselves as being in control of the forces that importantly affect their lives" p. 340 [6]. Sense of Coherence (SOC) refers to a construct based on stress theory developed by Antonovsky, according to whom SOC represents a global orientation that reflects the extent to which stressors in the internal and external environment are

**<sup>\*</sup>Address correspondence to Christina Halford:** Researcher at the Department of Public Health and Caring Sciences, Uppsala University, Box 564, 751 22 Uppsala, Sweden; Tel: +46 18 471 6572; Fax: +46 18 471 6675; E-mail: christina.halford@pubcare.uu.se

perceived as (1) structured, predictable and comprehensible, (2) challenges worthy of engagement and investment, and (3) the extent to which internal or external resources needed to handle stressors are perceived as available [7].

## **AIM**

The aim of the present chapter was to review existing literature on association between measures of cortisol in saliva and psychological constructs in terms of perceived stress and psychological resources (self-esteem, mastery, LoC, SOC). The evaluation against different measures of saliva cortisol is based on the question of whether seemingly divergent results may be functions of differences in methods used and theoretical assumptions made.

## **METHOD**

Electronic searches were performed in PubMed and PsychInfo data bases, covering the period up to October 1, 2009. Searches on perceived stress were based on the following search terms: (cortisol AND saliva\* AND "perceived stress"). The search terms used for psychological resources were (cortisol AND saliva\* AND self-esteem, "locus of control", mastery, "sense of coherence"). English-language, full-length articles, published in peer-reviewed journals, based on adult study populations, reporting direct statistical analyses of associations between cortisol in saliva and measures of perceived stress or psychological resources were included. Results based on patient populations, or with the primary aim of investigating associations in pregnant women, were excluded. When studies included analyses of healthy control groups, results from analyses relating to the healthy control groups were included.

Searches based on the search terms (cortisol AND saliva\* AND "perceived stress") generated 95 papers. For perceived stress, only papers measuring perceived stress based on the PSS [2] were included. Based on a review of the titles and abstracts, and when relevant by reading the full-length article, 18 papers were finally included.

Searches on associations between cortisol in saliva and psychological resource constructs based on the search terms (cortisol AND saliva\* AND self-esteem, "locus of control", mastery, "sense of coherence", respectively) generated 54 papers. Based on a review of the titles and abstracts, and when relevant by reading the full-length article, based on the inclusion and exclusion criteria, 11 papers were finally included.

In the following analyses, findings were considered significant if p values were <0.05. As most studies involved small numbers of participants and thus seemingly low statistical power, we also included marginally significant results (0.05<p<0.10) if they were reported; these are denoted by arrows in parentheses in the tables.

### **RESULTS**

### **Perceived Stress**

In total, 18 articles on possible associations between cortisol in saliva and the PSS fulfilled inclusion criteria [8-25]. A brief summary of the results (indicated as a positive association, a negative association, or a nonsignificant finding) is presented in Table **1a**. Study design, statistical approach, main results and discussion for each of the 18 articles are presented in Table **2a**.

Generally, few significant associations were found between PSS and cortisol in the papers examined. Of the 18 articles, six found significant associations, three articles reported trends towards significant associations (0.05<p<0.10) and nine failed to find any significant associations between PSS and cortisol measures. Most studies reported more than one measure of cortisol, ranging from one to four.

For single time points, 1 out of 14 analyses showed significant results. The significant finding was seen for samples taken at midday, whereas three other studies showed nonsignificant findings, *i.e.*, 1/4 (25%) of midday single measures showed significant results. Among deviation measures, all of which measured diurnal deviation, 2/13 results were significant and another two marginally significant. Thus, among the nine measures of diurnal variation, four or 44% did report significant or marginally significant results. For measures of Area Under the Curve (AUC), 3/8 (38%) analyses showed significant results; morning (increase), 1/2 (50%); morning (ground), 2/2 (100%); morning-evening (increase), 0/2 (0%); and laboratory (increase), 0/1 (0%). Only one study was included with results from a dexamethasone challenge test, reporting 1/1 (100%) significant result.

**Table 1a:** Summary of main findings of associations between measures salivary cortisol and Perceived Stress Scale (PSS) sorted by year of publication


*Abbreviations:* a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity (first column)/recovery (second column); c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, C-S, cross-sectional. Arrows in parentheses denoting marginally significant findings.

a Significant effect only for women with high internalized racism.

b Marginal effect (*p*=0.053), and significant (0.02 when suspected non-adherents were excluded).

#### **70** *The Role of Saliva Cortisol Measurement in Health and Disease Halford et al.*

**Table 2a:** Studies on PSS sorted by year of publication 



#### **72** *The Role of Saliva Cortisol Measurement in Health and Disease Halford et al.*

#### 


#### *Perceived Stress, Psychological Resources and Salivary The Role of Saliva Cortisol Measurement in Health and Disease* **73**


#### **74** *The Role of Saliva Cortisol Measurement in Health and Disease Halford et al.*

#### 


#### *Perceived Stress, Psychological Resources and Salivary The Role of Saliva Cortisol Measurement in Health and Disease* **75**


Analyses/methods/study design described in relation to relevance for examinations of association between PSS and cortisol only. May differ from main study presentation. 

\* Association between cortisol and PSS was not the primary aim; the results are presented only for analyses regarding this association. 

\*\* Analyses/results only presented for healthy subgroups, and regarding relation between PSS and cortisol.

*Abbreviations:* AURC, Area under response curve; BMI, body mass index; DAUC, Day(time) area under curve (12 hours); EIA,enzyme immunoassay; ELISA, enzyme-linked immunosorbent assay; HPLC, high performance liquid chromatography; HS, high stress; INR, internalized racism; LS, low stress; MANOVA, multivariate analysis of variance; OC, oral contraceptives; RIA, radioimmunoassay; SES, socioeconomic status; SIST, Stress Inducing Speech Task; TBS,:VE, vital exhaustion.

The specific findings among significant results were as follows. The only significant finding for a single time point was when groups with high and low PSS were compared; a lower mean cortisol level in the afternoon was found for the high PSS group [23]. With regard to diurnal deviation measures, in a study of school employees and a study including African-Caribbean women, high PSS levels were associated with a flatter cortisol curve [15, 20]. In line with these results, in a study of ranchers high PSS was related to smaller daytime decline [24]. However, this finding had marginal statistical significance (*p*=0.053), and the instrument used was not the typical PSS scale. In contrast with these results, in a study on a group of healthy women (control study group), a tendency (*p*=0.07) to a positive relationship between slope and PSS was seen [14].

In a study of psychology students [16], morning AUC with respect to ground was inversely and independently related to PSS. Similarly, in a normal population sample, the high PSS group had lower cortisol secretion during the morning (AUC with respect to increase) compared with a group with low PSS [23]. In contrast, in a group of healthy students, persons with higher AUC during morning hours reported significantly more stress on the PSS measure [12]. In a group of teachers, high levels of perceived stress were associated with higher cortisol levels after awakening on the day after dexamethasone intake, but no associations during days without previous dexamethasone intake [10].

### **Psychological Resources**

Eleven articles fulfilled the inclusion criteria [26-36]. The proportion of significant associations (in any direction) observed in the 11 papers investigating associations between different psychological resource construct variables and cortisol was 1/9 (11%) for single measures, 2/4 (50%) for standardized baseline measures, 4/22 (18%) for deviation measures, and 3/8 (37.5%) for AUC measures. A brief summary of the results (indicated as a positive association, a negative association, or a nonsignificant finding) are presented in Table **1b**. Study design, statistical approach, main results and discussion for each of the 11 articles are briefly presented in Table **2b**.

## *Questionnaire for Competence and Control*

Associations between combined measures of self-esteem and locus of control (SEC) were investigated in three studies, based on study populations comprising 16 men, 20 men, and 52 women and men, respectively [26, 27, 31].

One study examined associations between SEC and cortisol AUC in relation to a laboratory stress test in healthy students. An inverse association between cortisol AUC and the subscale "self-concept" (*r*=-0.69, *p*=0.003), and a positive association between cortisol AUC and the subscale "chance" (*r*=0.51, *p*<0.05) were observed. There were no significant associations between cortisol AUC and the subscales "internality" or "powerful others" [26].

Two studies investigated deviations in cortisol levels during a laboratory stress test in relation to SEC in healthy young men [27, 31]. One study reported a significant inverse effect of SEC on cortisol reactivity, *i.e.*, low SEC was associated with a stronger cortisol response [31]. In the other study, a significant negative effect of selfesteem on cortisol levels was observed only in a low self-esteem/failure condition subgroup [27].

**Table 1b:** Summary of main findings of associations between measures salivary cortisol and psychological resources sorted by year of publication. Note; For mastery and self esteem, first column under b4 denotes baseline levels before stress



*Abbreviations:* a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity (first column)/recovery (second column); c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, C-S, cross-sectional; E Experimental; SC; Positive selfconcept; INT internality, POC Powerful others; FAT Chance.

#### **78** *The Role of Saliva Cortisol Measurement in Health and Disease Halford et al.*

**Table 2b:** Studies on psychological resources sorted by year of publication 


#### *Perceived Stress, Psychological Resources and Salivary The Role of Saliva Cortisol Measurement in Health and Disease* **79**


#### **80** *The Role of Saliva Cortisol Measurement in Health and Disease Halford et al.*

#### 



*Abbreviations:* ANSIE, Adult Nowicki Strickland internal external control scale; BP, blood pressure; MI, myocardial infarction; PC, Perceived control; SEC, self-esteem and locus of control (questionnaire of competence and control; Krampen 1991).

#### **Locus of Control**

Analyses of associations between LoC and cortisol were investigated in two studies, based on study populations of 100 and 48 healthy volunteers and students (women and men), respectively [28, 30].

Both studies investigated cortisol response to a laboratory stress test. In one study, LoC was inversely associated with cortisol reactivity to the cold pressor test, *i.e.*, higher scores for internal LoC were associated with poorer cortisol response (*r*=0.22, *p*<0.05), but there were no significant associations between LoC and cortisol response to a mental arithmetic test [28]. In the second study, no significant associations were observed between LoC and cortisol levels in relation to a stress induction task based on exposure to aversive noise and a working memory math test (*r*=-0.28, *p*=0.08). However, when comparing participants in a perceived control and a non-control condition, internal locus of control in the perceived control condition was associated with a stronger cortisol response (*r*=0.29, *p*<0.01) while in non-control condition, cortisol response did not vary as a function of locus of control [30].

#### **Mastery**

Associations between mastery and cortisol (single time point measures, deviation measures, and cortisol AUC) were investigated in four studies based on normal populations consisting of 183 men [32], 781 women and men [33], 257 women and men [34], and 59 women [35].

Associations between mastery and single time point measures were investigated in two studies (five measures) [32, 34]. In one of these studies, a significant positive association between mastery and cortisol at awakening was reported (*r*=0.13, *p*<0.05) [34]. Single cortisol measurements based on morning, afternoon, and evening samples, were not statistically significantly associated with mastery in either of the two studies [32, 34].

All four studies investigated associations between mastery and deviation measures [32-35]. Two studies examined associations between mastery and morning deviation; none of these were significant [34, 35]. Three studies investigated associations between mastery and diurnal deviation, of which two studies reported significant positive associations between mastery and diurnal slope (*r*=0.12, *p*<0.05) [33], (*r*=0.13, *p*<0.05) [34], and one study found no significant associations [35].

One study investigated cortisol response in relation to laboratory stress, and reported an inverse association between mastery and baseline cortisol (*r*=-0.16, *p*=0.04), with no significant associations observed between mastery and cortisol reactivity [32].

Associations between mastery and cortisol AUC based on several samples over the day were investigated in two studies [33, 35]. One study found no significant associations between mastery and cortisol AUCground [33]; the other study reported an inverse association between mastery and AUC with respect to ground (*r*=- 0.29, *p*<0.05) but not to AUC with respect to increase [35].

### **Self-Esteem**

Associations between cortisol and self-esteem were investigated in four studies, based on study populations of 208 women and men [29], 183 men [32], 257 women and men [34], and 48 women [36].

None of the studies reported any significant associations between self-esteem and cortisol. Two studies reported analyses based on single time point measures at awakening, one of which also included an evening measure [34, 36]. All four studies investigated associations between self-esteem and cortisol deviation measures [29, 34, 36], three of which investigated deviations in morning or morning-evening cortisol levels; two studies investigated cortisol deviation in relation to a laboratory stress test. One study reported results based on analyses of associations between self-esteem and awakening cortisol AUCground levels.

#### **Sense of Coherence**

One study, based on a study population of 183 men, investigated associations between sense of coherence and salivary cortisol [32]. Analyses were based on a laboratory stress test. Inverse significant associations between sense of coherence and baseline levels of cortisol (*r*=-0.18, *p*=0.02) before a laboratory stress test, and no significant associations between sense of coherence and cortisol reactivity to the laboratory stressor were reported.

#### **DISCUSSION**

#### **Perceived Stress**

There is a large proportion (more than half of the studies) of nonsignificant findings among the statistical analyses reported. However, the largest proportion of significant findings was seen for the AUC measure with significant associations for 38% of the studies (3/8 papers); negative associations were found in two studies (morning AUC) and a positive association in one study (morning AUC with respect to ground). For diurnal deviation measures, 44% were significant or marginally significant; three of these had a negative association with diurnal deviation.

The inconsistency found for diurnal deviation arose from one study with marginally significant results, suggesting that stronger diurnal deviation is related to high PSS. Two studies showed the opposite findings. The positive associations were seen among healthy controls in a case-control study of patients with breast cancer (in which the patients showed flat diurnal deviation) [14]; while negative associations were seen among people in a stressful context. Significant findings were reported for employees in a primary school, who the authors characterized as being under "chronic stress" [20] and among African women with internalized racism [15]. Marginally significant findings were reported among Colorado ranchers [24]. These data do underpin the importance of subject's earlier experiences and context when evaluating cortisol response to an acute stress.

This relationship between long-term and acute stress is discussed in several of the papers in which the nature of the PSS scale was discussed as a potential explanation for the lack of significant findings. Van Eck *et al.* [8] suggest that individual differences in current distress, especially anticipatory distress, may be more important determinants of cortisol secretion than PSS level, because the latter is a measure of more long-term distress. This is also in agreement with Schwartz *et al.* [13] and Putterman *et al.* [17], who discuss that PSS may not reflect current stressful events but, rather, stress levels over the previous month; the cortisol sampling may reflect changes in HPA activation over approximately 30 min. Thus, the PSS may tap stress more generically, and may not be so sensitive to the more subtle demands associated with cortisol levels.

Hellhammer *et al.* [37] concluded that a missing or poor association between perceived stress and salivary cortisol is not surprising, considering the complex interplay of neurobiological events that link perceived stress to HPA activation. Methodological difficulties related to the assessment of perceived stress by selfreport instruments are also mentioned as possible explanations for a lacking covariance of perceived stress and salivary cortisol. Because several additional variables, such as adrenal sensitivity, capacity, and cortisol binding, also affect salivary cortisol levels, perceived stress can only be expected to be moderately associated with cortisol.

We could not find any apparent similarities for study design or methods, either in the studies that found significant associations or in the studies that failed to show effects. In several of the papers, the analyses of associations between cortisol and PSS were not the primary aim. Hence, the lack of significant results was sometimes not further discussed.

Although the present review gives little empiric support to the hypothesis that PSS is related to HPA activity, earlier studies have demonstrated that chronic stress does affect an individual's ability to respond to acute stress [8].

#### **Psychological Resources**

Based on the inclusion and exclusion criteria adopted, 11 studies were identified in which direct associations between cortisol in saliva and psychological resource constructs were investigated.

Most of the statistical analyses on associations between different measures of cortisol and psychological resource constructs were nonsignificant. This was especially so for single time point measures and for the resource construct self-esteem. Associations between psychological resource constructs and a standardized single measurement, in terms of baseline measures before stress testing, were reported in two studies, in one of which mastery and SOC were significantly associated with lower baseline cortisol levels [32]. There were no associations observed between cortisol and self-esteem in either study [32, 36].

Regarding deviation measures, 0/5 measures of Cortisol Awakening Response (CAR) were significant for mastery and self-esteem alike. For diurnal deviation measures, 2/3 studies reported significant associations with mastery. In two studies, higher levels of mastery were associated with a steeper cortisol diurnal slope (one study of 781 participants based on results from one sampling day [33], and one study of 257 participants based on three sampling days [34]) and nonsignificant findings for self-esteem. In laboratory stress testing, 2/9 analyses were significant, both showing an inverse association between psychological resources and cortisol response, one for SEC and one for LoC [28, 31].

Thus, consistencies in results on the one hand relate to nonsignificant results regarding associations between psychological resources and cortisol levels based on single time point measures or CAR. On the other hand, consistencies are seen for significant findings of lower cortisol at standardized rest for baseline, and for steeper diurnal deviation measures.

The present studies highlight a number of methodological issues of potential importance for further investigations on associations between psychological resources and cortisol. Cortisol is secreted in response to the daily life cycle of activity and rest and in response to internal and external stressors. The sensitivity of the HPA axis to internal and external events is reflected in large intra- as well as inter-individual variability of cortisol [37, 38]. Most of the studies identified were based on relatively small study populations. Power calculations were generally not presented, and whether or not nonsignificant results are due to lack of power or lack of effect size can not be determined. All studies on psychological resources were on normal populations, and thus medical conditions of participants did not seem to explain any of observed differences in results.

Aggregation of cortisol data over several days has been demonstrated to lead to increasingly consistent patterns in associations between psychological measures and cortisol, and it is suggested that data aggregation is needed for identification of the trait component of the stress response [26, 38]. In a recent study on the reliability of CAR, the results suggest that measurements during 2 days are necessary for reliable AUC with respect to ground measures, and 6 days are necessary to achieve reliable AUC with respect to increase measures, with state factors biasing data based on single day measures [38]. In parallel, in studies of diurnal deviation, evidence of increased reliability of results was seen comparing mean values over 3 days compared with single day values [26]. Of the six studies using ambulatory saliva sampling, three studies sampled over 1 or 2 days.

In two of the studies, nonsignificant results were reported for associations between psychological resources and cortisol in the study population as a whole. Significant associations were observed in the subgroup analyses [27, 30], in one study, only if the subject was in a failure condition [27], and in another study, only among participants who believed they were in control of the laboratory stressor [30].

## **CONCLUSIONS**

For both PSS and for psychological resources, most results of associations with saliva cortisol were nonsignificant particularly for single measures and for cortisol awakening response. For PSS the largest proportion of significant findings (38%) was seen for morning AUC, however with conflicting results. For psychological resource constructs, mastery and sense of coherence were related to lower cortisol level at baseline in standardized rest and high mastery was related to steeper diurnal slope in two studies. For selfesteem, no associations showed significant results. Differences in results may to a large extent be dependent on theoretical assumptions made and methods used.

#### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **CHAPTER 5**

# **Biological Markers and Salivary Cortisol**

#### **Åse Marie Hansen1,\*, Lars-Gunnar Gunnarsson2 , Anette Harris3 , Nanna Hurwitz Eller<sup>4</sup> , Peter Garvin5 and Anne Helene Garde6**

*1 Senior reseacher at the National Research Centre for the Working Environment, Lersø Parkallé 105, 2100 København Ø, Copenhagen, Denmark; <sup>2</sup> Medical Doctor in Occupational and Environmental Medicine at Örebro, University Hospital, Örebro, Sweden; <sup>3</sup> Researcher at Uni Health and Department of Psychology, Bergen University, Bergen, Norway; <sup>4</sup> Medical Doctor in Occupational and Environmental Medicine at Bispebjerg Hospital, Copenhagen, Denmark; <sup>5</sup> Post-doc at the Department of Medical and Health Sciences, Linköping University, Sweden and <sup>6</sup> Senior reseacher at the National Research Centre for the Working Environment, Copenhagen, Denmark.* 

**Abstract:** This chapter focuses on salivary cortisol in relation to biological markers. Specifically, associations with conventional cardiovascular risk factors and metabolic abnormalities (body mass index, waist circumference, waist/hip ratio, lipid status, glucose, blood pressure, heart rate and heart rate variability), markers related to inflammation (C-reactive protein, cytokines and tumor necrosis factor-alpha) and other stress hormones (adrenaline and noradrenaline) were studied. The focus was on healthy adult populations; studies on patient populations and pregnant women were excluded. Studies on genome variations and pharmacological interventions were also excluded. After meeting all exclusion criteria, 42 papers remained. In total, 273 associations between salivary cortisol and any of the markers mentioned were studied, comprising 241 associations on metabolic abnormalities, 30 on inflammation, and 2 on stress hormones. Of the salivary cortisol measures reported for evaluations of all markers tested were 136 (49%) single time points, 100 (37%) deviations, 36 (13%) AUC, and 1 (1%) dexamethasone test. Of these, 72 (26%) were statistically significant, and 201 (74%) indicated non-significant findings.Several of the markers tested showed low or no association with any of the measurements of salivary cortisol. The number of studies exploring the association between cortisol in saliva and markers for inflammation is low, which limits the possibility of interpretation. The number of studies on adrenaline and noradrenaline is also low. To sum up, the proportion of non-significant findings was considerable. This may be due to a large number of studies with relatively small study populations. This is true for metabolic abnormalities, markers related to inflammation as well as other stress hormones. Further studies on inflammatory markers and approaches designed to study variability in other systems in relation to cortisol variability are required.

**Keywords:** Salivary cortisol, body mass index, waist circumference, waist/hip ratio, lipid status, glucose, blood pressure, heart rate, C-reactive protein, cytokines, adrenaline, noradrenaline.

### **INTRODUCTION**

In the last decade, the technique of using ambulatory saliva sampling has become increasingly popular in field research and clinical studies. The non-invasive method is easy to administer and analyze, and therefore allows implementation in large-scale study designs. However, as with other biological, behavioral, and psychological measurements, the possibility of answering any research question is dependent on when and how measurements are made. Cortisol has considerable day-to-day and diurnal variation. Therefore, a fair number of saliva samples are needed to illustrate the general capacity of the Hypothalamic-Pituitary-Adrenal (HPA) axis. The HPA axis, a major part of the neuroendocrine system, controls physiologic response to stress and regulates many bodily processes, including digestion, the immune system, mood and emotions, sexuality, and energy storage and expenditure [1, 2].

Allostasis is the process of achieving stability, or homeostasis, through physiologic or behavioral change

**<sup>\*</sup>Address correspondence to Åse Marie Hansen:** Senior reseacher at the National Research Centre for the Working Environment, Lersø Parkallé 105, 2100 København Ø, Copenhagen, Denmark; Tel: +45 39 16 52 00; Fax: +45 39 16 52 01; E-mail: aamh@arbejdsmiljoforskning.dk

#### **88** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*

[3, 4]. This can be carried out by means of alteration in HPA axis hormones, the autonomic nervous system, cytokines, or a number of other systems. A shift in the HPA axis is generally adaptive in the short term [5]. However, it has been suggested that a long-term shift may have deleterious health effects [3, 4]. In particular, alterations in the HPA axis have been suggested as a plausible mechanism linking stress with metabolic abnormalities [6]. It has also been suggested that the immunoregulating effects of cortisol may be diminished by a long-term increase in cortisol levels [7]. Further, it has been suggested that a flat diurnal curve (low decrease from morning to evening and high evening values) may represent alack of recovery or sustained activation that may be associated to negative health outcome [8, 9], whereas a high as well as a low awakening response may reflect the individuals expectations to the upcoming day [8, 10]. The effects of cortisol are well described in several experimental studies, but it is unclear to what extent salivary cortisol in observational studies mirrors other biological measures associated with metabolic abnormalities or inflammation.

This chapter primarily aims to describe associations between measures of cortisol in saliva and other biological markers in non-clinical settings. We have focused on markers related to metabolic abnormalities (with particular interest in cardiovascular risk factors), inflammation, and other stress hormones. The evaluation of the literature was based on single cortisol measurements, the sum or mean over the day, diurnal variability, Cortisol Awakening Response (CAR), Area Under the Curve (AUC), reactivity and recovery from stress tests and the dexamethasone suppression test.

### **AIM**

To examine to what extent associations between cortisol measurements and other biological measures can be found, and which of the measurements are of highest relevance. The evaluation of the literature was based on the following question: is it possible that the seemingly divergent results of the studies involving cortisol assessments and biological markers are functions of differences in the theoretic assumptions made and methods used.

### **METHOD**

### **Search Strategies**

In a first step, an online search of the NCBI PubMed database (National Library of Medicine, National Institutes of Health, Bethesda, MD, USA-http://www.ncbi.nlm.nih.gov/PubMed) was conducted. The search covered the time period up to October 2009 (allowing e-publications if a full paper was published electronically prior to journal publication). Search terms were selected with reference to relevant PubMed terms and key words (see detailed description for each of the biological markers below), in combination with salivary cortisol in its truncated form ("saliva\*"). The limitations were set only to include studies matching "Human", "English" and "Adults" (aged 19 years old or more).

In a second step, studies on patient populations were excluded (*e.g.*, cancer, diabetes, and major depressive disorder). Studies on genome variations, pregnant women, and pharmacological interventions were also excluded.

In a third step, all articles retrieved from each search were briefly read. If no direct statistical analysis between salivary cortisol and the explored biomarker were presented in tables, figures, or text, the paper was excluded. Intervention studies (other than pharmacological) were included if associations with the biomarker of interest were present before the intervention. However, the effects on salivary cortisol in response to the intervention are not included in this review. Papers were also excluded if another (prior) publication from the same study material was already included in the evaluation.

### **Body Mass Index**

The term "body mass index" in combination with truncated salivary cortisol yielded 110 hits. In addition, 8 hits were found in the search using "metabolic." After meeting all exclusion criteria, the final number of papers was reduced to 24.

### **Waist Circumference**

The term "waist circumference" in combination with truncated salivary cortisol yielded 13 hits. In addition, 3 hits were found in the search using "metabolic", giving 16 papers. Of these, 7 papers remained after exclusion.

### **Waist/Hip Ratio**

To identify papers on cortisol in relation to cholesterol, the following search terms were used in combination with truncated salivary cortisol: "waist hip ratio" (16 hits), "waist-hip ratio" (16 hits, same as previous), "waist-to-hip" (19 hits, of which all was covered by the previous mentioned), and "WHR" (14 hits, all of which were included in the previous searches). In addition, 3 papers from the search on body mass index were included, as they presented associations between waist/hip ratio and cortisol. In total, 38 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 11 papers.

### **Cholesterol**

To identify papers of cortisol in relation to cholesterol, the following search terms were used in combination with truncated salivary cortisol: "cholesterol" (29 hits), "HDL" (18 hits, of which one was not covered by the previous search terms), "LDL" (8 hits, of which one was not covered by the previous search terms), "lipids" (56 hits), , "metabolic" (81 hits), "metabol\*" (8 hits) and "metabolite" (12 hits) and "apolipoprotein" (4 hits). In total, 131 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 6 papers.

## **Triglycerides**

To identify papers on cortisol in relation to triglycerides, the search term "triglycerides" was used in combination with truncated salivary cortisol (20 hits). "TG" as a search term did not yield any extra papers. In addition, 3 papers from the search on cholesterol were included, as they presented associations between triglycerides and cortisol. In total, 23 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 5 papers.

### **Plasma or Blood Glucose**

To identify papers on cortisol in relation to triglycerides, the search term "glucose" was used in combination with truncated salivary cortisol (68 hits); "blood sugar" as a search term did not yield any extra papers. After meeting all exclusion criteria, the final number was reduced to 5 papers.

### **Blood Pressure**

To identify papers on cortisol in relation to blood pressure, the search term "blood pressure" (216 hits) was used in combination with truncated salivary cortisol. "hypertens\*" (11 hits) was also used. In total, 224 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 14 papers.

### **Heart Rate**

To identify papers on cortisol in relation to heart rate, the search term "heart rate" was used in combination with truncated salivary cortisol (270 hits). Almost all those met the exclusion criteria. The final number was reduced to 3 papers.

For several of the parameters mentioned above, two additional papers were found [11, 12]. These two, conducted within the same research group, used a method of measuring cortisol that could not be fitted into the overview of the cortisol measurements used in this book. For that reason, these are excluded from the overview. However, they are reflected upon in the discussion on cortisol and metabolic abnormalities.

### **Heart Rate Variability**

To identify papers on cortisol in relation to heart rate variability, the search term "heart rate variability" was used in combination with truncated salivary cortisol. In total, 32 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 4 papers.

### **Interleukins and Other Markers Related with Inflammation**

To identify papers on cortisol in relation to inflammation, the search term "interleukin" (61 hits), "cytokine" (71 hits), "CRP" (5 hits), and "C-reactive protein" (13 hits) were used in different searches in combination with truncated salivary cortisol. After meeting all exclusion criteria, the final number was reduced to 11 papers.

## **Adrenaline**

To identify papers on cortisol in relation to adrenaline, the search terms "adrenaline" and "epinephrine" were used in combination with truncated salivary cortisol. "The latter did not yield any extra papers. In total, 69 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 2 papers.

## **Noradrenaline**

To identify papers on cortisol in relation to adrenaline, the search terms "noradrenaline" and "noraepinephrine" were used in combination with truncated salivary cortisol. "The latter did not yield any extra papers. In total, 72 unique papers were identified in the first step. After meeting all exclusion criteria, the final number was reduced to 2 papers.

## **RESULTS**

After meeting all exclusion criteria, 42 papers were included. In total, 273 associations between salivary cortisol and any of the markers mentioned were studied, comprising 241 associations on metabolic abnormalities, 30 on inflammation, and 2 on stress hormones. Of the salivary cortisol measures reported for evaluations of all markers tested were 136 (49%) single time points, 100 (37%) deviations, 36 (13%) AUC, and 1 (1%) dexamethasone test. Of these, 72 (26%) were statistically significant, and 201 (74%) indicated non-significant findings.

## **Body Mass Index**

### *Quantitative Analysis on the Evaluated Studies*

In the 24 studies [13-36], there were 60 analyses on the relationship with salivary cortisol (see Table **1a**). Of these, 29 were on single time points, 23 on deviations, and 8 on AUC. In total, 14 of the analyses (23%) showed significant associations with salivary cortisol, whereas the other 46 (77%) showed non-significant findings. The significant findings were mainly clustered in the following three categories:


### *Consistency of the Material*

Regarding a possible association between a lower cortisol level in the morning and BMI, there are no contradictions in the results. However, analyses suggesting such an association is derived from only five studies [20, 22, 23, 32, 34]. These studies point in the same direction, suggesting a lower cortisol level in

the early phase of the diurnal cycle. None of the studies evaluating awakening values found any associations with BMI. The possible association with a lower cortisol level in the early phase of the diurnal cycle seems to be valid regardless of the use of cross-sectional designs [23, 34] or a prospective design with cortisol levels as the outcome [22].

#### *Methodological or Contextual Explanation on Divergent Findings*

Overall, only two significant positive associations were found. One might be attributed to a subpopulation analysis [25]. The most striking contradictory finding was found in a large-scale population sample (Whitehall-II, *n*=2873) [29], in which associations were evaluated using an averaged cortisol level throughout the day based on 7 samples from awakening to bedtime.

Most of the other studies were based on fewer samples. Computing an average of 7 samples devalues the potential impact of lower levels at the early stages of the diurnal cycle. Thus, the seemingly contradictory findings may be accurate, given that values taken at time points later in the diurnal cycle were slightly higher among subjects with higher BMI (not reported in the study). On the other hand, there is no support in the other studies evaluating cortisol levels in the evening; all reported non-significant findings. Coutinho *et al.* reported a marginal significant negative association (*p*=0.063) between a sample taken at 23:00 h and BMI [27].

### **Waist Circumference**

### *Quantitative Analysis on the Evaluated Studies*

In the 7 studies [13, 21, 26, 27, 31, 37, 38] there were 16 analyses on the relationship with salivary cortisol (see Table **1a**). Of these, were 9 on single time points, 6 on deviations and 1 on AUC. In total, 4 of the analyses (25%) showed negative associations with salivary cortisol, whereas the other 12 (75%) showed non-significant findings. However, the negative findings were clustered in the same category where a lower deviation in a daily slope was associated with higher waist circumference (3/3; 100%), and a lower AUC in the morning was associated with a higher waist circumference (1/1; 100%).

### *Consistency of the Material*

Relatively few associations were found in the material. The low proportion of non-significant findings is explained by the non-significant findings for single time points. The significant findings included points in the same direction as BMI, implying that there may be an association between low values of cortisol in the morning/midday and waist circumference.

### *Methodological or Contextual Explanation on Divergent Findings*

There are no major contradictions in the overview. The reported results are either non-significant findings or negative associations between cortisol levels and waist circumference. However, all studies presented are fairly small, thus increasing the risk for beta errors.

### **Waist/Hip Ratio**

### *Quantitative Analysis on the Studies Evaluated*

In the 11 studies [13, 15-18, 22, 24, 25, 27, 29, 30], there were 31 analyses on the relationship with salivary cortisol (see Table **1a**). Of these, 15 were on single time points, 12 on deviations, and 4 on AUC. Seven of the analyses (31%) showed negative associations with salivary cortisol, 4 (13%) showed positive associations, and the other 20 (65%) showed non-significant findings. The negative significant findings were clustered in measure including morning values, either as single time points (3/4, 75%), low deviation at midday (1/1, 100%) or low AUC including morning values (1/1, 100%).

### *Consistency of the Material*

There were relatively few associations found in the material. Negative associations suggesting a lower cortisol level in the early phase of the diurnal cycle were found in 4 independent studies. The consistency of positive associations is weaker, partly because they are occur less, partly because the same group presenting positive associations also presented non-significant findings for similar cortisol measurements in different study populations [17, 24, 29].

## *Methodological or Contextual Explanation on Divergent Findings*

A similar pattern to that found for BMI emerged. Only 4 significant positive associations were found, 3 of which might be attributed to a sub-population analyses [17]. The most striking contradictory finding was found in a large-scale population sample (Whitehall-II, *n*=2873) [29], in which associations were evaluated using an averaged cortisol level throughout the day based on 7 samples from awakening to bedtime. Most of the other studies were based on fewer samples. Computing an average of 7 samples devalues the potential impact of lower levels at the early stages of the diurnal cycle. Thus, the seemingly contradictory findings may be accurate, given that values taken at time points later in the diurnal cycle are slightly higher among subjects with higher waist/hip ratio (not reported in the study). As with BMI, there is no support in the other studies evaluating cortisol levels in the evening; all reported non-significant findings [15, 17, 27].

**Table 1a:** Summary of the main findings of associations between measures of salivary cortisol and body mass index, waist circumference, and waist/hip ratio sorted by year of publication



a Significant associations only in group with obese women (BMI>=30 kg/m2 ).

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-C, Case-control; C-S, crosssectional; Exp, experimental; Pros, prospective.

#### **Cholesterol**

#### *Quantitative Analysis on the Evaluated Studies*

In the 6 studies [14, 15, 17, 26, 30, 37], there were 47 analyses on the relationship with salivary cortisol (see Table 1b; 6 on total cholesterol, 6 on Low-Density Lipoprotein (LDL), 27 on High-Density Lipoprotein (HDL), and 8 on total cholesterol/HDL ratio). Of these, 7 associations were significant (15%); the other 40 (85%) showed non-significant findings. The only potential cluster was that a high total cholesterol/HDL ratio and a high LDL level were associated with a higher increase in cortisol under the laboratory stress test (2/4, 50%).

## *Consistency of the Material*

Most studies did not reveal any association between different measurements of cholesterol and salivary cortisol. Of all 27 analyses on single time points, only 1 was significant. The significant association was based on a sub-population analysis [15]. Thus, given the low proportion of significant associations, this literature does not provide support for a strong relationship.

## *Methodological or Contextual Explanation on Divergent Findings*

The findings on cholesterol and cortisol in a laboratory stress test were derived from the same study [14]. Although a high proportion was positively associated with an increase of cortisol under a laboratory stress test, the picture is not clear, as no associations could be found with total cholesterol and HDL in the same study.

## **Triglycerides**

## *Quantitative Analysis on the Evaluated Studies*

In the 5 studies [14, 18, 26, 30, 37], there were 15 analyses on the relationship with salivary cortisol (see Table **1b**). Of these, 9 were on single time points, 4 on deviation, and 2 on AUC.

## *Consistency of the Material*

None of the studies evaluated showed a significant association with triglycerides.

## *Methodological or Contextual Explanation on Divergent Findings*

The studies were performed under similar circumstances. Although the studies were cross-sectional, ruling out the possibility of discussing causality, the non-significant findings strongly suggest that there are no associations between circulatory levels of triglycerides and salivary cortisol.

**Table 1b:** Summary of the main findings of associations between measures of salivary cortisol and lipid status sorted by year of publication



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Exp, experimental.

#### **Plasma or Blood Glucose**

#### *Quantitative Analysis on the Studies Evaluated*

In the 5 studies [18, 26, 30, 37, 39], there were 17 analyses on the relationship with salivary cortisol (see Table **1c**). Of these, 10 were on single time points, 4 on deviation, and 3 on AUC.

**Table 1c:** Summary of the main findings of associations between measures of salivary cortisol and blood glucose sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional.

### *Consistency of the Material*

There was only one significant finding presented (6%), where a single time point in the morning were positively associated with fasting blood glucose. None of the other studies evaluated showed a significant association with blood sugar.

### *Methodological or Contextual Explanation on Divergent Findings*

The studies were performed under similar circumstances. Although the studies were cross-sectional, ruling out the possibility of discussing causality, the amount of non-significant findings suggest that there are no strong associations between levels of fasting blood sugar and salivary cortisol.

#### **Blood Pressure**

#### *Quantitative Analysis on the Studies Evaluated*

Among the 14 studies [14-16, 18, 20, 26, 30, 37, 40-45], there were 53 analyses on the relationship with salivary cortisol (see Table **1d**; 18 on hypertension and 35 on systolic and diastolic blood pressure). Of these, 20 associations were significant (38%); the other 33 (62%) showed non-significant findings.

### *Consistency of the Material*

There were no clear trends on blood pressure and cortisol measurements. The high percentage of significant findings is somewhat devalued by significant findings pointing in opposite directions. If anything, the results suggest that an increase in cortisol in a laboratory stress test is associated with increased blood pressure. This is shown in 4 out of 12 associations (33%). In addition 5 out of 15 reported findings on morning cortisol levels had a positive association (33%)

#### *Methodological or Contextual Explanation on Divergent Findings*

The literature does not provide any clear explanations on why the results vary. One possible explanation for divergent findings has been proposed by Weitz and colleagues [20]. The authors have reported that low birth weight may modify the association between blood pressure and salivary cortisol such that subjects with low birth weight show a clear negative correlation with salivary morning cortisol, whereas no correlation can be found in subjects with normal birth weight [20]. However, their hypothesis is hard to evaluate in the included studies in this overview, as most studies do not report data on birth weight.

**Table 1d:** Summary of the main findings of associations between measures of salivary cortisol and blood pressure sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional.



a Significant associations with blood pressure during stress task, but not with blood pressure at rest. b

Significant associations only in group with low birth weight.

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Exp, experimental.

### **Heart Rate**

All 3 studies evaluated [16, 43, 44] used a laboratory stress test (see Table **1e**). One out of 4 associations tested (25%) between an increase in cortisol level after the stress test and heart rate were significant [43]. However, this was true in only 1 of the 2 stressors tested in the study. Due to the low number, it is hard to further evaluate the consistency of the material.

## **Heart Rate Variability**

### *Quantitative Analysis on the Evaluated Studies*

In the 4 studies [16, 46-48], there were 7 analyses on the relationship with salivary cortisol (see Table **1e**). Of these, 2 were on single time points and 5 on deviations.

## *Consistency of the Material*

Both studies investigating cortisol in a laboratory stress test found a negative association between increase in cortisol level and heart rate variability.

## *Methodological or Contextual Explanation on Divergent Findings*

Heart rate variability can be divided into several components. These are not used uniformly in the literature, which makes it hard to compare different studies on heart rate. Even if there were more comparable studies, it may be relevant to investigate whether the associations are influenced by the stress level among subjects. A recent publication suggests that the associations between level of cortisol are associated with both heart rate and heart rate variability under stressful conditions, but that these associations are attenuated in periods of low stress [49].



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Exp, experimental.

### **Interleukins and Other Markers Related with Inflammation**

### *Quantitative Analysis on the Evaluated Studies*

Combining all inflammatory markers, 30 analyses were presented in 10 studies [16, 29-31, 50-55] (see Table **1f**). Of these, were 12 on single time points, 11 on deviation, and 7 on AUC. Twelve of the analyses (40%) were significant.

## *Consistency of the Material*

Two clusters without contradictory findings arise: High cortisol output throughout the day may be associated with higher average levels of inflammatory markers [29, 30]. 80% of the associations were significant, however derived from two studies only [29, 30]. There was also a possible cluster of significant negative findings in the laboratory stress tests (33 %), indicating that the ability to react with cortisol secretion on a stress test are associated with lower levels of inflammatory markers [16, 51]. However, this inference is derived from two studies only.

### *Methodological or Contextual Explanation on Divergent Findings*

One question raised is the effect on cortisol and cytokine levels following acute stress. It is generally expected that an increase in cortisol levels will lower the level of cytokines. One explanatory factor for some of the non-significant findings may be that some commonly used acute stress tests may actually be too mild to affect the cortisol level to any major extent [56]. Thus, the possibility of finding any strong associations between cortisol and levels of cytokines following an acute stressor may be limited.

In addition, there is a large natural fluctuation of cytokines, depending on ongoing inflammations. Glaser *et al.* [50] have demonstrated that cortisol levels in the morning not is associated with cytokines a normal day but is inversely associated with cytokine levels measured 24 hours after an experimentally induced wound. Thus, occurrence of acute inflammation in some participants but not in others might add to the complexity when studying cortisol in relation to cytokines.

**Table 1f:** Summary of the main findings of associations between measures of salivary cortisol and inflammatory markers sorted by year of publication


 Morning values were not associated with cytokines prior to the experiment, but high cortisol level was associated with low cytokine levels 24 hours after an induced wound

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory stress test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory stress test increase/ground; d, DST, dexamethasone suppression test; C-S, cross-sectional; Exp, experimental.

#### **Adrenaline and Noradrenaline**

Only 2 studies were included in this literature study. Krantz *et al.* [44] reported that there were no associations between levels of salivary cortisol following a laboratory stress test and levels of urinary catecholamines. Cohen *et al.* [57] concluded that there is a non-association between an AUC throughout the day (with respect to ground) and urinary catecholamines in a population-based sample.

#### **DISCUSSION**

#### **General Remarks**

Before interpreting the results further, there are 4 aspects that should be considered.

#### **100** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*

First, we have focused our evaluation on biological correlates in healthy populations. This could be part of the explanation for the large number of non-significant findings. It is possible that studies on populations with diabetes would yield a different association between cortisol and glucose than the ones in this evaluation. The focus on healthy populations also contributes to the high loss of articles comparison with the number of hits in the first step of our search. A review on some of these articles can be found in the chapter on somatic outcome.

Second, we have studied only salivary cortisol. The aim was not to determine physiological correlates with cortisol, but rather to test the feasibility of using salivary cortisol in different contexts.

Third, a large proportion of the papers are based on relatively small study populations. This leads to the possibility of a high number of beta errors in our non-significant findings. On the other hand, the results may suffer from publication bias, where non-significant findings are not reported explicitly in some papers, even though analyses were done on cortisol and the biomarker under investigation.

Fourth, the search strategies used may be somewhat incomplete. It is likely that associations between any of the biomarkers and cortisol in saliva have been studied and presented in papers that could not be identified in our search.

### **Metabolic Abnormalities and Cortisol**

The large number of non-significant findings may come as a surprise. This is in contrast to the widespread hypotheses that there is an association between cortisol and the metabolic syndrome. One of the more wellcited research groups [6, 11, 12] base their conclusions on a design that differs from the most commonly used design, as presented in Table **1**. Instead of any of the suggested measures, the authors base their calculations on "stress-induced cortisol secretion", combining 7 cortisol measurements throughout the day plus a dexamethasone test the following morning. The authors present non-significant findings with cortisol values and all of the metabolic parameters tested (including all criteria for metabolic syndrome). When the intra-individual variance (or the inverse intra-individual variance) between all time points is taken into account by weighting the correlations, the associations with the metabolic parameters become significant. This statistical approach has a considerable impact on the results and alters the conclusions in this study. This approach has not been used in any of the other papers examined in this chapter. Hence, the findings from this study that suggest an association between cortisol and metabolic abnormalities cannot easily be compared with the other studies.

One seemingly contradictory finding is that a high BMI and a high waist circumference are associated with a lower cortisol peak than normal. Based on experimental studies, several authors have proposed an opposite association, namely that an excess of cortisol would lead to an accumulation of abdominal fat. The feedback mechanisms to the HPA axis may be of relevance in observational studies. The high levels of cortisol potentially leading to an accumulation of fat will be counterbalanced by either lowering the sensitivity of receptors to circulating cortisol [58] or lowering the levels of cortisol.

An interesting design that occurs rarely in the literature but still may be of relevance is to consider the dynamics of a system. This is a central point of view in a study by Holt-Lunstad and colleagues on the association between a diurnal variation in cortisol and a nocturnal dip in blood pressure [45]. Going back to the theoretic assumptions in the concept of allostasis denoting stability through change [3, 4], studies linking dynamic capacity in 2 or more systems may be of particular importance. Approaches designed to study variability in other systems in relation to cortisol variability should be encouraged to increase knowledge about the role of cortisol in health and disease.

#### *Biological Markers and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **101**

**Table 2:** Studies included on body mass index, waist circumference, waist/hip ratio, lipid status, glucose, blood pressure, heart rate, heart rate variability, inflammatory markers and other stress hormones sorted by first appearance in text in this chapter 


#### **102** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*



#### *Biological Markers and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **103**


#### **104** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*

#### 


#### *Biological Markers and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **105**


#### **106** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*

#### 


#### *Biological Markers and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **107**


#### **108** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*

#### 


#### *Biological Markers and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **109**


#### **110** *The Role of Saliva Cortisol Measurement in Health and Disease Hansen et al.*


*Abbreviations*: AUC, Area under curve; BMI, body mass index; CAR, cortisol awakening response; CRP, C-reactive protein; C-C, case-control; C-S, cross-sectional; DBP, diastolic blood pressure; ELISA, enzyme linked immonusorbent assay; Excl, exclusions; Exp, experimental; HDL, high-density lipoprotein; HPA, hypothalamic-pituitary-adrenal axis;; HPLC, highperformance liquid chromatography; HRV, heart rate variability; IL, interleukin; LDL, low-density lipoprotein; m/w, men/women; n.a., not applicable; n.s. not stated; P rate, participation rate; Pros, prospective; RIA, radioimmunoassay; SBP, sysstolic blood pressure; SD, standard deviation; SEM, standard error of mean; TNF, tissue necrosis factor; TSST, Trier social stress test; WHR, Waist-hip ratio.

#### **Inflammation and Cortisol**

It is well known that cortisol exerts anti-inflammatory effects. The discovery in the late 1940s that a synthesized cortisol derivate could reverse inflammation in rheumatoid arthritis [59] led to a Nobel Prize. The anti-inflammatory properties have resulted in widespread clinical use of exogenous cortisol (or hydrocortisone as the synthetic form is named).

Thus, the associations in this overview may have been expected to be stronger than shown. In addition to the general problem with possible beta errors due to low statistical power in some analyses, it should be remembered that cytokines are measured in very low concentrations (pg/ml). Thus, statistical analyses on cortisol and cytokines are much more vulnerable to possible insensitivity of laboratory methods than statistical analyses on the other markers covered in this chapter. Moreover, there might also be a considerable variation of cytokine levels within subjects. For instance, von Känel and colleagues compared levels of interleukin-6 (IL-6) measured at a laboratory visit and then a visit 2 weeks later, and found no correlations between IL-6 concentrations at the two time points [60].

However, a couple of studies reported that a higher capability to react with cortisol secretion on a stress test and good capability to recover after a stress test are associated with lower levels of inflammatory markers [16, 50, 51]. Thus, the reported positive association between a higher cortisol output throughout the day and IL-6 [29, 30] is in line with earlier research on the role of cortisol in immunoregulation in studies on patient populations [7, 61]. Fanatidis and colleagues proposed that "inappropriately normal" cortisol levels due to limited capability to respond with increased cortisol levels may not be sufficient to limit an ongoing inflammation [61]. Raison and Miller describe a situation "when not enough is too much", with increased levels of cortisol due to downregulation of receptors on target cells, making glucocorticoid signaling in immunoregulation insufficient [7].

The total numbers on cytokines and other inflammatory markers are few, especially compared with the number of studies focusing on some or several aspects of metabolic abnormalities. This implies that the main paradigm in stress research on cortisol and its physiologic effects has focused on energy supply rather than regulation of inflammation. More research on the association between cortisol in saliva and inflammatory markers is needed to elucidate whether salivary cortisol may be a marker of relevance in clinical settings regarding diseases related to inflammation such as autoimmune diseases and coronary events.

### **Adrenaline, Noradrenaline and Cortisol**

The number of studies reviewed is low perhaps because it is unusual to draw blood and collect saliva samples in the same study design (in comparison, there were 2090 hits on PubMed for a search on "cortisol and "adrenaline" but omitting "saliva\*"). It is therefore not easy to draw any general conclusions on associations between salivary cortisol and adrenaline or noradrenaline. The few studies included in this overview do not provide support for a strong relationship.

In recent research, salivary alpha-amylase has been suggested as a good proxy for activity in the sympathetic adrenal medullary axis [62, 63]. This opens up the possibility to explore the association between the two main stress hormone systems in observational studies, using non-invasive procedures. A few studies have investigated the association between salivary alpha-amylase and salivary cortisol [64, 65]. None have reported any significant correlations between the two stress hormone systems, for single time points or change over the day. This is in line with the studies presented in this chapter evaluating salivary cortisol in relation to urinary catecholamines.

### **Limitations**

A number of problems can be listed when studying the association between salivary cortisol and other physiological measures. Most of the other physiological measures were measured in other matrices than saliva, typically blood or urine. Using different matrices may pose problems *e.g.* due to half-life and timing of samples. When looking at well-designed studies of cortisol, measurement of salivary cortisol has been found to be an excellent indicator of unbound concentrations of cortisol in serum [66-69]. The studies show that the correlation between mean saliva cortisol and mean serum cortisol were approx. r = 0.6 and with a mean cortisol concentration in serum 10-20 times higher than the concentration measured in saliva. Furthermore, even though there is a close correspondence regarding circadian fluctuations of cortisol in saliva and plasma [70], the half lives of various biological measures are not easily comparable to cortisol and per se dependent on the biological fluid. In most of the studies included in the present chapter the sampling of saliva and blood or urine was most often carried out at different time points. A profound diurnal variation will affect the results, particularly if there is a large variation in the time difference between samples or when during the day samples are taken. *E.g.* the difference in concentrations of cortisol between two saliva samples taken on hour apart is much larger in the morning compared to the evening due to the diurnal variation in cortisol.

Comparison between concentrations of cortisol in saliva and other physiological measures in urine poses yet another problem. Whereas concentrations in saliva is affected only by the past few minutes concentrations measured in urine represent the mean excretion since the past urine void.

The sampling design is very closely related to whether the topic of concern is acute or long term stress or a mixture. If the biological measure has a low diurnal and monthly variation and there is an effect of a stressor, it is relevant to talk about a stable measure over time. In contrast, a biological measure with a large diurnal and monthly variation will be influenced by a number of daily hassles. However, the general level of which the measure is fluctuating around may be influenced by a long term perception of stress [71].

### **CONCLUSIONS**

The number of non-significant findings was considerable. This is true for metabolic abnormalities, markers related to inflammation as well as other stress hormones. This overview may suffer from a large number of studies with relatively small study populations. With regard to metabolic abnormalities, the data point to an association between a lower cortisol peak than normal and high BMI or high waist circumference. Further studies on inflammatory markers are needed to elucidate the association with cortisol in saliva.

### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **CHAPTER 6**

# **Sleep and Salivary Cortisol**

#### **Anne Helene Garde1,\*, Berndt Karlson2 , Åse Marie Hansen3 , Roger Persson4 and Torbjörn Åkerstedt5**

*1 Senior reseacher at the National Research Centre for the Working Environment, Lersø Parkallé 105, 2100 København Ø, Copenhagen, Denmark; <sup>2</sup> Doctor in Occupational and Environmental Medicine,at the Department of Public Health and Clinical Medicine, Umeå University, Sweden; <sup>3</sup> Senior reseacher at the National Research Centre for the Working Environment, Copenhagen, Denmark; <sup>4</sup> Senior reseacher at the National Research Centre for the Working Environment, Copenhagen, Denmark and <sup>5</sup> Professor at the Stress Research Institute, Stockholm University, Stockholm, Sweden* 

**Abstract:** The aim of the present chapter was to analyze whether measures of cortisol in saliva were associated with measures of sleep and to explore if divergent results were related to underlying differences in theoretic assumptions and methods. Measures of sleep quality included sleep duration, overall sleep quality, difficulty falling asleep, disturbed sleep, and sleep deprivation. Twenty-three papers were found to fulfil the inclusion criteria. Cortisol measures were grouped into single time points at different times during the day, deviations at different time periods during the day, reactivity and recovery after a standardized laboratory test, area under the curve and response to dexamethasone test. A large proportion of the studies included showed nonsignificant findings, which, in several cases, may be a result of low power. The most consistent results were a positive association between sleep duration and single measures of salivary cortisol at awakening, which was observed in 3 studies. In these studies, sleep duration was also associated with low evening cortisol levels, steep diurnal deviation of cortisol and/or high area under the curve. Together these findings suggest that longer sleep duration is related to a more dynamic cortisol secretion. Two of the 6 studies on disturbed or restless sleep showed relations to flat diurnal deviation and low laboratory stress test reactivity. This to some extent corroborates the findings on sleep duration. However, the many nonsignificant findings as well as the theoretical and methodological differences (*e.g.*, heterogeneity in measures) complicate comparisons. Conflicting results may be at least partially due to differences in methods and underlying assumptions.

**Keywords:** Salivary cortisol, sleep, sleep quality, sleep duration, sleep deprivation, difficulty falling asleep, single time point measures, deviations measures, area under the curve, dexamethasone.

#### **INTRODUCTION**

The stress response can be described as an increase in arousal in response to a real or anticipated perturbation of homeostasis [1]. The Hypothalamus-Pituitary-Adrenal Cortex (HPA) axis is one of the main stress systems with cortisol as a main actor [2, 3]. The underlying anatomy of the stress response is closely interconnected with the anatomy that regulates sleep and wakefulness [4, 5]. Emotional and cognitive arousal may therefore provide inputs that override the normal circadian and homeostatic processes that otherwise govern sleep and wakefulness in normally healthy humans [4, 6]. The interconnectedness also makes sleep a potent factor that may modulate most components of the endocrine system [6]. To summarize, there is a possible bidirectionality between stress and sleep.

Cortisol levels have a circadian peak early in the morning, show a decline throughout the day and are near the limits of detection in the late evening [6]. The secretion of cortisol is inhibited at sleep onset, and during the early part of the sleep period, and cortisol concentrations continue to decrease until a few hours before normal waking time when they start to rise again [6-8].

In experimental studies, induced sleep deprivation lead to higher cortisol concentrations the subsequent

**Margareta Kristenson, Peter Garvin and Ulf Lundberg (Eds) © 2012 The Author(s). Published by Bentham Science Publishers**

**<sup>\*</sup>Address correspondence to Anne Helene Garde:** Senior reseacher at the National Research Centre for the Working Environment, Lersø Parkallé 105, 2100 København Ø, Copenhagen, Denmark; Tel: +45 39 16 52 00; Fax: +45 39 16 52 01; E-mail: ahg@nrcwe.dk

evening [8] and HPA axis hormones such as cortisol-releasing hormone had a negative effect on sleep quality with increased episodes of rapid eye movement sleep and inhibited Slow-Wave Sleep (SWS). In contrast, cortisol has been shown to promote SWS [9].

Although the theoretic and empirical evidence of a close interconnectedness between sleep and HPA axis hormones is strong, there are still several unknowns with regard to understanding the interplay between stress reactions and sleep. As in other areas of stress research, findings have been disparate on these interactions.

## **AIM**

The aim of the present chapter was to analyze whether measures of cortisol in saliva were associated with measures of sleep and to see if possible divergent results were functions of differences in assumptions made and methods used.

## **METHOD**

In a first step, an online search of the NCBI PubMed database (National Library of Medicine, National Institutes of Health, Bethesda, MD, USA-http://www.ncbi.nlm.nih.gov/PubMed) was conducted. The search covered the time period up to October 1, 2009. The search terms were "sleep AND (saliva OR salivary) AND cortisol". One hundred and eight-eight papers were found after limiting the search to papers written in English and studies on humans. Of these, 69 were selected for further scrutiny based on the titles and abstracts. They were supplemented with hand searches. In this step, studies were only included in this review if the study group comprised healthy adults and the study included specific statistical analyses of the association between sleep and cortisol.

Measures of sleep quality included (1) sleep duration, (2) overall sleep quality, (3) difficulty falling asleep, (4) disturbed sleep, (5) premature awakening, and (6) sleep deprivation. Sleep duration is a well-defined measure of the number of hours a person sleeps. It may be assessed from self-reports, actigraphy, or polysomnography (PSG). Reports on sleep quality such as ease of awakening, sleep efficiency, and sufficient sleep by use of questionnaire, logbook or actigraphy were all considered as indicators of overall sleep quality. Sleep quality may be related to sleep problems and divided into categories related to different parts of the sleep: difficulty falling asleep, disturbed sleep (difficulties maintaining sleep), and premature awakening. Difficulty falling asleep covered ease of sleep (inverted), speed of sleep onset (inverted), sleep latency, but not sleep onset, and time of falling asleep. Disturbed sleep covered restless sleep, nocturnal awakenings, time awake after sleep onset, number of microarousals during the night, and number of wake periods after sleep onset. In studies of sleep deprivation participants are actively kept awake.

In the following analyses findings were considered significant if p-values were <0.05. As most of the studies had small numbers and seemingly low statistical power, we also included marginally significant results (0.05<p<0.10) denoted by arrows in parentheses in Table **1**.

## **RESULTS**

In total 23 papers fulfilled the inclusion criteria. A brief summary of the results (indicated as arrows denoting positive associations, or negative association and zero for a nonsignificant finding) are presented in Table **1**. More detailed information on study design, statistical approach, main results, and discussion for each of the 23 papers is presented in Table **2**.

Results are presented for each sleep measure. Cortisol measures were grouped as follows. Single time points at: a1, awakening; a2, morning; a3, midday; a4, evening; a5, all day. Deviations during: b1, morning; b2, midday; b3, morning to evening; b4, laboratory test. Area Under the Curve (AUC): c1, morning (increase/ground). Suppression test: d, response to dexamethasone (DST). No studies were found for premature awakening.

## **Sleep Duration**

Thirteen papers were found to test the association between salivary cortisol and sleep duration [10-22]. In the 13 papers there were 37 analyses on relationships between measures of salivary cortisol and sleep duration. The proportion of significant relationships were 4/16 (25%) for single time points, 6/12 (50%) for deviations, 2/8 (25%) for AUC and 0/1 (0%) for dexamethasone test.

The most consistent results were a positive association between sleep duration and a single measure of salivary cortisol at awakening found in 3 studies [19-21]. In these studies, sleep duration was also associated with low evening cortisol levels [19], steep diurnal deviation of cortisol [19, 20], and with high AUC [21].

In 7 studies the authors failed to find any statistically significant associations between single measures of cortisol and sleep duration [11, 12, 15-17, 22]. The size of these studies was, in general, very small.

The association between sleep duration measures and deviations in cortisol measures was investigated in 7 studies. Morning deviations in cortisol concentrations were found to be positively associated with sleep length in an experimental study of 16 young people (8 morningness and 8 eveningness) using PSG [16]. In 2 ambulatory studies with more than 200 participants [10, 14] and a study of 2761 civil servants using selfreports negative associations to morning deviation in cortisol concentrations [20] were found.

Two studies showed a positive association between self-reported sleep duration and diurnal deviation of cortisol [19, 20]. In 4 other studies, no significant associations were found [11, 12, 15, 22], although tendencies were observed in 1 [22].

Morning AUC was the only AUC investigated in relation to sleep duration [12, 13, 18, 21, 22]. One study, a case study with 50 days of sampling, showed a positive relationship. In contrast, 1 study, which used an insomnia scale and defined sleep duration as "more than six hours sleep", showed a negative relationship. Two out of 4 studies had only nonsignificant findings.

### **Overall Sleep Quality**

Associations between sleep quality and measures of salivary cortisol were assessed in 8 studies [11, 12, 15, 17, 19, 21, 23, 24]. In the 8 papers there were 28 analyses on relationships between measures of salivary cortisol and overall sleep quality. The proportion of significant relationships was 5/21 (24%) for single time points, 1/5 (20%) for deviations, and 0/2 (0%) for AUC. Sleep quality was measured mainly by use of self-reports, but also PSG [13].

The most consistent pattern, a positive association to a single measure at awakening [11] or in the morning [17, 23], was observed in 3 studies. However, 5 other studies found no associations with a single morning or awakening cortisol measure [12, 15, 19, 21, 24]. In 4 studies, sleep quality was examined in relation to single measures in the afternoon or an evening measure; no associations were found [11, 15, 19, 24]. No significant associations were seen for sleep quality and deviations in cortisol concentrations [12, 19, 24].

One study found a positive relationship between stress reactivity and sleep quality measured as sleep efficiency by actigraphy, but not by self-reports [15]. One study examined associations between sleep quality and morning AUC, and found no significant relationship [21].

### **Difficulty Falling Asleep**

Three studies assessed a total of 10 associations between salivary cortisol and difficulty falling asleep [15, 23, 25]. The proportion of significant relationships was 0/5 (0%) for single time points, 2/3 (67%) for deviations, and 1/2 (50%) for AUC. Difficulty falling asleep was assessed by use of actigraphy and selfreports (ease of sleep (inverted), speed of sleep onset (inverted), sleep latency, and time to fall asleep). The studies all used different types of cortisol measures.

Only 1 of the 3 studies reported significant associations, and the results were mixed [25]. In the same study the association between self-reported difficulty falling asleep in terms of ease of sleep was positively related to slope, whereas speed of sleep onset was negatively related [25]. High self-reported difficulty falling asleep was related to high AUC morning [25]. No other significant associations were observed between self-reported ease of sleep and measures of cortisol [15, 23, 25].

## **Disturbed Sleep/Restless Sleep**

Disturbed or restless sleep was examined in 6 studies [11, 12, 15, 20, 21, 26] analyzing a total of 22 relationships. The proportion of significant relationships was 3/13 (23%) for single time points, 4/7 (57%) for deviations, and 0/2 (0%) for AUC. Disturbed sleep was assessed as the number of microarousals during the night using PSG, forced awakening, actigraphy, and self-reports (restless sleep, nocturnal awakenings, time awake after sleep onset, and number of wake periods after sleep onset).

Four studies included associations with a single cortisol measure at awakening or in the morning: 1 found a positive association with the number of microarousals [12], 1 found a negative association with selfreported frequency of nightly awakenings, but no association with self-reported wake time after sleep onset [11], and 2 found no associations [20, 21]. No associations were observed for single measures of cortisol later in the day [11, 15].

One study investigated the relation between disturbed sleep and diurnal deviation and found a negative association [20]. No significant findings were seen in the 3 studies that investigated the relationship between morning deviations of cortisol and disturbed sleep in terms of nightly microarousals[12], forced awakenings [26], and sleep disturbance [20]. One study investigated the effect of disturbed sleep the night before a laboratory stress test, and found negative associations with reactivity [15].

AUC in the morning was tested in relation to disturbed sleep on a day to day basis in a case study with 50 days of sampling; and no significant associations were found [21].

## **Sleep Deprivation**

Six studies investigated a total of 8 associations between sleep deprivation and measures of salivary cortisol with mixed results [27-32]. The proportion of significant relationships was 2/5 (40%) for single time points, 1/3 (33%) for deviations, and 0/0 (0%) for AUC. The studies used either 1 night of sleep deprivation [28- 30, 32] or 5-6 nights of only 4 h sleep [27, 31].

In 1 study sleep restriction was associated with increased concentrations of cortisol in the evening and smaller decline in cortisol during the afternoon [27]. In another study it was found that cortisol concentrations were higher in the afternoon after sleep deprivation [29]. In 4 studies using cortisol concentrations in the morning, evening, and during the day following sleep deprivation, no associations were observed [28, 30-32].

**Table 1:** Summary of main findings of associations between measures salivary cortisol and studied domains sorted by year of publication



*Abbreviations:* a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; AG, actigraphy; AUC, Area under the curve (increase vs ground) ; C, case; C-C, case-control; C-S, cross-sectional; Exp, experimental; M, men; Pros, prospective; PSG, polysomnography; SR, selfreported; W, women. indicates that the slope is steeper. a

Sleep length >6 h.

b Significant finding only in evening types.

#### *Sleep and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **121**


**Table 2:** Descriptives of the articles on salivary cortisol and sleep parameters sorted by domain of sleep parameter and year of publication 

#### **122** *The Role of Saliva Cortisol Measurement in Health and Disease Garde et al.*

Design: C-S

M/W: 102/117

Group: Healthy

No.: 219

Schlotz 2004

*Sleep length:* 

?

?

*Method:*

[14]

#### Age: 48.6 (24-83) years Days: 7 consecutive Samples per day: 4 Times for sampling: Awakening, +15, 30, and 60 min, Setting: Ambulatory (at home) RIA Measurement(s): a1. Cortisol on awakening b1. Mean increase from awakening Cortisol data: Continuous Statistics: ANOVA with repeated measures No association between sleep duration and cortisol on awakening (a1) Positive association between sleep duration and mean cortisol increase from awakening (b1) Days: 1 Samples per day: 4 Times for sampling: Base line cortisol before stress test (14:00 h), post test, +30 min and 45 min post test Setting: Laboratory with stress test Immunoassay After stress test the participants were asked to relax and read general interest magazines Measurement(s): a3. Single measure at baseline (14:00 h) b4. Reactivity to test Cortisol data: Logarithmic (base 10) Statistics: Pearson´s correlations, univariate analysis and partial correlations adjusting for baseline cortisol Positive association between cortisol reactivity to test reactivity to test reactivity to test (14:00 h) Days: 6 days (?) Samples per day: 2 LIA (IBL) No smoking, no teeth Measurement(s): a2. Single measures at Cortisol data: Continous? Statistics: ANCOVA with TST had positive association with b1 after night sleep


#### *Sleep and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **123**


#### **124** *The Role of Saliva Cortisol Measurement in Health and Disease Garde et al.*

*Difficulty falling asleep:*

sleep onset

*Disturbed sleep:* 

sleep registration

Three experimentally

call). The wake up in the

*Method:*

*Method:*

Lasikiewicz

2008 [25]

Dettenborn

2007 [26]

#### Ease of sleep and speed of Questionnaire (Leeds Sleep Evaluation Questionnaire) Design: C-S No.: 147 M/W::68/79 Age: :mean age 46.2 years (±7,2) Group: volunteers Days: 1 (*n*=64) or 3 (*n*=83) Samples per day: 8 Times for sampling: Awakening, +15, 30, 45 min, +3, 6, 9, 12 h Setting: Ambulatory Immunoassay method Not to consume food or drink other than water in relation to sample collection. Avoid teeth brushing and vascular leakage Measurement(s): Mean of same time point on consecutive days a5. b4. Deviation evening from 45 min post awakening (slope) c1. AUC not specified Cortisol data: Log transformed Statistics: Pearson's correlation. Cluster analysis (M)ANOVA Confounders: Age, gender Association between higher ease of sleep (less difficulty falling asleep) and low AUC Association between high ease of sleep (less difficulty falling asleep) and less steep slope (b3) Association between high speed of sleep onset (less difficulty falling asleep) and more steep slope (b3) No association between ease of sleep, speed of sleep onset and diurnal mean (a5) induced awakenings (phone morning was optional or set up by alarm clock. No other Design: Exp No.: 13 M/W: 0/13 Age: 24 years Days: 3 intervention nights + 3 reference nights Samples per day: 8 on intervention nights and 2 on recovery nights Times for sampling: Awakening and +15 min in the morning Setting: Ambulatory CLIA Measurement(s): b1. Repeated measures Cortisol data: continous Statistics: ANOVA and ANCOVA Confounders: Oral contraceptives, thyroid hormone The morning CAR after disturbed nights was not different from CAR on undisturbed nights There was a lack of HPA axis activation by forced nightly awakenings Design: Exp No.: 11 M/W: 11/0 Age: 18-27 years Group: Young healthy volunteers Days: 3 Samples per day: 12-20 Times for sampling: Every 30 min between 15.00 and bedtime Setting: Laboratory. 3 nights with 8 h in bed, 6 nights with 4 RIA a4. Single evening concentration b2. Deviation between 16:00 h and 21:00 h Cortisol data: Continuous Statistics: ANOVA for repeated measures Higher evening cortisol concentration after sleep restriction Lower rate of decrease in the


#### 

#### *Sleep and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **125**


*Abbreviations*: AUC, Area under the curve; C-C, case-control; C-S, cross-sectional; CAR, cortisol awakening response; CLIA, chemiluminescence-assay; DST, Dexamethasone test; ELISA, Enzyme Linked Immuno-Sorbant Assay; Exp, experimental; GLM, generalized linear model; LIA, luminescence immunoassay; M, Male; PSG, polysomnography; PSQI, Pittsburgh Sleep Quality Index; P rate, Response rate; RIA, radioimmunoassay; TST, total sleep time, W, women.

#### **DISCUSSION**

The aim of the present chapter was to analyze whether measures of cortisol in saliva were associated with measures of sleep to see if divergent results were functions of differences in theoretic assumptions made and methods used. Relatively few papers were first identified (n=188), and only 23 papers met the final inclusion criteria.

The most consistent results were a positive association between sleep duration and a single measure of salivary cortisol at awakening, which was observed in 3 studies [19-21]. In these studies, sleep duration was also associated with low evening cortisol levels [19], steep diurnal deviation of cortisol [19, 20], and with high AUC [21]. Together these findings suggest that longer sleep duration is related to a more dynamic cortisol secretion.

However, long sleep duration is also associated with a lower CAR [10, 14, 20]. Since a lower CAR implies a less dynamic response these observations seems to contradict the suggestion that longer sleep duration is related to a more dynamic cortisol secretion. However, a lower CAR needs not to be mutually exclusive with a dynamic cortisol secretion as assessed by the decline in concentrations during the entire day. Indeed, a lower CAR might just reflect the diurnal rhythmicity of cortisol secretion. For example, to the extent that participants, who sleep long wake up later than usual, they are likely to take their first sample at a time when the concentrations of cortisol already have risen due to the normal diurnal rhythm. Hence, there is simply less room to obtain a high CAR as the morning value has been inflated by the underlying diurnal rhythm. However, there might also be several other explanations and this question deserves more attention in future studies.

Two of the 6 studies on disturbed sleep showed that it was associated with less diurnal deviation [20], and lower reactivity to a laboratory stress test [15]. This to some extent corroborates the findings on sleep duration. However, the many nonsignificant findings as well as the theoretic and methodological differences (*e.g.* heterogeneity in measures) complicate comparisons.

As expected, statistically significant associations were more often reported in studies with a large number of participants or a high sampling frequency. Among the papers included, statistically significant findings were generally seen in studies with more than 100 participants [10, 14, 19, 20], or samples (case study) [21]. In the large studies, it was more common to focus on general sleep patterns, whereas the case study registered day to day variations within the same person. The results suggest that a relationship between sleep and salivary cortisol is observable both within and between subjects, but requires many observations due to high variability in both cortisol and sleep measures. The use of small study samples and non-optimal measurement procedures might lead to studies with low power and failure to detect statistically significant results.

Even if the inclusion criteria drastically reduced the number of papers, the papers included covered several different types of measures and indicators for cortisol and sleep. This heterogeneity in measures and methods appears to generate a mixed pattern of results that are also conflicting at times. For example, in the same study a positive relationship was observed between stress reactivity and sleep quality measured as sleep efficiency by actigraphy, but not by self-reports [15]. As differences between objectively measured and self-reports of sleep have been observed previously, this conflict may be partially related to the selected instrument for measurement of sleep quality [33].

There are several methodological problems in studies of sleep and saliva cortisol. Because saliva sampling requires that the person is awake, it is virtually impossible to sample saliva during sleep and to establish the relationship with different sleep stages or other processes that may occur during sleep. It is often not practically feasible in field studies to obtain direct measurements during the sleep period with, for example, PSG or endocrine measures. Even if it is technologically possible to use such measures, the procedures and commitments necessary for successful implementation have often been considered to be cumbersome and too expensive to be considered as a realistic option. Stress researchers often have to rely on less sophisticated and simpler approaches such as motion logging (*e.g.*, actigraphy) and subjective reports of various aspects of sleep (*e.g.*, bedtime, awakening time, and perceived quality of sleep).

Many of the studies were not conducted primarily to evaluate the relationship between sleep and salivary cortisol. Even so, all studies included in the present chapter have used statistical tests to address the overall question about the relationship between sleep and saliva cortisol. Considering that research have been driven by many points of departures, it is not surprising that most of the result do not act in concert.

#### **CONCLUSIONS**

In total 23 studies examining sleep quality in relation to salivary cortisol measures were identified. There was a large proportion of nonsignificant findings and many operational definitions of sleep quality and cortisol secretion. Because many of the studies included were small and entailed few measurements, there is reason to believe that the nonsignificant findings partly reflect low statistical power.

The most consistent results were our observation of a positive association between sleep duration and a single measure of salivary cortisol at awakening, which was observed in 3 studies. In these studies, sleep duration was also associated with low evening cortisol levels, steep diurnal deviation of cortisol, and/or with high AUC. Together these findings suggest that longer sleep duration is related to a more dynamic cortisol secretion. Two of the 6 studies on disturbed or restless sleep showed a relationship with flat diurnal deviation and low laboratory stress test reactivity. This to some extent corroborates the findings regarding sleep duration. However, the many nonsignificant findings as well as the theoretic and methodological differences (*e.g.*, heterogeneity in measures) complicate comparisons. Conflicting results may be at least partially due to differences in methods and underlying assumptions.

#### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **CHAPTER 7**

# **Mental Health and Salivary Cortisol**

#### **Ingibjörg H. Jonsdottir1,\* Christina Halford2 and Frida Eek<sup>3</sup>**

*1 Associate professor at the Institute of Stress Medicine, Carl Skottsbergs gata 22 B, 413 19 Gothenburg. Sweden; <sup>2</sup> Researcher at the Department of Public Health and Caring Sciences, Uppsala University, Sweden and <sup>3</sup> Associate professor at the Division of Occupational and Environmental Medicine, Lund University, Sweden.* 

**Abstract:** The aim of this chapter was to analyze associations between measures of cortisol in saliva and mental health and to see if divergent results were functions of the methods used. Measures of mental health outcome included Major Depressive Disorder (MDD), symptoms of depression, and symptoms of anxiety, Burnout (BO), and Vital Exhaustion (VE). Only studies on otherwise healthy individuals were included. Cortisol measures were grouped into single time point measures, measures of deviations, laboratory test responses, Area Under the Curve (AUC), and response to dexamethasone. Some consistency is seen for MDD, mainly higher mean levels. The results regarding single measures and depressive mood are less consistent, but the overall picture for depression shows poorer diurnal deviation and response to stress. Inconsistency among papers studying depression seems to be related mainly to the study population. Very few significant findings were found for anxiety, therefore cortisol does not seem to be strongly related to anxiety. Most of the statistical analysis does not show a significant relationship between BO and cortisol, and when these are present, the results are inconsistent. One explanation seems to be the measures of BO used, probably due to the different conceptual basis for BO. VE measured using the Maastricht Questionnaire seems to be related to a poorer cortisol response to stress and poorer diurnal deviation. The coexistence of BO and VE in many studies does make it difficult to conclude how the different concepts are related to cortisol. However, an interesting difference appeared between MDD and VE in response to dexamethasone administration, showing lower suppression in MDD patients and higher suppression in VE patients. A general conclusion for all mental health measures is that a large proportion of non-significant findings are due to low power and few sampling days combined with low contrasts between study groups and within study populations. Generally, deviation measures such as diurnal deviation seem to be more valid measures compared with single measures to capture possible changes in the hypothalamus-pituitaryadrenal axis, measured using salivary cortisol.

**Keywords:** Salivary cortisol, depression, anxiety, major depression disorder, burnout, vital exhaustion, single time point measures, deviations measures, area under the curve, laboratory test, dexamethasone.

#### **INTRODUCTION**

Mental health consequences of long-term stress exposure can vary greatly between individuals, and symptoms of depression, anxiety, Burnout (BO), and Vital Exhaustion (VE) are often, but not always, related to psychosocial stress exposure. There is a great discrepancy in the literature regarding the relationship between the Hypothalamus-Pituitary-Adrenal (HPA) axis function and mental health outcomes. This review only deals with salivary cortisol and the aim is to analyze if the existing literature, studying the above-mentioned mental health symptoms and salivary cortisol, show consistent findings and if not, can the discrepancy be explained. Previous literature on depression shows inconsistent findings and the finding on hypercortisolemia in patients with depression is not always confirmed in the literature [1, 2]. In this chapter, the studies are separated depending on whether patients with Major Depressive Disorder (MDD) were studied or if the study deals with the relationship between salivary cortisol and depressive mood.

Anxiety has been shown to be characterized by hypocortisolemia and supersuppression after

**Margareta Kristenson, Peter Garvin and Ulf Lundberg (Eds) © 2012 The Author(s). Published by Bentham Science Publishers**

**<sup>\*</sup>Address correspondence to Ingibjörg Jonsdottir:** Associate professor at the Institute of Stress Medicine, Carl Skottsbergs gata 22 B, 413 19 Gothenburg. Sweden; Tel: +46 31 342 0716; Fax: +46 31 41 42 73; E-mail: Ingibjorg.jonsdottir@vgregion.se

dexamethasone [3] but there is a discrepancy in the literature regarding the relationship between HPA axis function and anxiety, one possible explanation being that anxiety is a multifacetted phenomenon. Cortisol Awakening Response (CAR) does not seem to be significantly affected by anxiety according to the metaanalysis by Chida and Steptoe [2] and HPA axis reactivity in response to a stressor does not seem to be significantly affected by anxiety [4].

The constructs of BO and VE are suggested to reflect responses, in terms of exhaustion, due to long-term stressor exposure. BO is a mental condition defined as a result of continuous and long-term stress exposure, particularly related to psychosocial factors at work [5, 6]. However, the theoretic basis for the term burnout differs between the available self-report instruments constructed to assess BO, and different instruments do seem to measure quite different aspects depending on the theoretic base for the instrument. However, they share the fact that they were mainly developed for research on work-related stress. The most widely used instrument is the Maslach Burnout Inventory (MBI) and the conceptual basis for MBI is thus often considered as synonymous with the construct BO [7]. Maslach and colleagues originally defined BO as a psychological syndrome of emotional exhaustion, depersonalization (later replaced with the construct cynicism) and reduced effectiveness or personal accomplishment [6]. Another conceptual approach was presented by Melamed and coworkers, viewing BO as a multidimensional construct consisting of emotional exhaustion, physical fatigue, and cognitive weariness, which together represents the core component of BO [5, 8]. The concept of VE is characterized by unusual fatigue, loss of mental and physical energy, increased irritability, and feeling of demoralization. The concept was developed in the search for premorbid psychological characteristics of people who developed myocardial infarction, and has thereafter been used to define the psychological state viewed as chronic stress [9, 10]. The concepts of BO and VE do share several features and many studies do not distinguish between these two concepts. Furthermore, within the same concept, *e.g.*, BO, different measures can differ considerably and the dissimilarities could thus be larger within the concept BO compared with dissimilarities between the BO and VE concepts. Based on the definitions used, the concepts of MDD, depressive state, BO and VE, all share several conceptual similarities; *e.g.*, high correlation between the scales used to measure BO and VE is seen [11], and between scale scores of depressive state and VE [12]. In this review, BO and VE are treated partly as the same construct when the general interpretations are done, but both constructs are also initially divided into respective measures as these can differ considerably as previously mentioned. Kudielka and coworkers have recently reviewed the literature on cortisol measurements in BO and VE, concluding that there seems to be a considerable divergence on data regarding HPA axis functioning in chronically distressed individuals [11]. As both hypoactivity and hyperactivity of the HPA axis has been reported in studies on BO or VE, the direction of the supposed dysregulation of the HPA axis remains inconsistent.

Another related concept is chronic fatigue which is not considered in this review, mainly because it is difficult to discriminate chronic fatigue and chronic fatigue syndrome, which is sometimes, but not always, due to postviral infection or other somatic conditions. Also, several symptoms related to chronic fatigue, *e.g.,* muscle pain, joint pain, sore throat, and tender cervical nodes, are of different character compared with BO, even though fatigue is the core feature as for both BO and VE. Furthermore, chronic fatigue syndrome is not as distinctively related to psychosocial stress as the concepts of BO and VE.

#### **AIM**

The aim of this chapter is to analyze if measures of cortisol in saliva are associated with depression, anxiety, BO, and VE, all of which are measures of mental health frequently related to chronic psychosocial stress. Furthermore, we aim to see if the divergent results in studies involving cortisol assessments and mental health are functions of differences in the theoretic assumptions made and methods used.

#### **METHOD**

#### **Search Strategies**

For all outcome measures, an electronic search was performed in the NCBI PubMed database (National Library of Medicine, National Institutes of Health, Bethesda, MD, USA-http://www.ncbi.nlm.nih.gov/PubMed). The database PsycINFO was also searched for relevant papers not found in the PubMed database. The search covered the time period up to October 1, 2009. Only full-length articles published in English in peer-reviewed journals, based on adult study populations, including direct statistical analyses of associations between cortisol and the actual outcome measure, were included. Studies were selected in two steps, with the first step based on titles and abstracts and, when relevant, by reading the full-length article.

### **Depression**

Search terms "depression" and truncated "salivary cortisol" generated 324 abstracts. The exploration and results were divided into two sections: one included papers studying MDD according to DSM-IV and salivary cortisol and the second part included papers studying depression measured with different self-rated scales and salivary cortisol in otherwise healthy individuals. Thus, studies on subjects with apparent somatic diseases (*e.g.*, cancer and traumatic brain injury) or psychiatric illness (*e.g.*, posttraumatic stress disorder (PTSD) and bulimia) other than anxiety were excluded. Studies on pregnancy and postnatal depression were also excluded. Intervention studies (*e.g.*, medications) were included if comparison between groups was available before the intervention. However, the effects on salivary cortisol in response to the intervention are not included in this review. Papers studying a possible relationship between salivary cortisol and vulnerability to developing future depression were not included. In some cases, the same research group published different data sets using the same patient material and in the cases where the salivary cortisol data were repeated, only one of the papers is included in this review. Studies using instruments measuring mood state (*e.g.*, Profile of Mood States Questionnaire) were not included. The final number of papers included was 21 studying MDD and salivary cortisol and 14 studying the relationship between scores measuring symptoms of depression and cortisol. Reasons for exclusions were: 116 papers because the patients had comorbid conditions (such as cancer and cardiovascular diseases), in 103 papers statistical analyses between depression and cortisol were not presented; 24 papers studied pregnancy or postnatal depression; in 13 papers, cortisol was measured in children. Twenty-seven papers studied MDD and cortisol in saliva, but did not include a control group and 6 papers studied previous or future depression.

### **Anxiety**

The search terms were "anxiety" and truncated "salivary cortisol". The search on PubMed generated 281 abstracts. In order to limit variation in the anxiety measures, we decided to limit the inclusion to three scales used to measure symptoms of general anxiety levels: the Spielberger State-Trait Anxiety Inventory (STAI), the Beck Anxiety Inventory (BAI) and Hospital Anxiety and Depression (HAD) scale. Only papers on otherwise healthy individuals were included. Thus, studies on subjects with apparent somatic diseases (*e.g.*, cancer and traumatic brain injury) or psychiatric illness (*e.g.*, PTSD and bulimia) other than comorbid depression were excluded. Studies reporting cortisol responses to psychosocial stress were included if anxiety levels were related to salivary cortisol in the statistical analysis. Studies of salivary cortisol and its relation to mood changes (state anxiety) were not included. Results from intervention studies (*e.g.*, medication) were included if in the statistical analysis the relation between trait anxiety and salivary cortisol was measured before the intervention (basal levels). Studies of acute anxiety scores and salivary cortisol levels due to, for example, parachute jumping or other similar physical stress were not included. The final number of papers included was 17 and all except 1 used STAI to measure anxiety levels.

### **Burnout and Vital Exhaustion**

The following search terms were used: "cortisol", "saliva\*", "burnout" and "exhaustion". The first step identified 31 papers investigating associations between cortisol and BO and 28 studies investigating associations between cortisol and exhaustion. Studies based on patient populations (other than BO), pregnant women, and studies investigating effects of intervention on cortisol were excluded, as were studies in which analyses of direct associations between BO or exhaustion and cortisol were not presented. Concerning exhaustion, studies primarily investigating the effects of physical exhaustion on cortisol were excluded. Finally, 21 papers were included, 12 of which investigated associations between BO measures and cortisol in saliva, and 7 with associations with exhaustion measured using the Maastricht Questionnaire (MQ) [9] or the exhaustion subscale of MBI.

### **RESULTS**

Results are presented for each mental health measure. Cortisol measures are grouped into: single time points at: a1, awakening; a2, morning; a3, midday; a4, evening; a5, all day. Deviations during: b1, morning; b2, midday; b3, morning to evening; b4, laboratory test. AUC: c1, morning; c2, midday; c3, morning to evening; c4. laboratory test. Dexamethasone suppression test: d, response to dexamethasone (DST).

#### **MDD and Depressive Mood**

Twenty-one papers were included covering MDD, and diagnosis based on DSM-IV was used in these studies to confirm depression. In the 14 papers studying depressive mood, 6 different instrument were used: Becks Depression Inventory (BDI) (5 papers), Center for Epidemiologic Studies Depression Scale (CES-D) (4 papers), the Zung Self-Rating Depression Scale (2 papers), Hamilton Depression Inventory (HDI) (1 paper), HAD scale (1 paper) and the Patient Health Questionnaire (PHQ) (1 paper). A brief summary of the results (indicated as a positive association, a negative association, or a non-significant finding) is presented in Tables **1a** and **1b**. The study design, statistical approach, main results, and discussion for each of the 35 articles are briefly presented in Table **2a** (MDD) and Table **2b** (depressive mood).

As seen in Tables **1a** and **1b**, there is a large variation in the salivary cortisol measures used in different studies. Among the 21 studies on MDD, most of the measures (39 measures) were from single time points, and 7, 10 and 6 measures were on deviations, AUC, and after DST, respectively. The proportion of significant relationships (in any direction) among the 21 papers studying MDD was 20/39 (51%) for single measure, 3/7 (43%) for deviation measures, 6/10 (60%) for AUC measures, and 4/6 (67%) for the DST studies.

Among the 14 studies on depressive mood, the proportions of significant relationship were 8/20 (40%) for single measures, 4/10 (40%) for deviation measures, and 1/2 (50%) for AUC measures. No DST suppression test study was found for depressive mood.

The only consistency among studies on MDD was reactivity on laboratory stress (one study) or daily event measured in ambulatory settings (one study), as both studies showed poorer ability to respond with cortisol to acute stress. The only study included in this review, studying response to stressor in relation to depressive mood, also showed that response to naturalistic stressors was poorer in women with higher scores of depressive mood.

No single measure of salivary cortisol showed full consistency regarding either significant or nonsignificant results for either of the groups (MDD or depressive mood). The most consistent significant findings among the MDD papers were single evening measures and single morning measures; 5 out of 10 measures of evening cortisol (50%) showed increased cortisol levels and 10 out of 20 morning measures (50%) showed increased levels. Among the papers studying single measures in relation to depressive mood, no clear consistency can be found, and there are too few measures for each time point to draw any firm conclusions regarding depressive mood and single measures of salivary cortisol.

For deviation measures, 1 study examining CAR in MDD patients showed non-significant findings, 1 study of 4 on diurnal deviation was significant and showed an inverse relation for MDD and cortisol. Both studies on stress reactivity were significant (one on men only), showing poorer cortisol response among MDD patients. The same pattern is seen for the studies on depressive mood, as the 3 studies examining CAR did show non-significant findings; 3 out of 6 studies on diurnal deviation were significant, all showing an inverse relation between depressive mood and cortisol. Three out of 5 studies measuring salivary cortisol after dexamethasone administration reported that MDD patients showed less suppression, usually calculated as more non-suppressors among the patients.

Comparing results from case-control studies on MDD patients and population studies on depressive mood, two measures showed full consistency. Thus, as all three studies on cortisol reactivity in response to stress

showed that depression or depressive mood was related to poorer response to stress and the four studies including CAR deviation measures reported non-significant findings. Full consistency across MDD and depressive mood studies was also found among significant findings for the deviation between morning and evening, as all studies reporting significant relation, showed that flatter diurnal curve was related to depression or depressive mood.

**Table 1a:** Summary of main findings of associations between measures of salivary cortisol and MDD sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; C-C, Case-Control; C-S, Cross-sectional. a

Information regarding males and females only reported for the patient group. b

Measured after DST 16:00 h . No suppressors among the patients (compared with none among controls) but statistical analyses not presented. c Lack of response to negative daily event in patients with MDD compared with controls.

d Significant in males only.

e Significant in women only.

f Level of cortisol not significant (0), cortisol suppression ratio significantly different. More suppression in the patient.

#### **Anxiety**

In total, 17 papers studying the relation between salivary cortisol and general anxiety trait were included. A brief summary of the results (indicated as a positive association, a negative association or a non-significant finding) is presented in Table **1c**. Study design, statistical approach, main results and discussion for each of the 17 papers are briefly presented in Table **2c**.



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; C-C, Case-Control; C-S, Cross-sectional.; Exp, Experimental.

The proportion of significant relationships among the 33 measures presented in the papers studying anxiety and salivary cortisol was as follows: 2/12 (17%) for single measure (both measures in the same study), 8/18 (44%) for deviation measures, and 1/3 (33%) for AUC in response to a laboratory stress test. No study included dexamethasone administration.

As for depression, the cortisol measures varied among the studies and there were relatively few studies on each measure making it difficult to draw firm conclusions. Full consistency was seen for single measures in the afternoon, single measures in the evening and the means of all day measurements of cortisol, as all studies reported non-significant findings. Significant findings among single time point measures were seen only for morning measures and in one study only (showed higher cortisol among military men). For deviation measures, only 1 study examined CAR, showing lower CAR among more anxious men but not among women. Diurnal deviation measures showed significant findings in 3 of 4 studies: a negative relationship between anxiety and cortisol among women attending a breast cancer clinic (receiving a benign diagnosis) and positive relationships in two studies among white-collar middle-aged men and among military men; the latter were the same group as those with high morning levels.

Stress reactivity was found to be higher in anxious women (only in the follicular phase); 2 studies found blunted reactivity in anxious individuals compared with controls. One study showed that pre-test values were positively correlated to anxiety, but no relation was found for stress reactivity.

**Table 1c:** Summary of main findings of associations between measures salivary cortisol and anxiety sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; C-C, Case-Control; C-S, Cross-sectional.; Exp, Experimental, STAI, State-trait Anxiety Inventory; FP, Folicular phase; LP, Luteal phase.

#### **Burnout and Vital Exhaustion**

The final analysis includes 13 articles on BO. Four studies used the Shirom Melamed Burnout Questionnaire (SMBQ), 8 studies used the MBI and one combined MBI and Teachers Burnout Scale (TBI). The final analysis on VE included 9 articles, 6 of which used the MQ on VE and 3 used the exhaustion subscale of MBI. One study is presented twice as two different instruments were used to measure vital exhaustion [60]. A brief summary of the results (indicated as a positive association, a negative association, or a non-significant finding) are presented in Table **1d**. Study design, statistical approach, main results, and discussion for each of the articles are briefly presented in Table **2d**.

With regard to BO (Tables **1d** and **1e**), there is a large variation of the salivary cortisol measures used in different studies. Among the studies measuring BO using SMBQ, the proportion of significant relationships among the 19 statistical analyses performed was 5/19 (26%), all in the same direction; *i.e.*, BO measured with SMBQ was positively related to salivary cortisol. The proportion of significant relationships (in any direction) among the 9 articles studying BO using MBI was 10/20 (50%) for single measure showing both positive and negative relationship with BO. All deviation measures showed non-significant results except for the 1 studying acute stress responses to laboratory stress, showing higher reactivity in BO. Two studies using AUC measures, and 3 studies on DST response showed non-significant results.

The proportion of significant relationships (in any direction) among the 6 articles studying VE using MQ was 2/14 (14%) for single measures, 3/7 (42%) for deviation measures, 0/1 study using AUC measures; the only DST study showed a significant relationship between VE and lower cortisol level, showing higher suppression. Among the studies measuring VE with the MBI-EE subscale or an electronic diary using an exhaustion item from MBI-EE, the proportion of significant relationships (in any direction) among the 3 articles was 0 for single measures (only 1 measure presented), 2/5 for deviation measures, and 2/6 for DST.

Thus, with regard to BO, most of the significant findings for both SMBQ and MBI were seen for single measures. However, there was no clear tendency for any time point to be more relevant, and within each time point the direction of the relationship varied. The full consistencies found in the data set were all related to non-significant findings. Regarding VE, among single measures (MQ and MBI-EE), significant findings between cortisol and exhaustion were all in the negative direction. For deviation measures, the CAR measure findings were divergent, 1 positive and 1 negative; for diurnal measures all showed negative relationships, *i.e.*, flatter diurnal curve among people with VE. All significant DST results (3/7) showed higher suppression related to VE.


**Table 1d:** Summary of main findings of associations between measures of salivary cortisol and burnout sorted by instruments


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; C-C, Case-Control; C-S, Cross-sectional.; Exp, Experimental

a Slight positive association between cortisol at 07:00 h and exhaustion subscale, but no association between cortisol and DP, or PA subscales.

**Table 1e:** Summary of main findings of associations between measures of salivary cortisol and burnout and exhaustion sorted by instruments


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; C-C, Case-Control; C-S, Cross-sectional.;

a ESM: aggregated 2-week score.

b ESM: same-moment assessment.

#### **138** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*

**Table 2a:** Studies on MDD sorted by year of publication 


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **139**


#### **140** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*



#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **141**


#### **142** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **143**


#### **144** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*

**Table 2b:** Studies on depressive mood sorted by year of publication 


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **145**



#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **147**


**Table 2c:** Studies on anxiety sorted by year of publication 


#### **148** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **149**


#### **150** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **151**


**Table 2d:** Studies on burnout and exhaustion sorted by year of publication 


#### **152** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*

#### 


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **153**


#### **154** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*

#### 


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **155**


#### **156** *The Role of Saliva Cortisol Measurement in Health and Disease Jonsdottir et al.*


#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **157**



#### *Mental Health and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **159**


#### **DISCUSSION**

The aim of the present chapter was to analyze associations between several measures of mental health related to chronic stress and cortisol in saliva. Relatively few papers met the final inclusion criteria. Many papers have studied the relationship between depression and salivary cortisol, but the studies often includes patients with other somatic or mental diseases, making it difficult to conclude whether a true association exists, not related to other diseases.

#### **MDD and Depressive Mood**

For depression, when looking at the overall picture, there seems to be some consistency that salivary cortisol is increased in clinical populations of patients with MDD, particularly in the single morning and evening measures. The results from the laboratory stress studies also suggest that MDD is related to poorer ability to respond with cortisol to acute stress. Measures of salivary cortisol after dexamethasone administration demonstrated that patients with MDD showed less suppression, usually calculated as the number of non-suppressors with a larger proportion among the patients.

No time point, however, among the single measures showed full consistency regarding salivary cortisol levels in patients with MDD. In an attempt to explain the discrepancy between different studies, 4 main explanations are given: the type of the depression, the power of the study in terms of participants, the reliability of cortisol measures in terms of one or more days of sampling, and confounding factors, mainly medication.

Discussions about the type of depression seem to mainly involve melancholic or non-melancholic depression. Seven out of 21 papers studying MDD did not find any significant differences between patients and controls for measures of salivary cortisol [16, 19-21, 23, 25, 30]. The authors of 3 of these studies mentioned that one possible explanation could be that the HPA axis abnormalities are usually found in either melancholic or psychotic depression [19, 23, 25], but Copolov and coworkers did not find any differences between non-melancholic and melancholic patients regarding salivary cortisol levels [13]. Unfortunately, most of the papers do not specify the type of depression and thus it is difficult to speculate to what extent this explains the discrepancy between the studies. One study showed changes in the opposite direction and the authors suggest that this could be because the patients were a healthier population of individuals recruited from a general population through advertising [28] compared with other studies in which *e.g.* inpatients with MDD were included.

The power of the studies in terms of small sample size may be another important explanation for several nonsignificant findings [16]. The sample sizes in studies not showing significant differences between patients and controls are, in general, small, ranging from 14 to 44 patients. However, among those studies in which significant differences between the groups were found, small samples sizes were also found. Small sample size could in many cases, depending on the type of depression, increase the difficulty in identifying difference between patients and controls. However, power is also dependent on the reliability of the measures taken. Among the present studies, the number of days for ambulatory sampling was lower (median 2 measures) for studies without significant results compared with those with significant findings (median 3 measures). The combination of few cases and few days of sampling may add to this loss of power.

The authors of 3 of the 7 papers not showing any difference in salivary cortisol levels between patients and controls also suggested use of antidepressants as a possible explanation, because the HPA axis function has been normalized using antidepressants [19, 21, 25]. The fact that 3 of the studies showing hypercortisolemia in the patient group included patients taking antidepressants does not entirely support this notion. However, there are too few studies to draw firm conclusions but most of the studies showing hypercortisolemia did include antidepressant-free patients. Possibly, a combination of all these factors could explain much of the discrepancy between studies. Among the MDD studies, 5 of those showing hypercortisolemia in patients with MDD attempted to relate to salivary cortisol levels with scale scores in the patient group by using different measures of self-reported depression or, in one case, symptom severity; none of these correlations were found to be significant.

There are fewer studies on depressive mood and salivary cortisol measures included in our analysis. Of 9 studies, measuring single values or the mean of several measures at any time point during the day, 4 showed that depression was related to increased levels, and 2 of these divided the groups into high and low depression instead of relating to scale scores.

The results regarding depressive mood are less consistent for higher levels in singles measures and this is confirmed when the authors studying MDD attempted to relate cortisol level to scale scores. However, with regards to deviation measures, the results seem to be consistent with MDD; 3 of 6 studies showed flatter diurnal deviation. The only study measuring acute response to stress confirms the data from the MDD groups showing poorer response.

### **Anxiety**

An overall conclusion for all single measures is that cortisol levels do not seem to be related to anxiety. In the 2 studies measuring ambulatory salivary cortisol during the morning hours, only 1 study showed a significant relation to anxiety. This was also the only study that measured cortisol at several time points during the morning (3 time points) at fixed hours. The fact that the study group consisted of men on military service, where stressors are often more prominent than in ordinary life, may be more important [57].

Similarly, there were few significant findings regarding measurements of deviation, but also few studies. Only 1 study examined anxiety in relation to CAR and this showed lower CAR among more anxious men but not among women [45]. Diurnal deviation was studied in 4 studies. In 3 of these, significant relationships were found, 2 of which were positive and 1 negative. These differences occurred in different contexts; positive relations occurred among healthy and/or military men [34, 57], and negative relations among women attending a breast clinic [38]. However, the changes where relatively small and the results are not explained by a difference in the awakening or evening levels but rather a slight change in the diurnal profile during mid-morning or afternoon showing either higher or lower levels in relation to anxiety.

Six of the 17 studies assessed saliva cortisol in relation to a laboratory stress test. Among these, only 1 had a significant effect on baseline levels, *i.e.*, standardized rest before stress.

Three of 7 studies found significant results but in opposite directions. Thus, 2 showed a positive relationship (male students and healthy elderly men and women) and 1 showed a negative relationship (female students during the luteal phase). In addition, 1 out of 3 studies measuring AUC in response to stressors showed significant findings; with lower salivary cortisol among male students with high anxiety.

In 6 of 9 studies showing a significant relationship between anxiety level and salivary cortisol, regardless of the measure of cortisol, the groups were split into high and low anxiety [38, 50, 53, 57-59], either by median split (3 studies), using a cut-off (1 study) or by only including extreme groups, *e.g.*, subject with high or low anxiety (2 studies). All except 1 of the studies not showing any significant relation with anxiety treated the anxiety measure as a continuous variable using scales scores. This is in parallel with the conclusion drawn from the depression section that differences are more likely to be detected when the subjects are stratified into groups with higher contrasts. An additional explanation for non-significant findings could be low power and few sample days, because many studies included 1 day sampling. In conclusion, very few significant findings were found for single measurements. Using deviation measures, although there were only a few studies, most showed a significant relationship between anxiety and salivary cortisol, but in opposite directions. The observed divergence of results may be dependent on contextual factors, *i.e.*, the amount of external stressors and population characteristics. Thus, our paper may extend the previous conclusion made by Chida *et al.* that the HPA axis does not seem to be strongly affected by anxiety levels [2, 4].

### **Burnout**

The overall conclusion regarding BO is that most of the statistical analyses (59 analyses in 13 papers), irrespective of the character of the analysis, do not show any significant relationship between BO and measures of cortisol in saliva. However, most of the studies (9 out of 13) found a significant relationship between BO and at least one of the cortisol measures. Only four studies did not show any relationship with BO [63, 67, 69, 77]. Consequently, most of the authors rightfully concluded that BO is associated with dysregulation in HPA axis functioning. Clearly, however, when analyzing the findings, there seems to be a large discrepancy between different studies regarding the measures used and there is no salivary cortisol measure that clearly demonstrates the character of this supposed HPA axis dysregulation.

Full consistency in terms of non-significant findings is seen for the single midday measures, for all deviation measures from ambulatory saliva sampling, the morning deviation measures (9 studies), and for diurnal deviation measures (6 studies).

No clear picture regarding methodological issues can be seen, that could explain the discrepancy between different studies. There are probably several contributing factors, including the groups studied and the possible influence of comorbid conditions such as depression and in some cases PTSD that could explain some discrepancy between the studies. Differences related to the BO measure used are another possible explanation. It is clear that the BO concept is defined differently depending on which BO measure is being used [7]. Scrutinizing all single measures of cortisol among the papers studying BO, it seems that the type of BO measure could partly explain the discrepancy among the papers showing significant relationships. When significant, studies using the SMBQ showed a positive relationship with cortisol [8, 62]; there were divergent findings among the papers using MBI, as those papers showed both positive and negative associations [64-66, 68, 71, 72].

Unfortunately, many papers include too little or different information about the subjects, comorbidity, medications, adjustments, etc., making it difficult to use the information in the articles to explain the results. Thus, we confirm the conclusion by Kudielka and coworkers that the inconsistency between studies on BO cannot be easily explained [11].

### **Vital Exhaustion**

The pattern for VE is somewhat different from BO, as significant relationships were found for several deviation measures that were not seen for BO. Thus, 2 of 3 studies relating VE using MQ to diurnal variation showed significant relationships, both in the same direction, towards a flatter curve among VE subjects. The major difference between these 2 studies and the study not showing any relationship seems to be the subjects included. The studies showing a flatter day curve included subjects originally selected from a population sample; the study not showing a significant relationship included nurses with high levels of BO. The BO measures in that study were significantly related to increased levels of salivary cortisol, but no relation was seen with VE. Thus, there seems to be some differences between the VE measures and the BO measures regarding the relationship with changes in the HPA axis but this is also highly dependent on which subjects are included in the studies. In all 3 studies measuring VE with MBI-EE, BO was also present, and in 2 of these studies, the inclusion criterion was BO. Consequently, in these studies it is difficult to separate BO from VE, and it also shows that these measures are highly correlated. When significant, the DST suppression test shows consistent results, all pointing towards higher suppression among VE subjects. The only study showing a significant relation among the BO papers also showed higher suppression.

The findings for VE indicate that VE measured with MQ in the general population do suggests that exhaustion could be related to a flatter day curve and poorer response to stress test. The results for cortisol in relation to BO measured with SMBQ seem to be entirely the opposite as BO seems to be related to a higher level of cortisol. The results from the MBI BO measure are mixed and when exhaustion is measured with the MBI-EE subscale, BO is also present and it is thus difficult to conclude whether VE is related to cortisol independent from BO. From these results, BO and VE seem to some extent to be different entities, and the 2 BO measures also seem to be different entities. Even more interesting, 2 of the BO also seem to be different entities, strengtening what has been previously suggested [7].

### **CONCLUSIONS**

The relationship between cortisol measures in saliva and mental health, and consistency of the results, varies depending on the mental health measure. However, for all measures studied, one important notion is that the final number of studies included is relatively few and the power of studies is, in general, small.

For MDD, there seems to be some consistency for higher mean cortisol levels, poorer response to stress, and loss of feedback. In most of the studies, single measures of salivary cortisol were increased in clinical populations of patients with MDD, particularly in the morning and evening measures. The pattern of response to stress in ambulatory settings and in laboratory stress testing does suggest that MDD is related to a poorer ability to respond with cortisol to acute stress. However, this is seen in terms of a flatter diurnal deviation and in poorer laboratory stress reactivity, but not in terms of CAR. Responses to dexamethasone administration showed less suppression among MDD patients. The results regarding depressive mood are less consistent for higher levels in singles measures and this is confirmed when the authors studying MDD attempted to relate cortisol level to scale scores. However, with regards to deviation measures, the results seem to be consistent with MDD; 3 of 6 studies showed flatter diurnal deviation. The only study measuring acute response to stress confirms the data from the MDD groups showing poorer response.

In contrast, the anxiety results show poorer consistency and few studies are included. For single measurements, very few significant results were found; for deviation measures, the results were divergent. Thus, our paper may extend the previous conclusion made by Chida *et al.* that the HPA axis is not strongly affected by anxiety levels [2, 4].

Similarly, the overall conclusion regarding BO is that most statistical analyses, irrespective of the character of the analysis, do not show any significant relationship between BO and cortisol measures, and when these are seen, the results are inconclusive. The two BO measures seem to be partly responsible for this and there is a better consistency within the studies using SMBQ compared with those using MBI. One possible explanation is that the BO concept is defined differently depending on the measures being used [7].

For VE measured using MQ, the pattern of results seems to be different, suggesting that this measure of exhaustion, particularly in the general population, is related to a poorer cortisol response to stress, which is only seen for diurnal deviation and laboratory stress testing but not for CAR. The coexistence of BO and VE in many studies makes it difficult to conclude how the different concepts are related to cortisol. However, an interesting difference appeared between MDD and VE in response to dexamethasone administration. MDD patients tended to show lower suppression and VE patients higher suppression, which suggests a difference in the biological (HPA axis) underpinning of these conditions.

A general impression is that the large proportion of non-significant findings is a function of low power; because of small study samples, few sampling days resulting in low reliability of saliva measures, but also less sensitive measures in terms of low contrast between study groups and within study populations. All mental health measures included in this chapter can be considered to be a chronic state of symptoms or illness, related to long term exposure of psychosocial stress. A cautious conclusion is that this seems be related to poorer cortisol response to stress and a flatter deviation day curve, but more studies are needed with larger power to confirm this as many studies also show non-significant findings. An overall conclusion is also that for several mental health outcomes, deviations measures are more valid than single time point measures

### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **Somatic Disease and Salivary Cortisol**

# **Margareta Kristenson1,\* and Oskar Lundgren2**

*1 Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, 58183 Linköping, Sweden and <sup>2</sup> Ph.D. Student in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, Sweden* 

**Abstract:** Stress is a well-known predictor of somatic disease. Although most clearly demonstrated for Coronary Heart Disease (CHD), stress has also been shown to be involved in several other somatic diseases *e.g.* rheumatoid arthritis, cancer and for pain syndromes. The psychoneuroendocrine mechanisms of these effects have been examined in terms of cortisol levels and cortisol dynamics. The aim of this chapter is to investigate if there are associations between salivary cortisol and somatic disease in terms of cardiovascular disease (CVD), rheumatoid arthritis, cancer and pain, and whether divergent results can be explained by differences in the theoretic assumptions made and methods used. A literature research identified eight articles on CVD, four articles on cancer (all breast cancer), three papers on rheumatoid arthritis and 15 papers on the term pain. CVD, CHD and atherosclerosis were associated with low morning cortisol levels, high evening cortisol levels and a flat diurnal curve. Among patients with metastatic breast cancer, high evening levels and low diurnal deviation characterized patients compared with healthy controls, and low diurnal deviation predicted poorer survival. No relationships with salivary cortisol were found early in the breast cancer disease process. Patients with rheumatoid arthritis, especially with high disease activity, had higher evening levels and a poorer reactivity for laboratory stress. In most studies on pain, low morning cortisol, high evening cortisol, low cortisol awakening response and low diurnal deviation were associated with more pain. Fibromyalgia and pelvic pain among men were an exception. We found few studies where the relationship between salivary cortisol and somatic disease/illness was analyzed. However, among these, a relatively large proportion showed significant findings. The results suggest that, across outcomes, low morning cortisol levels, high evening cortisol levels and a low dynamic cortisol response to stress are related to poorer somatic outcome.

**Keywords:** Salivary cortisol, cardiovascular disorders, breast cancer, rheumatoid arthritis, pain, single time point measures, deviations measures, area under the curve, laboratory test, dexamethasone.

### **INTRODUCTION**

Psychosocial stress is a well-known determinant for several somatic outcomes [1]. This has especially been demonstrated for CHD, where both the incidence of disease and the risk of recurrent disease have been shown to be related to psychosocial factors and stress [1-3].

However, this has also been demonstrated for other somatic diseases *e.g.*, cancer [1, 4, 5], rheumatoid arthritis [1, 6, 7] and pain disorders [8, 9]. Generally, stressful events are thought to influence the pathogenesis of disease by causing negative affective states (*e.g.*, feelings of anxiety and depression), which, in turn, exert direct effects on biological processes or behavioral processes that influence disease risk [1].

One of the endocrine systems that is particularly reactive to psychological stress is the Hypothalamus-Pituitary-Adrenal (HPA) axis. Cortisol as the primary effector of HPA activation regulates a broad range of physiologic processes including anti-inflammatory responses. Although this stress hormone is important for survival, a disrupted balance of response may lead to higher vulnerability to disease [10] and have the potential to influence a variety of diseases including CHD, autoimmune diseases and cancer [1].

**<sup>\*</sup>Address correspondence to Margareta Kristenson:** Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, 58183 Linköping, Sweden; Tel: +46 10 103 5075; Fax: +46 10 103 1865; E-mail: margareta.kristenson@liu.se

Alternations in cortisol dynamics have been associated with tumor growth [5] and rheumatic disease [6-8] and have also been shown to affect the prognosis of treatment among cancer patients [4] and patients with pain [8, 9].

In several of the somatic diseases mentioned earlier, cortisol assessments have been performed to try to establish possible mechanisms for the observed relations to stressors. However, as for other measures in this book, the results have been conflicting, and both high and low cortisol have been found in relation to different somatic outcomes, if any relationship exists at all. One possible explanation for these inconsistencies in results is the differences in hypotheses and basic assumptions behind different studies. Such differences can be seen in terms of whether researchers look for high overall cortisol output or dynamic responses to acute stress.

## **AIM**

To investigate if there are associations between salivary cortisol and somatic disease, more specifically in relation to measures of CVD, cancer, rheumatoid and pain morbidity, and whether divergent results can be explained by differences in the theoretic assumptions made and methods used.

## **METHOD**

### **Search Strategies**

A literature search was conducted using the PubMed database for articles published before October 1st , 2009. The search was limited to human studies on adults more than 19 years of age in the English language. The search terms were truncated saliva\* AND cortisol AND somatic disease/diagnosis for the following diagnoses: CVD, Coronary Artery Disease (CAD), CHD, Acute Coronary Syndrome (ACS), atherosclerosis and stroke. The search for cancer included breast cancer, prostate cancer, intestinal and cervical cancer. The search for inflammatory disease was done for RA and rheumatoid arthritis. A search was also done for the term pain.

The search result was skimmed for relevant articles by reading the title and abstract from each hit. Articles that met the inclusion criteria with measures of salivary cortisol related to the measures of somatic disease or pain were then selected for inclusion in the literature review. These articles were then read in detail.

### **RESULTS**

The number of hits and final number of papers used in the review were: for cardiovascular disease; CAD (15 and 5), CHD (15 and 5), ACS (3 and 2), CVD (86 and 8); atherosclerosis (12 and 4), and stroke (10 and 1). Several of these findings were the same studies and the final number of articles on CVD was 8. For cancer diagnoses, the number of hits and final number of papers were: breast cancer (32 and 4), prostate cancer (3 and 0), intestinal cancer (3 and 0), and cervical cancer (1 and 0); in summary 4 studies met the inclusion criteria. For rheumatoid arthritis, 3 papers met the inclusion criteria, selected from the following search results: RA (9 and 2) and rheumatoid arthritis (8 and 3). The search term pain resulted in 76 hits and 14 papers. A total of 29 articles met the inclusion criteria and were hence included in the literature review.

### **CVD**

A total of 8 papers were found [11-18]. Four of these studies used measures of atherosclerosis as outcomes, *i.e.*, Intima Media Thickness (IMT) [11, 13], coronary calcification [14] and plaques in arteria carotis [18]; the other four used measures of symptomatic CVD, CAD or ACS [12, 15-17]. A brief summary of the results (indicated as a positive association, a negative association or a nonsignificant finding) is presented in Table **1a** and the details are given in Table **2a**.

Four of 14 results on cortisol measurements from single time points showed significant findings: none at awakening, 1 of 4 morning levels and 1 of 2 midday levels (both a negative relationship) and 2 of 4 evening levels (both showing positive relationships). Hurwitz Eller *et al.* [11] reported, in cross-sectional analyses,

lower morning levels among women (not men) with higher Intima Media Thickness (IMT) and Rosmond *et al.* [12] reported lower mean cortisol at 12:00 h to characterize a pattern predicting CVD, diabetes and hypertension over 5 years. Nijm *et al.* [15] found high evening levels among CAD patients, compared to healthy controls, and Dekker *et al.* [18] found that high levels at 17:00 h predicted more plaques.

Among deviation measures, 8 of 13 results showed significant findings. Six CAR results were reported from four studies and three of these showed significant findings. Two studies came from the same population from which Hurwitz Eller *et al.* reported significant relationships to IMT (for women only); a negative relationship in cross-sectional analysis [11] and a positive relationship in a prospective analysis [13]. Bhattachyyra *et al.* [17] reported higher CAR among CAD patients compared with patients investigated for suspected CAC without CAD; Matthews *et al.* [14] found no relationship between CAR and different severity of ACS among patients within 5 days of admission.

Five studies reported 6 measures on deviations throughout the day and 4 of these showed significant findings. In a prospective study Rosmond *et al.* [12] reported a significant relationship for low cortisol deviation at midday and higher CVD incidence over 5 years. Matthews *et al.* [14] reported that a flatter diurnal decline was related to a higher risk of coronary calcification and Nijm *et al.* [15] reported a flatter diurnal decline among CAD patients compared with controls. Bhattachyyra *et al.* [17] reported that, with no significant relationship between diurnal deviation and the presence of CAD, CAD patients with depression had a flatter diurnal decline compared with those without depression. As for measures of CAR, Whitehead *et al.* [16] found no significant relationship between diurnal deviation and severity of ACS.

Only one study reported results from laboratory stress test reactivity: Nijm *et al.* [15] reported a significantly poorer response to the stress test among CAD patients compared with controls. Two studies examined the Area Under the Curve (AUC). Matthews *et al.* [14] found no significant relationship between AUC and coronary calcification. Dekker *et al.* [18] demonstrated that a higher area under the daytime curve (AUC) was related to more carotid plaques.

These findings can be summarized as follows: CVD seems to be associated with low morning cortisol levels, high evening cortisol levels and lower deviation throughout the day.


**Table 1a:** Summary of the main findings of associations between salivary cortisol parameters and CVD, sorted by year of publication

*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, DST, dexamethasone suppression test; CC, casecontrol; CS, cross-sectional; Prosp, prospective.

\*Significant lower diurnal deviation among CAD patients with depression compared to CHD patients without.

### **Breast Cancer**

Four studies were identified [19-22] (see Table **1b** and Table **2b**). Of these, one studied survival of patients with metastatic breast cancer; the other three were case-control studies of breast cancer at different stages (one metastatic cancer, one newly diagnosed breast cancer and one at stage I-III).

Two results from single measures were reported; both were summary measures over the day; in one case a significant positive relationship (case- control study of metastatic breast cancer) was found; the other had nonsignificant findings (stage I-III).

One measure of CAR was reported and showed nonsignificant findings for a relationship for patients with newly diagnosed breast cancer.

All four studies reported results from diurnal cortisol deviations. Two of these showed significant relationships. Both were studies of metastatic breast cancer and in terms of a flat curve: Sephton *et al.* [19] as a determinant of survival over 7 years and Abercrombie *et al.* [20] as the difference between patients and healthy controls. The other two studies showed nonsignificant findings: Vedhara *et al.* [21] for patients with newly diagnosed breast cancer and Carlson *et al.* [22] for breast cancer stage I-III.

Three studies reported results from the AUC, both with respect to increase and ground, all with nonsignificant findings [19, 21, 22].

The findings for breast cancer can be summarized as follows: patients with metastatic cancer seem to be characterized by high evening cortisol levels and low diurnal deviation compared with healthy controls, and the low diurnal deviation predicted poorer survival. Among patients early in the disease process, no relationship with salivary cortisol was found.

**Table 1b:** Summary of the main findings of associations between salivary cortisol parameters and breast cancer, sorted by year of publication


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, DST, dexamethasone suppression test; BC, breast cancer; Prosp, prospective; CC, case-control; CS, cross-sectional.

### **Rheumatoid Arthritis**

Three studies reported salivary cortisol in relation to rheumatoid arthritis [23-25] (see Table **1c** and Table **2c**). Two studies reported results from summary measures over the day. Dekkers *et al.* [23] and Catley *et al.* [24] both found significant positive relations, *i.e.*, higher mean cortisol among patients with rheumatoid arthritis [23, 24]. However in the study by Dekkers *et al.* [23] this finding occurred in patients with high disease activity only, and not for patients with low disease activity. All three studies reported deviation measures, but none with significant findings. Dekkers *et al.* [23] found no relationship with to CAR and neither Catley *et al.* [24] or Eijsbouts *et al.* [25] found any significant relationship with diurnal deviation.

In summary, patients with rheumatoid arthritis, especially those with high disease activity, seem to have higher mean levels of salivary cortisol.



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, DST, dexamethasone suppression test; RA, rheumatoid arthritis; CC, case-control; CS, crosssectional.

\*Valid for difference between RA patients with high activity, compared to RA patients with low activity and to healthy controls.

#### **Pain**

Fourteen papers were identified [26-39] (see Table **1d** and Table **2d**). They included sciatic pain, pain in interstitial cystitis, pelvic pain, pain in fibromyalgia, chronic widespread pain, pain in the shoulder and neck, acute and chronic lumbar pain, facial pain and experimental pain.

Eight out of 12 results (11 studies) from single time points showed significant findings.

Two studies reported awakening levels. Geiss *et al.* [26] reported that patients who had persistent pain after discectomy had a lower cortisol level on awakening compared with patients with low postoperative complaints and with healthy controls. McLean *et al.* [28] found that, among patients with fibromyalgia, current pain symptoms were related to higher cortisol levels on awakening.

Four studies reported results from morning levels: two with positive associations, one with a negative association and one with nonsignificant findings. Lutgendorf *et al.* [27] reported that, among patients with interstitial cystitis, those with lower morning cortisol levels had more pain, and McBeth *et al.* [31] reported that low morning salivary cortisol was associated with higher risk of chronic widespread pain 15 months later. In the study of McLean *et al.* [28] on patients with fibromyalgia, a positive relationship between current pain symptoms and higher cortisol level was also seen 60 min after awakening. Ehrström *et al.* [36] reported no relationship with morning cortisol levels in a case-control study of localized vulvodynia.

Two studies reported midday levels. McLean *et al.* [28] found no relationship to pain among patients with fibromyalgia while Shell *et al.* reported that, in a cross-sectional analysis of healthy population that midday/evening cortisol levels were higher in the group with higher pain, however only among men (35).Two studies reported evening levels of cortisol, one with significant positive relationships. McBeth *et al.* [31] reported that high evening cortisol levels were related to higher risk of chronic widespread pain 15 months later, while McLean *et al.* [28] found no relationship to pain among patients with fibromyalgia. One study reported a summary measure of cortisol. McBeth *et al.* [30] reported, in a cross-sectional case-control study using a summary measure of morning and evening levels, that subjects with, or at risk of, chronic widespread pain were more likely to have lower cortisol scores.

For deviation measures 12 results were reported from 9 studies and 7 of these with significant findings. Six out of seven studies on CAR showed significant findings: 5 with negative associations and 1 with a positive relationship.

Giess *et al.* [26] reported that patients with persistent pain after discectomy had lower CAR than those without pain. Gaab *et al.* [29] showed that patients with chronic widespread pain had lower CAR compared with healthy controls. Ehrström *et al.* [36] reported that patients with localized vulvodynia had lower CAR than healthy controls. Sudhaus *et al.* [38] reported that patients with chronic low back pain and more nonverbal pain behavior had lower CAR and Fabian *et al.* [37] reported that a flattened CAR was related to greater pain intensity in experimental pain ratings. Anderson *et al.* [33] reported that men with chronic pelvic pain syndrome had higher CAR than age-matched pain-free controls.

Four studies reported results from diurnal deviation: one with marginally significant results and three with nonsignificant findings. Johansson *et al.* [34] demonstrated that among patients scheduled for lumbar disc surgery, those with low diurnal deviation had more leg pain and disability compared with the high diurnal variation group (p=0.10). Gaab *et al.* [29], using a short circadian profile found no difference between patients with chronic widespread pain compared with healthy controls and Galli *et al.* [39] found no differences in diurnal deviation between patients with chronic myogenous facial pain and controls and Anderson *et al.* found no differences between men with or without chronic pelvic pain syndrome (33).

One study reported data from laboratory stress testing. Wingelfeld *et al.* [32] found no difference between patients with chronic pelvic pain and fibromyalgia and healthy controls in the Trier Social Stress Test.

Three studies reported results in terms of AUC, all of them with nonsignificant findings; Wingelfeld *et al.* [32] found no difference between patients with chronic pelvic pain and fibromyalgia , Sudhaus found no relationship to low back pain (38) and Galli *et al.* found no significant relationship to chronic facial pain (39).

Four studies reported results from the dexamethasone test, all with significant findings (two with positive and two with negative associations). Gaab *et al.* [29] and Galli *et al.* [39] reported on prolonged/enhanced suppression among patients with whiplash-associated disorder and chronic myogenous facial pain, respectively. In a cross-sectional [30] and prospective (31) analyses of chronic widespread pain, McBeth *et al.* found higher serum cortisol levels the day after a dexamethasone test in the pain group.

In summary, in most of the studies low morning cortisol, high evening cortisol and low cortisol reactivity in terms of low CAR or low diurnal deviation were related to more pain. Exceptions were findings of higher cortisol being related to more pain among patients with fibromyalgia and men with pelvic pain.


**Table 1d:** Summary of the main findings of associations between salivary cortisol and pain sorted by year of publication

#### *Somatic Disease and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **173**


*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, DST, dexamethasone suppression test; FM, fibromyalgia; WAD, whiplash-associated disorder; CWP, chronic widespread pain; Exp; Experimental pain ratings LBP, low back pain; CMP; Chronic myogenous pain. \* Parenthesis denoting marginal significance of results.

#### **174** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson and Lundgren*

**Table 2a:** Studies on cardiovascular disease sorted by year of publication 


#### *Somatic Disease and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **175**


*Abbreviations:* ACS, acute coronary syndrome; BDI, Beck Depression Inventory; BMI, body mass index; CaC, coronary calcification; CAD, coronary artery disease; CHD, coronary heart disease; CLIA, chemiluminescence immunoassay; IMT, intima media thickness; RIA, radioimmunoassay; SBP, systolic blood pressure; SES, socioeconomic status.

#### **176** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson and Lundgren*

**Table 2b:** Studies on breast cancer sorted by year of publication 


#### *Somatic Disease and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **177**


*Abbreviations:* EIA, electroimmunoassay; ELISA, enzyme-linked immunosorbent assay; NK, natural killer; RIA, radioimmunoassay. 

#### **Table 2c:** Studies on rheumatoid arthritis sorted by year of publication


*Abbreviations:* ESR, eosinophil sedimentation rate; IL-6, interleukin 6; RA, rheumatoid arthritis.

#### **178** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson and Lundgren*

**Table 2d.** Studies on pain sorted by year of publication 


#### *Somatic Disease and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **179**


#### **180** *The Role of Saliva Cortisol Measurement in Health and Disease Kristenson and Lundgren*


#### *Somatic Disease and Salivary Cortisol The Role of Saliva Cortisol Measurement in Health and Disease* **181**


*Abbreviations:* ALBP, acute low back pain; CLBP, chronic low back pain; CLIA, chemiluminescence immunoassay; CPP, chronic pelvic pain; CWP, chronic widespread pain; FMA, fibromyalgia syndrome; OR, odds ratio; PCA, principal component analysis; RIA, radioimmunoassay; TSST, Trier Social Stress Test; WAD, whiplash-associated disorder.

#### **DISCUSSION**

The main impression from this literature review is that few studies have examined salivary cortisol in relation to somatic outcomes. Several nonsignificant findings were seen among these studies, however there was a relatively large proportion of significant findings and a parallel pattern of findings across diseases: *i.e.*, low morning cortisol, high evening cortisol and low cortisol reactivity related to disease or ill health.

However, not all findings followed this pattern. In a cross-sectional analysis CAR was negatively related to IMT among women [11], but in a follow-up of the same population the opposite finding (*i.e.*, a positive relationship) was reported [13]. Matthews *et al.* [14] found, in a young population with 8% prevalence of atherosclerosis, that an attenuated diurnal deviation was related to more coronary calcification, and Dekker *et al.* [18] found, in an elderly population with 75% prevalence of atherosclerosis, that high evening values and high AUC were related to more plaques in arteria carotis with no relationships for diurnal deviation.

These studies all used measures of atherosclerosis. This was also the case for the study of Bhattacharyya *et al.* [17], which examined patients being investigated for suspected CAD and definitive CAD defined by results from angiography. Patients with definitive CAD had higher CAR compared with patients without CAD. Thus, high evening levels and high AUC were positively related to measures of atherosclerosis and a possible summary of these findings is that the amount of atherosclerosis is positively related to high total cortisol.

Bhattacharyya *et al.* [17] also found that CAD patients with depression had flatter diurnal deviation and the authors concluded that this "flatter cortisol might contribute to the progression of coronary atherosclerosis".

Three studies had symptomatic CVD disease as the outcome: one prospective study of CVD [12], one casecontrol study of CAD [15], and one cross-sectional study on the severity of ACS [16]. For CVD event/CAD, the findings were consistent; the prospective study on CVD found low diurnal deviation related to more CVD [12]; the case-control study found low diurnal deviation and poor laboratory stress response among cases with CAD [15]; no relationship was seen for severity of ACS [16].

An acute coronary event is a result of a rupture of plaques and there is evidence that plaques characterized by more inflammation and higher levels of degrading enzymes are more vulnerable. In the study of Nijm *et al.* [15] the attenuated cortisol stress response was accompanied by higher levels of inflammatory markers and higher levels of degrading enzyme matrix metalloproteinase. Thus, the flat diurnal curve and/or attenuated response may be related to CAD *via* the quality of atherosclerosis, *i.e.* to more vulnerable plaques.

Four studies were found for cancer, all breast cancer. Two were on patients with metastatic cancer [19, 20] and two on early stage cancer [21, 22]. Results suggest that salivary cortisol is related to breast cancer disease at the metastatic stage but not in earlier stages. The two studies on metastatic breast cancer were in agreement that low diurnal deviation was related to patients with metastatic breast cancer and/or prognosis of disease [19, 20]. In addition, in the case-control study, patients had higher all-day cortisol [20].

High all-day cortisol was also seen in patients with rheumatoid arthritis compared with healthy controls. However, in one of these studies, this was seen only for patients with high disease activity in terms of sedimentation rate [23], and this was a result of higher levels in the afternoon and evening, with no differences in morning levels. In the second study, Catley *et al.* [24] concluded that the results provide additional evidence that the HPA axis is disturbed in rheumatoid arthritis, and also that there was no evidence that the differences were effects of differences in ongoing stress. In the third study, no significant relationship with salivary cortisol was seen [25].

The 14 papers on pain included a variety of patient groups; however, all studies had pain as outcome. The overall pattern suggests that low morning, high evening cortisol, flat CAR and/or diurnal deviation are related to more pain in interstitial cystitis, risk of chronic widespread pain, risk of pain in the neck and shoulders, localized vulvodynia, chronic low back pain, experimental pain ratings and sciatic pain.

Low morning cortisol was related to more pain among patients with interstitial cystitis [27], and higher risk of chronic widespread pain 15 months later [31]; high evening levels of cortisol were related to higher risk of chronic widespread pain 15 months later [31], and high midday/evening cortisol predicted pain in the neck and shoulders 12 months later, however, among men only [35].

In case-control studies on patients with chronic widespread pain [29], localized vulvodynia [36], chronic low back pain [38] compared with healthy controls and in experimental pain ratings, a flattened CAR was related to greater pain intensity [37]. In two studies, patients with sciatic pain showed better outcome after surgery if they had a low morning cortisol level and/or morning/diurnal deviation of cortisol [26, 33], which in both cases was suggested to be a function of the anti-inflammatory effect of cortisol being related to steep diurnal deviation.

Contrasting findings were seen in a study on fibromyalgia in which current pain symptoms were related to higher cortisol at awakening and 60 min after awakening [28] and for men with chronic pelvic pain syndrome who had higher CAR than age-matched pain-free controls [33]. A possible interpretation is that cortisol levels in these latter cases are reactions to pain and/or the stressful situation for these patients.

Most studies were cross-sectional; two were prospective. Most had small numbers of participants, which may have influenced the results. The two studies reporting only nonsignificant findings had the smallest number of participants (*n*=40 and *n*=42, respectively) [32, 39].

A summary of the findings suggests that, depending on whether the pain is a result of an inflammatory disease or a chronic pain process, different results are found.

#### **CONCLUSION**

Few studies were found where the relationship between salivary cortisol and somatic disease/illness was analyzed. Among these, a large proportion showed significant findings which, across outcomes (CVD disease, metastatic breast cancer, rheumatoid arthritis or pain), suggest that low morning cortisol levels, high evening cortisol levels and a low dynamic cortisol response to stress are related to poorer somatic health.

### **REFERENCES**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **CHAPTER 9**

# **Discussion and Concluding Remarks Based on the Scancort Group Review**

#### **Ulf Lundberg1,\*, Peter Garvin2 and Margareta Kristenson<sup>3</sup>**

*1 Professor Emeritus of Biological Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University, 106 91 Stockholm, Sweden and <sup>2</sup> Post-doc at the Department of Medical and Health Sciences, Linköping University, Sweden and <sup>3</sup> Professor in Social Medicine and Public Health Science at the Department of Medical and Health Sciences, Linköping University, Sweden* 

**Abstract:** The aim of this book was to evaluate the usefulness of salivary cortisol as a biomarker in various settings. Our hypothesis was that observed diversities in results can be a function of different kinds of assessments. In this chapter, we try to respond to this aim by giving a summary of the results from different cortisol measures in relation to the health-related variables and conditions investigated in this review. The overarching pattern shows a predominance of non-significant findings but also a couple of rather consistent trends emerged when comparing the results from different chapters. The most apparent is that single measures of absolute concentrations of salivary cortisol, for most health-related variables, seldom give significant findings; deviation measures, in terms of diurnal deviations and/or laboratory stress tests seem to be more strongly and consistently associated with a number of factors, such as Socioeconomic Status (SES), psychological characteristics, biological variables in terms of overweight and abdominal fat accumulation, and mental and somatic disease. Across disorders, the pattern related to ill-health/stress is generally characterized by a flatter diurnal cortisol curve, which in most cases is due to attenuated morning and/or increased evening levels, or a reduced response to a laboratory stress test. For some specific questions, single mean values seem to provide valuable information, but in all cases a careful design in terms of power and standardization is important. Thus, salivary cortisol can be a useful biomarker in many settings, if caution is taken in the choice of methods used.

**Keywords:** Salivary cortisol, adult, conclusions, nonsignificant findings, deviation measures, diurnal curve, laboratory test, dexamethasone, biomarker, disease.

#### **INTRODUCTION**

As emphasized in Chapter **1**, salivary cortisol is a very popular measure in research on stress, health and disease. However, results from different studies are sometimes contradictory and confusing and there has been frustration over diverging findings where both high and low cortisol levels have been associated with the same condition. Our hypothesis was that the observed diversities could be a function of different frames of reference, and especially different kinds of assessments. The aim of this book was to perform a critical review of the existing empirical literature on salivary cortisol, to evaluate the usefulness of salivary cortisol as a biomarker in various settings. More specifically, we investigated the results that we found for different cortisol measures in relation to different health-related variables and somatic and mental conditions.

This review is based on articles published up to October 1, 2009. The analysis of the papers has been done by sorting the evidence for each variable and condition according to the methods used for different measures of cortisol: single time points, deviation/slope, Area Under the Curve (AUC) measurements and laboratory and dexamethasone testing according to standardized schemes. The variables and conditions in this review comprise demography (age, sex, SES, ethnicity), psychosocial work stress, perceived stress, psychological resources, biological markers of Cardiovascular Disease (CVD) or inflammation and hormones, sleep, mental health, and somatic disease (CVD, rheumatoid arthritis, breast cancer, and pain).

**<sup>\*</sup>Address correspondence to Ulf Lundberg:** Professor Emeritus of Biological Psychology at the Department of Psychology and Centre for Health Equity Studies (CHESS), Stockholm University, 106 91 Stockholm, Sweden; Tel: +46 8 163 874; Fax: +46 8 167 847; E-mail: ul@psychology.su.se

The presentation of the results of significant findings defines positive as an association with higher single cortisol mean levels, steeper deviation, larger AUC, or greater reactivity in a laboratory stress test and a higher level of cortisol after dexamethasone suppression. In concordance, negative associations are set as lower single cortisol mean levels, a less steep deviation, smaller AUC, or lower reactivity in a laboratory stress test and a lower level of cortisol after dexamethasone suppression.

## **SUMMARY OF THE RESULTS**

The overarching pattern from this review shows a predominance of non-significant findings. For some measurements and settings, very few or no significant findings seem to arise, or significant findings were seen but with opposing messages; *i.e.*, high or low levels/deviations for the same variable or condition. In both cases these results suggest that this measure of biomarker, in these settings, is not informative. The results are summarized in the present chapter and the implications for choices of methods in different settings are discussed. In some cases non-significant findings were also common in cases where significant associations were expected on the basis of general physiology or assumptions. Possible explanations for these non-significant or noncoherent findings are also discussed in the present chapter.

However, a number of significant findings were also seen, which, after sorting the results according to comparable measures, did show a coherent pattern in several studies, suggesting that some types of cortisol measurements seem to be more informative for some types of correlates and conditions and even across different variables. On the basis of the conclusions from this discussion, some general recommendations are given for future studies on salivary cortisol as a biomarker in research on stress, health, and disease.

First a short summary of the main findings for each cortisol measure (see Table **1**) is given, followed by a brief summary of the findings for each health-related variable/condition.


**Table 1:** Overview of the findings in each chapter of the book

#### **188** *The Role of Saliva Cortisol Measurement in Health and Disease Lundberg et al.*


#### *Discussion and Concluding Remarks The Role of Saliva Cortisol Measurement in Health and Disease* **189**



#### *Discussion and Concluding Remarks The Role of Saliva Cortisol Measurement in Health and Disease* **191**



*Abbreviations*: a1, awake; a2, morning; a3, midday; a4, evening; a5, all day; b1, morning; b2, midday; b3, morning to evening; b4, laboratory test reactivity/recovery; c1, morning increase/ground; c2, midday increase/ground; c3, morning to evening increase/ground; c4, laboratory test increase/ground; d, DST, dexamethasone suppression test.

a Results for HDL cholesterol and heart rate variability have been inverted in this table as low HDL and low heart rate variability is considered a cardiovascular risk factor.

### **SUMMARY OF THE RESULTS WITH REGARD TO THE TYPE OF MEASURE (TABLE 1)**

In total 1188 analyses were reported. Of these 598 were single measures (50%), 292 were deviation measures (25%), 130 were AUC measures (11 %), 107 were taken in laboratory tests (9%) and 28 followed a dexamethasone test (2%). The overall proportion of significant findings was 30%, ranging from 0% to 60% among the different measures.

### **Single Time Points**

The overall proportion of significant findings for single time points was 27%, ranging from 16% (awakening) to 39% (morning levels).

### *Awakening*

One hundred and twenty-two analyses were found. Overall, the proportion of significant findings (in any direction) was 16%, which is the lowest proportion amongst the five single time points evaluated. Levels at awakening seem to be of most relevance for measures of sleep, *i.e.*, sleep duration and overall sleep quality with a positive association with a single measure at awakening (42% significant). Also, burnout was positively associated with a single measure at awakening (40% significant). Otherwise only very few significant findings, or divergent findings were seen.

## *Morning Value*

One hundred and seventy-five analyses were found. The overall proportion of significant findings (in any direction) was 39%, which is the highest proportion amongst the five single time points evaluated. The highest consistency for positive associations with morning values was found for depression (52%) and anxiety (50%). The highest consistency for negative associations with morning values was found for BMI and waist/hip ratio (53%) and pain (50%).

### *Midday*

Ninety-nine analyses were found for cortisol levels around midday. The overall proportion of significant findings was 22%. The highest consistency for positive associations with midday values was found for depression (42%) and increasing age (66%). However, the latter is based on three studies only. None of the tested variables had clear results for negative associations with cortisol levels at midday.

### *Evening*

One hundred and twenty-two analyses were found. The overall proportion of significant findings was 21%. The highest consistency for positive associations with evening values was found for depression (42%) and somatic disease (50%). The highest consistency for negative associations with evening values was found for high SES (43%).

## *Summary Measures Over the Day*

Eighty analyses were found. The overall proportion of significant findings was 30%. The highest consistency for positive associations with all-day means was found for somatic disease (50%), depression (37%), burnout (40%), and inflammatory markers (80%). However, all these proportions are based on a low number of studies. In addition, a moderate proportion of significant positive associations (with no negative associations found in the studies evaluated) were found for BMI (22%).

## *Deviations*

The overall proportion of significant findings was 36%, ranging from 26% (morning deviation/Cortisol Awakening Response (CAR)) to 55% (midday deviation, including peak and later).

## *Morning Deviation/CAR*

One hundred and thirty-nine analyses were found. Overall, the proportion of significant findings (in any direction) was 26%, which is the lowest proportion amongst the evaluated deviations. There were no clear positive relationships between CAR and any of the factors evaluated. The highest consistency for positive associations with a high morning deviation was found in sex analyses; women had higher CAR than men in three out of ten studies (30%). The highest consistency for negative associations with CAR was found for pain (72%), although one analysis suggested that high perceived pain is associated with a higher CAR. In addition, a moderate proportion of significant negative associations (with no positive associations found in the studies evaluated) were found for BMI and waist/hip ratio (25%).

## *Midday Deviation*

Twenty analyses were found. Overall, the proportion of significant findings (in any direction) was 55%, which is the highest proportion amongst the five deviations evaluated. The low number of studies makes it hard to fully interpret the results. However, the proportion of significant associations is to a large extent explained by analyses on BMI and waist/hip ratio, where all four out of four analyses showed a negative relationship, *i.e.*, lower deviation between two time points at midday associated with higher BMI and/or waist/hip ratio. Thus, there is a high consistency for BMI and waist/hip ratio with regard to midday deviations. A lower midday deviation was also associated with somatic disease and disturbed sleep, and the midday deviation was lower among men, but these are all based on single studies and should be interpreted cautiously.

## *Diurnal Deviation*

One hundred and thirty-three analyses were found. Overall, the proportion of significant findings (in any direction) was 37%. The highest consistency for positive associations with diurnal deviations was found for high SES (64%), psychosocial resources (50%) and sleep quality (50%). The highest consistency for negative associations with diurnal deviations was found for somatic disease (54%) and depression (40%).

## **Laboratory Stress Test**

One hundred and seven analyses were found for cortisol response/reactivity in a laboratory stress test setting. Overall, the proportion of significant findings (in any direction) was 45%. The highest consistency for positive associations with a cortisol response was found for sleep quality and sleep duration (75%). The highest consistency for negative associations with a cortisol response was found for disturbed sleep (75%), and depression (37%) and inflammatory markers (33%).

There were nine analyses found for cortisol recovery after a laboratory stress test. Overall, the proportion of significant findings (in any direction) was 44%. The low number of studies makes it hard to fully interpret the results. Two out of four studies reported that a poorer recovery is associated with higher levels of inflammatory markers, one out of four studies reported that a poorer recovery is associated with cardiovascular risk factors (other than BMI and waist/hip ratio).

### **AUC**

The overall proportion of significant findings was 27%, ranging from 9% (AUC with respect to increase throughout the day) to 60% (AUC at midday, including peak and later).

## *AUC Morning/CAR*

Fifty-five analyses were found for AUC with respect to increase following awakening. Overall, the proportion of significant findings (in any direction) was 22%. There are too few studies to evaluate consistency. The highest proportions of positive associations with AUC with respect to increase were found for disturbed sleep (50%) and depression (50%). However, both these proportions are based on single studies and should be interpreted cautiously. The highest proportions of negative associations with AUC with respect to increase were found for perceived stress (50%) and increased age (33%). Again, all these proportions are based on single studies and should be interpreted cautiously.

Forty-five analyses were found for AUC with respect to ground following awakening. Overall, the proportion of significant findings (in any direction) was 27%. There are too few studies to evaluate consistency. The highest proportions of positive associations with AUC with respect to increase were found for depression (66%), based on two significant associations. The highest proportions of negative associations with AUC with respect to ground were found for pain (50%) and perceived stress (50%). However, these proportions are based on single studies and should be interpreted cautiously.

## *AUC Midday*

Five analyses were found for AUC with respect to increase during the early part of the day. Overall, the proportion of significant findings (in any direction) was 60%, which is the highest proportion amongst the AUC measures evaluated. These were fully explained by three analyses on BMI and waist/hip ratio in which a higher BMI was negatively associated with AUC.

Fourteen analyses were found for AUC with respect to ground during the early part of the day. Overall, the proportion of significant findings (in any direction) was 14%. The two significant findings were: one where higher BMI was associated with a lower AUC; and one where psychosocial strain in the work environment was associated with a higher AUC. In the latter, an additional 11 studies showed non-significant findings.

## *AUC Throughout the Day*

Eleven analyses were found for AUC with respect to increase throughout the day. Overall, the proportion of significant findings (in any direction) was 9%. The only significant finding of the analyses was a higher AUC associated with somatic disease. However, there were five additional studies on somatic disease showing non-significant findings. Twenty-four analyses were found for AUC with respect to ground throughout the day. Overall, the proportion of significant findings (in any direction) was 25%. The highest consistency was found for a negative association between high SES and AUC (50%). The other two significant findings were a negative association between psychosocial resources and AUC (50%), and a positive association between perceived stress and AUC (100%).

### *AUC in Laboratory Stress Tests*

Nine analyses were found for AUC with respect to increase in laboratory stress tests. Four of them were significant. The highest consistency was found for inflammatory markers, showing an inverse association with AUC for salivary cortisol during a stress test (40%).

Ten analyses were found for AUC with respect to ground in laboratory stress tests. Overall, the proportion of significant findings (in any direction) was 50%. There are too few studies to evaluate consistency. Positive associations with AUC with respect to ground were found for depression (50%) and anxiety (50%). Negative associations with AUC with respect to ground were found for high SES (100%). Psychosocial resources had a study reporting a positive significant association (25%) as well as negative significant association (25%). However, all these proportions are based on single studies and should be interpreted cautiously.

### **Dexamethasone Suppression Test**

Twenty-eight analyses were found for dexamethasone suppression tests. Overall, the proportion of significant findings (in any direction) was 50%. The highest consistency for positive associations with a high cortisol level after suppression with dexamethasone was found for depression (33%). The highest consistency for negative associations with a high cortisol level after suppression with dexamethasone was found for vital exhaustion (42%). Studies on pain were inconclusive, with one study reporting a positive association (25%) and two studies reporting a negative association (50%). There are too few studies to evaluate consistency for the other factors. A positive association with a high cortisol level after suppression with dexamethasone was found for increased age (100%), perceived stress (100%), and hypertension (100%). A negative association with a high cortisol level after suppression with dexamethasone was found for high SES status (50%) and burnout (25%). However, all these proportions are based on a low number of studies and should be interpreted cautiously.

## **SUMMARY OF THE RESULTS WITH REGARD TO HEALTH-RELATED VARIABLES AND CONDITIONS**

### **Demographic Variables (SES, Age, Sex, Ethnicity) and Salivary Cortisol**

SES was usually defined in terms of education and/or income. In general, the results indicate that high SES is associated with a steeper cortisol slope and participants with high SES tend to have slightly lower mean cortisol levels, mainly due to lower levels in the evening. Health behavior, smoking in particular, seems to play an important role in this relationship, but a similar pattern with a flatter diurnal rhythm among low SES individuals has also been found among nonsmokers. This suggests a more normal and healthy cortisol secretion in individuals with high SES, which is in line with their generally better health and longer life expectancy compared with individuals with low SES. However, inconsistent and non-significant findings were also reported. Studies on ethnicity followed a clear trend: Caucasian study populations had a higher diurnal variation than African American study populations. It is also suggested that Caucasians had higher diurnal variation than Hispanics, who, in turn, had higher diurnal variation than African Americans. This ethnicity ladder is in congruence with a translation to a socioeconomic ladder, and give further support to the findings on SES, emphasizing that a higher status seem to be associated with a higher diurnal variation.

A small increase in cortisol levels during the later part of the day was seen with age, but no consistent sex differences in cortisol levels or responses were found.

### **Psychosocial Work Stress and Salivary Cortisol**

With regard to work related psychosocial stress, most analyses of the association with salivary cortisol were found to be non-significant, but some significant positive findings were found between high work stress exposure and high cortisol levels. No specific cortisol measure or statistical analysis could be related to more consistent significant findings. Furthermore, the two main measures of work stress, that is, the demand–control and the effort–reward–imbalance models, did not differ in terms of significant findings in relation to salivary cortisol. In Chapter **3**, a quality index was calculated for each study and it was found that the more recent studies tended to be of higher quality. However, a finding of concern was that lowquality studies tended to produce more significant findings than high-quality studies.

Only a limited number of studies has been performed on the possible influence of work stress on deviation measures, usually CAR. This means that no conclusions can be made regarding the use of such measures compared with single time point measures. A tentative stress-induced change in the diurnal pattern, in terms of decreased morning levels and increased evening levels, could lead to inconsistent findings from single measures of cortisol obtained at different times of the day (*e.g.*, morning vs evening measures) or if a mean cortisol level is calculated for the whole day.

Another possible explanations for the relative lack of consistent significant associations between work stress and single measures of salivary cortisol could be that the stress induced by regular work conditions was mild, and/or that most occupational groups were relatively homogenous with regard to stress levels (*e.g.*, by having the same occupation) and, therefore, there was too little variability to reveal possible associations. In addition, study groups were usually relatively small.

A third conclusion is that the effects of a stressful work environment are not mediated by cortisol, at least not as measured by saliva cortisol measures.

In summary, the main pattern from this review shows non-significant associations between salivary cortisol and work stress. This is consistent with previous studies on work stress and cortisol measures based on urine samples [1-3], which show that routine work conditions are associated with only a small increase or no increase at all in mean cortisol during the day, whereas another stress hormone, urinary adrenaline, is consistently increased 50–100% during ordinary work [4]. Only one study has used dexamethasone administration and found higher suppression in relation to high work stress (defined by low reward, high burnout, and high vital exhaustion). If vital exhaustion was an important factor in this relationship, this finding is consistent with results reported for the dexamethasone suppression test above (p. 15).

### **Perceived Stress, Psychological Resources and Salivary Cortisol**

Despite the fact that cortisol is an important stress hormone and known to increase in response to acute experimental stress, more than half of the studies failed to find a significant association between perceived stress and salivary cortisol. Furthermore, some of the significant associations found seem to be inconsistent with general expectations. For example, some studies showed that perceived stress was related to lower AUC cortisol, especially in the morning. One study found significantly lower cortisol in the afternoon (single time point) among individuals reporting high perceived stress. One study of the effects of dexamethasone showed, as expected, less suppression on the Hypothalamo-Pituitary-Adrenal (HPA) axis (response after awakening) in teachers with high levels of perceived stress. In conclusion, measures based on the Perceived Stress Scale (PSS) are not consistently related to salivary cortisol. In view of this, it is important to remember that PSS represents perceived stress during a longer period of time (weeks), whereas cortisol generally has been measured at a single time point and, thus, reflects a momentary value.

For measures of psychological resources, such as sense of coherence, internal locus of control, mastery, and self-esteem, many non-significant findings were seen, especially for single time measurements. Also 5/5 associations with CAR were non-significant. However, significant findings were seen for other deviation measures, but were different depending on the resource measure. High external locus of control and low levels of a combined measure of self-esteem and locus of control were related to a stronger cortisol response to a laboratory stress test. High scale scores of sense of coherence and of mastery were related to significantly lower cortisol baseline levels and two out of three studies on diurnal deviation showed a significant positive relationship, *i.e.*, steeper deviation related to high mastery. However, no significant associations were found between self-esteem and cortisol. In conclusion, for measures of psychological resources, significant findings differed depending on the measure used; there were few significant findings for single measures and CAR, but there were significant findings in relation to diurnal deviations or laboratory conditions in terms of reactivity or baseline levels, however not for self-esteem.

#### **Biological Markers and Salivary Cortisol**

The chapter on salivary cortisol in relation to a number of other biological markers generated very few significant associations and, for most markers, no consistent pattern appeared. However, for some markers there seems to be a consistency, which is also in line with expectations.

As several biomarkers tested do represent different time periods compared with salivary cortisol, strong correlations with single measures of cortisol cannot be expected. For example, BMI and waist circumference are measures that are very stable over time, urinary adrenaline and noradrenaline represent mean measures over a couple of hours, whereas blood pressure and heart rate represent momentary values that fluctuate considerably over time. Additional possible explanations for the lack of associations are the small number of studies performed with each of the biomarkers and the small study samples.

However, among the few significant findings found, higher BMI was related to lower cortisol in the morning (but not at awakening), higher total measure over the day, to a lower deviation and a lower AUC around midday. Similarly, higher waist circumference was significantly related to lower deviation in the daily slope of the cortisol curve in two out of three studies. Results from measurements of waist/hip ratio showed a similar pattern with lower cortisol level in the early phase of the diurnal cycle among participants with higher waist/hip ratio. Blood pressure and heart rate responses to a stress test indicate positive associations with cortisol deviation, whereas a negative association was found for heart rate variability. Total cholesterol, triglycerides, and glucose levels did not show any consistent association with any cortisol measure.

Thus, the response pattern for variables characterizing the metabolic syndrome (high BMI, abdominal fat accumulation, *etc.*) indicates a less dynamic activity of the HPA axis. Compared with single measures of blood pressure and heart rate, cardiovascular reactivity seems more associated with cortisol secretion under experimental stress. Low heart rate variability is a risk factor for cardiovascular disorders and tended to be associated with higher increase in cortisol levels.

With regard to inflammatory markers, there were fewer significant associations than expected, but the associations found were in line with the immunosuppressive function of cortisol. A couple of studies reported that a higher capability to react with cortisol secretion on a stress test and a good capability to recover after a stress test are associated with lower levels of inflammatory markers [5-7]. This, in combination with the reported positive association between a higher cortisol output throughout the day and interleukin-6 [8], are in line with earlier research on the role of cortisol in immunoregulation in studies on patient populations [9, 10]. Fanatidis *et al.* [9] have proposed that "inappropriately normal" cortisol levels due to limited capability to respond with increased cortisol levels may not be sufficient to limit an ongoing inflammation. Raison and Miller [10] describe a situation "when not enough is too much", with increased levels of cortisol due to downregulation of receptors on target cells, making the glucocorticoid signaling in immunoregulation insufficient.

The number of studies on salivary cortisol and plasma catecholamines is low. Adrenaline and noradrenaline cannot (so far) be measured reliably in saliva. Also, the stress response in terms of catecholamine secretion is more rapid (less than a minute after exposure) than the cortisol response (30–40 min to reach a peak after stress exposure). Moreover, studies using urinary levels of catecholamines and cortisol collected over a longer time adjacent to a laboratory stress test indicate only moderate positive correlations in response to stress [11,12], and it is suggested that there is sometimes a dissociation of the activation of these two systems [13,14]. In line with this, earlier studies on laboratory stress testing have demonstrated that longterm stress affects immunologic but not cardiovascular responsiveness to acute stress in humans [15] and Schommer *et al.* [16] found a dissociation between reactivity of the HPA axis and the sympathetic adrenal medullary system to repeated psychosocial stress. As noted earlier and in line with CATS (positive outcome expectancies), the HPA axis seems to deactivate quickly to normal repeated stress exposure, whereas the sympathetic nervous system continues to respond.

In summary, the number of well-controlled studies found on the relationship between salivary cortisol and other stress hormones and inflammatory markers is low. Although cortisol is known to be involved in many central biological processes of importance for health and disease, more studies are needed on other stress hormones and inflammatory markers to fully elucidate the feasibility and usefulness of salivary cortisol in that context.

#### **Sleep and Salivary Cortisol**

Surprisingly few studies have been performed on sleep and salivary cortisol and the measures of sleep vary considerably. The most consistent finding from the present review was a positive association between sleep duration and a single measure of salivary cortisol at awakening. Sleep duration also tended to be associated with low evening cortisol levels, and self-reported sleep quality was positively correlated with cortisol in the morning. Disturbed sleep tended to be associated with higher cortisol over the day and a more flat diurnal cortisol pattern. The general pattern from these studies indicates that a more dynamic cortisol response (high morning and low evening levels) is positively related to sleep quality. However, as time of awakening usually was not controlled, it is possible that individuals who occasionally are sleeping longer get up later and, therefore, have a higher cortisol level, due to the typical morning increase, but a consistently later wakeup should not influence the cortisol morning response.

Because sleep in terms of quality and quantity is such an important prerequisite for long-term health and for protection against stressful conditions, the conclusion from this chapter is that more specific studies on different indicators of sleep quality and amount of sleep in relation to salivary cortisol, controlled for time of wake up, are badly needed. Reduced sleep to less than 6 hours per night is known to have important metabolic consequences and increase the risk of health problems such as type 2 diabetes, increased susceptibility to infections [17], and musculoskeletal disorders [18]. Cortisol is likely to play an important role in these relationships but the results from the present review on sleep deprivation and salivary cortisol were inconsistent.

## **Mental Health and Salivary Cortisol**

A large proportion of non-significant findings were seen for measures of mental health. Some consistency is seen for Major Depressive Disorder (MDD), mainly higher mean levels. The results regarding single measures and depressive mood are less consistent, but the overall picture for depression shows poorer diurnal deviation and response to stress.

Inconsistency among papers studying depression seems to be related mainly to the study population, with stronger effects for more depressed individuals. A recent review study [19] showing a link between depression and HPA hyperactivity, but with great variation across patient groups, is in line with the present findings.

A recent study based on 408 population-based midlife women investigated the relationship between a measure of depressive symptoms and salivary cortisol obtained at 18:00 h, 21:00 h, and immediately on awakening the next morning [20]. It was found that the diurnal cortisol slope was significantly flatter for women with high depressive scores than for less depressed women, after adjustment for a number of possible confounders, except sleep. A flatter curve for individuals with higher depressive scores is consistent with the studies reviewed in Chapter **7**, but according to Knight *et al.* [20] their finding is mainly based on lower morning cortisol levels among women with more depressive symptoms rather than increased evening levels. This group of women represents individuals with relatively mild symptoms and very few individuals were suffering from MDD.

Anxiety was related to cortisol levels only in studies comparing groups with high versus low anxiety, but not when anxiety scores were treated as a continuous variable. However, very few significant findings were found for anxiety, and findings from different studies, and across different measures used, were not consistent. Thus, saliva cortisol does not seem to be strongly related to anxiety.

The proportion of significant relationships (in any direction) among the six articles studying Vital Exhaustion (VE) using the Maastricht Questionnaire was 2/14 (14%) for single measures, 3/7 (42%) for deviation measures, 0/1 study using AUC measures. Chapter **7** also indicates that VE was related to poor cortisol response to stress and/or a flatter diurnal deviation and to higher suppression after dexamethasone administration. Most of the statistical analyses do not show a significant relationship between burnout and cortisol, and when these are present, the results are inconsistent. One explanation seems to be the measures of burnout used, probably due to the different conceptual basis for burnout. VE measured using the Maastricht Questionnaire seems to be related to a poorer cortisol response to stress and poorer diurnal deviation. The coexistence of burnout and VE in many studies does make it difficult to conclude how the different concepts are related to cortisol.

A seemingly contradictory result arises between depression and vital exhaustion. Despite often being intracorrelated to a high extent, depression is positively correlated and vital exhaustion negatively correlated with high levels of cortisol after suppression with dexamethasone. It is reported that the higher cortisol values due to nonsuppression in depressed subjects is explained by a small proportion of the subjects. For most subjects, cortisol levels are suppressed by dexamethasone administration, even in the depressed groups. However, the differences might pinpoint a physiologic difference amongst depressed patients, where some patients with clinical depression exhibit generally higher cortisol levels throughout the day and an insufficient feedback on cortisol secretion. Vital exhaustion, on the other hand, is associated with lower levels throughout the day, and a prolonged suppression after dexamethasone administration. In view of these findings, it is of interest to note that Lindeberg *et al.* (2008) found a flatter diurnal cortisol curve (smaller deviation between morning peak and evening values) related to more exhaustion measured by the inverted SF-36 vitality scale [21].

A general conclusion for all mental health measures is that a large proportion of non-significant findings are due to low power and few sampling days combined with low contrasts between study groups and within study populations. Generally, deviation measures such as diurnal deviation seem to be more valid measures compared with single measures to capture possible changes in the HPA axis measured using salivary cortisol.

#### **Somatic Disease and Salivary Cortisol**

Few studies were found for somatic disease, however, among these a rather high proportion showed significant findings.

Salivary cortisol was related to CVD, breast cancer, Rheumatoid Arthritis (RA), and pain syndromes but, again, many non-significant findings appeared. The main pattern for CVD was associations with low morning and high evening cortisol and a flat diurnal curve and low stress reactivity. With regard to breast cancer, a similar pattern of associations was found, that is, high evening cortisol and low diurnal variation, however only for breast cancer patients with metastases. RA patients, especially those with high disease activity, were found to be characterized by higher evening levels of cortisol. Pain conditions were more inconclusive for single time points, but tended to be associated with high midday and evening levels, low CAR, and low diurnal variation in cortisol.

In conclusion, a low diurnal variation in cortisol seems to be the most typical finding in relation to somatic disease.

The present review was limited to the somatic diseases described above. However, cortisol is assumed to play an important role also in many other somatic diseases. For example, Cohen *et al.* [22] have shown that individuals with high cortisol reactivity to experimental stress are more susceptible to upper respiratory illness when exposed to stressful life events.

### **METHODOLOGICAL DISCUSSION**

The most important aim of the present review was to investigate if inconsistent results from different studies could be explained by different theoretical assumptions or by different ways to obtain and statistically analyse cortisol measures (single values, deviations, AUC, with samples either collected in a natural setting with ambulatory sampling or in laboratory stress tests). From the pattern in each chapter in this book, it was not possible to detect any consistent differences in results depending on the type of measure or theoretical assumption, generally due to insufficient number of studies for each type of measure. A predominant finding in each chapter, except for MDD, was that associations investigated between salivary cortisol and other variables were non-significant. Most studies are based on single time point cortisol measures and relatively few on more dynamic measures, such as slope of the curve, AUC, or on repeated measures over several days or reactivity and recovery of cortisol activity in response to acute stress exposure or dexamethasone administration. This makes it difficult to find consistent patterns and make conclusions regarding the feasibility of these different methods.

However, combining results across methods and over the different chapters, the overall pattern of findings indicates a trend. Although less clear for work stress, deviation measures of cortisol, such as diurnal pattern, seem to reveal somewhat stronger and more consistent associations than single time point measures. This was seen for SES, psychological resources, biological markers, including overweight and abdominal fat accumulation, sleep, and mental and somatic disorders.

Thus, a pattern that seems fairly consistent is that high exposure (low SES), low psychological resources, poor sleep, and various mental and somatic disorders are associated with a flatter diurnal cortisol curve. In most cases this is due to attenuated morning and increased evening cortisol, but in terms of major depression, increased cortisol levels were also found in the morning (but not at awakening). A flatter diurnal curve or reduced slope over the day is generally considered to indicate a dysregulation of the HPA axis. In view of the two stress models presented in the introduction, this is consistent with the assumption that certain long-term stress conditions, characterized by helplessness and hopelessness according to CATS due to negative outcome expectancies, and chronic or repeated stress according to the Allostatic Load Model, initiate/reflect pathologic biological processes.

For some measures and some settings, few significant results were seen, but when seen the results were inconclusive. This was the case for most of the single time measures but also for several other measures. This was especially the case for awakening cortisol, and most of the single cortisol levels for all health measures except for major depression and sleep duration. These measures, therefore, do not seem to be informative biomarkers. For some measures, *e.g.*, the PSS scale and for work stress, few significant findings were seen.

A non-significant finding does not necessarily mean that no association exists. Methodological shortcomings can in some cases explain the lack of significant findings. Many studies were performed with single time point cortisol measures and samples of less than 50 individuals, which after taking sex, differences in stress levels, various confounders, and subgroups into consideration, often become too small to reveal possible differences even if they exist. Low statistical power also increases the risk of significant chance findings and, consequently, inconsistent results.

A major weakness of many studies included in this review is the low statistical power. In view of the great natural variability in cortisol levels during the day, between different days, and between individuals, a large number of measurements and individuals are necessary to be able to demonstrate associations that may exist but are of low effect size.

In most cases just a few or a single cortisol measure has been obtained, usually on one single day. Because of the large day-to-day variation of cortisol, measures from more than one day are needed for good reliability of the measure. This is also the case for good validity, when the research question is about the ability to respond and relax, in general, among people with chronic mental health conditions, such as depression, anxiety, burnout, psychological resources (sense of coherence, locus of control), or in relation to demographic variables. It seems likely that a more successful approach would be to relate these conditions to cortisol levels measured on several occasions over several days or weeks or in response to acute stress under carefully controlled conditions. As an example, the study by Cohen *et al.* [23] revealed small but significant differences in mean cortisol levels measured over several days in relation different levels of SES defined by education and income. An alternative approach to study possible associations between cortisol and stable conditions could be the use of urinary cortisol, which can be measured more easily by collecting urine over longer periods of time. However, such mean measures do not reveal information about the dynamics of the HPA axis.

In addition, strict control of confounders and compliance (*e.g.*, time of sampling) among participants is of critical importance for valid findings. As shown in the tables, many studies were performed without proper control of important confounders and of compliance and, in some cases, the possible influence of these factors was not even discussed. Some confounders are more important than others. For example, demographic factors such as age and even sex seem to have little or moderate influence on the relation between cortisol and other variables, whereas other factors such as time of the day, medication and cigarette smoking may have more pronounced effects.

A particular problem is the compliance among participants when obtaining a saliva sample at awakening, when experimenter control is very unusual and difficult to arrange. Even small variations in the time of saliva sampling in relation to wake up cause large variations in cortisol levels, which may produce misleading results. This may be one reason for the inclusive results from CAR measures.

The comparability of study groups is also important, and comorbidity, which is very common in many disorders, may cause different results in different studies depending on the composition of the patient group. Somatic disease often causes mental disorders and mental disorders may cause somatic symptoms, which could make conclusions regarding associations with specific conditions unclear. For example, in Chapter 8 it was found that patients with coronary artery disease differed in cortisol levels depending on their amount of depression. Lack of control of certain forms of medication may also contribute to inconsistent findings among patient groups.

Severity of the condition could also be of importance as indicated by the somewhat different results reported in Chapter **7** between cortisol and depressive symptoms versus major clinical depression. Also the duration of the condition could be of importance. Possible associations with the HPA axis of certain mental and physical health conditions may vary over an early and a late phase of a disease. It has been hypothesized that under certain stressful conditions the HPA axis first responds with overactivity (hyperresponsiveness) which gradually changes to underactivity (hypo-responsiveness), depending on changes in regulatory mechanisms (*e.g.*, increased or decreased receptor expression or number of receptors) [24], but empirical evidence for this assumption is weak. Theoretically, if several mental conditions are characterized by a gradual change in HPA activity from hyper- to hypoactivity, studies based on patients who have suffered from their condition longer or shorter periods of time are expected to give different results. If patients with different levels of chronicity with regard to their conditions are mixed in the same study, nonsignificant findings would be expected.

Comparisons within too homogeneous populations are also unlikely to reveal significant findings. Weak positive or most often non-significant associations were found between psychosocial work stress and cortisol levels (Chapter **3**). As these measures are well known to predict cardiovascular and other disease, these results could be surprising. A possible explanation could be the characteristics of the study populations in terms of healthy worker effects. According to the CATS model, ordinary workers have positive expectancies about their ability to handle their daily work demands. This is assumed to initially induce a short-term (phasic) activation of the HPA axis, followed by a rapid return to baseline and successive reduction in the stress response with repeated exposure to the same conditions. This represents a normal, healthy, and economic response pattern, as also suggested by the Allostatic Load Model as well as the adaptation stage of Selye's GAS. This means that the investigations on single individuals who continue to respond (chronic strain) to repeated work stress, who do not respond at all, or are unable to relax in the afternoon, may be of particular interest in order to identify work conditions that may cause health problems.

Many of the studies were not designed for investigating correlations between salivary cortisol and other biological or psychological variables. The design of the study was in these cases aimed at other purposes, which may explain the low statistical power for the analyses of interest for the present review. This means that the general conditions for investigating the relationships with salivary cortisol were not optimal in these studies.

An additional important factor to consider in this context is the possibility of publication bias. Studies showing significant findings are assumed to be more likely to be published compared with studies showing non-significant findings. This would mean that the actual number of non-significant findings could be even greater than found in the present review. However, in several cases the correlations reported in this review were only one part of studies with other aims and, therefore, the non-significant findings did not influence the decision to publish very much. Nevertheless, it is likely that even more non-significant results would have been obtained if unpublished studies had also been included in this review.

It is important to see that our method of comparing studies is pragmatic, and, to a large extent, the results are mainly dependent on significant findings and the direction of the results. The number of participants varies considerably between studies and, in a few cases, one single study was based on more participants than all other studies investigating the same association added together. An example is the Whitehall II study by Steptoe *et al.* [8] with 2873 participants. This means that one single study could outweigh all the others and makes a simple calculation of the number of significant findings in relation to the number of non-significant findings misleading. However, the way in which data are handled is also of importance. In the Whitehall study average cortisol levels were calculated from awakening to bedtime and, thus, do not reflect changes in diurnal pattern. For example, the same mean level could be obtained by high morning and low evening levels as by low morning and high evening levels. A steep curve represents a normal healthy diurnal secretion of cortisol, whereas a flatter curve indicates a dysregulation of the HPA axis. Thus, even a study based on a large number of participants may suffer from other weaknesses and, therefore, may not be more informative than smaller studies. Additional weaknesses in large studies could be lack of sufficient variation in the variables investigated, too few cortisol measurements, or insufficient control of confounders.

#### **CONCLUDING COMMENTS AND RECOMMENDATIONS**

Despite several shortcomings mentioned above, some general conclusions and recommendations for future studies can be made. The overall conclusion is that the studies reviewed in this book show very few significant associations between salivary cortisol and the psychological and biological variables and health conditions investigated. The significant associations found were not very strong but, assuming that publication bias was not a major issue, certain associations between salivary cortisol and health-related conditions seem to exist. A pattern that emerges when summarizing the findings from all chapters is that mental and somatic disorders tend to be associated with a rather flat diurnal cortisol curve, which was due to (except for major depression) lower cortisol levels in the morning and increased levels in the evening. These conditions, including depression, also tended to be associated with an attenuated cortisol response to stress exposure and in the few cases with the dexamethasone stress test, a lack of suppression of HPA activity (except for vital exhaustion). Psychological resources including high SES, on the other hand, seem to be associated with a steeper diurnal cortisol curve and in a few cases lower mean cortisol levels.

In order to use salivary cortisol as a biomarker of psychological conditions and mental and somatic health and in relation to other biological variables, it is concluded that single absolute levels are usually not very informative. Repeated measurements over several days, where means and measures of deviation/slope over the day can be obtained, and large groups of participants, are necessary to investigate these relationships in view of the great natural variability in cortisol levels during and between days and between individuals. For some associations, mean levels of cortisol can be relevant, but in most cases the dynamic properties of the HPA axis are likely to be more important, except perhaps for CAR.

For example, the diurnal slope measured over two or more days and the response to a stress test or dexamethasone administration are recommended as more useful measures of the dynamic function of the HPA axis. In addition, strict control of relevant confounders and compliance (*e.g.*, time of sampling) among participants is of critical importance for valid findings. As shown in the tables, many studies were performed without proper control of important confounders and of compliance and, in some cases, the possible influence of these factors was not even discussed. A particular problem is the compliance among participants when obtaining a saliva sample on awakening, when experimenter control is very unusual and difficult to arrange. Even if the participant reports that he or she took the sample immediately on awakening, it is known that a large proportion of the participants deviated considerably in their time of saliva sampling.

As indicated by this review, many of the associations calculated between salivary cortisol and other variables were not part of the main aim of the study, but rather the result of secondary analyses. In future studies, the design of studies aimed at investigating the association between salivary cortisol and stress, mental and somatic disorders and health-related biological variables should be based on established theoretical formulations such as CATS and Allostatic Load and be optimal with regard to the possibilities to accept or reject the hypotheses regarding the associations of interest. This includes sufficient number of participants to give enough statistical power, strict control over possible confounders and compliance among participants, and repeated measurements at specific hours of the day and, when relevant, measures of responsiveness and recovery of activity in the HPA axis during and after stress exposure and/or use of the dexamethasone suppression test.

Other important points to consider are that when patient groups are compared with healthy controls, patients should be diagnosed with reliable instruments and be rather homogeneous with regard to medication and how long they have had their symptoms.When studying the association between salivary cortisol and mental and somatic disease, the severity and duration of the condition could be of importance for the findings. Another problem is comorbidity, which is very common in many disorders, and which may cause different results in different studies depending on the composition of the patient group. Lack of control of certain forms of medication may also contribute to inconsistent findings among patient groups.

Investigations of the possible effects of different levels of work stress on cortisol secretion should be performed with individuals exposed to a great variety of stress levels and not with a homogeneous group of workers at a specific work place (having the same work tasks).

As pointed out in the introduction of this book, salivary cortisol is a very convenient measure and, therefore, has been used extensively. This could easily lead to misuse of this measure. The surprisingly great number of non-significant findings reported cannot be ignored and indicates that many of the associations investigated in this review do not exist or are very weak. However, lack of significant findings and misleading results could in some cases also be due to lack of control over a number of factors important for adequate testing of the hypotheses. It is obvious that the use of salivary cortisol as a biomarker of stress and health is more complicated than first assumed and, therefore, these associations need to be investigated under more optimal and carefully controlled conditions and be based on established theoretical models before valid conclusions can be drawn. However, this measure could be a very important tool in further understanding the links between stress, health and disease.

### **REFERENCES**


#### **204** *The Role of Saliva Cortisol Measurement in Health and Disease Lundberg et al.*


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode

# **Index**


**Margareta Kristenson, Peter Garvin and Ulf Lundberg (Eds) © 2012 The Author(s). Published by Bentham Science Publishers**


#### **Index**


© 2012 The Author(s). Published by Bentham Science Publisher. This is an open access chapter published under CC BY 4.0 https://creativecommons.org/licenses/by/4.0/legalcode