Associations of Salivary Cortisol Levels with Metabolic Syndrome and ...

ORIGINAL E n d o c r i n e

ARTICLE R e s e a r c h

Associations of Salivary Cortisol Levels with Metabolic Syndrome and Its Components: The Multi-Ethnic Study of Atherosclerosis A. S. DeSantis, A. V. DiezRoux, A. Hajat, S. H. Golden, N. S. Jenny, B. N. Sanchez, S. Shea, and T. E. Seeman Center for Social Epidemiology and Population Health (A.S.D., A.V.D., A.H.), Department of Epidemiology, University of Michigan, Ann Arbor, Michigan 48109; Department of Medicine–Endocrinology (S.H.G.), Department of Epidemiology, Johns Hopkins University, Baltimore, Maryland 21205; Department of Pathology (N.S.J.), College of Medicine, University of Vermont, Burlington, Vermont 05405; Department of Biostatistics (B.N.S.), University of Michigan, Ann Arbor, Michigan 48109; Department of Epidemiology (S.S.), College of Physicians and Surgeons, Columbia University, New York, New York 10032; and Department of Epidemiology (T.E.S.), Department of Medicine, University of California-Los Angeles, Los Angeles, California 90095 Context: Prior research has identified associations between social-environmental factors and metabolic syndrome (MetS) components. The physiological mechanisms underlying these associations are not fully understood, but alterations in activity of the hypothalamic-pituitary-adrenal axis, a stress-responsive biological system, have been hypothesized to play a role. Objective: The aim of the study was to determine whether MetS diagnosis and specific clusters of MetS components (waist circumference, high-density lipoproteins, glucose, and blood pressure) are associated with cortisol levels. Design and Setting: We conducted cross-sectional analyses of data from the Multi-Ethnic Study of Atherosclerosis (MESA) study in the general community. Patients or Other Participants: We studied a population-based sample of 726 adults (ages 48 to 89 yr) who do not have clinical diabetes. Intervention(s): There were no interventions. Main Outcome Measure(s): Cortisol awakening response, cortisol decline across the waking day, and total cortisol output were analyzed (using 18 timed measures of salivary cortisol over 3 d). Results: Overall, we found little evidence that the presence of MetS or its components is related to cortisol output or patterns. Contrary to expectation, the presence of MetS was associated with lower rather than higher area under the curve, and no consistent pattern was observed when MetS components or subsets of components were examined in relation to cortisol. Conclusions: Our findings do not support the hypothesis that differences in level or diurnal pattern of salivary cortisol output are associated with MetS among persons without clinical diabetes. (J Clin Endocrinol Metab 96: 3483–3492, 2011)

N

umerous studies have identified associations between social or psychosocial factors, such as lower socioeconomic status and increased work stress, and metabolic syndrome (MetS) components (1, 2). The physiological

mechanisms underlying these associations have not been fully explained, but alterations in activity of the hypothalamic-pituitary-adrenal (HPA) axis, a stress-responsive biological system, have been proposed as a potential con-

ISSN Print 0021-972X ISSN Online 1945-7197 Printed in U.S.A. Copyright © 2011 by The Endocrine Society doi: 10.1210/jc.2011-0483 Received February 22, 2011. Accepted August 2, 2011. First Published Online August 31, 2011

Abbreviations: AUC, Area under the curve; BMI, body mass index; CAR, cortisol awakening response; HDL, high-density lipoprotein; HPA, hypothalamic-pituitary-adrenal; MetS, metabolic syndrome; SES, socioeconomic status; WC, waist circumference.

J Clin Endocrinol Metab, November 2011, 96(11):3483–3492

jcem.endojournals.org

3483

3484

DeSantis et al.

Cortisol and Metabolic Syndrome

tributor to observed links between adverse psychosocial conditions and poorer metabolic functioning (3, 4). When an individual encounters psychological or physical stressors, the HPA axis responds by increasing cortisol production (5–7). Cortisol levels also follow a strong circadian rhythm, peaking shortly after awakening, declining over the day, and reaching a nadir around 2400 h (8, 9). Because the HPA axis affects metabolic system activity as a part of its regular function (5), it is plausible that atypical HPA axis activity could contribute to metabolic alterations. Higher fasting cortisol levels have been associated with higher systolic blood pressure, insulin resistance, fasting glucose, triglycerides, and central adiposity in population-based samples (10 –13). Individuals with extremely high cortisol secretion (e.g. Cushing’s syndrome patients) typically have higher levels of triglycerides, lower levels of high-density lipoprotein (HDL), and increased risk of hypertension (13). Postprandial and stress-related cortisol increases have also been found to be associated with adverse lipid profiles (14) and abdominal obesity (15). Moreover, atypical cortisol rhythms (i.e. low morning levels and decreased variability in cortisol levels across the day) have been associated with metabolic function and lipid profiles (16). Decreased cortisol variability across the day is associated with failure to experience nocturnal blood pressure dipping (17). Several studies have investigated associations between cortisol levels and type 2 diabetes mellitus, with most finding that diabetes is associated with higher cortisol levels at various times of day (18 –21). Higher cortisol levels have also been related to higher fasting glucose levels, glycosylated hemoglobin, and systolic and diastolic blood pressures in patients with type 2 diabetes (21). Chronic diabetes complications have also been associated with fasting cortisol levels, 24-h urinary free cortisol, and higher cortisol after dexamethasone suppression tests (22). However, fewer studies have investigated associations of cortisol with MetS diagnosis or MetS components among persons without diabetes. The presence of these associations would lend support for an etiological role of psychosocial stress in the development of the metabolic alterations linked to diabetes. Moreover, identification of specific MetS components or clusters of MetS components that are more strongly associated with cortisol would provide clues regarding the key physiological pathways involved. We used a large population-based sample with repeat measures of cortisol over 3 d and well-characterized metabolic markers to investigate the following: 1) is there an association between selected features of diurnal cortisol rhythms or total daily cortisol output and the presence of the MetS? and 2) are there specific clusters of components


(e.g. hypertension, glucose, and triglycerides) that are more strongly associated with cortisol parameters? We hypothesized that the presence of MetS would be associated with increased cortisol output and flatter decline across the day. In exploratory analyses, we also investigated whether any components or sets of components were especially predictive of cortisol patterns.

Subjects and Methods The data used in these analyses come from the Multi-Ethnic Study of Atherosclerosis (MESA), a longitudinal study designed to prospectively examine risk factors for subclinical cardiovascular disease and its progression to clinical disease. At baseline, the main study included 6814 participants between 45 and 84 yr old without clinical cardiovascular disease from six U.S. cities. Various population-based approaches (e.g. sampling from lists of area residents, Medicare and Medicaid services lists, and random digit dialing) were used to recruit participants (22). In 2003, an ancillary study, the MESA Stress Study, was initiated in conjunction with exams 3 and 4 at two MESA study sites (New York and Los Angeles) (23). Participants were enrolled in the ancillary study until approximately 500 participants were enrolled at each of the two sites, resulting in an approximately random sample of Black, White, and Hispanic participants in each site. The demographics of the MESA Stress Study were similar to those of the larger parent study, although there were slightly fewer persons in the oldest age category (75⫹ yr old) (18.2% in the parent study vs. 12.1% in the ancillary study) and slightly more males (47.6 vs. 44.7%) and more participants with some college education (29.7 vs. 23.9%) in the Stress Study. The Stress Study excluded Chinese participants but included more minority participants, due to the location of participating centers (New York and Los Angeles). MESA and MESA Stress protocols obtained Institutional Review Board approval at participating sites, and all participants provided written informed consent.

Cortisol measures and sampling protocol Participants were asked to provide six salivary cortisol samples per day (upon awakening, 30 min after awakening, 1000 h, 1200 h (or before lunch in the event that lunch occurred before 1200 h), 1800 h, and bedtime) over 3 weekdays. Samples were collected using Salivette collection tubes and stored at ⫺20 C in the laboratory until being assayed. Saliva samples were centrifuged at 3000 rpm for 3 min and assayed for cortisol levels using a chemiluminescence assay with high sensitivity of 44.72 nmol/liter (IBL International, Hamburg, Germany). Average intra- and interassay coefficients of variation were below 8%. Before analyses, original cortisol values (in nmol/liter) were logarithmically transformed to adjust for a positive skew in the distribution. Participants were informed that compliance with sampling protocols was monitored using time-tracking devices, which time-stamp bottle openings (trackcaps, MEMS 36; Clia Waived, San Diego, CA). Use of such devices is associated with increased compliance with sampling protocols (24). Participants also provided self-reports of wake times.


MetS assessment Criteria for MetS in these analyses were based on the National Cholesterol Education Program (NCEP) guidelines (25). MetS was determined to be present if three or more of the following MetS components were present: elevated triglycerides (ⱖ150 mg/dl); low fasting serum HDL cholesterol (ⱕ40 and 50 mg/dl among males and females, respectively); elevated fasting plasma glucose (ⱖ110 mg/dl); elevated systolic or diastolic blood pressure (ⱖ130/85 mm Hg or antihypertensive treatment); and increased waist circumference (WC) (⬎102 or ⬎88 cm in males and females, respectively). WC was measured and fasting blood samples drawn by research staff during the MESA exam (between 0730 and 1030 h). Serum was stored at ⫺70 C and shipped on dry ice to the University of Vermont Central Blood Analysis Laboratory. WC was measured at the umbilicus in duplicate, and values were averaged. In addition to the presence vs. absence of MetS, we also investigated whether a count of the number of MetS components (0 to 5) and whether particular combinations of two, three, four, or five components were associated with cortisol measures. Finally, we created a summary score of MetS components (by summing the standardized levels of each component and dividing by five), which we analyzed in relation to cortisol measures. Participants who had a fasting glucose of at least 126 mg/dl or were taking insulin or oral diabetes medication were considered diabetic and were excluded.

Covariates All analyses were adjusted for demographic characteristics (e.g. age, race/ethnicity, sex) previously found to be related to cortisol levels (23, 24, 26). Race/ethnicity categories included Black/African-American, Hispanic/Latino, and non-Hispanic White. Participants reported on their assets from the following list: owning one or more cars, owning or paying mortgage on a home, owning land, or owning investments (e.g. stocks, bonds). A five-point wealth index (range ⫽ 0 to 4), was created from responses (one point for ownership of each of these assets). Income was originally reported on a 13-category scale, which was later divided into quintiles. Quintiles were coded from 0 to 4 and summed with the assets variable to create an income-wealth index (23). Controls for study day and site were also included. Analyses were also adjusted for health behaviors (physical activity and smoking), due to prior research indicating that they may serve as confounders of cortisol-MetS associations. Intentional physical activity was assessed by daily diary on the days of saliva sampling. Participants reported on whether or not they had engaged in vigorous exercise (yes/no) and the amount of time (in minutes) they spent doing so. Smoking status (current, former, or nonsmoker) was assessed by questionnaire.

Exclusion criteria In total, 1002 participants were enrolled in the MESA Stress Study. Days for which there were insufficient trackcap data on sampling times (n ⫽ 127), no saliva samples (n ⫽ 15), or no data on wake times (n ⫽ 5) were excluded. Participants taking corticosteroid medications (n ⫽ 35) and those who failed to provide valid data on all five MetS components and at least 12 valid cortisol measurements were excluded. Individuals with clinical diabetes (n ⫽ 150) were excluded because these analyses aimed to examine associations between cortisol activity and MetS among participants without diabetes. Individuals with impaired


3485

fasting glucose based on American Diabetes Association criteria (110 –125 mg/dl) were included. The final analytical sample included 726 participants who provided 2170 d of data.

Analyses The presence of MetS and individual MetS components were analyzed in relation to the following: the cortisol awakening response (CAR), i.e. the difference between the 30-min and wake-up samples (a measure of reactivity and influenced by physiological mechanisms distinct from those that regulate the rest of the diurnal cycle) (27); the decline between wake-up and bedtime (excluding the “wake-up ⫹ 30 min” sample); and the total daily cortisol output [area under the curve (AUC)]. To estimate associations of MetS and its components with cortisol slope/decline, CAR, and AUC, we first created summary measures of each cortisol parameter for each participant for each day. Day-level rates of decline were calculated by regressing logged cortisol levels on time since waking in hours, excluding the second (30-min) sample for each individual person-day. Daylevel CAR were determined by subtracting wake-up cortisol levels from the values of the second sample (approximately 30 min after waking). Only samples provided between 15 and 60 min of waking were included in CAR analyses for the day. Finally, daylevel total cortisol output (AUC) was calculated by measuring the size of a series of “polygons” between each set of time points based on the sampling protocol: 0 to 30 min after awakening; 30 min after awakening and 1000 h; 1000 to 1200 h (or before lunch in the event that lunch occurred before 1200 h); 1200 to 1800 h; and 1800 h to 16 h after awakening. Data for the day-level cortisol parameters (decline, CAR, and AUC) were then analyzed in relation to the presence of MetS, MetS components, and clusters (of two or more components) using multilevel models with random intercepts that account for within-person correlations across the 3 d. Race, gender, age, socioeconomic status (SES), wake time, smoking status, physical activity, and study day and site were included as covariates in all models. Following prior work (28), we inferred that a particular grouping of components was associated with a feature of the curve if there was an interaction (or synergism) between its components on an additive (linear) scale. For each cluster of two or more components out of the five, we ran a model that included the main effects of all five components separately, plus a dummy variable (1/0) that reflected whether the grouping (cluster) was or was not present. The statistical significance of the latter term (equivalent to an interaction term between the components in the cluster) served to indicate whether the combination of factors in the cluster predicted above and beyond the factors separately. For example, a positive, statistically significant interaction term indicates that the predicted value of the cortisol feature in the presence of the cluster exceeds what would be expected from summing the individual values associated with each of the components. Main analyses treated all components as dichotomous, but sensitivity analyses were also conducted using continuous versions. Because analyses of clusters involve multiple models and comparisons, the Hochberg method was used to account for multiple comparisons and to determine levels of statistical significance of the interaction (cluster) terms (29). For comparability with prior work, other P values are not adjusted for multiple comparisons. In addition to analyzing isolated day-level cortisol parameters, in sensitivity analyses, we also examined associations of the presence of MetS, the summary MetS score, and the number of

3486

DeSantis et al.



TABLE 1. Selected sample characteristics and presence of MetS by demographic, socioeconomic, and behavioral characteristics: the MESA stress study % or mean (SD) Demographic factors Race/ethnicity (% distribution) Black Hispanic White Age category (% distribution) 48 –54 yr 55– 64 yr 65–74 yr 75– 89 yr SES index (% distribution)a 1 2 3 4 5 6 7 8 9 Health behaviors Nonsmokers (%) Former smokers (%) Current smokers (%) Activity levels, mean (SD) Mean levels of MetS components, mean (SD) Glucose (mg/dl) Triglycerides (mg/dl) WC (cm) HDL (mg/dl) Systolic BP (mm Hg) Diastolic BP (mm Hg) % with MetS components (dichotomized as defined below) High glucose High triglycerides High WC Low HDL Hypertensive No. of components, mean (SD) Any combination of three or more componentsb THG THB THW TGB TGW TBW HGB HGW HBW GBW TGWH TGWB TGBH TWBH GWHB All five components

% with MetS

26 51 22

23 35 20

17 31 33 19

26 27 29 32

6 13 16 13 12 12 12 9 6

44 31 38 33 35 23 22 12 17

52 38 10 10 min (21.20)

29 28 24

93.6 (9.8) 124.6 (70.0) 98.9 (14.4) 51.5 (9.8) 122.1 (20.3) 70.0 (9.9) 7 28 55 35 46 1.71 (1.22) 28 0.01 1.5 5.4 0.4 0.4 2.5 0.0 0.8 5.1 1.1 1.0 1.1 1.4 6.5 1.4 0.7

86 69 46 65 41

BP, Blood pressure. a

Participants were asked to report on their assets from a selected list that included the following: owning one or more cars, owning or paying mortgage on a home, owning land, or owning an investment (e.g. stocks, bonds, etc.), which was used to create a five-point wealth index (range ⫽ 0 to 4). Scores from this index were combined with reports of total family income, which was divided into quintiles (range ⫽ 0 to 4). Scores were added to create an income-wealth index (range ⫽ 1 to 9) that has been used in prior MESA analyses (24). b

Key for clusters of three or four components is as follows: T, high triglycerides; G, high glucose; W, high WC; H, low HDL; B, hypertensive.



3487

TABLE 2. Associations of MetS diagnosis with CAR, decline across the day, and AUC Cortisol parameter

MetS

Summary score

No. of MetS components

Difference Confidence interval Difference Confidence interval Difference Confidence interval CAR 1.67% ⫺6.66, 10.26 0.87% ⫺5.79, 7.64 ⫺1.55% ⫺4.83, 1.65 Rate of hourly decline 0.87% ⫺4.83, 1.65 0.20% ⫺0.37, 0.78 0.09% ⫺0.18, 0.35 AUC (nmol/liter 䡠 d) ⫺530.28* ⫺991, ⫺69 ⫺457.8⫹ 918, 2.63 ⫺209.14* ⫺385, ⫺33.5 All analyses included the following covariates: age, race/ethnicity, gender, SES, physical activity, smoking, wake time, study day, and study site. MetS columns present the percentage differences in the size of the CAR and rate of hourly cortisol decline, as well as the difference in total cortisol output between participants with MetS diagnoses (Y/N) compared to those without. Positive percentage differences in the CAR indicate a more pronounced CAR. Positive percentage differences in the decline indicate a less pronounced, flatter decline. Summary score columns present the percentage differences in the size of the CAR and rate of hourly cortisol decline, as well as the difference in total cortisol output, associated with each one-unit increase in MetS summary scores (i.e. the sum of the standardized continuous versions of all five MetS components). No. of MetS components columns present the percentage differences in the size of the CAR and rate of hourly cortisol decline, as well as the difference in total cortisol output, associated with each additional MetS component (high glucose, low HDL, high WC, hypertension, and high triglycerides; range ⫽ 0 to 5). ⫹

P ⬍ 0.10.

* P ⬍ 0.05.

MetS components with cortisol levels using spline models that separate the declines into two periods, following prior work (23, 30). We modeled logarithmically transformed cortisol levels as a function of time since waking, using linear regression mixed model splines with “knots” located at 30 and 120 min after wake time (23), with results stratified by gender. These models were run for analyses of MetS only because they rapidly become unwieldy and difficult to interpret when many interactions are included (as would be necessitated in analyses involving multiple MetS components). In addition, to facilitate comparisons with other work, we also examined whether MetS diagnosis was associated with cortisol values for specific individual samples (e.g. wake-up and bedtime).

scores (calculated by summing the standardized versions of the continuous versions of all five MetS components); or with the total number of MetS components (range, 0 to 5) (Table 2). Because cortisol measures were log-transformed, exponentiated coefficients from CAR and decline models can be interpreted as percentage differences in the magnitude of the CAR or the hourly decline. A positive value for the percentage differences indicates more pronounced CAR or a flatter decline. AUC was not log-transformed, so coefficients can be interpreted as mean differences in AUC values on the original scale. There were no associations of MetS diagnosis, MetS summary scores, or number of MetS components with the magnitude of CAR

Results The 726 participants included 191 Blacks (26%), 373 Hispanics (51%), and 162 non-Hispanic Whites (22%) (Table 1), of whom 355 (49%) were male and 371 (51%) were female. Participants ranged from 48 to 89 yr old. Twelve percent of participants met criteria for MetS, and average levels of MetS components are summarized in Table 1. Cortisol levels increased an average of 35% (SD ⫽ 0.51) between waking up and 30 min after awakening. On average, cortisol levels decreased by 11% (SD ⫽ 0.05) per hour when the second (CAR) sample was excluded. The average total daily cortisol output (AUC) was 6958.7 nmol/liter 䡠 d (SD ⫽ 3436). Intraclass correlation coefficients across the 3 d were 0.30, 0.35, and 0.53 for the CAR, decline, and AUC, respectively. Intraclass correlation coefficients for CAR, decline, and AUC were 0.23, 0.33, and 0.55, respectively, among those with MetS; and 0.33, 0.47, and 0.53 among those without MetS. We first explored whether cortisol activity was associated with the presence of MetS diagnosis; MetS summary

Cortisol (nmol/L) 25

20

15

10

5

0 Wakeup W+30

10 AM

Noon

6 PM

Time Since Wakeup Metabolic Syndrome

No

Yes

FIG. 1. Cortisol profile across the waking day stratified by the presence or absence of MetS.

Bedtime

3488

DeSantis et al.



or the hourly rates of cortisol decline, after adjusting for age, race/ethnicity, gender, SES, physical activity, smoking, wake time, and study day and site. MetS diagnosis and the number of MetS components were associated with significantly reduced AUC; summary scores were associated with marginally lower AUC (P ⬍ 0.10). In spline models, the presence of MetS, higher summary z-scores of MetS components, and higher number of MetS components were all associated with lower wake-up cortisol levels but were not associated with the size of the CAR or early or late decline (data not shown). Figure 1 shows unadjusted cortisol levels among participants with and without MetS. In the next set of analyses (Tables 3–5), we investigated whether any of the individual MetS components, pairs of

components, or clusters of three or more MetS components were significantly associated with the size of the CAR, the decline over the waking day, or the AUC. Associations of individual components with cortisol parameters were examined before and after adjustment for each other (top two rows of Tables 3–5). Analyses for each cluster were based on regression models that included each component in the cluster (indicated by “X” in Tables 3–5) plus an indicator for the presence of the cluster as described in the Subjects and Methods. All models were also adjusted for race/ethnicity, age, gender, SES, physical activity, smoking status, wake time, study day, and study site. None of the individual components (before or after adjustment for each other), pairs of components, or clusters of three or more components were significantly associated with

TABLE 3. Percentage differences in CAR associated with MetS components % Differencea,b Two componentsc

BP ⫺1.34 ⫺0.44 X X X X

HDL ⫺3.33 ⫺3.45

Triglycerides 0.03 5.96

Glucose ⫺0.22 0.03 X

X

X X X

X

X

0.21

X X X

X X X

X

X X X X X X Four components

X X X X

X X X X X

X X X X X X X X

X X X X X X

X X X X

X

X

% Difference

⫺10.73 3.41 0.57 2.25 ⫺5.02 13.58 26.95 ⫺8.33 ⫺16.70 3.60 28.25 55.27 ⫺6.27 ⫺ 17.48 8.84 ⫺19.73 11.73 4.92 5.34 ⫺24.19 15.81 ⫺6.22 ⫺13.75 36.48

X

X

Three components

WC ⫺8.16⫹ ⫺8.43⫹

X X X

X X X X X X X X X X

X X X X X X

Five components X BP, Blood pressure. a

Top row indicates percentage difference in the size of the CAR associated with the presence of each individual component adjusted for age, race/ethnicity, gender, SES, physical activity, smoking, wake time, study day, and study site. Each component is included in a separate model.

b Second row indicates percentage difference in the size of the CAR associated with the presence of each individual component in model adjusted for all other components shown plus covariates listed above.

Remaining rows show results for specific pairs or clusters of components. The components in the pair or cluster are indicated by ⬙X⬙ in the corresponding columns. The rightmost column indicates the percentage differences in the size of the CAR associated with the presence of the pair or the cluster. For example, in the first row under ⬙Two components,⬙ the CAR was 10.7% lower among persons with both high blood pressure and high glucose compared to those who did not have high levels of both components. In the first row under ⬙Four components,⬙ persons who simultaneously had high BP, low HDL, high glucose, and high WC had a 24% lower CAR than those who did not have all four components. Each model is adjusted for main effects of all five MetS components and covariates listed above. c

⫹

P ⬍ 0.10.



3489

TABLE 4. Percentage differences in cortisol decline associated with MetS components % Differencea,b Two componentsc

BP 0.42 0.41 X X X X

HDL 0.23 0.34

Triglyceride ⫺0.47 ⫺0.67

Glucose ⫺0.06 ⫺0.08 X

X X X X X

X X X

X

Three components X X X X X X Four components

Five components

X X X X X

WC 0.31 0.32

X X X X X

X X X X X

X X X X X X X X X X

X X X X X

X

X X X X X X X X

X X X X X X

X X X X X X X X X X

% Difference

⫺0.35 0.35 1.30 ⫺0.70 1.38 4.04* 0.52 ⫺0.73 0.19 ⫺0.71 ⫺8.93d* 0.47 ⫺0.88 ⫺ ⫺2.81 ⫺0.70 1.31 0.28 ⫺0.19 1.72 3.99* 2.57 ⫺0.10 ⫺0.05 2.78 2.55⫹


Top row indicates percentage difference associated with the presence of each individual component adjusted for age, race/ethnicity, gender, SES, physical activity, smoking, wake time, study day, and study site. Each component is included in a separate model.

b

Second row indicates percentage difference associated with the presence of each individual component in model adjusted for all other components shown plus covariates listed above. c Remaining rows show results for specific pairs or clusters of components. The components in the pair or cluster are indicated by ⬙X⬙ in the corresponding columns. The rightmost column indicates the percentage differences in the hourly rate of cortisol decline associated with the presence of the pair or the cluster. For example, in the first row under ⬙Two components,⬙ the hourly cortisol decline was 0.35% lower among persons with both high blood pressure and high glucose compared to those who did not have high levels of both components. In the first row under ⬙Four components⬙ persons who simultaneously had high BP, low HDL, high glucose, and high WC had a 4% flatter decline than those who did not have all four components. Each model is adjusted for main effects of all five MetS components and covariates listed above. d ⫹

Cluster contained only one participant. P ⬍ 0.10.

* P ⬍ 0.05.

the CAR (Table 3). None of the individual MetS components (before or after adjustment for each other) were significantly associated with the rate of cortisol decline across the day (Table 4). The pairing of high glucose with low HDL levels was associated with flatter cortisol decline (4%), but no other pair of MetS components was significantly associated with cortisol. Among three-component clusters, no cluster with at least 10 participants was associated with cortisol decline (high levels of glucose and triglycerides, in conjunction with low HDL levels, predicted steeper decline, but this “group” included only one individual and must be interpreted with extreme caution). The four-component cluster that excluded

high triglycerides was associated with flatter declines (4%). Having all five MetS components predicted marginally flatter decline (P ⬍ 0.10). High WC and hypertension were associated with lower AUC before, but not after, adjustment for each other. No other pairs or clusters of MetS components were associated with total cortisol output (Table 5). We also examined whether MetS diagnosis was associated with values for individual cortisol samples. It was associated with 16% lower cortisol levels upon awakening and 13.3% lower cortisol levels 30 min after awakening, but it was not significantly associated with cortisol at other times of day.

3490

DeSantis et al.



TABLE 5. Mean differences in AUC (nmol/liter 䡠 d) associated with MetS components Difference (nmol/liter 䡠 d)a,b Two componentsc

BP ⫺549.71* ⫺490.15 X X X X

HDL ⫺327.25 ⫺150.38

Triglyceride ⫺352.46 ⫺189.56

Glucose ⫺228.68 ⫺49.43 X

X X X X X

X X X

X Three components X X X X X X Four components

Five components

X X X X X

WC ⫺525.38* ⫺424.47

X X X X X

X X X X X X X X

X X X X X X X

X X X X X

X X X

X X X X X X X X X X X X

X X X X X X X X X X

Mean difference (nmol/liter 䡠 d)

718.68 ⫺4.72 ⫺486.85 ⫺749.05 216.56 1628.77 ⫺218.32 ⫺278.39 67.96 ⫺1015.15 ⫺718.79 ⫺872.26 ⫺272.61 ⫺ ⫺2210.65 3342.60 ⫺2235.47 907.07 ⫺231.95 1651.77 2378.97 ⫺164.53 90.41 ⫺761.32 ⫺109.56 434.41


Top row indicates difference associated with the presence of each individual component adjusted for age, race/ethnicity, gender, SES, physical activity, smoking, wake time study day, and study site. Each component is included in a separate model.

b Second row indicates difference associated with the presence of each individual component in model adjusted for all other components shown plus covariates listed above. c Remaining rows show results for specific pairs or clusters of components. The components in the pair or cluster are indicated by ⬙X⬙ in the corresponding columns. The rightmost column indicates the mean differences in the average daily AUC (in nmol/liter 䡠 d) associated with the presence of the pair or the cluster. For example, in the first row under ⬙Two components,⬙ the average AUC was 718.68 nmol/liter 䡠 d higher among persons with both high blood pressure and high glucose compared to those who did not have high levels of both components. In the first row under ⬙Four components,⬙ persons who simultaneously had high BP, low HDL, high glucose, and high WC had an average total AUC that was 2378.97 nmol/liter 䡠 d higher than those who did not have all four components. Each model is adjusted for main effects of all five MetS components and covariates listed above.

* P ⬍ 0.05.

Discussion Overall, we found little evidence that the presence of MetS or its components is related to cortisol output or patterns. Contrary to expectation, the presence of MetS (as well as higher levels of MetS components) appeared to be associated with lower rather than higher AUC. MetS, MetS components, or subsets of components were not consistently related to other features of the cortisol curve. The dual combination of hypertension and high glucose, as well as the cluster of four components that excludes triglycerides, predicted flatter decline, but no other associations of clusters of MetS components with any of the outcomes investigated were observed. Prior studies have reported that individual MetS components and HPA axis activity are related, although many

studies have examined individual MetS components without controlling for the others (3, 4, 10 –16). Specifically, higher WC has been associated with higher fasting cortisol levels and enhanced stress reactivity, as well as decreased cortisol variability across the day (10, 15, 17). Some evidence suggests that hypertension is associated with fasting cortisol levels and/or cortisol reactivity to stressors (15), and dyslipidemia has been associated with higher cortisol levels at particular times of day (10, 15). We did not find consistent associations of any of these factors with cortisol patterns among persons without diabetes. Body mass index (BMI) and obesity have been investigated frequently in relation to various cortisol parameters (30 –34), but they were not analyzed here because BMI is not among the MetS components as defined by


NCEP. Studies of associations of BMI and central adiposity with cortisol have been equivocal, with some reporting that persons of higher BMI and adiposity have higher AUC, but others reporting that persons of higher BMI and WC have lower awakening cortisol levels and/or decreased daily cortisol output (15, 33). The latter findings are in accordance with our results showing that MetS is associated with lower cortisol AUC and awakening cortisol levels. Research on diabetes and HPA axis activity has also reported higher total cortisol output among diabetics (18 –21), particularly those with chronic diabetes complications (21). However, these prior studies primarily included smaller and often clinical samples, with atypical HPA axis activity (Cushing’s syndrome) and/or persons with specific metabolic risk factors such as obesity (30 –34). It is plausible that associations of cortisol with clinical diabetes differ from associations of cortisol with MetS in persons without diabetes. Ongoing MESA analyses are examining associations of cortisol with clinical diabetes. The specific cortisol parameters we examined differ from those of most prior studies. Many prior studies have examined cortisol reactivity, rather than basal levels, and/or examined cortisol at specific times of day, rather than patterns or profiles of cortisol patterns over the day (13, 14, 16, 20, 31–34). Moreover, many studies have analyzed plasma or urinary, rather than salivary, cortisol levels (11, 33–35). Some have also analyzed these associations specifically among persons with insulin resistance, in overweight or obese persons, and/or persons born with low birth weight (14, 31–35). MESA study participants were selected for their absence of clinical cardiovascular disease at baseline, making them a relatively healthy sample given their age (and racial/socioeconomic background). Furthermore, participants with clinical diabetes were intentionally excluded from these analyses to examine associations between cortisol and MetS among “healthy” individuals. This undoubtedly reduced variability in the distributions of metabolic indicators and reduced the number of participants with adverse metabolic conditions. Additionally, although we had more data than many studies, 3 d may be insufficient to capture stable cortisol parameters. Nonetheless, the large population-based sample and the use of detailed measures of MetS components and multiple cortisol measures over 3 d contribute to the strength of the current study. In summary, we found no evidence that the presence of the MetS is consistently or strongly associated with cortisol parameters among healthy adults. These results suggest that cortisol profiles (at least to the extent that they could be captured with our data) are not strongly related


3491

to the development of the metabolic alterations that precede the development of diabetes. Contrary to expectation, the presence of MetS was associated with lower AUC. There was some suggestive evidence that selected sets of components were associated with flatter declines over the day, but these associations showed no clear pattern by clusters of components and need to be replicated in other samples. Additional longitudinal research is necessary to assess whether alterations of cortisol precede the development of clinical diabetes or whether diabetes itself affects HPA axis functioning.

Acknowledgments The authors thank the other investigators, the staff, and the participants of the Multi-Ethnic Study of Atherosclerosis (MESA) for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org. Address all correspondence and requests for reprints to: A. S. DeSantis, 1415 Washington Heights, SPH Tower 1, University of Michigan, Ann Arbor, Michigan 48109. E-mail: [email protected]. MESA was supported by contracts N01-HC-95159 through N01-HC-95169 from the National Heart, Lung, and Blood Institute (NHLBI). NHLBI had no further role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication. The MESA Stress Study was supported by additional National Institutes of Health Grants 1R01HL101161 and R21 DA024273. This work was also partly supported by the National Center for Minority Health and Health Disparities of the National Institutes of Health Grant P60 MD002249. A.S.D. drafted the paper and conducted all analyses; A.D.R. supervised the work and assisted with the writing of the paper; A.H. assisted with analyses and interpretation of results; S.H.G., N.S.J., B.N.S., S.S., and T.E.S. provided critical feedback on successive drafts. Disclosure Summary: The authors report no potential conflicts of interest and have nothing to disclose.

References 1. Brunner EJ, Marmot MG, Nanchahal K, Shipley MJ, Stansfeld SA, Juneja M, Alberti KG 1997 Social inequality in coronary risk: central obesity and the metabolic syndrome. Evidence from the Whitehall II study. Diabetologia 40:1341–1349 2. Chandola T, Brunner E, Marmot M 2006 Chronic stress at work and the metabolic syndrome: prospective study. BMJ 332:521–525 3. Black PH 2003 The inflammatory response is an integral part of the stress response: implications for atherosclerosis, insulin resistance, type II diabetes and metabolic syndrome X. Brain Behav Immun 17:350 –364 4. Brunner EJ, Hemingway H, Walker BR, Page M, Clarke P, Juneja M, Shipley MJ, Kumari M, Andrew R, Seckl JR, Papadopoulos A, Checkley S, Rumley A, Lowe GD, Stansfeld SA, Marmot MG 2002 Adrenocortical, autonomic, and inflammatory causes of the meta-

3492

5.

6.

7.

8.

9. 10. 11.

12.

13.

14.

15.

16.

17.

18.

19. 20.

21.

DeSantis et al.


bolic syndrome: nested case-control study. Circulation 106:2659 – 2665 Sapolsky RM, Krey LC, McEwen BS 1986 The neuroendocrinology of stress and aging: the glucocorticoid cascade hypothesis. Endocr Rev 7:284 –301 Dickerson SS, Kemeny ME 2004 Acute stressors and cortisol responses: a theoretical integration and synthesis of laboratory research. Psychol Bull 130:355–391 Adam EK, Hawkley LC, Kudielka BM, Cacioppo JT 2006 Day-today dynamics of experience-cortisol associations in a populationbased sample of older adults. Proc Natl Acad Sci USA 103:17058 – 17063 Pruessner JC, Wolf OT, Hellhammer DH, Buske-Kirschbaum A, von Auer K, Jobst S, Kaspers F, Kirschbaum C 1997 Free cortisol levels after awakening: a reliable biological marker for the assessment of adrenocortical activity. Life Sci 61:2539 –2549 Kirschbaum C, Hellhammer DH 1989 Salivary cortisol in psychobiological research: an overview. Neuropsychobiology 22:150 –169 Reynolds RM, Walker BR 2003 Human insulin resistance: the role of glucocorticoids. Diabetes Obes Metab 5:5–12 Fraser R, Ingram MC, Anderson NH, Morrison C, Davies E, Connell JM 1999 Cortisol effects on body mass, blood pressure, and cholesterol in the general population. Hypertension 33:1364 –1368 Ward AM, Fall CH, Stein CE, Kumaran K, Veena SR, Wood PJ, Syddall HE, Phillips DI 2003 Cortisol and the metabolic syndrome in South Asians. Clin Endocrinol (Oxf) 58:500 –505 Anagnostis P, Athyros VG, Tziomalos K, Karagiannis A, Mikhailidis DP 2009 Clinical review: the pathogenetic role of cortisol in the metabolic syndrome: a hypothesis. J Clin Endocrinol Metab 94: 2692–2701 Rosmond R, Dallman MF, Bjo¨rntorp P 1998 Stress-related cortisol secretion in men: relationships with abdominal obesity and endocrine, metabolic and hemodynamic abnormalities. J Clin Endocrinol Metab 83:1853–1859 Duclos M, Marquez Pereira P, Barat P, Gatta B, Roger P 2005 Increased cortisol bioavailability, abdominal obesity, and the metabolic syndrome in obese women. Obes Res 13:1157–1166 Rosmond R, Bjo¨rntorp P 2000 The hypothalamic-pituitary-adrenal axis activity as a predictor of cardiovascular disease, type 2 diabetes, and stroke. J Internal Med 247:188 –197 Holt-Lunstad J, Steffen PR 2007 Diurnal cortisol variation is associated with nocturnal blood pressure dipping. Psychosom Med 69: 339 –343 Cameron OG, Thomas B, Tiongco D, Hariharan M, Greden JF 1987 Hypercortisolism in diabetes mellitus. Diabetes Care 10:662– 664 Welles B 2006 Glucocorticoids in type 2 diabetes mellitus and the metabolic syndrome. Drug Dev Res 67:570 –573 Chiodini I, Torlontano M, Scillitani A, Arosio M, Bacci S, Di Lembo S, Epaminonda P, Augello G, Enrini R, Ambrosi B, Adda G, Trischitta V 2005 Association of subclinical hypercortisolism with type 2 diabetes mellitus: a case-control study in hospitalized patients. Eur J Endocrinol 153:837– 844 Oltmanns KM, Dodt B, Schultes B, Raspe HH, Schweiger U, Born J, Fehm HL, Peters A 2006 Cortisol correlates with metabolic dis-


22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33.

34.

35. 36.

turbances in a population of type 2 diabetics. Eur J Endocrinol 154: 325–331 Bild DE, Bluemke DA, Burke GL, Detrano R, Diez Roux AV, Folsom AR, Greenland P, Jacob Jr DR, Kronmal R, Liu K, Nelson JC, O’Leary D, Saad MF, Shea S, Szklo M, Tracy RP 2002 Multi-ethnic study of atherosclerosis: objectives and design. Am J Epidemiol 156: 871– 881 Hajat A, Diez-Roux A, Franklin TG, Seeman T, Shrager S, Ranjit N, Castro C, Watson K, Sanchez B, Kirschbaum C 2010 Socioeconomic and race/ethnic differences in daily salivary cortisol profiles: the Multiethnic Study of Atherosclerosis. Psychoneuroendocrinology 35:932–943 Kudielka BM, Broderick JE, Kirschbaum C 2003 Compliance with saliva sampling protocols: electronic monitoring reveals invalid cortisol daytime profiles in noncompliant subjects. Psychosom Med 65:313–319 2001 Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA 285:2486 –97 Kudielka BM, Kirschbaum C 2003 Awakening cortisol responses are influenced by health status and awakening time but not by menstrual cycle phase. Psychoneuroendocrinology 28:35– 47 Clow A, Thorn L, Evans P, Hucklebridge F 2004 The awakening cortisol response: methodological issues and significance. Stress 7:29 –37 Golden SH, Folsom AR, Coresh J, Sharrett AR, Szklo M, Brancati F 2002 Risk factor groupings related to insulin resistance and their synergistic effects on subclinical atherosclerosis: the Atherosclerosis Risk in Communities Study. Diabetes 51:3069 –3076 Hochberg Y 1988 A sharper Bonferroni procedure for multiple tests of significance. Biometrika 75:800 – 802 Ranjit N, Diez-Roux AV, Sanchez B, Seeman T, Shea S, Shrager S, Watson K 2009 Association of salivary cortisol circadian pattern with cynical hostility: multi-ethnic study of atherosclerosis. Psychosom Med 71:748 –755 Sen Y, Aygun D, Yilmaz E, Ayar A 2008 Children and adolescents with obesity and the metabolic syndrome have high circulating cortisol levels. Neuro Endocrinol Lett 29:141–145 Weigensberg MJ, Toledo-Corral CM, Goran MI 2008 Association between the metabolic syndrome and serum cortisol in overweight Latino youth. J Clin Endocrinol Metab 93:1372–1378 Phillips DI, Barker DJ, Fall CH, Seckl JR, Whorwood CB, Wood PJ, Walker BR 1998 Elevated plasma cortisol concentrations: a link between low birth weight and the insulin resistance syndrome? J Clin Endocrinol Metab 83:757–760 Misra M, Bredella MA, Tsai P, Mendes N, Miller KK, Klibanski A 2008 Lower growth hormone and higher cortisol are associated with greater visceral adiposity, intramyocellular lipids, and insulin resistance in overweight girls. Am J Physiol Endocrinol Metab 295: E385–E392 Bjo¨rntorp P, Rosmond R 2000 Obesity and cortisol. Nutrition 16: 924 –936 Cleveland WS, Devlin SJ, Grosse E 1988 Regression by local fitting: methods, properties, and computational algorithms. J Econometrics 37:87–114