Diurnal Cortisol Patterns, Future Diabetes, and Impaired Glucose Metabolism in the Whitehall II Cohort Study

Context:

The hypothalamic pituitary-adrenal axis is thought to play a role in type 2 diabetes (T2D). However, evidence for an association between cortisol and future glucose disturbance is sparse.

Objective:

The aim was to examine the association of diurnal cortisol secretion with future T2D and impaired glucose metabolism in a community-dwelling population.

Design:

This is a prospective cohort study of salivary cortisol measured at the 2002–2004 clinical examination of the Whitehall II study, United Kingdom. We measured cortisol (nmol/l) from six saliva samples obtained over the course of a day: at waking, +30 minutes, +2.5 hours, +8 hours, +12 hours, and bedtime. Participants who were normoglycemic in 2002–2004 (phase 7) were reexamined in 2012–2013 (phase 11).

Setting:

The occupational cohort was originally recruited in 1985–1988.

Participants:

A total of 3270 men and women with an average age of 60.85 years at phase 7 (2002–2004).

Outcome Measures:

Incident T2D and impaired fasting glucose in 2012–2013 were measured.

Results:

Raised evening cortisol at phase 7 was predictive of new-onset T2D at phase 11 (odds ratio [OR], 1.18; 95% confidence interval [CI], 1.01–1.37) with a trend for a flatter slope in participants with incident T2D (odds ratio, 1.15; 95% CI, 0.99–1.33). When expanding this analysis to a broader category of glucose disturbance we found that a flattened diurnal cortisol slope at phase 7 was predictive of future impaired fasting glucose or T2D at phase 11 (OR, 1.12; 95% CI, 1.02–1.22), as was high bedtime cortisol (OR, 1.10; 95% CI, 1.01–1.20).

Conclusions:

In this nonclinical population, alterations in diurnal cortisol patterns were predictive of future glucose disturbance.

Type 2 diabetes (T2D) is a chronic metabolic disorder characterized by hyperglycemia resulting from insulin resistance and β-cell dysfunction (1). The hyperglycemia of T2D develops gradually (2), and evidence suggests that the health risk accompanying raised glucose is continuous (3). Therefore, intermediate states of hyperglycemia that are higher than normal but do not met the diagnostic criteria for T2D have been defined (1). These “prediabetes” states are significant because individuals with glucose concentrations in this range have an elevated risk of developing T2D and diabetes complications (4, 5).

Cortisol (a product of the hypothalamic pituitary adrenal [HPA] axis) plays a role in many processes relevant to T2D. Pathological (6) and experimental (7) exposure to excessive cortisol is related to metabolic disturbances such as hypertension, hyperlipidemia, and central obesity, which predispose individuals to prediabetes and overt T2D. In cross-sectional studies with healthy individuals, raised cortisol concentrations assessed from plasma samples (8) and 24-hour urinary free samples (9) have been associated with raised plasma glucose (8) and insulin resistance (8, 9). Prolonged hypercortisolism as seen in Cushing's syndrome (10) and in glucocorticoid-treated patients (11) increases susceptibility for hyperglycemia and manifest T2D.

Cortisol has a distinctive diurnal pattern. It is typically characterized by high cortisol concentrations on waking, followed by a rise that peaks 30 minutes after waking (termed the cortisol awakening response [CAR]) and a subsequent decline over the day (12). Several studies have investigated the cross-sectional association between daily cortisol secretion and diabetes. However, the findings are mixed. In the largest study to date of 3508 individuals, we found that participants with T2D had a flatter slope in cortisol across the day (13). This corroborates the findings of Lederbogen et al (14), who observed an association between flatter daily cortisol profiles and T2D in a community cohort. In both studies, individuals with T2D had significantly higher evening cortisol levels compared with nondiabetic controls (13, 14). In contrast, Champaneri et al (15) and Bruehl et al (16) found a blunted CAR in individuals with T2D relative to controls, but no association for cortisol slope, whereas Vreeburg et al (17) found no association between any component of the diurnal cortisol curve and T2D.

Longitudinal evidence relating neuroendocrine dysfunction with impaired fasting glucose (IFG, a form of prediabetes) or T2D is sparse. In the Longitudinal Aging Study of Amsterdam (LASA) morning and evening salivary cortisol were measured in 998 initially healthy people. Raised evening cortisol was associated with future T2D in female participants, but no associations were found for men (18).

To date, no study has examined the relationship between the diurnal cortisol profile and future glucose status in an initially healthy population. We therefore sought to examine these associations in the Whitehall II cohort. In keeping with our cross-sectional findings, we hypothesized that a flatter diurnal cortisol slope and raised evening cortisol levels at phase 7 (2002–2004) would predict new onset IFG and T2D at phase 11 (2012–13).

Methods and Measures

Participants

We used data from phases 7 (2002–2004) and 11 (2012–2013) clinical examinations of the Whitehall II study. This cohort of 10 308 participants was initially recruited from 20 London-based civil service departments between 1985 and 1988. Saliva collection for the assessment of cortisol started partway through phase 7 and 90.1% (n = 4608) of participants that were asked to collect saliva returned samples. Of those, 3508 participants had complete information on cortisol measures and diabetes status (for further details, please refer to (13)). Participants who had prevalent diabetes at phase 7 (n = 238) were excluded from the analysis, giving a final sample of 3270 participants. Ethical approval was obtained from the University College London Medical School committee on the ethics of human research. All participants gave full informed consent to take part in the study.

Cortisol collection and analysis

Participants were asked to take six samples in Salivettes over the course of a normal weekday at waking, after 30 minutes, 2.5 hours, 8 hours, and 12 hours, and at bedtime. They were requested not brush their teeth or consume any food or beverages for 15 minutes before sample collection. Salivettes were centrifuged at 3000 × g for 5 minutes, resulting in a clear supernatant of low viscosity. Cortisol levels were measured using a commercial chemiluminescence immunoassay (CLIA; IBL Hamburg, Germany). The lower concentration limit of the assay was 0.44 nmol/liter and the intra- and interassay coefficients of variance were less than 8%. Any sample larger than 50 nmol/liter was reanalyzed.

Type 2 diabetes and impaired fasting glucose

New-onset T2D cumulating from the end of phase 7 (2002–2004) to phase 11 (2012–2013) was defined as a fasting glucose ≥7.0 mmol/l, by reported doctor-diagnosed diabetes, or diabetes medication usage. IFG was classified as a fasting glucose between 5.6 and 6.9 mmol/liter (1). Blood glucose was measured using the glucose oxidase method (19) on a YSI model 2300 Stat Plus Analyzer (YSI Corporation, Yellow Springs, OH; mean coefficient of variation, 1.4–3.1%) (20).

Covariates

Age, sex, and current or most recent civil service employment grade, a measure of social position, were assessed by questionnaire. Smoking status was defined as current smokers vs noncurrent smokers. Height was assessed using a stadiometer with the head in the Frankfort plane; weight was assessed using a digital scale (Tanita, Yiewsley, Middlesex, UK). Body mass index (BMI) was calculated as weight (in kilograms)/height (in meters) squared. History of coronary heart disease (CHD) was assessed in phase 7. Participants provided information on medication use; this were subsequently coded using the British National Formulary (21). Cardiovascular medication usage was defined as the use of β-blockers, antihypertensives, lipid-lowering drugs, nitrates, or antiplatelet medications.

Statistical analysis

We removed participants with cortisol values outside 3 SD from the mean and those taking steroid medications from the analyses (n = 171). Despite this, cortisol data were skewed and were logged for analysis. The CAR was calculated by subtracting cortisol at time 1 (waking) from cortisol at time 2 (30 minutes after waking). Analyses are conventionally limited to samples collected within 10 minutes of waking (sample 1 taken ≥10 minutes, n = 579) because of reduced CAR in those with longer delays (22). We did not find a difference in sample delays by new-onset diabetes or IFG, so all participants were retained. Participants were asked to refrain from eating 15 minutes before sample collection; there was high adherence to this protocol. We checked whether eating between samples 1 and 2 affected the pattern of results. We did not find a relationship between eating behavior on the morning of sampling and new-onset T2D or IFG, so all samples were retained for analysis. The slope of the decline in cortisol levels over the day was calculated by regressing cortisol values on time after waking for samples 1 (waking), (2.5 hours), 4 (8 hours), 5 (12 hours), and 6 (bedtime) (13, 23). It is thought that the CAR and slope are under different neurobiological control systems (12). Therefore, sample 2 was not included to ensure the CAR did not obscure the slope calculation. More negative slopes indicate a more rapid decline in cortisol levels, whereas slope values closer to zero reflect flatter diurnal rhythms. Descriptive characteristics of the sample were compared using univariate ANOVA for continuous variables and χ² tests for categorical variables. Z scores (mean = 0; SD = 1) were created for waking cortisol, the CAR, slope, and evening cortisol. The associations between the cortisol measures and new-onset diabetes or IFG at phase were assessed using logistic regression. Age, sex, grade of employment, smoking, BMI, IFG at phase 7, cardiovascular medication, and history of CHD were included as covariates in all analyses. Participants with missing covariate information were excluded from the analyses. We have previously shown a nonlinear relationship between BMI and cortisol slope (24); therefore, BMI was categorized using the cut-point of 23, where the relationship between slope and BMI changes. Including BMI as a continuous variable did not change the pattern of results. Sex differences in the relationship between cortisol and diabetes have been reported previously (15, 18). We investigated whether sex interacted with the cortisol measures, but found no significant associations, so interaction terms were not included in the final models and men and women were analyzed in combination. Results are presented as adjusted odds ratios (OR) with 95% confidence intervals (CI). The slope estimates were generated using 1 MLWin, version 2.10 beta 6; all other analyses were conducted using SPSS, version 21 (SPSS, Chicago, IL).

Results

We restricted our study group to those with complete information on cortisol measures and removed participants with prevalent diabetes at phase 7 (n = 238), giving a final sample of 3270. We excluded people who agreed to provide cortisol samples, but did not send any samples or did not send a complete set of samples (n = 1637). Participants with prevalent diabetes were removed from the excluded group for comparative purposes (n = 179). The characteristics of participants included and excluded from the analyses are displayed in Table 1. The group included in the analyses was more likely to be male and had fewer participants in the lowest civil service employment grades. They were less likely to smoke, take cardiovascular medication, and have a history of CHD than the excluded group.

For the purposes of the study, the participants were divided into three groups based on their glucose status at phase 11. The characteristics of the groups are displayed in Table 2. A total of 210 (6.4%) participants had new-onset diabetes and 518 (15.8%) individuals had IFG at phase 11. The individuals with diabetes were older, more likely to be male, and more likely to be the lowest civil service employment grades compared with normoglycemic individuals. They had higher BMI and were more likely to have a history of CHD and take cardiovascular medication than normoglycemic or IFG participants. Late saliva collection and unadjusted cortisol values did not significantly differ by glucose status.

Table 3 shows the ORs for incident diabetes and combined diabetes and IFG per 1 SD increase in scores of the cortisol measures. We found no association between the CAR and incident diabetes (P = .318). For slope across the day, we observed a trend for a flatter slope in participants with new-onset diabetes (P = .075). A flattened cortisol slope can be due to low waking or high evening cortisol values. No association was observed between waking cortisol and T2D (P = .745). In contrast, evening cortisol was predictive of incident diabetes (P = .035). This association was robust to adjustment for all covariates. Participants with new-onset diabetes had higher evening cortisol values ( = 2.43, SD = 2.07) than participants without diabetes ( = 2.34, SD = 2.99) controlling for covariates.

To further explore the relationships between the cortisol measures and future glucose status and in particular the trend for a flattened slope in those with T2D, we analyzed the prospective association between cortisol measures and glucose status in the combined group of participants with new-onset diabetes or IFG at phase 11. Again, no associations were detected between the CAR and incident diabetes or IFG (P = .35). However, slope was significantly predictive of new-onset diabetes or IFG at phase 11, controlling for covariates (P = .015). Participants with incident diabetes or IFG had a flatter slope in cortisol across the day at baseline ( = −0.128, SD = 0.023) compared with normoglycemic controls ( = −0.130, SD = 0.022). No significant association was found between waking cortisol and incident diabetes or IFG (P = .064), but evening cortisol was associated with future diabetes or IFG (P = .044) adjusting for covariates. Participants with diabetes or IFG had higher evening cortisol values ( = 2.40, SD = 2.67) than normoglycemic individuals ( = 2.33, SD = 3.02), controlling for covariates. It is possible that raised cortisol levels later in the day contributed to the difference in slope between the two groups.

Discussion

This study investigated the longitudinal association between components of the diurnal cortisol profile and future T2D and impaired fasting glucose a sample of community-dwelling adults. We found that raised evening cortisol levels were predictive of new-onset T2D approximately 9 years later. For slope across the day, we observed a trend for a flatter slope in participants with incident diabetes. We found that a flattened diurnal cortisol slope at phase 7 was predictive of the combined outcome of future IFG or T2D at phase 11. Evening cortisol was also associated with future diabetes or IFG. No associations emerged for the CAR or waking cortisol.

To our knowledge, this is the first study to investigate the prospective associations between the diurnal cortisol profile and future T2D as well as impaired fasting glucose. However, the association between morning and evening cortisol and future T2D has been investigated previously in the LASA cohort (18). Our finding that evening cortisol at phase 7 is predictive of new-onset diabetes at phase 11 corroborates and extends the results of that study. In the LASA cohort, raised evening cortisol concentrations were associated with future T2D, but only in female participants. We checked for an interaction between gender and cortisol measures, but found no significant associations. Our study is considerably larger than the LASA cohort study; this may account for our ability to detect an association between evening cortisol and incident diabetes in both genders. Raised evening cortisol has been associated with T2D in cross-sectional studies (13, 14). The present study adds to these findings by showing that evening cortisol is not only elevated in people with prevalent diabetes, but also is predictive of new-onset diabetes in initially healthy individuals (13). No associations between waking cortisol and incident T2D were detected in the study. This finding is in keeping with the LASA study (18) and our previous cross-sectional work (13).

No associations emerged for the CAR in the present study. The lack of a prospective association with the CAR corroborates the results from our cross-sectional analysis (13). We observed a borderline significant result for a flatter cortisol slope in participants with new-onset T2D. Expanding this analysis to a broader category of glucose disturbance, we found that participants with IFG or T2D had a flatter cortisol slope compared with normoglycemic controls. The analysis of the T2D group alone may have been underpowered to detect a significant effect. This result builds upon our cross-sectional work (13) and suggests that a flatter slope in cortisol might also be predictive of future disturbances in glucose metabolism in an initially healthy population.

The reasons for the alterations in cortisol secretion observed in this study are unknown and the mechanism by which T2D and IFG are related to the HPA axis remains unclear. Cortisol plays an important role in many processes relevant to IFG and T2D. One major function of cortisol is to raise glucose through gluconeogenesis. By acting via glucocorticoid receptors, which are expressed on β cells, cortisol directly reduces insulin sensitivity and decreases insulin secretion. Cortisol can induce lipolysis and the release of fatty free acids into the bloodstream and the buildup of triglycerides in adipose tissue (25). Both of these factors are associated with an increased risk of diabetes, although it is unclear whether these associations are causal (26, 27). Changes in diurnal cortisol secretion have been associated with disorders that are possible complications of T2D and impaired glucose metabolism. Previous work with this cohort has shown that flatter diurnal cortisol rhythms and raised evening cortisol levels are predictive of cardiovascular mortality (23).

Our cross-sectional work has associated obesity with a flatter cortisol slope and raised evening cortisol concentrations (24). Obesity is risk factor for T2D (28), and weight loss is the treatment of choice for prediabetes (5). Visceral adipose tissue expresses high concentrations of glucocorticoid receptors (29), and it is thought that adipocytes are a source of cortisol. The enzyme 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1) regulates glucocorticoid metabolism at the tissue level by catalyzing the conversion of the inactive form of glucocorticoids to active cortisol. Transgenic mice overexpressing 11β-HSD1 have heightened adipose tissue levels of corticosterone (30). In humans, increased 11β-HSD1 activity has been associated with features of the metabolic syndrome (31), and inhibitors of 11β-HSD1 are being trialed as a potential treatment for T2D (32). BMI was associated with cortisol slope and evening cortisol and was predictive of new onset T2D as well as the broader category of glucose disturbance in our analysis. Additionally, participants with new-onset IFG and T2D had higher BMIs than the normoglycemic individuals in our sample. Despite this, the relationship between future glucose status and these cortisol measures was robust to adjustment for BMI. This suggests that in this cohort obesity is not the mechanism by which diurnal cortisol secretions increases the risk of T2D and IFG.

Abnormal daily cortisol rhythms may disrupt immune and inflammatory processes. Inflammatory cytokines play a role in T2D, and heightened cytokine concentrations are predictive of T2D development (33). Cortisol plays a role in regulating inflammation, and it has been suggested that glucocorticoids might dysregulate immunity via circadian-immune communication (34). In addition to changes in immune and inflammatory processes, it is possible that disturbances in circadian rhythms may act on IFG and T2D through the alteration of glucose metabolism. Experimental work suggests that circadian disruption increases both fasting and postprandial plasma glucose levels through insufficient insulin secretion (35). Future research should assess whether changes in immune and inflammatory processes and alterations in glucose metabolism might underlie the associations observed in this study.

A further potential pathway that might link cortisol slope and evening cortisol with future glucose disturbance is stress. An increasing body of literature suggests that psychosocial stress factors increase the risk of T2D (36). Stress stimulates the HPA axis, inducing cortisol release (37) and acute and chronic stressors have been linked with a flatter cortisol slope across the day and raised evening cortisol concentrations (38). Acute stress on the day of saliva sampling was not associated with slope or evening cortisol in this analysis (data not shown), and our results were robust to adjustment for this measure. However, it is plausible that our results could be due to long-term changes in circadian regulation as a result of chronic stress that predisposed participants to future IFG and T2D.

Our results should be interpreted in light of some limitations. The Whitehall II study is an occupational cohort of civil servants and therefore is not representative of the general population. The cohort participants are predominately of white ethnicity, and because of ethnic differences in the pathogenesis of T2D, the current analysis was restricted to white individuals. Therefore, these findings are not necessarily generalizable to other populations. Glucose was only measured at the Whitehall clinical waves; therefore, it was not possible to assess when exactly incident IFG and T2D occurred in the follow-up period. The effects observed in our analyses were small. However, these patterns are thought to be representative of chronic differences that are present on an everyday basis. Our findings suggest that even modest differences in daily cortisol secretion may have negative effects on glucose status over time. We only assessed cortisol over a single day, and it is suggested that this may obscure the CAR to situational rather than chronic correlates (39). This may have contributed to our inability to detect an association between the CAR and future glucose status in the present analysis. This is an observational study, so we are not able to determine causal relationships, and other factors such as food intake, chronic stress, or other psychological factors may have contributed to or independently driven the association between cortisol and later glucose status. We do not have information on the amount or type of food participants ate on the day of cortisol collection. The night release of cortisol was not assessed and therefore we could not evaluate total 24-hour circadian cortisol exposure. We relied on self-report for the timing of sample collection because evidence suggests that people are usually accurate in reporting this information (40). The prevalence of “late” collection was comparable to previously reported rates (22).

Despite these considerations, our findings indicate that diurnal cortisol secretion is associated with future T2D and impaired glucose metabolism in a nonclinical population. It is plausible that neuroendocrine dysfunction is related to the pathophysiology of T2D, but the precise mechanisms through which changes in cortisol secretion impairs glucose metabolism remain to be determined.

Acknowledgments

This work was supported by the British Heart Foundation (to R.A.H. and A.S.); the Medical Research Council, the US National Institutes of Health (R01HL036310; R01AG034454), and a professorial fellowship from the Economic and Social Research Council (to M.Ki); and the Economic and Social Research Council International Centre for Life Course Studies in Society and Health (RES-596-28-0001) (to M.Ku). The Whitehall II study is supported by the British Heart Foundation, Medical Research Council, National Heart, Lung and Blood Institute (HL36310), and National Institute on Aging (AG13196). The funding sources had no role in the design, conduct, or reporting of this study.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• 11β-HSD1

  11β-hydroxysteroid dehydrogenase type 1

 
• BMI

  body mass index

 
• CAR

  cortisol awakening response

 
• CHD

  coronary heart disease

 
• CI

  confidence interval

 
• HPA

  hypothalamic pituitary adrenal

 
• IFG

  impaired fasting glucose

 
• LASA

  Longitudinal Aging Study of Amsterdam

 
• OR

  odds ratio

 
• T2D

  type 2 diabetes

References