Psychological Distress and Incidence of Type 2 Diabetes in High-Risk and Low-Risk Populations: The Whitehall II Cohort Study

Recruitment to the Whitehall II Study took place between 1985 and 1988 among all office staff, aged 35–55 years, in 20 London-based Civil Service departments (22). The response rate was 73% (6,895 men and 3,413 women). Informed consent was obtained from all participants, and the University College London Medical School Committee on the Ethics of Human Research approved the protocol.

The target population for the current study was a sample of 5,932 participants (4,189 men and 1,743 women, mean age 54.6 years) for whom data were collected in at least two of the following cycles: from 1991–1993 to 1997–1999, from 1997–1999 to 2002–2004, and from 2002–2004 to 2007–2009. Included participants had complete data on type 2 diabetes, psychological distress, FRS (21), and covariates (age, sex, socioeconomic status [SES], ethnicity, antidepressant use, smoking, and physical activity) at baseline. Those included in the analyses were free of diabetes at the baseline cycle(s) and had data on their prediabetes status and FRS (Fig. 1). Each participant could thus contribute to a minimum of one and a maximum of three cycles. The 5,932 eligible participants produced 13,207 person-observations; mean follow-up time between baseline and follow-up for each cycle was 5.46 (SD 0.51) years.

At each clinical phase, venous blood samples were taken from individuals who were fasting ≥8 h (those whose clinic visit was in the afternoon had a light fat-free breakfast and they were asked to fast for ≥5 h) before undergoing a standard 2-h 75-g oral glucose tolerance test (OGTT) (1). Diabetes was defined as fasting glucose ≥7.0 mmol/L, 2-h postload glucose ≥11.1 mmol/L (23–26), self-reported doctor-diagnosed diabetes, or use of diabetes medication. Blood samples were handled at all phases using similar standard protocols, and baseline cases were excluded from the prospective analyses. Prediabetes was defined as impaired fasting glucose (IFG; a fasting glucose of 5.6–6.9 mmol/L and 2-h glucose <7.8 mmol/L) and/or impaired glucose tolerance (a fasting glucose <7 mmol/L and a 2-h postload glucose of 7.8–11.0 mmol/L) (23).

The FRS is based on a previously published detailed algorithm (21) with the following components: IFG, overweight or obesity, low level of HDL, high level of triglycerides, elevated blood pressure or antihypertensive medication, and parental diabetes. The total score ranges from 0 to 30. In the present analysis, participants with impaired glucose tolerance received the same score as those with IFG (10 points in both cases). In the Framingham Offspring Study, participants with >19 points had an 8-year incidence of type 2 diabetes >15% (21); thus we used this score to determine high risk status. Participants with normoglycemia scored 0–18 in the FRS and were further classified into the low-risk group (0–9 points) and intermediate-risk group (10–19), the latter range corresponding to the lower (i.e., intermediate)-risk group (10–19 points) among participants with prediabetes. Based on the baseline information on prediabetes status and the FRS, participants were classified into four groups: 1) no prediabetes, low FRS (0–9); 2) no prediabetes, intermediate FRS (10–19); 3) prediabetes, intermediate FRS (10–19); and 4) prediabetes, high FRS (>19).

The 30-item General Health Questionnaire (GHQ-30) was used to assess psychological distress (27). The GHQ is a screening instrument designed to detect common psychological symptoms, such as depression and anxiety. It is widely used in population-based surveys and trials and has been validated in the Whitehall II Study (28). Each questionnaire item inquires about a specific symptom; response categories are scored as either 1 or 0 to indicate presence or absence of the symptom. A total score of 5 or more led to individuals being defined as GHQ-symptom “cases” and scores 0–4 as “noncases” (28).

Sociodemographic covariates included age, sex, SES (based on the last known occupational grade and divided into high, intermediate, and low), and ethnicity (white, South Asian, or other) and were all based on survey responses (22). Antidepressant use (yes/no) and smoking (yes/no) were based on self-reported information at the baseline survey of each cycle. Physical activity was assessed at cycle 1 based on answers to questions about the frequency and duration of participation in mildly energetic (e.g., weeding, general housework, bicycle repair), moderately energetic (e.g., dancing, cycling, leisurely swimming), and vigorous physical activity (e.g., running, hard swimming, playing squash). At cycles 2 and 3, the questionnaire included 20 items on frequency and duration of participation in different physical activities (e.g., walking, cycling, sports) that were used to compute hours per week of each intensity level. Participants were classified as active (>2.5 h/week of moderate physical activity or >1 h/week of vigorous physical activity), inactive (<1 h/week of moderate physical activity and <1 h/week of vigorous physical activity), or moderately active (if not active or inactive) (29).

After harmonization of data across cycles, we examined associations between psychological distress at baseline and incident type 2 diabetes at follow-up for each cycle. We used generalized estimation equations (GEEs) with a logistic link to control for intraindividual correlation between repeated measurements to estimate odds ratios (ORs) and their 95% CIs. GEE analysis was used because repeated measurements were nested within participants (i.e., the same individuals could contribute more than one observation to the dataset), and the GEE method takes into account nonindependence of the within-participant observations when estimating standard errors. We first analyzed the association between psychological distress and the incidence of type 2 diabetes in the total cohort. Then we grouped the participants according to their baseline prediabetes status, FRS, and psychological distress into eight groups as follows: 1) normoglycemia, FRS 0–9, no distress (reference group); 2) normoglycemia, score 0–9, distress; 3) normoglycemia, score 10–19, no distress; 4) normoglycemia, score 10–19, distress; 5) prediabetes, score 10–19, no distress; 6) prediabetes, score 10–19, distress; 7) prediabetes, score >19, no distress; and 8) prediabetes, score >19, distress. Models were adjusted for age, sex, SES, ethnicity, antidepressant use, smoking, and physical activity. SAS version 9.2 (SAS, Cary, NC) was used for all analyses.

Table 1 shows the characteristics of participants at the baseline of each study cycle and in total. The proportion of participants who were male and white, had high SES, were without psychological distress, and had incident type 2 diabetes was greater at the last cycle than at the first. Antidepressant use was more prevalent and smoking less prevalent and both high and low physical activity were more prevalent in the latter cycles. Overall 5-year incidence for type 2 diabetes was 4.2%.

Associations between baseline covariates for participants with normoglycemia and prediabetes by psychological distress are presented in Supplementary Table 1. Irrespective of the participant’s prediabetes status, psychological distress was associated with younger age, female sex, intermediate SES, nonwhite ethnicity, antidepressant use, smoking, and low physical activity.

In the total cohort, psychological distress did not predict type 2 diabetes after adjustment for age, sex, SES, and ethnicity (OR 1.16 [95% CI 0.94–1.42]; data not shown). We found no interaction between sex and psychological distress (P = 0.37) or between ethnicity and psychological distress (P = 0.91) in the prediction of type 2 diabetes.

We then examined whether combinations of prediabetes status, FRS, and psychological distress predicted the incidence of type 2 diabetes. Figure 2 shows the unadjusted incidences. Type 2 diabetes incidence was low among participants with normoglycemia and a low risk score, irrespective of the presence (1.9%) or absence (1.6%) of psychological distress. Among the normoglycemic participants with an intermediate risk score of 10–19, 7.6% of distressed and 6.0% of nondistressed people had type 2 diabetes at follow-up; the confidence intervals suggesting no association with psychological distress. Similarly, among participants with prediabetes and an FRS of 10–19, no difference was found in the incidence of type 2 diabetes between those with (15.6%) and without psychological distress (15.0%). However, among participants with prediabetes and a high FRS (>19), 40.9% of those with psychological distress developed type 2 diabetes at follow-up compared with only 28.5% of those without psychological distress.

Results of the multivariable-adjusted logistic regression models confirm those from the unadjusted analysis by showing a strong dose-response association between prediabetes and FRS status, and risk of incident type 2 diabetes (Table 2). Moreover, comparisons indicate no difference regarding the association between psychological distress and type 2 diabetes among normoglycemic participants or among those with prediabetes and a lower FRS, whereas among participants with prediabetes and a high FRS (>19), psychological distress was associated with a doubling of the risk of type 2 diabetes. A statistically significant interaction (P = 0.039) was found when comparing the prediabetes–high FRS group with the other groups combined, as regards the association between psychological distress and incident type 2 diabetes.

The findings were replicated in sensitivity analysis using an alternative, less stringent cut point of >18 to define high FRS (Supplementary Table 2).

We examined whether the association between psychological distress and incident type 2 diabetes differed between populations at different baseline risk levels of type 2 diabetes as assessed by the presence of prediabetes and the level of FRS. The FRS is based on traditional type 2 diabetes risk factors: prediabetes, overweight or obesity, low HDL level, increased level of triglycerides, hypertension, and a history of parental diabetes. In the current study, the score was a strong predictor of incident type 2 diabetes.

Our main finding was that psychological distress is associated with a doubling of diabetes risk in a high-risk populations. In the overall analysis, psychological distress was not significantly associated with type 2 diabetes. Subsequent stratified analysis revealed no association between psychological distress and type 2 diabetes in normoglycemic participants irrespective of the risk score and in participants with prediabetes and a lower FRS. However, in the group of participants with prediabetes and high risk score (>19), the 5-year incidence of diabetes was 28.5% in those without psychological distress and 40.9% in those with psychological distress at baseline. In the multivariate-adjusted model, the OR was twofold increased among those with psychological distress compared with those without. The findings were replicated using a lower cut point (>18) for defining a high FRS.

Earlier literature suggests that the relationship between psychosocial factors and type 2 diabetes may be complex (16,17). We found no overall association between psychological distress and type 2 diabetes. Earlier findings on the association between psychological factors, such as general stress and work stress, and incident type 2 diabetes have been mixed, including both null and positive findings (10–17). Inconsistencies in earlier research may be due not only to the use of different stress and distress indicators across studies but also, as our present findings suggest, to a failure to recognize that the “nondiabetes” group might be heterogeneous in terms of vulnerability to distress; those at an “advanced” stage of prediabetes may be more affected by the adverse metabolic effects of psychological distress than those at lower risk levels (1,17). Indeed, the effects of psychological stress factors have been suggested to be synergistic (17). Rather than a general risk factor in all diabetes-free populations, it might be a stage-specific risk factor that has a much stronger effect among those at an advanced stage of progression toward manifest type 2 diabetes. In line with our results, an earlier report from the Whitehall II Study showed that work stress predicted type 2 diabetes among obese, but not nonobese, women (14).

Plausible pathways through which psychological distress may accelerate progression to type 2 diabetes among high-risk individuals include health risk behaviors and direct physiological pathways, such as long-term dysregulation of the hypothalamic-pituitary-adrenal axis, leading to increased levels of glucocorticoids, especially cortisol, and changes in immune system activity, leading to increased concentrations of proinflammatory cytokines (1,8,16,17). There is some evidence supporting inflammation and lifestyle factors as mediators between depressive symptoms and incident type 2 diabetes (8).

Healthcare implications of this study include the notion that psychological distress might hamper the outcomes of intensive lifestyle and other treatment interventions targeted at high-risk groups (2,3). Psychological distress symptoms, such as stress, anxiety, depression, feelings of hopelessness, and sleep disturbances, have previously been shown to hinder commitment to major, long-term lifestyle changes and reduce adherence to pharmacological treatments (30).

There are limitations to our study. First, although the study had a high response rate in the successive data collection phases, loss to follow-up accumulated over time, as in most long-term cohort studies. However, large differences in missing data as a function of psychological distress and type 2 diabetes seem unlikely. Second, participants of the Whitehall II Study are from an occupational cohort that is likely to cover a “healthier” end of the variation in health status compared with the general population, which limits the generalizability of our findings. Third, we used the GHQ-30 to detect psychological distress symptoms. As this instrument was not designed to make a psychiatric diagnosis of depression or anxiety, we cannot exclude the possibility of confounding by unmeasured depression or anxiety disorders. However, this is unlikely to be a major source of bias because the GHQ-30 has been shown to be a valid screening instrument for mental disorders, particularly depression in the Whitehall II Study (28). The strengths of this study are its large sample size and use of accurate, repeat assessments of all examined variables, use of the FRS based on clinical measurements, and ascertainment of diabetes based on the standard 75-g OGTT at each clinical study cycle (23,24). However, part of the incident type 2 diabetes identification was based on self-report, although antihyperglycemic medication was confirmed by asking the participants to take their medications or list of medications to the study clinic. Of those participants who had self-reported diagnosis of diabetes without evidence on the use of antihyperglycemic medications (33.4% of incident cases), a substantial proportion was confirmed by a repeat OGTT or by antihyperglycemic medication at the subsequent phase, leaving only 23.6% of all incident diabetes cases unconfirmed.

In summary, this observational study suggests that psychological distress may be related to accelerated progression of late-stage prediabetes to clinical diabetes. Further investigations are needed to examine mechanisms linking psychological distress to onset of type 2 diabetes in this group. Current prevention guidelines do not generally consider psychological factors such as stress or depression (31), although some recognize them as contributing factors to be taken into account in efforts to prevent type 2 diabetes (32). Given the high comorbidity between psychological problems and diabetes and the accumulation of evidence on the role of psychological distress as a predictor of type 2 diabetes, it is important to consider whether more attention should be paid to psychological status among high-risk populations in addition to lifestyle modifications.

Funding. The Whitehall II Study is supported by grants from the Medical Research Council (K013351); the British Heart Foundation; the National Heart, Lung, and Blood Institute; National Institutes of Health (NIH) (R01-HL-036310); and the National Institute on Aging, NIH (R01-AG-013196 and R01-AG-034454). M.V. is supported by the Academy of Finland (258598 and 265174); A.G.T. by the TÁMOP 4.2.4.A/1-11-1-2012-0001 National Excellence Program (a research fellowship cofinanced by the European Union and the European Social Fund); T.N.A. by the National Heart, Lung, and Blood Institute (R01-HL-036310) and by the Languedoc-Roussillon Region (Chercheur d’Avenir Grant 2011); and A.S.-M. by the National Institute on Aging, NIH (R01-AG-013196 and R01-AG-034454). M.K. is supported by the Medical Research Council (K013351), the NIH (R01-HL-036310 and R01-AG-034454), and a professorial fellowship from the Economic and Social Research Council.

The study sponsors had no role in study design, analysis, or interpretation of the data or preparation, review, or approval of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.V. wrote the manuscript and researched data. J.E.F. wrote, reviewed, and edited the manuscript. A.G.T. reviewed and edited the manuscript. T.N.A. reviewed and edited the manuscript. J.V. reviewed and edited the manuscript. A.S.-M. reviewed and edited the manuscript. M.K. helped in study design and wrote, reviewed, and edited the manuscript. M.V. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.