Working overtime predicted the onset of a major depressive episode in a middle-aged cohort of British civil servants, followed for an average of 5.8 years. Working 11 or more hours a day was associated with a 2.3- to 2.5-fold risk of an MDE when compared with working a standard 7–8 hours a day. This association was robust to adjustment for a range of socio-demographic, life-style, and work-related factors at baseline.
The main strengths of our study are its relatively large sample size, the prospective design and the use of the standardized CIDI interview which has been shown to be a valid measure of DSM-III-R non-psychotic disorders 
. However, some limitations are noteworthy. First, the CIDI interview was only available at follow-up so baseline cases had to be excluded based on GHQ-30 caseness. However, the GHQ is a well-established scale for the evaluation of psychological morbidity in general population samples. In relation to diagnosed mental disorders, especially mood and anxiety disorders, the GHQ has shown good clinical validity 
. As the GHQ-30 also detects a range of minor psychiatric disorders, such as subclinical depression, it is possible that our baseline exclusion of GHQ-30 cases is over zealous.
Second, being based on observational data, this investigation cannot rule out the possibility of residual confounding by other, unmeasured or imprecisely measured predictors of new-onset depression. We also found that the statistically significant association between 11+ weekly hours of work and the onset of depression was hidden in models not adjusted for individual SES. As SES was inversely associated with depression and positively associated with working hours, its effect can be considered as suppression, ‘a situation in which the magnitude of the relationship between an independent variable and a dependent variable becomes larger when a third variable is included’ 
. In terms of prevention, revealing the relevance of long working hours as a risk factor among high-SES employees who otherwise have lower risk of depression seems important.
A further limitation relates to our inability to assess interaction effects due to the relatively small numbers of new-onset cases of MDE. For example, it is possible there are sex differences in the association between long working hours and mental health 
. Some positive work characteristics, such as high control or high rewards at work, may buffer an employee against the adverse health effects of long working hours 
. On the other hand, working long hours may also mean higher exposure to adverse working conditions. Further studies with larger sample sizes are needed to examine possible interaction effects in relation to clinical depression.
The rate of depression at follow-up was 3.1%. The median 1-year prevalence of depression across studies in general populations has been higher, approximately 5% 
. This might be because only participants free of psychological distress at baseline were selected for follow-up. Furthermore, the Whitehall II study comprised a working population which has previously been shown to be healthier than the general population 
. As all civil servants are white collar workers, it remains to be investigated whether our findings are generalizable to blue-collar workers and employees in the private sector.
Predictors of the onset of depression in our study were younger age, female sex, lower occupational grade, chronic physical disease, and moderate alcohol use (the odds ratio for depression in participants who used alcohol beyond the recommended limits was 2.19; in the expected direction but not statistically significant at conventional levels). We found no robust associations between marital status, smoking, job strain, or work social support and the onset of depression. Earlier literature suggests quite strong evidence for age (early or mid-adulthood), female sex, chronic physical disease such as CHD, binge drinking, smoking, low SES, and negative stressful life events as predictors of depression 
. Smoking intensity in white-collar employees, such as those in our study, is very likely to be lower than in the general population, which may be one explanation for the lack of an association with depression. Work stress and work social support have been suggested to be associated with depression 
, but a recent review focusing on clinical disorders 
suggested that the association is inconclusive. However, in other analyses in this cohort we have found associations between job strain and depression, especially when the exposure to job strain was repeated 
Our sensitivity analysis of all follow-up participants irrespective of employment status revealed similar, although a weaker association between overtime work and depression compared to findings from analyses restricted to those employed at follow-up. These slightly weaker associations were expected given the possible misclassification of working hours during follow-up among those who were no longer employed. Some of those not employed may have stopped working several years before the onset of depression.
Our findings are in accordance with some observations from previous studies showing a positive association between long working hours and depression 
but contrast with other reports of null findings 
. Mixed findings in this field of research may relate to various operationalisations of long working hours, that is, in some studies the cut-point has been 40 
or 45 hours 
or “more than one hour weekly overtime work” versus less 
and possibly to the different work cultures in which these studies were carried out. Furthermore, some studies included part-time workers in the reference group although it is known that people with pre-existing health problems may choose part-time jobs 
. For a better comparison between studies, a consistent definition for overtime and long working hours is needed in the future.
Plausible explanations of why long working hours are associated with the development of depression can not be drawn directly from our study. Serial adjustment for socio-demographic factors, physical disease, smoking, alcohol use, job strain, and social support at work, had little effect on the association or even strengthened it. Long working hours may in part affect mental health through factors not measured in our study, such as work-family conflicts 
, difficulties in unwinding after work 
or prolonged increased cortisol levels 
. The effect of long working hours on mental health may also be different in women and men 
. To date, the exact aetiology of depression is not known, but it is widely assumed that it is multifactorial involving genetic, biological, and psychosocial factors 
In conclusion, this study suggests an association between long working hours and the onset of major depressive episode. Large-scale population-based cohort studies are needed to examine whether the association can be found in other contexts than Civil Service and intervention studies are needed to examine whether interventions designed to reduce working hours would alter depression risk in working populations.