Obesity and psychotropic medication: a prospective register linkage study among midlife women and men

The Helsinki Health Study baseline mail surveys were conducted in 2000–2002 among 40–60-year-old employees of the City of Helsinki, Finland. There were 8960 respondents (response rate 67 %). Corresponding to the gender distribution in the Finnish municipal sector, 78 % of these respondents were women. Men, younger employees and manual workers were slightly underrepresented, but according to our non-response analyses, the data satisfactorily represent the target population [22]. Data on prescribed reimbursed psychotropic medication purchases from 1995 to 2009 were derived from registers of the Social Insurance Institution of Finland. The follow-up started from the day of returning the survey questionnaire and lasted until the end of 2009, the date of purchase of the psychotropic medication, or death (n = 176). The mean follow-up time was 6.4 years. The register data included the date of purchase, the type of prescribed medication and the amount in defined daily doses (DDD). Seventy-four per cent of the participants gave consent to the register linkages. According to our non-response analyses, employees from lower occupational and income classes were somewhat underrepresented, but the differences were small and are unlikely to cause a major bias in the results [23].

The 319 persons who were taking psychotropic medication at baseline were excluded. Further exclusions included 15 pregnant women, 58 underweight (BMI <18.5 kg/m²) individuals and 135 persons with missing information on some of the variables. Thus the final analysis included 6098 participants, of whom 78 % were women.

The ethics committees of the Department of Public Health, the University of Helsinki and the health authorities of the City of Helsinki approved the Helsinki Health Study protocol.

Cross-tabulation was initially used to describe the data. Kaplan-Meier curves were calculated to estimate the proportions of different weight groups with no psychotropic medication. Hazard ratios (HR) and 95-per-cent confidence intervals (95 % CI) for the first purchase of any psychotropic medication during the follow-up were calculated using Cox regression analysis. In addition, antidepressants and sedatives were examined separately. Normal weight was the reference category in all the analyses. Model 1 was adjusted for age and model 2 for age and previous psychotropic medication. Model 3 was also adjusted for covariates, including socio-economic position, marital status, physical and mental strenuousness of work, physical activity, smoking, and drinking behaviour. Model 4 was additionally adjusted for physical functioning and somatic ill-health. Women and men were examined separately. We used Schoenfeld’s residuals to confirm Cox proportional hazards assumptions. The analyses were performed using IBM SPSS Statistics 22.

We examined the association between body mass index and subsequent psychotropic medication among midlife employees. Obesity and especially severe obesity were associated with having some psychotropic medication among the men, but the association attenuated following adjustment for previous psychotropic medication and other covariates. However, the association between severe obesity and sedatives among the men remained after adjustment. We found no association between BMI and subsequent psychotropic medication among the women.

The majority of previous studies indicate an association between obesity and subsequent mental ill-health, findings that have been particularly consistent among women. We found no associations among the women, however, and only weak associations among the men. There are some differences to be noted when the results of this study based on prescribed, reimbursed psychotropic medication purchases are compared with earlier studies based mainly on self-reported mental-health data. First, our measure may be less sensitive to mental ill-health in that a medical prescription is based on a professional examination made by a physician in addition to symptoms reported by the patient. However, the threshold for prescribing psychotropic medication is probably quite low in Finland: the majority of common mental disorders are treated by general practitioners and access to reimbursed rehabilitative psychotherapy is restricted, and is first considered after 3 months of other appropriate treatment. Second, psychotropic medication for the obese may be affected by physicians’ preconceptions of obese people. It has been reported that physicians may consider obese people less motivated and compliant than patients of normal weight [29], and it is therefore possible that mental health among the obese is less likely to be treated as a medical problem. Third, obese people are more likely than those of normal weight to show other somatic health problems that require medical treatment, whereby psychotropic medication might be avoided due to the risk of drug interaction. Fourth, the threshold for consulting a doctor on account of mental ill-health might be higher among the obese than among those of normal weight. According to the results of one study, 84 % of obese participants agreed that “weight is blamed for most medical problems” [30]. However, the obese may have more regular contact with healthcare professionals on account of their somatic health, and thus their mental disorders could more easily be noticed.

In contrast to the majority of previous studies, the participants in our study were employed at baseline. There is one earlier study from the US that also focused on employees, but only female nurses were included and the participants were older, with an average age of 65 and most of them were probably no longer employed [13]. It is possible that work environments protect the obese from developing mental ill-health, given that the employed have been shown to have better health than the non-employed [31]. A recent study, using the same data, examined joint associations of BMI and leisure-time physical activity with psychotropic medication and suggested that physical activity dominates the association over normal-weight [32]. It is possible that employed people have healthier lifestyle e.g. less alcohol use and more exercise, which might partly protect them against mental ill-health. In further studies it would be interesting to compare employees with non-employed in order to study the differences in life style and the possible protective effect of employment.

Another difference from the majority of earlier studies is that our cohort focused on midlife people. Age has been discussed as both a protective and a risk factor for mental ill-health among the obese. It is suggested in a recent longitudinal study that the associations could be weaker among adults on account of improved emotional regulation and less stigmatisation associated with weight by age [15]. However, different views have also been presented, and a Finnish study has identified an association between advancing age and lower levels of body satisfaction on many aspects of body image [14, 33]. The study also showed that women associated their body mass index with their body image more strongly than men. It was therefore unexpected in our study that the association between obesity and mental ill-health was stronger among the men. However, the prevalence of psychotropic medication was higher among the women in our cohort, and it is possible that mental ill-health is perceived and treated differently among women and men, which might weaken the associations among women.

In contrast to the majority of earlier research, some recent studies [14‐17] somewhat in line with our results, report no association between obesity and subsequent mental ill-health among women. It could be that the increased prevalence of obesity has changed people’s attitudes and made obesity more acceptable. Reducing the stigma attached to weight could positively affect self-esteem among the obese and protect against mental ill-health [34]. However, an extensive review did not produce evidence of a decrease in the stigmatisation of obesity [29], although it did show some differences between European countries compared with the US and Australia. Given such differences between countries, it would be interesting in further studies to focus on how the neighbourhood environment and culture contribute to the stigmatisation of weight and thus the association between obesity and mental health.

The main strength of this study lies in the use of extensive prospective data (n = 6098) including employed women and men with different occupations and backgrounds. Another strength is the long follow-up time (a mean of 6.4 years). A third strong point is the use of reliable register data on prescribed psychotropic medication, including all prescribed reimbursed psychotropic medication purchased in pharmacies in Finland. Prescribed psychotropic medication from registers reflects mental health and provides data that avoids self-report bias. The register data on psychotropic medication have previously been used in order to examine associations between mental health and e.g. physical activity, workplace bullying, traffic noise, and marital status [20, 35‐37].

However, our study population included only one, albeit large, municipal employer and is restricted to midlife employees. Height and weight measurements as well as the covariates are based on self-reports: people tend to underestimate their weight and to overestimate their height. However, self-reported BMI has been shown to predict sickness absence equally accurately as measured BMI [38]. The severely obese were few in number (n = 209), but the results did not differ from the ten per cent with the highest BMI. We also divided BMI into quartiles in the sensitivity analyses, instead of following the WHO recommendation, but this did not affect the results, either. The baseline-survey response rate was acceptable, but non-response remains a problem, and those under psychotropic medication may be overrepresented among the non-respondents.

A limitation with the data on medication is that the register does not include information on psychotropic medication that was prescribed but not purchased, or not reimbursed. We also lack information on the clinical indication of the prescribed medication. In some cases antidepressants are used to treat insomnia, chronic pain, premenstrual syndrome and anxiety disorder, whereas sedatives can be used during sudden life crises to treat insomnia or seizures, as well as anxiety. According to our sensitivity analyses, single prescriptions were evenly distributed over the weight groups, and even when 365 DDD was used as a cut-off point to indicate long-term psychotropic medication the results were similar. Given the overlapping of the clinical indication of the medication we decided to focus on psychotropic medication in general, and in addition examine antidepressants and sedatives separately. Because the indication of sedatives is varied, we carried out additional analyses separating anxiolytics (N05B) and pure sedatives (N05C, including sedatives and hypnotics). We found no association with anxiolytics among the men, but there was an association with pure sedatives. Sedative use could be related to sleep disorders such as obstructive sleep apnoea, which is a relatively common but frequently undiagnosed disorder among obese, middle-aged men [39]. Further studies are therefore needed to examine the association between psychotropic medication and sleep disorders, especially among obese men.

The Kaplan-Meier curves (Fig. 1) were steeper during the first year, suggesting a time-dependent association. However, when tested on the basis of Schoenfeld’s residuals the Cox proportional hazard assumptions held. Nonetheless, we carried out a sensitivity analysis in which the follow-up was cut off at 5 years, and the associations were only marginally stronger and again disappeared following adjustment. It is possible that adjusting for physical functioning and somatic health leads to over-adjustment in that physical health could also mediate the examined associations. For this reason we did not include physical health in model 3, although even when adjusted only for covariates in model 3, the statistical significance was lost.