Does obesity really protect against psychological distress? Examining the ‘fat-jolly’ versus ‘fat-sad’ hypotheses using Mendelian randomization


Mika Kivimäki, PhD, Department of Epidemiology and Public Health, 1-19 Torrington Place, University College London, London WC1E 6BT, UK.
(fax: +44(0)20 74196732; e-mail:

The association between obesity and psychological distress has been an issue of controversy for decades. While several studies suggest that obesity increases the risk of distress, this has not been observed in all cohort studies and some investigators have reported that obese persons have a reduced risk of psychological distress.

In this issue of the Journal of Internal Medicine, Lawlor et al. [1] present interesting results to address this puzzle by utilizing two genetic variants that are robustly associated with adiposity: FTO rs9939609 and MC4R rs17782313. This is a Mendelian randomization (MR) study with the aim of improving causal inference from observational data which are often confounded by third variables, some of which are measured, others not. The MR approach is predicated upon the random assortment of alleles at the time of gamete formation, which leads to population distributions of genetic variants that are, generally, independent of the environmental exposures commonly confounding risk factor–disorder associations [2]. These unconfounded genetic differences in risk factor levels (in this case adiposity) should translate into genuine differences in disorder occurrence (i.e., psychological distress) if the exposure is truly a causal risk factor.

Lawlor et al. [1] demonstrate that both their genetic ‘instruments’ of obesity –FTO rs9939609 and MC4R rs17782313 – were associated with a reduced risk of self-reported psychological distress and antidepressant medication use in a large cohort of over 50 000 Danish men and women. According to the best estimate of their MR analysis, people with normal weight had over three times greater odds of stress and anxiety compared to their obese counterparts. In 2008, a meta-analysis of 15 conventional cohort studies examining the relationship between obesity and depression (total N = 59 000) found the opposite findings such that obesity, and to a lesser extent overweight, at baseline increased the risk of depression onset at follow-up, with summary odds ratios of 1.6 and 1.3, respectively [3].

Thus, MR seems to support a ‘fat-jolly’ hypothesis whereas conventional analysis has favoured a ‘fat-sad’ hypothesis. The present paper reflects this apparent discordance: on running a standard multivariate regression analysis without genetic instruments, Lawlor et al. [1] report that the odds for antidepressant medication use, stress and anxiety were greatest among the obese. The findings from MR are provocative because they seem to suggest that there is a powerful ‘unknown factor’ that reverses the strong true protective effect of obesity identified in MR to a potentially spurious heightened risk in observational data.

Let us assume that the standard analysis reveals spurious findings and that the MR reveals the truth. From a public health perspective, this could have some worrying implications. Antidepressants are currently one of the most prescribed medicines world-wide [4], and depressive disorders are projected to be among the three leading causes of burden of disease in 2030 [5]. If the efforts to reduce the global obesity epidemic were successful, the expected decrease in premature mortality and cardiovascular disease would, according to the MR findings, be accompanied by an increase in mental health problems and in consumption of antidepressants.

The problems of conventional epidemiologic analysis – bias and confounding – are well known, but MR is also not without its shortcomings. First, the genetic instruments typically represent only a limited range of adiposity values: we expect that the difference in mean BMI between extreme genotype groups was <2 kg m−2 in Lawlor’s analysis. The influence of BMI (ranging from an ‘underweight’ of BMI < 18.5 to an ‘obese’ of BMI > 30) on distress is therefore likely to have been imprecisely estimated despite the large sample size. Second, genetic instruments should be associated with the disease outcome only via their association with the exposure (here adiposity). In the present MR study, unexpectedly, adjustment for BMI and waist circumference did not attenuate the associations of FTO and MC4R with psychological distress [1]. If this is because the genetic variants affect distress via other pathways besides obesity (e.g. due to pleiotrophy or developmental compensation) or because the gene-distress association was observed by chance, the MR estimates are likely to be biased. Third, the findings did not reproduce a corresponding MR analysis of 2981 male British civil servants (the Whitehall II study) [6]. In that cohort we found the FTO genotype to be positively associated with both adiposity (measured four times over a 19-year period in the adult life course) and common mental disorders; long-term obesity thus seemed to increase rather than decrease the likelihood of symptoms of depression and anxiety in that MR analysis.

Like empirical papers, commentaries rarely conclude that ‘no further studies are needed’. The analysis of Lawlor et al. clearly makes a substantial contribution to this controversial field of research. However, their findings also call for more research that extends, and possibly challenges, conventional analysis of obesity and psychological distress to achieve the final verdict on the validity of the ‘fat-sad’ versus ‘fat-jolly’ hypotheses. Recently, a consortium of genome-wide association studies confirmed 14 previously reported obesity susceptibility loci and identified 18 new loci associated with body mass index [7]. These 32 genetic variants should be utilized in future MR analyses. Using an allele score of these independent loci as the instrument is likely to improve its specificity to the adiposity trait. Furthermore, multiple genetic variants would capture a wider range of adiposity than single variants, so increasing analytic power. With the help of MR and related statistical methods, these advances in genetics should also benefit research on the effects of modifiable risk factors other than obesity and psychological distress.

Conflict of interest statement

No conflict of interest was declared.