Association between depression, anxiety and weight change in young adults

The present study analysed data from the Childhood Determinants of Adult Health (CDAH) study. Details of the study design and participants were previously reported [26, 27]. In brief, the CDAH study started with the follow-up of 8498 schoolchildren (7–15 years) who participated in the 1985 Australian Schools Health and Fitness Survey (ASHFS) [27]. The CDAH study was set up to investigate the contribution of childhood factors to the risk of developing cardiovascular diseases in later life [27]. At baseline (2004–2006), and at follow-up (2009–2011) participants completed assessments of mood disorder, anxiety disorders and their lifestyle behaviours. The present study included 11,638 participants who have complete anxiety and BMI data at baseline and at follow-up, and 1646 who have complete baseline and follow-up data on BMI and mood disorder (Fig. 1) [28]. Participants were 26–36 years at baseline and 31–41 years at follow-up.

At baseline and follow-up, participants reported recent sociodemographic information, including occupation, education, and marital status. Details of the baseline and follow-up measurements and procedures have been described previously [26, 28]. Family history of cardiometabolic disease (myocardial infarction, stroke, high blood pressure, coronary heart disease and diabetes), lifestyle behaviours and risk factors, including smoking, and alcohol consumption were recorded using self-administered questionnaire [26].

When completing the Food Frequency and Habits questionnaire, participants were asked to refer to their diet during the 12 months prior to the interview. The combined Dietary Guideline Index (DGI) was used to generate a dietary score. It assesses adherence to the Dietary Guidelines for Australian Adults and covers consumption of foods from the core food groups, along with other items such as use of salt and fluid intake. The DGI includes 15 items that are scored using 127 items from a food frequency questionnaire [12]. Higher scores indicated closer adherence to dietary guidelines. The International Physical Activity Questionnaire (IPAQ) was used to assess the duration of moderate- and vigorous-intensity leisure, work, active commuting, and yard or household physical activity in the past week. Total physical activity is calculated by summing all four domain sub-scores according to the IPAQ scoring protocol [29].

Residential postcode was used to derive area-level socio-economic status based on the Australian Bureau of Statistics Postal Area Index of Relative Socio-Economic Disadvantage. The Henderson Interview Schedule for Social Interaction (ISSI) asks respondents about the number of people with whom they share strong affectual attachments and about the perceived adequacy of these relationships were used to derive a Social Support Index (SSI) [30]. We used the NEO Five-Factor Inventory (NEO-FFI) to assess the personality (Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness) of the participants [31, 32].

Serum C-reactive protein (CRP) at baseline was measured from fasting venous blood samples using an automated analyser and a highly sensitive turbidimetric immunoassay kit. Fasting plasma glucose was measured enzymatically with the Olympus AU5400 automated analyser. Blood pressure was measured at follow-up using a digital automatic monitor.

At baseline, BMI was calculated as weight divided by the square of height. For those with self-reported height and weight data, a correction factor was applied based on a validation study from the same cohort [33]. At follow-up, all participants reported their BMI, and therefore the correction factor applied to all BMI. We classified persons into three BMI categories: Normal weight (18.50–24.99), overweight (25 ≤ BMI ≤ 29.99) and obese (BMI ≥ 30).

At follow-up, lifetime history of DSM-IV disorder were obtained from the Composite International Diagnostic Interview (CIDI-Auto 2.1 version) via telephone interview with trained interviewers [26].

“Presence of mood disorder history at baseline” was defined as an episode of depression/dysthymia that occurred prior to baseline; and “presence of anxiety history at baseline” as an episode of anxiety that occurred prior to baseline. We categorized those with a diagnosis of mood disorders, or anxiety disorder between baseline and CDAH2 as having ‘episode’ (first or recurrent events) of these disorders, or having ‘first episode’ in those without history mood or anxiety disorders who suffered from these disorders for the first time since baseline.

Of the total 634 males and 1012 females who have complete data for BMI and presence of mood disorder history at baseline and at follow-up, 85 (13.4%) males and 229 (22.6%) females had a history of mood disorder at baseline. In males, after adjustment for covariates in the base model, including age, education, family history of cardiometabolic diseases and social support score, history of mood disorder at baseline was associated with an increase in BMI (β = 0.77, 95% CI: 0.14–1.40, P = 0.017). Serial adjustment for potential mediators, including smoking status, adherence to DGI, moderate-to-vigorous physical activity reduced the average BMI increase due to history of mood disorder to 0.70 (β = 0.70, 95% CI: 0.01–1.39). This suggests that 0.07 kg/m² of the average increase in BMI is mediated by these factors. However, antidepressant use has no effect on the association between history of mood disorder and BMI change (β = 0.77, 95% CI: 0.05–1.49). In males, extra food consumption, moderate or vigorous physical activity, dietary adherence and smoking status appeared to be the strongest mediators (Table 2).

Among females, a history of mood disorder was associated with a smaller increase in BMI (β = 0.53, 95% CI: 0.00–1.06, p = 0.048), which was attenuated (β = 0.39, 95% CI: − 0.16-0.95) after further adjustment for participants health behaviours. The full list of the effect of individual covariates on the associations between history of mood disorder and BMI change in males and females is presented in Table 2.

In the base model, history of anxiety was associated with a non-significant increase in BMI in males (β = 0.16, 95% CI: − 0.45-0.76) and females (β = 0.01, 95% CI: − 0.59-0.61). Further adjustment for potential mediators, including C-reactive protein, antidepressant use, physical activity, diet and smoking did not significantly change the associations between history of anxiety and BMI change in both males (β = 0.19, 95% CI: − 0.44-0.62) and females (β = 0.16, 95% CI: − 0.48-0.80). The results are presented as Additional file 2.

In females (RR per kg/m²: 1.04, 95% CI: 1.01–1.08), BMI at baseline tended to increase the risk of episode of mood disorder between baseline and follow-up, and mood disorder diagnosed in the 12 months prior to CDAH2 (RR per kg/m²: 1.03, 95% CI: 1.00–1.07). Serial adjustment strengthened both the risk of mood disorder (RR per kg/m²: 1.07, 95% CI: 1.00–1.15) and risk of mood disorder diagnosed in the 12 months prior to CDAH2 (RR per kg/m²: 1.06, 95% CI: 1.00–1.13) (Table 3).

In males, BMI at baseline did not predict the risk of mood disorder (RR per kg/m²: 0.97, 95% CI: 0.90–1.04), or mood disorder diagnosed in the 12 months prior to CDAH2 (RR per kg/m²: 0.94, 95% CI: 0.86, 1.03) in males. These findings persisted after adjustment for potential mediators including health behaviours, diet and physical activity. In females, adjusting for body weight satisfaction increased the risk of an episode of mood disorder (RR per kg/m²: 1.08 95% CI: 1.00–1.16) but not mood disorder diagnosed in the 12 months prior to CDAH2 (RR per kg/m²: 1.11 (0.94–1.30). In males, body weight satisfaction was not associated with episode of mood disorder or mood disorder diagnosed in the 12 months prior to CDAH2 (Table 3).

After adjusting for confounding factors in the base model, BMI at baseline did not significantly predict episode of anxiety in males (RR per kg/m²: 0.96, 95% CI: 0.87–1.07) or females (RR per kg/m²: 0.99, 95% CI: 0.94–1.04). Further adjustment for potential mediators, including body weight satisfaction, did not significantly change the associations between baseline BMI and episode of anxiety in both males (RR per kg/m²: 1.07, 95%CI: 0.76–1.50) and females (RR per kg/m²: 1.00, 95%CI: 0.93–1.09). The finding of no significant association between baseline BMI and episode of anxiety did not change when the analyses was limited to those diagnosed with anxiety in the previous 12 months of the CDAH2 in both males and females. The results are presented as Additional file 3.

This study adds to the relatively small number of prospective studies examining the temporal relationship between mood and anxiety disorders and change in BMI while controlling for the mediation effects of a wide range of variables. It is one of few studies to examine the effect of a wide range of potential confounding or mediating variables on these associations. While a bidirectional association between anxiety and BMI change is rarely reported, evidence regarding bidirectional association between mood disorder and BMI change is inconsistent [23, 24, 38]. There are several possible reasons for the inconsistent results among the studies including variation in the populations studied in terms of age at onset, duration of episodes, duration of study follow-up, method of mental disorder assessment, and severity of these disorders [21, 39, 40]. Recent evidence suggest that the unfavourable effect of depression on development of weight gain and the effect of weight gain on development of depression may be reinforced by follow-up time, suggesting that the associations between these disorders may be less apparent in studies with short follow-up [21]. Mood disorders with onset at younger ages tend to have a more unfavourable course and are associated with greater comorbidity than those with a late-onset [39]. Previous studies have also shown that younger adults are more likely to seek care for their mental illness, reducing the duration of an episode and recurrence rates [40‐42], onset of mood disorder-related unhealthy lifestyle and comorbid conditions [40, 42].

The positive association between history of mood disorder and subsequent increase in BMI in males is consistent with the findings of some studies [43, 44]. The increase in BMI among males attributed to mood disorder is relatively small, although it could be a clinically significant increase in light of the wide range of adverse health effects with even a small increase in BMI above the normal range. For example, a one- kg/m² increase in BMI raises heart failure risk by 17% [45], and type 2 diabetes by 15 and 11% in men and women, respectively, independent of subsequent weight changes [46].

There could be several explanations for our findings that the association between history of disorder and BMI was weaker in females than in males. Firstly, the statistical power to detect association is reduced due to variation introduced when applying methods to address large scale attrition in this cohort. The effect size estimate for females is smaller than that for males. This might be explained by the fact that females are more likely to report mental health disorders and seek mental health services [41], which may improve their quality of life and response to mood disorders [47]. According to the 2007 Australian health survey, more women (41%) with a 12-month mental disorder accessed services for mental health problems than men (28%) [48]. In the UK, more than 750,000 people were referred for counselling for anxiety and depression in 2012–2013, of whom 62% were women [41]. Women also tend to have a better coping strategies and favourable outcomes than men, which could possibly attenuate the potential BMI increase among women with history of mood disorder [49, 50]. In addition, evidence shows that depression-related unhealthy lifestyles, which in turn could lead to BMI change, are more common in males than in females [51].

We found that the association between baseline BMI and elevated risk of mood disorder was limited to females. This is consistent with findings from previous studies, which have demonstrated a relationship between BMI and subsequent risk of mood disorder in females but not in males [52, 53]. A longitudinal study of 4410 participants in the US found that overweight in adolescence was associated with higher odds of developing depressive symptoms (odds ratio:1.74) in later adulthood in females, but not in males [53]. One possible explanation is that, in women, higher BMI is related with negative body image perception and satisfaction, which in turn, is associated with higher risk of mood disorders [54]. According to previous studies, weight dissatisfaction is both common and has stronger effect on depression in middle-aged than in younger adults [55, 56].

Our findings on the mediating role of dietary pattern and health behaviours, for mood disorder and BMI, are consistent with the common factors associated with both conditions. Studies suggest that people living with mental health disorders are significantly more likely to engage in low levels of physical activity, more likely to smoke cigarettes and have poor diet quality [57, 58], which in turn are known to increase the risk of weight gain. On the other hand, weight gain increases risk of declining physical activity and diet quality over time [59]. Furthermore, increase in body weight has been shown to activate chronic low-grade inflammation and low-grade systemic inflammation contributes to the development of mental disorders [60, 61]. From a public health point of view, the mediating effect of physical activity and dietary patterns on the risk of BMI and mood disorder suggest that interventions targeting these risk factors may not only promote physical health but also improve mental health.

The finding that history of anxiety was not associated with subsequent changes in BMI and vice-versa, is consistent with reports from previous studies [62‐64], although positive associations have also been reported [38, 65, 66]. The reasons why previous studies have been inconsistent regarding the association between anxiety and change in BMI are not clear. One possible explanation could be the heterogeneity among the studies in terms of duration of follow-up, participant characteristics, and confounding variables. One previous study suggest that the association between depression and weight gain increases with increasing duration of follow-up, which could also be the case for anxiety [21]. Some of the studies that reported a positive association between weight or BMI and anxiety had longer duration of follow-up (up to 10 to 20 years) [38, 65]. Race may also play a role in the association between anxiety and weight gain. A population-based study among 103,557 residents of Olmsted County, Minnesota found that anxiety was significantly associated with obesity in Blacks but not in Asian or Hispanic groups [67]. However, data on race was not collected in our study, however, the majority of participants in the current study were Caucasian.

This study has some limitations. First, there was a substantial loss to follow-up from the original ASHFS random sample to the analysis sample. Study participants included in the present analyses were more likely to be employed, educated beyond high school, and have a healthier lifestyle [28]. However, inverse probability weighting, including variables that predicted non-response, was applied to minimize the effect of selection bias due to loss to follow-up. Participants were young-adults, and application of the findings to older populations in the community may be limited. Due to self-reporting, some misclassification of anxiety or mood disorder and under or over estimation of body weight is inevitable. However, both misclassification and under or over reporting are probably non-differential across the different study outcomes. Controlling for the confounding effect in the analyses may not entirely remove the confounding effect of some variables, such as diet and physical activity, which might have been measured without sufficient accuracy. Adjusting for change in predictor variables from baseline to follow-up may have potentially introduced reverse causation bias. However, the conclusions were consistent when the outcomes were BMI recorded at last follow-up and anxiety and mood disorder in the last 12 months. Furthermore, the multiple-testing issue may also be a limitation of this study. A substantial proportion of the participants were overweight or obese at baseline [33], which might have underestimated the change in BMI attributed to anxiety or mood disorders. Although correction for self-reporting error has been applied, the use of self-reported height and weight in some participants may have resulted in under-reporting of BMI.

Strengths of this study include its well-characterized cohort of younger adults, its use of adequate sample size, and its exploration of the mediation and confounding effect of a wide range of variables. Both lifestyle-related risk factors, and anxiety and mood disorders were measured using standardised data collection tools commonly used for epidemiological studies. The fact that the participants were younger individuals means that the findings are less likely to be confounded by chronic conditions. The findings were consistent when the analyses were restricted to those who have recent episode of anxiety or mood disorders.