Use of Marginal Structural Models to Examine the Bidirectional Association between Depression and Excess Body Weight

Abstract

Background and aims – There is mixed evidence about the bidirectional association between excess body weight (EBW) and depression during adolescence. This study estimated the effects of cumulative depression or excess body weight during early adolescence (age 12-14) on the likelihood of excess body weight or depression during late adolescence (age 17-18). We also examined the effects of depression or excess body weight at a specific time point in adolescence on the likelihood of excess body weight or depression at the subsequent adolescent time point. Effect modification by sex for each association was also examined. Design: We conducted a prospective cohort study of youth with repeated measurement. The study sample included 521 adolescents from public middle schools in Seattle, Washington, USA. Measured height and weight from five waves were used to determine BMI percentile based on CDC standards. The child-report version of the Mood and Feelings Questionnaire from five waves was used to assess depressive symptoms. Marginal structural models (MSMs) with stabilized inverse probability weighting (SIPW) that accounted for both time-varying and fixed covariates were used to estimate the strength of associations between excess body weight and depression. Result: The prevalence of EBW ranged from 23.43% to 31.45%, and the prevalence of elevated depressive symptoms ranged from 13.36% to 23.71% across time points. Finding from MSM analyses indicated that cumulative elevated depressive symptoms during early adolescence had no statistically significant effect on excess body weight during late adolescence (RR=1.08; 95% CI: 0.72, 1.62; p-value=0.70). Similarly, findings from MSM analyses indicated that cumulative EBW during early adolescence had no statistically significant effect on elevated depressive symptoms during late adolescence (RR=1.01; 95% CI: 0.66, 1.53; p-value=0.98). Likewise, MSM analyses indicated that there was no statistically significant lagged effect of elevated depressive symptoms on EBW at the subsequent time point during adolescence (RR=1.17; 95% CI: 0.88, 1.55; p-value=0.27). There was also no statistically significant lagged effect of EBW on elevated depressive symptoms at the subsequent time point during adolescence (RR=1.13; 0.85, 1.51; P-value=0.39). Sex did not modify any of these observed associations (interaction p-value > 0.05). Conclusion: We found no statistically significant bidirectional cumulative or lagged association between overweight and depression in adolescence, and sex did not modify the associations.