A Genetic IV Analysis to Assess the Effect of Body Mass Index on Depression in the Health And Retirement Study

TitleA Genetic IV Analysis to Assess the Effect of Body Mass Index on Depression in the Health And Retirement Study
Publication TypeJournal Article
Year of Publication2013
AuthorsWalter, S, Mejía-Guevara, I, Tchetgen Tchetgen, EJ, Koenen, KC, Kawachi, I, Kubzansky, LD, M. Glymour, M
JournalAm J Epidemiol
KeywordsGenetics, Health Conditions and Status, Public Policy

Background: The relationship between adiposity and mental health remains controversial despite extensive prior observational research. We use a genetic risk score (GRS) as an instrumental variable (IV) to estimate effects of increases in body mass index (BMI) on depressive symptoms. Methods: Data are from 11842 (whites = 10322, blacks = 1520) individuals from the nationally representative Health and Retirement Study with data on a 10-item Center for Epidemiologic Studies Depression (CESD) scale and genetics collected in 2006 or 2008. Based on 32 Single Nucleotide Polymorphisms (SNPs) known to predict BMI, we calculated a GRS as the sum of the number of risk alleles multiplied by the per allele effect size on BMI (based on prior publica- tions). We conducted over-identification tests with 4 IVs using subsets of the SNPs organized based on biological pathways (adiposity, appetite, and cardio-pulmonary factors) by which the genes might influence adiposity. Self-reported BMI prior to the assessment of CESD was used for validation of the external instrument. The CESD score was regressed on each genetic IV plus population stratification eigenvectors to derive the IV effect estimate. We compared these to conventional effect estimates derived from regressing CESD on self-reported BMI. Results: The GRS significantly predicted BMI in blacks (r2=0.016, p 0.001) and whites (r2=0.010, p 0.001). Self-reported BMI significantly predicted CESD, more strongly in whites (0.043, 95 -CI: 0.036, 0.05) than blacks (0.021, CI: 0.004, 0.038). IV effect estimates were large and statistically significant for whites (0.0 96, CI: 0.026, 0.167) but had very wide CIs for blacks (0.016, CI: -0.224, 0.256) . Over-identification tests showed no evidence of possible pleiotropy in the genetic risk score. Conclusions: Our IV analysis reveals that genotypes associated with BMI are also associated with depressive symptoms in whites, supporting a causal effect of BMI on depression.

Endnote Keywords

public Health/occupational health/Body Mass Index/Genetic Markers/risk alleles/genetic risk score/Depression/biological pathways/mental Health/genetics/genetics

Endnote ID


Citation Key7900