Fig 3. BMI_CNS and BMI_non-CNS genetic scores associations with quality, quantity, and timing of workplace purchases.

Association results are adjusted betas reflecting difference in Healthy Purchasing Score (percentage), items purchased (units over 3-month period), or timing (in minutes) between highest (Q4) and lowest (Q1, reference) quartile of BMI genetic scores adjusted for age, sex, seasonality, and 5 principal components of ancestry. Higher purchasing score = healthier purchases (0%–100%). For interpretation purposes, adjusted betas are from models in which outcomes are untransformed, whereas P values are derived from models in which outcomes are inverse normalized. P values are unadjusted for multiple testing, and false discovery rate–corrected P values (P_adj) are presented in S2 Table. Based on the biological functions of genes in or near the 97 previously identified BMI loci, the BMI_CNS PRS and BMI_non-CNS PRS comprise 54 variants previously classified as CNS-related and 43 variants previously classified as non-CNS-related, respectively. BMI, body mass index; CI, confidence interval; CNS, central nervous system; GPS, genome-wide polygenic score; int, interaction; PRS, polygenic risk score.