Figure 3

Scatterplot of adjusted versus unadjusted within couple correlation estimates across all metabolites. In adjusted analyses in males and females separately, metabolites (in SD units) were regressed on covariates (age, educational attainment, ever smoking, physical activity, family history of cardiometabolic disease, alcohol consumption, BMI, and ethnicity) and the residuals of those models use to estimate the adjusted Pearson correlation coefficients within couples. Unadjusted analyses were the Pearson correlation coefficients of the metabolites in SD units. The green dots highlight the individual metabolites. The grey dashed line reflects the reference line (slope 1, intercept 0) and the red dashed line is the best line of fit.