These files include summary results for the proinsulin meta-analysis (in press, Broadaway et al 2023 AJHG). The meta-analysis was composed of 16 European-ancestry studies. The model adjusting for BMI had maximum sample size of 45,826, while the model not adjusting for BMI had a maximum sample size of 45,861 individuals.
Brief methods:
Fasting proinsulin values (pmol/L) were natural logarithm transformed and adjusted for age, sex, population structure, and natural logarithm of fasting insulin. Covariate adjustment was performed in two steps: first modeling natural logarithm of fasting proinsulin on all covariates, then inverse normal transforming the residuals, and then modeling the inverse normally transformed residuals on the covariates again. We performed a fixed-effects inverse-variance weighted meta-analysis using METAL, and applied double genomic-control (GC). Variants must be represented by at least one quarter of the maximum sample size, in at least two studies, and have an overall MAF > 0.005 for inclusion.