Statistical analyses

Baseline characteristics of study participants are presented as median and interquartile ranges [IQR]. Log-rank test (chi2) was used to assess the correlation between genotype and binomial variables. Cox proportional hazard regression model, adjusted for population stratification based on MDS coordinates, was used to examine associations between genotype and efavirenz discontinuation. Primary analyses considered three metabolizer groups (slow, intermediate, and extensive). Post hoc analyses considered 12 ordinal strata of plasma efavirenz exposure (defined by various combinations of the four CYP2B6 and CYP2A6 SNPs). Analyses were also performed separately in subgroups stratified for self-reported race/ethnicity, excluding individuals in whom self-reported race disagreed with clustering based on MDS coordinates. All analyses used a 5% two-sided significance level and were performed using Stata/IC version 10.0.