Sensitivity to bias due to SLCO1B3 c.334T>G SNP
There is no biologically plausible reason to expect prevalence of TT/TGSLCO1B3 c.334T>G genotype (estimated in Figure S4 at 31% in Caucasian Europeans) to differ between ABCG2 c.421C>A variant carriers (“treated”) and wt subjects (“controls”). However, one could assume a range of different imbalances in prevalence of TT/TG subjects occurring by chance. Table S1 illustrates scenarios in which a considerable imbalance occurred by chance - prevalence of TT/TG (associated with higher MPA AUC0-12) among ABCG2 c.421C>Avariant carriers (treated) is at the upper limit of the 95% prediction interval (Figure S4) and is 40%, while in wt subjects (controls) it is at the lower limit of the prediction interval and is 25%. It also illustrates scenarios with an even greater imbalance with 2.5-fold difference in prevalence of TT/TG in ABCG2 c.421 variant carriers (50%) and wt controls (20%). Even if the effect of TT/TG is much higher than estimated in Figure S2 (i.e., it is ROM=1.25, which should be considered markedly higher than the pooled estimate, since it was based on raw, unadjusted [reported] values) – the effect of the variant ABCG2 c.421 allele would still be at least GMR=1.31 (vs. the estimated 1.40).
Table S1 . Sensitivity analysis of the effect of ABCG2 c.421C>A variant allele on MPA AUCτ,ss(estimated in the main analysis as GMR=1.40) to account for hypothetical bias arising from not accounting for the SLCO1B3 c.334T>G SNP. The effect of the TT/TG (vs. GG)SLCO1B3 c.334 genotype is estimated at ROM=1.135 (95%CI 0.949-1.361) (Figure S2) and prevalence of the TT/TG genotype is estimated at 31% (Figure S4). Shown are bias-corrected effects (GMRs) of the ABCG2 c.421C>A variant (vs. wt) assuming considerable imbalance between variant carriers and wt controls in prevalence of the SLCO1B3 c.334 TT/TG genotype, and assuming TT/TG effect as estimated (i.e., 1.15) and higher, up to 1.25.