Supplemental Methods A – Methods to achieve conditional exchangeability
Currently, ABCG2 is not considered in the context of MPA (or MPAG) pharmacokinetics [1] and there is no explicit evidence that either MPA or MPAG are ABCG2 substrates – but there is also no explicit evidence that they are not [2]. We aimed to estimate effect of theABCG2 c.421C>A (rs2231142) SNP (results in reduced transporter numbers and function), i.e., of carrying a variant allele, on steady-state exposure to MPA. As reviewed, studies (so far) have failed to detect association between this SNP and exposure to MPA (as AUCτ or as trough concentrations) [3]. However, in one study in Japanese renal transplant recipients, variant carriers (n=44) had higher (adjusted for MMF dose; expressed per 1000 mg) MPAG AUC0-12 than 36 wt controls (median 1540 vs. 1195 mg × h/L; P=0.029) [4]. Authors suggested that ABCG2 might be included in biliary excretion of MPAG [4]. By analogy with reports of association between OATP1B1 and OATP1B3 polymorphisms (MPAG is a substrate, MPA is not) [1] and systemic bioavailability of MPA (reviewed in [3]), the ABCG2 c.421 SNP might (as well) reflect on the systemic exposure to MPA. Current “failures” in this respect might be due to methodological study characteristics, and/or could indicate that even if it existed, the effect was mild-moderate, i.e., not robust enough as to be spotted under certain methodological circumstances.
Figures S1A and S1B schematically represent the setting in which the effect of variant c.421C>A allele was to be assessed and measures undertaken to control for confounding. We followed the concepts developed and presented by Pearl [5] and further elaborated by VanderWeele [6, 7], and implemented in R package dagitty[8]. Major elements are depicted in Figure S1A (some are based on explicit in vitro and/or in vivo evidence, some are implied based on circumstantial evidence and are partly hypothetical): 1. For simplicity, in the main text ABCG2 c.421C>ASNP is designated as a binary treatment (variant allele=1, treated; wt homozygous=0, control). Its immediate consequence is reduced number (and activity) of ABCG2 (increased degradation). Hence, the actual treatment (or exposure) is ABCG2 activity (may be dichotomized as “reduced” [variant allele]=treatment; and “preserved” [wt homozygous]= control). However, in vivoABCG2 activity cannot be measured, hence the true exposure remains unobserved, and we use ABCG2 c.421 genotype as an instrumental variable – it has no other effect on the outcome or on any other variable (in vitro , both CsA and tacrolimus are potent ABCG2 inhibitors [9], but neither is an ABCG2 substrate [10, 11]) beyond that conveyed through reduced transporter function; 2. Steady-state MPA pharmacokinetic indicators (MPA PK) are continuous outcomes; 3. The connection between the exposure (ABCG2 activity,c.421 SNP) and the outcome might be a direct one (assumes that MPA is an ABCG2 substrate), and/or an indirect one (as suggested in the Japanese study [4]) with MPAG as a hypothetical mediator (unmeasured in the present study). The setting is such that it can estimate the total effect of exposure (instrument) on the outcome.