Figure S1B . The same DAG as in figure S1A, but with depicted outcome/exposure ancestors controlled for by different means depicted as open, black-outlined circles - minimal sufficient adjustment set. Black circle is the instrument, gray/black-outlined circles depict unmeasured exposure defined by the instrument (ABCG2 activity) and a hypothetical mediator (MPAG levels) on a causal path - direct (thick full black arrow) and/or indirect (dashed black arrow) – between the instrument (unmeasured exposure) and the outcome. Completely gray circles are unmeasured variables – enzyme/transporter activities that are (presumable) mediators on some of the biasing paths (descendants of outcome/exposure ancestors). They did not need to be adjusted for, since their ancestors were included in the adjustment set. One suggested confounder and one ancestor of the outcome (based on some of the literature data) remained unmeasured – the SLCO1B3 c.334 SNP andUGT1A9 c.98 SNP. Two green circles depict ancestors of exposure that did not need to be in the adjustment set, since their descendant – CNI concentration – is adjusted for (see text for details).
xiii) Additionally – estimated creatinine clearance – by matching and statistical adjustment; xiv) Hypoalbuminemia – by inclusion criteria: patients had to have serum albumin >31 g/L; xv) Medical conditions that may interfere with bioavailability of both MPA and CNIs – by inclusion criteria: patients had to be generally “well-doing” (free of surgical complications, infections, relevant metabolic, gastrointestional, hepatic or cardiovascular conditions); xvi) Drugs interfering with MPA pharmacokinetics – by inclusion criteria: patients had to be free of treatment known to affect MPA during the pre-study days and on the study day (antacids, phosphate binders, proton pump inhibitors, antibiotics, rifampicin, oral iron, calcium or magnesium; also, co-medication was restricted by exclusion based on comorbidities); xvii) Drugs interfering with CNIs – by inclusion/exclusion criteria: no specific list was formed, but co-medication was restricted already regarding MPA, and comorbidities; also, CNI concentrations were accounted for in matching; xviii) Drugs interfering with ABCG2 activity (apart from CNIs) – by inclusion/exclusion criteria: no specific list was formed, but considering the known ABCG2 substrates and inhibitors [11], relevant co-medication was practically completely restricted already regarding MPA and comorbidities.
Variables that did not need to be included in the adjustment set (depicted as green circles in Figure S1B): i) CNI dose – as it is represented by its descendant, CNI concentrations; ii) The same applies to the two genotyped CYP3A4/5 SNPs. CYPs are of marginal relevance for MPA clearance, hence the two genotyped CYPs and other possible SNPs affecting CYP3A4/5 or other CYPs and/or transporters are of relevance primarily for their effect on CNI concentrations [1, 10] – and are thus controlled for by adjustment for CNI concentrations. By the same logic, drugs affecting CNI concentrations (by any means) did not necessarily need to be included in the adjustment set, but by comorbidity and co-medication restrictions pertinent to MPA, they were, to a great extent. As reviewed [3], in vitro data and a few studies in humans (vs. several “negative” studies) suggest thatSLCO1B3 c.344 SNP and UGT1A9 c.98 SNP may be relevant for exposure to MPA – in the present study, these SNP was not genotyped, i.e., remained “unobserved” (indicated as gray circles [for unobserved/unmeasured variables] in Figure S1B.