Background/Purpose: Prior observational studies suggest that SLE patients receiving hydroxychloroquine (HCQ) may have reduced risk of infections, cardiovascular disease and end-stage renal disease (ESRD). However, adherence to HCQ was not assessed. HCQ is the standard of care for SLE, and therefore randomized trials comparing outcomes among users vs. nonusers are not feasible. We implemented inverse probability weighting within a marginal structural modeling framework to assess whether SLE patients who adhered to HCQ (vs. nonadherers) had reduced risks of disease-related adverse outcomes, adjusting for time-varying confounding and healthy adherer effects.

Methods: We identified Medicaid beneficiaries (2000-2010) from 28 states with SLE, defined by ≥2 SLE ICD-9 codes, who initiated HCQ (index date) and had no use in the prior 6 months (baseline period). We examined 3 outcomes: ESRD, serious infections requiring hospitalization, and myocardial infarction/cerebrovascular accident (MI/CVA), excluding these outcomes during the baseline period. For our primary analyses, we updated monthly adherence (80% of days/month covered = adherent), and used inverse probability weights to account for prior adherence, time-varying confounders (comorbidities, medications), and censoring (death, disenrollment, retinal toxicity). We constructed marginal structural pooled logistic models for each outcome. In secondary analyses, we adjusted only for baseline, time-fixed covariates using Fine and Gray proportional hazards models accounting for the competing risk of death, to examine the association between adherence 90-days post index date (proportion of days covered (PDC) ≥80% = adherent), and outcomes after this period. We also assessed the relationship between adherence and motor vehicle accidents (MVAs), a control outcome; no association would support potential biologic effects in our main models.  

Results: We identified >20,000 HCQ initiators with SLE with mean (± SD) follow-up of >4 (± 3) years. For all cohorts, the mean (± SD) PDC during the 90-days post index date was 62% (± 26); 31% had PDCs≥80%. In our marginal structural models, comparing adherers vs. nonadherers, we observed reduced risks of ESRD (OR 0.73, 95% CI 0.61-0.89) and serious infections (OR 0.84, 95% CI 0.79-0.90), and borderline reduced risk of MI/CVA (OR 0.92 95% CI 0.85-1.00) (Table). There was no association between adherence and MVAs, our negative control (OR 0.95 95% CI 0.75-1.19). In secondary analyses adjusted only for baseline covariates, results were similar but attenuated for ESRD and serious infections.

Conclusion: We observed modestly reduced risks of ESRD, serious infections and MI/CVA among HCQ adherers vs. nonadherers. Our analyses are limited by lack of data on reasons for HCQ discontinuation and the challenge of distinguishing between the biologic effects of HCQ vs. unmeasurable healthy adherer effects.