Background: Prior observational findings based on analyses of Medicaid data from economically disadvantaged Americans suggest that potassium supplementation upon loop diuretic initiation improves survival. We sought to confirm or refute this association among commercially-insured Americans, a generally healthier population with less multimorbidity.
Objectives: To emulate a target trial examining empiric potassium supplementation and all-cause mortality among new users of furosemide
Methods: Within US Optum data, we conducted a retrospective cohort study of furosemide initiators. Exposure was potassium dispensing (yes/no) concurrent with furosemide initiation. The outcome was all-cause mortality. The primary analysis emulated an intention-to-treat (ITT) effect estimate; the secondary analysis emulated an as-treated effect estimate. We used inverse probability of treatment weighting (IPTW) to control for confounding, with weights calculated from high-dimensional propensity scores. Outcomes were modeled with Cox proportional hazards regression. We assessed for effect measure modification and trimmed extreme weights in additional analyses. In the subset of subjects with prescriber information, we performed instrumental variable (IV) analyses using physician preference for empiric potassium as the instrument.
Results: We identified 511,532 and 320,703 new users of furosemide < 40 (low-dose) and ≥40 (high-dose) mg/day with 21.5% and 35.3% being empirically treated with potassium, respectively. Median age was 70.6–72.9 years. Approximately half of users were female (49.6%–60.6%). ITT IPTW-hazard ratios for death with potassium (vs. no potassium) were 1.02 (95%CI 1.01–1.04) and 1.02 (0.999–1.03) among users of furosemide < 40 and ≥40 mg/day, respectively. No association was observed in the trimmed or as-treated analyses. In subgroup analyses, the effect of empiric potassium supplementation varied in those with vs. without baseline heart failure. Adjusted IV analyses suggested a risk increase of 0.9 (0.1–1.8) and a risk reduction of 1.8 (0.8–2.8) deaths/100 subjects in the furosemide < 40 and ≥40mg/day cohorts, respectively.
Conclusions: We observed slightly reduced mortality with empiric potassium among high-dose furosemide users in the IV analysis, but slightly increased mortality among low-dose furosemide users in IPTW and IV analyses. An effect of empiric potassium supplementation on mortality may be minimal; however, given the high prevalence of furosemide use, a large pragmatic trial of empiric potassium supplementation may be warranted.
