There is increased interest in studying the effects of medication adherence on health outcomes. However, if patients appropriately stop treatment because of side effects and treatment failure, it is neither possible nor clinically meaningful to estimate the effect of full medication adherence.
We present an analysis designed to estimate the effect of nonmedical (preventable) discontinuation of cinacalcet, an oral medication approved to treat secondary hyperparathyroidism in patients with end-stage renal disease on dialysis on mortality and heart failure. The approach involves artificially censoring patients who discontinue treatment for a reason that does not appear to be related to an adverse effect of treatment. We address potential bias from informative censoring through inverse-probability of censoring weighted estimation.
Although the analysis is subject to possible residual confounding by the healthy adherer effect and other limitations, we find that potentially preventable discontinuation associates with 2.9 excess deaths at 1 year per 100 patients treated (95% confidence interval, 2.4, 3.5), and 4.6 excess deaths at 2 years (95% confidence interval, 3.5, 5.5). The association between cinacalcet persistence and heart failure hospitalization risk was sensitive to the outcome definition.
Inverse-probability of censoring weighted estimation can be used to estimate the effect of potentially preventable treatment discontinuation in populations where treatment can be stopped for both medical and nonmedical reasons. Estimates from such approaches may represent an upper bound of what would be achievable by an adherence improvement intervention.
From the aUniversity of North Carolina at Chapel Hill, Chapel Hill, NC; and bAmgen, Inc., Thousand Oaks, CA.
Submitted June 14, 2016; accepted September 28, 2017.
Disclosure: This study was supported by a research contract with Amgen, Inc, Thousand Oaks, CA. Before submission for peer review, the article was reviewed by the sponsor. Comments were sent to the authors, who are solely responsible for the final version. The analysis, interpretation, and reporting of these data are the responsibility of the authors. M.A.B. has received investigator-initiated research funding from the National Institutes of Health and through contracts with the Agency for Healthcare Research and Quality’s DEcIDE program and the Patient Centered Outcomes Research Institute. Within the past three years, he has received research support from Amgen and AstraZeneca and has served as a scientific advisor for RxAnte, Amgen, Merck, GSK, and UCB. Merck, GSK, and UCB provide financial support to the UNC Center for Pharmacoepidemiology, which has supported some of Dr. Brookhart’s students. D.R. has received investigator-initiated grant support from Amgen, Inc. A.V.K. has received investigator-initiated grant support from Amgen, Inc. A.V.K. served on a Fresenius Advisory Board. P.D. and B.B. are employees of Amgen, Inc. and are Amgen stockholders. L.W. is a former employee of Amgen and an Amgen stockholder.
Part of the research was based on data from the US Renal Data System. At present, these data can be obtained through an appropriate data use agreement with the National Institute of Diabetes, and Digestive, and Kidney Diseases. The researchers also used clinical data from DaVita, Inc. These data must be obtained through a license with DaVita Clinical Research.
Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com).
Correspondence: M. Alan Brookhart, Department of Epidemiology, Gillings School of Global Public Health, UNC-Chapel Hill, McGavran-Greenberg, CB #7435, Chapel Hill, NC 27599-7435. E-mail: firstname.lastname@example.org.