Investigating the effect of Gleevec list price on adherence and outcomes in Medicare patients with chronic myeloid leukemia.

Abstract

Gleevec is a specialty drug that revolutionized the treatment of chronic myeloid leukemia (CML). Its relatively early introduction and high list price have made it emblematic of drug pricing concerns in the US. In Medicare Part D, out-of-pocket payments (OOP) in the catastrophic coverage phase are co-insurance based and a direct function of list price. The financial impact of this benefit structure on CML patients taking Gleevec may have significant downstream consequences in the form of reduced adherence and suboptimal health outcomes. This dissertation comprises two aims focused on better understanding the complex relationship between list-price-based OOP, adherence, and outcomes in newly-diagnosed CML patients on Gleevec who are enrolled in Medicare Part D. In the first aim, a fuzzy donut regression discontinuity design was used to estimate the causal effect of Medicare Part D enrollment at age 65 on adherence and OOP. Results were then combined to calculate the corresponding price elasticity of demand (PED) and determine the degree of price sensitivity in these patients with respect to Gleevec OOP. In the second aim, we applied the alternating conditional estimation (ACE) algorithm alongside a backward elimination procedure to identify time dependence and nonlinearity in the relationship between adherence and time-to-remission. An extended Cox model allowing for the flexible modeling of each effect via unpenalized B-splines was subsequently fit and assessed. Results from aim 1 indicate that, while there is a large and significant increase in initial OOP of $232 for patients on Part D at diagnosis, the effect on adherence is a smaller and non-significant decrease of 6 percentage points. The corresponding PED of -0.02 indicates that the demand for Gleevec with respect to OOP is highly inelastic for these patients. Aim 2 provides insight into the shape of the dose-response curve (DRC) for Gleevec and demonstrates how the strength of this curve varies over the first year following treatment initiation. Results suggest that the DRC is non-linearly monotonically increasing and that the strength of this effect on time-to-remission is strongest in the first three to four months of treatment. This study provides evidence regarding potential drivers of the substantial non-adherence observed in CML patients, as their minimal responsiveness to OOP indicates that other factors may be playing a more important role. The finding that cumulative adherence is most important early on can potentially inform monitoring and intervention strategies for meeting treatment milestones. Results from both aims can additionally be used to help guide the design of future studies regarding important analytic decisions and assumptions.