Abstract
Quantitative benefit-risk assessment (qBRA) can reveal how patients balance benefits and risks of cancer treatments. To align with qBRA good practice guidelines, researchers must address challenges including attribute value dependence, double counting, attribute dominance, and uncertainty associated with immature clinical trial outcomes. We present a case study illustrating these challenges in a qBRA of treatment preferences among patients with Philadelphia chromosome-positive acute lymphoblastic leukemia. Preferences were elicited using a discrete choice experiment (DCE). First, we explain how we mitigated potential dominance of survival outcomes by narrowing the range of overall survival (OS) durations that each participant considered. Second, we describe how we acknowledged the conceptual interaction between OS and duration of remission (DOR) attributes and tested for a statistical interaction. Third, we detail how we conducted qBRA with uncertain efficacy data using bivariate sensitivity analysis. Bivariate sensitivity analysis based on DCE-elicited preferences and head-to-head clinical performance data showed that if the considered treatments - ponatinib + chemotherapy and imatinib + chemotherapy - had equivalent efficacy, 52.9% (95% CI: 52.5%-53.4%) of DCE participants would be expected to choose ponatinib over imatinib. If ponatinib offered 10-month longer DOR and 20-month longer OS vs. imatinib, 71.6% (95% CI: 67.2%-76.0%) would choose ponatinib. Probabilistic sensitivity analyses showed that the probability of ≥ 70% of patients preferring ponatinib is 77.5% if ponatinib offers 15-month longer OS and DOR and 93.0% if it offers 45-month longer OS and DOR. Preference heterogeneity analyses identified that the overall choice probability results hold for all subgroups in nearly all scenarios.