Use of Minimal Residual Disease Status to Reduce Uncertainty in Estimating Long-term Survival Outcomes for Newly Diagnosed Multiple Myeloma Patients.

Abstract

Background: Demonstrating the cost-effectiveness of new treatments for multiple myeloma (MM) often relies on the extrapolation of overall survival (OS) trial data. This method can introduce uncertainty in long-term survival estimates if OS data are immature, as is often the case in newly diagnosed MM (NDMM). We explore the use of the relationship between minimal residual disease (MRD) status and OS to reduce uncertainty of long-term survival outcomes. Objectives: To evaluate if uncertainty in long-term modeled outcomes in NDMM is reduced using a response-based partitioned survival model (PSM), whereby patients were categorized as MRD-positive or -negative, relative to a standard PSM, when OS data are immature. Methods: Standard and response-based PSMs, estimating patient life-years (LYs) over a lifetime horizon, were developed for NDMM patients treated with bortezomib, thalidomide, and dexamethasone (BTd) with or without daratumumab as induction and consolidation therapy. In the standard PSM, LYs were determined by extrapolations from individual patient data from CASSIOPEIA. In the response-based PSM, survival was dependent on MRD status at the time of the response assessment via a landmark analysis. Cox-proportional hazard ratios from external sources and CASSIOPEIA informed the relationship for OS between MRD-positive and MRD-negative, and between patients receiving BTd and daratumumab plus BTd, respectively. Uncertainty was assessed by comparing LYs and OS extrapolations from deterministic and probabilistic analyses. Results: This response-based PSM demonstrated reduced uncertainty in long-term survival outcomes compared with the standard PSM (range across extrapolations of 3.4 and 7.7 LYs for daratumumab plus BTd and BTd, respectively, vs 14.8 and 11.8 LYs for the standard PSM). It also estimated a narrower interquartile range of LYs in the probabilistic analyses for the majority of parametric extrapolations. Discussion: Alternative methods to estimate long-term survival outcomes, such as a response-based PSM, can reduce uncertainty in modeling predictions around cost-effectiveness estimates for health technology assessment bodies and payers, thereby supporting faster market access for novel therapies with immature survival data. Conclusions: Use of MRD status in a response-based PSM reduces uncertainty in modeling long-term survival in patients with NDMM and provides a greater number of clinically plausible extrapolations compared with a standard PSM.