OBJECTIVES: For rare diseases like acute lymphoblastic leukemia (ALL), trials frequently present limitations including small samples and single arm design. Markov cohort models (MCM) can include intermediate endpoints affecting survival but are data intensive. In this context, we propose to assess the value of using an MCM instead of a partitioned survival model (PSM) that only relies on relapse-free survival (RFS) and overall survival (OS) in evaluating cost effectiveness of blinatumomab versus chemotherapy.

METHODS: In the PSM, RFS and OS were based on parametric distributions fit to patient failure-time data from the BLAST trial and a historical comparator study. To validate PSM findings, an MCM, which explicitly estimates the contribution of minimal residual disease (MRD) and hematopoietic stem-cell transplant (HSCT) on survival, was developed. Transition probabilities for HSCT, relapse, and death, were estimated using the same data. We evaluated the sensitivity of both models around parameters and assumptions.

RESULTS: Both models led to similar incremental cost-effectiveness ratios for blinatumomab versus chemotherapy: $102,016/QALY in the PSM and $118,659/QALY in the MCM. Incremental costs (PSM: $261,876; MCM: $242,940) and QALYs (PSM: 2.57; MCM: 2.05) were comparable. The lower QALY difference in the MCM (versus the PSM) is due to lower estimations of post-relapse survival for blinatumomab. The lower incremental cost in the MCM is due to the explicit allocation of MRD-related costs, leading to higher cost offsets. Probabilistic sensitivity analyses showed little variability due to parameter uncertainty in both models.

CONCLUSIONS: Using both models, blinatumomab is cost effective versus chemotherapy in ALL patients with MRD for US healthcare payers. The PSM structure cannot explicitly account for treatment effects on survival mediated by MRD response or HSCT. However, the MCM requires more assumptions due to limited data for estimating transition probabilities. When data is limited, a PSM is recommended as the primary approach.