Evaluating treatments in rare indications warrants a Bayesian approach

Evaluating efficacy and real-world effectiveness for novel therapies targeting rare mutations or patient subpopulations with unmet needs is a growing challenge in health economics and outcomes research (HEOR). In these settings it may be difficult to recruit enough patients to run adequately powered randomized clinical trials, resulting in greater reliance on single-arm trials or basket trial designs. Additionally, evidence networks for performing network meta-analysis may be sparse or disconnected when comparing available treatments in narrower patient populations. These challenges create an increased need for use of appropriate methods for handling small sample sizes, structural modelling assumptions and more nuanced decision rules to arrive at "best-available evidence" on comparative and non-comparative efficacy/effectiveness. We advocate for greater use of Bayesian methods to address these challenges as they can facilitate efficient and transparent borrowing of information across varied data sources under flexible modelling assumptions, probabilistic sensitivity analysis to assess model assumptions, and more nuanced decision-making where limited power reduces the utility of classical frequentist hypothesis testing. We illustrate how Bayesian methods have been recently used to overcome several challenges of rare indications in HEOR, including approaches to borrowing information from external data sources, evaluation of efficacy in basket trials, and incorporating non-randomized studies into network meta-analysis. Lastly, we provide several recommendations for HEOR practitioners on appropriate use of Bayesian methods to address challenges in the rare disease setting.

Indirect Treatment Comparison of Daratumumab, Pomalidomide, and Dexamethasone Versus Standard of Care in Patients with Difficult-to-Treat Relapsed/Refractory Multiple Myeloma

INTRODUCTION

The phase 3 APOLLO study demonstrated significantly better progression-free survival (PFS) and clinical responses with daratumumab, pomalidomide, and dexamethasone (D-Pd) versus pomalidomide and dexamethasone (Pd) in patients with relapsed/refractory multiple myeloma (RRMM). On the basis of these results and those from the phase 1b EQUULEUS trial, D-Pd was approved in this patient population. In the absence of head-to-head data comparing D-Pd with further standard of care (SOC) therapies, indirect treatment comparisons (ITCs) can provide important information to help optimize treatment selection. The objective of this study was to indirectly compare PFS improvement with D-Pd versus daratumumab, bortezomib, and dexamethasone (D-Vd) and D-Pd versus bortezomib and dexamethasone (Vd) in patients with RRMM.

METHODS

Patient-level data were from APOLLO, EQUULEUS, and CASTOR. Three methods of adjusting imbalances in baseline characteristics including stabilized inverse probability of treatment weighting (sIPTW), cardinality matching (CM), and propensity score matching (PSM) were initially considered. CM offers mathematically guaranteed largest matched sample meeting pre-specified maximum standardized mean difference criteria for matching covariates. sIPTW and PSM were based on propensity scores derived from logistic regression. Feasibility assessment of the PSM method returned too low effective sample size to support a meaningful comparison. CM was chosen as the base case and sIPTW as a sensitivity analysis.

RESULTS

After harmonized eligibility criteria were applied, 253, 104, and 122 patients from the D-Pd, D-Vd, and Vd cohorts, respectively, were included in the ITC analyses. Some imbalances in baseline characteristics were identified between D-Pd and D-Vd/Vd cohorts that remained after adjustment. PFS hazard ratios showed significant improvement for D-Pd over D-Vd and Vd for CM and sIPTW analyses.

CONCLUSIONS

Results showed consistent PFS benefit for D-Pd versus D-Vd and Vd regardless of the adjustment technique used. These findings support the use of D-Pd versus D-Vd or Vd in patients with difficult-to-treat RRMM.

TRIAL REGISTRATION

NCT03180736; NCT02136134, NCT01998971.