Indirect Comparisons to Assess the Relative Efficacy of Carfilzomib + Lenalidomide + Dexamethasone Versus Panobinostat + Bortezomib + Dexamethasone and Bortezomib + Dexamethasone: A Matching Adjusted Indirect Comparison

Background: Several novel treatments have recently been approved for the treatment of relapsed multiple myeloma (RMM). In the absence of head-to-head comparisons between these novel treatments, clinicians and payers must rely on statistical indirect comparisons. The objective of this analysis is to derive measures of relative effectiveness for carfilzomib + lenalidomide + dexamethasone (KRd) against bortezomib + dexamethasone (Vd), and the recently approved combination of panobinostat + bortezomib + dexamethasone (PVd) in patients with RMM who have been treated with at least one prior therapy.

Methods: A matching-adjusted indirect comparison (MAIC) (Signorovitch, 2010) for progression-free survival (PFS) and overall survival (OS) was conducted between the KRd arm of the phase III study ASPIRE (Stewart et al., 2015) versus the PVd and Vd arms of the phase III study PANORAMA 1 (San-Miguel et al., 2014). The MAIC utilized patient level data from ASPIRE, and adjusted for reported patient population differences. An MAIC uses a propensity score type equation to assign case weights to the KRd patients so that their weighted baseline characteristics match the baseline of the PVd or Vd population. This re-weighting process attempts to answer the question: What would the outcomes be if KRd had been administered to a population matching the characteristics of the PVd or Vd arms? Adjustments were made for age, gender, ECOG status, history of autologous stem cell transplant, disease duration, number of prior regimens, ISS stage, prior bortezomib use, and creatinine clearance rate. Cox PH models were fitted to estimate hazard ratios (HRs) for PFS and OS. Weibull survival curves best fit the adjusted survival data and were used to estimate median survival times. A simulated treatment comparison (STC) (Ishak et al., 2015), which adjusts for reported patient population differences using regression equations, was conducted as a cross validation.

Results: The KRd arm in ASPIRE included 396 patients and the PVd and Vd arms in PANORAMA 1 included 387 and 381 patients, respectively. After successfully matching, the effective sample size of the KRd population was 131 for the PVd comparison and 138 for the Vd comparison. See Figure 1 for the MAIC adjusted PFS and OS Kaplan-Meier curves. Hazard ratios (95% CIs) from the Cox models for PFS and OS outcomes were 0.317 (0.228, 0.44) and 0.582 (0.394, 0.86) for KRd vs PVd, respectively and 0.208 (0.153, 0.283) and 0.472 (0.324, 0.688) for KRd vs Vd, respectively. Corresponding hazard ratios from the STC were similar and validate the MAIC results.

Estimates of median PFS and OS in months were 29.5 and 65.2 for KRd compared to 12.0 and 40.9 for PVd, respectively. Corresponding estimates were 29.7 and 57.3 for KRd compared to 8.2 and 33.0 for Vd.

Figure 1.

Conclusion: This MAIC analysis suggests that KRd provides a consistent and statistically significant PFS and OS benefit relative to PVd and Vd in RMM patients who have been treated with at least one prior therapy. Beyond the patient characteristics available from PANORAMA 1, other variables that may potentially influence outcomes were not adjusted for in the analysis. This analysis did not compare KRd to PVd in patients who have received at least 2 prior regimens including bortezmib and an IMiD (panobinostat's FDA-approved indication) due to lack of published data on the baseline characteristics of this patient subset studied in PANORAMA 1.