A Matching-Adjusted Indirect Comparison of Nivolumab Versus Brentuximab Vedotin for Relapsed/Refractory Classical Hodgkin Lymphoma after Failure of Autologous Hematopoietic Cell Transplantation

Introduction: Historically, patients with relapsed/refractory classical Hodgkin lymphoma (R/R cHL) who relapse after autologous hematopoietic cell transplantation (auto-HCT) have poor outcomes. In the phase 2 CheckMate 205 study (NCT02181738) of patients with R/R cHL and prior auto-HCT, patients in Cohort A were brentuximab vedotin (BV) naive, and those in Cohorts B and C had prior BV exposure. Nivolumab (nivo), an anti-PD-1 immune checkpoint inhibitor monoclonal antibody, was associated with a high response rate (71% Cohorts A+B+C; 65% Cohort A) and durable remissions (median duration of response 18 and 25 months, respectively). Notably, the 2-year overall survival (OS) rates were 87% in Cohorts A+B+C and 90% in Cohort A; the median OS was not reached (median follow-up 33 months; Armand et al. ASH 2018). These results appear better than those in prior studies in this patient population. In the phase 2 pivotal trial of BV (NCT00848926), patients with R/R cHL had a 2-year OS rate of approximately 65% (Chen et al. Blood 2016). Without direct head-to-head randomized trials, cross-trial comparisons have limitations primarily due to differences in patient populations and trial design, and the survival benefits of various treatments are difficult to distinguish. We used matching-adjusted indirect comparison (MAIC) to balance patient populations and then assess the efficacy and survival benefit of nivo relative to BV in patients with R/R cHL for whom auto-HCT had failed. Methods: Individual patient data (IPD) from patients receiving nivo in the CheckMate 205 study were matched to summary data from patients in the BV pivotal trial reported by Chen et al. IPD from Cohort A and combined Cohorts A+B+C of the CheckMate 205 study were re-weighted to match relevant baseline characteristics reported for BV. Pseudo-IPD were generated for BV from KM curves applying a published algorithm. Treatment outcomes (progression-free survival [PFS] and OS) were then compared across balanced trial populations and assessed using hazard ratios (HRs) generated via Cox regression and differences in the area under the curve (ΔAUC) from best-fitting parametric curves. For nivo, 2 well-fitting OS curves were selected for comparison, which conveyed optimistic and conservative assumptions. Restricted AUC at a specific time point is identical to mean survival time at that time point, thus ΔAUC was used as a summary metric to compare survival time between groups. Confidence intervals (CIs) for