Validation of Cure Assumptions when Analyzing ZUMA-7 Follow-up Data of Axicabtagene Ciloleucel and Standard of Care Therapy in Second-Line Relapsed/Refractory Large B-cell Lymphoma.

backgroundZUMA-7 is the largest randomized controlled trial (RCT) for chimeric antigen receptor (CAR) T-cell therapy, which compared axicabtagene ciloleucel (axi-cel) to historical standard of care (HSoC) for the treatment of relapsed or refractory diffuse large B-cell lymphoma (R/R DLBCL). Both arms of ZUMA-7 contained potentially curative treatments; however, differences in the treatment completion rate and timing to receive definitive treatment led to differences in the extent and timing of cure. Mixture cure modeling (MCM) has been suggested as a superior method in designing and powering clinical studies of curative therapies but also for extrapolation of long-term outcomes in simulation studies. The aim of this study was to evaluate the appropriateness of MCM in second line (2L) therapy for DLBCL by analyzing extrapolations from other cost-effectiveness analyses; to use data cuts of ZUMA-7 trial to validate MCMs and describe differences that may have arisen due to treatment specific timing and extent of cure.

methodsA targeted literature review (TLR) was conducted to identify cost-effectiveness modeling studies that have used ZUMA-7 clinical trial data to extrapolate survival of axi-cel over the long-term. Overall survival (OS) individual patient-level data from the ZUMA-7 trial for axi-cel and historical HSoC from both primary event-free survival (EFS; median follow-up: 24.9 months) and primary OS cutoff (median follow-up: 47.2 months) were used in this analysis. Three intermediate datasets were simulated to compare differences in cure fractions and accuracy of extrapolation predictions among data cutoffs. Differences between trial arms were evaluated.

resultsThe TLR demonstrated the extrapolation using ZUMA-7 OS data cutoff provided lower difference margins with respect to long-term observed data, than external extrapolations. Within the models tested, MCM provided the best fit based on the percentage inaccuracy of extrapolations which ranged 2%-27% in the axi-cel group. Based on observed primary EFS cutoff, cure fractions ranged in both treatment arms (axi-cel: 24%-54% and HSoC: 35%-49%). The range of cure fractions narrowed following extrapolations performed on the more mature observed OS cutoff (axi-cel: 50%-54% and HSoC: 41%-50%).

conclusionReplicating prior findings, MCM methods provide the more accurate estimate of long-term OS compared with standard parametric modeling for patients with R/R DLBCL treated with CAR T-cell therapy. However, these findings also highlight the importance of obtaining appropriate clinical benchmarking to ensure clinical plausibility and face validity of long-term extrapolations intended to informed cost-effectiveness analysis and inform decision making.