Umbilical Cord Blood or HLA-Haploidentical Transplantation: Real-World Outcomes versus Randomized Trial Outcomes

Randomized clinical trials offer the highest-quality data for modifying clinical practice. Results of a phase III randomized trial of nonmyeloablative transplantation for adults with high-risk hematologic malignancies with 2 umbilical cord blood (UCB) units (n = 183) or HLA-haploidentical relative bone marrow (Haplo-BM; n = 154) revealed a 2-year progression-free survival (PFS) of 41% after Haplo-BM transplantation and 35% after 2-unit UCB transplantation (P = .41), with overall survival (OS) of 57% and 46%, respectively (P = .04). We sought to examine the generalizability of BMT CTN 1101 to a contemporaneous cohort beyond the trial's prespecified 2-year outcomes. All transplantations were performed between June 2012 and June 2018 in the United States. We hypothesized that the results of a rigorous phase III randomized trial would be generalizable. Changes in graft selection for HLA-haploidentical relative transplantation during the trial period allowed comparison of outcomes after transplantation with Haplo-BM with those after haploidentical peripheral blood (Haplo-PB). The trial's broad eligibility criteria were applied to the data source of the Center for International Blood and Marrow Transplant Research to select nontrial subjects. Extended follow-up of trial subjects was obtained from this data source. Three separate analyses were performed: (1) trial subjects beyond the trial's 2-year endpoint; (2) comparison of trial subjects with a contemporaneous cohort of nontrial subjects (195 2-unit UCB, 358 Haplo-BM, and 403 Haplo-PB); and (3) comparison of nontrial subjects by donor and graft type. Multivariate analyses were performed using Cox proportional hazards models for comparison of outcomes by treatment groups. With longer follow-up of the trial cohorts, 5-year PFS (37% versus 29%; P = .08) and OS (42% versus 36%; P = .06) were not significantly different between the treatment groups. We then compared the trial results with outcomes of comparable real-world transplantations. Five-year OS did not differ between trial and nontrial 2-unit UCB transplantations (36% versus 41%; P = .48) or between trial and nontrial Haplo-BM transplantations (42% versus 47%; P = .80), confirming generalizability. The randomized trial did not accrue as planned and therefore lacked the statistical power to detect a 15% difference in PFS. With substantially larger numbers of nontrial Haplo-BM transplantations, 5-year survival was higher after nontrial Haplo-BM compared with trial 2-unit UCB (47% versus 36%; P = .012). Nontrial patients who underwent Haplo-PB transplantation had higher 5-year survival (54%) compared with trial Haplo-BM (hazard ratio [HR], 0.76; P = .044) and nontrial Haplo-BM (HR, 0.78; P = .026). Similarly, survival was better after Haplo-PB compared with trial UCB (HR, 0.57; P < .0001) and nontrial UCB (HR, 0.63; P = .0002). When considering alternative donor low-intensity conditioning regimen transplantation, a haploidentical relative is preferred, and PB is the preferred graft source.