In this issue of Blood Advances, a retrospective analysis of patients who underwent allogeneic blood or marrow transplantation (alloBMT) with posttransplant cyclophosphamide (PTCy) by the Center for International Blood and Marrow Research (CIBMTR), as reported by Nath et al,1 found that overall survival (OS) was not significantly different between recipients of older matched sibling donors (MSDs) and younger alternative donors (matched unrelated donors [MUDs], haploidentical related [haplo] donors, and mismatched unrelated donors [mMUDs]).
MSDs have historically been considered the “perfect” donors for alloBMT and prioritized over alternative donors because of lower rates of graft-versus-host disease (GVHD) and improved survival. However, the development and wide implementation of PTCy resulted in improved GVHD control, decreased nonrelapse mortality (NRM), and improved survival with alternative donors.2,3 Moreover, within a specific donor type, MUDs, mMUDs, or haplo donors, younger donors are now consistently found to produce improved outcomes.3-5 Younger donor age is associated with both decreased NRM4,5 and relapse.3 The younger the donor the better: down to age 18 years for MUDs4 and adult-sized adolescents for haplo donors.5 Other donor characteristics such as sex, parity, cytomegalovirus serostatus, and ABO matching appear to have limited effects on outcomes compared to donor age.4,5 Nath et al also found that disease-free survival (DFS) was better with younger MUDs than other donor types, leading the authors to suggest that, in the setting of PTCy, younger MUDs may be the preferred donor choice when available.
Many retrospective studies now show that haplo alloBMT using PTCy produces similar results to MUD and even MSD transplants.2 However, these analyses generally compared haplo donors getting PTCy to matched donors getting calcineurin inhibitor (CNI)–based GVHD prophylaxis. Because PTCy also improves outcomes compared to CNI-based GVHD prophylaxis with matched donors,6 the CIBMTR analyzed outcomes after MUD and haplo alloBMT that both used PTCy.7 They reported that, with reduced intensity conditioning, GVHD, NRM, and DFS were better using MUDs and PTCy.7 Interestingly, there were no significant differences when myeloablative conditioning was used. Notably, there were several prognostic factor imbalances between the haplo and MUDs, including race, era of BMT, and importantly donor age, with the latter not included in the multivariable analysis performed.7 However, a reanalysis of these CIBMTR data using propensity score matching to correct for the imbalances including donor age showed no significant differences in NRM, OS, or DFS.8 There are now several other retrospective analyses comparing matched donor and partially mismatched transplants in the setting of PTCy (see table). Although some (see table; Gooptu et al,7 Mussetti et al, and Nath et al) show improved outcomes with MUDs over haplo donors, others (see table; Sanz et al 2000 and 2001, Ambinder et al,8 and Mehta et al) found similar outcomes. The analysis by Mehta et al actually showed better outcomes for young haplo B-leader–matched donors than older MSDs (see table). Furthermore, Nath et al found no significant differences in outcomes between young haplo and mMUDs, and a recent separate analysis by the CIBMTR (see table, Shaffer et al) showed no difference in outcomes between MUDs and mMUDs when PTCy was used. Thus, these retrospective data suggest any differences in outcomes among donor types are likely to be minor using modern transplant platforms and young donors. Determining the best donor would obviously be best addressed in prospective randomized trials. However, such trials are unlikely to ever occur due to several challenges, such as the need for multiple donor types being available in a timely fashion.
Large retrospective analyses by the CIBMTR and European Society for Blood and Marrow Research, although not substitutes for well-designed prospective randomized trials, have been important tools in the field, especially in areas in which randomized trials are lacking. To leverage the power of these large registries, rigorous statistical methods, such as multivariable regression analyses and propensity score matching, are needed to account for confounding factors associated with retrospective analyses.8 However, these statistical tools can be problematic when applied to groups that are highly imbalanced for important confounders, as is often the case for heterogeneous BMT data sets. Thus, it is not surprising when such retrospective analyses come to somewhat different conclusions even when the data sets are large. This is especially true when outcome differences between the groups are small (see table).
In contrast, donor age has consistently been shown to be the most important donor characteristic in retrospective analyses; the paper by Nath et al is further evidence that young age should now be part of “perfect” donor algorithm. A young donor is almost always available through unrelated registries when including mMUDs. Similarly, most older patients have an available young haplo donor: either a child (100% chance of being haplo), niece, nephew, or grandchild (all 50% chance).
AlloBMT has made enormous progress since the IBMTR’s report at the end of the last millennium showing dismal outcomes for mismatched transplants.9 Even MUDs in that report were associated with NRMs >50%.9 A large prospective clinical trial from the Blood and Marrow Transplant Clinical Trials Network (BMT CTN 1702) recently found no significant differences in OS, NRM, relapse, or GVHD based on donor type.10 The study concluded that patients unlikely to find an MUD should proceed directly to alloBMT using the best available alternative donor, rather than accepting any delay to identify an MUD.10 Thus, all patients in need now have an effective donor option, be it an MSD, MUD, haplo donor, or mMUD. Moreover, most patients will have multiple donor options. A discussion regarding selecting the best donor from among several available donor types was likely unimaginable by most in the field even a few years ago. Considerations such as timing, graft source (eg, marrow is generally preferred over peripheral blood for nonmalignant disease), and possibly specific genetic considerations (see table, Mehti et al) may ultimately decide the best young donor to choose. Accordingly, a “one-size-fits-all” approach for selecting a donor should probably be replaced by personalized donor selection, because the perfect HLA match no longer appears to represent the ideal donor.
Conflict-of-interest disclosure: R.J.J. declares no competing financial interests.