Haploidentical Versus HLA-matched Unrelated Donors for Hematopoietic Cell Transplantation in Acute Leukemia: A Multicenter Observational Study in a Developing Country Facilitated by Collaboration Between the CIBMTR and the Brazilian Society of Cellular Therapy and Bone Marrow Transplantation (SBTMO)

Dear Editor,

Hematopoietic cell transplantation (HCT) continues to expand worldwide, with roughly equal contributions from matched-related donors (MRD), matched unrelated donors (MUD), and haploidentical (Haplo) donors, but the latter has markedly increased in recent years, becoming a key alternative [1].

Latin America presents unique epidemiologic and genetic features compared to Europe and North America [2]. Despite the Brazilian donor registry (REDOME) being the world’s third-largest, limitations often hinder timely access to matched unrelated donors, making Haplo HCT a practical and cost-effective strategy [3]. A previous retrospective multicenter study in Brazil (2014 to 2018) compared HCT outcomes in acute leukemias with MSD, MUD, and Haplo [4]; however, Haplo HCT had not yet been widely adopted in Brazil, and local experience was limited. Our study compared the results of Haplo with those of MUD performed in Brazil between 2018 and 2021.

This study was conducted in collaboration with the Sociedade Brasileira de Transplante de Medula Ossea e Terapia Celular (SBTMO, the Brazilian Society for Bone Marrow Transplantation) and supported by the Center for International Blood and Marrow Transplant Research (CIBMTR). PTCy was mandatory for the haploidentical transplantation. ATG was allowed (the only brand available in Brazil is Thymoglobulin). All patients had acute myeloid or lymphoblastic leukemia in complete remission and were reported to the CIBMTR, which provided the curated database. The primary objective was to compare overall survival (OS) between MUD and Haplo. We hypothesized that, during this more contemporary period, outcomes after Haplo HCT would approach those with MUD HCT.

Survival curves were generated using the Kaplan–Meier method, and cumulative incidence functions were estimated using the Gray method. Comparisons were performed with the log-rank and Gray tests. Multivariable analyses (MVA) were performed using Cox proportional hazards models. Model selection was based on the lowest Akaike Information Criterion (AIC). Variables included were donor type (in all models), age, Karnofsky Performance Status (KPS), disease, disease status, donor and recipient sex, graft source, and conditioning regimen. Confounders were identified as variables causing at least a 10% change in the donor-type coefcient when added to the model. The following outcomes were analyzed: overall survival (OS, which was the primary outcome), relapse-free survival (RFS), GVHD-, relapse-free survival (GRFS), relapse (REL), nonrelapse mortality (NRM), grades II to IV and III to IV acute GVHD (II to IV aGVHD and III to IV aGVHD), and chronic GVHD (cGVHD). We considered P-values <.05 significant, except for the stratified analyses (CR1 and CR2+), for which we applied the Bonferroni correction (P-value <.025). Analyses were conducted using R, version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria).

Of 480 patients were included. CMV-negative donors, bone marrow grafts, and a myeloablative conditioning regimen were more frequent in the MUD group. Time from diagnosis to transplantation was longer for MUD recipients, particularly among CR1 patients. Other characteristics were well-balanced (Table S1).

With a median follow-up of 29 months, 2-year OS was 61% for Haplo and 66% for MUD (P = .068, Figure 1). In MVA, Haplo was not associated with inferior OS compared with MUD (HR = 1.31, 95% CI 0.95 to 1.79, P = .10, Figure S1).

Figure 1.

Overall and relapse-free survival.

Of 2-year RFS were 57% for Haplo and 62% for MUD (P = .16; in multivariable, HR = 1.21, 95% CI 0.88 to 1.67, P = .24; Figures 1 and S1). GRFS did not differ (44% for Haplo versus 47% for MUD, P = .57; in MVA, HR = 1.14, 95% CI 0.90 to 1.50, P = .36). There was a trend towards better OS and RFS in patients in CR1 with MUD in univariable analyses (P = .046 and P = 032, respectively), but not confirmed in MVA (OS: HR = 1.50, P = .08, and RFS: HR = 1.53, P = .07, for Haplo compared with MUD). The 2-year relapse incidence was 19% for Haplo and 14% for MUD (P = .080; multivariable HR = 1.52, 95% CI 0.90 to 2.56; P = .11). The 2-year incidence of NRM was 24% for both Haplo and MUD (P = .89), as confirmed in MVA (HR = 0.94, 95% CI: 0.62 to 1.43; P = .79; Figures S1 and S2).

At 6 months, II to IV aGVHD occurred in 31% of Haplo and 34% of MUD (P = .44), while III to IV aGVHD, in 8% and 13%, respectively (P = .048). In MVA, Haplo was not associated with a significant difference in aGVHD risk (HR = 0.86, 95% CI: 0.62 to 1.19; P = .37, for II to IV aGVHD and HR = 0.59, 95% CI: 0.32 to 1.06, P = .08, for III to IV aGVHD, Figures S1 and S2). The 2-year cGVHD was 15% for Haplo and 16% for MUD (P = .83; MVA, HR=1.13, 95% CI: 0.69 to 1.83; P = .63; Figures S1 and S2).

Our results demonstrate comparable outcomes between Haplo and MUD HCT in a developing country that had continuously invested in an unrelated donor registry and rapidly incorporated PTCy-based haploidentical HCT.

Haplo was not inferior to MUD with respect to OS, RFS, or GRFS. Relapse rates did not differ significantly after adjusting for time to transplantation. Among CR1 patients, we could not rule out better OS and RFS with MUD. Gooptu et al. [5], in a homogeneous cohort, previously reported small differences in outcomes favoring the PTCy-based MUD over the PTCy-based haploidentical HCT. However, these differences may reflect residual confounding rather than biological disparity, and prior studies have reported comparable outcomes. Our findings in the CR1 should be interpreted with caution given the nature of this analysis and the borderline P-values and could be explained by the shorter time to transplantation among Haplo in this subgroup, indicating either preferential referral of unmeasured higher-risk patients to Haplo, aiming for a faster HCT, or loss of these higher-risk patients referred to MUD due to relapse before the inherent bureaucratic procedures for MUD are finished.

Both groups exhibited similar GVHD risks. Earlier reports indicated that haploidentical HCT with PTCy results in a lower incidence of chronic GVHD than MSD or MUD using conventional prophylaxis. However, in those studies, the prevalence of peripheral blood stem cell (PBSC) grafts is usually substantially lower in the Haplo group, which partially explains the findings. Our study showed a higher frequency of PBSC grafts in the Haplo group, but 97% of MUD patients received ATG- or PTCy-based GVHD prophylaxis, suggesting that PTCy-based GVHD prophylaxis in Haplo HCT offers no additional clinical advantage over ATG-based regimens in MUD HCT.

Compared with the earlier Brazilian retrospective study conducted during the initial implementation of Haplo HCT [4], the earlier study suggested a higher relapse risk with Haplo HCT in CR2+ patients, but the current study, conducted in later years, did not confirm this, suggesting that the former study’s findings may have been due to reduced-intensity conditioning or survival bias, emphasizing the need to minimize biases when comparing donor sources, especially in resource-limited settings where access to unrelated grafts may delay transplantation. Collectively, both studies support the feasibility and effectiveness of Haplo and MUD HCT strategies in Brazil, each reflecting the evolution of transplantation practice over time.

This study has several limitations inherent to its observational, registry-based design. Brazil’s socioeconomic diversity may introduce unmeasured heterogeneity. Nevertheless, the transplantation data are highly consistent: nearly all MUD recipients received ATG, and all Haplo recipients received PTCy. The large, contemporary cohort and methodological rigor are key strengths of our study. Given the logistical and ethical challenges, a randomized comparison of Haplo versus MUD HCT is unlikely soon.

In summary, both Haplo and MUD donors are suitable options for patients without an MSD. Donor selection should prioritize non-HLA factors, anti-HLA, and donor availability. Finally, this study demonstrates the feasibility of multicenter observational studies in a developing country through collaboration between national societies, like the SBTMO, and international partners, such as the CIBMTR.

Supplementary Material

SUPPLEMENTARY MATERIALS

Supplementary material associated with this article can be found in the online version at doi:10.1016/j.jtct.2026.06.033.