Highlights
– In a CIBMTR registry cohort of 10,025 adults with acute leukemia or MDS, increasing unrelated donor age was associated with slightly worse overall survival when calcineurin inhibitor (CNI)–based GVHD prophylaxis was used, but this association was not seen when posttransplant cyclophosphamide (PTCy)–based prophylaxis was applied.
– The attenuation of donor-age effect in the PTCy group was largely explained by a lack of association between older donor age and nonrelapse mortality (NRM), suggesting PTCy may mitigate alloimmune complications that disproportionately affect survival with older donors.
– Findings were robust across multiple analytic approaches (LASSO-penalized Cox models, IPTW, and XGBoost), but as an observational registry study they require external validation before changing donor selection policies.
Background and clinical context
Selecting an optimal unrelated donor (URD) for allogeneic hematopoietic cell transplantation (HCT) balances human leukocyte antigen (HLA) compatibility, donor availability, and donor characteristics that influence recipient outcomes. Registries and clinical practice have long prioritized younger donors because multiple observational studies found older donor age is associated with higher graft-versus-host disease (GVHD), nonrelapse mortality, and worse overall survival (OS). These effects likely reflect age-related changes in immune cell composition and function, thymic involution, increased comorbidity burden, and accumulation of alloimmunologic memory.
Concurrently, posttransplant cyclophosphamide (PTCy) has emerged from haploidentical transplantation experience to become a widely used strategy for GVHD prophylaxis across donor types, including matched unrelated donors (MUDs) and mismatched unrelated donors (MMUDs). PTCy selectively eliminates proliferating alloreactive T cells early posttransplant while sparing regulatory elements and hematopoietic stem cells, thereby reducing acute and chronic GVHD in many series. Whether PTCy changes the relevance of donor age for recipient outcomes has been an open and clinically important question: if PTCy mitigates age-associated alloimmune risk, donor age could become a less restrictive criterion, potentially enlarging the donor pool.
Study design
The referenced study is a multicenter cohort analysis using Center for International Blood and Marrow Transplant Research (CIBMTR) registry data from January 2017 through June 2021, with data analyzed January–June 2025. Inclusion criteria were adult recipients of allogeneic HCT for acute leukemia or myelodysplastic syndrome (MDS) using an unrelated donor who was either fully matched at 8/8 HLA loci (MUD) or mismatched at 7/8 loci (MMUD).
Primary exposure variables were GVHD prophylaxis strategy (PTCy-based versus conventional calcineurin inhibitor [CNI]–based) and donor age (modeled both continuously per year and categorically). The primary endpoint was overall survival. The investigators used a multipronged analytic approach to assess robustness and reduce bias: LASSO-penalized Cox proportional hazards models to select covariates, inverse probability of treatment weighting (IPTW) including standard and overlap weighting to balance covariates across prophylaxis groups, and XGBoost machine learning to examine nonlinearities and interactions.
Cohort composition: 10,025 patients (mean recipient age 56.5 years; 43.7% female). Distribution by donor/HLA/prophylaxis groups was: 7,272 MUD-CNI (72.5%), 1,681 MUD-PTCy (16.8%), 613 MMUD-PTCy (6.1%), and 459 MMUD-CNI (4.6%).
Key findings
Primary result
– Increasing donor age was associated with worse overall survival in recipients who received conventional CNI-based prophylaxis: for MUD-CNI cohorts the hazard ratio (HR) per additional donor year ranged roughly 1.004 to 1.009, and for MMUD-CNI cohorts the HR range was larger (1.022–1.034 per year). By contrast, in the combined PTCy cohort (MUD-PTCy and MMUD-PTCy pooled) donor age showed no meaningful association with OS (HR approximately 1.001–1.007 per year).
Robustness across analytic approaches
– These patterns persisted across multiple modeling strategies including standard and overlap-weighted IPTW analyses, LASSO-penalized Cox models that adjusted for a broad set of recipient, disease, donor, and transplant characteristics, and XGBoost (a gradient boosting machine learning method) to probe for nonlinear effects and interactions. This convergence supports that the differing associations are unlikely to be an artifact of a single analytic method.
Secondary analyses and drivers
– The attenuation of the donor-age effect in the PTCy group was primarily driven by a lack of association between donor age and nonrelapse mortality (NRM). Put differently, older donor age predicted higher NRM (and thus worse OS) in CNI cohorts but not in PTCy cohorts. Associations with relapse risk were less consistent and did not meaningfully explain the overall differences.
Effect sizes and clinical meaning
– The per-year HRs are numerically small but aggregate over decades of donor-age differences. For example, an HR of 1.01 per donor year translates to ~1.1-fold increased hazard over a 10-year donor-age gap; an HR of 1.03 per year (seen in some MMUD-CNI analyses) becomes clinically more substantial over similar spans. The findings therefore have practical relevance when practitioners consider donors decades apart in age.
Subgroup observations
– The effect of donor age appeared more pronounced in the CNI-prophylaxis, mismatched donor setting (MMUD-CNI) than in 8/8 matched settings, aligning with the general notion that mismatch amplifies alloimmune risk. PTCy seemed to blunt this amplification.
Expert commentary and mechanistic considerations
Biologic plausibility
– Older donor age could worsen recipient outcomes through multiple, non–mutually exclusive mechanisms: altered T-cell repertoire and function, increased frequency of memory/effector T cells that mount stronger alloresponses, increased incidence of donor clonal hematopoiesis, and age-linked cytokine milieu differences. PTCy, with its ability to selectively purge proliferating alloreactive lymphocytes early posttransplant and promote regulatory T-cell reconstitution, plausibly dampens these donor-age–related alloimmune processes and thereby reduces NRM driven by severe GVHD, infections secondary to GVHD and its therapy, or other immune complications.
Limitations and alternative explanations
– Observational registry data remain subject to confounding by indication and unmeasured variables. Transplant centers may choose PTCy for specific donor-recipient constellations (for example, MMUDs or particular graft sources), and subtle selection biases could influence results even with advanced statistical adjustment.
– Donor-related variables beyond chronological age—such as donor sex, parity, cytomegalovirus (CMV) serostatus, history of infections, choice of peripheral blood vs bone marrow graft, and donor comorbidity—may co-vary with age and influence outcomes. If such factors were incompletely captured or unbalanced, they could contribute to observed associations.
– Follow-up time in the registry cohort (patients transplanted 2017–2021) may limit assessment of very late effects. PTCy is relatively newer in matched-unrelated settings; adoption patterns changed over that interval and calendar-time effects could confound associations despite IPTW.
Implications for practice and policy
– If validated, the finding that PTCy attenuates the adverse impact of older unrelated donors could relax strict age-based donor prioritization rules and increase the usable donor pool, shortening time to transplant for some patients. This would be particularly relevant when a slightly older but readily available URD could be selected over waiting for a younger donor.
– Nevertheless, it is premature to recommend broad, immediate changes to donor selection policies. The study supports equipoise and provides compelling observational evidence, but prospective validation and external replication (including in different registries and prospective comparative cohorts) are needed.
Research opportunities
– Mechanistic correlative studies could compare reconstituting T-cell phenotypes, regulatory T-cell recovery, cytokine profiles, and incidence of clonal hematopoiesis of indeterminate potential (CHIP) among recipients of older vs younger donors across prophylaxis strategies.
– Randomized or pragmatic trials comparing PTCy vs CNI prophylaxis stratified by donor age could provide higher-level evidence. Alternatively, prospective registry-based studies with richer donor biologic annotation (CMV, CHIP, immunophenotyping) could help disambiguate mechanisms.
Conclusions and practice points
This large CIBMTR registry cohort suggests that PTCy-based GVHD prophylaxis may neutralize the previously observed adverse association between older unrelated donor age and recipient overall survival, largely by removing the link between donor age and nonrelapse mortality. The observation is biologically plausible and consistent across multiple advanced analytic methods, but it is not definitive proof of causality.
Until external validation and further mechanistic study, clinicians should consider these findings as an important signal rather than a practice-changing mandate. In clinical scenarios where donor age is a central concern and PTCy is a feasible prophylactic strategy, the data support considering PTCy as part of a plan that may permit acceptance of somewhat older unrelated donors—particularly if donor availability and timeliness are critical. Policy changes in registry donor prioritization should await replication.
Funding and clinicaltrials.gov
– The reported analysis used data from the Center for International Blood and Marrow Transplant Research (CIBMTR). Specific funding statements and trial registrations, if any, are reported in the primary publication.
References
1. Mehta RS, Sparapani RA, Kanakry CG, McCurdy SR, Saultz J, Lazaryan A, Milano F, Lee SJ. Unrelated Donor Age and Recipient Outcomes After Posttransplant Cyclophosphamide vs Conventional Prophylaxis. JAMA Oncol. 2025 Nov 6:e254551. doi: 10.1001/jamaoncol.2025.4551. Epub ahead of print. PMID: 41196629; PMCID: PMC12593671.
(Readers seeking deeper methodological detail, subgroup breakdowns, and adjusted estimates should consult the full JAMA Oncology article cited above.)
