Haploidentical versus matched unrelated donor transplantation with post-transplant cyclophosphamide: a platform-dependent machine learning analysis of donor age.

This retrospective registry analysis (2017–2021) evaluated 4258 adults with acute leukemia undergoing first allogeneic hematopoietic cell transplantation using post-transplant cyclophosphamide, comparing matched unrelated donors (MUD) with haploidentical donors and focusing on donor age as a platform-specific risk factor.

Using machine learning models (Random Survival Forests, DeepSurv) and multiple robust regression approaches (inverse probability of treatment weighting, 1:1 propensity score matching, Elastic-Net Cox), the authors found that the impact of donor age on overall survival differed substantially by donor type:

• On the MUD + PTCy platform, survival remained relatively stable with increasing donor age; an additional 1% absolute mortality risk (number-needed-to-harm = 100) was not observed until donor age reached 50 years.
• On the haploidentical + PTCy platform, the same 1% added mortality risk threshold was reached at a much younger donor age of 38 years, indicating greater age sensitivity.

After adjustment for confounders, the MUD + PTCy approach was associated with improved overall survival compared with haploidentical + PTCy (adjusted HR 0.85; 95% CI 0.75–0.97; P = 0.01).

Clinically, the study provides a quantitative framework suggesting that, within PTCy-based transplant platforms, a somewhat older fully matched unrelated donor may be preferable to a younger haploidentical donor, and supports re-examination of rigid upper age limits for unrelated donors.