A 52-year-old man with acute myeloid leukemia needs a bone marrow transplant. His 28-year-old son is willing to donate - half-matched, young, healthy, ready to go tomorrow. Meanwhile, somewhere in the donor registry, a 47-year-old stranger is a perfect genetic match. The transplant team huddles. Who gets the call?
This is the kind of decision that keeps hematologists up at night, and a new study just handed them a surprisingly clear answer.
The Great Donor Dilemma
Here's the setup. When someone needs an allogeneic hematopoietic cell transplant (that's a bone marrow transplant from another person, for those of you not fluent in hematology-speak), finding the right donor is like online dating but with significantly higher stakes. You want someone whose immune system proteins - called HLA markers - match yours as closely as possible. A perfect 8-out-of-8 match from a stranger on the registry? That's a matched unrelated donor, or MUD. Your parent, sibling, or kid who shares exactly half your markers? That's a haploidentical donor, or "haplo" if you're into the whole brevity thing.
For years, the conventional wisdom was: younger donor = better outcomes. Registries like the NMDP actively prefer donors aged 18-35, and doctors request them in that range nearly 80% of the time. So when you've got a young family member offering up their stem cells versus a middle-aged stranger from across the country, the choice seems obvious, right?
Not so fast.
Machine Learning Crashes the Party
Researchers led by Rohtesh Mehta analyzed data from 4,258 adults with acute leukemia who received transplants with post-transplant cyclophosphamide (PTCy) between 2017 and 2021. PTCy, by the way, is the drug given after transplant that basically tells rogue donor immune cells to calm down - it's revolutionized how we prevent graft-versus-host disease, that nasty complication where donated cells decide your body is the enemy (Bolaños-Meade et al., NEJM, 2023).
Instead of just running the usual statistics, the team unleashed Random Survival Forests and DeepSurv - machine learning models that can sniff out non-linear patterns humans would miss. And what they found was, well, kind of wild.
Donor age hits different depending on donor type.
For MUD-PTCy transplants, age barely mattered. A matched unrelated donor could be up to 50 years old before their age added even a 1% bump in mortality risk. Fifty! That's a Number-Needed-to-Harm of 100, meaning you'd have to transplant from 100 fifty-year-old MUDs before one extra death could be attributed to age. The MUD platform is basically the Nokia 3310 of transplant - remarkably hard to break.
The haplo platform? Way more fragile. That same 1% risk threshold hit at donor age 38. Twelve years younger. The difference is stark, and the machine learning models and traditional regression approaches agreed on it across the board.
Why the Age Gap?
The researchers think it comes down to immunological complexity. When your donor is only half-matched, the transplanted immune system has a much bigger job - navigating all those mismatched HLA proteins while trying to rebuild your blood system from scratch. Older immune cells are less nimble at that balancing act. With a fully matched donor, there's less immunological chaos to manage, so the system tolerates an aging donor just fine.
Think of it this way: if you're assembling IKEA furniture with a partner who speaks your language (MUD), it doesn't matter much if they're a bit slower. But if you're building that same bookshelf with someone who speaks a completely different dialect (haplo), you really need them sharp and quick.
What This Actually Means for Patients
The bottom line: MUD-PTCy transplants came with an overall survival advantage - an adjusted hazard ratio of 0.85 (95% CI: 0.75-0.97; P = 0.01). That's a 15% reduction in the risk of death compared to haplo-PTCy (Mehta et al., Leukemia, 2026).
This tracks with prior work. A 2023 study in Transplant Cell Therapy found that even younger haplo donors had inferior survival compared to younger MUDs for AML patients on PTCy (Mehta et al., 2023). And a 2025 JAMA Oncology analysis of over 10,000 patients showed PTCy essentially neutralized the negative effects of older unrelated donors on survival, primarily by eliminating the age-related bump in non-relapse mortality (Mehta et al., JAMA Oncol, 2025).
So that 47-year-old stranger on the registry? Probably the better call for our hypothetical patient. And those rigid upper age limits that donor registries impose? This study makes a strong case for loosening them - at least when PTCy is on the menu.
The Bigger Picture
This is really a story about precision over convention. For decades, "pick the youngest donor" was a reasonable heuristic. But it turns out that rule was built on data from a pre-PTCy world. PTCy changed the transplant landscape so dramatically that old assumptions about donor age need recalibrating - and they need recalibrating differently depending on which donor platform you're using.
The researchers have essentially given clinicians a quantitative framework: if you have a well-matched unrelated donor under 50, age shouldn't scare you off. If your best option is a haplo donor pushing 40, maybe keep searching.
Not bad for a Tuesday.
References
-
Mehta RS, Kanakry CG, Nawas M, et al. Haploidentical versus matched unrelated donor transplantation with post-transplant cyclophosphamide: a platform-dependent machine learning analysis of donor age. Leukemia. 2026. DOI: 10.1038/s41375-026-02903-8. PMID: 41986623
-
Bolaños-Meade J, Hamadani M, Wu J, et al. Post-transplantation cyclophosphamide-based graft-versus-host disease prophylaxis. N Engl J Med. 2023;388(25):2338-2348. DOI: 10.1056/NEJMoa2215943. PMID: 37342922
-
Mehta RS, Ramdial J, Marin D, et al. Impact of donor age in haploidentical-post-transplantation cyclophosphamide versus matched unrelated donor post-transplantation cyclophosphamide hematopoietic stem cell transplantation in patients with acute myeloid leukemia. Transplant Cell Ther. 2023;29(6):377.e1-377.e7. DOI: 10.1016/j.jtct.2023.03.028. PMID: 36990221
-
Mehta RS, Sparapani RA, Kanakry CG, et al. Unrelated donor age and recipient outcomes after posttransplant cyclophosphamide vs conventional prophylaxis. JAMA Oncol. 2025. DOI: 10.1001/jamaoncol.2025.4551. PMID: 41196629
-
Kanakry CG, Fuchs EJ, Luznik L. Modern approaches to HLA-haploidentical blood or marrow transplantation. Nat Rev Clin Oncol. 2016;13(1):10-24. DOI: 10.1038/nrclinonc.2015.128. PMCID: PMC4955091
Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.