Optimal Donor Selection Across Multiple Outcomes For Hematopoietic Stem Cell Transplantation By Bayesian...

Area of Science:

• Hematology
• Immunology
• Biostatistics

Background:

• Allogeneic hematopoietic cell transplantation (HCT) offers curative potential for hematologic malignancies but faces challenges from donor selection complexities.
• Standard HLA gene matching is established, but optimal selection among 8/8 allele-matched unrelated donors requires prioritizing additional factors like HLA-DPB1/DQB1, donor sex, CMV status, and age.
• Donor characteristics influence various post-transplant outcomes, including survival, relapse, graft failure, and chronic graft-versus-host disease, impacting overall event-free survival (EFS).

Approach:

• Developed a general methodology using Bayesian nonparametric machine learning to model trade-offs between multiple outcomes for optimal treatment decisions.
• Applied this approach to HCT donor selection, utilizing a large registry dataset from the Center for International Blood and Marrow Transplant Research (CIBMTR).
• The model aims to optimize both overall survival and event-free survival (EFS) by considering complex donor attributes.

Key Points:

• Identified a data-driven donor selection strategy balancing multiple post-transplant outcomes.
• The methodology addresses the variability in donor selection practices across different centers and patient cases.
• Bayesian nonparametric machine learning provides a robust framework for optimizing complex medical decisions with multiple competing objectives.

Conclusions:

• The proposed approach recommends a donor selection strategy favoring the youngest male donor.
• An exception is made when a female donor is substantially younger, indicating a potential for better outcomes.
• This data-driven strategy aims to enhance patient survival and event-free survival in allogeneic HCT.