The optimal conditioning intensity of stem cell transplantation for acute myeloid leukemia in complete remission

This study aimed to identify patient groups in which myeloablative conditioning (MAC) or reduced-intensity conditioning (RIC) regimens induced superior progression-free survival (PFS) in patients with acute myeloid leukemia (AML) in complete remission (CR) using a machine-learning approach. Our study included 3273 patients aged 40–69 with AML in CR. The patients were divided into training (N = 2020) and validation cohorts (N = 1253). We employed a machine learning-based group identification model in the training cohort. Subsequently, in the validation cohort, we estimated the impact of the optimal conditioning group compared with the non-optimal conditioning group on PFS using an inverse probability weight analysis. The developed model was consistent with the eight factors and combinations, and the high score suggested that RIC was more appropriate than MAC. In the validation cohort, 127 patients with high scores and who received RIC and 769 patients with low scores and who received MAC were categorized into the optimal conditioning group (896, 71.5%). The weighted hazard ratio for PFS was 0.73 (95% confidence interval: 0.57–0.94) in the optimal conditioning group compared with the non-optimal conditioning group (P = 0.016). In conclusion, we developed an easy-to-use model that helps the physician choose a patient-specific conditioning regimen for patients with AML in CR.