https://orcid.org/0000-0002-3207-1490
Key Points
Computational evaluation of axi-cel across UC Health confirms its real-world benefit and toxicity burden in diverse patients.
An ML model predicts early relapse using seven factors collected 24 hours post-infusion, aiding early intervention to improve outcomes.
Accumulating real-world (RW) evidence of axi-cel has demonstrated comparable performance to the pivotal trials. However, nearly 57% of patients eventually experience relapse, with most requiring additional therapies. Being able to identify patients with risk of early relapse enables clinicians to consider additional interventions to further extend survival outcomes. This study aimed to first comprehensively evaluate the RW performance of axi-cel in the multi-center University of California (UC) Health Systems using automated computational approaches. Secondly, we developed a decision tree machine learning (ML) model to identify patients with risks of early relapse within six months. 416 adult patients with DLBCL receiving axi-cel between 2017-2024 were included in the study. The median PFS, and OS were 10.1, and 54.4 months; the 18-month PFS and OS rates were 41.6 and 65.5%, respectively. Severe CRS and ICANS were observed in 18.8 and 32.5% of patients, respectively. The decision tree ML model, relying on age and six routinely measured laboratory tests (LDH, CRP, ferritin, hematocrit, platelet count, and PT), achieved a high AUROC score of 0.82. The decision curve analysis indicated positive net benefit of the model across a broad range of decision thresholds (0-0.7), indicating potential clinical utility in diverse scenarios. This study further confirmed the RW performance of axi-cel in diverse populations. Our ML model represented a novel approach to identify patients that may benefit from additional interventions to further extend survival outcomes. Following prospective confirmation study, our model and decision tree approach may support clinical decision making in patients with high-risk DLBCL.
