The added value of multi-state modelling in a randomized controlled trial

Clofarabine is an active antileukemic drug for subgroups of patients with acute myeloid leukemia (AML). Multi-state models can provide additional insights to supplement the original intention-to-treat analysis of randomized controlled trials (RCT). We re-analyzed the HOVON102/SAKK30/09 phase III RCT for newly diagnosed AML patients, which randomized between standard induction chemotherapy with or without clofarabine. Using multi-state models, we evaluated the effects of induction chemotherapy outcomes (complete remission [CR], measurable residual disease [MRD]), and post-remission therapy with allogeneic stem cell transplantation [alloSCT] on relapse and death. Through the latter a consistent reduction in the hazard of relapse in the clofarabine arm compared to the standard arm was found, which occurred irrespective of MRD status or post-remission treatment with alloSCT, demonstrating a strong and persistent antileukemic effect of clofarabine. During the time period between achieving CR and possible post-remission treatment with alloSCT, non-relapse mortality was higher in patients receiving clofarabine. An overall net benefit of treatment with clofarabine was identified using the composite endpoint current leukemia-free survival (CLFS). In conclusion, these results enforce and extend the earlier reported beneficial effect of clofarabine in AML and show that multi-state models further detail the effect of treatment on competing and series of events. publishedVersion