Matching-Adjusted Indirect Treatment Comparison of Bosutinib, Dasatinib, Nilotinib and Ponatinib on Survival for Second Line Chronic Phase Chronic Myeloid Leukemia Patients

Background: In clinical trials of second line and later line therapies for chronic myeloid leukemia (CML), to date only single arm trials have been conducted for the available treatments (i.e. Bosutinib, Dasatinib, Nilotinib and Ponatinib). These trials included heterogeneous patient populations in terms of line of treatment, stage of disease and other patient- and disease baseline characteristics that may impact the outcomes of the treatment. This hampers any direct or indirect comparisons of these second line CML treatments in terms of efficacy measures such as progression-free survival (PFS) and overall survival (OS). In this situation (i.e. lack of common comparator), matching-adjusted indirect treatment comparisons (MAICs) can be applied.

Objective: The aim of this research was to compare the efficacy (PFS and OS) of Bosutinib, Dasatinib, Nilotinib and Ponatinib in second line chronic phase CML patients by means of MAICs. Patient data are reweighted such that their medians on demographic and clinical characteristics match those of the aggregated comparator trial, in order to adjust for cross-trial differences in those baseline characteristics. Outcomes of the reweighted patient level dataset are then compared with those of the aggregated comparator trial.

Methods: A systematic literature review of second-line or later CML Phase II and Phase III clinical trials identified 2 trials for Bosutinib, 11 trials for Dasatinib, 8 trials for Nilotinib and 2 trials for Ponatinib. For the MAIC, only second line chronic phase trials with three or more years of follow-up (to allow for sufficient differentiation of treatments over time) were considered, resulting in 1 trial for Bosutinib (NCT00261846), 1 trial for Dasatinib (CA180-034), 1 trial for Nilotinib (NCT00109707) and no trials for Ponatinib. To adjust for cross-trial differences, patient data from the Bosutinib trial were subject to the inclusion and exclusion criteria reported in the comparator trials, and were weighted to match all available summary baseline characteristics reported in the 1 Dasatinib trial and the 1 Nilotinib trial. The baseline characteristics considered for reweighing were as follows: age, gender, Imatinib resistant/intolerant, disease duration, and prior stem cell transplantation. After the MAICs, PFS and OS were compared based on hazard ratios (HR) derived by Cox proportional hazard (PH) regressions using the reweighed Bosutinib dataset. The Cox PH assumption was tested with log cumulative hazard plots, Schoenfeld residuals and its corresponding statistical test. In case the PH assumption did not hold, Cox regressions with time varying coefficients were tested and the relative restricted mean survival time (RMST) was estimated.

Results: Comparing Bosutinib to Nilotinib resulted in HRs for PFS and OS of 2.0 (<0.01) and 1.4 (p=0.109) respectively. When compared to Dasatinib, Bosutinib showed a non-significant HR of 1.3 (p=0.30) in favor of Bosutinib for OS and a significant 1.6 HR (p<0.01) for PFS. In the PFS and OS analyses compared to Dasatinib, the PH assumption was violated based on the statistical test or almost violated according to the visual plots. Therefore, we examined the relative restricted mean survival times, which were 1.123 (p=0.02) for PFS and 1.025 (p=0.41) for OS.

Discussion: After performing MAICs to adjust for cross-trial differences in baseline characteristics, Bosutinib appeared to have a significantly greater PFS than Nilotinib (p<0.01). Based on the relative RMST analyses, Bosutinib also appeared to have a significantly greater PFS than Dasatinib, at the p<0.05 level. For OS the findings were numerically positive in favor of Bosutinib, but not statistically significant, compared to both. However, qualitatively, Bosutinib appears to be an equally effective second line chronic phase CML therapy. The greater PFS compared to Nilotinib and Dasatinib, especially at the end of the follow up period for Bosutinib, may translate into future OS benefits not yet observed.

A limitation of our study is that we could only correct for observed differences between the trials, and not for unobserved ones. Also, the reweighing was performed for all relevant reported baseline characteristics; not all baseline characteristics were reported in all trials. Both limitations could have an impact on the outcomes and therefore the study conclusions.