An Indirect Comparison between Bosutinib, Nilotinib and Dasatinib in First-Line Chronic Phase Chronic Myeloid Leukemia

Introduction

Several randomized controlled trials (RCTs) compare second generation tyrosine kinase inhibitors (TKIs) with imatinib (IM) for the first-line (1L) treatment of chronic phase chronic myeloid leukemia (CP-CML). However, an examination indicated cross-trial heterogeneity in terms of disease and patient characteristics (baseline risk score, age, trial region, and post-baseline IM dose escalation) and outcomes reported (type and time of evaluation). For this reason, a matching adjusted indirect treatment comparison (MAIC) can be used to assess comparative efficacy. The objective of this study was to compare the efficacy of bosutinib (BOS) with nilotinib (NIL) and dasatinib (DAS) in 1L CP-CML via unanchored MAICs.

Methods

As well as heterogeneity in baseline characteristics, differences in the amount of IM dose escalations in earlier clinical pivotal trials (ENESTnd: 15.9% at 12 months; DASISION 20% at 24 months) relative to more recent ones (BFORE: 27.5% and 30.9% at 12 and 24 months, respectively) were observed. Because differences in IM dose-escalation regimens are trial design matters, they cannot be adjusted for using anchored MAICs. Therefore, an unanchored MAIC (which dismisses the common comparator arm, in this case IM) was used.

Unanchored MAICs comparing BOS with NIL and DAS were performed to match the summary statistics in the BFORE trial to the reported summary statistics in the ENESTnd (NIL) and the DASISION (DAS) trials. Outcomes assessed included major molecular response (MMR) at 12 months and at any time after 12 months, complete cytogenetic response (CCyR) by 12 months, and deep molecular response (MR4) at 12 months. Statistical significance was assessed based on confidence intervals (CI; significant when CI excludes 1). The sample size of the matching-adjusted BOS population was estimated using the effective sample size (ESS). No adjustments for multiple comparisons were made.

Results

The MAICs showed an MMR at 12 months of 49.32% (vs 47.13% before the MAIC, N=244, ESS=186) for BOS vs 44% for NIL (OR 1.24, 95% CI [0.87-1.77], N=282) and an MMR at any time of 58.09% (vs 57.79% before MAIC, N=244, ESS=142) for BOS vs 57% for NIL (OR 1.05, 95% CI [0.72-1.5], N=282), and 58.52% (vs 58.16% before MAIC, N=239, ESS=192) for BOS vs 52% for DAS (OR 1.3, 95% CI [0.9-1.88], N=259). BOS had a CCyR by 12 months of 82.59% (vs 82.38% before MAIC, N=244) vs 80% for NIL (OR 1.19, 95% CI [0.75-1.87], N=282, ESS=195) and 80.47% (vs 82.01% before MAIC, N=239, ESS=185) vs 83% for DAS (OR 0.84, 95% CI [0.53-1.35], N=259). MR4 at 12 months was 20.34% (vs 20.49% before MAIC, N=244, ESS=185) for BOS vs 12% for NIL (OR 1.87, 95% CI [1.15-3.05], N=282). MMR and MR4 at month 12 were not reported for DAS.

Discussion

The general MAIC limitations (impact of having access to only marginal covariate distribution, choice of scale for indirect comparisons, choice of target population and sampling variation in the target population) apply to our analyses as well. Nevertheless, indirect comparisons using unanchored MAICs were conducted because cross-trial heterogeneity in patient baseline characteristics could bias the results of a naïve comparison and increases in IM dose escalation over time could bias the results of an anchored MAIC. The unanchored MAICs indicated that BOS was numerically higher than NIL for all response rates examined, however, except for MR4, none were statistically significant. BOS had a numerically higher MMR at any time and a numerically lower CCyR by 12 months compared with DAS, however, neither were statistically significant. BOS therefore seems to be an equally effective TKI as NIL and DAS in the 1L CP-CML setting. These findings, based on statistical methods, should be confirmed by clinical research.