Early switch to second-line tyrosine kinase inhibitor in chronic myeloid leukemia patients failing to achieve early molecular response

Failure to achieve early molecular response (EMR, BCR-ABL1 copies ≤ 10% by the international scale) 3 months after initiation of tyrosine kinase inhibitor (TKI) therapy has been shown to be a poor prognostic marker in chronic myeloid leukemia in chronic phase (CML-CP).1 However, whether the risk of adverse outcomes of EMR failure can be reduced by a switch in TKI therapy and when the switch should occur are unknown. Furthermore, patients with EMR failure may be a heterogeneous group that can be further stratified by long and short BCR-ABL1 halving time (HT).2 Thus, patients whose BCR-ABL1 copy numbers at 3 months are halved or near-EMR may not be need an early switch.3 We conducted an observational study to address the role of early switch of TKI therapy in EMR failure and long HT.

We evaluated the association of switch within 6 months to second-line TKI for EMR failure with the achievement of major molecular response (MMR, BCR-ABL1 copies ≤ 0.1% by the international scale) among patients enrolled in the Quebec CML Registry (see detailed methods in the Supporting Information). Among 283 evaluable patients, 59 (20.8%) failed to achieve EMR while this milestone was achieved in 224. Baseline characteristics of patients with and without EMR were comparable (Table S1 in the Supporting Information). EMR failure was associated with a reduced probability of MMR (adjusted hazard ratio [HR], 0.46; 95% confidence intervals [CI], 0.33–0.63).

In the subset with EMR failure, early switch occurred in 23 (39.0%) patients, while 36 (61.0%) remained on frontline TKI through 6 months from EMR failure. Compared with non-early switchers, early switchers were more likely to be male, to have received imatinib in the front line and to have been treated at lower volume centers (Table S2 in the Supporting Information).

Patients with EMR failure were followed for a mean (standard deviation) of 19.1 (15.0) months from EMR determination. During 1125 months of follow-up, 45 patients achieved MMR. To avoid immortal time bias, we treated the switch to second-line TKI as a time-updated covariate such that patients were considered as “no switch” up to the time a switch occurred. The HR for achieving MMR was adjusted for age, sex, and first-line TKI (imatinib vs non-imatinib). The adjusted HR for MMR with early switch at 6 months was 1.63 (95% CI, 0.83–3.23), compared with no early switch. The adjusted HR for MMR with a switch within 3 months was similar (HR 1.47; 95% CI, 0.70–3.09), but was below the null with a switch up to 12 months (HR 0.79; 95% CI, 0.41–1.54) (Table 1).

Among the 59 patients with EMR failure, baseline BCR-ABL1 copy number was available for 46. The median HT in this population was 65 days using the method suggested by Branford et al.2 The adjusted HR for MMR with early switch within 6 months in patients with short HT was 0.63 (95% CI, 0.15–2.60), while in patients with long HT it was 2.40 (95% CI, 0.66–8.71) (Table 1).

Finally, after entering all the covariates in the model in a sensitivity analysis, the adjusted HR for MMR with early switch was 2.06 (95% CI, 0.85–4.98).

Our study suggests that switch in TKI therapy within 6 months is associated with a higher likelihood of achieving MMR in patients with EMR failure or long HT. We could not find any other studies reporting the rate of early switch among patients with EMR failure in routine CML care nor the comparative effectiveness of this approach. In our study, 17/23 (74%) of early switchers achieved MMR during follow-up, which compares favorably with the 21/54 (39%) who switched from imatinib to nilotinib in the Therapeutic Intensification in De Novo Leukaemia (TIDEL)-II trial.4 One possible explanation for this difference may come from the benefit suggested in this analysis to a switch no later than 6 months from EMR failure, whereas in TIDEL II there was heterogeneity in the timing of switch.

The proportion of patients with EMR failure reported herein is very similar to that reported in several trials (18%).2 Further, the prognostic importance of EMR failure was reiterated in our study despite a surprisingly high rate of MMR (72.7% at 30 months) in the subset with EMR failure. This rate is higher than that reported in imatinib-treated patients (41% at 4 years),2 or in nilotinib-treated patients (29% at 2 years)5 enrolled in clinical trials.

The strengths of our study include the use of time-dependent exposure definition, which avoids the misclassification of observation time in patients who switched at various time points after EMR failure. Further, we used only EMR determination using the international scale, which is comparable with trial populations and with the current standard of care. Lastly, we modeled early switch only in poor responders to frontline TKI and avoided comparisons with non-early switch in patients with good response, which would have led to confounding by indication, ie, comparing early switchers with patients who benefited from frontline TKI (which would have diluted the effect).

Our study also has several limitations. In adjusted models, we were unable to demonstrate a statistically significant increase in MMR with early switch to second-line TKI, perhaps due to residual confounding in a heterogeneous population and small sample size. To avoid overfitting, we adjusted our primary analysis only for age, sex and first-line TKI. Notably, with inclusion of several additional covariates, the estimate for benefit with early switch only increased. Our sample size allowed us only to reject a probability lower than 0.83 (the lower CI limit), suggesting cautious interpretation of our findings. Lastly, we could not adjust our models for Sokal scores. Nevertheless, EMR failure and HT have been repeatedly shown to be more predictive of patient trajectories than Sokal.2, 6

In summary, this first study of the comparative effectiveness of early switch in EMR failure in routine care suggests that this strategy is associated with a greater likelihood of achieving MMR particularly in patients with long HT. Our findings provide a rationale to conduct a prospective investigation of early switch in this population.

AUTHOR CONTRIBUTIONS

Study design: Kilil-Drori, Azoulay, Assouline

Data analysis: Kilil-Drori, Yin

Manuscript writing: All authors

Data review: All authors

Data collection: Harnois, Gratton, Del Corpo, Klil-Drori

Obtained financial support: Assouline