Response to ‘Adverse cardiac outcomes in patients with chronic myeloid leukaemia treated with tyrosine kinase inhibitors’

Abstract

To the Editor Leong and colleagues performed an important retrospective cohort study to assess the risk of MACE between patients with and without chronic myeloid leukaemia (CML), with subgroup analysis by patients present before and after the market introduction of tyrosine kinase inhibitors (TKIs) (cutoff at 2001). However, the cohort design has several limitations. While the cutpoint by TKI marketing at first seems logical, TKIs as a group have drastically changed since the early 2000s. For example, TKI access was often restricted among older adults because initial studies showed the greatest survival benefit in younger patients. Also, many patients who were diagnosed prior to 2001 were eventually treated with TKIs. Furthermore, latergeneration TKIs have both greater oncological response and more adverse reactions (including cardiac toxicity and congestive heart failure) when compared with firstgeneration TKIs. Notably, the authors thoughtfully compared patients receiving imatinib with patients receiving dasatinib or nilotinib. Existing literature has established greater side effects with latergeneration TKIs, although the current study does not support this finding (and importantly does not include patients treated with the more potent thirdgeneration TKIs). The authors should consider subgroup analyses for firstgeneration versus secondgeneration TKIs, with separate analysis for any secondgeneration recipients previously treated with firstgeneration or nonTKI therapies. Ultimately, the authors found significant protective HRs among those with CML prior to 2001 and nonsignificant protective HRs among those with CML from 2001 onwards. In the majority of included figures, however, the Cox models are not proportional, with strong divergence as age increases. If the HR is not constant over time, then the proportional hazards assumption is violated and it is difficult to interpret these data. The restricted means technique can provide more clinically meaningful estimates in this situation. Another option would be to report perioddependent HRs (eg, compare 50–60 years with 60–70 years subgroups). Although this may risk differential selection bias among patients with CML in the later cohorts who are less susceptible to MACE, the HR would be biased towards the null. Given the extremes in the authors’ figure curves despite the assumed bias towards the null, there will likely be a significant difference in hazards among the older groups and none among the younger groups. Accordingly, older patients may benefit more from cautions about TKIrelated MACE, and further studies are needed to better guide TKI strategies (specifically drug duration and generation) among geriatric populations. Again, we commend the authors for their excellent contribution and hope it will urge research centres to establish more robust trials that can more directly delineate the relationship between initiation TKIs and acceleration of cardiovascular disease.