Frontline nilotinib vs dasatinib in newly diagnosed chronic phase chronic myeloid leukemia: A propensity score analysis based on AIFA monitoring registries

Abstract Introduction Frontline second generation tyrosine kinase inhibitors increased the rate of deep molecular responses (DMR) but were also mined by increased related adverse events. A single Asiatic trial randomized the two drugs, dasatinib and nilotinib without differences in efficacy. Aim the aim of this analysis is to evaluate the time to discontinuation among the two drug-defined groups after balancing for the available baseline characteristics. Methods A nationwide real-world retrospective cohort study including all newly diagnosed chronic phase CML adult Italian patients who started dasatinib or nilotinib from January 2013 to December 2016. Dasatinib and nilotinb posology and administration were in accordance to EU SmPC recommendations. The decision to prescribe either of the two drugs was based on treating physician clinical judgement. The main outcome considered is time to treatment discontinuation (TTD). TTD was compared among the two groups after balancing for the available baseline characteristics. Median TTD and follow-up were obtained using the Kaplan-Meier and Reverse Kaplan-Meier respectively. Baseline characteristics of the two groups was balanced using weights obtained from a gradient boosting machine algorithm (gbm). The methodology is based on an average treatment effect for the treated (ATE) Inverse Probability Treatment Weighting (IPTW) (REF) approach to estimate propensity scores from the gbm model and obtain the relative weights. The variables included in the model were: age (continuous, years), sex, Sokal risk class and center experience (more/less). The weights were then used to adjust TTD between the two groups and to fit Cox proportional hazard models both in the overall population and in the subgroups of interest, to obtain the adjusted hazard ratios and their 95% confidence intervals of the comparison of dasatinib and nilotinib. Results Overall, 1526 patients were analyzed (628 treated with dasatinib and 898 with nilotinib), with a male prevalence (56.2%) and a median age of 54.1 years (57.3 in dasatinib cohort and 51.9 in nilotinib group). Overall the majority of patients were aged <65 years (74.1%), with 9.6% of patients aged between 65 and 70 years, and 16.3% aged >70 years. Sokal score identify a low risk in 40.8%, intermediate 37.8% and a high risk in 21.4% (high Sokal in 26.4% with dasatinib and 17.8% with nilotinib, respectively). According to geographic site of prescription, 33% were treated in south of Italy, 30% in the north, 26.7% in the center and 9.7% in the islands. After a median follow-up of 4.9 years and post-adjustment, the median time in treatment was 53.5 months with dasatinib and 70.5 with nilotinib (HR 0.61, 95% CI 0.54-0.69, p<0.0001). The TTD at 24 and 48 months with nilotinib was 0.79 (0.77-0.82) and 0.68 (0.65-0.72), respectively and with dasatinib was 0.73 (0.70-0.77) and 0.55 (0.51-0.60), respectively. A comparison between the two drugs for selected covariate categories, showed, among all selected variables, a lower risk of treatment discontinuation associated to the nilotinib treatment. In particular, the rate of discontinuation is higher for dasatinib compared to nilotinib for high Sokal risk (p=0.0039), male gender (p<0.001), age 65-69 years (p<0.001) and older patients aged >70 years (p<0.001). Conclusions With the limitation of a real-world analysis based on AIFA registries we showed a longer probability of treatment with frontline nilotinib. Considering the future introduction of new drugs for newly diagnosed CML, this data suggests the need for further investigations on treatment discontinuation causes to evaluate baseline clinical findings supporting the use of selected drugs.