The treatment landscape of patients with chronic myeloid leukemia (CML) changed dramatically after the approval of imatinib, the first tyrosine kinase inhibitor (TKI), in 2001 (1). Since then, several newer TKIs have also been approved, and 3 different TKIs are currently available as front-line treatments in Italy (2). Due to the remarkable efficacy of TKIs therapy, the life expectancy of newly diagnosed chronic phase (CP) patients is now near to that of age-matched individuals in the general population (3). The improved outcome and long-term safety of these drugs have mainly been demonstrated in sponsored randomized controlled trials (RCTs) (4). However, many questions related to the prognosis and the optimal management of newly diagnosed CML patients remain unanswered, and it is widely accepted that these questions could be addressed by evaluating large prospective cohorts in real world clinical practice. Prognostic evaluation of baseline features has been reported in small single country series (5, 6) or in large datasets including 20 countries in Europe, such as the EUTOS registry (7). Until 2016, three different clinical prognostic scores were in use in clinical practice (Sokal, Euro, Eutos scores) (8), before the EUTOS Long Term survival (ELTS) score (9) was introduced, that stratifies patients in three different risk groups, with significantly different probabilities of dying of CML. The EUTOS score ability to discriminate CML patients in terms of long-term overall survival has been validated several times, but predominantly in patients treated with front-line imatinib (10–12). The score was recently suggested by the European Leukemia Net (ELN) panel in the updated recommendations (13) as a helpful tool to predict the rate of deaths related to CML in TKIs-treated patients.

The Italian CML GIMEMA network promoted an observational cohort study in January 2013, to collect a large series of consecutive newly diagnosed patients and evaluate the management and the long-term outcomes in a real-world perspective.

The aim of this article is to analyze the impact of the baseline patients’ characteristics and their front-line treatments on long term overall survival and CML-related deaths.

The CML Italian GIMEMA network (including 68 Hematology Centers from 19 Italian regions) prospectively recorded, in a dedicated web-based database (https://www.epiclin.it/lmc), the clinical and biological features of all newly diagnosed adult (>18 years) Italian Ph+ CML patients in each phase of disease, diagnosed from January 2013 onwards. All consecutive patients were included, without any exclusion criteria and regardless of their participation in any other clinical trial, to limit the eligibility criteria selection bias typical of experimental studies. All centers followed the ELN guidelines currently available and their updates (13, 14) for the management of patients, without any other predefined recommendations, including the selection of first line TKI treatment and the subsequent monitoring every 3 months.

The study was approved by the local ethics committees and other competent authorities; all patients were registered after obtaining prior informed consent.

Standardized information on all newly diagnosed patients was collected and entered in the database by local staff and centrally checked for completeness and coherence. Baseline information included sociodemographic, clinical, and standard laboratory data. Comorbidities were evaluated by medical staff before the start of treatment according to the Charlson Comorbidity index (CCI) (15). CML was classified as chronic phase (CP), accelerated phase (AP) and blast phase (BP) according to the ELN criteria (13). Cytogenetic analysis was performed according to banding analysis at baseline, as well as qualitative molecular analysis to define the type of BCR-ABL transcript. Sokal (16) and ELTS (9) scores were calculated as previously reported. Any front-line treatment was recorded. Overall survival (OS) and cause-specific survival (CML-related deaths, other causes of death) were calculated from the date of diagnosis. Cause of death was clinically defined as leukemia-related when it occurred after progression to accelerated phase (AP) or blast phase (BP). All other deaths were classified as leukemia-unrelated, and the specific causes of death were recorded.

In this article we present the baseline characteristics (age, sex, comorbidities, prognostic scores), the type of TKI initially prescribed (Imatinib - IMA or 2^nd generation TKIs -2GTKI) and their impact on survival, overall and by cause of death. For OS analyses, Kaplan–Meier curves were calculated. To quantify differences in survival probabilities between groups, the log-rank test was applied. Cox proportional hazard regression models were applied to analyze the influence of all the variables considered on OS. Harrell’s C statistic was calculated to assess the discrimination ability of the two prognostic scores.

To analyze the association of prognostic factors and of first-line treatments on CML mortality, deaths due to other causes were considered competitive events. Separate cumulative incidence curves for CML and other causes mortality were estimated with the Gooley method. Adjusted sub–Hazard Ratios (sHR) for prognostic factors and treatment were estimated separately for CML and other causes mortality with the Fine and Gray method.

In the last decade, the availability of different first line TKIs has changed the clinical management of CML patients, allowing a better diagnostic work-up and careful evaluation of baseline comorbidities to select the best therapeutic option. In fact, the consequence of TKI-related improved survival is an increased probability of dying of other, unrelated causes: for this reason, it is of paramount importance to analyze specific causes of deaths separately, and avoid specific drug-related off-target effects.

In this study we report the overall survival and CML-related death probability of a large CML population, prospectively enrolled in a multicentric Italian observational study, together with estimates of the role of major prognostic factors and of first line treatments. Prescription preferences by Italian centers clearly showed that imatinib was prevalently chosen for older patients, whose median age was 69 years older, and who had an increased burden of comorbidities, whereas a 2GTKI was reserved to the younger and healthier population. The 4 years OS for the complete cohort was 89%, a result not so different from those reported in randomized controlled trials (RCT) of younger, more selected patients.

Previous reports analyzed large CML cohorts outside clinical trials but prevalently treated with imatinib: the EUTOS group reported a CML population collected in 20 predefined countries and regions in Europe between January 2008 and December 2013, showing an OS probability at 30 months of 92%, with a risk of dying in remission of 1% after 24 months (17). In this population, the ELTS score was tested and showed a significant difference in OS, namely 96%, 89% and 84% in the low, intermediate, and high-risk groups, respectively (8). The EUTOS group reported a subsequent validation of the ELTS score in 2949 CML patients, of whom 236 died, 89 of CML-related causes (12). The overall probability of dying of CML was 5%: applying the ELTS score both the intermediate and high-risk groups had significantly higher probabilities of dying of the disease as compared to the corresponding risk groups defined by the Sokal score. The results of our study are in line with the report by the EUTOS group but are based on a population in which about 50% of patients were treated with a 2GTKI rather than mainly with imatinib.

Since the introduction of imatinib as front-line treatment, older age appears to have lost much of its prognostic relevance. Several experiences have been reported based on age stratification: among them, the MDACC experience showed that older patients had similar cytogenetic response rates and survival compared to younger patients in chronic phase, whereas a worse survival was reported for patients in an advanced phase of the disease (18). Characteristics of CML and rates of responses vary according to age, as demonstrated in a large analysis including 2784 adult patients: the frequency of splenomegaly was more evident in younger patients, as also a high-risk stratification according to prognostic scores, and lower rates of cytogenetic responses with a higher risk of progression (19). In contrast, the German group reported that younger patients do well with imatinib despite baseline prognostic indicators (20). Age may have an influence on the initial dose and compliance to imatinib but the real impact on overall survival was related to comorbidities and to a higher risk of death from other causes, unrelated to CML, confirming the findings of a study on 181 patients aged over 75 years observed in real life practice (21). However, in the EUTOS registry, older age, more peripheral blasts, an enlarged spleen, and low platelet counts were significantly associated with an increased probability of dying of CML (8).

Comorbidities may affect survival and the choice of treatment in CML. A first observation analyzed the Charlson comorbidity index in 125 older CML patients treated with dasatinib in relation to compliance and the onset of pleural effusions, showing a direct association between a higher comorbidity score and drug-related side effects (22). The role of comorbidities and the prognostic role of Charlson index stratification on CML outcomes was further assessed by the German group: 1519 patients entered this analysis, and no differences were detected in terms of cumulative incidences of accelerated and blast phase or remission rates in the different groups. Indeed, higher scores according to the Charlson index were significantly associated with lower overall survival probabilities, with an 8-year survival of only 46.4% in patients with score >7 versus 93.6% in patients with score 2 (23). The presence of comorbidities may also be correlated with the risk of developing adverse events with TKIs. In particular, this association is highlighted by the increasing risk of arterio-occlusive events in patients with other pre-existing cardiovascular risk factors (18). As shown also in our study, in CML patients the assessment of comorbidities and the baseline age may affect TKI selection and dictates close follow-up to optimize the TKI dose.

The estimated crude effects of the variables analyzed were all strongly associated to OS, with a HR of 0.22 in favor of the 2GTKI. Even adjusting the comparison between these drugs by applying a multivariable Cox model, the protective effect of a 2GTKI on overall survival remained remarkable, with an HR of 0.50. However, the proportion of patients who died of CML-related causes was similar in the two cohorts: 1.8% in the imatinib cohort and 1.7% in the 2GTKI cohort. Furthermore, considering the CML-related deaths separately from those due to other causes, treated as competing events, no difference could be detected between treatments in terms of the cumulative risk of CML mortality: all the difference observed for OS was attributable to an increased mortality due to other causes. The analyses performed with the Fine and Gray model, that can account for competing events, clarified the role of the study variables on CML-related death or other causes death. These results did not confirm any advantage of 2GTKIs on CML mortality, in line with the results of the meta-analyses of randomized trials (4). However, these agents induced faster and deeper molecular responses but without differences in overall survival as compared to imatinib (24–26). The rationale to start a 2GTKI was supported by a subanalysis of the ENESTnd, that reported an increased rate of sustained deep molecular response with the 2GTKI when started as front-line treatment as compared to imatinib, allowing an increased proportion of patients to become candidates for a possible discontinuation over time (24). However, the optimization of the dose to reduce the possible long-term off-target events, in particular the cardiovascular side effects associated to these drugs, and the effect of treatment discontinuation, are still a matter of debate.

The principal strength of this large prospective and representative national cohort is that the TKI choice was performed according to a “patient-centered approach”, considering at baseline the prognostic role of age, concomitant comorbidities, prognostic score stratification and all the possible concomitant factors that could have influenced the adherence in the long-term (27, 28). Although the comparison between treatments has been adjusted for a set of pre-defined important prognostic factors, and the causes of death analyzed separately, the observational study design does not allow the role of uncontrolled or residual confounding to be excluded.

In conclusion, the analysis conducted showed that some specific clinical factors could be predictive of long-term overall survival in CML patients treated with TKIs. In particular, the comorbidity profile and the stratification by the ELTS score have to be considered at baseline as the mainstay on which to base decisions about the therapeutic strategy best suited to each patient. No difference between IMA and 2GTKI was observed for CML-related mortality.

The data analyzed in this study is subject to the following licenses/restrictions: Observational study. Requests to access these datasets should be directed to gianni.ciccone@cpo.it.

The studies involving human participants were reviewed and approved by University of Bari. The patients/participants provided their written informed consent to participate in this study.

GSp, PP, MBr, GSag, and GCi collected, analyzed and wrote the manuscript. GCa and CM made the statistical analysis. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.