Imatinib as the first‐line treatment of patients with chronic myeloid leukemia diagnosed in the chronic phase: Can we compare real life data to the results from clinical trials?

Abstract

Imatinib (IM) dramatically improved the prognosis of chronic myeloid leukemia (CML), particularly with newly diagnosed patients in a chronic phase (CP) [1]. The most robust source of data about IM efficacy in this setting is the IRIS trial. However, every day clinical practice data are still scarce. We analyzed IM efficacy and safety in the first-line therapy of 152 consecutive adult CP-CML patients from a defined region. The estimated 4-year cumulative incidences of complete hematologic, complete cytogenetic, major, and complete molecular responses were 95.3%, 80.6%, 65.4%, and 39.2%, respectively. The 4-year probability of overall and progression-free survival (PFS) defined as with the IRIS [2] was 91.5% and 78.1%, respectively. We thus confirmed very good IM efficacy also in patients not participating in clinical trials. However, the estimated 4-year event-free survival (EFS), which also counted failure events according to valid recommendations [3] or IM discontinuation due to intolerance, was only 60.7%. The 4-year probability of an alternative treatment-free survival, our newly defined parameter, which better reflects the proportion of patients remaining on IM despite an event, was 67.6%. Therefore, more appropriate selection and unification of survival analyses end-points is desirable to describe and compare IM real efficacy. Based on results from the International Randomized Study of Interferon versus STI571 (IRIS), IM is currently recommended as front-line therapy for patients with CML in CP [2-5]. However, clinical trial results extrapolation to an entire patient population is difficult and sometimes nearly impossible due to significant differences between in-study and out-study patient populations, as was shown with CML [6] and also with other malignant diseases [7]. Despite several years of experience with IM use in first-line CML treatment, real-life data have been scarce [8], and some data did not confirm excellent results achieved during the IRIS trial [9]. Another factor making clinical trial results extrapolation and comparison difficult is a wide variability of survival analyses end-point definitions [2, 4, 8, 10, 11]. In the Czech Republic, the care of CML patients is concentrated at only six specialized centers, which cover a virtually complete population from a defined region. Informative real life data can thus be obtained. Between July 2003 and July 2009, 152 consecutive adult patients with CML in CP received IM as first-line therapy in two of the largest hematooncological centers in the Czech Republic. The baseline characteristics are presented in Table I. The median follow-up on IM treatment was 31.2 months (range, 5.7–68.1). The median of mean actually administered daily dose was 400 mg. Dose was escalated to 600–800 mg/day in 23 (15.1%) of patients, mainly because of inadequate response. The dose was reduced or treatment interrupted with 93 (61.2%) of patients, mainly because of side effects. In total, 36 patients (23.7%) permanently discontinued IM after a median of 16.0 months (range, 1.0–51.1) since the start of therapy for various reasons: Progression or IM failure (n = 20), elective allogeneic transplantation (n = 4), IM intolerance (n = 9), and death from non-CML-related causes (n = 3). IM efficacy is presented in Fig. 1. At first, treatment responses on hematologic, cytogenetic, and molecular levels achieved at given time points are presented (Fig. 1A). Patient disposition reflecting treatment response definition according to European LeukemiaNet (ELN) recommendations [3] is shown in Fig. 1B. The 4-year cumulative incidence of CHR was 95.3% (95% CI: 91.7–98.9%), CCgR 80.6% (95% CI: 72.9–88.2%), major molecular response (MMolR) 65.4% (95% CI: 56.7–74.1%), and complete molecular response (CMolR) 39.2% (95% CI: 27.6–50.7%) (Fig. 1C). Profile of therapeutic responses reached during follow-up. A: CHR, CgR, and MolR response rates at given time points. B: Disposition of treatment responses according to ELN criteria [3]. C: Cumulative incidence of responses to IM. The 4-year cumulative incidence of CHR was 95.3%, CCgR 80.6%, MMolR 65.4%, and CMolR 39.2%. In total, nonhematological and hematological toxicities of all grades occurred in 119 (78.3%) and 95 (62.5%) patients, respectively. The toxicity profile was comparable to published experience, with most adverse events being of grades 1 or 2 and manageable by dose reduction or treatment interruption. Grades 3 and 4 of nonhematological and hematological toxicities occurred in 13 (8.5%) and 19 (12.5%) patients, respectively. Figure 2A shows a comprehensive set of time-to-event analyses. Estimated overall (OS) and transformation-free survival (TFS) at 4 years were 91.5% (95% CI: 84.8–98.2%) and 88.4% (95% CI: 80.4–96.3%), respectively. In total, seven patients died (4.6%) in a median of 27 months (range, 13.1–38.8) since the start of IM—three patients from CML-unrelated reasons and four patients from disease progression. Probability of PFS—defined according to first IRIS report [2]—was 78.1% (95% CI: 68.0–88.2%) at 4 years. However, this IRIS definition has not included events, which fulfilled criteria for failure according to ELN [3] and also discontinuations of IM due to intolerance. After including these important events into the definition of EFS, the 4-year probability of being in CCgR still receiving IM was only 60.7% (95% CI: 48.3–73.2%). At the time of data analysis, 10 patients among those with events (40; 26.3%) continued on IM, and the remaining 30 patients stopped IM before the date of analysis. The reasons for continuing on IM despite the event were various, including patient preference to stay on IM in case of less than CCgR at 18 months and attempt to achieve response with IM dose escalation. For the estimation of patients proportion actually receiving IM in spite of possible event, we performed analysis called alternative treatment-free survival (ATFS). The 4-year probability of remaining on first-line IM treatment in real clinical practice was 67.6% (95% CI: 56.9–78.3%). Comprehensive set of time-to-event analyses. A: Estimated OS, TFS, PFS, ATFS, and EFS for entire group of analyzed patients (N = 152) at 4 years were 91.5%, 88.4%, 78.1%, 67.6%, and 60.7%, respectively. B: Estimated 4-years probability of OS, TFS, PFS, ATFS, and EFS for the subset of patients achieving either partial or complete cytogenetic response at 6 months (N = 94) was 90.9%, 90.1%, 77.7%, 74.0%, and 67.1%, respectively. The 4-year probability of being alive, progression-free, alternative treatment-free, and event-free was calculated also for the subset of patients achieving at least partial cytogenetic response or CCgR at 6 months of IM therapy (N = 94) (Fig. 2B). With OS of 90.9% (95% CI: 81.6–100%) and PFS of 77.7% (95% CI: 63.9–91.6%), there were virtually no differences in comparison with the entire group of analyzed patients (N = 152). On the other hand, EFS (67.1%; 95% CI: 51.2–83%) and ATFS (74%; 95% CI: 60.4–87.5%) calculated for the subset of optimal responders at 6 months showed better results than for all analyzed patients. With median follow-up of 60 months, the cumulative incidences of CHR and CCgR reported with the IRIS trial were 98% and 87%, respectively, and the 5-year estimated OS, PFS, and EFS were 89%, 93%, and 83%, respectively [4]. It should be noted that PFS and EFS definitions used in the 5-year update [4] differ from those published in the first report on IRIS trial results [2]. Originally, PFS was defined as survival without evidence of AP or BC, loss of CHR, loss of MCgR, increasing white blood cell count (in patients who never achieved CHR), or death from any cause while on IM treatment [2], and we also used this original definition for our analysis purpose. EFS was not originally defined at all. In the work published by Druker et al. [4], PFS means survival without progression into AP or BC (which was defined as TFS in this presented report), and EFS is the former PFS. Another confusing issue is the fact that a considerable proportion of patients (28%) discontinued first-line IM therapy for a variety of reasons, which were censored and thus not counted as events in survival analyses [4]. De Lavallade et al. [8] performed an intention-to-treat analysis on 204 consecutive patients with newly diagnosed CML treated with IM with the aim to validate the IRIS data. Five-year cumulative incidences of CHR, CCgR, MMolR, and CMolR were 98.5%, 82.7%, 50.1%, and 8.3%, respectively, and the 5-year estimated OS and PFS (defined as in IRIS first report) [2] were 83.2% and 82.7%, respectively. On the opposite to this single-center experience, Lucas et al. [9] in a population-based analysis of 68 patients with IM as their first-line therapy from the defined area of 2 million people did not verify these results, with overall CCgR and MMolR rates at 24 months only 51% and 31%, respectively. In our analysis, with 4-year estimated probability of CHR, CCgR, MMolR, and CMolR reaching 95.3%, 80.6%, 65.4%, and 39.2%, respectively, and 4-year estimated OS and PFS 91.5% and 78.1%, respectively, our results were similar to those achieved in IRIS with comparable median follow-up [4] and to those presented in a single-institution report [8]. In terms of the cumulative incidence of MMolR and CMolR, our results were even better than reported in the latter. As for the tolerability of IM, there were no meaningful differences between our report and others cited. Similar or even improved IM efficacy results among our patients can be explained in part by the shorter follow-up and in part by the fact that the care of such patients is concentrated in a limited number of experienced centers in the Czech Republic where well-standardized techniques for treatment efficacy monitoring are used, and current recommendations for the management of CML patients are maintained. This real-life situation is also reflected in the time-to-event analyses in our report. PFS, as originally defined in the first IRIS trial report [2], does not contain the ELN criteria for IM failure or its discontinuation due to intolerance [3]. Using the more appropriate EFS definition in our analysis, the 4-year probability of being in CCgR while remaining on IM was 60.7%, which was consistent with the Hammersmith study data [8] with 5-year probability of 62.7% remaining event-free. Both EFS results thus impair the optimistic expectations about IM efficacy. In our cohort, a significant proportion of the patients with event (10/40; 25%) continued on IM at the time of the latest assessment. In this light, our newly defined parameter—ATFS (4-year probability 67.6%)—drew a more realistic picture of the probability of remaining on first line IM treatment in every day clinical practice. In summary, we confirmed very good IM efficacy and tolerability in first line CML therapy as well as with the patient population not participating in clinical trials that were treated at well-managed specialized centers. However, in spite of similar results to those obtained in the large multicenter trial in terms of OS, TFS, and PFS—with more appropriate selection of events for calculation of EFS and ATFS—overestimation of time-to-event analysis in IRIS trial is obvious. The data concerning all consecutive adult patients with newly diagnosed CP-CML treated with IM in the two largest hematological centers in the Czech Republic between July 2003 and July 2009 were stored in a detailed database called INFINITY (tyrosine kinase Inhibitors in First and following CML Treatment). The database design was reviewed and approved by the ethics committee at each participating center, and the patients were required to give written informed consent. Patients were treated with IM 400 mg orally once daily and had no prior treatment for CML except hydroxyurea, anagrelide, or leukapheresis. IM dose was modified mainly because of toxicity [12] or inadequate response [3]. Cytogenetic and molecular studies were performed according to ELN recommendations [3, 5, 13]. Conventional cytogenetic analysis was done using the G-banding technique, and at least 20 metaphases were analyzed. A FISH analysis was used for cytogenetic response evaluation only in cases of poor quality samples or an insufficient number of assessable metaphases [14]. An optimized multiplex reverse transcription-polymerase chain reaction (RT-PCR) was adopted from Cross et al. [15] to determine the type of BCR-ABL transcript. Real-time quantitative PCR (RQ RT-PCR), standardized within the international program EUTOS for CML [16], was performed according to Europe Against Cancer recommendations [17, 18]. B2M or ABL were applied as housekeeping genes in the Prague and Brno laboratories, respectively. The CMolR was defined here as undetectable BCR-ABL transcript by qualitative nested PCR with the sensitivity of 1 BCR-ABL positive cell in 106 leukocytes. Treatment responses were evaluated and defined according to valid recommendations [3]; toxicity of IM was assessed according to CTCAEv.3. OS was defined as the time from the start of IM to death from any cause in spite of IM discontinuation. TFS was defined as survival without evidence of AP or BC or death from any cause during IM therapy. PFS was defined as in the IRIS trial [2], that is, survival without evidence of AP or BC, loss of CHR, loss of MCgR, increasing white blood cell count (in patients who had never had CHR), or death from any cause while on IM treatment, whichever came first. In EFS, events were defined as a progression (the same as in PFS, as described earlier), loss of CCgR, failure to achieve CHR at 6 months, MCgR at 12 months, and CCgR at 18 months, or intolerance of IM as the cause of its discontinuation, whichever came first [3, 8, 10]. ATFS was defined as time since start of IM to change to any alternative treatment or death from any cause during IM therapy. Survival probabilities were estimated using the Kaplan–Meier method. Patients who underwent stem cell transplantation were censored in all above-mentioned analyses with the exception of ATFS. Point estimates were supplied with corresponding 95% confidence intervals. The probabilities of hematologic, cytogenetic, and molecular responses were estimated using the cumulative incidence method. Analyses were performed using SPSS software (version 18.0.1), MedCalc (version 10), and R software for statistical computing and graphics (version 2.10.1). In particular, we thank data-managers Blanka Dobesova and Jana Rysava for the data collection. We also thank Rich Zimmerman for the English correction of the manuscript. This work was supported by CELL—The CzEch Leukemia Study Group for Life.