Prognosis of patients with chronic myeloid leukemia presenting in advanced phase is defined mainly by blast count, but also by age, chromosomal aberrations and hemoglobin

Chronic myeloid leukemia (CML) is usually diagnosed in chronic phase, yet there is a small percentage of patients that is diagnosed in accelerated phase or blast crisis. Due to this rarity, little is known about the prognosis of these patients. Our aim was to identify prognostic factors for this cohort. We identified 283 patients in the EUTOS population‐based and out‐study registries that were diagnosed in advanced phase. Nearly all patients were treated with tyrosine kinase inhibitors. Median survival in this heterogeneous cohort was 8.2 years. When comparing patients with more than 30% blasts to those with 20‐29% blasts, the hazard ratio (HR) was 1.32 (95%‐confidence interval (CI): [0.7‐2.6]). Patients with 20‐29% blasts had a significantly higher risk than patients with less than 20% blasts (HR: 2.24, 95%‐CI: [1.2‐4.0], P = .008). We found that the blast count was the most important prognostic factor; however, age, hemoglobin, basophils and other chromosomal aberrations should be considered as well. The ELTS score was able to define two groups (high risk vs non‐high risk) with an HR of 3.01 (95%‐CI: [1.81‐5.00], P < .001). Regarding the contrasting definitions of blast crisis, our data clearly supported the 20% cut‐off over the 30% cut‐off in this cohort. Based on our results, we conclude that a one‐phase rather than a two‐phase categorization of de novo advanced phase CML patients is appropriate.

treated with tyrosine kinase inhibitors. Median survival in this heterogeneous cohort was 8.2 years. When comparing patients with more than 30% blasts to those with 20-29% blasts, the hazard ratio (HR) was 1.32 (95%-confidence interval (CI): [0. 7-2.6]). Patients with 20-29% blasts had a significantly higher risk than patients with less than 20% blasts (HR: 2.24, 95%-CI: [1.2-4.0], P = .008). We found that the blast count was the most important prognostic factor; however, age, hemoglobin, basophils and other chromosomal aberrations should be considered as well. The ELTS score was able to define two groups (high risk vs non-high risk) with an HR of 3.01 (95%-CI: [1.81-5.00], P < .001). Regarding the contrasting definitions of blast crisis, our data clearly supported the 20% cut-off over the 30% cut-off in this cohort. Based on our results, we conclude that a one-phase rather than a two-phase categorization of de novo advanced phase CML patients is appropriate.

| INTRODUCTION
Chronic myeloid leukemia (CML) is a myeloproliferative disorder characterized by a reciprocal chromosomal translocation resulting in the BCR-ABL1 fusion gene. The disease is traditionally described in three distinct clinical phases: chronic phase (CP), accelerated phase (AP) and blast crisis (BC). 1 The exact mechanism that underlies the transformation from a CP to aggressive BC still remains a key biological question for the future. 2 The accumulation of distinct additional cytogenetic abnormalities (ACAs) 3,4 and a variety of mutations [5][6][7] were associated with disease progression. Current models of disease progression predict that increased BCR-ABL1 expression plays an important role in the secondary molecular and chromosomal changes which precede disease transformation. 8 Although the triphasic course of CML is generally well recognized, the precise definition of these three phases varies greatly in the T A B L E 1 Staging classification systems for CML ELN definitions 23 WHO definitions 33 Chronic phase None of the following criteria As these data suggest that patients diagnosed in the advanced phase of the disease seem to be a very heterogeneous group in terms of their prognosis, it is essential to identify characteristics that are predictive for their outcome. All of the well-established CML prognostic scores like the Sokal, 19 the Euro, 20 EUTOS 21 or the ELTS score 22 were developed and validated for CML patients in chronic phase. However, whether these scores are useful to categorize patients who are diagnosed in advanced phase of CML has not been investigated so far. Therefore, the aim of our work was to analyze the outcome of patients diagnosed in AP or BC and to identify prognostic factors associated with their survival.

| Definitions
In the registries, accelerated phase and blast crisis were defined according to the ELN criteria. 23 Particularly, the cut-off between AP and BC was 30% blasts in the blood or bone marrow. Of note, investigators were asked to provide the disease phase of the patients, but not all variables involved in the definition of the phases were part of the case report form. As the registry was to be manageable also for smaller centers in whole Europe, the data set had to be restricted. Particularly, data on parameters in bone marrow were not collected. All results with respect to immature cells reported in this work relate to peripheral blood. Clonal chromosomal abnormalities in Ph + cells were counted as ACAs. Only patients who were presenting in advanced phase were considered.

| Statistical analysis
Correlation between candidate variables was investigated by use of Spearman's rank correlation coefficient and scatter plots, Mann-Whitney U test, or Fisher's exact test, whichever was appropriate.
Overall survival was counted from the date of diagnosis. Survival time of patients alive at the last follow-up was censored. As none of the patients was (by definition) transplanted in first chronic phase, patients were not censored at the date of an allogeneic stem cell transplantation. Survival probabilities were estimated using the Kaplan-Meier estimator and compared using the log-rank test. For the multivariate analysis, the Cox proportional hazards model was used.
All P values <.05 were considered significant. Model selection was done using Akaike's Information Criterion. 24 Due to the exploratory character of this work, no P value adjustment was applied; thus all Pvalues have to be interpreted descriptively. An external validation of these results by another research group would be welcome. The description of the registry can be found elsewhere. 26 The PBR comprised data on 188 patients that were diagnosed in advanced phase. 15 In the OSR, we identified 117 patients diagnosed in advanced phase. For this analysis, we asked for an update of the followup. As a result, three patients had to be excluded due to lacking samples before start of treatment, and two entries proved to be double records of the same patient. Also, for six patients no follow-up was available and for 11 patients, it was not known whether they were in advanced phase at diagnosis or not. Thus, the final sample size consisted of 283 patients, of whom 203 were diagnosed in AP and 80 were diagnosed in BC. The flowchart is given in supplementary Figure 1.

| Regional distribution
The patients' regional distribution is shown in Table S1. Of note, only the Czech Republic, Poland, Romania, Russia, Spain, and the United Kingdom were part of the OSR. In the PBR, all participating countries except Slovenia contributed at least one patient to the updated advanced phase data set. In the PBR, either the whole country (for smaller countries with a population of less than 10 millions) or selected regions with a maximum of 10 million inhabitants (for larger countries) were part of the registry. 26 Thus, in this study, the number of patients per country cannot be directly connected to the country's population. Furthermore, a comparison between countries was not possible, as in some countries, the registration was required by law, while in others, the active participation of the physicians was needed.

| Baseline values
The sample consisted of 151 male (53%) and 132 female patients (47%). Median age was 51 years (range:  but not between the high-risk and the low-risk group. The Euro score 20 did not provide any prognostic discrimination of survival. The EUTOS score 21 was able to identify two different groups; however the high-risk group had a considerably better survival probability than the low-risk group. Both in patients with AP as well as with BC, the ELTS score 22

| DISCUSSION
In this cohort of advanced phase patients, the main objectives were the description of survival probabilities in patients diagnosed in AP or BC, and to consider potential prognostic factors related to outcome.
To our knowledge, this study based on data of centers all over Europe is the first large study on this cohort.
The median survival of the AP patients in the present data was much better than in patients that developed an AP from a CP. This was reported in trials from the GIMEMA group where median survival was 37 months, 27 from the MD Anderson group with a 4-year survival probability of 53%, 17 and from China with a 6-year survival probability of 51%. 28 It seems that patients with de novo AP have a different biological background as compared with patients transferring from CP to AP while under therapy. For some of the patients here, de novo diagnosis of AP might have been due to a very late diagnosis of CML, however, without anyone carrying the burden of a history of resistance to TKI. In contrast to patients that progressed from CP to AP because they were not responding to TKI, newly diagnosed patients in AP were reported to show good response when initially treated with TKI, albeit this was observed in small studies. 29,30 In conclusion, AP at diagnosis should be clearly distinguished from AP after CP.
Also the survival of patients with de novo diagnosis of BC was longer than for patients with a BC that developed during the course of the disease. After transformation from CP to BC, median survival in the German CML-study IV was 7.9 months. 31 In a study by the MD Anderson group, 32 median survival for de novo BC patients was more than 2 years, while median survival for BC patients originally diagnosed in CP or AP was less than 1 year. In a population-based setting, the Swedish CML register has recently published a median survival of de novo BC patients of 1.6 years. 30 Even though the differences were smaller than for the AP patients, transferring results from de novo BC patients to patients that develop a BC in the course of the disease does not seem appropriate.
Unfortunately, no information on bone marrow was available, as this information was not collected within the registries. For this reason, we refrained from the attempt to combine statistically significant and clinically relevant factors associated with overall survival in a prognostic sys-  (Figures 3 and 4). Lacking a more appropriate prognostic tool, the ELTS score could be applied until further notice. However, a prognostic score developed in advanced phase patients remains a desirable aim for the future.
Apart from the lack of data on bone marrow, our patient sample was subject to further limitations. Data came from two observational registries where treatment was not standardized, and we expect that the percentage of patients treated with second-generation TKI in this cohort might have increased in the meantime. Besides, the vast differences in the incidence of CML diagnoses in advanced phases indicate considerable heterogeneity between the countries. This heterogeneity could be due to a more or a less strict application of the criteria defining the advanced phases, or due to the tendency in a certain country to see a physician earlier rather than later. With the lack of data on bone marrow, it was not possible to verify the phase reported by the investigators for each of the patients. On the upside, heterogeneity means also strength.
Results do not depend on data from just one study group or even one single center, but are based on patients from all over Europe supporting their generalizability. Furthermore, to our knowledge, this is the largest data set including patients diagnosed in advanced phase.
In this study, the ELN criteria 23  between chronic phase and blast crisis at diagnosis. As seen in our data, patients with late chronic phase will be identified as high-risk patients by the ELTS score and thus, will still receive particular attention by physicians. To achieve a better understanding of CML blast crisis, and to improve treatments and outcomes, an international blast crisis registry was launched. It is a major aim of this collaborative project to systematically collect baseline, treatment and outcome data of patients diagnosed with CML blast crises. This is to get more information on the biology, prognosis and treatment of this disease. The registry could also provide an empirical base for future national and international trials.