Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Chronic myelogenous leukemia

Prognosis of long-term survival considering disease-specific death in patients with chronic myeloid leukemia

Abstract

In patients with chronic myeloid leukemia (CML), first-line imatinib treatment leads to 8-year overall survival (OS) probabilities above 80%. Many patients die of reasons unrelated to CML. This work tackled the reassessment of prognosis under particular consideration of the probabilities of dying of CML. Analyses were based on 2290 patients with chronic phase CML treated with imatinib in six clinical trials. ‘Death due to CML’ was defined by death after disease progression. At 8 years, OS was 89%. Of 208 deceased patients, 44% died of CML. Higher age, more peripheral blasts, bigger spleen and low platelet counts were significantly associated with increased probabilities of dying of CML and determined a new long-term survival score with three prognostic groups. Compared with the low-risk group, the patients of the intermediate- and the high-risk group had significantly higher probabilities of dying of CML. The score was successfully validated in an independent sample of 1120 patients. In both samples, the new score differentiated probabilities of dying of CML better than the Sokal, Euro and the European Treatment and Outcome Study (EUTOS) score. The new score identified 61% low-risk patients with excellent long-term outcome and 12% high-risk patients. The new score supports the prospective assessment of long-term antileukemic efficacy and risk-adapted treatment.

Introduction

For patients with Philadelphia chromosome-positive chronic phase chronic myeloid leukemia (CML), the use of the BCR–ABL1 tyrosine kinase inhibitor (TKI) imatinib led to 8-year overall survival (OS) probabilities of >80%.1, 2, 3 A consequence of the improved survival induced by imatinib is the increased probability of dying of causes other than CML. Of 152 deaths recorded in the German CML-study IV, 41% were assessed as ‘not directly CML related’.2 Investigators wonder to what extent survival probabilities still depend on CML. Owing to the therapeutic success, large patient samples and long observation times are needed to distinguish imatinib-treated patient groups with significantly different OS. This is even more of a concern when the aim is to discriminate between causes of death.

To study treatment outcome, a registry of CML patients was established by the European LeukemiaNet (ELN) and maintained within the framework of the European Treatment and Outcome Study for CML (EUTOS).4 The registry is divided into the in-study and the out-study section.5 For both sections, patients were prospectively registered. As the only contrast to the in-study section, patients of the out-study section were not enrolled in controlled clinical trials. Now, with median observation times above 5.5 years, the two study sections have allowed a more detailed consideration of long-term survival.

The main topic of our work was the prognosis of dying because of CML. Baseline prognostic scores serve a variety of important tasks, not only outcome prediction for individual patients but also selection and development of risk-adjusted treatments, adjustment of imbalances between treatment groups in clinical trials and comparative assessment of outcomes of different studies.6, 7 Since 2000, treatment allocation and/or analysis based on risk stratification have been an essential component of the design of all major studies investigating CML treatment.8, 9, 10, 11, 12, 13, 14

Based on patient characteristics at diagnosis of CML, the Sokal,15 the Euro7 and the EUTOS4 score are established prognostic scores covered by the ELN management recommendations.16 Calculated before any therapy is initiated, the scores were meant for predicting treatment outcome of Philadelphia chromosome-positive patients in non-blastic15 or chronic phase.4, 7 The Sokal and the Euro score were developed to discriminate three risk groups with significantly different OS probabilities for chemotherapy-treated15 and interferon alpha-treated7 patients, respectively. Introduced to distinguish complete cytogenetic remission probabilities at 18 months in imatinib-treated patients, the EUTOS score was also able to identify two groups with significantly different OS.5 Now, for the first time, each score was evaluated for its ability to discriminate probabilities of dying of CML in large multinational patient samples. The possibility of further discrimination improvement was assessed by the development of a new score in the in-study data. The new score was tested in the out-study data.

Patients and methods

Patients

The in-study and out-study sections of the ELN/EUTOS CML registry contain individual data on adult patients who were prospectively enrolled between 2002 and 2006.4, 5 Further patient eligibility criteria for both registry sections were diagnosis of Philadelphia chromosome-positive and/or BCR–ABL1-positive CML in chronic phase, transcript type b2a2 and/or b3a2, and any form of imatinib-based treatment within 6 months from diagnosis.4, 5 In the analysis described here, the out-study data were solely used for external validation of a new score. For all analyses, the registries’ data as of 7 September 2014 were used.

Definitions and end points

Survival time was calculated from the start date of imatinib treatment to death or to the latest follow-up date. Progression-free survival time was calculated like survival time but ended with the observation of progression. Progression was defined by the observation of accelerated phase or blast crisis, with both phases determined according to the ELN criteria.16 Any survival end point was censored at the time of allogeneic hematopoietic stem cell transplantation (HSCT) in first chronic phase; because of transplant-related mortality, survival probabilities after HSCT in first chronic phase were considered not to be representative for survival probabilities had TKI treatment been continued. Chronic phase was defined by the absence of progression.16 With the limited survival prospects after progression both when receiving TKI or HSCT, survival end points were not censored at the time of allogeneic HSCT in advanced phase when TKI treatment had already failed. Only death after recorded disease progression was regarded as ‘death due to CML’. Death without prior disease progression was rated as ‘death unrelated to CML’. For details regarding the calculation of the Sokal,15 the Euro7 and the EUTOS4 score, see Supplementary Table 1.

Statistical analysis

OS and progression-free survival were calculated by the Kaplan–Meier method and compared by the log-rank test.17 When differentiating competing causes of death, cumulative incidence probabilities of dying were obtained using the Aalen-Johansen estimator18, 19 and compared by the Gray test.20 To estimate relative survival, OS probabilities in our data were adjusted by survival probabilities from matched population data according to the method of Pohar-Perme.21, 22 For each year of start of imatinib treatment and each country of study origin given in the in-study registry, population data were downloaded from the Human Mortality Database (www.mortality.org) on 5 May 2014.23 The Swedish data were used for the Nordic study group since the origin of an individual patient of the Nordic study group was not included in the registry and no relevant differences in survival between the populations of the Nordic countries (Denmark, Finland, Norway and Sweden) were found in the mortality database. Apart from year of treatment start and country, the population data were matched to our patient sample with regard to sex and age at treatment start. Relative survival was interpretable as the survival resulting from the excess hazard of dying because of CML.

Under consideration of fractional polynomials,24 the influence of potential prognostic factors on the subdistribution hazard of dying because of CML was analyzed by the Fine-Gray model.25 Candidate prognostic factors were age, sex, spleen size below costal margin, hemoglobin, white blood cell count, blasts, basophils, eosinophils (all in peripheral blood) and platelet count. For units, see Table 1. Multiple regression modeling was based on patients with complete data on all candidate factors. For variable selection, the Akaike information criterion was applied.26 To aid medical decision making, the linear predictor of the final model was categorized into risk groups. For identification of the cutoff(s), bootstrap resampling,27 the minimal P-value approach28 and a kernel density estimator29 were used. The ability of the new score to provide risk groups with significantly different cumulative probabilities of dying of CML was assessed. Besides variable selection and the use of the minimal P-value approach, for the two-sided P-values the unadjusted significance level of 0.05 was applied at all statistical tests.

Table 1 Characteristics of the 2290 in-study registry patients at diagnosis

Code availability

Most analyses were undertaken with SAS (version 9.2, SAS Institute Inc., Cary, NC, USA). Fractional polynomials modeling was performed with a particular SAS macro.30 For estimation of relative survival probabilities and the use of the Gray test and the Fine-Gray model (with the Akaike information criterion), functions of the programming software R (version 3.0.3.3) were applied.22, 26, 31 For more details regarding regression modeling and cutoff identification, see Supplementary methods.

Ethics

All studies complied with the Declaration of Helsinki. They were approved by the local human investigations committee and performed in accordance with the legal requirements of the corresponding country. Informed consent was obtained from all patients.

Results

Baseline characteristics and survival in 2290 in-study patients

Of 2410 patients captured by the in-study registry, 2290 fulfilled the inclusion criteria and came from a Dutch32 (n=117), French11 (n=562), German12 (n=735), Italian33 (n=544), Nordic34 (n=137) or Spanish35 (n=195) study group. Median age was 51 years; 60% were male (Table 1). Allogeneic HSCT in first chronic phase was performed in 99 patients (4%). With a median follow-up of 6.3 years (range: 0.1–9.7 years), the 8-year probability of OS was 89% (95% confidence interval (CI): 87–90%) and of progression-free survival 87% (95% CI: 85–89%). After adjustment by matched population data, the 8-year relative survival probability with CML was 96% (95% CI: 93–97%) (Figure 1a).

Figure 1
figure1

Overall, progression-free, relative survival (a) and cumulative incidence probabilities of causes of death related and unrelated to CML (b) in 2290 patients from the in-study registry. At 2, 5 and 8 years, horizontal crossbars indicate the upper and lower limit of the 95% CI for the estimated probability.

Causes and cumulative incidence probabilities of dying

Without prior HSCT in first chronic phase, 208 of 2290 patients (9%) had died. Cause of death was CML in 92 cases (44%), non-CML-related in 104 cases (50%) and unknown in 12 cases (6%). Of the non-related cases, 50 patients (24% of 208) died of a second cancer, 21 (10%) of a cardiovascular event, 12 of an infection (6%) and 21 (10%) of other causes. More details on the other causes are listed in Supplementary Table 2. The 8-year cumulative incidence probability of CML-related death was 4% (95% CI: 4–5%) and of non-CML-related death 7% (95% CI: 5–8%), including the unknown causes where no progression before death was observed (Figure 1b).

Influence of treatment on the probabilities of dying of CML

Stratified for study group, the kind of allocated first-line imatinib treatment (Table 1) had no significant influence on the cumulative incidence probabilities of dying of CML. Compared with imatinib 400 mg monotherapy, it made no difference whether allocation was to imatinib monotherapy given in a higher dose than 400 mg or to imatinib 400 mg in combination with another treatment.

Probabilities of dying of CML in the risk groups of the established scores

The ability to discriminate probabilities of dying of CML was examined in the 2205 registry patients whose risk group was evaluable for all three prognostic scores. All three scores identified a high-risk group with significantly higher cumulative incidence probabilities of dying of CML than any of the non-high-risk groups (always P<0.001, apart from P=0.031 for intermediate- versus high-risk group according to the Sokal score, Figure 2). Neither the Sokal nor the Euro score found significantly different probabilities of dying of CML between low- and intermediate-risk patients. With the low-risk group as reference, the subdistribution hazard ratios (SHRs) of the intermediate- and high-risk group were 1.642 (95% CI: 0.955–2.823) and 2.796 (95% CI: 1.625–4.810) with the Sokal score and 0.836 (95% CI: 0.507–1.377) and 3.452 (95% CI: 2.029–5.872) with the Euro score (Table 2). Using the EUTOS score, the SHR of the high- to the low-risk group was 2.428 (95% CI: 1.446–4.076).

Figure 2
figure2

Cumulative incidence probabilities of dying because of CML in 2205 patients from the in-study registry stratified for the risk groups according to the Sokal score (a), the Euro score (b) and the EUTOS score (c). At 2, 5 and 8 years, horizontal crossbars indicate the upper and lower limit of the 95% CI for the estimated probability.

Table 2 Probabilities of dying of CML according to the risk groups of four different prognostic scores

Development of a new score in the in-study data

Based on all 2205 evaluable patients, the cumulative incidence probabilities of dying of CML were significantly increased by higher age (P=0.009), a bigger spleen size below costal margin (P<0.001), a higher percentage of peripheral blasts (P=0.005) and low platelet counts (P=0.022, Table 3). After polynomial transformation, age and platelet count had been rounded to three decimal places. Together with their estimated regression coefficients, the four factors were combined in a new prognostic score:

Table 3 Significant prognostic factors for death due to CML in 2205 patients evaluable for the variables of the best model

ELTS score=0.0025 × (age/10)3 + 0.0615 × spleen size below costal margin + 0.1052 × blasts in peripheral blood + 0.4104 × (platelet count/1000)0.5.

For a detailed description of regression results for ‘death unrelated to CML’, see Supplementary results and Supplementary Tables 3 and 4.

Probabilities of dying of CML in three risk groups of the new score

The EUTOS long-term survival (ELTS) score was rounded to four decimal places. Subsequently, one thousand bootstrap samples of the size n=2205 were drawn from the original data set. After P-value adjustment for multiple testing, in 997 samples, a statistically significant best cutoff was identified. A smoothing function of these 997 cutoffs resulted in a three-headed kernel density (Supplementary Figure 1). The cutoffs at the two highest peaks led to the definition of a low-risk group (ELTS score 1.5680, n=1349, 61%), an intermediate-risk group (1.5680⩽2.2185, n=596, 27%), and a high-risk group (ELTS score>2.2185, n=260, 12%). Referring to the low-risk group, the patients of the intermediate-risk group (SHR: 2.996 (95% CI: 1.800–4.987), P=0.014)) and of the high-risk group (SHR: 5.627 (95% CI: 3.271–9.681), P<0.001)) had significantly higher probabilities of dying of CML (Figure 3). Also, the probabilities between the intermediate- and the high-risk group were significantly different (P<0.001).

Figure 3
figure3

Cumulative incidence probabilities of dying because of CML in 2205 patients from the in-study registry stratified for the risk groups according to the ELTS score. At 2, 5 and 8 years, horizontal crossbars indicate the upper and lower limit of the 95% CI for the estimated probability.

External validation in the out-study data

In the out-study data, 1120 patients fulfilled the inclusion criteria and were evaluable with respect to all scores. The out-study patients came from groups located in Madrid (n=168), Moscow and St Petersburg (n=456), from the national Polish registry (n=200), from two Czech registries (Camelia and Infinity) (n=285, including 89 patients from Slovakia) and from a Romanian registry (n=11).5 Median age was 49 years; 52% were men. Median follow-up in the out-study patients was 5.6 years (range: 0.1–10.4 years). Compared with the low-risk group (n=681, 61%), the ELTS score provided an intermediate-risk group (SHR: 2.040 (95% CI: 1.039–4.005), P=0.035) and a high-risk group (SHR: 6.746 (95% CI: 3.614–12.592), P<0.001) with significantly greater probabilities of dying of CML (Figure 4a). Again, the probabilities of the intermediate-risk patients (n=302, 27%) and the high-risk patients (n=137, 12%) were significantly different (P<0.001). Although significant differences were observed for the comparisons between the high-risk patients and each of the non-high-risk groups for both the Sokal and the Euro score (always P<0.001), neither score could significantly discriminate probabilities of dying of CML between low- and intermediate-risk patients (Table 2 and Supplementary Figure 2). The EUTOS score identified significantly greater probabilities of dying of CML in the high-risk group as compared with the low-risk group (P=0.002).

Figure 4
figure4

Cumulative incidence probabilities of dying because of CML in 1120 patients from the out-study registry (a) and OS in 2205 patients from the in-study registry (b). At 2, 5 and 8 years, horizontal crossbars indicate the upper and lower limit of the 95% CI for the estimated probability. With too few patients under observation at 8 years, 5-year probabilities were given in panel a.

OS probabilities

Considering death from any cause, the intermediate- and high-risk group of the new ELTS score had significantly lower survival probabilities than the low-risk group (in both cases P<0.001, Figure 4b). The corresponding Cox hazard ratios were 2.781 (95% CI: 2.024–3.830) and 3.282 (95% CI: 2.207–4.814). With the Sokal and the Euro score, both hazard ratios were also significant (all P<0.005, Table 4). Neither the ELTS score, nor the Sokal or the Euro score showed significantly different survival probabilities between intermediate- and high-risk patients. The EUTOS score was not able to significantly distinguish OS probabilities.

Table 4 Overall survival probabilities according to the risk groups of four different prognostic scores in 2205 in-study patients

End of imatinib treatment, second-line treatment

Besides the patients who received HSCT, 571 of 2290 patients (25%) terminated imatinib treatment in first chronic phase. The application of a second-line TKI was reported in 254 patients (44% of 571). In the remaining 317 cases, just an ending of imatinib treatment but no start of another TKI treatment was stated. As a consequence, the performance of the ELTS score was also investigated in the 2205 evaluable patients when follow-up was censored at the start of a second-line TKI or, if this was unknown, 6 months after ending imatinib. Assuming the start of a second-line treatment within 6 months after stopping imatinib treatment for most of the patients seemed reasonable. Of the patients where both dates were available, 84% started second-line-treatment within half a year after imatinib treatment had been terminated. Despite reduced numbers of events, the ELTS score was still able to differentiate probabilities of CML-specific death and of OS. Patients in the low-risk group had significantly lower probabilities of dying of CML (8-year probability: 2% (95% CI: 1–3%)), and higher OS probabilities (8-year probability: 94% (95% CI: 91–93%)) than patients in the intermediate- or high-risk group (always P<0.005). Ending of imatinib treatment was often an individual decision and not systematically documented. Patients remaining with imatinib treatment may be a positive selection. However, considering only the 243 evaluable patients (of 254) with a reported use of a second-line TKI, after starting second-line TKI, the 5-year probability of dying of CML was 2% (95% CI: 0.4–7%) in the low-risk group and the 5-year OS probability was 96% (95% CI: 90–99%). Adding the 277 evaluable patients (of 317) and their follow-up starting from 6 months after ending imatinib treatment, results in the low-risk group were alike: the 5-year probability of dying of CML was 2% (95% CI: 1–4%) and the 5-year OS probability was 94% (95% CI: 90–97%).

A score calculator is accessible via the link http://www.leukemia-net.org/content/leukemias/cml/elts_score/index_eng.html.

Discussion

In a recently reported population-based study, median age at diagnosis was 55 years.36 With additional diseases in the elderly and the therapeutic success of TKIs, about half of the patients die of reasons unrelated to CML. For an adequate assessment of treatment efficacy, it was most appropriate to focus on the event ‘death due to CML’. The new long-term survival score presented here deals with the current situation and was developed to differentiate risk groups with clinically relevant differences in the probabilities of dying of CML.

In 2290 patients of the in-study registry, the 8-year relative survival probability of 96% corresponded to a 4% probability of dying because of CML. After 5 years, relative survival was temporarily increasing. Here, the hazard of dying was higher in the matched general population than in the in-study data. This observation can be explained by nowadays relatively high survival probabilities of CML patients and by the exclusion criteria for the studies preventing the inclusion of frail patients with certain serious comorbidities. Thus, some overestimation of the relative survival probabilities is likely. Without using population data, at 8 years, a 4% probability of dying of CML was also estimated from the in-study data alone. However, in contrast to the relative survival approach, a cause of death assessment was necessary.

In our work, CML-related death was defined as death after disease progression. Without prior progression, in some cases, death because of infection, toxicities or second cancer may also be attributable to CML. Still, in considering only death after disease progression as ‘death due to CML’, the definition was clear-cut. Furthermore, in regression modeling, estimates are most influenced by the cases counted as ‘the event of interest’. Using the prerequisite of progression, it was ensured that only real deaths due to CML had this maximum influence. At the same time, the influence of ambiguous and unrelated cases on the estimates was limited and bias reduced.

Investigating the association of the candidate variables with the cumulative incidence probabilities of CML-specific death, the same prognostic factors as for the Sokal score were identified: age, spleen size, peripheral blasts and platelet count. However, large variations in the estimated coefficients and cutoffs resulted in significantly different risk group distributions with the ELTS score. Age was a significant prognostic factor in the Sokal, the Euro and the new ELTS score but not in the EUTOS score.4, 7, 15 This can be explained by the fact that the first EUTOS score was developed for short-term response and not for long-term survival.4, 33 Recent studies reported a poorer prognostic profile in young adults (<30 years of age) as compared with other age groups.37, 38 However, with TKI treatment, neither Kalmanti et al.37 nor Castagnetti et al.38 observed less favorable long-term survival outcome for young adults in comparison with patients aged between 30 and 59 years or with elderly patients at least 60 years of age. Instead, with similar survival results in the two younger groups, Castagnetti discovered significantly lower OS probabilities in the elderly.38 This observation corresponds to our modeling of age in the ELTS score: although younger age has a rather low impact on the ELTS score as compared with blasts or spleen size, because of its exponentiation by the power of 3, older age has a much higher weight. Remarkably, with the ELTS score, the significant influence of age was identified when using only CML-specific death as an event.

In both registry sections, only the ELTS score identified three significantly different risk groups regarding cumulative incidences of CML-specific death. The first EUTOS score was not designed to distinguish intermediate- and low-risk patients.4 However, using the new ELTS score, a significant discrimination of the probabilities of dying of CML between these two groups was confirmed. Furthermore, the SHRs of high- to low-risk patients were more than twice as high as with the EUTOS score where intermediate- and low risk were not separated. These results and the fact, that the first EUTOS score could not identify significantly different OS probabilities in the in-study registry, support the view that for long-term outcome, a homogeneous low-risk group consisting of about 90% of the patients, as suggested by the first EUTOS score, does not exist. Instead, intermediate-risk patients should be separated from a low-risk group, which has particularly excellent prospects for survival. For long-term outcome, a score distinguishing three risk groups should be preferred.

The ELTS score identified an absolute proportion of 20% more low-risk patients than the Sokal or Euro score. As these patients identified by the new score had similar (slightly lower) probabilities of dying of CML as the low-risk groups of the two other scores, the suggestion of this much larger low-risk group is justified. This superiority was confirmed by the results in the out-study patients and for OS. At the same time, smaller SHRs showed that neither the Sokal nor the Euro score identified a high-risk group, which differed more from the low-risk group than in the case of the ELTS score. The new score identified more high-risk patients than the Euro score (397 versus 353, both registry sections combined). Survival outcome of high-risk patients was comparable between the two scores. As had been shown for interferon alpha treatment,39 the Sokal score allocated too many actual non-high-risk patients to its high-risk group resulting in the lowest probabilities of dying of CML when compared with the other scores. For the identification of high-risk patients for first-line imatinib treatment, the Sokal score is not a good choice.

Validation in the out-study patients was successful. Also, when considering OS in the in-study data, the ELTS score showed significant differences. Successful validation and the significant differences also for OS both indicate that the score provides a most useful discrimination of long-term survival and no relevant bias was introduced by our restriction on and definition of CML-related deaths. With OS intrinsically considering all unrelated deaths, the significant differences between high- and low-risk patients suggest that the prognostic performance of the ELTS score will not be compromised if the definition of ‘CML-related death’ is extended by some cases where no prior progression was observed.

In summary, judging from the incidence probabilities and the distribution of the patients into the risk groups, of all scores, the new prognostic score demonstrated the best discrimination of the probabilities for CML-specific death in both registry samples. The ELTS score is the first long-term survival score developed in patients treated with imatinib and considering the end point ‘CML-related death’. Much improved survival with TKI treatment as compared with previous drug therapy and an end point, which in particular was incomparable with short-term remission outcome, suggested the necessity of a new score. A better prognostic discrimination than with the earlier established scores underlined this necessity.

Calculation of the ELTS score is hardly more complicated than the calculation of the Sokal score. From the estimation of cumulative remission probabilities, the concept of competing risks is well-known among investigators. Detailed guidance on understanding and calculating the corresponding cumulative incidence probabilities was provided.40, 41

Whether differentiating between causes of death or not, in the patients of the low-risk groups (both samples: 61%), the new score showed a very promising long-term outcome when starting treatment with imatinib; an outcome that would be difficult to improve and which is clinically highly relevant—considering that imatinib-based treatment is the primary choice in many countries. The use of imatinib will be further boosted by its availability as a generic drug. Significantly worse results for (overall) survival make an upfront comparison between different TKIs highly desirable for the high-risk patients (both samples: 12%) and worthwhile also for the intermediate-risk patients (27%). An examination of the performance of the new score in patients with first-line treatment other than imatinib will be welcome.

The new ELTS score supports the prospective assessment of long-term antileukemic efficacy and OS of patients with first-line imatinib therapy. We hope that the ELTS score will be considered for the risk-stratified planning, analysis and outcome interpretation of clinical trials. It provides a sound basis for the development of risk-adapted treatment.

References

  1. 1

    Deininger M, O'Brien SG, Guilhot F, Goldman JM, Hochhaus A, Hughes TP et al. International randomized study of interferon vs STI571 (IRIS) 8-year follow up: sustained survival and low risk for progression or events in patients with newly diagnosed chronic myeloid leukemia in chronic phase (CML-CP) treated with imatinib. Blood 2009; 114: 462 (abstr 1126).

    Google Scholar 

  2. 2

    Hehlmann R, Muller MC, Lauseker M, Hanfstein B, Fabarius A, Schreiber A et al. Deep molecular response is reached by the majority of patients treated with imatinib, predicts survival, and is achieved more quickly by optimized high-dose imatinib: results from the randomized CML-study IV. J Clin Oncol 2014; 32: 415–423.

    Article  Google Scholar 

  3. 3

    Castagnetti F, Gugliotta G, Breccia M, Specchia G, Intermesoli T, Capucci A et al. Frontline treatment with imatinib mesylate in chronic myeloid leukemia patients in early chronic phase: a very long-term analysis by the GIMEMA CML Working Party. Blood 2013; 122: 258 (abstr 258).

    Google Scholar 

  4. 4

    Hasford J, Baccarani M, Hoffmann V, Guilhot J, Saussele S, Rosti G et al. Predicting complete cytogenetic response and subsequent progression-free survival in 2060 patients with CML on imatinib treatment: the EUTOS score. Blood 2011; 118: 686–692.

    CAS  Article  Google Scholar 

  5. 5

    Hoffmann VS, Baccarani M, Lindoerfer D, Castagnetti F, Turkina A, Zaritsky A et al. The EUTOS prognostic score: review and validation in 1288 patients with CML treated frontline with imatinib. Leukemia 2013; 27: 2016–2022.

    CAS  Article  Google Scholar 

  6. 6

    Byar DP . Identification of prognostic factors. In: Buyse ME, Staquet MJ, Sylvester RJ (eds). Cancer Clinical Trials—Methods and Practice. Oxford University Press: Oxford, 1984; pp 423–441.

    Google Scholar 

  7. 7

    Hasford J, Pfirrmann M, Hehlmann R, Allan NC, Baccarani M, Kluin-Nelemans JC et al. A new prognostic score for survival of patients with chronic myeloid leukemia treated with interferon alfa. Writing Committee for the Collaborative CML Prognostic Factors Project Group. J Natl Cancer Inst 1998; 90: 850–858.

    CAS  Article  Google Scholar 

  8. 8

    O'Brien SG, Guilhot F, Larson RA, Gathmann I, Baccarani M, Cervantes F et al. Imatinib compared with interferon and low-dose cytarabine for newly diagnosed chronic-phase chronic myeloid leukemia. N Engl J Med 2003; 348: 994–1004.

    CAS  Article  Google Scholar 

  9. 9

    Saglio G, Kim DW, Issaragrisil S, le Coutre P, Etienne G, Lobo C et al. Nilotinib versus imatinib for newly diagnosed chronic myeloid leukemia. N Engl J Med 2010; 362: 2251–2259.

    CAS  Article  Google Scholar 

  10. 10

    Kantarjian H, Shah NP, Hochhaus A, Cortes J, Shah S, Ayala M et al. Dasatinib versus imatinib in newly diagnosed chronic-phase chronic myeloid leukemia. N Engl J Med 2010; 362: 2260–2270.

    CAS  Article  Google Scholar 

  11. 11

    Preudhomme C, Guilhot J, Nicolini FE, Guerci-Bresler A, Rigal-Huguet F, Maloisel F et al. Imatinib plus peginterferon alfa-2a in chronic myeloid leukemia. N Engl J Med 2010; 363: 2511–2521.

    CAS  Article  Google Scholar 

  12. 12

    Hehlmann R, Lauseker M, Jung-Munkwitz S, Leitner A, Muller MC, Pletsch N et al. Tolerability-adapted imatinib 800 mg/d versus 400 mg/d versus 400 mg/d plus interferon-{alpha} in newly diagnosed chronic myeloid leukemia. J Clin Oncol 2011; 29: 1634–1642.

    CAS  Article  Google Scholar 

  13. 13

    Cortes JE, Kim D-W, Kantarjian HM, Brümmendorf TH, Dyagil I, Griskevicius L et al. Bosutinib versus imatinib in newly diagnosed chronic-phase chronic myeloid leukemia: results from the BELA trial. J Clin Oncol 2012; 30: 3486–3492.

    CAS  Article  Google Scholar 

  14. 14

    Radich JP, Kopecky KJ, Appelbaum FR, Kamel-Reid S, Stock W, Malnassy G et al. A randomized trial of dasatinib 100 mg versus imatinib 400 mg in newly diagnosed chronic-phase chronic myeloid leukemia. Blood 2012; 120: 3898–3905.

    CAS  Article  Google Scholar 

  15. 15

    Sokal JE, Cox EB, Baccarani M, Tura S, Gomez GA, Robertson JE et al. Prognostic discrimination in ‘good-risk’ chronic granulocytic leukemia. Blood 1984; 63: 789–799.

    CAS  PubMed  PubMed Central  Google Scholar 

  16. 16

    Baccarani M, Deininger MW, Rosti G, Hochhaus A, Soverini S, Apperley JF et al. European LeukemiaNet recommendations for the management of chronic myeloid leukemia. 2013 Blood 2013; 122: 872–884.

    CAS  Article  Google Scholar 

  17. 17

    Therneau TM, Grambsch PM . Modeling Survival Data: Extending the Cox Model. Springer: New York, 2000.

    Book  Google Scholar 

  18. 18

    Kalbfleisch JD, Prentice RL . The Statistical Analysis of Failure Time Data. Wiley: New York, 1980.

    Google Scholar 

  19. 19

    Choudhury JB . Non-parametric confidence interval estimation for competing risks analysis: application to contraceptive data. Stat Med 2002; 21: 1129–1144.

    Article  Google Scholar 

  20. 20

    Gray RJ . A class of k-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16: 1141–1154.

    Article  Google Scholar 

  21. 21

    Pohar-Perme M, Stare J, Estève J . On estimation in relative survival. Biometrics 2012; 68: 113–120.

    Article  Google Scholar 

  22. 22

    Pohar M, Stare J . Relative survival analysis in R. Comput Meth Prog Bio 2006; 81: 272–278.

    Article  Google Scholar 

  23. 23

    Human Mortality Database: University of California, Berkeley (USA), and Max Planck Institute of Demographic Research (Germany): the human mortality database provides mortality and population data on various countries. 05/2014 update. Available at http://www.mortality.org/ or http://www.humanmortality.de/.

  24. 24

    Royston P, Sauerbrei W . Multivariable Model-Building: A Pragmatic Approach to Regression Analysis Based on Fractional Polynomials for Continuous Variables. John Wiley & Sons: Chichester, 2008.

    Book  Google Scholar 

  25. 25

    Fine JP, Gray RJ . A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94: 496–509.

    Article  Google Scholar 

  26. 26

    Kuk D, Varadhan R . Model selection in competing risks regression. Stat Med 2013; 32: 3077–3088.

    Article  Google Scholar 

  27. 27

    Davison AC, Hinkley DV . Bootstrap Methods and Their application. Cambridge University Press: Cambridge, 1997.

    Book  Google Scholar 

  28. 28

    Altman DG, Lausen B, Sauerbrei W, Schumacher M . Dangers of using ‘optimal’ cutpoints in the evaluation of prognostic factors. J Natl Cancer Inst 1994; 86: 829–835.

    CAS  Article  Google Scholar 

  29. 29

    Silverman BW . Density Estimation for Statistics and Data Analysis. Chapman and Hall:: London, 1986.

    Book  Google Scholar 

  30. 30

    Sauerbrei W, Meier-Hirmer C, Benner A, Royston P . Multivariable regression model building by using fractional polynomials: description of SAS, STATA and R programs. Comput Stat Data An 2006; 50: 3464–3485.

    Article  Google Scholar 

  31. 31

    Scrucca L, Santucci A, Aversa F . Regression modeling of competing risk using R: an in depth guide for clinicians. Bone Marrow Transpl 2010; 45: 1388–1395.

    CAS  Article  Google Scholar 

  32. 32

    Deenik W, van der Holt B, Verhoef GE, Smit WM, Kersten MJ, Kluin-Nelemans HC et al. Dose-finding study of imatinib in combination with intravenous cytarabine: feasibility in newly diagnosed patients with chronic myeloid leukemia. Blood 2008; 111: 2581–2588.

    CAS  Article  Google Scholar 

  33. 33

    Gugliotta G, Castagnetti F, Palandri F, Breccia M, Intermesoli T, Capucci A et al. Frontline imatinib treatment of chronic myeloid leukemia: no impact of age on outcome, a survey by the GIMEMA CML Working Party. Blood 2011; 117: 5591–5599.

    CAS  Article  Google Scholar 

  34. 34

    Simonsson B, Gedde-Dahl T, Markevärn B, Remes K, Stentoft J, Almqvist A et al. Combination of pegylated IFN-α2b with imatinib increases molecular response rates in patients with low- or intermediate-risk chronic myeloid leukemia. Blood 2011; 118: 3228–3235.

    CAS  Article  Google Scholar 

  35. 35

    Cervantes F, López-Garrido P, Montero M-I, Jonte F, Martínez J, Hernández-Boluda J-C et al. Early intervention during imatinib therapy in patients with newly diagnosed chronic-phase chronic myeloid leukemia: a study of the Spanish PETHEMA group. Haematologica 2010; 95: 1317–1324.

    CAS  Article  Google Scholar 

  36. 36

    Hoffmann VS, Baccarani M, Hasford J, Lindoerfer D, Burgstaller S, Sertic D et al. The EUTOS population-based registry: incidence and clinical characteristics of 2904 CML patients in 20 European Countries. Leukemia 2015; 29: 1336–1343.

    CAS  Article  Google Scholar 

  37. 37

    Kalmanti L, Saussele S, Lauseker M, Proetel U, Muller MC, Hanfstein B et al. Younger patients with chronic myeloid leukemia do well in spite of poor prognostic indicators: results from the randomized CML study IV. Ann Hematol 2014; 93: 71–80.

    Article  Google Scholar 

  38. 38

    Castagnetti F, Gugliotta G, Baccarani M, Breccia M, Specchia G, Levato L et al. Differences among young adults, adults and elderly chronic myeloid leukemia patients. Ann Oncol 2015; 26: 185–192.

    CAS  Article  Google Scholar 

  39. 39

    Hasford J, Pfirrmann M, Hehlmann R, Shepherd P, Guilhot F, Mahon FX et al. Prognostic factors. In: Carella AM, Daley G, Eaves C, Goldman J, Hehlmann R (eds). Chronic Myeloid Leukemia – Biology and Treatment. Martin Dunitz: London, 2001; pp 205–223.

    Google Scholar 

  40. 40

    Scrucca L, Santucci A, Aversa F . Competing risk analysis using R: an easy guide for clinicians. Bone Marrow Transpl 2007; 40: 381–387.

    CAS  Article  Google Scholar 

  41. 41

    Pfirrmann M, Hochhaus A, Lauseker M, Saussele S, Hehlmann R, Hasford J . Recommendations to meet statistical challenges arising from endpoints beyond overall survival in clinical trials on chronic myeloid leukemia. Leukemia 2011; 25: 1433–1438.

    CAS  Article  Google Scholar 

Download references

Acknowledgements

The valuable assistance of Doris Lindörfer, Annett Schmitt, Christina Niedermeier and Kirsi Manz is gratefully appreciated. We would like to thank Sina Hehn for providing link and tool in order to calculate the ELTS score. The EUTOS is a common project of the European LeukemiaNet and Novartis Oncology Europe. Novartis Oncology Europe provided financial support for the EUTOS project.

Author contributions

MP, JH and BS designed the work. All authors were involved in the collection and assembling of the data. MP performed the statistical analysis. MP VSH, RH, MB, JH and JG had an important role in interpreting the results. MP wrote the first draft of the manuscript; all authors participated in the revision. All others approved the final version. MP has had full access to the data in the study and final responsibility for the decision to submit for publication.

Author information

Affiliations

Authors

Corresponding author

Correspondence to M Pfirrmann.

Ethics declarations

Competing interests

MP acted as a consultant for and received honoraria from Novartis and Bristol-Myers Squibb. MB served on the speakers’ bureau of and received honoraria from ARIAD, Bristol-Myers Squibb, Novartis and Pfizer and acted as a consultant for ARIAD and Novartis. SS acted as a consultant for Novartis, and Pfizer, received honoraria ARIAD, Bristol-Myers Squibb, Novartis and Pfizer, and received research funding from Bristol-Myers Squibb and Novartis. FC served on the speakers’ bureau of and received honoraria from ARIAD, Bristol-Myers Squibb and Novartis and acted as a consultant for ARIAD and Pfizer. GO acted as a consultant for ARIAD, Bristol-Myers Squibb, Celgene, Johnson & Johnson, Novartis and Roche and received research funding from Celgene and Novartis. VSH received research funding from Novartis. F Castagnetti acted as a consultant for and received honoraria from ARIAD, Bristol-Myers Squibb, Novartis and Pfizer. JH received honoraria from Bristol-Myers Squibb and research funding from Novartis; RH received research funding from Bristol-Myers Squibb and Novartis. BS acted as a consultant for and received honoraria from Novartis and Bristol-Myers Squibb.

Additional information

Supplementary Information accompanies this paper on the Leukemia website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Pfirrmann, M., Baccarani, M., Saussele, S. et al. Prognosis of long-term survival considering disease-specific death in patients with chronic myeloid leukemia. Leukemia 30, 48–56 (2016). https://doi.org/10.1038/leu.2015.261

Download citation

Further reading

Search

Quick links