Chronic myelomonocytic leukaemia (CMML) is a clonal hematopoietic stem cell disorder that is regarded as myeloproliferative/myelodysplastic overlap disorder according to the 2016 revision to the World Health Organization (WHO) classification of myeloid neoplasms and acute leukaemia. Besides the absence of the Philadelphia chromosome and/or the BCR-ABL fusion gene, evidence of dysplasia in at least one myeloid lineage and less than 20% myeloblasts, monoblasts or promonocytes in the peripheral blood or bone marrow and a persistent monocytosis >1000/µl with monocytes accounting for ≥10% of the white blood cell count and exclusion of secondary causes thereof are required to establish the diagnosis of CMML. If myelodysplastic features are missing, the detection of an acquired clonal or molecular genetic abnormality in hematopoietic stem cells and/or a persistent monocytosis for more than 3 months without other causes can still lead to the diagnosis of CMML. Rearrangements of the platelet-derived growth factor receptor alpha (PDGFRA) gene, of the platelet-derived growth factor receptor beta (PDGFRB) gene and the fibroblast growth factor receptor 1 (FGR1) gene or PCM1-JAK2 fusions must be excluded if eosinophilia is present [1
While cytogenetic abnormalities are only found in approximately 30% of patients with CMML, molecular abnormalities have been reported in up to 90% of CMML cases [2
]. Prognostic tools, such as the MD Anderson prognostic score (MDAPS) divide treatment-naïve CMML patients into the risk groups “low”, “intermediate-1”, “intermediate-2” and “high” with a median overall survival (OS) of 24, 15, 8 and 5 months, respectively [6
]. Several other CMML-specific scores, e.g. modified MDAPS (MDAPS M1), CMML-specific prognostic scoring system (CPSS) and the Mayo prognostic model, are used to predict OS [7
]. The Düsseldorf score, international prognostic scoring system (IPSS) and the revised IPSS (IPSS-R) were primarily applied to estimate OS in MDS but also included myelodysplastic CMML (MD-CMML) patients [10
]. Molecular abnormalities, such as ASXL1, NRAS, RUNX1 and SETBP1 mutations impact on OS and have already been included in molecular prognostic risk models in CMML [5
Current therapeutic options in CMML are limited. A best supportive care strategy may include transfusion of blood products, administration of erythropoiesis-stimulating agents or myeloid growth factors, antibiotics, antiviral medication, iron chelation therapy as well as cytoreduction with either hydroxyurea or etoposide [15
]. In the only randomized clinical trial performed and published in CMML to date including 105 patients, oral hydroxyurea proved to be superior to oral etoposide in terms of response rate and median OS [16
]. In retrospective analyses, the hypomethylating agents azacitidine and decitabine yielded a median OS of 13.2 and 19.0 months in CMML, respectively [17
]. The approval of both substances by the U.S. Food and Drug Administration (FDA) for the treatment of CMML is based on the results of two phase III trials, albeit the number of included CMML patients (14 in each study) was low [19
]. In contrast to the U.S. the European Medicines Agency (EMA) only approved azacitidine for a subset of CMML patients, for those with MD-CMML and a bone marrow blast percentage of 10–29%, which therefore includes a subset of patients who fulfil the criteria of acute myeloid leukaemia according to the WHO classification, namely those with 20–29% bone marrow blasts. Up to now, decitabine has neither been approved for the treatment of CMML, nor for the treatment of myelodysplastic syndromes (MDS) in Europe. Even with intensive chemotherapy protocols the median OS does not exceed 44 weeks and allogeneic stem cell transplantation represents the only curative treatment approach in CMML [21
Treatment criteria in CMML have not been established so far and therefore treatment indications are mainly based on expert opinion or consensus statements from expert panels including severe anemia (hemoglobin <10 g/dl), blast percentage >5% in the peripheral blood, immature myeloid cells including myeloblasts, promyelocytes, myelocytes and metamyelocytes >10% in the peripheral blood, platelet count <50 G/l, white blood cell count >30 G/l, extramedullary disease manifestations, presence of B‑symptoms and symptomatic splenomegaly [15
]. According to the current National Comprehensive Cancer Network guidelines for MDS, which subsume CMML, symptomatic anemia, clinically relevant thrombocytopenia and/or neutropenia or increased bone marrow blasts demonstrate potential treatment indications for CMML-specific therapy [24
Apart from the abovementioned treatment indications and due to the heterogeneity of this disease as well as the broad spectrum of the clinical course of CMML, the time to first treatment (TTFT) with intensive chemotherapy regimens, hypomethylating agents or cytoreductive therapy, such as hydroxyurea varies considerably among patients. Defining treatment indications as well as choosing the optimal time point for treatment initiation pose common challenges in clinical practice, and analyses of a potential effect of differences in TTFT on OS are scarce or lacking in CMML.
The aim of this single centre retrospective study was (a) to evaluate which parameters at the initial diagnosis of CMML were relevant to the time point of treatment initiation based on established treatment indications in clinical practice at the center in Salzburg, (b) to propose a simplified risk model for TTFT in CMML, and (c) to validate this risk model in an external test cohort.
Patients and methods
This retrospective analysis was approved by the Ethics Committee of the provincial government of Salzburg, Austria (reference number 415-EP/39/11) and was based on the data of 55 unselected consecutive CMML patients (training set) diagnosed and/or treated at our tertiary oncology center in Salzburg, Austria, between 2004 and 2015. Those CMML patients who received azacitidine during the course of the disease were included in the Austrian Registry on Hypomethylating Agents (NCT01595295) of the working group on pharmaceutical tumor treatment (Arbeitsgemeinschaft Medikamentöse Tumortherapie, AGMT; www.agmt.at
). The external independent validation set consisted of 65 CMML patients from 12 Austrian hospitals included in the Austrian Registry on Hypomethylating Agents, with the majority (n
= 60) derived from the data base of the Austrian Registry on Hypomethylating Agents. All patients alive at the time point of data acquisition signed an informed consent to allow the collection of personal data. The diagnosis was established according to the 2008 WHO classification of tumors of hematopoietic and lymphoid tissues [25
]. The OS was calculated from the date of first diagnosis until date of death or date of last known follow-up, TTFT was defined as the time period between initial diagnosis and first CMML-specific treatment with either hydroxyurea or azacitidine.
Estimates on TTFT distributions were based on the Kaplan-Meier method and the log-rank test was used to compare Kaplan-Meier survival curves. We analyzed the impact of various clinical baseline factors and cytogenetic abnormalities at initial diagnosis on TTFT, which in part had already been incorporated into established prognostic scores for OS [6
]. The optimal cut-offs for discerning the treatment status (untreated or treated) of patients at the end of the follow-up time were calculated based on receiver operating characteristics (ROC) analyses and the Youden index J, which represents the maximum of sensitivityc
for all cut points in the ROC curve [26
]. In order to avoid variable redundancy, the WHO classification was not included in the univariate and multivariate analysis, as this is determined by the percentage of bone marrow blasts and peripheral blood blasts: CMML-1 (blasts <5% in the peripheral blood and blasts <10% in the bone marrow) and CMML-2 (peripheral blood blasts 5–19% and/or 10–19% bone marrow blasts[25
]). Parameters which proved statistically significant in univariate analysis (p
< 0.05) were included in multivariate analysis. Statistical analyses were carried out using the IBM® SPSS® statistics software, version 20.
The optimal time point to initiate treatment in CMML may pose a challenge to the treating physician. For example, while there is a general consensus that symptomatic cytopenia in CMML demonstrates a treatment indication, initial management of the latter often involves application of erythropoiesis-stimulating agents or blood products without immediately initiating treatment with hypomethylating agents or chemotherapy. Whether such a treatment delay is associated with a better or worse clinical outcome has not been prospectively tested in CMML so far. In clinical practice at our center in Salzburg, the decision to start treatment with either hydroxyurea or azacitidine is mainly based on (i) expert panel recommendations, (ii) the National Comprehensive Cancer Network guidelines for MDS and (iii) the clinical presentation of the individual patients [15
]. With an analysis of 55 unselected consecutive CMML patients we evaluated which baseline factors impact on treatment initiation with either azacitidine or hydroxyurea in CMML. By testing the influence of individual parameters applied in established prognostic scores and further variables, lactate dehydrogenase, bone marrow blast percentage and platelets at the initial diagnosis were significantly associated with TTFT. Because a single cut-off was desirable for further statistical analyses and comparison with other dichotomous variables, we calculated cut-offs for parameters measured on a continuous scale with ROC analysis and the Youden index. These cut-off values were in line with a clinically relevant graduation, for example, the calculated cut-off for lactate dehydrogenase (≥223 U/l) represents the upper limit of normal and the cut-off for the platelets (<55 G/l) closely matches the definition of grade 3 thrombocytopenia according to the Common Terminology Criteria for Adverse Events (Version 4.0; [27
]). Elevated lactate dehydrogenase at the initial diagnosis was the strongest predictor for the time interval to systemic treatment initiation with either hydroxyurea or azacitidine, followed by increased bone marrow blast percentage and thrombocytopenia. Elevated lactate dehydrogenase and an increased bone marrow blast percentage are clinical parameters that may indicate pending disease progression in CMML.
Interestingly, neither the hemoglobin level nor red blood cell transfusion-dependence was significantly associated with TTFT in multivariate analysis. This is probably because the initial management of symptomatic anaemia with erythropoiesis-stimulating agents might defer the initiation of treatment with azacitidine or hydroxyurea. In the training set, 26% of CMML patients initially received erythropoiesis-stimulating agents, whereas the highest frequency (44%) was documented in the low-risk group according to our TTFT prediction model. The use of thrombopoietin receptor agonists has not been approved for CMML patients with severe thrombocytopenia so far and as a consequence treatment with either azacitidine or hydroxyurea might be initiated earlier in comparison to patients who present with symptomatic anaemia, who also have erythropoiesis-stimulating agents as a therapeutic option.
In our institution, we generally use a watch and wait strategy in patients with CMML without red blood cell or platelet transfusion dependence and with a hemoglobin level ≥10.0 g/dl and a white blood cell count <20 G/l. Erythropoiesis-stimulating agents are used in patients with CMML-0 or CMML-1 without red blood cell or platelet transfusion dependence and with a hemoglobin level <10.0 g/dl. Hydroxyurea is generally initiated in CMML-0 and CMML-1 patients with a white blood cell count ≥20 G/l with leukocyte dynamics (i. e. rapid increase in in white blood cell count). We generally initiate azacitidine as front-line therapy in patients with CMML-0 or CMML-1 red blood cell and/or platelet transfusion dependence, and/or PLT counts <50 G/l with platelet dynamics (i. e. rapidly dropping platelet counts), if they do not remain stable over a course of several months.
Our proposed TTFT risk model might identify CMML patients who are likely to require early treatment initiation and may be considered for early interventional trials. Another clinical implication of the proposed TTFT risk score is the ability to identify patients who will not require treatment initiation for a longer period of time or who will never require treatment initiation with hypomethylating agents or hydroxyurea and this in turn might help to individualize routine follow-up intervals. We acknowledge the fact that some OS risk scores incorporate molecular information such as ASXL1, NRAS, RUNX1 and SETBP1 mutations in order to estimate OS in CMML [5
]. We could not include data of molecular aberrations in our analysis as data collection started in 2004 and molecular studies were not routinely performed at this time. We aimed at creating a simplified score using easily available clinical parameters, which specifically estimates TTFT in CMML. In the great majority (91%) of patients in the validation set, treatment with either azacitidine or hydroxyurea has been initiated, while this was only the case in 62% in the training set. This fact is due to the inclusion criterion of azacitidine treatment for patients included in the Austrian Registry on Hypomethylating Agents, from which most patients (n
= 60) in the validation set were recruited. Due to this selection bias we could observe more treatment events in the low-risk group of the validation set in comparison to the training set (Fig. 1
a,b); however, Kaplan-Meier curves for the intermediate-risk and high-risk groups showed striking similarity between the training set and the validation set.
In summary, we were able to demonstrate that lactate dehydrogenase, bone marrow blast percentage and platelets at the initial diagnosis are the most relevant parameters for the time to first treatment initiation with either azacitidine or hydroxyurea in our CMML cohort. Based on these three parameters, we propose a TTFT risk score for treatment-naïve CMML patients with clinical implications, such as identifying CMML patients for early investigational trials or to tailor individual follow-up intervals. The validity of our model has been confirmed in an external separate CMML set of 65 patients.
Compliance with ethical guidelines
Conflict of interest
F. Huemer received travel support from Roche and Merck. L. Weiss reports receiving honoraria from Bayer, Ipsen, Lilly and Shire; travel support from Abbvie, BMS, Pfizer and Pharmamar; research funding from Takeda. A. Egle is a consultant for Celgene and reports receiving honoraria from Celgene. K. Geissler has been a member of advisory boards for Celgene. A. Zebisch reports receiving honoraria from Celgene and Novartis. S. Burgstaller has been a consultant for Celgene and Novartis, a member on the Board of Directors or advisory committees for Celgene and Novartis and reports receiving research funding from Celgene and honoraria from AOP Orphan Pharmaceuticals, Celgene, Mundipharma and Novartis. R. Stauder has been a consultant and a member on the Board of Directors or advisory committees for Celgene and reports receiving research funding and honoraria from Celgene, Teva (Ratiopharm) and Novartis. W.R. Sperr has been a consultant for Celgene. Alois Lang has been a consultant for Celgene. M. Pfeilstöcker has been a consultant for Celgene and Novartis and reports receiving honoraria from Celgene, Novartis and Janssen-Cilag. S. Machherndl-Spandl has been a member of an advisory board for Celgene. R. Greil reports receiving honoraria from Bristol-Myers-Squibb, Cephalon, Amgen, Eisai, Mundipharma, Merck, Janssen-Cilag, Genentech, Novartis, AstraZeneca, Boehringer Ingelheim, Pfizer, Roche and Sanofi Aventis, research funding from Cephalon, Celgene, Amgen, Mundipharma, Genentech, Pfizer, GSK and Ratiopharm and has been a consultant for Bristol-Myers-Squibb, Cephalon and Celgene. L. Pleyer has been a consultant for Agios, Celgene, Bristol-Myers Squibb and Novartis and reports receiving honoraria and travel support from Agios, Celgene, Bristol-Myers Squibb, Novartis and AOP Orphan Pharmaceuticals. V. Faber, D. Neureiter, D. Voskova, A. Pichler, M. Stampfl declare that no competing interests exist.
This retrospective analysis was approved by the Ethics Committee of the provincial government of Salzburg, Austria (reference number 415-EP/39/11). All patients alive at the time point of data acquisition signed an informed consent to allow the collection of personal data.