Features of TP53-mutated patients with chronic myelomonocytic leukemia in a national (ABCMML) and international cohort (cBIOPORTAL)

Chronic myelomonocytic leukemia (CMML) is a rare, phenotypically and genetically diverse hematologic cancer that affects the elderly and has an inherent risk of developing into secondary acute myeloid leukemia (AML). According to the French American British (FAB) criteria, CMML was initially separated into two groups based on the presence of myeloproliferation: myeloproliferative disorder (MP-CMML; WBC count > 13 × 10⁹/L) and myelodysplastic syndrome (MD-CMML; WBC count < 13 × 10⁹/L) [1, 2]. The World Health Organization (WHO) classified CMML as belonging to the mixed category MDS/MPN in 2002 because it possesses characteristics of both an MDS and an MPN [3]. Two groups recently reported updated diagnostic criteria for CMML following the 2016 revision to the World Health Organization’s classification of myeloid neoplasms and acute leukemia [4‐6]. The outcome of CMML patients can vary greatly, indicating that a number of factors may influence how the disease progresses and what causes these patients to die [7‐13].

The Austrian Biodatabase for CMML (ABCMML) was recently reported. Patients with CMML have had their epidemiologic, hematologic, biochemical, clinical, immunophenotypic, cytogenetic, molecular, and biologic data gathered from various Austrian centers for 40 years [14]. It has been demonstrated to be a representative and practical source of real-world data for biomedical research.

Because of the molecular heterogeneity of CMML, it is critical to understand the meaning of molecular characteristics so that the patient can be provided with the best care possible for their unique circumstances. The effects of molecular aberrations on the clinical outcome and phenotype of disease have been examined in a few studies, but the majority of these studies’ conclusions were not confirmed by separate cohorts. However, until a prognostic parameter’s usefulness has been established, it should not be used in clinical settings. Evaluation of a prognostic parameter’s performance in a sample different from the one used to build the model is known as external validation [15].

Big data containing a huge number of datasets from large international consortium efforts are now available for many cancer entities including CMML. The cBIOPORTAL platform is such a collection of big data aiming to build a platform to support clinical decisions for personalized cancer treatment [16]. Moreover, due to the large number of well-characterized patients, it is a perfect source of data for validation of findings in traditional, sometimes much smaller patient cohorts. In this study we used data from CMML patients documented in cBIOPORTAL to validate the features of TP53-mutated CMML patients who have been analyzed in the ABCMML.

Five to ten percent of cases of AML and de novo myelodysplastic syndrome (MDS) have TP53 mutations [17]. In contrast, up to 30–40% of patients with therapy-related MDS and AML have TP53 mutations. Missense mutations found in a few common hotspots account for the majority of inactivating mutations seen in MDS and AML. A loss-of-function effect on normal p53 function is determined by TP53 missense mutations in conjunction with truncating mutations or chromosomal loss of TP53. The selection pressure of chemotherapy or MDM2 inhibitor therapy causes TP53-mutant clones to grow clonally. Current chemotherapy is ineffective against TP53-mutant clones. A complex karyotype and a generally poor prognosis are linked to TP53 mutations.

There are some studies analyzing the prevalence and role of TP53 mutations in CMML. The prevalence of these mutations has a wide range, from 1–8.3%. In one Chinese study of CMML patients, the prevalence of TP53 mutations was 8.3% [13]. On the other hand, the prevalence was 1% in other series [18]. In our study, the prevalence of TP53 mutations was 1.6% in the ABCMML cohort and 3.7% in the cBIOPORTAL cohort. Differences regarding the prevalence of TP53 mutations may partly be due to different proportions of therapy-related CMML cases in reported series. In an American study the prevalence of molecular aberrations of TP53 was 4.3% in de novo CMML and 11.8% in therapy-related CMML [19]. Interestingly, this group also evaluated a cohort of 189 patients with CMML-associated AML [20]. They found that transformation occurs through distinct trajectories characterized by genomic profiles and clonal evolution: monocytic (Mo-AML, 53%), immature myeloid (My-AML, 43%), or erythroid AML (Ery-AML, 2%), which were defined by complex karyotypes and TP53 mutations.

There is only limited information on the clinical outcome of TP53 mutations in CMML series. In a study reported by Gurney, 31/1315 (2.4%) patients had concurrent TP53 alterations. There was a highly significant inferior impact of TP53 alteration in this study [21]. In another study of the effect of TP53 mutation variant allele frequency on phenotype and outcomes in myelodysplastic syndromes, only TP53 mutations with VAF  40% were an independent factor for inferior survival in multivariate analysis [22]. In the cBIOPORTAL cohort, TP53-mutated patients had significantly inferior survival and AML-free survival as compared to nonmutated patients. In the ABCMML cohort, survival and AML-free survival were numerically shorter in TP53-mutated patients, but this did not reach significance. Looking at the median survival numbers, the shape of the Kaplan–Meier curves, and the low number of mutated patients, it is likely that the results would have reached significance with higher patient numbers, indicating the necessity of sufficient patient numbers for the comparison of CMML patients regarding outcome and phenotype according to their molecular subtype, particularly in the case of rare mutations.

The correlation of phenotypic features with the mutational status in CMML patients has been described previously [18], but regarding TP53 there are only limited data. In our study decreased platelet counts, decreased Hb values, and the presence of blast cells in PB were significantly associated with TP53 mutations in the cBIOPORTAL cohort but not in the ABCMML cohort.

Limitations of this study include the fact that the proportion of patients with leukocytes  13 G/L was significantly higher in the ABCMML cohort as compared to the cBIOPORTAL cohort. The reason for this imbalance is not completely clear. Increased laboratory screening in recent years in asymptomatic persons may detect some diseases including CMML in an earlier phase than in the past. Therefore, older patient series may be enriched in patients with more advanced disease as compared to more recent series. In fact, we have seen a significant drop in the proportion of patients with MP-CMML from 66% to 48% since 2010 in the ABCMML database (unpublished data).

Changes in the diagnostic criteria of CMML over time since its first description in 1982 represent another limitation of the ABCMML database, suggesting that this patient group is more heterogenous as compared to the cBIOPORTAL group which contains patients that were included over a shorter period of time. Furthermore, it needs to be considered that a proportion of patients in ABCMML, in particular older patients, did not consent to have BM puncture. However, we do not think that this greatly affected diagnostic accuracy, since persistent peripheral blood monocytosis is the most important diagnostic feature, and a genoclinical model has been recently described that uses mutational data, peripheral blood values, and clinical variables to predict the MDS vs. CMML diagnosis with high accuracy in the absence of a BM biopsy result [23]. Moreover, somatic mutations associated with CMML were not only detected in CMML patients confirmed by BM biopsy but also in 57% of patients with nondiagnostic BM features. Interestingly, the OS in non-diagnostic-mutated patients was indistinguishable from CMML, suggesting that the mutational spectrum is a much more sensitive parameter for the detection of myeloid malignancies as compared to BM morphology [24].

In recent years, health care management has changed from a disease-centered model to a patient-centered model [25]. The adoption of big data characterized by a large amount of digital data which are continually generated by people within clinical care and everyday life will enable implementation of personalized and precise medicine based on personalized information. Moreover, these data can be used for validation of findings from national cohorts, as we have shown in this study, and thus make an important contribution to optimizing patient management.