An adaptive survival prediction model in CMML Free

In this issue of Blood, Tefferi et al¹ describe a mutation-agnostic survival prediction model for patients with chronic myelomonocytic leukemia (CMML), based on circulating blasts ≥2%, leukocytosis ≥13 × 10⁹/L, and severity of anemia to formulate a clinical stratification tool, acronymized to “BLAST.” This model is resource adaptive as it is further refined using molecular and cytogenetic data (BLAST-mol).

CMML, the most common subtype of myelodysplastic/myeloproliferative (MDS/MPN) overlap syndromes, is clinically heterogenous with features of both myeloproliferation (monocytic leukocytosis, extramedullary hematopoiesis) and myelodysplasia (dysplastic blood and/or marrow morphology, cytopenias). Its diagnostic criteria include persistent peripheral monocytosis ≥10% of total white blood cells (WBCs), with the absolute monocyte count threshold lowered to 0.5 × 10⁹/L to include oligomonocytic CMML by both the World Health Organization (WHO) and the International Consensus Classification (ICC); the presence of ≥1 somatic mutations on a next-generation sequencing (NGS) panel and/or clonal cytogenetic abnormality; morphologic dysplasia in the blood and/or marrow; and an increase in the fraction of classical monocytes (CD14⁺/CD16⁻) to >94% of total monocytes by flow cytometry.^2-4 Per the ICC criteria, the presence of cytopenias (hemoglobin <12 g/dL in female patients and <13 g/dL in male patients, absolute neutrophil count <1.8 × 10⁹/L, and/or platelet count <150 × 10⁹/L) helps differentiate CMML from the precursor entity of clonal monocytosis of undetermined significance (CMUS).^2,3 

Allogeneic hematopoietic stem cell transplantation (HSCT) is the only treatment modality with curative potential in CMML, but may decrease life expectancy if used early in patients with lower-risk disease.⁵ Currently available nontransplant therapies (many used off-label) include erythropoiesis-stimulating or maturation agents, hydroxyurea, hypomethylating agents (HMAs), thrombopoietin receptor agonists, JAK inhibitors, and acute myeloid leukemia (AML)-like therapy (eg, intensive chemotherapy or HMA + venetoclax) in patients with high blast percentage in the blood or marrow. Those therapies have limited and short-lived efficacy and lack disease modification. HMAs (azacitidine, decitabine, decitabine/cedazuridine) are the only Food and Drug Administration–approved therapies in CMML and do not affect mutational burden or clonal evolution.⁶ A recent phase 3 trial failed to show superiority of azacitidine in improving event-free survival (death, AML transformation, or progression) compared with hydroxyurea in proliferative CMML, due to toxicity.⁷ Patients treated with hydroxyurea had numerically longer median overall survival (21.9 months) compared with those treated with azacitidine (18.4 months, P = .67).⁷ This underscores the importance of prioritizing clinical trials in this disease.

The variability in clinical presentation, disease course, and survival poses a challenge and warrants the use of risk stratification models to help in the selection of treatment, clinical trial enrollment, and timing of HSCT. Several risk-scoring models were developed and were either MDS-specific models that included patients with dysplastic CMML (WBC <13 × 10⁹/L), like the International Prognostic Scoring System for MDS (IPSS), IPSS-Revised (IPSS-R), IPSS-Molecular (IPSS-M), or CMML-specific models, for example, CMML-specific prognostic scoring system (CPSS), Mayo, MD Anderson prognostic score, with contemporary CMML models integrating mutational data such as Mayo Molecular, Groupe Francophone des Myelodysplasies, and molecular CPSS (CPSS-Mol).^8,9 There is no consensus on the best prognostic system to use in clinical practice or for eligibility in clinical trials. In the clinic, where I am privileged to have multigene NGS panels readily available, I have been using IPSS-M and CPSS-Mol for patients with dysplastic CMML and CPSS-Mol for those with proliferative CMML (WBC ≥13 × 10⁹/L) to help guide timing of HSCT.

This latest model, BLAST, which was developed using the Mayo Clinic database (n = 457) and validated using cohorts from MD Anderson Cancer Center (n = 449) and University of Milan (n = 203), uses simple variables that are readily available in low-resource settings. BLAST exhibited comparable predictive accuracy to currently used molecular models, CPSS-mol and IPSS-M. There was still a value for including molecular and cytogenetic abnormalities to help guide decision-making in non-high-risk patients, as the presence of high-risk mutations or patients with unfavorable karyotypes identified in the low-risk group who demonstrated a more aggressive disease course. Conversely, the presence of favorable mutations (in PHF6, TET2) predicted a more indolent course in patients assigned to the intermediate-risk group. Nevertheless, using molecular risk variables alone failed to significantly differentiate the 3 risk groups in the validation cohorts (low vs intermediate in the MD Anderson cohort and intermediate vs high in the Milan cohort), and validation of the favorable prognostic significance of PHF6 and TET2 mutations in other large cohorts is needed given the enrichment of otherwise low-frequency PHF6-mutated patients in the Mayo cohort.

Applying a risk stratification model for a complex disease comes with many challenges, shortcomings, and questions when caring for patients with CMML: There is subjectivity in identifying blasts and promonocytes (blast equivalents) included in this model, and clinicians should be vigilant to those cells misclassified as “reactive lymphocytes” or reported under “immature granulocytes” by automated cell counters. Can BLAST and BLAST-mol confirm the benefit of HSCT in the intermediate- and high-risk groups and predict their survival posttransplant?¹⁰ Should patients with intermediate- and high-risk disease proceed with HSCT soon after diagnosis or wait till they show signs of disease progression? Patients with oligomonocytic CMML are currently included under CMML in the ICC/WHO criteria, but would they be better stratified using BLAST/BLAST-mol, IPSS-M, or CPSS-Mol? As this model used the ICC diagnostic criteria for CMML, which requires a variant allele frequency (VAF) ≥10%, will BLAST-mol show similar prognostic accuracy in patients with VAF <10%? Can BLAST/BLAST-mol be applied to patients with CMUS? Can this new model identify patients with treatment-related CMML that are relatively “lower risk” and benefit from watchful waiting? Can BLAST/BLAST-mol be used in other MDS/MPN overlap syndromes that share similar mutational profiles such as MDS/MPN with neutrophilia (atypical chronic myeloid leukemia) and MDS/MPN, not otherwise specified? Lastly, given the lack of clear disease-modifying activity and potential toxicity of HMAs, patients with CMML should not reflexively receive such therapy solely based on assignment to the “high-risk” group.

Drug development and clinical trials in CMML have not kept pace with other myeloid malignancies and are mostly extrapolated from MDS, MPNs, or AML or from small cohorts of patients with CMML enrolled in MDS/AML trials looking for a “signal.” This beseeches academia, industry, regulatory authorities, and patient advocacy groups to develop innovative collaborations for CMML-specific clinical trials. The quest to find a disease-modifying therapy in CMML will hopefully flip the ratio of prognostic models to effective therapies.

Conflict-of-interest disclosure: W.S. has served as a consultant for, and received research funding for the conduct of clinical trials from, Incyte, Inc.