Machine Learning Improves Risk Stratification in Myelofibrosis: An Analysis of the Spanish Registry of Myelofibrosis.

Myelofibrosis (MF) is a myeloproliferative neoplasm (MPN) with heterogeneous clinical course. Allogeneic hematopoietic cell transplantation remains the only curative therapy, but its morbidity and mortality require careful candidate selection. Therefore, accurate disease risk prognostication is critical for treatment decision-making. We obtained registry data from patients diagnosed with MF in 60 Spanish institutions (N = 1386). These were randomly divided into a training set (80%) and a test set (20%). A machine learning (ML) technique (random forest) was used to model overall survival (OS) and leukemia-free survival (LFS) in the training set, and the results were validated in the test set. We derived the AIPSS-MF (Artificial Intelligence Prognostic Scoring System for Myelofibrosis) model, which was based on 8 clinical variables at diagnosis and achieved high accuracy in predicting OS (training set c-index, 0.750; test set c-index, 0.744) and LFS (training set c-index, 0.697; test set c-index, 0.703). No improvement was obtained with the inclusion of MPN driver mutations in the model. We were unable to adequately assess the potential benefit of including adverse cytogenetics or high-risk mutations due to the lack of these data in many patients. AIPSS-MF was superior to the IPSS regardless of MF subtype and age range and outperformed the MYSEC-PM in patients with secondary MF. In conclusion, we have developed a prediction model based exclusively on clinical variables that provides individualized prognostic estimates in patients with primary and secondary MF. The use of AIPSS-MF in combination with predictive models that incorporate genetic information may improve disease risk stratification.

Genomics improves risk stratification of adults with T-cell acute lymphoblastic leukemia enrolled in measurable residual disease-oriented trials.

Genetic information has been crucial to understand the pathogenesis of T-cell acute lymphoblastic leukemia (T-ALL) at diagnosis and at relapse, but still nowadays has a limited value in a clinical context. Few genetic markers are associated with the outcome of T-ALL patients, independently of measurable residual disease (MRD) status after therapy. In addition, the prognostic relevance of genetic features may be modulated by the specific treatment used. We analyzed the genetic profile of 145 T-ALL patients by targeted deep sequencing. Genomic information was integrated with the clinicalbiological and survival data of a subset of 116 adult patients enrolled in two consecutive MRD-oriented trials of the Spanish PETHEMA (Programa Español de Tratamientos en Hematología) group. Genetic analysis revealed a mutational profile defined by DNMT3A/ N/KRAS/ MSH2/ U2AF1 gene mutations that identified refractory/resistant patients. Mutations in the DMNT3A gene were also found in the non-leukemic cell fraction of patients with T-ALL, revealing a possible mutational-driven clonal hematopoiesis event to prime T-ALL in elderly. The prognostic impact of this adverse genetic profile was independent of MRD status on day +35 of induction therapy. The combined worse-outcome genetic signature and MRD on day +35 allowed risk stratification of T-ALL into standard or high-risk groups with significantly different 5- year overall survival (OS) of 52% (95% confidence interval: 37-67) and 17% (95% confidence interval: 1-33), respectively. These results confirm the relevance of the tumor genetic profile in predicting patient outcome in adult T-ALL and highlight the need for novel gene-targeted chemotherapeutic schedules to improve the OS of poor-prognosis T-ALL patients.

CNL and aCML should be considered as a single entity based on molecular profiles and outcomes.

Chronic neutrophilic leukemia (CNL) and atypical chronic myeloid leukemia (aCML) are rare myeloid disorders that are challenging with regard to diagnosis and clinical management. To study the similarities and differences between these disorders, we undertook a multicenter international study of one of the largest case series (CNL, n = 24; aCML, n = 37 cases, respectively), focusing on the clinical and mutational profiles (n = 53 with molecular data) of these diseases. We found no differences in clinical presentations or outcomes of both entities. As previously described, both CNL and aCML share a complex mutational profile with mutations in genes involved in epigenetic regulation, splicing, and signaling pathways. Apart from CSF3R, only EZH2 and TET2 were differentially mutated between them. The molecular profiles support the notion of CNL and aCML being a continuum of the same disease that may fit best within the myelodysplastic/myeloproliferative neoplasms. We identified 4 high-risk mutated genes, specifically CEBPA (ß = 2.26, hazard ratio [HR] = 9.54, P = .003), EZH2 (ß = 1.12, HR = 3.062, P = .009), NRAS (ß = 1.29, HR = 3.63, P = .048), and U2AF1 (ß = 1.75, HR = 5.74, P = .013) using multivariate analysis. Our findings underscore the relevance of molecular-risk classification in CNL/aCML as well as the importance of CSF3R mutations in these diseases.

Predictors of cardiovascular events and all-cause of death in patients with transfusion-dependent myelodysplastic syndrome.

Cardiovascular disease (CVD) involves the second cause of death in low-risk myelodysplastic syndrome (MDS) population. Prospective study to characterise the CVD and to identify predictors for the combined event (CE) cardiovascular event and/or all-cause mortality in transfusion dependent low-risk MDS patients. Thirty-one patients underwent a cardiac assessment including biomarkers and cardiac magnetic resonance (cMR) with parametric sequences (T1, T2 and T2* mapping) and myocardial deformation by feature tracking (FT) and were analysed for clonal hematopoiesis of indeterminate potential mutations. Cardiac assessment revealed high prevalence of unknown structural heart disease (51% cMR pathological findings). After 2·2 [0·44] years follow-up, 35·5% of patients suffered the CE: 16% death, 29% cardiovascular event. At multivariate analysis elevated NT-proBNP ≥ 486pg/ml (HR 96·7; 95%-CI 1·135-8243; P = 0·044), reduced native T1 time < 983ms (HR 44·8; 95%-CI 1·235-1623; P = 0·038) and higher left ventricular global longitudinal strain (LV-GLS) (HR 0·4; 95%-CI 0·196-0·973; P = 0·043) showed an independent prognostic value. These variables, together with the myocardial T2* time < 20ms, showed an additive prognostic value (Log Rank: 12·4; P = 0·001). In conclusion, low-risk MDS patients frequently suffer CVD. NT-proBNP value, native T1 relaxation time and longitudinal strain by FT are independent predictors of poor cardiovascular prognosis, thus, their determination would identify high-risk patients who could benefit from a cardiac treatment and follow-up.

Fluorescence in situ hybridization of TP53 for the detection of chromosome 17 abnormalities in myelodysplastic syndromes.

Conventional G-banding cytogenetics (CC) detects chromosome 17 (chr17) abnormalities in 2% of patients with de novo myelodysplastic syndromes (MDS). We used CC and fluorescence in situ hybridization (FISH) (LSI p53/17p13.1) to assess deletion of 17p in 531 patients with de novo MDS from the Spanish Group of Hematological Cytogenetics. FISH detected - 17 or 17p abnormalities in 13 cases (2.6%) in whom no 17p abnormalities were revealed by CC: 0.9% of patients with a normal karyotype, 0% in non-informative cytogenetics, 50% of patients with a chr17 abnormality without loss of 17p and 4.7% of cases with an abnormal karyotype not involving chr17. Our results suggest that applying FISH of 17p13 to identify the number of copies of the TP53 gene could be beneficial in patients with a complex karyotype. We recommend using FISH of 17p13 in young patients with a normal karyotype or non-informative cytogenetics, and always in isolated del(17p).

Response to lenalidomide in myelodysplastic syndromes with del(5q): influence of cytogenetics and mutations.

Lenalidomide is an effective drug in low-risk myelodysplastic syndromes (MDS) with isolated del(5q), although not all patients respond. Studies have suggested a role for TP53 mutations and karyotype complexity in disease progression and outcome. In order to assess the impact of complex karyotypes on treatment response and disease progression in 52 lenalidomide-treated patients with del(5q) MDS, conventional G-banding cytogenetics (CC), single nucleotide polymorphism array (SNP-A), and genomic sequencing methods were used. SNP-A analysis (with control sample, lymphocytes CD3+, in 30 cases) revealed 5q losses in all cases. Other recurrent abnormalities were infrequent and were not associated with lenalidomide responsiveness. Low karyotype complexity (by CC) and a high baseline platelet count (>280 × 10(9) /l) were associated with the achievement of haematological response (P = 0·020, P = 0·013 respectively). Unmutated TP53 status showed a tendency for haematological response (P = 0·061). Complete cytogenetic response was not observed in any of the mutated TP53 cases. By multivariate analysis, the most important predictor for lenalidomide treatment failure was a platelet count <280 × 10(9) /l (Odds Ratio = 6·17, P = 0·040). This study reveals the importance of a low baseline platelet count, karyotypic complexity and TP53 mutational status for response to lenalidomide treatment. It supports the molecular study of TP53 in MDS patients treated with lenalidomide.