Improving risk-stratification of patients with chronic lymphocytic leukemia using multivariate patient similarity networks. (April 2019)

• Record Type: Journal Article
• Authors: Turcsanyi, Peter; Kriegova, Eva; Kudelka, Milos; Radvansky, Martin; Kruzova, Lenka; Urbanova, Renata; Schneiderova, Petra; Urbankova, Helena; Papajik, Tomas
• Is Part Of: Leukemia research. Volume 79(2019)
• Page Start: 60
• Page End: 68
• Publication Date: 2019-04
• Subjects: BM bone marrow -- CLL chronic lymphocytic leukemia -- CLL-IPI CLL-International Prognostic Index -- CK complex karyotype -- del deletion -- HR-CLL high-risk chronic lymphocytic leukemia -- LN lymph nodes -- OS overall survival -- PB peripheral blood -- PFS progression-free survival -- PSN patient similarity networks -- TTFT time to first treatment; Prognostication -- Multivariate and network-based approaches -- Precision medicine -- CLL -- Risk patient subsets; Leukemia -- Periodicals; Leukemia -- Periodicals; Leucémie -- Périodiques; Leukemia; Periodicals; Electronic journals; Electronic journals; 616.9941905
• DOI: 10.1016/j.leukres.2019.02.005
• Languages: English
• ISSNs: 0145-2126

Abstract

Highlights: Patient similarity network (PSN) is a new approach for patient stratification. PSN clusters patients based on their similarities in various features. PSN supports heterogeneous data (demographic/laboratory/clinical data, genomics). PSN has excellent model interpretability utilizing selections of relevant features. PSN is suited for precision medicine. Abstract: Background: Better risk-stratification of patients with chronic lymphocytic leukemia (CLL) and identification of subsets of ultra-high-risk (HR)-CLL patients are crucial in the contemporary era of an expanded therapeutic armamentarium for CLL. Methods: A multivariate patient similarity network and clustering was applied to assess the prognostic values of routine genetic, laboratory, and clinical factors and to identify subsets of ultra-HR-CLL patients. The study cohort consisted of 116 HR-CLL patients (F/M 36/80, median age 63 yrs) carrying del(11q), del(17p)/ TP53 mutations and/or complex karyotype (CK) at the time of diagnosis. Results: Three major subsets based on the presence of key prognostic variables as genetic aberrations, bulky lymphadenopathy, splenomegaly, and gender: profile (P)-I (n = 34, men/women with CK + no del(17p)/ TP53 mutations), P-II (n = 47, predominantly men with del(11q) + no CK + no del(17p)/ TP53 mutations), and P-III (n = 35, men/women with del(17p)/ TP53 mutations, with/without del(11q) and CK) were revealed. Subanalysis of major subsets identified three ultra-HR-CLL groups: men with TP53 disruption with/without CK, women with TP53 disruption with CK and men/women with CK + del(11q) with poor short-term outcomes (25% deaths/12 mo). Besides confirming the combinations of known risk-factors, the used patient similarity network added further refinement of subsets of HR-CLL patients who may profit from different targeted drugs. Conclusions: This study showed for the first time in hemato-oncology the usefulness of the multivariate patient similarity networks for stratification of HR-CLL patients. This approach shows the potential for clinical implementation of precision medicine, which is especially important in view of an armamentarium of novel targeted drugs.