A predictive model for bone marrow disease in cytopenia based on noninvasive procedures

Abstract

Abstract Bone marrow specimens are the core of the diagnostic workup of patients with cytopenia. To explore whether next-generation sequencing (NGS) could be used to rule out malignancy without bone marrow specimens, we incorporated NGS in a model to predict presence of disease in the bone marrow of patients with unexplained cytopenia. We analyzed the occurrence of mutations in 508 patients with cytopenia, referred for primary workup of a suspected hematologic malignancy from 2015 to 2020. We divided patients into a discovery (n = 340) and validation (n = 168) cohort. Targeted sequencing, bone marrow biopsy, and complete blood count were performed in all patients. Mutations were identified in 267 (53%) and abnormal bone marrow morphology in 188 (37%) patients. Patients with isolated neutropenia had the lowest frequency of both mutations (21%) and abnormal bone marrow morphology (5%). The median number of mutations per patient was 2 in patients with abnormal bone marrow morphology compared with 0 in patients with a nondiagnostic bone marrow morphology (P < .001). In a multivariable logistic regression, mutations in TET2, SF3B1, U2AF1, TP53, and RUNX1 were significantly associated with abnormal bone marrow morphology. In the validation cohort, a model combining mutational status and clinical data identified 34 patients (20%) without abnormal bone marrow morphology with a sensitivity of 100% (95% confidence interval: 93%-100%). Overall, we show that NGS combined with clinical data can predict the presence of abnormal bone marrow morphology in patients with unexplained cytopenia and thus can be used to assess the need of a bone marrow biopsy.