OAB-056: A machine learning model based on tumor and immune biomarkers to predict undetectable measurable residual disease (MRD) in transplant-eligible multiple myeloma (MM)

Background There is expectation of using biomarkers to personalize treatment. Yet, a successful treatment selection cannot be confirmed before 5 or 10 years of progression-free survival (PFS). Treatment individualization based on the probability of an individual patient to achieve undetectable MRD with a singular regimen, could represent a new model towards personalized treatment with fast assessment of its success. This idea has not been investigated previously. Methods We sought to define a machine learning model to predict undetectable MRD in newly-diagnosed transplant-eligible MM patients (n=278). The training (n=152) and internal validation cohort (n=60) consisted of 212 active MM patients enrolled in the GEM2012MENOS65 clinical trial. The external validation cohort was defined by 66 high-risk smoldering MM patients enrolled in the GEM-CESAR clinical trial, which treatment differed only by the substitution of bortezomib by carfilzomib during induction and consolidation. Patients were included in the study based on data availability of 17 parameters (p≤.05) associated with MRD outcomes. Results We started by investigating patients’ MRD status after VRD induction, HDT/ASCT and VRD consolidation (GEM2012MENOS65) according to their ISS and Revised-ISS, LDH levels, and cytogenetic alterations. High LDH levels and del(17p13), two features relatively infrequent at diagnosis, were the only parameters associated with lower rates of undetectable MRD. The ISS and R-ISS were not predictive. Therefore, we aimed to evaluate other disease features associated with MRD outcomes and develop more effective models based on machine learning logistic regression. The most effective one resulted from integrating cytogenetic [t(4;14) and/or del(17p13)], tumor burden (plasma cell clonality in bone marrow and circulating tumor cells in peripheral blood) and immune related (myeloid precursors, mature B cells, intermediate neutrophils, eosinophils, CD27negCD38pos T cells and CD56brightCD27neg NK cells) biomarkers. Data obtained for an individual patient can be substituted into our formula, which results in a numerical probability of achieving undetectable (>0.5) vs persistent (0.685 or Conclusions We demonstrated that selecting a regimen based on probable MRD outcomes, and confirming soon after if that probability was accurate, is a possible new approach towards individualized treatment in MM. The model is available at www.MRDpredictor.com to facilitate its use in clinical practice.