B-082 Evaluation of Minimal Residual Disease Detection Modalities in Multiple Myeloma: SPEP, Immunofixation, NGS, and Flow Cytometry

Abstract

Minimal residual disease (MRD) detection is increasingly central to treatment decisions in multiple myeloma (MM). Current IMWG guidelines recommend bone marrow–based MRD assessment using either next-generation sequencing (NGS) or flow cytometry, yet discordant results between these modalities create interpretive and operational challenges. This study evaluates the concordance between NGS and flow cytometry MRD testing and examines their performance alongside serum protein electrophoresis (SPE) and immunofixation electrophoresis (IFE), with additional assessment of a combined MRD classifier to support clinical decision-making. A retrospective dataset was constructed using Temple University Health System (TUHS) derived MRD results from flow cytometry, SPE, and IFE results, together with clonoSEQ NGS MRD data, were standardized into binary MRD classifications and used to construct a multimodal retrospective dataset. Analytical procedures included detection-rate estimation, inter-assay Pearson correlations, and confusion-matrix–based performance metrics for relapse prediction. Random Forest and multinomial logistic regression models were trained to classify remission, relapse, and unspecified status, with performance evaluated using sensitivity, specificity, F1 score, balanced accuracy, and one-vs-all AUC. Among 79 TUHS multiple myeloma cases, MRD detection rates differed substantially across modalities: FC detected residual disease in 26.6% of patients, NGS in 60.3%, SPE in 67.1%, and IFE in 75.9%. Concordance between methods was modest (r = 0.07–0.61), with the strongest correlations observed between SPE and IFE. A combined MRD classifier based on majority positivity demonstrated balanced performance (sensitivity 0.78, specificity 0.82, F1 = 0.61), outperforming most individual assays. FC showed high sensitivity (1.00) but low specificity, whereas SPE yielded the highest overall balanced accuracy (0.83). Discordance among MRD assays largely due to the lower ∼10?4 sensitivity of local flow cytometry relative to NGS highlights the utility of a combined MRD classifier to stabilize relapse assessment. Although literature reports ∼85% concordance between NGS and flow cytometry, sensitivity gaps can still yield divergent results. Once data collection is complete, training and validation/testing splits will be applied to further validate this multimodal approach.