P-135: Evaluating Racial Differences in the Systemic Impact of Monoclonal Protein in Multiple Myeloma Using Machine Learning

Introduction: Multiple myeloma (MM) is a hematologic malignancy characterized by the clonal proliferation of plasma cells, leading to systemic effects on immune function, hematopoiesis, renal function, bone metabolism, and electrolyte homeostasis. While the clinical outcomes for MM have significantly improved with advancements in therapeutic strategies, the influence of race on disease biology and treatment response remains an intriguing and critical area of investigation. Racial disparities in healthcare outcomes have been well-documented, and understanding these disparities in the context of MM is essential for optimizing patient care and treatment strategies. This study aims to employ a machine learning approach to evaluate the systemic impact of monoclonal protein (M-protein) and explore potential racial differences between African American and Caucasian patients.



Methods: The upper limit of observed M-spike was 3.5 g/dL. Forty-three clinical and laboratory variables were selected as predictors of M-spike values and were fed into a machine-learning model using the random forest algorithm. The model also included two lagged variables representing each subject's two preceding M-spike values. The dataset was randomly divided into a training set (80%) and a test set (20%). A regression tree was built using the training set and validated with the test set. Bootstrapping was employed to generate data sets (n=500) for model validation. The importance of each variable was assessed by excluding it from the model and evaluating the impact on the mean squared error (MSE).

Results: The training set comprised 749 patient-based observations. The residual distribution of the random forest model indicated that nearly all M-spike values predicted using the 43 variables were distributed equally on either side of zero. The inclusion of race as a variable did not significantly change the residual plots. The weighted value of each of the 43 independent variables was determined by individually removing a variable from the model and measuring its effect on the MSE. Key variables, including the first lagged M-spike, serum total protein, second lagged M-spike, serum IgG, serum IgM, and serum IgA, had the greatest impact on the model. When race was added to the model, it had a low weighted value compared to other variables. The Pearson correlation coefficient between predicted and actual M-spike values was 0.96, and the Spearman correlation coefficient was 0.91. With the inclusion of race, these coefficients were 0.95 and 0.94, respectively, indicating no significant difference.

Conclusions: Our machine learning analysis supports the hypothesis that there is no significant difference in the systemic impact of MM between African American and Caucasian patients, as measured by M-spike levels. These findings highlight the robustness of the model and suggest that racial differences may not play a major role in the systemic effects of MM.