Machine learning to identify racial and socioeconomic disparities in the onset and tumor distribution of second primary malignancies in multiple myeloma.

Abstract

7551 Background: As survival in Multiple Myeloma (MM) improved with novel therapies, the burden of second primary malignancies (SPM) has emerged as a significant competing risk. While racial disparities in MM outcomes are established, it remains unclear whether the the timing and type of SPM differ by race or socioeconomic status. Methods: We analyzed 7,513 MM survivors from the SEER Research database (2000–2022) who developed a confirmed, invasive SPM (excluding non-melanoma skin cancers). To account for the shorter follow-up duration in the modern era, latency drivers were assessed using Fine-Gray competing risk models which adjust for censoring. Racial and socioeconomic differences were quantified using logistic regression adjusted for age, sex, and median household income. Random Survival Forests were used to identify predictors of SPM latency with variable importance ( VIMP) indicating the contributor for each factor. Results: The mean time-to-SPM decreased from 5.18 years (2000–2010) to 2.61 years (2011–2022) ( p < 0.001 ). Black patients developed SPMs earlier than White patients (Average Direct Effect: -0.33 years, p < 0.001 ) and at a younger mean age for secondary Prostate cancer (68.5 vs. 71.6 years, p < 0.001 ). In Random Survival Forest modeling, biological age (VIMP=0.15) was the primary predictor of latency, followed by Stem Cell Transplant and Radiation utilization (VIMP=0.03). Multivariable linear regression indicated that patients with low household income (<$60k) developed SPMs 0.34 years faster than wealthier peers ( p = 0.009 ). Black patients had significantly lower odds of therapy-related hematologic malignancies (OR 0.62, 95% CI 0.49–0.77; p < 0.001 ) but higher odds of solid tumors (OR 1.62, 95% CI 1.30–2.04; p < 0.001 ). This solid tumor disparity persisted after adjusting for age and income ( p=0.27 ) and was driven by Prostate (OR 1.79, 95% CI 1.54–2.09) and Breast cancer (OR 1.31, 95% CI 1.06–1.62). Post-SPM mortality hazard was higher in Black patients compared to White patients (HR 1.15, p < 0.001 ). Conclusions: Our findings showed racial and socioeconomic factors impact both the timing and type of SPM in MM survivors. Black and low-income patients develop SPM earlier and are more likely to develop solid tumors, especially prostate and breast. These findings suggests need for surveillance strategies and earlier focused screening for high risk populations to address disparities in MM survivorship.