Assessing differences in miRNA profiles by race and ethnicity: Implications for creating an equitable ovarian cancer early detection test.

Abstract Introduction: Assessing serum micro RNA (miRNA) profiles is an emerging technique for earlier detection of ovarian cancer. A panel of miRNAs for ovarian cancer prediction has been proposed, but it is unknown whether race and ethnicity may influence their expression. The goal of this study was to determine how race affects expression of miRNAs associated with ovarian cancer risk to ensure creation of an equitable test. Methods: Serum samples from 1586 women from the Mass General Brigham BioBank were analyzed for the presence of 179 different miRNAs. The Abcam Fireplex® assay was used to assess miRNA expression determined by flow cytometry, measured in mean fluorescence units (MFI) and log2 normalized prior to analysis. Demographic characteristics and medical histories were abstracted from linked electronic medical records. Patients with missing race/ethnicity were excluded. Differences in miRNA profiles by race/ethnicity (white, non-Hispanic vs non-white) were assessed using dimension reduction. The miRNA expression data was mapped to two-dimensions using constrained least-squares, then the reduced dimension data was used to train a simple neural network to predict race/ethnicity. In addition, pairwise t-test analyses were performed to analyze racial/ethnic differences among eight miRNAs previously reported to be informative for prediction of ovarian cancer. Results: The racial/ethnic composition of the study population was similar to the state population (76.9% white, 13.4% Hispanic, 5.2% black, 2.6% Asian, 1.9% Other). Non-white patients were on average younger (41.9 years ± 13.2 vs 51.3 years ± 15.1, p<0.01) and had fewer comorbid conditions compared to white patients (3.5 comorbidities ± 2.7 vs 4.6 comorbidities ± 2.8, p<0.01). The dimension reduction and classification analysis showed a significant effect of race on miRNA profiles, with miRNAs able to predict race at an AUC of 0.71 (95% CI 0.68-0.74) after 10-fold cross validation. The AUC remained consistent when stratified for age, menopausal status, and comorbidities. Among eight miRNAs highly predictive of ovarian cancer, seven significantly varied by race (hsa-mir-150-5p, hsa-mir-200c-3p, hsa-mir-23b-3p, has-mir-29a-3p, hsa-mir-320c, hsa-mir-320d, and hsa-mir-32-5p, p<0.01). Hsa-mir-200c-3p was the most significantly different by race (7.19 mean log2 MFI among non-white vs 6.94 mean log2 MFI among white patients, p<0.01). Upon analyzing each racial subtype and variation in hsa-mir-200c-3p, black and Hispanic patients had relatively similar mean values which were significantly different from values observed among white patients. Conclusions: miRNA expression is significantly influenced by race and ethnicity, with differences remaining largely consistent even after controlling for confounders. Understanding baseline differences in biomarker test characteristics prior to clinical implementation of novel diagnostic tools is essential to ensuring instruments perform comparably across diverse populations. Citation Format: Stephanie Alimena, James Webber, Laura Wollborn, Marta Williams, Chad B. Sussman, Joyce Y. Wang, Julia Spiegel, Kevin M. Elias. Assessing differences in miRNA profiles by race and ethnicity: Implications for creating an equitable ovarian cancer early detection test [abstract]. In: Proceedings of the 15th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2022 Sep 16-19; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2022;31(1 Suppl):Abstract nr PR001.