Genomic and ancestry-associated determinants of survival in gynecologic cancers

Survival outcomes for gynecologic cancers vary across populations; however, the genomic and other factors that contribute to these differences are poorly defined. We leveraged a large, real-world dataset to examine how genetic ancestry, tumor genomics, and histology are associated with overall survival (OS) in endometrial (EC), ovarian (OC), and cervical cancer (CC). This retrospective analysis used integrated data from whole-exome sequencing (WES), medical claims, mortality records, and electronic health records in Natera's proprietary de-identified Real-World Database. Tumor and germline WES data were obtained as part of the workflow for a personalized, tumor-informed mPCR-NGS circulating-tumor DNA assay (Signatera TM , Natera Inc.). Variants, microsatellite instability (MSI), and genetic ancestry were analyzed. Ancestry was inferred by EthSeq and classified into predefined groups: African (AFR), American, East Asian (EAS), European (EUR), and South Asian. Cancer histologies were extracted from clinical records using natural language processing. OS associations were evaluated using Cox proportional hazards models. A total of 2,574 EC, 2,425 OC, and 773 CC patients were included. In EC, AFR ancestry was associated with worse OS compared to EUR ancestry (HR 1.56, p=0.013), especially in stage I/II patients (HR 2.1, p=0.03). The most frequent variants in EC were in PTEN (46.5% of patients), PIK3CA (40.6%), TP53 (38.0%), and ARID1A (37.8%). Variants in TP53 (stage I/II: HR 2.1, p<0.001; stage III/IV: HR 2.1, p<0.0001) and KMT2B (stage III/IV: HR 1.48, p=0.04) were linked with worse OS, while variants in PTEN (stage I/II: HR 0.5, p=0.002; stage III/IV: HR 0.44, p<0.0001) and ARID1A (stage I/II: HR 0.6, p=0.02; stage III/IV: HR 0.5, p<0.0001) were associated with better OS. Microsatellite stable (MSS) EC tumors had a worse OS than MSI-High (MSS vs MSI-H: HR 2.0, p<0.001). In OC, stage I AFR patients had worse OS compared to EUR patients (HR 7.5, p=0.04). OC tumor variants were most common in TP53 (59.7%), PIK3CA (11.4%) , and ARID1A (10.5%), consistent with established OC biology. No associations with OS and variants, MSI, or histology were identified. In CC, AFR ancestry was associated with worse OS vs EUR (HR 2.0, p=0.04). PIK3CA variants were common (29.5% of patients) and associated with worse OS in stage II/III/IV patients (HR 1.9, p=0.015). No OS differences by histology were observed. Across cancers, AFR and EAS ancestry showed distinct histologic distributions compared with EUR. In conclusion, genetic ancestry was independently associated with OS across gynecologic cancers, particularly among AFR-ancestry patients with EC and CC. Specific genomic features (e.g., MSI status and TP53 , PTEN , ARID1A, PIK3CA, and KMT2B variants) further define risk. Further study of ancestry-informed genomic and pathologic data into clinical research may reduce survival disparities in gynecologic malignancies.