Multicohort Development and Validation of a Reproducible Vaginal Microbiome Model for Endometrial Cancer Detection

Associations between the vaginal microbiota and gynecological cancers have been documented, but inconsistencies across studies limit clinical translation. Endometrial cancer (EC), the most common gynecological malignancy in high-income countries, currently lacks a screening strategy. This retrospective multicohort study reanalyzes vaginal 16S rRNA data from five independent cohorts (n = 265) to identify microbial signatures associated with EC and develop a predictive machine-learning model. The study included women with histologically confirmed EC and controls with benign gynecologic conditions. Microbial diversity was higher in EC samples, with host characteristics influencing community composition. Peptoniphilus was reproducibly enriched in EC. An ensemble classifier identified all EC pathologies in a held-out test set with an area under the ROC curve of 0.93 (95% CI: 0.71 - 0.93), along with a sensitivity and negative predictive value of 1. These findings support the potential of vaginal microbiome profiling as a minimally invasive approach for early EC detection.