A first-of-its-kind multi-omic assay using lipids and proteins to identify early-stage ovarian cancer in symptomatic individuals.

5552 Background: Ovarian cancer (OC) is often diagnosed at advanced stages, leading to one of the highest mortality rates for gynecologic malignancies. While 85% of patients are symptomatic even at early stages, symptoms are vague and overlap with benign conditions. Existing diagnostic tools such as CA125 and HE4 lack sufficient sensitivity and specificity to detect early-stage disease, even in symptomatic individuals. As a result, there is a critical unmet need for a highly sensitive diagnostic tool designed for patients presenting with vague abdominal symptoms. Methods: Here, we report the first large-scale, prospective U.S. clinical trial evaluating a targeted multi-omic assay of symptomatic women, its intended-use population. Assay performance was assessed in a subset of an independent, prospectively recruited, multi-site study enrolled across the United States, supplemented with samples from a commercial vendor (N=346). The cohort comprised patients diagnosed with OC across stages and subtypes (N=116: 50 early-stage I/II, 66 late-stage III/IV) and a symptomatic control group (N=230), comprising individuals with benign adnexal masses (BAN; N=116) and symptomatic individuals with vague abdominal symptoms who were followed longitudinally for 12 months to confirm absence of malignancy (symptomatic normal; N=114). Serum lipids including selected phospholipids, sphingolipids, and ceramides were quantified by mass spectrometry. Proteins were measured using clinically validated immunoassays. An ensemble machine learning modeling approach was applied using targeted feature subsets from a refined panel of 23 lipid and 5 protein markers, with outputs weighted to reflect anticipated disease prevalence in the intended-use population. Results: The assay demonstrates consistently high diagnostic performance, achieving 96.5% sensitivity with 80.6% specificity for OC vs. symptomatic control, and 92% sensitivity for early-stage OC in the same comparison. Integrating lipid and protein biomarkers into a targeted multi-omic model therefore results in a substantial improvement in OC detection, especially early-stage, over current standard-of-care biomarkers. Conclusions: In a large, prospectively collected cohort reflecting real-world symptomatic patients, this first-of-its-kind, targeted multi-omic assay accurately distinguishes OC from non-malignant conditions in symptomatic women with dramatic improvement in early-stage detection. These findings support the feasibility and clinical utility of a multi-omic assay in addressing a longstanding diagnostic gap for symptomatic women, clearing a path to earlier triage and diagnosis.