Ovarian cancer is the deadliest gynecologic cancer in the United States. When detected early, the 5-year survival rate is 92%, although most cases remain undetected until the late stages where 5-year survival rates are 30%. Serum biomarkers may hold promise. Although many markers have been proposed and multivariate diagnostic models were built to fit the data on small, disparate sample sets, there has been no systematic evaluation of these markers on a single, large, well-defined sample set. To address this, we evaluated the dysregulation of 204 molecules in a sample set consisting of serum from 294 patients, collected from multiple collection sites, under a welldefined Gynecologic Oncology Group protocol. The population, weighted with early-stage cancers to assess biomarker value for early detection, contained all stages of ovarian cancer and common benign gynecologic conditions. The panel of serum molecules was assayed using rigorously qualified, high-throughput, multiplexed immunoassays and evaluated for their independent ovarian cancer diagnostic potential. Seventy-seven biomarkers were dysregulated in the ovarian cancer samples, although cancer antigen 125, C-reactive protein, epidermal growth factor receptor, interleukin 10, interleukin 8, connective tissue growth factor, haptoglobin, and tissue inhibitor of metalloproteinase 1 stood out as the most informative. When analyzed by cancer subtype and stage, there were differences in the relative value of biomarkers. In this study, using a large sample cohort, we show that some of the reported ovarian cancer biomarkers are more robust than others, and we identify additional informative candidates. These findings may guide the development of multivariate diagnostic models, which should be tested on additional, prospectively collected samples. (Cancer Epidemiol Biomarkers Prev 2008;17(10):2872 -81)