Validation of a Multivariate Serum Profile for Epithelial Ovarian Cancer Using a Prospective Multi-Site Collection

Abstract

In previous studies we described the use of a retrospective collection of ovarian cancer and benign disease samples, in combination with a large set of multiplexed immunoassays and a multivariate pattern recognition algorithm, to develop an 11-biomarker classification profile that is predictive for the presence of epithelial ovarian cancer. In this study, customized, Luminex-based multiplexed immunoassay kits were GMP-manufactured and the classification profile was refined from 11 to 8 biomarkers (CA-125, epidermal growth factor receptor, CA 19-9, C-reactive protein, tenascin C, apolipoprotein AI, apolipoprotein CIII, and myoglobin). The customized kits and the 8-biomarker profile were then validated in a double-blinded manner using prospective samples collected from women scheduled for surgery, with a gynecologic oncologist, for suspicion of having ovarian cancer. The performance observed in model development held in validation, demonstrating 81.1% sensitivity (95% CI 72.6 – 87.9%) for invasive epithelial ovarian cancer and 85.4% specificity (95% CI 81.1 – 88.9%) for benign ovarian conditions. The specificity for normal healthy women was 95.6% (95% CI 83.6 – 99.2%). These results have encouraged us to undertake a second validation study arm, currently in progress, to examine the performance of the 8-biomarker profile on the population of women not under the surgical care of a gynecologic oncologist.