Risk prediction model for epithelial ovarian cancer using molecular markers and clinical characteristics

BACKGROUND: A high-quality risk prediction model is urgently needed for the clinical management of ovarian cancer. However most existing models are solely based on clinical parameters, and molecular classifications in recent reports are still being debated. This study aimed to establish a risk prediction model by using both clinicopathological and molecular factors (the synthetic model) for epithelial ovarian cancer. METHODS: A retrospective cohort study was conducted in epithelial ovarian cancer patients (n = 161) treated with primary debulking surgery and adjuvant chemotherapy. The expression level of 15 selected molecular markers were measured using immunohistochemistry. A risk model was developed using COX regression analysis with overall survival as the primary outcome. A simplified scoring system for each prognostic factor was based on its coefficient. Independent validation (n = 40) was conducted to evaluate the performance of the model. RESULTS: A total of 10 out of 15 molecular markers were significantly associated with clinical characteristics and overall survival. The synthetic model performed better than the clinicopathological risk model or the molecular risk model alone, as assessed by analysis of the receiver-operating characteristics curve area and the Youden index. The synthetic model included parity (>3), peritoneal metastasis, stage, tumor type, residual disease, and expression of human epidermal growth factor receptor 2 (HER2), epidermal growth factor receptor (EGFR), breast cancer 1 (BRCA1), murine sarcoma viral oncogene homolog B (BRAF) and Kirsten rat sarcoma viral oncogene homolog (KRAS). CONCLUSIONS: Our synthetic risk model may more accurately predict survival of epithelial ovarian cancer patients than current models. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1186/s13048-015-0195-6) contains supplementary material, which is available to authorized users.