Can we predict a high risk group in stage I epithelial ovarian cancer?

A retrospective review of 373 patients with stage I invasive epithelial ovarian cancer was undertaken over a 5 year period to develop a model to characterize the patient at high risk. Actuarial 5-year survival was 70%. To identify factors with an independent effect on 5-year survival, a logistic regression analysis was performed. Adjuvant chemotherapy, histologic grade and peritoneal washings, were identified as independent variables. A model to determine the predictivity of survival was created using a learning sample (2/3 of the cases) and the model was then used to reclassify a validation sample (1/3 of the cases). Using all the independent variables, outcome was predicted correctly in 78% of cases. However the model failed to improve identification of those at risk of recurrence (specificity of 53%).