Differentiating between borderline and invasive malignancies in ovarian tumors using a multivariate logistic regression model

Abstract

Objective: The objective of this study was to build a model to differentiate between borderline and invasive ovarian tumors. Materials and Methods: We performed a retrospective study involving 148 patients with borderline or invasive ovarian tumors in our institute between 1997 and 2012. Clinical and pathologic data were collected. Logistic regression was used to build the model. Results: The model was created based on the following variables (p < 0.05): menopausal status; preoperative serum level of cancer antigen 125; the greatest diameter of the tumor; and the presence of solid parts on ultrasound imaging. The sensitivity and specificity of the model were 94.6% [95% confidence interval (CI), 0.887–1] and 78.3% (95% CI, 0.614–0.952) for patients aged ≥ 50 years, and 76.0% (95% CI, 0.622–0.903) and 60.0% (95% CI, 0.438–0.762) for those aged < 50 years, respectively. The performance of the model was tested using cross-validation. Conclusion: Differentiation between borderline and invasive ovarian tumors can be achieved using a model based on the following criteria: menopausal status; cancer antigen 125 level; and ultrasound parameters. The model is helpful to oncologists and patients in the initial evaluation phase of ovarian tumors.

Keywords

• borderline ovarian tumor
• cancer antigen 125
• invasive ovarian tumor
• logistic regression