Objectives: We aimed to develop a risk model to predict a risk of suboptimal cytoreduction in primary surgery of ovarian cancer.
Methods: The clinical records and computed tomography (CT) data of 358 patients with stages II-IV epithelial ovarian cancer were reviewed. Tumor spread patterns identified by principal component analysis, CA-125, and a newly developed surgical skill index were integrated into a logistic model along with other variables. Internal validation was performed using bootstrapped re-sampling and calibration was assessed by goodness-of-fit test.
Results: Among the 358 patients, optimal cytoreduction, which was defined as no residual tumor, was achieved in 145 patients (40.5%). The surgical capacity of an individual institution was estimated by a surgical skill index, which was the frequency of complex surgeries in patients with advanced disease. In a multivariate model, two distinctive CT patterns of tumor spread (diffuse spread pattern and upper abdominal extension pattern), a surgical skill index, and serum CA-125 independently predicted a risk of suboptimal cytoreduction (P=0.006, P=0.013, P=0.031, and P=0.001, respectively). The model showed a C-statistic of .73 (95% confidence interval .67 to .79), which was significantly higher than tumor stage or ascites. Rigorous internal validation by bootstrapped re-sampling successfully confirmed the model.
Conclusions: We identified two distinct tumor spread patterns of ovarian cancer, which can be integrated to improve a prediction model. Our model may be useful in patient referral or clinical trials for patient stratification.
Keywords: Computed tomography; Cytoreductive surgery; Optimal cytoreduction; Ovarian cancer; Spread pattern; Surgical skill index.
© 2013.