Background: Primary cytoreduction surgery followed by chemotherapy is the cornerstone treatment for epithelial ovarian cancer (EOC). In patients with a low probability of optimal primary surgical debulking, neoadjuvant chemotherapy (NACT) followed by interval debulking increases the chance of optimal surgery. The aim of this study was to develop a model to identify preoperative predictors for suboptimal cytoreduction. Methods: Medical records of patients with EOC who underwent primary cytoreductive surgery in a referral tertiary gyneco-oncology center were reviewed from 2007 to 2017. Data were collected on a range of characteristics including demographic features, comorbidities, serum tumor markers, hematologic markers, preoperative imaging, surgical procedures, and pathologic reports. Univariate and multivariate analyses were performed to clarify the ability of preoperative factors to predict suboptimal primary surgery. Results: The majority of patients (71.3%) who underwent primary cytoreductive surgery were optimally debulked. Based on the Youden index, the best cut-off point for the serum CA125 level to distinguish suboptimal debulking was 420U/ml with 0.730 (95%CI:0.559 to 0.862) sensitivity and 0.783 (0.684 to 0.862) specificity. Multiple logistic regression results showed that serum CA125 level >420 U/ ml (p value <0.001), the presence of liver metastasis on preoperative imaging (p value: 0.041) and ascites (p value: 0.032) or massive ascites (p value:0.010) significantly increased the risk of suboptimal debulking (logit p = 2.36 CA125 level +1.85 Liverinvolvement +1.68 presence of Ascites+ 2.28 Massive Ascites). Conclusion:The present study suggests that a serum CA125 level >420 U/ml, the presence of ascites or massive ascites and liver metastasis are strong predictors of suboptimal primary surgery in cases of EOC. Based on the constructed model, with any of these 4 factors, the probability of suboptimal debulking in EOC is more than 80%.
Keywords: Optimal debulking; neoadjuant chemotherapy; CA125; prediction model.
Creative Commons Attribution License