Development and validation of a risk-calculator for adverse perioperative outcomes for women with ovarian cancer

Abstract

Primary cytoreduction followed by platinum-based chemotherapy is the primary treatment for advanced ovarian cancer. However, neoadjuvant chemotherapy followed by interval debulking is an alternative option, particularly in those who may be poor surgical candidates. The objective of this study was to determine factors associated with short-term, significant perioperative morbidity and mortality for women undergoing surgery for ovarian cancer and to create a nomogram to predict the risk of adverse perioperative outcomes. We used the National Surgical Quality Improvement Program database to identify women with ovarian, fallopian tube, or primary peritoneal cancer who underwent surgery from 2011 to 2015. Demographic factors, clinical characteristics, comorbidity, functional status, and the extent of surgery were used to predict the risk of severe perioperative complications or death using multivariable models. Multiple imputation methods were employed for missing data. A nomogram was developed based on the final model. The discrimination ability of the model was assessed with a calibration plot and discrimination concordance index. We identified a total of 7029 patients. Overall, 5.8% of patients experienced a Clavien-Dindo IV complication, 9.8% of patients were readmitted, 3.0% of patients required a reoperation, and 0.9% of patients died within 30 days. Among the baseline variables assessed, increasing age, emergent surgery, ascites, bleeding disorder, low albumin, higher American Society of Anesthesiology classification score, and a higher extended procedure score were associated with serious perioperative morbidity or mortality. Of these factors, performance of ≥3 cytoreductive procedures (adjusted odds ratio 4.53, 95% confidence interval 3.01-6.82), American Society of Anesthesiology classification score ≥ class 4 (adjusted odds ratio 2.89, 95% confidence interval 1.17-7.14), bleeding disorder (adjusted odds ratio 2.73, 95% confidence interval 1.82-4.10), and age ≥80 years (adjusted odds ratio 2.46, 95% confidence interval 1.66-3.63) were most strongly associated with risk of an event. The final nomogram included the above variables and had an internal discrimination concordance index of 0.71, with accurate predictions in an internal validation set, indicating a 71% correct identification of patients across all possible pairs. Women undergoing surgery for ovarian cancer are at significant risk for the occurrence of adverse perioperative outcomes. Using readily identifiable characteristics, this nomogram can predict adverse outcomes.