Using a machine learning algorithm to predict outcome of primary cytoreductive surgery in advanced ovarian cancer

This retrospective cohort study developed a random forest model to predict optimal cytoreduction and no gross residual disease among patients with advanced ovarian cancer undergoing primary cytoreductive surgery. The study included 151 patients assigned to training, validation, and test cohorts. For optimal cytoreduction, the model achieved areas under the curve of 99.8% in training, 89.6% in validation, and 89.0% in testing; its leading contributors included CA-125, albumin, diaphragmatic disease, age, and ascites. The results suggest that the model may assist treatment decision-making for appropriately selected patients.

Journal of Surgical Oncology, 127(3), 465–472