Nomogram Models for Predicting Risk and Prognosis of Newly Diagnosed Ovarian Cancer Patients with Liver Metastases - A Large Population-Based Real-World Study

Background: Previous studies about liver metastases (LM) in newly diagnosed ovarian cancer (NDOC) patients based on Surveillance, Epidemiology, and End Results (SEER) program disregarded selection bias of missing data. Methods: We identified Data of NDOC patients from SEER between 2010 and 2016, presented a comprehensive description of this dataset, and limited possible biases due to missing data by applying multiple imputation (MI). We determined predictive factors for underlying LM development in NDOC patients and evaluated prognostic factors in NDOC patients with LM (OCLM). We then established predictive nomograms, assessed by the concordance index, calibration curve, decision curve analysis (DCA), and clinical impact curves (CIC). Results: The amount of missing data for different variables in SEER dataset ranges from 0 to 36.11%. The results between complete dataset and MI datasets are similar. LM prevalence in NDOC patients was 7.18%, and median overall survival for OCLM patients was 11 months. The C-index of risk nomogram for LM development in the training cohort (TC) and validation cohort (VC) were 0.764 and 0.759, respectively. The C-index and integrated area under curve within five years of prognostic nomogram for OCLM patients in the TC and VC were 0.743 and 0.773, 0.714 and 0.733, respectively. For both nomograms, DCA revealed favorable clinical use and calibration curves suggested good consistency. Conclusion: The risk nomogram is expected to aid clinicians in identifying high-risk groups of LM development in NDOC patients for intensive screening. The prognostic nomogram could facilitate individualized prediction and stratification for clinical trials in OCLM patients.