Nomograms for Predicting Risk and Prognosis of Liver Metastases in Ovarian Cancer Patients.

Abstract

Aims: Liver metastases (LiM) commonly manifest in ovarian cancer (OC) patients. We intended to establish nomograms for predicting the risk and prognostic factors in OCLiM patients.

Methods: Data from the SEER database (Nov 2022, Sub 1992-2020) were analyzed, excluding patients with missing data on liver metastases, survival months, race, AJCC T stage, marital status, rural/urban status, and metastases to bone, brain, or lung. Logistic and Cox regression analyses identified risk and prognostic factors for liver metastases. Predictive nomograms were developed from the multivariable regression results. The nomograms were evaluated using Harrell's C-index, ROC curve, calibration curve, DCA, NRI, and IDI. Moreover, the efficacy of the treatment in the new risk stratification subgroups was demonstrated by Kaplan-Meier (KM) curves.

Results: Among 17,056 OC patients, 5.67% (n = 967) had liver metastases. Nomograms were constructed based on identified risk and prognostic factors, with dynamic web-based nomograms developed for clinical use. The nomogram demonstrated C-index values of 81.9% (training) and 82.9% (validation) for predicting liver metastases. For OS and CSS, the C-index values were 73.3% and 73.7% (training), and 73.3% and 72.8% (validation), respectively. The ROC curves for OS at 1-, 3-, 5-year showed AUC values of 84.1%, 79.8%, 75.9% (training) and 82.9%, 78.5%, 82.2% (validation), respectively. For CSS, AUC values at 1-, 3-, and 5-year were 84.5%, 80.2%, 76.1% (training) and 82.6%, 78.0, 82.0% (validation), respectively. The calibration and DCA curves confirmed favorable performance. NRI and IDI analyses showed superiority over the Grade and AJCC stage systems. Surgery improved prognosis in the low-risk group, while chemotherapy was more effective in both low- and medium-risk groups.

Conclusions: we developed nomograms and risk stratification systems to assist clinicians in the individualized prediction, risk stratification, and prognostic assessment of OCLiM patients.