319P - Survival prediction for ovarian cancer patients from Taiwan cancer registry data

Background

Ovarian cancer (OC) is a grave disease and is one of the top 10 causes of cancer-related deaths in women both worldwide and specifically in Taiwan. OC is difficult to get diagnosed early leading to its high mortality rate. OC demonstrates heterogeneity with its subtypes demonstrating unique incidence and survival rates, which also vary among populations with distinct genetic backgrounds. Hence, this study aims to introduce and validate stratification models that can potentially play pivotal role in enhancing the prevention and treatment strategies for OC among Taiwanese patients.

Methods

Patients registered in the Taiwan Cancer Registry (TCR), diagnosed between January 1, 2009, and December 31, 2015, were analyzed. Follow-up data was collected until December 31, 2017. Two distinct survival prediction models were formulated: Model 1 incorporated clinical variables from TCR, that overlapped with Surveillance, Epidemiology, and End Results (SEER) dataset. Model 2 included additional cancer-specific variables from TCR, with the intention of any potential enhancement in prediction accuracy. For external validation patients of White, Black, and Asian ancestry from SEER, collected within the identical study-period as TCR, were employed.

Results

Cox-proportional hazards regression analyses were performed with death as the primary outcome. In Model 1, significant factors included age, histology subtype, tumor-grade, pathological M, Pathological N, and lymph-node-ratio. While in Model 2, significant variables were age, histology-subtype, tumor-grade, pathological T, pathological M, CA125 levels, and residual tumor. Evaluation revealed C-index > 0.7 for both models. Calibration analysis demonstrated that the proportional difference between predicted and observed survival was largely <5%.

Conclusions

Model 1 and Model 2 exhibited strong and robust predictive capabilities for survival of OC patients. Notably, no significant racial disparities in predictions were observed. Therefore, these models hold potential for utilization in clinical treatment settings, facilitating informed decision-making between patients and their healthcare providers.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Health Promotion Administration, Ministry of Health and Welfare, Taipei, Taiwan.

Disclosure

All authors have declared no conflicts of interest.