791P - Development and validation of nomograms to predict survival in patients with high-grade serous ovarian cancer

Background

The current American Joint Committee on Cancer TNM and FIGO systems are without sufficient details encompassing the variety of ovarian cancer. To develop and validate nomograms for overall survival (OS) and cancer-specific survival (CSS) in patients with high-grade serous ovarian cancer (HGSOC).

Methods

The collection of data from California, Connecticut, Detroit, Georgia, Hawaii Iowa was formed as the development cohort. A cohort of data from Kentucky, Louisiana, New Jersey, New Mexico, Seattle, and Utah formed the external validation cohorts. We developed the nomograms using the Cox prediction model. We applied five-method comparisons to screen the best model for both nomograms, and the nomogram models underwent successful independent external validations to support their reliability. We assessed nomogram model performance by examining overall accuracy (Brier score), calibration (calibration plots and Hosmer–Lemeshow calibration test), and discrimination (Harrell C index).

Results

The development cohorts included 2465 patients which were split into two parts (7:3) for the development (n=1726) and internal validation (n=739) for the model, and the 5-year overall survival was 18·6% (458/2465). The OS nomogram showed that Scope Reg LN Sur (0-100), AJCC_N (0-89.37), Systemic Sur Seq (0-74.25), the year of diagnosis (0-69.15), and Surg Prim Site(0-51.52) shared the larger contribution to prognosis. Followed by AJCC_ M (0–48.42) and age (0–45.31). The C index of 5-year OS and CSS nomograms was 0.715 (95% CI 0.688–0.741) and 0.719 (0.693–0.745) in the development cohort, 0.695 (0.646–0.744) and 0.702 (0.653–0.750) in the internal validation cohorts. The external validation cohorts included 961 patients, were 0.684 (0.644–0.724) and 0.687 (0.648–0.727). The nomograms performed well according to overall accuracy, discrimination, and calibration when they were applied to the validation cohorts.

Conclusions

Our nomograms are credible methods that can be used to predict OS and CSS in patients with HGSOC. They should be offered to clinicians to assess patient prognosis, make prognosis-based decisions and be conducive to stratifying patients in clinical trials or population-based analyses, and inform patients clinically.

Clinical trial identification

Editorial acknowledgement

We gratefully thank the SEER database for providing data support. we also express sincere gratitude to Steve for guiding us on how to handle the data of alive or dead of other causes to perform Cause-specific survival analysis. At the same time, we thank the Free Statistics team for providing technical guidance and valuable tools for data analysis and visualization.

Legal entity responsible for the study

J. Liu.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.