Article Text

Download PDFPDF
1115 A systematic review of models predicting cytoreductive outcome in epithelial ovarian cancer
  1. Amy Hawarden
  1. Mater Misericordiae University Hospital, Dublin, Ireland


Introduction/Background The volume of remaining disease following cytoreductive surgery in stage III and IV high-grade serous epithelial cancer (HGSEOC) is an independent prognostic marker of survival. Primary cytoreductive surgery (PCRS) resulting in no residual disease (R0) translates to optimal rates of overall and progression free survival (OS and PFS), and conversely when residual disease remains patients suffer worse outcomes. To improve outcome, it is paramount to achieve the correct pathway selection for all patients. Pre-operative multi-modal decision-making tools aiming to predict surgical outcome exist in the literature, however few are used in practice and none consistently. A systematic review of published models and their performance was conducted.

Methodology Multimodal prognostic models predicting outcome of PCRS in patients with HGSEOC stage II-IV between 2010–2023 were identified. Research question was formulated and designed using the PICOS strategy. Data extraction was performed using CHARMS, and risk of bias was assessed using PROBAST.

Results 33 publications met eligibility criteria and were included in the final review. The publications described the initial development of 33, and the external validation of 16 pre-existing models. All were developed using retrospective data and patient numbers varied between 31 and 668. Prediction modalities included patient and disease characteristics, as well as biological, genomic, radiological and laparoscopic predictors. Model performance varied greatly, with area under the curve performance ranging between 0.73 and 0.91. Laparoscopic based models performed most consistently. Performance of models decreased with external validation in all cases. PROBAST analysis revealed a high risk of bias in 93% of studies.

Conclusion Most models displayed high risk of bias. Therefore, none would be appropriate to be used routinely in clinical practice in their current form outside the development cohort. Using all available predictors during model development, as well as incorporating surgical heterogeneity may improve performance and warrants further study.

Disclosures Nil.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.