End-of-life prognostic models in advanced cancer: a scoping review of model development, validation, and impact
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background Accurate survival estimation is important for clinical decision-making in advanced cancer. Prognostic models can support prediction of end-of-life and should ideally undergo staged evaluation across a development framework encompassing model development, validation, and impact evaluation before clinical implementation. No synthesis currently maps existing models across this framework, making it difficult to determine which tools are available and the extent to which they are ready for clinical use. This review aimed to identify end-of-life prediction models for advanced cancer and map them across this development framework. Methods This scoping review included studies reporting prognostic models combining at least two predictor variables to predict end-of-life (≤ 12 months) in adults with advanced cancer. MEDLINE, Embase, CINAHL, and the Cochrane Library were searched from inception to August 2025. Studies were classified as model development, internal validation, external validation, or impact evaluation, with findings summarised descriptively and narratively. Results A total of 104 studies were included: 17 (16%) reported development only, 34 (33%) internal validation, and 53 (51%) external validation. Across all studies, 74 distinct models were identified: 8 (11%) had undergone development only, 31 (42%) had been internally validated, and 35 (47%) externally validated. However, most externally validated models had been evaluated only once and within a single country, limiting confidence in their generalisability. No models had undergone impact evaluation. Conclusions Although many end-of-life prognostic models exist for advanced cancer, fewer than half have been externally validated and none have been assessed for clinical impact. Their validity, generalisability, and usefulness therefore remain uncertain. Future research should prioritise independent external validation and robust impact evaluation before routine clinical implementation.
