Contemporary challenges in long-term survival prediction in resected pancreatic ductal adenocarcinoma: a transatlantic multicenter development and validation of prognostic models.

BACKGROUND

Predicting long-term survival (>5 years; LTS) in resected pancreatic ductal adenocarcinoma (PDAC) remains challenging. The aim of this study was to train and evaluate LTS prediction models.

METHODS

We retrospectively included patients with PDAC who underwent resection between 2012 and 2019 from the databases at New York University, the Johns Hopkins Hospital, University of Verona Hospital Trust, and the Dutch Pancreatic Cancer Group. Two models were developed for the (1) post-operative and (2) post-adjuvant treatment phase. Training involved the full dataset followed by internal-external cross validation.

RESULTS

4084 patients with resected PDAC were included. The estimated rate of LTS in (1) was 22% (95% CI: 21-24%) and in (2) 24% (95% CI: 22-26%). For model (1) 5-year performance metrics were an AUC of 0.68 (0.60-0.75), O/E ratio of 0.91 (0.45-1.82), and slope of 1.07 (0.56-2.05). Model (2) achieved 5-year AUC of 0.70 (0.64-0.75), O/E ratio of 0.96 (0.60-1.54), and slope of 1.16 (0.60-1.23).

CONCLUSIONS

Our models achieved only modest performance despite a large, granular dataset and rigorous statistical methods, demonstrating the need for novel prognostic biomarkers.