Validation and modification of the PAncreatic surgery Composite Endpoint (PACE) model: A single-institution study from Southeast Asia

Abstract

Background

The PAncreatic surgery Composite Endpoint was developed as a standardized, clinically relevant outcome measure for pancreatectomies, and it has been shown to correlate with prolonged length of stay and 90-day mortality. Although the PAncreatic surgery Composite Endpoint combines key complications of postoperative pancreatic fistula, postpancreatectomy hemorrhage, and reoperation/reintervention into a single measure, delayed gastric emptying has not been integrated. We aimed to validate PAncreatic surgery Composite Endpoint and assess whether adding delayed gastric emptying (modified PAncreatic surgery Composite Endpoint) enhances its predictive performance in a Southeast Asian cohort.

Methods

Prolonged length of stay was defined as ≥20 days (the 75th percentile of length of stay). PAncreatic surgery Composite Endpoint and modified PAncreatic surgery Composite Endpoint were evaluated as binary indicators using logistic regression analyses to assess their associations with prolonged length of stay and 90-day mortality. Model performance was evaluated using the area under the receiver operating characteristic curve, sensitivity, specificity, and calibration curves.

Results

A total of 575 patients underwent pancreatectomy at Singapore General Hospital between 2018 and 2023. When predicting prolonged length of stay, modified PAncreatic surgery Composite Endpoint (area under the receiver operating characteristic curve, 0.781; sensitivity, 79.2%; specificity, 77.0%) outperformed PAncreatic surgery Composite Endpoint (area under the receiver operating characteristic curve, 0.669; sensitivity, 53.0%; specificity, 80.8%), demonstrating stronger discrimination in both pancreatoduodenectomy and distal pancreatectomy subcohorts and retaining good calibration. For postoperative 90-day mortality, PAncreatic surgery Composite Endpoint and modified PAncreatic surgery Composite Endpoint demonstrated moderate, comparable performance (area under the receiver operating characteristic curves of 0.682 and 0.679, respectively) and exhibited good calibration.

Conclusion

The PAncreatic surgery Composite Endpoint and modified PAncreatic surgery Composite Endpoint models are robust and clinically relevant tools for predicting prolonged length of stay and 90-day mortality after pancreatectomy. Adding delayed gastric emptying to the modified PAncreatic surgery Composite Endpoint model enhances its predictive performance for prolonged length of stay.