OBJECTIVES: This study aimed to assess whether routine transabdominal ultrasonography (US) is clinically helpful for the early detection of postoperative pancreatic fistula (PF). METHODS: In a prospective cohort of patients undergoing partial pancreatectomy, US was performed on postoperative day (POD) 3. Potential predictors of PF, including amylase value in drains (AVD) on POD 1, were investigated. A tree-based classification model of the independent predictors of PF was also performed. RESULTS: One hundred seventy-three patients were analyzed. A peripancreatic collection on US and an AVD 5000 U/L or greater on POD 1 were predictors of PF. In the tree-based classification model, patients were stratified by AVD on POD 1. For values less than 5000 U/L (incidence of PF, 11.3%), US had a sensitivity of 23.1% and a specificity of 97.5%. For AVD 5000 U/L or greater (incidence of PF, 70.7%), sensitivity was 46.3% and specificity was 100%. CONCLUSIONS: Despite the presence of a peripancreatic collection as a predictor of PF, US-as a diagnostic test-resulted to be highly specific but poorly sensitive even in the tree-based classification model. Therefore, its role does not seem to be clinically relevant and does not add value to AVD on POD 1, which remains the most powerful and relevant early predictor of PF.
OBJECTIVES: This study aimed to assess whether routine transabdominal ultrasonography (US) is clinically helpful for the early detection of postoperative pancreatic fistula (PF). METHODS: In a prospective cohort of patients undergoing partial pancreatectomy, US was performed on postoperative day (POD) 3. Potential predictors of PF, including amylase value in drains (AVD) on POD 1, were investigated. A tree-based classification model of the independent predictors of PF was also performed. RESULTS: One hundred seventy-three patients were analyzed. A peripancreatic collection on US and an AVD 5000 U/L or greater on POD 1 were predictors of PF. In the tree-based classification model, patients were stratified by AVD on POD 1. For values less than 5000 U/L (incidence of PF, 11.3%), US had a sensitivity of 23.1% and a specificity of 97.5%. For AVD 5000 U/L or greater (incidence of PF, 70.7%), sensitivity was 46.3% and specificity was 100%. CONCLUSIONS: Despite the presence of a peripancreatic collection as a predictor of PF, US-as a diagnostic test-resulted to be highly specific but poorly sensitive even in the tree-based classification model. Therefore, its role does not seem to be clinically relevant and does not add value to AVD on POD 1, which remains the most powerful and relevant early predictor of PF.