Background: Intraoperative frozen section (IFS) is routinely utilized by many surgeons during pancreaticoduodenectomy. However, its utility has not been rigorously studied.
Methods: Patients who underwent pancreaticoduodenectomy between 2006 and 2015 were identified from institutional data. Measures of diagnostic accuracy of frozen section and multivariate logistic regression are reported.
Results: The cohort included 1076 patients. Of resected specimens, 73.3% were malignant. IFS and final pathologic review (the gold standard) were discrepant for (1) pathologic diagnosis or (2) resection margin status in 5.3% and 3.3% of cases. The sensitivity, specificity, and accuracy of IFS for histologic determination of malignancy were 97.2%, 95.3%, and 96.7% respectively. For resection margins, they were 92.3%, 99.3%, and 96.8%, respectively. Positive bile duct and neck margins were revised intraoperatively 62% and 65% of the time, respectively; positive uncinate margins were never resected but led surgeons to avoid revision of a second positive margin in 13% of cases (4.2% of all PDA). Operative changes were rarely noted in the presence of benign disease (n = 11, 1.0%); conversion to total pancreatectomy based on positive margins was performed in just 13 cases (1.2%). Upon multivariable analysis, a positive neck margin proved to be the greatest predictor for a revised resection margin (AOR 16.9 [4.8-59.8]), whereas a positive uncinate margin or a diagnosis of chronic pancreatitis was protective against IFS-driven operative changes (AOR 0.25 [0.09-0.73]; AOR 0.16 [0.13-0.19]).
Conclusions: IFS is highly accurate and guides reresection of margins. However, selective omission of IFS may be justified for cases where benign disease is suspected.
Keywords: Decision analysis; Frozen section; Intraoperative period; Pancreaticoduodenectomy.