Interobserver Variability and Challenges in Intraoperative Frozen Section Evaluation of Pancreatic Margins in Pancreatectomy Specimens.

OBJECTIVE

Frozen-section evaluation of the pancreatic margin is challenging. We aimed to determine interobserver variability among gastrointestinal pathologists for the assessment of frozen sections of pancreatic margins with marked chronic pancreatitis and to determine the challenging histological features in discrepant cases.

METHODS

We identified 45 patients who underwent pancreas resection for pancreatic ductal adenocarcinoma and showed marked chronic pancreatitis at pancreatic margin. Deidentified first levels of frozen-sections of the pancreatic margins from all cases were independently reviewed by 5 experienced gastrointestinal pathologists for the presence of carcinoma and/or high-grade dysplasia.

RESULTS

Interobserver agreement among pathologists was calculated as kappa coefficients ([Formula: see text]). A consensus diagnosis for discordant cases was obtained after group review and discussion. Interobserver agreement for adenocarcinoma diagnosis was 87%, and there was "substantial agreement" (Fleiss [Formula: see text]=0.78, P<0.01) and "almost perfect agreement" (Brennan-Prediger [Formula: see text]=0.86, P<0.01). Using the final diagnosis based on frozen and permanent sections as the gold standard and the concordant read of at least 3 of 5 pathologists for comparison, the diagnosis of adenocarcinoma was made in frozen-sections of pancreas margins, with accuracy 98%, sensitivity 83%, specificity 100%, negative predictive value 97%, positive predictive value 100%, false negative rate 9%, and false positive rate 0%.

CONCLUSIONS

We showed excellent interobserver agreement among gastrointestinal pathologists for diagnosis of adenocarcinoma on frozen sections of pancreatic margins with marked chronic pancreatitis. Missed adenocarcinoma at the margin was mainly caused by freezing or cautery artifacts or by overlooking a tiny focus of perineural invasion in a background of marked chronic pancreatitis. The evaluation of deeper levels led to perfect agreement.