Frozen section diagnosis of pancreatic lesions

The clinical and radiologic diagnosis of pancreatic cancer and the safety of pancreatic resections have improved. These improvements, together with the indication for resection in some cases of complicated chronic pancreatitis, have reduced the necessity for confirmed preoperative tissue diagnosis. We investigated the clinical use and accuracy of frozen section diagnosis for pancreatic lesions.

We searched archival files for the years 1989-2000 for patients with pancreatic lesions who had received a diagnosis based on frozen section results. We compared the diagnosis of all frozen section slides with that of the permanent sections and reviewed the clinical follow-up notes. We evaluated histologic features useful in differentiating between malignant and benign pancreatic lesions.

A total of 538 patients underwent surgical biopsy and/or resection for suspected pancreatic lesions. Frozen section was requested in 131 cases (284 frozen sections). Ninety cases had frozen section of the pancreatic lesions, 70 cases had frozen section of metastatic sites, and 29 cases had frozen section of surgical margins. Of the 90 cases in which frozen section of the pancreatic lesions was requested, malignancy was diagnosed in 44, a benign lesion was diagnosed in 37, and the diagnosis was atypical and deferred in 9. In total, 3 false-negative frozen sections and 1 false-positive frozen section were identified for respective rates of 1.2% and 0.3%. In all cases in which the frozen section diagnosis was deferred or was inconsistent with the operative impression, and the surgeon acted on his/her impression, the operative diagnoses were subsequently confirmed by additional permanent sections and/or clinical follow-up. The most useful histologic features for the diagnosis of pancreatic adenocarcinoma in frozen sections were variation in nuclear size of at least 4:1, disorganized duct distribution, incomplete duct lumen, and infiltrating single cells.

Frozen sections are useful in conjunction with the impression at surgery for the management of patients with pancreatic lesions. Frozen sections of resection margins were 100% accurate; frozen sections of pancreatic lesions or metastatic sites were accurate in 98.3% of cases. We found an acceptable rate of deferred frozen section (6.6%). The experienced surgeon's impression of malignancy is reliable in cases in which frozen section is deferred or has negative findings.