Detection of liver metastases from pancreatic cancer using FDG PET

We evaluated the potential of the glucose analog [18F]fluorodeoxyglucose (FDG) as a PET tracer for the hepatic staging in 168 patients designated for resective pancreatic surgery.

Metastatic liver disease was confirmed or excluded during surgery or with CT follow-up for at least 6 mo. Proven metastases were then retrospectively identified on preoperative CT (gold standard). Hepatic PET scans of all patients were interpreted blindly. Any focal FDG uptake was considered malignant. Both proven hepatic metastases and suspicious hepatic PET lesions were then compared, lesion by lesion, with CT. Standardized uptake values (SUV) and tumor-to-liver ratios (T/L) were determined for the most intense lesion of each patient.

Sensitivity of FDG PET was 68% (15 of 22 patients). The lesion detection rate was 97% (28 of 29 metastases) for lesions >1 cm and 43% (16 of 37 metastases) for lesions < or = 1 cm. Specificity was 95% (138 of 146 patients). Six of eight patients with false-positive results had marked intrahepatic cholestasis (versus 3 of 15 patients with true-positive lesions), one had an infrahepatic abscess and one had a right basal pulmonary metastasis. The SUV and T/L were 4.6+/-1.4 and 2.3+/-1.1, respectively, for malignant lesions and 4.1+/-1.5 and 1.9+/-0.3, respectively, for false-positive lesions and therefore are of limited value.

FDG PET provides reliable hepatic staging for lesions >1 cm. False-positive results are associated with the presence of marked intrahepatic cholestasis. For lesions < or = 1 cm, FDG PET can define malignancy in 43% of suspicious CT lesions in the absence of dilated bile ducts.