Ultrasonography has been reported in some studies to have a sensitivity of 94 percent in the diagnosis of pancreatic cancer. However, other studies suggest a sensitivity of no better than 70 percent. Postulated reasons for this discrepancy include differences in the sonographer's experience and methodologic limitations. Karlson and associates evaluated the accuracy of ultrasound examination in the diagnosis of pancreatic tumors in a consecutive series of 919 patients seen during a two-year period.

An ultrasonographic diagnosis of small pancreatic tumors was based on the identification of a poorly reflecting and attenuating pancreatic mass. Larger tumors were more heterogeneous echogenically and showed well-defined irregular or lobulated margins. Information on tumor diagnosis and deaths was obtained from cancer and death registries. Histologic verification of the diagnosis of pancreatic cancer was available in 84 percent of the patients.

In 140 (15.2 percent) of the patients, cancer was diagnosed within one year of the ultrasound examination. The tumors included primary exocrine and endocrine tumors (102 patients), cancer of the common bile duct, duodenal papilla and duodenum (17 patients) and pancreatic metastasis (21 patients). The head of the pancreas was most commonly involved, followed by the body and tail. The recorded mean size of the tumors was 4 cm.

Ultrasound examination demonstrated tumors in 124 (88.6 percent) of the 140 patients with pancreatic cancer. The sensitivity of ultrasonography in the detection of exocrine pancreatic cancer was 90 percent (79 of 88 patients). The diagnosis of cancer was missed by ultrasound examination in 16 patients, including nine with exocrine pancreatic cancer, five with cancer of the common bile duct and two with cancer of the duodenal papilla.

In the entire group of patients without tumors, nine examinations were falsely positive, giving a specificity of 98.8 percent (770 to 779 patients). False-positive results were obtained in patients with pancreatitis and pancreatic cysts.

The positive predictive value and negative predictive value of ultrasonography in the detection of all pancreatic tumors were 93.2 percent (124 of 133 patients) and 98 percent (770 of 786 patients), respectively. In nine of the patients in whom pancreatic cancer was initially missed by ultrasound examination, the diagnosis was made within two months by means of computed tomography in two patients, by repeat ultrasound examination in one patient, by laparotomy in four patients, by autopsy in one patient and by an unknown procedure in one patient.

The authors also reviewed the records of 94 ultrasound examinations. The ultrasound examinations were performed by three experienced sonographers in 73 percent of the cases; the other examinations were often supervised by the experienced sonographers. Neither the specific request to examine the patient because of suspected tumor nor a history of common symptoms of pancreatic cancer influenced the choice of experienced versus inexperienced sonographer.

The authors conclude that the results support the general recommendation to use ultrasonography as the primary imaging modality when pancreatic cancer is suspected. They note that the accuracy of the diagnosis seems to be influenced by the sonographer's experience.