With the advent of preoperative radiotherapy for pancreas cancers, their measurement by imaging is rendered more important, so that outcome data from various treatment programs may be compared. Two radiologists have examined tumor size measurements from 29 patients by CT scans obtained within a week of measurement in the Pathology Department after resection. The radiologists assessed these scans independently from one another and blinded from the pathologic measurement. The largest diameter of their readings was compared to the largest diameter of the tumor measured by the Pathology technician. The correlation between radiologists (P < 3 x 10(-8) was excellent. Correlation between the average of the two radiologic estimates of greatest tumor diameter and actual tissue measurement is excellent for tumors (n = 21) greater than 2 and less than 5 cm in diameter (P < 0.03), but of four specimens measured in the Pathology laboratory as less than 2 cm, all were measured by the radiologists as being at least 1.5 cm larger. Twenty of the 29 cancers (69%) were measured by two radiologists to be within 1 cm of the actual diameter. The nine specimens producing the greatest errors were two with diffuse mucinous tumors throughout the gland (radiologists undercalled by 2.5 cm), one 7 cm mass with a cystic center composed of necrotic cells (overcalled by 2.5 cm, perhaps because of deformity or release of fluid in Pathology), four small masses (1, 1.5, 1.5, and 1.5 cm in diameter), smaller than the normal diameter of the pancreas head, and two tumors with inexplicably inaccurate size estimates.(ABSTRACT TRUNCATED AT 250 WORDS)