Objective: Our purpose was to determine whether abdominal CT can detect pancreatic cancer before its clinical diagnosis.
Subjects and methods: Two radiologists interpreted in a blinded manner 62 CT scans in 28 pancreatic cancer patients that were obtained before histologic diagnosis and 89 CT scans in 89 control subjects and noted specific CT findings. The presence of pancreatic cancer was characterized as definite, suspicious, low probability, or normal. The scans of the pancreatic cancer patients were divided into four groups on the basis of the time interval preceding cancer diagnosis (0-2, 2-6, 6-18, or > 18 months), and one scan (closest to 18 months) was selected per patient per time interval. Sensitivity and specificity for pancreatic cancer and interobserver agreement for CT findings were calculated.
Results: Radiologists agreed that CT findings definite or suspicious for pancreatic cancer were present in 50% of the scans obtained 2-6 and 6-18 months before the diagnosis of pancreatic cancer (3/6 and 4/8 scans, respectively), but they noted such CT findings in only 7% (1/15) of the scans obtained more than 18 months before diagnosis. Pancreatic duct dilatation and cutoff were early CT findings identified by both radiologists and were associated with near-perfect and substantial interobserver agreement (kappa = 0.84 and 0.76, respectively). Ninety-five percent confidence intervals of specificity for tumor absence ranged from 92% to 100%.
Conclusion: CT can detect a significant proportion of asymptomatic incident pancreatic cancers before the clinical diagnosis of pancreatic cancer. CT should be considered in screening at-risk patient populations. Pancreatic duct dilatation and cutoff are early findings associated with the development of pancreatic cancer and can be detected on CT with a high degree of reproducibility.