The endoscopic diagnosis of duodenal ulcer disease. A randomized clinical trial of bias and of interobserver variation.

Abstract

In a randomized design we examined whether endoscopists are biased by knowledge of the radiologic diagnosis of duodenal ulcer and deformity of the duodenal bulb when recording the corresponding endoscopic diagnoses. A total of 156 patients had a barium meal and were subsequently randomized into 2 groups. In 74 of the cases the 2 endoscopists knew the result of the X-ray examination when doing the endoscopy; in 82 of the cases they did not. One endoscopist was significantly biased by his knowledge of the radiologic diagnosis of deformity of the duodenal bulb. Neither of the endoscopists was biased by his knowledge of the radiologic diagnosis of duodenal ulcer. In addition, the interobserver variation between the two endoscopists with regard to the endoscopic diagnoses of duodenal ulcer, deformity of the duodenal bulb, and duodenitis was examined. The interobserver variation was expressed by the overall agreement and by the kappa statistics, which adjusts the overall agreement for expected chance agreement. For duodenal ulcer, deformity of the duodenal bulb, and duodenitis, the overall agreements and kappa values were 0.91, 0.78, and 0.75, and 0.54, 0.42, and 0.33, respectively.