Observer variation and discriminatory value of biopsy features in inflammatory bowel disease

If skilled histopathologists disagree over the same biopsy specimen, at least one must have an incorrect interpretation. Thus, disagreement is associated with, although not the cause of, diagnostic error. The present study aimed to determine the magnitude of variation among 10 observers with a special interest in gastrointestinal histopathology. They independently interpreted the same biopsy specimens for morphological features which may discriminate between patients with Crohn's disease and ulcerative colitis and normal subjects. Thirty of 41 features had agreement measures significantly better than expected by chance (p < 0.05). The range of agreement in the 45 observer pairs over the final diagnosis was 65-76%. There was good agreement in discriminating between normal slides and those showing confirmed inflammatory bowel disease. For normal slides, however, the term nonspecific inflammation was often applied and without any consistency. In addition, true Crohn's disease slides were often and consistently thought to be ulcerative colitis. Having identified 11 important discriminatory morphological features, two multiple regression analyses were then carried out to produce a scoring system for inflammatory bowel disease. These results suggest there is considerable room for improvement in the reliability of colonic biopsy specimen interpretation and that this could probably be achieved using more exact definitions of morphological features and diseases.