Background: Inflammatory bowel disease (IBD)--and Crohn's disease (CD) and ulcerative colitis (UC) in particular--could be more reliably diagnosed by using biopsy criteria incorporating the colorectal distribution of specific histologic features. The aim of this study was to elucidate criteria distinguishing IBD from other forms of colitis (non-IBD), and CD from UC on the basis of multiple colorectal biopsies.
Methods: We examined multiple biopsy specimens (mean, 6.1) from 299 consecutive Japanese subjects with active colitis and performed multiple logistic regression analyses on 70 histologic features, from which 2 equations were constructed for the probabilities (P(IBD) and P(CD)) of a) IBD (versus non-IBD), and b) CD (versus UC), respectively, being present. On the basis of a receiver-operating characteristic curve, we determined four cut-off values for P(IBD) and constructed the criteria, consisting of the five categories 'definite IBD', 'probable IBD', 'unknown', 'probable non-IBD', and 'definite non-IBD'. The criteria for CD versus UC were constructed in a similar manner. Their validities were evaluated using 132 Canadian subjects.
Results: The statistically significant histologic features were as follows: for IBD, crypt architectural abnormalities, basal plasmacytosis with severe chronic inflammation, and distal Paneth cell metaplasia; for CD, segmental crypt architectural abnormalities and mucin depletion, mucin preservation at the active sites, and focal chronic inflammation without crypt atrophy. In the categories of probable IBD and probable non-IBD, both sensitivities and specificities exceeded 97%. Probable CD and probable UC showed high specificities of more than 97%, and their sensitivities were 94% and 89%, respectively. Kappa statistics showed these criteria to be sufficiently reproducible.
Conclusions: Specific histological features together with their distribution can reliably diagnose IBD, distinguish CD from UC, and provide an estimate of the probability of the underlying disease being present.