We have examined prospectively whether it was possible for a multicenter group (12 centers) to standardize data collection and to achieve a reasonable level of reproducibility of colonscopic findings in Crohn's disease. Each colonoscopy was performed simultaneously by a pair of endoscopists who did not communicate with each other until they had completed a standardized form. A preliminary period was necessary to achieve an agreement on items to be recorded and on lesion definition. Thereafter 112 colonoscopies were carried out by 18 intracenter endoscopist pairs (both endoscopists from the same center) and 21 colonoscopies by 14 intercenter endoscopist pairs (endoscopists from different centers) in patients with Crohn's disease. The following data were recorded: (1) for each segment involved--rectum, sigmoid, and left colon, transverse colon, right colon, and ileum--the nature of elementary lesions (by ticking on a list of nine items) and an estimation of the surface of lesions and of ulcerations on two linear analog scales; and (2) a global assessment of lesion severity on a five-point scale and on a linear analog scale. Reproducibility of lesion detection was fair to excellent (kappa values greater than 0.54, P less than 0.05 or less) except, in the intercenter study, for erythema and swollen mucosa. Agreement on estimation of diseased and ulcerated segmental surfaces was good (intraclass correlation, coefficient, r, greater than 0.85; P less than 0.001) in both intra- and intercenter studies. Global assessment of lesion severity on a five-point scale (intra- and intercenter kappa = 0.64 and 0.55, respectively; P less than 0.001) and on a linear analog scale [intra- and intercenter r = 0.86 (P less than 0.001) and 0.58 (P less than 0.01), respectively was reproducible. In conclusion, reproducible endoscopic data can be collected by a cooperative multicentric group and thus be used in controlled therapeutic trial in Crohn's disease.