The aim of this study is to evaluate the consistency between routine methods for coding urinary bladder tumours in eight Italian cancer registries and the European Network of cancer registries (ENCR) criteria. Furthermore, it aims to evaluating the impact of the discordance on survival data. Eight cancer registries took part in the study: Ferrara, Florence, Macerata, Ragusa, Romagna, Sassari, Turin and Varese. The first 100 cases of neoplasm of the urinary bladder incident in the years 1993-1994 were identified from the files of each registry. The original pathology reports were made available. A working group considered eligible to the study 699 cases of microscopically confirmed transitional carcinoma (ICD-O morphology code 812-813). Using the ENCR criteria, each of these was classified according to morphology code (8120 vs. 8130) and behaviour (1/ uncertain, /2 non-invasive, 3/ invasive). Information of tumour behaviour was classified as follows: (i) present, when expressly stated in the original report, (ii) deducible, when not expressly stated but suggested by the pathologist's description, and (iii) absent, when impossible to determine on the basis of the original pathology report. The working group classification of tumour behaviour and the classification of the registry of origin were compared. There was a full concordance in the case of complete agreement on the morphology code, and partial concordance when only the invasive or non-invasive behaviour code was agreed upon. As much as 92.5% cases were microscopically confirmed. Tumour behaviour was expressly stated in the original report of 69.2% cases, not stated but suggested by the pathologist's description in 21.2% cases, and impossible to determine in 9.6%. Agreement between the panel and the registry of origin was complete in 71.2% cases and partial in 12.3% while there was a complete discordance in 16.5% cases. The panel interpreted as non-invasive 111 cases coded as invasive by the registry of origin. Conversely, it was estimated that 24% cases included in incidence data were non-invasive. This article discusses the impact of misclassification on survival data.