A prospective study was undertaken to determine the reproducibility of clinical staging based on digital rectal examination (DRE) in prostate carcinoma. We evaluated 48 consecutive patients diagnosed with localized prostatic cancer. Four urologists performed DRE and sorted the patients according to the 1992 American Joint Committee on Cancer Classification for prostate cancer. Both the percentage observed total agreement among each couple of two different observers and the interobserver variability (Kappa index) were analyzed. The percentage observed total agreement among observers in distinguishing five clinical subcategories (T1c, T2a, T2b, T2c, and T3a) ranged between 38-60% (mean 49%) and the Kappa index showed interobserver agreement was poor (overall Kappa = 0.3 1). All four examiners agreed in assigning the same subcategory in only 21 % of cases, and 90% of them were T I. If only categories are distinguished (T I, T2, or T3), the percentage observed total agreement rises to 60-71% (mean 66%) and the interexaminer agreement improves to good (overall Kappa = 0.4 1). Accurate pathologic staging was obtained in every patient and the percentage observed agreement between every examiner and the pathologist was calculated, excluding cases interpreted as T I c. Regarding subcategories, clinicopathologic agreement ranges between 17-46%. If only categories T2 and9T3 are distinguished, agreement rises to 57-69%. In summary, the ability to reproduce clinical staging based on DRE among multiple examiners is disappointingly low and understandably correlates poorly with pathologic stage.