OBJECTIVE--To determine whether the number of patients required in a therapeutic trial that uses progression of radiological abnormalities as the outcome measure would be similar for multiple centres. METHODS--The progression of radiological damage to the fingers and wrists of patients with rheumatoid arthritis in five centres, three in North America and two in Europe, was examined. The reproducibility of repeated readings by the same and multiple observers was examined. The number of patients required in a two group trial was calculated for several combinations of power and significance. RESULTS--Scoring progression of radiological abnormalities in sequential films taken between 0.5 and 2.1 years was found to be highly reproducible. When the scores of a single reader were used the rate of change of radiological scores was similar in all centres. Based on the mean progression rate for all centres it was estimated that 153 patients in each group would be required to assure 90% power for detecting a 50% slowing of radiological progression at a significance of 0.05. Review of the experience in three trials showed a large variability in the radiological progression rates. CONCLUSION--The progression of scores for radiological damage in rheumatoid arthritis is relatively uniform in North America and Europe and thus the number of patients required in a trial would be similar. Experience in three trials showed that patient selection is of paramount importance in setting up a successful study.