OBJECTIVE: Patients in a clinical trial who withdraw before the scheduled end date are a serious problem in economic evaluations. The method to deal with data from these patients can have important impact on outcomes in terms of resource use and costs, especially when withdrawal rates differ between treatment groups. The aim of this study was to compare the impact of various methods for dealing with censored data on the total costs and on the difference in costs between treatment groups. METHODS: Five methods for dealing with censored data were applied to data from 519 patients with chronic disease participating in a one-year randomized clinical trial. These five methods are complete case analysis, linear extrapolation, hot-decking, predicted regression, and multiple imputation. RESULTS: Fifteen percent of the patients in treatment group A and 21% of the patients in treatment group B withdrew from the study before the scheduled end date. Mean costs per patient varied from €889 (SE: 94) in the complete-case analysis to €1400 (SE: 189) after predicted regression. Cost differences between treatment groups varied from €14 in the complete-case analysis, to €243 after multiple imputation, to €372 after predicted regression. Hot-decking, multiple imputation and predicted regression were sensitive to the selection of covariates. CONCLUSION: The various methods had a considerable impact on total costs and on the difference in costs between treatment groups. In economic evaluations more attention should be paid to methods for dealing with censored patients and the impact of these different methods on the CE-ratio.