An interlaboratory study on the reproducibility of the CEA (Roche) RIA Test was carried out. Four different plasma pools of approximately 2, 3, 6, and 12 micrograms/l CEA were tested over a period of 4 weeks with 4 different lots of reagents in order to determine the interassay variances. At the same time we compared the lately introduced column technique with the dialysis and ultrafiltration method. Best results were obtained with the column technique which also showed best reproducibility. Only 1.4% of samples showed deviations greater than 5% between the mean of CEA duplicates and single CEA values, and these were omitted from the evaluation. On the other hand about 15% of the corresponding dialysis results showed deviations greater than 5% and were excluded from the evaluation. The methods compared showed a good correlation with a coefficient of 0.96, but the average values for the CEA determination, using the columns technique were lower than those obtained from dialysis. Interassay variances were greater for the dialysis procedures, i.e. 1.88 +/- 0.81, 3.25 +/- 0.83, 5.81 +/- 1.09, and 11-91 +/- 1.23 compared with 1.77 +/- 0.54, 2.63 +/- 0.68, 4.89 +/- 0.79, and 11.16 +/- 1.23 for the column technique. There were no systematic changes of the CEA values over the period of 4 weeks, thus giving optimal conditions for a follow up of patients.