A comparison of four different commercial immunometric thyrotropin (TSH) assays (Amerlite R TSH-30 Ultrasensitive assay from Kodak, BeriLux R hTSH from Behring Werke, Delfia R hTSH Ultra from Wallac and IMX R Ultrasensitive hTSH from Abbott) was made by measuring serum TSH in 81 consecutive patients referred to hospital for various reasons with a serum TSH less than 0.8 mIU/I in the IMX assay. The analytical and functional assay sensitivities of each of the assays were analysed. Even though three of the methods had a sensitivity corresponding to third generation assays, we could only demonstrate an overall agreement of serum TSH when comparing two of the kits. The measurements in Delfia Ultra and Berilux showed good agreement ( P = 0.7, paired t-test and bias = 0.003 mIU/1), while the comparisons between the other assays showed different measurements ( P < 0.00001, paired t-test and bias more than 0.07 mIU/1). Differences in the calibrators used in the assays might explain some of the discrepancy, although all methods were calibrated according to the same international standard. Also, differences in the specificity of the TSH monoclonal antibodies used in the assays might be an evident explanation and further studies of the specificity of the monoclonal antibodies are needed. An international collaborative study to clarify reasons for the differences between the TSH assays and to standardize the measurements is recommended.