A large evaluation study, analytical as well as clinical, was performed on four published improvements for the determination of creatinine in serum. Two of the methods were based on the Jaffé reaction, the other two were enzymatic methods. Analytically, all four methods showed a similar performance. In the analysis of numerous specimens from advanced care departments, however, the methods performed differently. The two enzymatic procedures scored better than the two Jaffé methods, when compared with an HPLC-based reference method. However, even the best methods produced outliers and a larger scatter than that obtained with "normal" specimens.