Currently, metallo-beta-lactamase-producing gram-negative bacilli resistant to imipenem are increasing world-wide. The disk diffusion method recommended by NCCLS may not be appropriate for the detection of low-level ceftazidime-resistance. Therefore, a screening method with higher sensitivity should be developed for the detection of metallo-beta-lactamase producing strains. The aim of this study was to standardize a method for the detection of metallo-beta-lactamases and determine the prevalence of metallo-beta-lactamase producing ceftazidime-resistant gram-negative bacilli. We used two screening methods; DDST (Arakawa et al.) and modified the combination-disk test by using a 2-MPA inhibitor and CAZ disk. Three-parts testing of both the DDST and combination-disk test were performed. In the first two parts, the appropriate condition of both methods was determined in 20 and 103 IMP and CAZresistant gram-negative bacilli. In the last part, the two methods were used to determine metallo-beta-lactamases in 550 CAZ-resistant gram-negative bacilli. The DDST, using 1 µl of 2-MPA disk with 1 cm apart from edge to edge between the two disks, showed the highest positive percentage and was more clear cut than other distances. For the combination-disk test, the appropriate difference inhibition zone diameter surrounding the CAZ disk containing 1 µl of 2-MPA and CAZ alone for metallobeta- lactamase positive, was > 8 mm. The accuracy of DDST and the combination-disk test for screening of the total 550 CAZ-resistant isolates was 97.5% and the statistical correlation was regarded as being significant. Although DDST was simple, inexpensive and able to be used for the detection of metallo-beta-lactamase in the routine laboratory, the method was highly subjective in the interpretation of results, with a great variation in the shape of the enhanced zone. Moreover, the inappropriate distance between the two disks may give a false-negative result. The combination-disk test has clear-cut criteria for positive and negative results and the distance between the two disks does not effect the results. This method cannot detect VIM-1 and VIM-2. False positives also occurred in the isolates that were susceptible to 2-MPA. However, the false positives in the combinationdisk test were not higher than the DDST. Therefore, the combination-disk test is recommended for screening metallo-beta-lactamases in the routine laboratory. The prevalence of metallo-beta-lactamases was 8.7% in this study.