Five years after the last survey of drug-resistant tuberculosis in Japan, a nationwide survey was conducted by the Tuberculosis Research Committee (Ryoken). A total of 78 hospitals in various districts of Japan participated in this cooperative study. Each collaborating laboratory sent all the isolated mycobacterial cultures during June 1 to November 30, 1997 to the Research Institute of Tuberculosis (RIT), which is one of the Supranational Reference Laboratories of the WHO/IUATLD Global Project on Anti tuberculosis Drug Resistance Surveillance. At RIT identification and drug susceptibility of Mycobacterium tuberculosis isolates were reexamined. The RIT received a total of 2, 167 cultures. Among them, 523 cultures were excluded from further examinations because of various reasons, such as growth of mycobacteria other than tubercle bacilli (MOTT, 453), mixed cultures of M. tuberculosis and MOTT (16), and contamination or non-viability (54). Thus drug susceptibility test results were available for 1, 644 cultures, including 47 from foreign-born people. In the local laboratories, the absolute concentration method using 1% Ogawa egg slant (standard method, 26 hospitals), its modified methods using a 48-well plate (Microtiter method, 29 hospitals) and a 16-well plate (Well-pack method, 7 hospitals), combination of above 2 or 3 methods (13 hospitals), and other method (3 hospitals) were used for drug susceptibility testing, and the proportion method using 1% Ogawa egg slant was used in the RIT. The results in the local labora tories were compared with those in the RIT. A high concordance rate (over 90 %) was seen in the testing for 1 ug/ml of isoniazid (INH), rifampin (RFP) and streptomycin (SM), but the rate was lower (under 90%) in the testing for 0.1 ti g/ml of INH andethambutol (EMB). However, there was no significant difference in the concordance rates according to the test drugs among methods for drug susceptibility testing used in the local laboratories. Median concordance rates between the results with the standard method, Microtiter method and Well-pack method in the local laboratories, and those in the RIT were 95.9%, 93.2% and 96.4% respectively. Relatively lower concordance rates were seen in the laboratories using the Microtiter method related to high overestimation rates (median overestimation rate of 5.3%), compared with 1.2% and 2.3% in the laboratories using the standard method and Well-pack method, respectively. However, relatively lower concordance rates (less than 90%) were seen in the laboratories using any of the three methods, indicating that there are variations among facilities. Part of the results concerning the resistance patterns to four first-line anti-tuberculosis drugs were reported elsewhere.
