[Anti-tuberculosis drug resistance survey in Japan, 2002: external quality assessment of results].

OBJECTIVE A nationwide anti-tuberculosis drug resistance survey for Mycobacterium tuberculosis was conducted by the Tuberculosis Research Committee (Ryoken) in Japan, 2002, to clarify the recent trend of drug resistant M. tuberculosis. The drug susceptibility testing (DST) results by participating laboratories were compared with the test results by the reference laboratory, in order to evaluate the quality of DST by the participating laboratories. METHOD Mycobacterium strains were collected from patients who were admitted to the 99 participating hospitals between June and November in 2002. For each isolate, DST was performed at participating facilities and also at the reference laboratory (Research Institute of Tuberculosis: RIT) for four first-line anti-tuberculosis drugs, i.e., isoniazid (INH), rifampicin (RFP), streptomycin (SM) and ethambutol (EB). Each participating laboratory performed the DST with its routine method. The DST method for M. tuberculosis performed at the reference laboratory was the simplified proportion method on the standard 1% Ogawa medium as described in the national guidelines, and the results were regarded as the judicial ones. The DST results of each isolate from the participating laboratories were compared with the judicial results from the reference laboratory. The accuracy of DST done by the participating laboratories was evaluated in terms of the following indices; sensitivity for detecting drug resistant strains, the specificity for susceptible strain, the overall agreement, and kappa coefficient were calculated to evaluate the performance of local laboratories. RESULTS A total of 3,122 M. tuberculosis strains were obtained out of 4,134 mycobacterial strains collected from the participating facilities. Fifty, 23 and 17 local laboratories used Bitspectre-SR (Kyokuto pharmaceuticals), Welpack S (Nichibi) and Ogawa media for DST, respectively. MGIT (Becton Dickinson) and BrothMIC MTB-I (Kyokuto pharmaceuticals) were used in four laboratories each. The sensitivity, specificity, efficiency and kappa coefficient for INH were 84.5%, 98.7%, 98.0%, and 0.798, respectively. Similarly for RFP, they were 90.3%, 99.7%, 99.5%, and 0.894, respectively. False susceptible results were frequently observed (2.1%) for SM compared with false resistance (0.5%), whereas the efficiency of SM was 97.4%. Similarly for EB, false resistances were frequently observed (2.6%) compared with false susceptibles (0.4%), whereas the efficiency of EB was 96.9%. The kappa coefficient for EB (0.470) was obviously lower than the others. The DST results with Ogawa method at local laboratories showed significantly lower sensitivity than those with Welpack S and MGIT. DISCUSSION The DST methods used at local laboratories were still mainly microtitre methods with egg-based solid media, but the number of laboratories using liquid DST methods increased in 2002 compared to 1997. The overall specificity and efficiency of DST for each anti-tuberculosis drug was over 95%, but the sensitivity was below 90.3%. Because of the frequency of drug resistance (up to 7.9% for SM in 2002), the efficiency and specificity may not be useful indicators for the quality assessment. The kappa coefficient for the agreement between local and reference laboratories DST was clearly low in the case of EB, except for the laboratories using MGIT where kappa coefficient was higher than 0.8. The quality improvement of DST for EB could be achieved through the standardisation and automatisation.