[External quality assessment of drug susceptibility testing for Mycobacterium tuberculosis]

The Committee for Mycobacterial Examinations has planned and implemented the third external quality assessment of drug susceptibility testing for Mycobacterium tuberculosis to the hospital and private laboratories.

The Committee delivered 20 M. tuberculosis strains, exactly pairs of 10 strains, which were evaluated and standardized for the drug resistance pattern in the WHO/ IUATLD supra-national laboratory network (SRLN). The agreement of the majority of SRLN laboratories was considered as the gold standard of susceptibility result for each strain tested. Each laboratory performed the drug susceptibility testing (DST) with its own routine method. The sensitivity to detect drug resistance, the specificity for susceptible strain, the efficiency of overall agreement, the reproducibility for each pair, and kappa coefficient were calculated to evaluate their performance. DST was performed for isoniazid (INH), rifampicin (RFP), streptomycin (SM) and ethambutol (EB).

A total of 48 results has been collected. The overall sensitivity, specificity, efficiency, reproducibility and kappa coefficient for each anti-tuberculosis drug tested were as follows respectively; 100%, 99.0% (62.5-100), 99.6% (85.0-100), 99.6% (90-100) and 0.991 for INH; 97.7% (83.3-100), 100%, 98.6% (90-100), 99.0% (90-100) and 0.972 for RFP; 87.5% (66.7-100), 99.0% (87.5-100), 92.1% (80-100), 97.5% (70-100) and 0.84 for SM; 99.5% (75.0-100), 97.9% (75.0-100), 98.5% (85.0-100), 97.9% (70-100) and 0.97 for EB. Regarding private laboratories, all indicators showed 100% for INH but 2 hospital laboratories showed less than 90% in sensitivity. As for RFP, one private laboratory (4.3%) and 3 hospital laboratories showed less than 90% in sensitivity. The major difference between private and hospital laboratories was seen in EB. One hospital laboratory (4.0%) showed less than 90% in sensitivity and three (12.0%) showed less than 90% in specificity, compared to none in private laboratories. Additionally, six hospital laboratories (24.0%) showed less than 90% in reproducibility. As for quality improvement, two private laboratories that showed poor performance in 2003 have improved their quality up to 100% in 2004.

The overall efficiency by private and hospital laboratories satisfied WHO criteria. However, it diverged in each category of laboratories and hospital laboratories tended to show poor performance compared to the private ones. The reason for the difference was not clear, but the routine workload, allocated time and cost for the panel testing might contribute to it. The sensitivity of SM was relatively low compared to the other drugs as it was observed in 2003. It was mainly due to two strains to which the participating laboratories showed poor agreement to the gold standard. The difference of critical drug concentration for SM in Löwenstein-Jensen and in 1% Ogawa medium might contribute to the discrepancies. As for quality improvement, two private laboratories with poor performance in 2003 have shown marked improvement after on-site evaluation. The results indicated the usefulness of external quality assessment for the maintenance and improvement of the quality of the test.