Introduction: Limited data exist on use of the microscopic-observation drug-susceptibility (MODS) assay among persons suspected of MDR-TB living in high HIV-prevalence settings. Methods: We retrospectively reviewed available clinical and drug susceptibility data for drug-resistant TB suspects referred for culture and drug-susceptibility testing between April 1, 2011 and March 1, 2012. The diagnostic accuracy of MODS was estimated against a reference standard including Löwenstein-Jensen (LJ) media and manual liquid (BACTEC MGIT) culture. The accuracy of MODS drug-susceptibility testing (DST) was assessed against a reference standard absolute concentration method. Results: One hundred thirty-eight sputum samples were collected from 99 drug-resistant TB suspects; in addition, six previously cultured MDR isolates were included for assessment of DST accuracy. Among persons with known HIV infection status, 39/59 (66%) were HIV-infected. Eighty-six percent of patients had a history of prior TB treatment, and 80% of individuals were on antituberculous treatment at the time of sample collection. M. tuberculosis was identified by reference standard culture among 34/98 (35%) MDR-TB suspects. Overall MODS sensitivity for M. tuberculosis detection was 85% (95% CI, 69-95%) and specificity was 93% (95% CI, 84-98%); diagnostic accuracy did not significantly differ by HIV infection status. Median time to positivity was significantly shorter for MODS (7 days; IQR 7-15 days) than MGIT (12 days; IQR 6-16 days) or LJ (28 days; IQR 21-35 days; p<0.001). Of 33 specimens with concurrent DST results, sensitivity of the MODS assay for detection of resistance to isoniazid, rifampin, and MDR-TB was 88% (95% CI, 68-97%), 96% (95% CI, 79-100%), and 91% (95% CI, 72-99%), respectively; specificity was 89% (95% CI, 52-100%), 89% (95% CI, 52-100%), and 90% (95% CI, 56-100%), respectively. Conclusion: In a high HIV-prevalence setting, MODS diagnosed TB and drug-resistant TB with high sensitivity and shorter turnaround time compared with standard culture and DST methods. © 2013 Makamure et al.
