Setting: Eleven countries/territories.
Objectives: Global information on the determinants of drug-resistant tuberculosis (TB) based on representative data is not available. We therefore studied the relationship between demographic characteristics, prior TB treatment, and human immunodeficiency virus (HIV) infection with anti-tuberculosis drug resistance.
Methods: Population-based representative data on new and previously treated patients with TB collected within an international drug resistance surveillance network.
Results: Of 9,615 patients, 8,222 (85.5%) were new cases of TB and 1,393 (14.5%) were previously treated cases. Compared with new cases, previously treated cases were significantly more likely to have resistance to one (OR = 2.5,95% CI 2.1-3.0; P < 0.001), two (OR = 4.6, 95%CI 3.7-5.6; P < 0.001), three (OR = 11.5, 95%CI 8.6-15.3; P < 0.001), and four (OR = 18.5, 95% CI 12.0-28.5; P < 0.001) drugs. An approximately linear increase in the likelihood of having multidrug-resistant tuberculosis (MDR-TB) was observed as the total time (measured in months) of prior anti-tuberculosis treatment increased (P < 0.001, chi2 for trend). In multivariate analysis, prior TB treatment for 6-11 months (OR = 7.6, 95% CI 2.6, 22.4; P < 0.001) and > or = 12 months (OR 13.7, 95% CI 4.5-41.6; P < 0.001), but not HIV positivity, was associated with MDR-TB.
Conclusion: This study shows that prior but ineffective treatment is a strong predictor of drug resistance, and that HIV is not an independent risk factor for MDR-TB. The association between length of treatment and drug resistance may reflect longer treatment as a result of treatment failure in patients with drug resistance; it may also reflect irregular prior treatment for TB, leading to drug resistance.