Tuberculosis diagnosed in a rural setting in Angola. Accuracy of follow-up sputum smears to predict outcome

Objective: To analyze treatment outcome and the accuracy of positive sputum smear at 2 months to predict treatment failure in a cohort of patients with tuberculosis (TB) in a rural setting in Angola. Design: Observational study of patients with TB from January 2009 to August 2010 in Hospital Nossa Senhora da Paz in Angola. A multivariate analysis was performed to identify variables associated with treatment failure and death. Sensitivity, specificity, positive and negative predictive values and likelihood ratios to define the accuracy of a positive sputum smear at 2 months to predict treatment failure were calculated. Results: One thousand four hundred and twenty-five patients were diagnosed with TB. Overall, 526 patients were cured from TB and 419 had treatment completed, so 945 (66.3%) patients achieved treatment success. The outcomes of the remaining patients were: 91 (6.4%) had treatment failure, 100 (7%) died, 49 (3.4%) interrupted treatment, and 240 (16.8%) were transferred out. Variables associated with a higher risk of treatment failure were previously treated patients (odds ratio, 2.36; 95% confidence interval, 1.32-4.2) and positive sputum smear at 2 months (odds ratio, 9.81; 95% confidence interval, 5.88-16.36). Among the group of 551 patients with sputum smear confirmed at diagnosis and specimens taken at 2 and 5 months, the positive predictive value (31%) and the positive likelihood ratio (3.21) of a positive sputum smear taken at 2 months to predict treatment failure were low. Conclusions: Patients with positive sputum smear at 2 months have a higher risk of treatment failure. However, this by itself is a poor predictor of treatment failure.