Objective: To identify predictors of initial sputum culture conversion, estimate the usefulness of persistent positive cultures at different time points in predicting treatment failure, and evaluate different definitions of culture conversion for predicting failure among patients with multidrug-resistant tuberculosis (MDR-TB) in five countries, 2000-2004.
Methods: Predictors of time to conversion were identified using multivariate Cox proportional hazards regression modeling. Receiver operating characteristic curves were plotted to visualize the effect of using different definitions of 'culture conversion' on the balance between sensitivity and specificity.
Results: Overall, 1209/1416 (85%) of patients with baseline positive cultures converted in a median of 3.0 months (interquartile range 2.0-5.0). Independent predictors of less likely conversion included baseline positive smear (hazard ratio [HR] 0.60, 95%CI 0.53-0.68), resistance to pyrazinamide (HR 0.82, 95%CI 0.70-0.96), fluoroquinolones (FQs; HR 0.65, 95%CI 0.51-0.83) or thioamide (HR 0.83, 95%CI 0.71-0.96), previous use of FQs (HR 0.71, 95%CI 0.60-0.83), poor outcome of previous anti-tuberculosis treatment (HR 0.69, 95%CI 0.54-0.88) and alcoholism (HR 0.74, 95%CI 0.63-0.87). The maximum combined sensitivity (84%) and specificity (94%) in predicting treatment failure was based on lack of culture conversion at month 9 of treatment, assuming conversion is defined as five consecutive negative cultures.
Conclusion: Patients with identified risk factors were less likely to achieve sputum culture conversion during MDR-TB treatment.