OBJECTIVES: Tuberculosis is frequent among poor and marginalized people whose limited tuberculosis-related knowledge may impair healthcare access. We characterised tuberculosis-related knowledge and associations with delayed treatment and treatment outcome. METHODS: Tuberculosis patients (n = 943), people being tested for suspected tuberculosis (n = 2020), and randomly selected healthy controls (n = 476) in 16 periurban shantytowns were interviewed characterizing: socio-demographic factors; tuberculosis risk-factors; and patients' treatment delay. Principle component analysis was used to generate a tuberculosis-related knowledge score. Patients were followed-up for median 7.7 years. Factors associated with tuberculosis treatment delay, treatment outcome and tuberculosis recurrence were assessed using linear, logistic and Cox regression. RESULTS: Tuberculosis-related knowledge was poor, especially in older people who had not completed schooling and had never been diagnosed with tuberculosis. Tuberculosis treatment delay was median 60 days and was more delayed for patients who were poorer, older, had more severe tuberculosis and in only unadjusted analysis with incomplete schooling and low tuberculosis-related knowledge (all p </= 0.03). Lower than median tuberculosis-related knowledge was associated with tuberculosis recurrence (unadjusted hazard ratio = 2.1, p = 0.008), and this association was independent of co-morbidities, disease severity and demographic factors (multiple regression adjusted hazard ratio = 2.6, p = 0.008). CONCLUSIONS: Low tuberculosis-related knowledge independently predicted tuberculosis recurrence. Thus health education may improve tuberculosis prognosis.