Background Tuberculosis remains a major public-health problem in Bangladesh, despite national efforts to improve case identification and treatment compliance. In 1984, BRAC (formerly the Bangladesh Rural Advancement Committee), a national, non-governmental organisation, began an experimental tuberculosis-control programme in one thana (subdistrict). Community health workers screened villagers for chronic cough and collected sputum samples for acid-fast bacillus (AFB) microscopy (phase one). Positive patients received 12 months of directly observed therapy. Phase two (1992-94) included another nine thanas and, in phase three (1995), eight more thanas were included. From 1995, the treatment was an 8-month oral regimen. Methods In 1995-96, we analysed all programme data from 1992 to 1995. First we analysed phases two (12-month therapy) and three (8-month therapy) separately for proportion cured, died, treatment failed, defaulted, migrated, and referred. Second, we did a cross-sectional survey of tuberculosis cases in more than 9000 randomly selected households in two phase-two thanas and one non-programme thana, and analysed the follow-up of all patients treated in the programme thanas. Findings In the phase-two analysis, 3497 (90%) of 3886 cases identified had accepted 12-month treatment. In phase three, all of 1741 identified cases accepted the 8-month regimen. 2833 (81.0%) and 1496 (85.9%) in phases two and three, respectively, were cured; 336 (9.6%) and 133 (7.6%) died. The relapse rate 2 or more years after treatment was discontinued was higher than the early relapse rate. The dropout rate was 3.1%. In the cross-sectional survey, the prevalence of tuberculosis in the two programme thanes was half of that in the comparison thana, where only government services were available (0.07 vs 0.15 per 1000). Interpretation The BRAC tuberculosis-control programme has successfully achieved high rates of case detection and treatment compliance, with a cure rate of at least 85% and a dropout rate of 3.1%. The prevalence survey suggested that at least half of all existing cases had been detected by the programme.
