Several studies have been carried out on the community based DOT in a variety of settings. However, although some have been very large, most of them have been relatively small. The Hlabisa TB Control Programme in rural south Africa has used community-based DOT extensively since mid 1991. A detailed analysis of the data belonging from 1991 to 1996 is done to find out reporting trends in adherence, timing of treatment interruption and risk factors for failing to complete therapy. The study was carried out in a population of 2.1 lakh zulu speaking people who are mostly farmers, labourers and pensioners with middle income and 69% literacy rate. HIV seroprevalence among adult TB patients increased from 36% in 1991 to 66% in 1997 and consequent to that annual case detection increased from 321 to 1250 by 1996. Of the 3610 surviving patients, 629 (17%) failed to complete treatment ranging from 11% in 1991-92 to 22% in 1996. Association of treatment interruption with age, sex, type of TB and HIV status was observed as follows: Age specific frequency distribution for treatment interruption was higher among those aged 25-34 years and significantly greater than among the patients aged 0-14 years and those aged 55 years and over. A similar age specific frequency distribution for treatment interruption was observed each year. Treatment interruption was higher in men than women. The interruption rate was similar among patients with smear positive pulmonary TB, smear negative and extra pulmonary disease. Treatment interruption was more frequent among patients known to be HIV infected (25%) than among those whose HIV status was unknown (17%) and those known to be HIV infected (12%). The pattern was observed each year and was unaffected by age or sex. The interruption of treatment among HIV infected and not tested for HIV patients was high when supervised by health worker. The interruption of treatment increased between 1991/92 – 1996 and was greatest among patients supervised at clinics. The single independent risk factor for treatment interruption was diagnosis between 1994-1996 compared with 1991-93 (odds ratio [OR] 1.9, 95% confidence interval [CI] 1.6-2.4). The second factor was known HIV- positive status versus known HIV-negative status (OR 1.8, 95% CI 1.4-2.4); supervised by village clinic with community worker (OR 1.9) and male versus female (OR 1.3). In conclusion, adherence to therapy in a community with high caseload, migration remains a challenge even with the community based DOTS.