An economic study was conducted alongside a clinical trial at three sites in Pakistan to establish the costs and effectiveness of different strategies for implementing directly observed treatment (DOT) for tuberculosis. Patients were randomly allocated to one of three arms: DOTS with direct observation by health workers (at health centres or by community health workers); DOTS with direct observation by family members; and DOTS without direct observation. The clinical trial found no statistically significant difference in cure rate for the different arms. The economic study collected data on the full range of health service costs and patient costs of the different treatment arms. Data were also disaggregated by gender, rural and urban patients, by treatment site and by economic categories, to investigate the costs of the different strategies, their cost-effectiveness and the impact that they might have on patient compliance with treatment. The study found that direct observation by health centre-based health workers was the least cost-effective of the strategies tested (US$310 per case cured). This is an interesting result, as this is the model recommended by the World Health Organization and International Union against Tuberculosis and Lung Disease. Attending health centres daily during the first 2 months generated high patient costs (direct and in terms of time lost), yet cure rates for this group fell below those of the non-observed group (58%, compared with 62%). One factor suggested by this study is that the high costs of attending may be deterring patients, and in particular, economically active patients who have most to lose from the time taken by direct observation. Without stronger evidence of benefits, it is hard to justify the costs to health services and patients that this type of direct observation imposes. The self-administered group came out as most cost-effective ($164 per case cured). The community health worker sub-group achieved the highest cure rates (67%), with a cost per case only slightly higher than the self-administered group ($172 per case cured). This approach should be investigated further, along with other approaches to improving patient compliance.