SETTING: In sub-Saharan Africa, tuberculosis (TB) has increased over the last two decades due to the human immunodeficiency virus pandemic. In Malawi, 20630 new TB patients were notified to the National Tuberculosis Programme in 1996, a fourfold increase since 1986. Due to this increase in cases and lack of resources (both human and monetary) it is becoming more difficult to ensure directly observed treatment (DOT) in the TB wards. METHODS: In Ntcheu district, Malawi, a new TB regimen was introduced from April 1996 in which patients received supervised treatment by either a health worker or a guardian (i.e., family member). Adherence to the different treatment options was measured by form checks, tablet counts, and tests for detecting isoniazid in the urine. Adherence was measured at 2, 4 and 8 weeks after onset of TB treatment. RESULTS: Overall adherence rate was 95-96%. Inpatients showed the highest adherence rates. Patients on guardian-based DOT (GB-DOT) (n = 35) showed 94% adherence, while patients on health centre based DOT (n = 40) showed more non-adherent behaviour: 11% according to monitoring forms, 14% according to tablet counts and 16% according to urine tests. DISCUSSION: The results suggest that decentralised care is a feasible option for anti-tuberculosis treatment and that guardians can supervise TB treatment just as well as health workers during the intensive phase of TB treatment.