OBJECTIVES: To assess the feasibility and effectiveness of voluntary counselling, HIV testing and adjunctive cotrimoxazole in reducing mortality in a cohort of tuberculosis (TB) patients registered under routine programme conditions in a rural district of Malawi. DESIGN: 'Before' and 'after' cohort study using historical controls. METHODS: Between 1 July 1999 and 30 June 2000 all TB patients were started on standardized anti-TB treatment, and offered voluntary counselling and HIV testing (VCT). Those found to be HIV-positive were offered cotrimoxazole at a dose of 480 mg twice daily, provided there were no contraindications. Side-effects were monitored clinically. End-of-treatment outcomes in this cohort (intervention group) were compared with a cohort registered between 1 July 1998 and 30 June 1999 in whom VCT and cotrimoxazole was not offered (control group). FINDINGS: A total of 1986 patients was registered in the study: 1061 in the intervention group and 925 in the control cohort. In the intervention group, 1019 (96%) patients were counselled pre-test, 964 (91%) underwent HIV testing and 938 (88%) were counselled post-test. The overall HIV-seroprevalence rate was 77%. A total of 693 patients were given cotrimoxazole of whom 14 (2%) manifested minor dermatological reactions. The adjusted relative risk of death in the intervention group compared with the control group was 0.81 (P < 0.001). The number needed to treat with VCT and adjunctive cotrimoxazole to prevent one death during anti-TB treatment was 12.5. INTERPRETATION: This study shows that VCT and adjunctive cotrimoxazole is feasible, safe and reduces mortality rates in TB patients under routine programme conditions.