OBJECTIVE: To determine the long-term effect of preventive therapy (PT) for tuberculosis on the rates of tuberculosis, mortality and HIV progression. METHODS: In a randomized controlled trial, 1053 HIV-positive Zambian adults received isoniazid (H) for 6 months, rifampicin plus pyrazinamide (RZ) for 3 months, or a placebo. CD4 percentage, neopterin, absolute lymphocyte count and haemoglobin were measured from enrolment (absolute CD4 cell counts from 12 months after enrolment). Because PT reduced the incidence of tuberculosis, eligible placebo subjects were offered H. Here, tuberculosis and mortality rates are compared in the three original arms (intention to treat) using data beyond the end of the trial (average follow-up 3 years; maximum 7 years). RESULTS: There were 102 cases of tuberculosis and 281 deaths (rates 3.6 and 9.0/100 person-years, respectively). There was no significant difference between the tuberculosis rates in the H and RZ groups at any time. The effect of H/RZ on tuberculosis diminished over time (P = 0.011) but the cumulative risk of tuberculosis in the first 2.5 years was significantly lower in the H/RZ group than the placebo group (rate ratio 0.55; 95% confidence interval 0.32-0.93; P = 0.028). There was no significant effect of PT on mortality or progression markers. Tuberculosis was associated with an increased mortality (adjusted rate ratio 1.96; 95% confidence interval 1.21-3.18; P = 0.006). CONCLUSIONS: Both PT regimens protect against tuberculosis for at least 2.5 years but appear to have no effect on HIV progression or mortality. These results may be used in cost-effectiveness models of PT.