Antiretroviral therapy and isoniazid preventive therapy (IPT) are both effective interventions to prevent HIV-associated tuberculosis, but work via different mechanisms. We propose that these two interventions might best be used as complementary strategies at different stages of HIV progression. At relatively high CD4-cell counts, IPT reduces tuberculosis risk by 64% (95% CI 39-78%) in patients with positive tuberculin skin tests, and is the key tuberculosis preventive intervention before patients are eligible for antiretroviral therapy. However, at low CD4-cell counts, reliable exclusion of active tuberculosis is difficult, fewer patients are eligible for IPT, and waning immune function might limit the durability of its effect. In such patients, antiretroviral therapy is the primary intervention needed, reducing tuberculosis incidence by 67% (95% CI 61-73%). However, tuberculosis risk during long-term antiretroviral therapy remains several times higher than background, especially in those with poor immune recovery. Patients might therefore derive additional benefit from combined use of IPT and antiretroviral therapy to simultaneously treat mycobacterial infection and restore tuberculosis-specific immune function. For those first presenting with advanced immunodeficiency, we propose that concurrent IPT might best be delayed until completion of the first few months of antiretroviral therapy, when active tuberculosis can be more reliably excluded. Data from randomised controlled trials are needed to underpin further development of public-health policy.