Antiretroviral therapy (ART) is the most effective preventive strategy against tuberculosis in people with HIV, substantially reducing the risk irrespective of baseline CD4 cell count, tuberculin skin test status, and antimycobacterial drug resistance. 1–3 This preventive effect occurs at individual, community, and national levels. 3, 4 However, ART is no silver bullet for the control of HIV-associated tuberculosis. 3 First, ART is often started late during HIV progression and many patients have tuberculosis before starting ART. 5 Second, data reveal that even during long-term ART, tuberculosis incidence remains much higher than among people not infected with HIV in South Africa. 6 However, few data are available from high-resource settings in which tuberculosis epidemiology differs. In The Lancet HIV, Gupta and colleagues7 report findings from a UK cohort of almost 80 000 HIV-positive people (2007–11) linked to the national tuberculosis register. Decreases in tuberculosis incidence during ART in all ethnic groups were consistent with other studies from around the world. 8 Long-term tub erculosis incidence in people taking ART were then compared with rates in 2009 in people without HIV. Rates in white patients receiving ART remained substantially higher than background, even among those with good immunological recovery. Rates in black Africans were much higher than rates in other ethnic groups, irrespective of HIV status and ART duration. 7 The degree of immune restoration achieved affects the pattern of opportunistic infections occurring during ART. Although the introduction of ART in the mid 1990s greatly reduced HIV-associated morbidity and mortality in many highincome countries, 9 tuberculosis remains an exception. Partial immune restoration is sufficient to prevent disease from low virulence opportunistic pathogens such as disseminated Mycobacterium avium. However, the more virulent pathogen Mycobacterium tuberculosis causes disease across the full spectrum of immunodeficiency, 10 such that risk is still increased even with partial restoration of specific functional immune responses. As a result, tuberculosis is the most common HIV-associated mycobacteriosis in Europe in the ART era, whereas in the pre-ART era disseminated M aviumwas more common. 11 A finding not emphasised by Gupta and colleagues was that 60% of HIV-associated tuberculosis diagnoses (n= 930) were coincident with (or occurred within 90 days of) HIV diagnosis, 7 showing that HIV often first presents with tuberculosis. These tuberculosis cases would better be prevented by improved HIV screening in asymptomatic individuals, which would lead to earlier diagnosis.

As shown previously, 8 the strongest risk factors for tuberculosis are low current CD4 cell count, detectable viral load, and non-white ethnic origin. 7 Gupta and colleagues did not examine the nadir CD4 cell count before ART as a risk factor: this is strongly associated with tuberculosis risk during the initial months of ART and also affects the subsequent trajectory of immune recovery. 12 Lower nadir CD4 cell counts are not only associated with impaired long-term restoration of the CD4 count13 but also with persistent functional deficits in antimycobacterial immune responses. 14 A comparison of the nadir CD4 cell counts between racial groups would have been interesting. Groups with poorer health-care access might have had lower nadir counts and this might in part account for some of the interethnic differences observed in tuberculosis risk. Gupta and colleagues’ data emphasise a need for other interventions to prevent tuberculosis in patients receiving long-term ART in the UK. ART and isoniazid …