Risk factors for active tuberculosis among adults on highly active antiretroviral therapy in Africa: a reply

Abstract

We thank Seyler and colleagues for their insightful thoughts concerning risk factors for the development of tuberculosis during HAART. Three studies addressing this question have been published and have differing findings [1–3]. The persistence of high rates of tuberculosis during HAART is an important issue, and an accurate identification of risk factors is key to the development of appropriate strategies to reduce this burden of disease. Studies from Cape Town, South Africa [1] and Europe/north America [3] found that more advanced pretreatment immunodeficiency and lack of response to HAART were the principal factors associated with the risk of tuberculosis during HAART. In contrast, Seyler and colleagues [2] found in a west African cohort in Abidjan that the only risk factor was a previous history of tuberculosis. We have previously suggested that the lack of an association between the risk of tuberculosis and pretreatment immunodeficiency in the Abidjan study may relate to the restricted composition of the cohort [4,5]. The inclusion of few patients with stage 1 or 2 disease or CD4 cell counts greater than 200 cells/μl will have diminished the power to assess CD4 cell count and clinical stage as risk factors. Explaining the discrepancy in findings concerning a history of previous tuberculosis as a risk factor is probably more complex. HIV-associated immunodeficiency is an important risk factor for the recurrence of tuberculosis, whether caused by relapse (resulting from incomplete sterilization of tuberculous lesions) or exogenous re-infection [6]. Evidence suggests that the restoration of Mycobacterium tuberculosis-specific immunity is incomplete during at least the first year of HAART [7], and it is therefore certainly plausible that a previous history of tuberculosis would persist as a significant risk factor for incident tuberculosis after the initiation of HAART. However, whether such an association is demonstrable in a given cohort is likely to be determined by the proportion of patients who have a past history of tuberculosis, the relative strength of association of other variables such as baseline CD4 cell count and their risk of re-exposure to tuberculosis. Whereas the underrecording of a previous history of tuberculosis may be a significant problem in community-based tuberculosis clinics [8], which are frequently overburdened in much of sub-Saharan Africa, we doubt that this is likely to be a significant factor among carefully documented patients in a study cohort, many of whom were in long-term medical care before antiretroviral therapy. Although the proportion of patients in the Cape Town cohort who had a previous history of tuberculosis was comparatively low (14%), this probably reflects the sociodemographic composition and immune profile of the cohort rather than reflecting underreporting. In marked contrast, approximately 50% of patients accessing a community-based antiretroviral treatment programme based within in a poor urban township in Cape Town (previously described in Lawn et al. [9]) have a history of previous tuberculosis. It may prove that study populations with markedly differing baseline characteristics and levels of community exposure to tuberculosis may have different risk factors associated with incident tuberculosis during HAART. The relative contribution of relapse versus re-infection as a cause of recurrent tuberculosis is also likely to differ between settings. Inadequate antituberculosis treatment increases the risk of tuberculosis relapse; in contrast, a higher incidence of tuberculosis in a community is associated with an increased likelihood of exogenous re-infection [6,10]. As patients with tuberculosis in Cape Town receive a 6-month rifampicin-containing regimen and the tuberculosis incidence rates in some communities are among the highest in the world [11], we suspect that a large proportion of recurrent tuberculosis is caused by exogenous re-infection. In this context, one would expect that antituberculosis treatment would be followed by a time-dependent reduction in tuberculosis incidence, and that the rate would gradually increase with time after the completion of antituberculosis treatment as the opportunity for re-exposure increases. Seyler and colleagues rightly point out that among patients with a history of previous tuberculosis, a greater proportion of those in the Cape Town cohort compared with the Abidjan cohort completed antituberculosis treatment in the year before the initiation of HAART. This may have conferred a time-dependent protective effect against tuberculosis during the first 1–2 years of HAART in the Cape Town cohort when the majority of cases of incident tuberculosis occurred. This may be an important factor contributing to the differing findings between the Abidjan and Cape Town studies. The potential benefit from the co-administration of isoniazid prophylaxis concurrently when initiating HAART may thus be diminished if a high proportion of enrolling patients have recently completed antituberculosis treatment. In conclusion, risk factors for tuberculosis during HAART may differ between study populations of differing composition and in different settings. Data from many different sites are needed to clarify the overall picture. Sponsorship: S.D.L. is funded by the Wellcome Trust, London, UK.