Reply to Ouattara et al.: past history of tuberculosis is not a risk factor for incident tuberculosis during antiretroviral treatment in South Africa

We thank Ouattara and colleagues for their letter concerning risk factors for incident tuberculosis during antiretroviral treatment (ART) in sub-Saharan Africa [1]. In a study from Abidjan that included 12 cases, Seyler et al. [2] identified a past history of tuberculosis as the sole risk factor for incident tuberculosis. We reported a larger number of cases (n = 27) within a hospital-based study cohort in Cape Town and, in contrast, a low baseline CD4 cell count and advanced World Health Organization (WHO) stage of disease were the principal risk factors [3]. In a second, much larger community-based study [4], we found that the current CD4 cell count was the sole independent risk factor for incident tuberculosis (n = 81). In both our studies, a history of previous tuberculosis was consistently found not to be a significant risk factor, agreeing with other unpublished studies from South Africa [5], Uganda [6] and Senegal [7]; a further study from Uganda reported a strong but statistically non-significant trend towards an association [8]. These cohorts differ in demographic characteristics, the baseline degree of immunodeficiency, socioeconomic status, the proportion of patients with previous tuberculosis treatment, exogenous tuberculosis infection pressure, and duration of follow-up. Patients with a previous history of tuberculosis may also differ with respect to tuberculosis treatment regimens received, treatment adherence, and rates of drug resistance, affecting the risk of recurrent tuberculosis. Variation between the results from different cohorts might thus be expected. However, understanding these differing findings is nevertheless important to permit the development of strategies to reduce incident tuberculosis during ART. The Abidjan data would favour isoniazid secondary prophylaxis after the completion of tuberculosis treatment. In contrast, our data suggest that the earlier initiation of ART with the maintenance of high CD4 cell counts is important or that adjunctive isoniazid prophylaxis might be used irrespective of a previous history of tuberculosis. The suggestion of Ouatarra et al. [1] that the data from our two studies in Cape Town are contradictory is clearly incorrect. Despite major differences in cohort characteristics, the studies were entirely consistent in finding no significant association between the risk of incident tuberculosis and a previous history of tuberculosis [3,4]. The 95% confidence intervals around the adjusted estimates from the two studies overlap the null and do not reach statistical significance. WHO stage is an important variable reflecting the risk of morbidity and mortality independent of that reflected by the CD4 cell count. We therefore included this variable in our analyses of tuberculosis risk. However, pulmonary and extrapulmonary tuberculosis are themselves WHO stage 3 and stage 4-defining conditions, respectively, and Ouattara and colleagues [1] question the inclusion of both the WHO stage and a history of previous tuberculosis in the same multivariate analysis. Although they overlap, these two variables reflect different information and were included separately. In our analyses, however, we found that the exclusion of one or either variables did not significantly affect the associations of the other. WHO stage was very strongly associated with the risk of incident tuberculosis in the first of our studies [3]. The large number of incident cases in the second study, however, provided the opportunity for us to include response to ART as a potentially important additional risk factor [4]. When the current CD4 cell count (reflecting both baseline and CD4 cell response) was included in the multivariate model, then no other baseline characteristics, including WHO stage, remained significantly associated with the risk of tuberculosis [4]. The finding that alterations in tuberculosis risk parallel changes in immune function over time obviously makes biological sense; indeed, we have found that mortality risk alters in a similar way [9]. Therefore, the suggestion by Ouattara et al. [1] that the data from our two studies are contradictory with regard to WHO stage is again incorrect. The second of our studies builds on the findings of the first and provides a more definitive analysis. We have previously suggested that a history of previous effective tuberculosis treatment among patients enrolling in our ART programme causes a time-dependent reduction in tuberculosis risk as a result of the sterilization of Mycobacterium tuberculosis infection [4,10]. Using overnight enzyme-linked immunospot assays and 7-day whole blood assays to detect M. tuberculosis-specific T-cell responses, we have recently generated laboratory data that are consistent with this hypothesis (S.D. Lawn, unpublished data). Results indicate that patients enrolling for ART who do not have active tuberculosis but have previously completed tuberculosis treatment have a much lower likelihood of having viable M. tuberculosis infection than those who have not previously received tuberculosis treatment. Such patients would remain at a lower risk of tuberculosis until re-infected exogenously. In summary, epidemiological data from our two study cohorts and more recent immunological studies are entirely consistent in showing that a previous history of tuberculosis is not associated with an increased risk of incident tuberculosis during ART in South Africa. Sponsorship: S.D.L. is funded by the Wellcome Trust, London, UK; R.W. is partly funded by the National Institutes of Health, USA, RO1 grant A1058736-01A1; L.M., L.G.B. and R.W. are all partly funded by the National Institutes of Health through a CIPRA grant 1U19AI53217-01