Patients who developed TB after ART initiation in our large urban ART cohort were more likely to have suboptimal immune reconstitution despite TB treatment. Active TB in the first year after ART initiation had long-lasting effects on immune restoration: the median CD4 count change at 2 years was significantly lower after correction for lower baseline CD4 count. These patients also fulfilled the WHO criteria for immunological failure more often. To our knowledge, the immunosuppressive effect of incident active TB after ART initiation has not been previously described.
The potential of virological and immunological response to ART seems to be equal in both resource-rich and resource-limited settings 
, although there are some reports that less overall immune recovery occurs more often in resource limited settings 
. The most important factor associated with suboptimal immune reconstitution in all settings is baseline CD4 count, but other factors have also shown to be associated, such as higher baseline HIV viral loads, male sex, older age and hepatitis C co-infection 
. The use of AZT-based regimens has also been implicated 
. We were able to analyse and identify an association with most of these factors in our study, except for viral load and hepatitis C co-infection as our setting does not allow for routine viral load and serologic testing. A higher baseline CD4 count was associated with a lower CD4 increase after two years of ART, which is consistent with previous reports that show that patients with the lowest baseline CD4 count tend to have the largest increase 
. In multiple linear regression analysis, the negative effect of incident TB during ART on immune restoration was independently associated.
The underlying mechanisms of immunological non-response are not completely understood, but it seems that a complex model of immune activation, T cell turnover and homeostatic regulation is responsible for CD4+ T cell loss 
. Excessive T cell destruction due to T cell activation plays an important role, and has been shown to persist even after virological suppression occurs 
. Higher levels of T cell activation have been reported in HIV-seronegative Africans 
, and have been attributed to frequent infections by various endemic pathogens, such as parasitic infections. It has been suggested that co-infections might limit the capacity for restoration of CD4 counts through an immunosuppressive effect on hematopoiesis and, most importantly, through augmentation of the HIV-induced heightened immune activation leading to widespread apoptosis of both HIV-infected and uninfected lymphocytes 
. In a small study done in Uganda, the presence of co-infection with highly prevalent and endemic pathogens in patients with and without ART was associated with increased CD4+ T cell activation independent of CD4 count and viral load 
. Moreover, in another study done also in Uganda, patients who developed an AIDS-defining condition according to the WHO clinical staging system within 12 weeks pre or post ART initiation, took longer to achieve an increase in CD4 count of 50 cells/mm3 
TB induced immunosuppression is a well-studied area: an investigation of T cell cytokine responses in HIV-negative pulmonary TB patients showed a persistent depressed tuberculin-induced IFN-γ response up to 18 months despite successful treatment, suggesting a long-lasting depletion or primary dysfunction of antigen-responsive T cells from the peripheral blood due to active TB 
. Furthermore, apoptosis of Mycobacterium tuberculosis
-reactive CD4+ and non-CD4+ T cells seems to be increased in pulmonary TB patients 
. Taken together, these results suggest that TB may have a long-lasting immunosuppressive effect, which could account for the persistent decreased CD4 count levels we found in our patients with incident TB. We hypothesize that the co-infection leads to a long-lasting augmentation of already heightened T cell activation in HIV-infected patients, followed by increased apoptosis of CD4 cells with decreased immune restoration.
Our data suggests that there may be a specific effect of incident TB after ART initiation on immune recovery. A Malawian study of 76 clinical ART failure cases showed that extrapulmonary TB and Kaposi's sarcoma were the most common conditions causing misclassification compared to virologic failure gold standard, but these factors were not significant in multivariable analysis 
. Other studies have reported that AIDS events (including TB) before or at the time of ART initiation are not associated with decreased immune recovery 
, although the numbers of patients analysed were small and this was not their primary study objective. There has also been a report suggesting the contrary 
. In a study from South Africa, there was no effect on immune recovery with prevalent and early incident TB together, although a relatively low proportion (25%) of patients were analysable at 48 weeks of follow-up 
A recent paper published by Lawn and colleagues showed that the time spent at lower CD4 count levels after ART initiation was the predominant predictor of TB risk at 4 months of ART and onwards in South Africa 
. To address this issue in our cohort, we restricted our analysis at 2 years after ART initiation to compare CD4 counts of patients who initiated TB treatment within the first 12 months of ART, and excluded all cases of active (and thus also recurrent) TB. Therefore, our results do not examine the duration of the effect of incident TB on immune recovery. The poor CD4+ T cell reconstitution could also represent a temporary delay in immune recovery followed by the same recovery rate as non-TB patients as suggested by the similarity in slope in . Patients with incident TB after ART would still spend a significantly longer amount of time at lower CD4 counts compared to patients who were TB free after ART.
As TB is the most frequent opportunistic infection in HIV-infected patients in our setting, it is possible that the decreased CD4 counts in patients after TB treatment may partly account for the high rates of misclassification of treatment failure based on the WHO immunological criteria, which have been reported in a number of recent articles from sub-Saharan Africa 
. The sensitivities of the WHO criteria ranged from 8% to 23% with specificities from 90% to 98%. In our study, patients with incident TB during ART were twice as likely to be classified as having immunological failure after 2 years on ART than those without, a finding which was confirmed in multivariable logistic regression to be independent of other associated factors. Misclassification of these patients as failing immunologically puts them at risk for inappropriate switching to second line therapy. The immunosuppressive effect was not seen with CD4 count percentages. The CD4 percent may be a better parameter to use in settings without access to viral load confirmation of failure, although this data highlights the need for more affordable and accessible viral load testing 
Our large retrospective cohort study also reported comparable incidence rates of ART-associated TB to those published from earlier work in high-endemic areas 
. As expected, incident TB during ART was a common occurrence in our study. Incidence rates were very high in the first three months, and decreased to a level which was still almost three fold higher than in the general Ugandan population 
. Mortality rates after ART initiation mirrored these rates. Identified risk factors for developing TB after ART initiation in our cohort were a low baseline CD4 count and male sex (both have been previously described 
) and highlight the importance of initiating ART earlier.
Our study had several limitations. Because of the retrospective design the data was often incomplete, although we validated our data with chart reviews. We excluded a large number of possible TB cases that could not be verified by chart review which may have introduced bias. Although our population was large, the presented data was collected at only one centre, possibly affecting the generalisability of our results. We are also aware of the possible under- and over-diagnosis of TB in our setting since mycobacterial culture is not routinely used to confirm cases of smear-negative TB which are frequent in HIV-positive populations. We do feel that our setting reflects the current practice in many clinics in sub-Saharan Africa. Finally, an alternative explanation for the lower CD4+ T cell count levels in patients with incident TB during ART is decreased adherence to ART during TB treatment due to high pill burden, and side-effects 
. This may have lead to lower adherence levels and thus slower immune recovery. Insensitive measurements of adherence in our clinic (self-report and pill counts) in combination with infrequent viral load testing made it difficult to exclude this possibility. Nonetheless, the majority of the TB patients who developed suboptimal immune reconstitution for whom a viral load was measured (6 of 8 patients) had an undetectable viral load.
In conclusion, in this large urban HIV clinic in Uganda the incidence of TB is very high in the first three months after ART initiation, especially in patients with very low CD4 counts, and likely contributes to high early mortality. Furthermore, incident TB after ART may have a long-lasting effect on immune recovery which, in turn, is associated with a higher risk of opportunistic infections, mortality and non-infectious complications of AIDS. The biologic importance of our finding of suboptimal CD4 recovery in patients with incident TB during ART in terms of virologic failure is unknown, and warrants further prospective investigation. Finally, ART effectively decreases TB incidence long term. Our findings highlight the importance of implementing strategies to initiate ART earlier in sub-Saharan Africa to decrease ART-associated active TB.