The extremely high TB prevalence in our study demonstrate that South Africa's dual HIV and TB challenge is not confined to urban settings; this rural cohort had a TB prevalence of 25.3%, similar to urban South Africa [8
] and higher than figures reported elsewhere in sub-Saharan Africa [7
HIV associated immunosuppression is known to increase the development of active TB in those latently infected and increase susceptibility to new TB infection [20
]. We show that people with a CD4 count of less than 50 cells/μl prior to ART initiation were more likely to have prevalent TB, which is consistent with previous findings [8
The incidence of TB following ART initiation was similar to that observed within an urban community in Cape Town [8
]. Despite similar cohort characteristics, the incidence within the first 3 months in our study, whilst within confidence intervals, is slightly lower. In that study, sputum culture was part of the work-up, whilst in our setting culture was performed infrequently. Given that much of the TB is pulmonary, this again highlights the need for more active and aggressive diagnosis within a rural setting.
The high incidence of TB in the initial three months of ART compared with later may reflect an immune reconstitution inflammatory syndrome (IRIS) to TB infection which was present, yet undiagnosed at the time of ART initiation; this phenomenon has been termed unmasking tuberculosis-associated IRIS [22
]. In this analysis, incidence refers to the incidence of diagnosis of disease rather than disease per se. The quoted figures for prevalence are therefore of prevalent diagnosed disease and the actual prevalence of disease might be higher. Incident calculations are not affected by high numbers of new TB diagnosis in the first two weeks, however, as this was only recorded in 4 cases; removal of these 4 from incidence calculations does not affect the final rate. Obviously it is impossible to comment on deceased or lost to follow-up cases.
The association between previous, prevalent and incident TB is complex and the subject of some debate [8
] and conflicting evidence [8
]. Our study is limited in the detail available about previous treatment and the number of follow-up CD4 results available; and therefore we were unable to confirm an association between higher CD4 attainment during ART and reduced TB incidence. However, we did find that previous TB was independently associated with prevalent TB at initiation, for which there are a number of possible explanations. In areas of high transmission, for example particular households or workplaces, re-infection may be common, particularly amongst the immunologically vulnerable. High re-infection rates may also be explained by nosocomial spread of TB within hospitals and clinics [25
]. Delayed diagnoses coupled with lack of infection control adds to the increased risk of re-infection. Previous poor treatment adherence is another potential explanation, as relapse of partially treated disease as immunodeficiency advances could account for retreatment. In the absence of routine culture based testing it is not possible to tell whether drug resistance TB could account for the significant number of individuals requiring retreatment, but community rates of drug resistance are high with 40-55% [26
] of individuals who gave culture specimens on retreatment showing evidence of some drug resistance, and 27% retreatment cases identified as MDR [27
]. Drug resistant TB in the HIV setting is of particular concern for the control of both epidemics [6
Separately analysing incidence in the initial three months and thereafter allows us to illustrate the factors strongly associated with incident TB. Male sex and younger age were not risk factors in the cohorts in Cape Town, Uganda and the Ivory Coast [8
] but were in the ART-LINC cohort [28
] and male sex, lower CD4 and lower BMI were associated with incident TB in Johannesburg [29
]. Male gender has been historically linked to increased TB incidence [30
] but in our study was associated with TB incidence only in the first three months post ART initiation. This may be attributable to illness behaviour; symptom denial at the time of presentation for ART initiation may be more common in men and therefore higher rates of immune reconstitution TB may be expected in men. Male sex in previous South African studies has been associated with increased delay in diagnosis of TB [31
]. Work related or social behavioural exposure such as travel in minibus taxis may be more common in men, and these may be areas of target for active case finding activities for both TB and HIV, it would be expected however, that this higher rate in men continue after three months which it does not in this study. After the initial three months, we found that only advanced HIV stage was significantly associated with incident TB, a finding which supports other evidence that more advanced disease at initiation is associated with increased risk of TB disease [28
The high TB incidence and high mortality in the first three months after initiation of ART necessitates improved active case finding through, amongst other things, optimised screening practices. The high mortality from TB in HIV infected individuals has been clearly illustrated in South Africa [32
]. The challenge can best be addressed by the integration of TB and HIV services [33
], especially in the most rural and under-resourced areas. Further, there is potential for investigating the use of targeted chemoprophylaxis with isoniazid possibly both pre and post ART initiation; although there remains significant concern about this approach in an area of high levels of TB drug resistance [34
Previous studies have shown similar virological outcomes in HIV patients on ART taking concurrent TB treatment and those not taking TB treatment [8
]. However, these were urban patients and follow up was less than two years. We have illustrated a non-significant trend towards detectable viral load in cases with incident TB. A longer period of follow-up might reveal significance. What is clear from our data is the strong association between prevalent and incident TB and mortality. This illustrates an opportunity to select groups in need of increased support and monitoring and advocates for the prevention of late presentation with illness on, and before ART.
Interestingly, the incidence of new TB cases was not geographically uniform across the sub-district and a cluster of unusually low TB incidence was detected surrounding the district hospital. These communities close to the hospital have easy access to diagnostic investigations such as chest x-rays, and were within easiest access of the first ART services in 2004/5. Both of these reasons may have contributed to earlier ART initiation in comparison to the rest of the sub-district. This in turn may have resulted in a reduced risk of incident TB, and earlier access to TB diagnosis and treatment leading to a reduction in the community's overall burden of TB disease.
However, 40% of patients geo-coded to the low-risk cluster reported living in the Hlabisa tribal area but only supplied a postal (and not a physical) address. It is therefore at least theoretically possible that a social rather than a geographical clustering pattern is being detected. This explanation is improbable, however, given the high population concentration in this area and the fact that individuals are likely to live close to the area where their post is collected. The use of geospatial techniques for analysing clustering of TB cases will aid understanding of the TB epidemic in rural sub-Saharan Africa (SSA), similar to the way in which mapping of HIV cases has highlighted not only areas where high transmission is occurring (along major transit routes [36
]) but also the targeting of areas where infection control interventions will be most effective. This analysis is particularly important in HIV and TB co-infection which remains a leading killer in SSA.
Our study has several limitations. The original purpose of this programme is service and not research and care is largely provided by nurses and counsellors. It is therefore possible that the information recorded does not give a full picture of TB incidence. Further, in this sub-district of 2280,000 people, where over 3,500 TB cases are diagnosed annually, and as many as 80% of pulmonary cases are smear negative (Tom Heller, personal communication March 2009), the possibility of over-diagnosing TB on clinical grounds cannot be dismissed. We are restricted by lack of information on the cause of mortality or the status of lost to follow-up patients. It is possible that several of our deceased patients had incident TB, and some of the defaulted patients may have been started on TB treatment.
In our study, the absence of data from the era prior to ART hinders drawing conclusions about the potential impact of ART on the risk of developing TB. Studies in urban South Africa suggest a reduced incidence of TB for HIV infected people taking ART before and after national roll-out [8
] and some have suggested that the reduction in TB incidence in high and low-income countries after HAART is as high as 70-90% [38
]. However there is lack of evidence for rural populations which face different challenges. The risk of TB whilst on ART during long term follow-up becomes stable in our setting at a rate six times higher than the overall TB incidence in KwaZulu-Natal [4
] which supports suggestions that ART alone will not be a fully effective TB control strategy for several reasons, including the lack of restoration of MTB immune response on ART to that of an HIV negative person coupled with prolonged lifespan after ART [38