The tuberculosis (TB) epidemic is poorly controlled in sub-Saharan Africa; the region that reports the highest TB incidence rates and highest prevalence of human immunodeficiency virus (HIV) infection [1, 2]. Prevention of mother to child transmission (PMTCT) programs are poorly established in many countries and in many instances fail to use potent antiretroviral regimens, resulting in huge numbers of HIV-infected children.

Childhood TB contributes significantly to the global TB case load (15–20% of cases), [3–5] especially in Africa where TB has been identified as a major respiratory cause of death in children [6]. The high disease burden results from ongoing TB transmission (poor epidemic control) and increased vulnerability as a result of HIV-induced immune compromise. Compared to HIV-uninfected children, HIV-infected children demonstrate greater morbidity and mortality from TB [5, 7–11], as well as increased risk of rapid disease progression [12], unsatisfactory treatment response [7, 8, 13–15], and TB recurrence [13]. In resource-limited settings, mortality rates among HIV-infected children diagnosed with TB range from 20–35% [5, 8, 9, 13–15], but the TB outcome in the era of highly active antiretroviral therapy (HAART) is less well documented.

With recent international efforts, HAART has become more widely available. The South African government supported the roll-out of HAART since 2004; all children with severe immune suppression and/or clinical stage 3 or 4 disease are eligible. HAART reduces the HIV viral load, which allows CD4 T-cell repletion and assists with the restoration of immunological function. Although complete functional immune reconstitution is rarely achieved, the beneficial effect of HAART in HIV-infected adults [16–19] and children [20–23] is well documented. In HIV-infected adults, HAART reduces the incidence of TB and the risk of TB recurrence [18, 24, 25], it also improves survival rates among TB patients [25–27]. A recent retrospective study from South Africa demonstrated a three-fold reduction in the incidence of TB in children on HAART compared to those not on HAART [28]. Restoration of specific antimycobacterial immune responses has been documented in children receiving HAART [29]; two recent case series described immune reconstitution inflammatory syndrome (IRIS)-related phenomena in children [30, 31], but the prevalence and presentation of these phenomena require better description.

TB is regarded as a World Health Organization (WHO) HIV stage 3 (intra-thoracic) or 4 (extra-thoracic) condition. However, the best time for the introduction of HAART following a TB diagnosis remains uncertain; WHO recommends introducing HAART at different time points, starting earlier in the more immunocompromised children or where response to TB treatment is poor [32]. Early introduction of HAART carries a higher risk of IRIS-related phenomena, which may be misinterpreted as poor treatment response, drug-related adverse events or TB/HIV disease progression [30, 33, 34]; there is limited data in children to guide clinical decision making.

WHO recommends standard TB treatment, irrespective of the child's HIV status [22]. However, some uncertainty about the optimal treatment duration remains, due to reports of increased TB recurrence in HIV-infected children not on HAART who received standard TB treatment [6, 7, 23]. More information is required regarding TB recurrence in HIV-infected children and the influence of HAART. There are also concerns regarding common toxicities, drug interactions and treatment adherence if TB therapy and HAART are given concurrently. The aims of this study were to describe the clinical presentation and outcome of all children on HAART, in whom TB was diagnosed either during pre-HAART screening or during post-HAART follow-up.

We conducted a retrospective descriptive study; data were collected by review of clinical files and chest radiographs. The files of all HIV-infected children (aged <13 years) who were initiated on HAART from 1 January 2003 through 31 December 2005 at Tygerberg Children's Hospital (TCH) were reviewed. We collected data on all TB episodes from the time of pre-HAART screening, defined as the 9 months prior to HAART initiation, until the end of the study period. Each TB episode was studied from treatment initiation until completion (where data was available), or death. Children who received single or dual antiretroviral therapy and children enrolled in TB prophylaxis trials were excluded.

During the study period, 48% of HIV-infected children received TB treatment. This confirms the high TB burden experienced by HIV-infected children in TB endemic areas. Children with TB were significantly more immunocompromised and malnourished than those without TB, which may relate to the severity of the underlying HIV disease and/or reflect the detrimental effects of the TB itself. Most children were only commenced on HAART once they had advanced HIV disease (83% had WHO advanced or severe immune suppression), reflecting the large number of children who had been awaiting access to HAART, as well as conservative eligibility criteria. The high proportion of children with previous TB, among those not diagnosed with TB during the study period, can be explained by their relatively high mean age (34 months) compared to those who were diagnosed with TB (mean age 19 months). In this TB-endemic setting it is expected that older children would be significantly more likely to have had a previous episode of TB.

Most TB episodes were diagnosed in the 9 months prior to HAART initiation, during which time a diagnosis of TB is actively excluded. The number of TB diagnoses made pre-HAART was five times greater than post-HAART, which is even greater than the three fold reduction reported by Martinson et al.[28]. When calculating the number of TB cases per 100 patient years there was an eight fold reduction in TB episodes (53.3 versus 6.4), comparing the 9 months pre-HAART period to the post-HAART period. This was not a manifestation of overzealous diagnosis in the pre-HAART period, as the striking observation was even more pronounced when only bacteriologically confirmed cases were taken into account (18.3 versus 1.8). The reduction achieved is similar to that recorded with the use of universal INH preventive therapy [35]. It must be pointed out that the number of TB cases among those not on HAART may have been inflated by the fact that TB was frequently the presenting disease that lead to the diagnosis of HIV-infection and HAART initiation, especially before the institution of effective PMTCT programs and routine testing of HIV-exposed babies.

The clinical presentation of TB in our cohort was similar to what has previously been described [5, 7–10] The presence or absence of a TB index case was recorded in less than 80% of children. This is an important oversight as careful scrutiny of the index case often provides important clues to identify drug-resistant disease in child contacts. Another interesting observation, although numbers were very small, is that the majority of index cases with documented sputum smear results were sputum smear-negative. Adults with sputum smear-negative pulmonary TB do transmit disease, albeit at a reduced rate. Current WHO guidelines do not recognize the potential transmission risk posed by exposure to a sputum smear-negative index case, especially if this is the child's mother or primary caregiver [36]. Weight loss or failure to thrive was the most frequent presenting symptom, but this may be caused by advanced HIV disease itself or by other HIV-related opportunistic infections. In addition accurate symptom definition, which is essential for optimal specificity [37], is impossible with retrospective analysis of routine clinical data.

The disease spectrum observed is comparable to that reported in HIV-uninfected children from the same area, although the number of children with disseminated (miliary) disease was significantly increased [9/134 (6.7%) vs 11/414(2.7%); OR 2.6, 95% CI 1.0–7.0] [38]. The timing of HAART introduction was largely dependent on the child's clinical condition; being introduced earlier in children with severe immune suppression. IRIS was not prominent in this group; in fact, disease progression in the absence of HAART often prompted its early initiation. The incidence of TB IRIS (7,4%) is slightly higher that that reported by Puthanakit et al.[30] and was higher among those with bacteriologically confirmed TB (4/46; 8.7%), compared to those without bacteriologically confirmed TB (6/88; 6.8%).

This study documents the frequency with which the management of TB in HIV-infected children deviates from standard protocol. Despite local [39] and international guidelines [40], therapy often has to be individualized. In this cohort, only 75/137 (55%) of TB episodes were treated with the standard 3-drug TB regimen; 33/137 (24%) for the standard 6 months. The proportion of children who received expanded treatment regimens was greater than that expected from the number of extra-thoracic and/or drug-resistant cases, which reflects extensive disease, high levels of co-morbidity and caution among clinicians who treat HIV-infected children with TB.

Adverse events that were probably drug-related occurred in only 7 children of whom 3 were on concurrent HAART. Although the absolute numbers are small it indicates that more than 5% of children may have experienced drug-related adverse events, with a further 4% experiencing non-drug-related adverse events. In total, adverse events necessitated treatment modification in nearly 10% of cases. This is important, as second line TB drugs are not as effective as first-line regimens, are more often associated with drug-related adverse events and result in longer treatment durations, with a concomitant increased risk of drug interactions and poor adherence. Although HAART was only suspended in a single child, stopping HAART for any period of time may impact negatively on the long-term prognosis of HIV-infected children.

The cohort in general did well, with 71% reporting positive outcomes. TB diagnosed after HAART initiation was associated with excellent outcomes. Although the reduced mortality recorded in children on HAART, compared to those not on HAART, at the time of TB diagnosis failed to reach statistical significance, no child who had been on HAART for more than 6 months prior to the TB episode died. Among the children who died, the majority had advanced HIV disease, and most died of HIV-related conditions other than TB. It is therefore not surprising that the group where HAART was introduced within 2 months of TB treatment had the worst outcomes, as the timing of HAART introduction was determined by the severity of the underlying HIV disease (clinical and immunological stage of the patient). These were young, severely immune suppressed and critically ill children in whom HAART was often given as a last resort. In this respect it is important to note that despite their grave clinical condition, 65% of these children survived with positive outcomes.

The early introduction of HAART was rarely (<10% of cases) associated with drug related adverse events or significant immune reconstitution phenomena. Two cases of BCG adenitis developed shortly after HAART initiation and probably represent BCG IRIS disease [41]. Although IRIS may have contributed to 4 deaths this remains highly speculative, as 2 deaths occurred relatively late (more than 5 months after HAART initiation) and the other 2 had progressive MDR TB; the progressive and drug-resistant nature of their disease was probably responsible for their demise.

Being retrospective, our study is limited by incomplete data and missing records. However, we were able to retrieve essential data in the vast majority of children (>95%) and believe that the information provided allows accurate assessment of the diagnostic challenges related to TB and HIV in everyday clinical practice. Although this is a hospital-recruited cohort, significant selection bias is unlikely as, during the study period, TCH functioned as the only point of access to HAART for children from the surrounding areas.

The study confirms the high incidence of TB among HIV-infected children in TB endemic settings. Although the management of TB in HIV-infected children may be complicated, the majority of children achieved a positive outcome. HAART initiation was followed by a substantial reduction in the incidence of TB, yet children only accessed HAART at an advanced stage of HIV disease. For HIV-infected children living in TB endemic areas, starting HAART at an earlier stage is likely to reduce TB-related morbidity and mortality.