This study demonstrates a high prevalence of MDR-TB in children in Johannesburg, South Africa. MDR-TB prevalence of 8.8% in cases on which drug-susceptibility testing was performed in our study are higher than the point-prevalence of MDR described in other African paediatric studies (2.3-6.7%) [4
]. Drug resistant-tuberculosis was prevalent in 16% of children with culture-confirmed tuberculosis, whose specimens underwent susceptibility testing. Notably, only four of the children with MDR-TB had a known exposure to an adult tuberculosis contact, none of whom were diagnosed with MDR-TB. Contact tracing in our context is poorly performed; a study at three Johannesburg academic hospitals (including CHBH and RMMCH) showed that 15% of 113 severely malnourished paediatric inpatients had contact with an adult recently diagnosed with tuberculosis, and none were offered IPT [9
]. Lack of MDR-TB household contacts in our cohort suggests transmission outside the home, but further epidemiologic patterns could not be ascertained. As childhood tuberculosis is a marker of ongoing adult disease and MTB transmission, our study suggests that MDR-TB transmission occurred within the community as well as within the child's household [10
]. As childhood tuberculosis also serves as a measure of the drug-susceptibility patterns of MTB being transmitted within a community, our study indicates that there is a high prevalence of MDR-TB circulating in our setting.
High rates of drug-resistant MTB may impact on tuberculosis preventive therapy strategies that have recently been recommended in the South African national TB guidelines (2009) [13
]. In children with INH mono-resistance or poly-resistance, without rifampicin resistance, 75% were HIV infected. The high prevalence of isoniazid resistance (14.2%; 95% CI, 9.0 to 20.9%) demonstrated in our study raises questions about the effectiveness of isoniazid preventive therapy in our community [14
]. The high prevalence of isoniazid resistance in this setting may also, in part, explain the lack of efficacy of primary isoniazid prophylaxis in improving tuberculosis-free survival in HIV-infected children in our setting [15
Current South African tuberculosis guidelines (2009) [13
] recommend tuberculosis microscopy and culture in adult tuberculosis suspects with smear-negative disease who are ill and not responding to antibiotics, regardless of HIV status. No drug susceptibility testing is recommended for smear-positive adult cases. One specimen for MTB culture and sensitivity is sent for high-risk TB suspects [13
]. In child tuberculosis suspects, smear and culture of sputum or gastric aspirates is recommended, and paediatric studies have demonstrated that it is possible to perform these investigations in many settings, in children of varying ages; however they are often not done routinely in outpatient settings [16
]. Currently, drug susceptibility testing is not routinely performed on MTB isolates cultured from children and drug susceptibility testing of isolates in children with tuberculosis in our cohort was performed at clinician discretion in an academic hospital setting, deviating from current local guidelines [13
]. Line probe assays have been demonstrated to provide drug susceptibility results within 5 - 7 days in high-throughput laboratories in microscopy-positive specimens, or once MTB is cultured; [18
] this test still requires evaluation in specimens derived from children. Future plans nationally in South Africa are to enhance the capacity for routine testing of all smear-positive adult and paediatric specimens using this assay, in order to report drug sensitivity patterns more rapidly.
High MDR-TB attributable mortality (30.8%) was evident in our cohort at a median of 2.8 months after investigation. Of the children who demised, three were HIV-infected on HAART and the fourth child was an HIV-uninfected ex-premature infant with associated congenital abnormalities. The three children who demised and were started on MDR treatment, started MDR-TB therapy at a median of 1.5 months after specimen submission, demonstrating no significant delay in access to specific therapy compared to the survivors. This mortality is higher than described in a previous study from the Western Cape where 10% of children with MDR-TB died over a four year study period, 50% of whom were HIV-infected; [19
] and compared to 3.5% described in children with culture-confirmed MDR-TB receiving community based treatment in Lima, Peru [20
]. In our study, one child died of MDR-TB meningitis and the other three children died of pneumonia and acute respiratory failure. All children died in hospital as they were too unstable to be transferred to a MDR-TB treatment facility. Appropriate isolation facilities are not widely available in many general hospital settings in South Africa, and therefore these children constituted an infection control dilemma when known to have MDR-TB, as well as prior to their MDR-TB diagnoses. Even if the diagnosis of MDR-TB is expedited through modern diagnostic techniques, increased risk of nosocomial MDR-TB transmission to fellow patients and healthcare workers is likely if adequate isolation facilities are not available [21
]. With increased MDR-TB rates evident especially in high HIV prevalence areas, infection control needs to be addressed urgently.
Our study has limitations, including the fact that we may have overestimated the true MDR-TB prevalence in children in Johannesburg, as patients attending referral hospitals may be at higher risk for drug-resistant tuberculosis compared to those investigated and treated at primary care facilities. In the Western Cape (2003 - 2005), however, there was no difference in drug susceptibility patterns between MTB isolates from hospitalized compared to community-based childhood TB cases [5
]. Furthermore, in our study, DST was performed in only 72.5% of children with culture-confirmed tuberculosis at clinician discretion, which may have introduced some bias. Our analysis also excluded all specimens with culture-confirmed mycobacterial isolates derived from lymph node aspirates, in order to control for the possibility of drug resistance being associated with culture of M. bovis-
] which is innately resistant to pyrazinamide and variably resistant to isoniazid [24
]. Drug susceptibility testing against pyrazinamide is not routinely performed in our setting and is only performed on clinician request. As we only had pyrazinamide resistance patterns for one specimen, we could not make any assumptions about whether these lymph node isolates were M.bovis
based on pyrazinamide resistance. As the study was retrospective and record-based, contact history data may be inaccurate. Detailed data on children with drug-susceptible MTB was not available for this study, as information was collected from laboratory records.
This study reports MTB and MDR-TB/HIV co- infection rates in the HAART era. In our cohort there were high rates of HIV co-infection. There was, however, no difference in HIV co-infection prevalence between children diagnosed with drug-susceptible and MDR-TB (P = 0.771), suggesting that HIV infection does not appear to be a risk factor for MDR-TB in this setting. There was also no significant difference between absolute CD4+ count, CD4+ percentage and HIV viral load in drug-susceptible and MDR-TB cases. Our study findings, nevertheless, highlight the importance of ascertaining the HIV-infection status of every child diagnosed with tuberculosis.