Article Text

Download PDFPDF

Severe manifestations of extrapulmonary tuberculosis in HIV-infected children initiating antiretroviral therapy before 2 years of age
  1. Elisabetta Walters1,
  2. Joanie Duvenhage1,
  3. Heather R Draper1,
  4. Anneke C Hesseling1,
  5. Susan S Van Wyk1,
  6. Mark F Cotton2,
  7. Helena Rabie2
  1. 1Department of Paediatrics and Child Health, Desmond Tutu TB Centre, Stellenbosch University, Cape Town, South Africa
  2. 2Department of Paediatrics and Child Health, Tygerberg Academic Hospital, Stellenbosch University, Cape Town, South Africa
  1. Correspondence to Dr Elisabetta Walters, Faculty of Medicine and Health Sciences, Desmond Tutu TB Centre, Stellenbosch University Tygerberg Campus, K-Floor, Clinical Building, Francie Van Zijl Drive, Tygerberg, Cape Town 7505, South Africa; ewal{at}


Background Early initiation of antiretroviral therapy (ART) in HIV-infected infants reduces mortality and opportunistic infections including tuberculosis (TB). However, young HIV-infected children remain at high risk of TB disease following mycobacterial infection. We document the spectrum of TB disease in HIV-infected children <2 years of age on ART.

Methods Retrospective cohort study; records of children <2 years of age initiating routine ART at Tygerberg Children's Hospital, Cape Town, January 2003–December 2010 were reviewed. Clinical data at ART initiation (baseline) and TB episodes after ART initiation, to June 2012, were recorded. TB immune reconstitution syndrome (TB-IRIS) and incident TB were defined as TB diagnosed within 3 months, and >3 months after, ART initiation respectively. Baseline characteristics were compared in children with TB-IRIS and those with incident TB.

Results In 494 children, median follow-up time on ART was 10.7 months. Fifty-five TB treatment episodes occurred after ART initiation: 23 (42%) TB-IRIS (incidence 21.9/100 person years (py)) and 32 (58%) incident TB (incidence 3.9/100 py). Children with TB-IRIS and those with incident TB had similar baseline characteristics. Eight of 10 cases of extrapulmonary TB were severe: 4 IRIS (2 meningitis, 1 disseminated, 1 pericarditis) and 4 incident cases (1 each miliary, meningitis, pericarditis and spinal). Fifty-one children (10%) died (mortality rate 5.96/100 py). Starting ART at <1 year of age approached significance as a risk factor for TB-IRIS (adjusted OR (AOR) 8.64, p=0.06); weight-for-age Z score <−2 predicted death (AOR 6.37, p<0.001).

Conclusions Severe TB manifestations were observed among young HIV-infected children on ART.

  • HIV
  • Antiretroviral Therapy
  • Infants
  • Tuberculosis

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic

  • Initiation of antiretroviral therapy (ART) before 3 months of age in HIV-infected infants, reduces the incidence of opportunistic infections, including tuberculosis (TB).

  • Young HIV-infected children initiating ART are at risk of immune reconstitution phenomena, including TB.

  • TB incidence among HIV-infected children on ART remains higher than for HIV-uninfected children in high TB-burden communities.

What this study adds

  • Risk of incident TB in HIV-infected infants and young children starting antiretroviral therapy (ART) was high at 3.9/100 child-years of follow-up.

  • Despite active TB screening around ART initiation, incidence of early post-ART TB was high, and disease was mostly severe/disseminated.


Approximately 1000 children are vertically infected with HIV daily worldwide.1 Initiation of antiretroviral therapy (ART) before 12 weeks of age reduces mortality by up to 76%, and lowers the incidence of opportunistic infections, including tuberculosis (TB) by 60%.2 However, TB incidence in South Africa among young HIV-infected children remains high at 121 cases/1000 child-years compared to 41/1000 child-years among HIV-exposed uninfected children.3 In South Africa, a setting with high TB and HIV burden, annual risk of TB infection reaches 4%4 ,5; up to 10% of infants from HIV-affected households are exposed to an adult TB case by 3 months of life.6 Children <2 years of age are more likely to progress after infection with Mycobacterium tuberculosis (M tb) to severe disease.7 In HIV-infected children, unidentified latent or active TB may cause immune reconstitution TB or ‘unmasking TB’ when ART is initiated.8 In high TB incidence settings, 6–8% of children develop TB within 6 months of starting ART,8 ,9 but the spectrum and manifestations of disease are poorly described. TB coinfection can result in impaired long-term HIV control, especially in the early stages of ART.8

Active screening before ART decreased early TB by up to 70% in Ugandan children.10 Isoniazid preventive therapy (IPT) is recommended by WHO for all HIV-infected children after exposure to an infectious TB source case, and as primary prevention to HIV-infected children aged >1 year.11 There are limited data on the routine delivery of IPT to HIV-infected children from highly endemic settings.

At Tygerberg Children's Hospital (TCH) in Cape Town, South Africa, children are screened for TB exposure routinely before ART initiation and at every visit. We report incidence and manifestations of TB disease, and IPT delivery among HIV-infected children <2 years of age routinely starting ART.


In this retrospective cohort study we identified children starting ART at <2 years of age between 1 January 2003 and 31 December 2010, from the electronic database of the Infectious Disease (ID) clinic at TCH. We obtained clinical information from hospital and clinic (outpatient) files, and laboratory data from the National Health Laboratory System database. TB episodes and IPT prescriptions until 30 June 2012 were noted. Mortality data were obtained from patient files and the provincial death registry. Children enrolled in trials of non-routine IPT were excluded. We previously reported clinical outcomes for part of this cohort12 ,13; we now report limited analysis of children <2 years of age at ART initiation with extended observation time over 9 years. Data were extracted into an Excel spreadsheet and analysed using Stata V.12.1 Special Edition software (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, Texas, USA: StataCorp LP).


First-line ART for children <3 years of age included two nucleoside reverse transcriptase inhibitors with a protease inhibitor, lopinavir-ritonavir (LPV/r). National guidelines until 2008 recommended switching LPV/r to standard-dose ritonavir (RTV) during TB cotreatment and using RTV rather than LPV/r in infants <6 months of age, switching to LPV/r at age 6 months.14 However, the ID clinic changed to universal LPV/r for young children <3 years and ‘super-boosting’ with additional RTV during TB cotreatment before the national guidance change. Before 2009, children <2 years of age started ART based on clinical and immunological criteria (WHO clinical stage 3/4 or CD4 percentage <25%); after 2009,2 children <2 years of age started ART as soon as possible after HIV diagnosis.

We defined TB episodes as any TB diagnosis presumed/confirmed to be M tb disease. TB was typically diagnosed on symptoms (prolonged cough/unexplained fever, poor growth, lethargy, and site-specific signs including persistent painless cervical lymphadenopathy), recent close exposure to a TB source case, reactive Mantoux skin test and suggestive imaging. Bacteriological tests were frequently requested on relevant samples: Auramine smear microscopy followed by the automated Mycobacterial Growth Indicator Tube culture method (MGIT; Becton-Dickinson, Sparks, Maryland, USA). Programmatically, TB is classified as pulmonary or extrapulmonary; if both are present, the disease is classified as pulmonary. We defined ‘severe disease’ as poor host control of M tb (uncontrolled disease), or the presence of complications.15 Timing of TB was categorised as ‘pre-ART’ if diagnosed before or on the day of ART initiation, ART-associated TB immune reconstitution or TB unmasking (grouped together as TB immune reconstitution syndrome (IRIS)) if diagnosed within 3 months of ART initiation and incident TB thereafter. We defined TB IRIS solely on timing post-ART based on: (1) insufficient data to apply the criteria by Meintjes et al16; (2) the high probability that TB within 3 months of ART represents an unmasking event rather than incident TB, as TB screening for a history of contact and symptoms with a low threshold for chest radiograph, and additional assessment is routine at TCH for children initiating ART and (3) the increasing probability that TB after 3 months would be an incident event due to high community transmission. Standard TB treatment followed WHO guidelines: daily isoniazid (INH), rifampicin and pyrazinamide for 2 months, followed by daily INH and rifampicin for 4 months. Ethambutol was added for extensive pulmonary or extrapulmonary TB (EPTB) and ethionamide for TB meningitis.17 In 2010, TB drug dosages were increased to new WHO recommendations.18

IPT was defined as a course of INH (any treatment duration) given for recent exposure to infectious TB case or for a positive tuberculin test, after excluding active TB.

‘Baseline’ characteristics denote clinical parameters at ART initiation, including age, gender, weight-for-age-Z-score (WAZ), CD4 count and percentage, and viral load (VL). We followed WHO immunological staging guidelines.19 WAZ was calculated using the WHO child growth standards.20

Study exit points included date of death, last date seen if lost to follow-up, and date of transfer to a different healthcare facility for HIV care. Children also exited the study if ART was suspended for >2 months for non-adherence, or if they entered an IPT trial.

Statistical analysis

The baseline and demographic characteristics at ART initiation were displayed by timing of TB in relation to ART initiation using frequencies and percentages for categorical and medians and IQR for continuous data. Categorical variables were analysed using χ2 or Fisher's exact tests when the frequencies were small or with very low percentages. Continuous variables were analysed using t test, if normally distributed or the Wilcoxon rank-sum test if not normally distributed. The person years (py) and median duration of time on study were calculated by exit reason, using ART initiation and study exit dates. Incidence was calculated for death, TB IRIS and incident TB, as the number of new cases per 100 py. Logistic regression models were generated for death, TB IRIS and incident TB. Age (<1 vs ≥1 year), sex, WAZ (<−2 vs ≥−2), VL, severe immunosuppression and prior TB were assessed for each outcome. Covariates with a univariable p value <0.20 were included in the multivariable analysis. Variables of clinical importance were included in the model regardless of the univariable p value. Adjusted ORs (AOR) were reported for covariates that were, or nearly were, associated with the outcome.


We identified 494 eligible children (figure 1). Total follow-up time was 855.54 years, median 0.89 (IQR, 0.28–2.49) years. Only 139 (28%) children remained in care at TCH until study completion (table 1), primarily due to an active process of down-referral to primary care adopted by the hospital (245 children were transferred during the study).

Table 1

Patient follow-up time by all patients, and by those lost to follow-up, transferred out and died (n=494)

Figure 1

Study cohort. TCH, Tygerberg Children's Hospital; ART, antiretroviral therapy; IPT, isoniazid preventive therapy; IRIS TB, immune reconstitution TB or unmasking TB. Note: Study period denotes 1 January 2003–30 June 2012.

In 161 (33%) children, TB occurred before ART initiation. After ART initiation 55 (11%) had TB: 23 had TB IRIS (4.7%) and 32 had incident TB (6.5%). TB IRIS incidence was 21.9/100 py of follow-up, with median time to diagnosis 1.0 (IQR 0.5–2.1) months. Incidence of new (non-IRIS, incident) TB was 3.9/100 py, with a median time to diagnosis 11.9 (IQR 7.2–20.8) months (figure 2). Seven (12.7%) children with post-ART TB had previous TB, seven (13%) had prior IPT and one had prior TB treatment and IPT. Table 2 reflects cohort baseline characteristics. Children with TB IRIS and those with incident TB did not differ by baseline characteristics (table 3). The only variable approaching significance as a risk factor for TB IRIS was starting ART at <1 year of age (AOR 8.65, 95%; CI 0.88 to 84.9); p=0.064), after controlling for prior TB, severe immunosuppression and WAZ. A history of prior TB was not associated with an increased risk of subsequent TB (AOR 3.27; 95% CI 0.30 to 36.07, p=0.33).

Table 2

Baseline demographic and clinical characteristics for all study patients

Table 3

Demographic and clinical characteristics of children with post-ART TB (TB IRIS and incident TB), TB prior to ART and children who had no TB during the observation period

Figure 2

Frequency distribution of time to incident tuberculosis (TB).

Children who did not develop TB initiated ART earlier (median age 5.7 (IQR 3.4–9.9) vs 9.7 (IQR 5.9–15.3) months, p<0.001), and were less severely immune suppressed (median CD4 19.5% vs 15.4%, p=0.003) compared to children with TB; baseline WAZ and VL were similar.

Nine of 46 (20%) post-ART TB episodes where culture was attempted were bacteriologically confirmed. EPTB was documented in 6/21 (29%) TB IRIS cases and 4/32 (13%) incident TB cases. In two cases, clinical records were insufficient to determine whether PTB or EPTB was present. Four EPTB IRIS and all incident EPTB cases were severe15: two meningitis, one disseminated and one pericarditis among IRIS cases; one each of miliary, meningitis, pericardial and spinal TB among incident cases.

During the observation period, 51 children (10.3%) died (mortality rate 5.96/100 py). Median time to death was 1.0 month (table 1) after ART initiation. The highest mortality was in the TB IRIS group (10.8 deaths/100 py) (table 4). There was no difference in mortality between culture-confirmed and culture-negative TB IRIS (p=0.37). No deaths were recorded after 2009. Most deaths were due to advanced HIV and other infections. Two deaths directly attributed to TB were in the TB IRIS group. Controlling for age, sex, severe immunosuppression and prior TB, WAZ<−2 significantly predicted death (AOR 6.37; 95% CI 2.43 to 16.69; p<0.001), while age, sex, immune status and prior TB did not.

Table 4

Deaths during study period

We recorded 127 courses of IPT: 10 before, 115 after and 2 at ART initiation; 19 children had >1 course. A 6-month course was completed in 66 (52%). Median duration of IPT was 5.5 (IQR 3.1–6.4) months. Eight children initially receiving IPT started TB treatment after a median 48.5 days (IQR: 31–64.5) of IPT. Only two children with culture-confirmed TB (both incident TB) had prior IPT: both were drug-susceptible.


We describe a cohort of young children who started ART at <2 years of age under routine conditions. Despite a median age at ART initiation of <8 months, most children were severely immune suppressed and malnourished. In many impoverished communities in Cape Town, transmission of TB is extremely high; HIV-infected infants are at highest risk of infection, with an estimated incidence of culture-confirmed TB 24-fold higher than their HIV-uninfected peers, and an overall incidence >1.5%.21 In this study, one-third of children had already had TB before ART, emphasising their vulnerability and the high TB exposure. Incidence of TB on ART remained two to fourfold higher than that of HIV-uninfected children in our community.22–24 Eight children (3.8%) had >1 TB episode: although lower than the recurrence rate in pre-ART times,25 ,26 it is high considering a median follow-up time of only 0.89 years.

A high rate of TB IRIS was seen despite active TB screening at ART initiation. However, few TB IRIS cases were bacteriologically confirmed and over-diagnosis is possible. Many infants starting ART with advanced immune suppression develop severe immune reconstitution phenomena or are critically ill8 ,27: multiple coinfections can occur and TB treatment is frequently initiated without diagnostic confirmation because of the high risk of death from untreated TB.28 Data on other coinfections were scanty and were therefore not included in this report. The highest mortality observed in children with TB IRIS is likely to be due to their serious clinical condition even before ART initiation. It is difficult to determine the impact of TB on death in our study. Although death was as frequent in children with TB (24/216-11%) as those without TB (28/286-10%), this finding could be biased. The entry point in this analysis was time of starting ART: some children with TB who died before ART initiation would be excluded, while some prior TB episodes may not have been documented. Additionally, many children with post-ART TB exited the study before completing TB treatment, resulting in incomplete outcome data. We could therefore only include TB prior to ART as a dependent variable in the death analysis. Severe malnutrition, as previously documented29–31 was the only significant predictor of death. While no deaths were reported after 2009, a decentralised model of HIV care was then actively being adopted in Cape Town, with children being transferred out to community clinics as soon as possible after ART initiation. We did not collect data after transfer, and may have missed deaths after study exit. However, local data suggest that mortality among HIV-infected children followed at community clinics is comparable with those in tertiary care (Morsheimer, (in progress)). Additionally, children who initiated ART and died during initial hospitalisation may not have been in the clinic database.

Our study is the first to describe the spectrum of TB in young children after ART initiation. We observed a higher proportion of EPTB among children with TB IRIS (29%) than in children with incident TB (13%). Most EPTB was severe or disseminated,15 and three children with IRIS died. The occurrence of TB adenitis may be low because many children with concurrent pulmonary TB are routinely documented as ‘PTB’ only. We could not classify the severity of pulmonary disease as many chest radiographs were unavailable for review. The substantial morbidity associated with severe TB in children with depressed immunity27 highlights the need for safer, more effective preventive strategies. During this study, approximately 25% of the cohort was prescribed IPT, mostly after ART initiation. This reflects regular contact with the ID clinic and ongoing TB surveillance, and highlights again the high TB exposure burden in our communities. Our study was not designed to assess the impact of IPT on subsequent TB, but it is notable that no children with TB IRIS received IPT, and only one had prior TB. The potential role of routine IPT to HIV-infected infants from high TB burden settings around the time of ART initiation for preventing TB IRIS needs further prospective study. IPT adherence was better than previously described in this setting,32 probably because of close follow-up.

Finally, a relatively low proportion (20%) had culture-confirmed TB. Diagnosing TB in young HIV-infected children is challenging.33 ,34 However, confirming a diagnosis is crucial, as underdiagnosis, a missed alternative diagnosis and unnecessary treatment can have serious consequences.

Limitations of our retrospective study include incomplete documentation of routine data. A substantial proportion of missing data may have introduced bias in the multivariable analysis. However, our results agree with similar studies.10 ,29 We had insufficient information to report on coinfections. However, the severe spectrum of EPTB in this cohort is a novel finding supporting the need for improved prevention and diagnostic strategies targeting young HIV-infected children in high TB-burden settings. Whether these findings are generalisable to other settings will require further studies to document the risk and disease spectrum of TB in children on ART.



  • This manuscript was written in partial fulfilment of a Master's Degree in Clinical Epidemiology, Division of Community Health, Stellenbosch University.

  • Contributors EW: conceived the research idea, developed a data collection tool, performed initial data collection, planned the analyses, interpreted results and was the primary author of the manuscript. JD: assisted with data collection, background literature review and initial drafting of the manuscript. HRD: assisted with planning of the analyses, performed the analyses on the statistical software, assisted with interpretation of results, contributed to the final manuscript and designed the tables. ACH and SSvW: assisted with planning of the analyses, contributed to the interpretation of results, and assisted with drafting of the manuscript. MFC: contributed to the interpretation of results, and assisted with drafting of the manuscript. HR: supervised the study, assisted with the formulation of the research idea, assisted with planning of the analyses, contributed to the interpretation of results, and assisted with drafting of the manuscript. All authors read and approved the final draft of the manuscript.

  • Competing interests None.

  • Ethics approval Ethical approval was granted by the Health Research Ethics Committee, Stellenbosch University, reference N06/08/167.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Unpublished data on the univariable and multivariable logistic regression analyses are available for interested researchers from the corresponding author.

Linked Articles