Risk of tuberculosis infection in young children exposed to multidrug-resistant tuberculosis in the TB-CHAMP multi-site randomised controlled trial

Susan E Purchase, Joanna Brigden, James A Seddon, Neil A Martinson, Lee Fairlie, Suzanne Staples, Thomas Wilkinson, Trinh Duong, H Simon Schaaf, Anneke C Hesseling, Risk of Tuberculosis Infection in Young Children Exposed to Multidrug-resistant Tuberculosis in the TB-CHAMP Multi-site Randomized Controlled Trial,Clinical Infectious Diseases, 2025;, ciaf284,https://doi.org/10.1093/cid/ciaf284 Young children have a high risk of developing tuberculosis (TB) disease following infection withMycobacterium tuberculosisin the absence of preventive treatment. Infection prevalence and risk factors for infection impact delivery of prevention strategies. We aimed to determine the prevalence of infection in child household contacts aged <5 years exposed to adults with confirmed pulmonary multidrug-resistant (MDR)-TB and to determine risk factors for infection. TB-CHAMP was a trial of MDR-TB prevention that recruited children younger than age 5 years, regardless ofM. tuberculosisinfection status. All children enrolled had an interferon-gamma release assay (IGRA) at baseline. We describedM. tuberculosisinfection prevalence, developed directed acyclic graphs to clarify causal relationships, and used modified Poisson regression models to assess the relationship between risk factors and IGRA positivity. Of 785 included children, 160 (20.4%) had a positive IGRA. Duration of cough and drug misuse in the index patient, age of the child, relationship between the child and the index patient, and study site were significantly associated with risk of infection. The prevalence of infection was lower than observed in previous studies. This may be related to improved diagnosis and treatment of MDR-TB in the study setting and/or test limitations and has implications for TB preventive treatment. When considering TB preventive treatment for child contacts, healthcare providers should be especially concerned about any young child exposed to an adult index patient who is his/her parent/primary caregiver, has a chronic cough, and/or a history of drug misuse. Young children have a high risk of developing tuberculosis (TB) disease following infection withMycobacterium tuberculosisin the absence of TB preventive treatment (TPT) [1]. Multidrug-resistant (MDR)-TB, caused byM. tuberculosisresistant to isoniazid and rifampicin, is threatening global TB control [2]. An estimated 2 million children younger than age 15 years are currently infected with MDR-M. tuberculosisand approximately 30 000 develop MDR-TB disease each year [3]. To design effective strategies to prevent MDR-TB, it is important to understand the risk of infection as measured by current diagnostic tools, and the factors that determine this risk. There is currently no diagnostic gold standard to measureM. tuberculosisinfection. Various commercially approved tests of infection are available, including tuberculin skin tests, interferon-gamma release assays (IGRAs) and tuberculosis antigen-based skin tests. Although children with a negative test of infection may have a higher risk of disease progression than previously appreciated [4], children with a positive test ofM. tuberculosisinfection have a substantially higher 2-year cumulative TB disease incidence than children with a negative result [1]. Estimates of the prevalence ofM. tuberculosisinfection in household contacts (HHCs) of adults with infectious TB disease vary greatly. Estimates of prevalence of infection for HHCs younger than age 5 years of drug-susceptible TB in low- and middle-income countries vary from 16% to 53% [5–7]. The prevalence of infection in HHCs younger than age 5 years of age exposed to drug-resistant TB seems comparable, with prevalence varying from 44% to 59% [8–11]. The risk ofM. tuberculosisinfection in close child contacts has been positively correlated with factors relating to the child, the index patient (IP) and the environment [10,12–15]. However, much of the work on risk ofM. tuberculosisinfection was completed more than a decade ago, with subsequent changes in the diagnosis and treatment of rifampicin-resistant (RR)/MDR-TB [16,17], scale-up of effective human immunodeficiency virus (HIV) test-and-treat strategies, and increased availability of more acceptable antiretroviral regimens; and, in South Africa, access to rapid molecular testing for TB. It is therefore important to understand the contemporary risk of having a positive test of infection [1] and the factors that modulate this risk in child household RR/MDR-TB contacts in settings with a high burden of TB and HIV. TB-CHAMP was a trial of MDR-TB prevention conducted in South Africa that recruited children younger than age 5 years of age regardless ofM. tuberculosisinfection status, and children aged 5–17 years with a positive IGRA or living with HIV. The trial investigated the efficacy and safety of 24 weeks of daily levofloxacin versus placebo. We estimated the prevalence ofM. tuberculosisinfection in child HHCs to be 40%, based on previous observational South African studies. The observed underlying incidence of TB disease in the control arm in TB-CHAMP was less than half of that expected, emphasizing the importance of understanding infection dynamics in children exposed to MDR-TB. We aimed to determine the prevalence ofM. tuberculosisinfection in child HHCs aged <5 years exposed to adults with infectious pulmonary MDR-TB in the household and to determine risk factors for infection in these child contacts, in this large prevention trial. TB-CHAMP was conducted at 5 sites across 6 provinces in South Africa, all serving poorly resourced communities. The trial was conducted at the Desmond Tutu TB Centre (Department of Paediatrics and Child Health, Stellenbosch University, Cape Town, Western Cape), the Perinatal HIV Research Unit (Matlosana Wits Health Consortium, Matlosana, North-West province), the Wits RHI Shandukani Research Centre (Johannesburg, Gauteng), and the Tuberculosis & HIV Investigative Network (Pietermaritzburg and Durban, KwaZulu-Natal). The fifth site that opened to accrual in the final year of the trial did not enroll children aged <5 years and is excluded from this analysis. South Africa has a high-burden country of TB, HIV-associated TB, and MDR-TB [2]. TB-CHAMP was a cluster-randomized, double-blind, placebo-controlled MDR-TB prevention trial, comparing levofloxacin (15–20 mg/kg) with placebo. Households were randomized 1:1 to either levofloxacin or placebo taken daily for 24 weeks. Follow-up was for 72 weeks in total [18]. The trial enrolled children between 27 September 2017 and 29 July 2022. Children <5 years were eligible regardless of their IGRA status at screening. All children enrolled had an IGRA (QuantiFERON-Gold Plus: Qiagen) collected at baseline before study drug was initiated. IGRAs were collected and transported according to manufacturer's specifications, then analyzed at a single central certified trial laboratory (BARC Laboratories, South Africa).M. tuberculosisinfection was defined as being QuantiFERON-Gold Plus positive, based on standard manufacturer guidelines. Adult IPs were identified following a routine diagnosis of confirmed pulmonary MDR-TB and recruited if there was at least 1 child aged <5 years living in the same household in the previous 6 months. Children aged <5 years were recruited if exposure to the MDR-TB IP had been substantial in the preceding 6 months. Before enrollment, children were evaluated for prevalent TB disease with history, examination, and plain-film chest radiography, and respiratory sampling in the case of symptoms or abnormal chest radiographs. Only children in whom TB disease had been confidently excluded were enrolled. All randomized participants aged <5 years with documented IGRA status at baseline were included in this analysis. At screening, demographic, medical history, and substance use data were collected for all IPs, and demographic, medical history, TB exposure history, and clinical data were collected for all child participants. Socioeconomic data for each household were systematically collected. The prevalence of IGRA positivity at baseline in child participants was described and compared between sites. Variables for analysis were identified based on biological plausibility and findings from prior studies [12,14,15]. These included factors relating to IPs (age, sex, duration of TB symptoms including cough, smoking, HIV status), child participants (age, sex, weight-for-age [WFA]z-score, duration of exposure to MDR-TB, relationship to the IP, HIV status, previous TPT or antibiotic use, recent hospitalization), household characteristics (number of household members, number of rooms, socioeconomic indicators) and study site. A per-household socioeconomic status (SES) score was developed by applying the published principle component analysis coefficients of questions in the South African Demographic and Health Survey 2016 [19], to applicable household characteristics of trial households. The TB-CHAMP household SES index was then translated to quintiles of Demographic and Health Survey households to enable a comparison of the SES of TB-CHAMP households to the general South African population, with the first quintile representing the lowest SES score (seeSupplementary Appendix 2, Table 2). Directed acyclic graphs were drawn to help clarify causal relationships and identify a priori confounders and mediators [20], which were adjusted for in multivariable models (Supplementary Figure 1). The proportion of participants who were IGRA positive was described by each factor. Modified Poisson regression models were used to assess the relationship with IGRA positivity in univariable and multivariable analyses, with robust standard errors derived using a clustered sandwich estimator to allow for household clustering [21]. In the multivariable analysis of each potential risk factor, the model was adjusted for a priori confounders and study site (Supplementary Figure 1). Analyses were performed with Stata version 16.0 (StataCorp. 2019. Stata Statistical Software: Release 16. StataCorp LLC, College Station, TX). Of 839 children enrolled, 815 (97.1%) had an IGRA at baseline and 785 (93.6%) were included in analysis. Of children excluded, 24 had indeterminate IGRA results and 6 were late screen failures (initially enrolled but later found to have had TB at baseline) (Figure 1). Baseline characteristics of child HHCs (Table 1) were relatively uniform across the 4 sites. Overall, 50% were girls; median age was 2.5 years (interquartile range: 1.3–3.8); 42% of children were enrolled during and after the COVID-19 pandemic. One percent of children had HIV, 36% were HIV-exposed and uninfected, and 94% of children had received bacillus Calmette-Guérin (BCG) vaccine at birth (indicated by a scar or vaccination card). The median WFA z-score was −0.4 (interquartile range: −1.2 to 0.3). Overall, only 26% of households fell into the poorest 2 (1st, 2nd) SES score quintiles. For TB exposure history of child participants, seeSupplementary Table 1. Flow of patients included in the analysis. Abbreviation: IGRA, interferon-gamma release assay. Baseline Characteristics and IGRA Results of Child Participants Aged <5 y With Known Baseline IGRA Status, by Study Site Abbreviations: HIV, human immunodeficiency virus; IGRA, interferon gamma release assay; IQR, interquartile range; SES, socioeconomic status. aAll treated for drug-sensitive tuberculosis, apart from 1 child with unknown relevant information. bStandardized to the World Health Organization reference. cSES status score derived from South Africa Demographic and Health Survey figures. SES quintiles range from the poorest (1st) to the wealthiest (5th). There were no households in the 5th quintile. dFor a small number of children aged <5 y without a test result at screening, test result up to wk 4 postrandomization was used. Baseline Characteristics and IGRA Results of Child Participants Aged <5 y With Known Baseline IGRA Status, by Study Site Abbreviations: HIV, human immunodeficiency virus; IGRA, interferon gamma release assay; IQR, interquartile range; SES, socioeconomic status. aAll treated for drug-sensitive tuberculosis, apart from 1 child with unknown relevant information. bStandardized to the World Health Organization reference. cSES status score derived from South Africa Demographic and Health Survey figures. SES quintiles range from the poorest (1st) to the wealthiest (5th). There were no households in the 5th quintile. dFor a small number of children aged <5 y without a test result at screening, test result up to wk 4 postrandomization was used. Of 785 children included, 160 (20.4%) had a positive IGRA result (Table 1). IGRA results varied between sites, from 13% positivity (Matlosana, periurban) to 26% (Desmond Tutu TB Centre, urban, densely populated). Of the 16 children younger than age 5 years who developed TB disease during follow-up, baseline IGRA status was positive in 6 and negative in 10; 6/160 (3.8%) of IGRA-positive children developed TB disease versus 10/625 (1.6%) of children with negative IGRA at baseline. Directed acyclic diagrams were drawn to identify a priori the confounding variables for estimating causal effects (Figure 2). Relationships between variables were found to be complex and interconnected. Some variables (infectiousness of the IP and the immune status of the child) could not be directly measured, but duration of cough in adults and presence of chronic illness/HIV/poor nutrition in child participants were used as proxies. Directed acyclic graphs of the characteristics of the index patient (IP), the household and child household contact (HHC). Abbreviations: HHC, household contact; IP, index patient; SES, socioeconomic status; TB, tuberculosis. In the univariable model, IPs’ age and sex were associated with prevalence of infection (Table 2). In the multivariable models, only drug misuse in the past 6 months (risk ratio [RR], 1.49; 95% confidence interval [CI]: 1.01–2.21;P= .047) and longer duration of cough (>4 weeks; RR, 1.65; 95% CI: 1.20–2.26; <4 weeks; RR, 1.06; 95% CI: .73–1.56;P= .006, when compared with no cough) remained associated. No relationship was found between alcohol use, smoking, or HIV status of the IP and prevalence of infection in the child. Forest plot showing variables associated with risk of infection on multivariate analysis. Abbreviations: CI, confidence interval; IP, index patient. Univariate and Multivariate Analysis of Factors Affecting Risk of TB Infection for Participants Aged <5 y With Known Baseline IGRA Status Abbreviations: ART, antiretroviral therapy; BCG, bacillus Calmette–Guérin; CI, confidence interval; HIV, human immunodeficiency virus; IP, index patient; SES, socioeconomic status; TB, tuberculosis. Univariate and Multivariate Analysis of Factors Affecting Risk of TB Infection for Participants Aged <5 y With Known Baseline IGRA Status Abbreviations: ART, antiretroviral therapy; BCG, bacillus Calmette–Guérin; CI, confidence interval; HIV, human immunodeficiency virus; IP, index patient; SES, socioeconomic status; TB, tuberculosis. In univariable and multivariable analyses, only increasing age was linked to prevalence of infection in children (multivariable: 1 to <3 years; RR, 1.54; 95% CI: .99–2.40; to <5 years; RR, 1.80; CI: 1.16–2.77;P= .030, when compared with <1 year). Sex, WFAz-score, HIV status, previous TB treatment, and BCG immunization status showed no association. Univariable analysis showed a strong association between the prevalence of infection and sleeping in the same room/bed as the IP as well as the number of hours of daily exposure. This was not maintained in the multivariable model. The relationship of the IP to the child and whether the IP was the primary caregiver were strongly associated in univariable analysis, with the effect for relationship remaining significant in multivariable analysis. The highest risk in multivariable analysis was seen when the IP was the father of the child (father: RR, 1.58; 95% CI: .82–3.05;P= .001—IP is the mother is used as reference group). There was substantially higher risk of infection if the IP was either the mother or father, when compared with other family/nonfamily members. Study site was significantly related to the prevalence of infection in univariable and multivariable analysis, with child contacts from the Cape Town site having the highest prevalence of infection, and Matlosana the lowest (Matlosana: RR, 0.52; 95% CI: .35–.79;P= .012 when compared with Desmond Tutu TB Centre). Overcrowding and household SES score did not influence prevalence of infection in these child contacts. We characterized the prevalence ofM. tuberculosisinfection in young children with MDR-TB exposure across diverse settings in South Africa. The observed prevalence ofM. tuberculosisinfection of 20% was lower than that observed in previous studies [8–11]. Factors that had a significant impact on prevalence of infection included duration of cough and drug misuse in the IP, the age of the child, the relationship between the child and the IP, and the trial site where the child was enrolled. There are several possible reasons to explain the lower-than-expectedM. tuberculosisinfection prevalence observed. South Africa was the first country to roll out Xpert MTB/RIF in 2011 and Xpert MTB/RIF Ultra in 2017 [22]. As these tools became increasingly available, it is likely that patients with RR/MDR-TB were diagnosed more rapidly, reducing duration of infectiousness. Since 2018, there has been widespread roll out of more effective MDR-TB treatment regimens (including bedaquiline and linezolid) in South Africa [16], which rapidly render IPs noninfectious [23,24]. Contact management strategies for HHCs of adults with TB disease have also improved in South Africa over the past decade [25]. These changes would reduce the duration of exposure for HHCs and thus their risk of infection. All children screened for the trial were rigorously investigated for TB disease at baseline, and those considered to have TB disease excluded. Thus, the prevalence ofM. tuberculosisinfection reported here is for well children only. Several of the studies reporting higher prevalence ofM. tuberculosissensitization in child MDR-TB contacts included children with infection and disease [10,11,14], although this is unlikely to account for the markedly lower infection prevalence we observed. Factors relating to the IP (age, chest radiograph disease severity, acid-fast bacilli smear-positive status, alcohol use, smoking), the child (age, sex, immune status, BCG vaccination status, and presence of other medical conditions), the level ofM. tuberculosisexposure (being the parent or sleeping in the same room, duration of exposure), and environment (SES, presence of overcrowding) [7,12,26,27] have all been correlated with increased prevalence ofM. tuberculosisinfection in children in previous studies. In our study, increased duration of cough and drug misuse in the IP, older age of the child and a close relationship between the child and the IP showed significant association on multivariable analysis, after controlling for potential confounders. Although univariable analysis showed a strong association between prevalence of infection and sleeping in the same room/bed as the IP and the hours of daily exposure, this was not maintained in multivariable analysis, likely due to the strong influence of relationship between the IP and child contact on the level of exposure. Univariable analysis also identified mothers who were IPs as posing the highest risk to child contacts, but the risk was higher for fathers in multivariable analysis. There is likely to be collinearity between some of the factors relating to exposure of the child to the IP, making it challenging to assess their associations with TB infection status individually. Infection rates varied considerably between sites, likely due to differences in risk factors (such as TB and HIV epidemiology, socioeconomic factors, healthcare practices, recruiting strategies, climate and genetic differences) that were not accounted for in our models. There appears to be an association between sites and the household SES score, which varied substantially between sites. In multivariable analysis, however, the standard errors of the estimates for the SES quintiles (as well as the corresponding globalP-value) were similar in the models with and without adjustment for sites, suggesting collinearity was not an issue here. Our finding that relative SES was not associated with young children's TB infection status is surprising. It is possible that there is threshold level of deprivation below which further deprivation does not incur a greater risk for infection. Additionally, the single numerical SES score derived from the national South African population may not be sufficiently sensitive to reflect specific SES factors associated with increased infection risk. Importantly, this finding does not indicate that SES has no impact on the risk ofM. tuberculosisinfection, but that further work is required to better reflect the complexities of SES in relation to clinical outcomes of exposure of a young child to a household IP with TB. Of note, of children who developed TB disease, more than half were IGRA negative at baseline. IGRAs are limited by reduced sensitivity in children younger than age 5 years, low predictive value for progression to TB disease, and multiple sources of variability if repeated [4,20,21]. IGRA results may have been negative due to very recent infection. There are several new tests ofM. tuberculosisinfection in the pipeline, but none are yet able to distinguish the continuum ofM. tuberculosisinfection or predict TB disease [22]. Biomarker-guided TPT is showing promise, with some signatures able to predict risk of disease progression [23]. More sensitive tools including novel biomarkers are needed to identify prevalence of infection and future disease progression to better guide prevention strategies. Despite the robust sample size and collection of detailed data, our analysis was cross-sectional and limited to children younger than age 5 years. Repeating IGRAs may have yielded test conversion. Given the modest number of incident TB end points, stratified analysis by IGRA status was limited, and we could not assess the predictive utility of IGRA for incident TB disease. We used routine microbiological data reported from the national TB program to determine eligibility of IPs, but these results were not captured, and we did not complete additional microbiological testing or chest radiography in IPs. We were thus unable to assess additional measures of infectiousness of IPs beyond duration of cough. Child contacts who were screened out with likely TB disease were excluded from analysis. Thus, children included in this trial do not represent all children exposed to adults with MDR-TB in the household. Finally, the SES score we used summarized complex SES information into a single score and may not reflect the complex dynamic of different dimensions of SES on TB infection risk. The World Health Organization now recommends levofloxacin as TPT in all close MDR-TB contacts once TB disease has been excluded, regardless ofM. tuberculosisinfection status [24]. Children younger than age 5 years remain a population of high priority given their risk of disease progression and severe forms of TB. The low prevalence ofM. tuberculosisinfection in child MDR-TB contacts seen in TB-CHAMP has implications. As infection is a prerequisite for disease, low infection prevalence implies less progression to TB disease. This also implies decreased absolute efficacy of TPT in a strategy where TPT is recommended for all contacts. That many IGRA-negative children developed TB disease in the trial suggests that the use of IGRA to predict disease progression is limited and better tests are needed to identify those at highest risk. Although TPT is now recommended for all MDR-TB contacts, children younger than age 5 years exposed to an infectious adult who is the parent/primary caregiver are at especially high risk of developing TB disease and should be prioritized in TB prevention programs. Supplementary materialsare available atClinical Infectious Diseasesonline. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Author Contributions. S. P. wrote the paper with support from J. A. S., A. C. H., H. S. S., T. D., and J. B. led the statistical analysis, with support and oversight from T. D. All authors reviewed the submitted and final version of the manuscript and approved of it. Acknowledgments. The authors would like to acknowledge the TB-CHAMP teams at all 4 sites and all participants and their caregivers. We would also like to thank our community advisory boards and local TB services. Data availability. Deidentified data set is available upon request, to researchers with approval for the proposed use of the data, and to policymakers. Ethical considerations. The trial was approved by the Health Research Ethics Committee of Stellenbosch University (M16/02/009) and the University of the Witwatersrand (160409), the South African Health Products Regulatory Agency (20160128), and the South African Department of Health (DOH-27-0117-5309) and was registered in the ISRCTN registry (ISRCTN92634082). Informed consent was provided by all IPs and participants’ parents or legal guardians. Data were stored using unique anonymized participant identifiers. Financial support.The TB-CHAMP trial was supported by UNITAID, through the BENEFIT Kids project grant [grant number 2019-36-SUN-MDR] to Stellenbosch University. UNITAID accelerates access to innovative health products and lays the foundations for their scale-up by countries and partners. This trial was also funded by a JGHT trial grant to Stellenbosch University [grant number MR/M007340/1], supported by the Department of Health and Social Care (DHSC), the Foreign, Commonwealth & Development Office (FCDO), the Global Challenges Research Fund (GCRF), the Medical Research Council (United Kingdom), and Wellcome. This UK-funded award is part of the EDCTP2 programme supported by the European Union. Additional funding was provided by the South African Medical Research Council for the TB-CHAMP trial grant to Stellenbosch University and the South African National Research Foundation (NRF) to A. C. H. (SARCHi chair). The Medical Research Council Clinical Trials Unit at University College London received core support from the U.K. Medical Research Council [grant numbers MC_UU_00004/04, MC_UU_00004/09]. The funders had no rule in the design, implementation and dissemination of these results. No funding was provided by Macleod's Pharmaceuticals. Study drug was procured for the trial. S. E. P. was supported by funding from the South African Medical Research Council through its Division of Research Capacity Development under the Bongani Mayosi National Health Scholars Programme from funding received from the Public Health Enhancement Fund/South African National Department of Health. The content hereof is the sole responsibility of the authors and does not necessarily represent the official views of the SAMRC. Google Scholar WorldCat Google Scholar Google Preview WorldCat Google Scholar WorldCat Google Scholar WorldCat Crossref Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar Google Preview WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar WorldCat Google Scholar Google Preview WorldCat Google Scholar WorldCat Google Scholar WorldCat Potential conflicts of interest.The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.