Bulletin of the World Health Organization

Prevention of tuberculosis in household members: estimates of children eligible for treatment

Yohhei Hamada a, Philippe Glaziou a, Charalambos Sismanidis a & Haileyesus Getahun a

a. Global Tuberculosis Programme, World Health Organization, 20 avenue Appia, 1211 Geneva 27, Switzerland.

Correspondence to Yohhei Hamada (email: yohei.hamada0@gmail.com).

(Submitted: 23 June 2018 – Revised version received: 24 April 2019 – Accepted: 06 May 2019 – Published online: 28 May 2019.)

Bulletin of the World Health Organization 2019;97:534-547D. doi: http://dx.doi.org/10.2471/BLT.18.218651

Introduction

The management of latent tuberculosis infection is a critical component of the World Health Organization’s (WHO’s) End TB Strategy. Given that between a quarter and a third of the global population is estimated to be infected with Mycobacteria tuberculosis, 13 the Strategy’s ambitious targets and the United Nations’ Sustainable Development Goals cannot be achieved without tackling the reservoir of latent infection.4 The risk of progression from tuberculosis infection to active disease is particularly high in young children, who are also at the greatest risk of severe and disseminated disease.5 As a result, treatment of tuberculosis infection (i.e. tuberculosis preventive treatment) is strongly recommended for children younger than 5 years who are household contacts of people with bacteriologically confirmed pulmonary tuberculosis.6 Accordingly, coverage of tuberculosis preventive treatment is one of the key indicators used to monitor the implementation of the End TB Strategy.7 In 2018, world leaders committed to providing 4 million child household contacts younger than 5 years with tuberculosis preventive treatment by 2022.8

A recent survey of policy and practice on latent tuberculosis infection in countries with a low tuberculosis burden and in African countries found that many lacked recording and reporting systems for infection.9,10 In 2016, WHO started collecting data on the number of children younger than 5 years globally who were household contacts of people with pulmonary tuberculosis and who had started tuberculosis preventive treatment.11 Although 118 countries, including 16 of the 30 countries with a high tuberculosis burden, reported data in 2017,11 there was a lack of clearly defined denominators for assessing coverage of preventive treatment, which makes planning and monitoring difficult.12

Consequently, the aim of this study was to use tuberculosis notification data from 2017 to estimate of the number of children younger than 5 years in individual countries who were household contacts of people with pulmonary tuberculosis and who were eligible for tuberculosis preventive treatment. This information should help countries implement and monitor preventive treatment.

Methods

Countries with a low tuberculosis burden comprised the 113 high-income or upper-middle-income countries in which the estimated annual incidence of tuberculosis disease in 2015 was fewer than 100 cases per 100 000 population, WHO’s 2015 guidelines on the management of latent tuberculosis infection are intended primarily for these countries.13,14 Countries with 100 or more cases per 100 000 population were regarded as having a high tuberculosis burden.

In countries with a high tuberculosis burden, the number of children eligible for tuberculosis preventive treatment was defined as the number younger than 5 years who are household contacts (hereafter referred to as child household contacts) of people with bacteriologically confirmed pulmonary tuberculosis and who do not themselves have active tuberculosis, regardless of whether they have a confirmed tuberculosis infection (in accordance with WHO guidelines on the management of tuberculosis in children).5 In countries with a low tuberculosis burden, the number of children eligible for tuberculosis preventive treatment was defined as the number of children younger than 5 years who are household contacts of people with bacteriologically confirmed pulmonary tuberculosis, who do not themselves have active tuberculosis and who have a confirmed tuberculosis infection, as indicated by a positive result on a standard tuberculin skin test or an interferon-gamma release assay. Consequently, the number of child household contacts eligible for tuberculosis preventive treatment, N, was calculated using:(1)in countries with a high tuberculosis burden; and(2)in countries with a low tuberculosis burden; where n was the number of notified cases of bacteriologically confirmed, pulmonary tuberculosis in the country, C was the average number of active tuberculosis cases per household with an index case, h was the average household size, p was the proportion of the national population that was younger than 5 years, T was the proportion of child household contacts who had active tuberculosis, and L was the prevalence of a confirmed latent tuberculosis infection among child household contacts. For countries with a high tuberculosis burden, L was not included in the calculation because eligibility for tuberculosis preventive treatment did not depend on confirmation of infection. We did not estimate numbers for countries or territories with a population under 300 000.

(1).
(2).

Table 1 details how we derived values for the parameters in these two equations. From the literature, we obtained country-specific values of n and p for 2017, country-specific values of h for different years and a global estimate of T. To obtain a global value for L, we updated a recent systematic review and meta-analysis, and to obtain a global value for C, we carried out a new systematic review of the literature from 1 January 2005 to 11 November 2017.18 For both the updated and new systematic reviews, we used the reference list of Fox et al.’s systematic review,18 which included publications up until 1 October 2011, and supplemented it with papers subsequently published up until 11 November 2017. The new systematic review did not consider publications before 2005 because we judged that earlier publications would not reflect the current situation. The following search string was used in PubMed® for both reviews: (tuberculosis[Title] OR “tuberculosis”[MeSH Terms] OR “mycobacterium tuberculosis”[MeSH Terms] OR “tuberculosis, pulmonary”[MeSH Terms]) AND ((“contact$”[All Fields]) OR (“contact tracing”[MeSH Terms]) OR “disease outbreaks”[MeSH Terms] OR “contact*”[Title] OR “spread”[Title] OR “contact screen*”[All Fields] OR “contact tracing”[Title] OR “disease transmission”[All Fields] OR “case find*”[Title] OR (cluster*[Title] AND analys*[Title]) OR “household*”[All Fields] OR “household contact*”[All Fields] OR (“case finding”[All Fields]) OR (“casefinding”[All Fields]) OR “case detection”[All Fields]).

For the updated and new systematic reviews: (i) household contacts were defined as people living in the same household or people who satisfied the definition of a household contact in the original publication; (ii) an index case was defined as the first identified case of new or recurrent tuberculosis disease in a person of any age in a specific household or as defined in the original publication; (iii) a person was defined as having a tuberculosis infection if the induration 48 to 72 hours after a tuberculin skin test was 10 mm or greater or, if this information was not available, the person satisfied the definition of a tuberculosis infection in the original publication; and (iv) a prevalent tuberculosis case was defined as a case of active disease that was diagnosed at the baseline visit during the study or within 3 months of diagnosis of the index case.

To obtain a global value for L, we included studies in the updated systematic review that reported the prevalence of tuberculosis infection among child contacts in countries with an annual incidence of tuberculosis under 100 cases per 100 000 population at the time of the study, according to WHO estimates.15 If an appropriate WHO estimate was not available, we used estimates from the published literature. We also included studies that reported data on children up to 4 or 6 years of age. The reasons for excluding studies are listed in Fig. 1.

Fig. 1. Flowchart for the selection of studies on the prevalence of latent tuberculosis infection among child household contacts, countries with a low tuberculosis burden, worldwide, 1964–2017
Fig. 1. Flowchart for the selection of studies on the prevalence of latent tuberculosis infection among child household contacts, countries with a low tuberculosis burden, worldwide, 1964–2017
Notes: We defined a child household contact as a child younger than 5 years living in the same household as a person with active tuberculosis disease. A low tuberculosis burden was defined as fewer than 100 cases per 100 000 population.

To obtain a global value for C, we included studies in the new systematic review that reported the number of index tuberculosis cases, the number of household contacts and the number of prevalent active tuberculosis cases among household contacts. We excluded studies if: (i) data on contacts other than household contacts were included; (ii) the number of cases or household contacts was less than 10; (iii) only child contacts were included (this would have led to an underestimate of the number of active tuberculosis cases in the household); or (iv) the study was not published in English (Fig. 2).

Fig. 2. Flowchart for the selection of studies on active tuberculosis cases in households with an index case, worldwide, 2005–2017
Fig. 2. Flowchart for the selection of studies on active tuberculosis cases in households with an index case, worldwide, 2005–2017

One author screened all titles and abstracts for relevance and then reviewed the full text of all potentially eligible articles. For both reviews, we extracted information on the country’s name, the year of the study, the definitions of index cases and household contacts, and the number of household contacts. For the updated systematic review, we obtained information about the number of child household contacts with a confirmed latent tuberculosis infection, the tuberculin skin test cut-off criterion for infection in a child contact, the child’s bacillus Calmette–Guérin (BCG) vaccination status and the age of index cases. For the new systematic review, we extracted information on the age and number of index cases and the number of active tuberculosis cases among household contacts. In evaluating the quality of individual studies, we used a checklist modified from an existing tool to assess issues related to contact investigations and tuberculosis infection.19

Data analysis

The meta-analysis of the prevalence of a confirmed latent tuberculosis infection among child household contacts (L) was conducted using a logistic-normal random-effects model.20 In the primary analysis, we did not consider the different definitions of tuberculosis infection used in the studies. The heterogeneity of study findings was assessed by visual inspection of forest plots and from the results of likelihood-ratio tests. Potential sources of heterogeneity were investigated in subgroup analyses that considered the following factors: (i) whether the index case tested positive or negative on smear microscopy; (ii) the tuberculin skin test cut-off value (i.e. 10 mm or more versus other values); (iii) the year of study publication (i.e. before 2000 or later); (iv) the country’s income status (i.e. whether high- or upper-middle-income);21 and (v) BCG vaccination coverage.

The average number of active tuberculosis cases per household with an index case (C) was estimated as follows. For each study, the average number of active tuberculosis cases among contacts in each household was calculated by dividing the number of prevalent active tuberculosis cases among household contacts by the number of index cases, which was assumed to be equal to the number of households. Data were pooled using mixed-effects, Poisson regression models. Subsequently, the average number of tuberculosis cases per household was calculated as the pooled average number of tuberculosis cases among contacts in each household plus one to account for the index case. The heterogeneity of study findings was assessed by visual inspection of forest plots and the effect of the national tuberculosis burden on estimates was assessed in a subgroup analysis. We also conducted a sensitivity analysis by excluding an outlier value for the number of tuberculosis cases per household to assess its influence on the pooled estimate.

We did not evaluate publication bias using statistical tests (e.g. Begg’s test or Egger’s test) or funnel plots because their utility has not been established in the meta-analyses of proportions obtained from observational studies.18,22 We considered uncertainty in: (i) the prevalence of tuberculosis infection in child contacts; (ii) the number of tuberculosis cases per household; and (iii) the proportion of child household contacts with active tuberculosis disease. We ignored uncertainty in population size estimates from the United Nations Population Division. Errors were propagated using a second-order Taylor series expansion.23,24 All statistical analyses were performed using Stata v. 13.1 (StataCorp LP., College Station, United States of America) and R v. 3.4.4 (The R Foundation, Vienna, Austria).

Results

Our systematic review of the prevalence of a latent tuberculosis infection among child household contacts younger than 5 years (L) in countries with a low tuberculosis burden included 17 studies (Fig. 1 and Table 2).2541 Nine of the 17 (52.9%) were conducted in high-income countries. The presence of a tuberculosis infection was defined as an induration of 10 mm or more on the tuberculin skin test in 11 studies, whereas the other six used different criteria: (i) one used an induration cut-off of 5 mm; (ii) three used multiple induration cut-offs, ranging from 5 to 15 mm depending on BCG vaccination status, the infectiousness of the index case or the study site; (iii) one used a Heaf grade of 2, 3 or 4; and (iv) one did not specify the criterion. The median prevalence of latent tuberculosis infection among child contacts was 26.4% (interquartile range: 11.1–42.2). Twelve studies included children who had received a BCG vaccination, one included only unvaccinated children and BCG vaccination status was not specified in four studies. There was substantial heterogeneity across the studies. The pooled prevalence of latent tuberculosis infection among child contacts younger than 5 years was 27.9% (95% confidence interval, CI: 18.8–39.4; Fig. 3). None of the subgroup analyses found significant differences between subgroups.

Fig. 3. Forest plot of the prevalence of latent tuberculosis infection among child household contacts, countries with a low tuberculosis burden, worldwide, 1964–2017
Fig. 3. Forest plot of the prevalence of latent tuberculosis infection among child household contacts, countries with a low tuberculosis burden, worldwide, 1964–2017
CI: confidence interval.
Notes: We defined a child household contact as a child younger than 5 years living in the same household as a person with active tuberculosis disease. A low tuberculosis burden was defined as fewer than 100 cases per 100 000 population.

Our systematic review of the number of active tuberculosis cases per household with an index case (C) included 58 studies (Fig. 2 and Table 3).35,37,4297 Of the 58, 16 (27.6%) were conducted in countries with a low tuberculosis burden. The number of active tuberculosis cases among contacts in each household ranged from 0 to 0.33, except for one study that reported a value of 0.93.35 The pooled number of active tuberculosis cases among contacts in each household was 0.06 (95% CI: 0.04–0.07). Consequently, the average number of active tuberculosis cases per household was 1.06 once the index case had been included. There was no significant difference between countries with a low or high tuberculosis burden (P = 0.33). Furthermore, excluding the one outlier reduced the average number of cases per household by only 0.002.

Using the values we obtained for L and C with the values of other parameters from the literature (Table 1), we estimated that the number of child household contacts younger than 5 years who were eligible for tuberculosis preventive treatment in 2017 ranged from less than one in four countries (i.e. Bahamas, Iceland, Luxembourg and Malta) to 350 000 (95% uncertainty interval, UI: 320 000–380 000) in India (Table 4; available at: http://www.who.int/bulletin/volumes/96/8/18-218651). Globally, the estimated number of child contacts eligible for preventive treatment was 1.27 million (95% UI: 1.24 to 1.31). Viewed regionally, the highest estimate was for the WHO South-East Asia Region: 510 000 (95% UI: 450 000–580 000; Table 5).

Discussion

We estimated that 1.27 million children younger than 5 years who were household contacts of people with bacteriologically confirmed pulmonary tuberculosis were eligible for preventive treatment globally in 2017. According to the WHO Global tuberculosis report 2018, countries reported that 292 182 child contacts received preventive treatment in 2017, which makes the best estimate of the global coverage of preventive treatment in children only 23%.98

Our study has several limitations. First, our estimate of the number of child household contacts was based on the number of notified bacteriologically confirmed tuberculosis cases. However, 3.6 million of the estimated 10.0 million people with incident tuberculosis globally in 2017 were neither reported nor enrolled in tuberculosis care.98 Consequently, our estimates are conservative, there would be substantially more eligible child contacts if all incident tuberculosis cases were considered. Second, we used national values for the average household size and for the proportion of the population younger than 5 years to estimate the number of child contacts. It is possible that the composition of households with a tuberculosis case may have differed from the national average and thus people with tuberculosis may have lived with a different number of children younger than 5 years from the national average. Furthermore, we did not consider people with tuberculosis who lived in a prison or nursing home. Doing so would have reduced the estimated number of child contacts, especially in countries where where number of tuberculosis cases among the prison and nursing home populations was high.the prison and nursing home populations were high. Third, we used the value for the average number of tuberculosis cases per household from our new systematic review for all countries, even though it may have varied between countries.

Fourth, in our updated systematic review, we observed substantial heterogeneity across studies in the prevalence of a latent tuberculosis infection among child household contacts in countries with a low tuberculosis burden. This heterogeneity probably reflects differences between studies in characteristic, such as the study population, setting, incidence of tuberculosis, the tuberculin skin test cut-off used and BCG status. We were unable to identify the source of the heterogeneity because the number of studies included in our subgroup analyses was small. Moreover, our estimates of the number of child household contacts eligible for preventive treatment in these countries were derived using an average value for the prevalence of a confirmed tuberculosis infection among child contacts, whereas the prevalence may have varied between countries. Using country-specific values would have given more accurate estimates. Nevertheless, as countries with a low tuberculosis burden accounted for only 14% of notified tuberculosis cases globally in 2017,14,98 their impact on our global estimate was small.

Fifth, we assumed that children were judged eligible for tuberculosis preventive treatment according to WHO guidelines.5 However, eligibility criteria may have varied between countries according to national policy. Sixth, we used a value for the proportion of child household contacts of a tuberculosis case who had active tuberculosis themselves (T) that was derived from a modelling study in 22 countries with a high tuberculosis burden,17 which together accounted for 80% of the global burden. However, the prevalence of active disease among household contacts in these countries was likely to have been higher than in others. Consequently, by using this proportion, we may have underestimated the number of child household contacts without active tuberculosis disease who were, therefore, eligible for preventive treatment. Our estimates of the number of children eligible for preventive treatment need to be validated using national data on the number of child contacts from well-functioning surveillance systems or surveys. These data could also be used to assess the coverage of preventive treatment directly, which should give more accurate figures than our modelling estimates with their inherent limitations. Nevertheless, in the absence of such data, our estimates should help galvanize efforts to implement, and monitor the progress of, tuberculosis preventive treatment among child contacts.

In conclusion, using our estimate of the number of children younger than 5 years eligible for tuberculosis preventive treatment, we calculated that the coverage of preventive treatment in children in 2017 was only 23%. Despite its proven efficacy, tuberculosis preventive treatment is still being underutilized. As the End TB Strategy targets can only be achieved by addressing the pool of tuberculosis infection, urgent action is needed to scale up the implementation of preventive treatment.


Competing interests:

None declared.

References