Elsevier

The Lancet

Volume 364, Issue 9452, 18–31 December 2004, Pages 2196-2203
The Lancet

Articles
Diagnosis of tuberculosis in South African children with a T cell-based assay: a prospective cohort study

https://doi.org/10.1016/S0140-6736(04)17592-2Get rights and content

Summary

Background

Childhood tuberculosis often presents non-specifically and is a common differential diagnosis in high prevalence areas. Current diagnostic tools have poor sensitivity and cannot reliably exclude tuberculosis, so overdiagnosis is common. HIV co-infection exacerbates this problem and accounts for an increasing proportion of paediatric tuberculosis worldwide.

Methods

We assessed the usefulness of a T-cell-based rapid blood test for Mycobacterium tuberculosis infection, the enzyme-linked immunospot assay (ELISPOT), in routine clinical practice. We did a prospective blinded study of 293 African children with suspected tuberculosis in kwaZulu-Natal, a region with high HIV prevalence. Children had full clinical assessment, ELISPOT, and a tuberculin skin test. Test results were compared with final clinical and microbiological diagnoses.

Results

In children with tuberculosis, sensitivity of ELISPOT was 83% (95% CI 75–89, n=133), significantly higher (p<0·001) than the 63% (54–72) sensitivity of tuberculin skin test (n=116). Sensitivity of tuberculin skin test fell significantly in children younger than 3 years (to 51%), with HIV co-infection (36%), or with malnutrition (44%). Sensitivity of ELISPOT, which was not significantly adversely affected by these factors, was 85%, 73%, and 78%, respectively in these subgroups. In 116 children with both test results available, sensitivity of the two tests combined was 91% (85–95).

Conclusions

Diagnostic sensitivity of ELISPOT is higher than that of the skin test and is less affected by factors frequently associated with childhood tuberculosis in developing countries. Used together with the skin test, ELISPOT provides a clinically useful diagnostic sensitivity in African children with suspected tuberculosis.

Introduction

The worldwide resurgence of tuberculosis is most severe in areas of high HIV prevalence, especially sub-Saharan Africa.1, 2, 3, 4 In resource-poor regions with a high prevalence of tuberculosis, children carry a large and increasing proportion of the overall burden,5, 6 and the rise in the proportion of children with tuberculosis who are infected with HIV has been well documented in Africa.7, 8, 9 An autopsy series of Zambian children dying of respiratory disease showed that 20% had tuberculosis.10

Because of its non-specific clinical presentation, tuberculosis is one of the most common differential diagnoses in sick children in high prevalence areas.6 However, current tests have poor sensitivity and cannot reliably exclude the diagnosis11, 12 and, since untreated paediatric tuberculosis carries a high mortality, overdiagnosis is common.13 This problem is exacerbated in areas with a high HIV prevalence,6, 11 because HIV disease and tuberculosis share many clinical and radiological features, and it is often impossible to exclude tuberculosis in HIV-infected children. Moreover, untreated tuberculosis in HIV-infected children is rapidly fatal. As a result, empirical therapy is common and often results in inappropriate treatment and overloading of already stretched tuberculosis services.13

A diagnosis of active tuberculosis is confirmed by positive acid-fast stains or positive cultures for Mycobacterium tuberculosis from clinical specimens. However, suitable specimens are difficult to obtain, since a high proportion of childhood tuberculosis is extrapulmonary or involves exclusively intrathoracic lymph nodes,9, 14, 15 requiring invasive procedures, and young children with pulmonary tuberculosis rarely produce sputum.12, 16 Fewer than 20% of specimens obtained from children with tuberculosis are positive by acid-fast stain.9, 12, 14, 17 Facilities for mycobacterial culture are often scarce in regions where tuberculosis is common, and even when available, culture results are frequently negative and usually too late to affect initial management. As a result, few children treated for tuberculosis have a microbiological diagnosis.11, 12

The tuberculin skin test (TST), which provides evidence of M tuberculosis infection, is widely used to lend support to clinical and radiological findings in the assessment of children with suspected tuberculosis. However, the sensitivity of TST in children with active tuberculosis is low, and is even lower in children with disseminated tuberculosis, malnutrition, or HIV infection, which are all common in regions with a high prevalence of tuberculosis.5, 9, 15 Hence, in children with suspected tuberculosis, a negative TST result does not help to exclude the diagnosis. The specificity of TST is also compromised because of cross-reactivity with BCG vaccine and environmental mycobacterial exposure,18 both of which are widespread in the developing world.19 This difficulty results in inappropriate treatment of recently vaccinated children with false-positive TST results.

A rapid blood test for M tuberculosis infection, the enzyme-linked immunospot assay (ELISPOT), might represent a more accurate alternative to TST.20 The assay detects T cells specific for antigens in M tuberculosis but absent from BCG and most environmental mycobacteria.21 The result is therefore not confounded by previous BCG vaccination, and the test is more specific than TST.21, 22, 23 A study in HIV-negative adults with culture-confirmed active tuberculosis suggested that ELISPOT has a higher sensitivity than TST,21 and its sensitivity remained high (92%) in HIV-positive adults with tuberculosis.24 Hence, since M tuberculosis infection is a necessary prerequisite for tuberculosis disease, a negative result in patients with suspected tuberculosis might help to rule out the diagnosis.20 These attributes suggest that ELISPOT might be able to improve diagnosis of paediatric tuberculosis.

We aimed to answer two questions. First, can this T cell-based approach improve diagnostic assessment of African children with suspected tuberculosis and high rates of HIV co-infection in routine clinical practice? Second, is the assay robust enough for a district hospital laboratory in a developing country? We prospectively studied unselected children presenting with suspected active tuberculosis in routine clinical practice. We did ELISPOT and TST at the time of initial assessment, and we compared test results with the final clinical and microbiological diagnoses. The study took place in kwaZulu-Natal, South Africa, where the tuberculosis notification rate is 700 per 100 000. Most children were recruited from a district hospital where 28% of children attending paediatric outpatient clinics are HIV-positive and 27% are malnourished.25 Thus, the spectrum of disease in our study population reflects that in most of sub-Saharan Africa where the burden of HIV and tuberculosis is very high, and represents a considerable diagnostic challenge.

Section snippets

Participants

Children were enrolled from November, 2000, through August, 2001, at two hospitals in kwaZulu-Natal: Ngwelezana, a district hospital serving a mixed rural and urban population, and King George V Hospital, a tuberculosis referral centre in Durban. All children aged up to 14 years who were being assessed for suspected tuberculosis were eligible for enrolment, and all had one or more of the following at presentation: fever or cough for more than 1 month; reported weight loss or failure to thrive;

References (33)

  • FA Drobniewski et al.

    Modern laboratory diagnosis of tuberculosis

    Lancet Infect Dis

    (2003)
  • EL Corbett et al.

    The growing burden of tuberculosis: global trends and interactions with the HIV epidemic

    Arch Intern Med

    (2003)
  • E Girardi et al.

    Impact of the HIV epidemic on the spread of other diseases: the case of tuberculosis

    AIDS

    (2000)
  • MC Raviglione et al.

    Tuberculosis and HIV: current status in Africa

    AIDS

    (1997)
  • PR Donald

    Childhood tuberculosis: out of control?

    Curr Opin Pulm Med

    (2002)
  • A van Rie et al.

    Childhood tuberculosis in an urban population in South Africa: burden and risk factor

    Arch Dis Child

    (1999)
  • Cited by (338)

    View all citing articles on Scopus
    View full text