Article Text


Diagnosing tuberculosis
  1. U Heininger
  1. Correspondence to:
    Professor Ulrich Heininger
    University Children’s Hospital Basel, PO Box, 4005 Basel, Switzerland;

Statistics from

Commentary on the paper by Swingler et al (see page 1153)

Although pulmonary tuberculosis is now reasonably well controlled in most industrialised countries, it still is a major and even increasing medical problem from a global perspective. The World Health Organisation estimated there were 8.8 million new cases worldwide in 2002, with case numbers currently growing by 2.4% per year.1 Many problems surround the fight against tuberculosis, including the emergence of multi-drug resistant Mycobacterium tuberculosis, compliance with treatment, and identification of new cases in order to avoid further spread of the disease. In children, who contract rather than transmit M tuberculosis, early diagnosis followed by appropriate treatment is important to reduce morbidity and mortality.2 However, the frequently non-specific symptoms of disease, interference of BCG immunisation with tuberculin skin testing, uncertainties with the interpretation of x ray findings, and the rare isolation of organisms from bronchial secretions are all factors that contribute to the difficulty in diagnosing pulmonary tuberculosis in children.

In this respect, the work published by Swingler and collaborators in this issue of the Archives of Disease in Childhood is to be applauded.3 This was a prospective cross sectional study from South Africa, an area with a yearly incidence of tuberculosis reaching almost 500/100 000. The diagnostic accuracy of conventional chest x rays (CXR) in detecting enlarged lymph nodes was assessed in children with clinically suspected pulmonary tuberculosis (see article in this issue for definition). In contrast to previous studies on this topic, not only was inter-observer agreement assessed, where agreement between observers does not guarantee that the diagnosis is correct, but spiral chest computerised tomography (CT) scans with contrast injection were used as the reference standard, where any lymph node measuring at least 1 cm in any dimension was considered a positive finding. The reference was established by a paediatric radiologist blinded to CXR findings. The expert panel for chest x rays consisted of three paediatricians and three primary care clinicians with a special interest in tuberculosis who were blinded to the CT results. Physicians knew about the suspected tuberculosis, but it is not stated clearly whether further medical details of individual patients were known to the reviewers or whether members of the panel were also in charge of clinical care for these patients. Furthermore, we do not know if indirect evidence of tuberculous lymphadenitis, such as obstructive hyperaeration or atelectasis, was searched for on the films. These limitations need to be kept in mind. Statistical methods were sound and 100 consecutive children were enrolled.

Overall, the accuracy of CXR interpretation was disappointing.

First, sensitivity and specificity were 67% and 59%, respectively. In other words, a third of enlarged lymph nodes, as determined by the reference method, were not detected on CXR and almost half of supposedly enlarged lymph nodes on CXR were not confirmed by CT. While the lack of sensitivity may delay or even prevent diagnosis, the lack of specificity could lead to unnecessary further diagnostic tests and/or unjustified treatment. But is this really the case? Doubtless, spiral CT is not available in most health care settings and radiographic imaging in cases of suspected tuberculosis rests on conventional CXR in most centres. However, no data are available which would confirm that tuberculosis is overlooked on initial x rays to such an extent nor do we know the magnitude of erroneous treatments for presumed pulmonary tuberculosis which in fact is not present in the patient. As a major limitation of this study, no data on the clinical course of the patients, any further imaging studies performed later in the course of suspected tuberculosis, or microbiological confirmation of tuberculosis is provided. This makes final interpretation of the accuracy of the reference standard impossible.

Second, in contrast to current textbook recommendations, assessment of both antero-posterior and lateral CXR did not improve sensitivity significantly.4 This has major implications because, if a second view does not improve detection of enlarged lymph nodes, radiation exposure could be reduced and resources thus be saved.

The validity of the reference standard, that is the CT scans, is most critical. A companion paper from members of the South African group spreads doubt about its reliability.5 Using the same cohort of 100 children, the paediatric radiologist’s lymph node assessments of the CT scans were controlled by three other radiologists independently. Unfortunately, the κ statistics, as a measure of agreement beyond chance, were only in the range of 0.6, which is generally considered to be only moderate.6

However, where CT is available, this could prove to be a useful tool in cases of children with suspected pulmonary tuberculosis in the future, as is suggested by the findings of this study. Whether CT should be performed instead of or in addition to conventional CXR will remain a matter of debate until results from further studies with long term follow up of patients allow us to better assess the sensitivity and specificity of both methods. In the meantime, I am afraid we are not much further advanced in radiographic imaging of suspected pulmonary tuberculosis.

Commentary on the paper by Swingler et al (see page 1153)


View Abstract


  • Competing interests: none declared

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.

Linked Articles