Reexamining the role of radiography in tuberculosis case finding.

IN THIS ISSUE of the Journal, Van’t Hoog and colleagues report high sensitivity of chest radiography in a tuberculosis (TB) prevalence survey in a resourcelimited area.1 It is time to reexamine the role of this century-old tool. Sputum microscopy has been the main tool for passive case fi nding among symptomatic patients in resource-limited settings. However, symptoms suggestive of TB were present in only 75 (61%) of 123 bacteriologically confi rmed cases in the above-mentioned survey. While sputum microscopy could only pick up 51 (41%) of these cases, 113 (94%) of 120 patients with an available chest radiograph showed some r adiological abnormality. Although interpretation of chest radiographs by qualifi ed medical personnel was found to have low sensitivity for bacteriologically confi rmed TB among human immunodefi ciency virus (HIV) infected patients in South Africa,2 lowering the detection threshold from ‘abnormality consistent with TB’ to ‘any abnormality’ allowed trained nonmedical clinical offi cers to pick out 95% of bacteriologically confi rmed TB cases, half of whom were HIV-infected.1 Installation of X-ray facilities requires considerable capital investment. However, the short turnover time and high throughput offer major attractions in the fi eld. Filmless techniques such as fl uoroscopy and digitisation further reduce the marginal costs in areas with a high patient load. In the 2000 TB prevalence survey in China, chest fl uoroscopy reduced the number of subjects requiring sputum examination from 365 097 to below 5000.3 Similarly, in a World Bankfunded DOTS project in China, chest fl uoroscopy reduced the number of symptomatic subjects requiring sputum examination by half,4 and an extra 32% decline in TB prevalence was observed in project areas compared to non-project areas from 1991 to 2000.3 In short, chest radiography has a defi nite potential in increasing the sensitivity and expanding the capacity of TB case fi nding, even in areas with a high prevalence of HIV infection. Although new molecular tools offer great promise for rapid diagnosis of TB, the current costs necessarily dictate their highly selective application in resource-limited settings. Chest radiography might therefore help to identify high-risk symptomatic patients with negative sputum smears for further work-up, thus maximising the overall costeffectiveness in case fi nding for both drug-susceptible and drug-resistant TB. Similarly, in settings with a high prevalence of active TB, chest radiography may also be a cost-effective solution, with higher sensitivity than sputum microscopy for active case fi nding. For example, chest X-ray examination of long-stay prisoners was found to give a 3.1% yield of smearnegative TB cases among those who had not undergone such an examination in 2 years.5 The increasing availability of mobile X-ray facilities and back-up generators has allowed the deployment of chest radiography for case fi nding even in relatively remote areas with an unstable electricity supply. As fi lm reading by non-medical personnel with suitable training could be highly sensitive in detecting TB cases,1 it might be worthwhile examining whether computer-aided screening of digitised images using standardised schemas could yield comparable results. With suitable technical refi nements and careful operational research, it is hoped that this centuryold diagnostic tool will fi nd an expanded role in TB case fi nding.