Abstract
BACKGROUND: The World Health Organization recommends systematic chest X-ray (CXR) screening for tuberculosis (TB) in silica-exposed workers. However, evidence on the accuracy of CXR screening in such populations is lacking. AIMS: To measure the accuracy of CXR screening for active TB in silica-exposed miners, in a population with a high prevalence of silicosis, post-TB lung disease and HIV. METHODS: A secondary analysis of data from a miner screening programme in Lesotho was undertaken. We measured the performance of CXR (in participants with and without cough) for 'abnormalities suggestive of TB' against Xpert MTB/RIF (Xpert). The sample size was 2572 and positive Xpert prevalence was 3%. RESULTS: CXR alone had high sensitivity (0.93, 95% confidence interval [CI] 0.87-0.99), but low specificity (0.41, 95% CI 0.39-0.42). Requiring cough and a positive CXR increased specificity (0.79, 95% CI 0.77-0.81), resulting in reduced sensitivity (0.41, 95% CI 0.30-0.52). There was no difference in CXR accuracy by HIV status. However, specificity was markedly reduced in the presence of silicosis (from 0.70, 95% CI 0.68-0.72, to 0.03, 95% CI 0.02-0.04) or past TB history (from 0.59, 95% CI 0.56-0.62 to 0.27, 95% CI 0.25-0.29). Throughout, positive predictive value remained very low (5%) and negative predictive value very high (99%). CONCLUSIONS: CXR screening accurately identifies TB-negative CXRs in this population, but post-TB lung disease and silicosis would result in a high proportion of Xpert-negative referrals and an increased risk of unneeded empirical treatment. Adapted screening algorithms, practitioner training and digital access to previous mining CXRs are needed.