Sensitivity of chest X-ray from community screening: a meta-reanalysis of TB prevalence survey data.

BACKGROUNDCommunity TB screening using chest X-ray (CXR) is recommended for high-burden settings. The sensitivity of CXR for community screening is regarded high but may be overestimated due to incorporation bias. We conducted a meta-reanalysis of data used to provide more accurate estimates for CXR sensitivity.METHODSWe considered 23 studies from a published systematic review and meta-analysis where participants were screened for TB using a combination of TB symptoms and CXR and received bacteriological sputum testing as the reference standard. We compared the sensitivity of CXR among symptom screen-positive study participants to that among all study participants.RESULTSTwenty-one studies were included for the meta-reanalysis. The pooled sensitivity (95% confidence interval) of CXR among symptom screen-positive participants was 86% (83%-90%), lower than the pooled sensitivity based on all study participants (94% [92%-96%]) reported in the original meta-analysis. The pooled difference in sensitivity between the two approaches was 6% (4%-8%).CONCLUSIONThe sensitivity of CXR among symptomatic participants was lower than the published sensitivity, indicating incorporation bias. The sensitivity among symptomatic participants should be considered the realistic maximum sensitivity for community screening; the combined sensitivity for symptomatic and asymptomatic TB is likely lower..