How much does the plain chest radiograph underestimate the extent of disease in subclinical pulmonary tuberculosis?

Subclinical pulmonary tuberculosis (PTB) is a recently described intermediate state of great interest, but about which little is known. This study sought to estimate the extent to which chest radiographs (CXRs), as compared to computed tomographic (CT) scans, under-detect key radiologic features in subclinical PTB, and to interpret the clinical and public health relevance of the differences. Diagnostic CXRs and CT scans of the thorax and neck in a 16-year cohort of subclinical PTB patients in Canada were re-acquired and read by two independent readers and arbitrated by a third reader. Logistic regression models were fit to determine how likely CXR features can be detected by CT scan versus CXR after adjustment for age and sex. Among 296 subclinical patients, CXRs were available in 286 (96.6%) and CT scans in 94 (32.9%). CXR features in patients with and without CT scans were comparable. Lung cavitation was 4.77 times (95% CI, 1.95–11.66), endobronchial spread 19.36 times (95% CI, 8.05–46.52), and moderate/far-advanced parenchymal disease 3.23 times (95% CI, 1.66–6.30), more common on CT scan than CXR. We conclude that the extent to which CXRs under-detect key radiologic features in subclinical PTB is substantial. This may have public health and treatment implications.