Assessment of airway compression on chest radiographs in children with pulmonary tuberculosis

Because small, pliable paediatric airways are easily compressed by enlarged lymph nodes, detection of radiographic airway compression might be an objective criterion for diagnosing pulmonary tuberculosis. To investigate the frequency and inter-observer agreement of airway compression on chest radiographs in children with pulmonary tuberculosis compared to those with a different lower respiratory tract infection. Chest radiographs of children with suspected pulmonary tuberculosis were read by two readers according to a standardised format and a third reader when there was disagreement. Radiographs of children with proven pulmonary tuberculosis were compared to those with a different lower respiratory tract infection. We evaluated frequency and location of radiographic airway compression. Findings were correlated with human immunodeficiency virus (HIV) status and age. We assessed inter-observer agreement using kappa statistics. We reviewed radiographs of 505 children (median age 25.9 months, interquartile range [IQR] 14.3-62.2). Radiographic airway compression occurred in 54/188 (28.7%) children with proven pulmonary tuberculosis and in 24/317 (7.6%) children with other types of lower respiratory tract infection (odds ratio [OR] 4.9; 95% confidence interval [CI] 2.9-8.3). A higher frequency of radiographic airway compression occurred in infants (22/101, or 21.8%) compared to older children (56/404, or 13.9%; OR 1.7; 95% CI 1.0-3.0). We found no association between airway compression and HIV infection. Inter-observer agreement ranged from none to fair (kappa of 0.0-0.4). There is a strong association between airway compression on chest radiographs and confirmed pulmonary tuberculosis. However this finding's clinical use as an objective criterion for diagnosis of pulmonary tuberculosis in children is limited by poor inter-observer agreement. (orig.)