Comparison of chest radiographs against minimum intensity projection reconstruction computed tomography scans for detection of airway stenosis in children with lymphobronchial tuberculosis

Ideally, suspected airway compression in symptomatic children with lymphobronchial tuberculosis (TB) would be diagnosed using modern computed tomography (CT) assisted by coronal minimum intensity projection (MinIP) reconstructions. However, in TB-endemic regions with limited resources, practitioners rely on conventional radiography for diagnosing TB and its complications. Furthermore, airway compression detected on conventional radiographs would upgrade a patient into the severe category according to the new World Health Organization guidelines, precluding the patient from shorter treatment protocols. The accuracy of conventional radiographs in the context of detecting airway compression in children with TB has not been specifically evaluated against an imaging gold standard.

We aimed to compare frontal chest radiographs against thick-slab angled coronal CT MinIP in identifying airway stenosis at ten specific sites and to determine observer agreement between the modalities regarding the degree of stenosis.

This retrospective cross-sectional study compared chest radiographs with standardized angled coronal CT MinIP in children with symptomatic lymphobronchial TB at ten predetermined airway locations. Chest radiographs were evaluated by one pediatric radiologist and CT MinIP reconstructions were independently interpreted by three readers. Sensitivity and specificity were calculated using CT MinIP as the gold standard. Stenosis was graded as 1 for mild (1–50%), 2 for moderate (51–75%), 3a for severe (76–99%), and 3b for total occlusion (100%). Agreement between the two modalities regarding severity of stenosis was calculated using the kappa coefficient for each affected site.

A total of 37 patients were included in the study. The median age of patients was 14.3 months (interquartile range 8.0–23.2). Three hundred and seventy individual bronchi (10 from each of the 37 patients) were evaluated for stenosis. Chest radiographs showed that 31 out of 37 (84%) patients had stenosis in at least one of ten evaluated sites, most commonly the left main bronchus and bronchus intermedius, and this was confirmed via CT MinIP. The gold standard (CT MinIP) demonstrated stenosis in at least one of ten sites in all 37 patients (100%). Left main bronchus stenosis was detected by chest radiography with a 92.9% sensitivity and 100% specificity. Sensitivity and specificity for bronchus intermedius stenosis were 80% and 75%, respectively. There was substantial agreement for grade of stenosis between chest radiographs and CT (kappa=0.67) for the left main bronchus and moderate agreement (kappa=0.58) for the bronchus intermedius. Severe stenosis was found in 78 bronchi on CT compared to 32 bronchi (Grade 3a: 9, Grade 3b: 23) on chest radiographs.

The diagnosis of pulmonary TB in children continues to rely heavily on imaging, and we have shown that in young children, chest radiographs had a high sensitivity and specificity for detecting airway stenosis at certain anatomical sites, when adequately visualized, resulting from tuberculous lymph node compression at left main bronchus and bronchus intermedius. For most sites, the interobserver agreement was poor. Stenosis of the left main bronchus and bronchus intermedius should be the focus of chest radiograph interpretation and can assist both diagnosis and classification of patients for treatment.