Do technical parameters affect the diagnostic accuracy of virtual bronchoscopy in patients with suspected airways...

Virtual bronchoscopy has gained popularity over the past decade as an alternative investigation to conventional bronchoscopy in the diagnosis, grading and monitoring of airway disease. The effect of technical parameters on diagnostic outcome from virtual bronchoscopy has not been determined. This meta-analysis aims to estimate accuracy of virtual compared to conventional bronchoscopy in patients with suspected airway stenosis, and evaluate the influence of technical parameters.

A MEDLINE search was used to identify relevant published studies. The primary endpoint was the "correct diagnosis" of stenotic lesions on virtual compared to conventional bronchoscopy. Secondary endpoints included the effects of the technical parameters (pitch, collimation, reconstruction interval, rendering method, and scanner type), and date of publication on the diagnostic accuracy of virtual bronchoscopy.

Thirteen studies containing 454 patients were identified. Meta-analysis showed good overall diagnostic performance with 85% calculated pooled sensitivity (95% CI 77-91%), 87% specificity (95% CI 81-92%) and area under the curve (AUC) of 0.947. Subgroups included collimation of 3mm or more (AUC 0.948), pitch of 1 (AUC 0.955), surface rendering technique (AUC 0.935), and reconstruction interval of more than 1.25 mm (AUC 0.914). There was no significant difference in accuracy accounting for publication date, scanner type or any of the above variables. Weighted regression analysis confirmed none of these variables could significantly account for study heterogeneity.

Virtual bronchoscopy performs well in the investigation of patients with suspected airway stenosis. Overall sensitivity and specificity and diagnostic odds ratio for diagnosis of airway stenosis were high. The effects of pitch, collimation, reconstruction interval, rendering technique, scanner type, and publication date on diagnostic accuracy were not significant.