Diagnostic accuracy of contrast-enhanced CT for neck abscesses: A systematic review and meta-analysis of positive predictive value

Abstract

To review the diagnostic accuracy of contrast-enhanced computed tomography (CT) in differentiating abscesses from cellulitis in patients with neck infections, using surgical findings as the reference standard. Previous studies in the last 32 years were searched from PubMed and Embase. Because of partial verification bias (only positive abscess findings are usually verified surgically), sensitivity and specificity estimates are unreliable, and we focused on positive predictive value (PPV). For all studies, PPV was calculated as the proportion of true positives out of all positives on imaging. To estimate pooled PPV, we used both the median with an interquartile range and a model-based estimate. For narrative purposes, we reviewed the utility of common morphological CT criteria for abscesses, such as central hypodensity, the size of the collection, bulging, rim enhancement, and presence of air, as well as sensitivity and specificity values reported by the original reports. 23 studies were found reporting 1453 patients, 14 studies in children (771 patients), two in adults (137 patients), and seven including all ages (545 patients). PPV ranged from 0.67 to 0.97 in individual studies, had a median of 0.84 (0.79-0.87), and a model-based pooled estimate of 0.83 (95% confidence interval 0.80-0.85). Most morphological CT criteria had considerable overlap between abscesses and cellulitis. The pooled estimate of PPV is 0.83 for diagnosing neck abscesses with CT. False positives may be due to limited soft tissue contrast resolution. Overall, none of the morphological criteria seem to be highly accurate for differentiation between abscess and cellulitis.