Accuracy of the central vein sign in diagnosis of multiple sclerosis: a systematic review and meta-analysis

Background: The central vein sign (CVS) is a promising imaging marker of multiple sclerosis (MS). We performed a systematic review and meta-analysis to evaluate the diagnostic accuracy of CVS-based rules in the differential diagnosis of MS and to identify the best cutoffs for these rules. Methods: PubMed, Embase, and Scopus were systematically searched for available evidence. Data extracted were entered into Bayesian models recommended by the Cochrane network. Summary sensitivity and specificity of CVS-based diagnostic rules across different positivity thresholds were calculated. A meta-regression for the role of gadolinium-based MRI protocols was also performed. Results: 3434 patients from 28 studies were included. Three CVS-based diagnostic rules were found: the first one considers the percentage of CVS + lesions (relative threshold method), the second one the presence of CVS in a given number of lesions selected in T2 sequences (select-n method), and the third one the presence of a given number of CVS + lesions in gradient-echo sequences (select-n* method). For relative threshold method, the best cutoff was 37.5% (sensitivity 97.3%, 95%CI 90.9–99.6%; specificity 90.4%, 95%CI 83.2–95.9%; Youden index 0.877); for select-n* method, 4 was the best threshold (sensitivity 87.1%, 95%CI 66.9–96.6%; specificity 88.2%, 95%CI 65.1–98.1%; Youden index 0.753). Use of gadolinium-based MRI protocols was irrelevant (for relative threshold method RDOR = 6.62, 95%CI 0.68–71.27; for select-n* method RDOR = 2.52, 95%CI 0.35–15.36). Conclusions: Relative threshold and select-n* methods are good predictors of MS diagnosis. This synthesis should support the use of CVS in clinical practice and prompt further research.