Diagnostic Performance and Interobserver Agreement of the Vesical Imaging-Reporting and Data System (VI-RADS) in Bladder Cancer Staging: A Systematic Review.

Background and

Objectives

The Vesical Imaging-Reporting and Data System (VI-RADS) represents a standardized approach for interpreting multiparametric magnetic resonance imaging (mp-MRI) in bladder cancer (BC) evaluation. This systematic review aimed to assess the VI-RADS' diagnostic performance and interobserver agreement in distinguishing muscle-invasive from non-muscle-invasive BC, a crucial differentiation for treatment planning. Materials and

Methods

A systematic literature search was conducted through PubMed, Google Scholar, and Web of Science, over an initial five-year time span, from VI-RADS' inception (May 2018) to November 2023. Studies reporting VI-RADS' diagnostic performance with histopathological confirmation and interobserver agreement data were included. The diagnostic accuracy was assessed using sensitivity and specificity, while interobserver agreement was evaluated using Cohen's κ coefficient.

Results

Nine studies comprising 1249 participants met the inclusion criteria. Using a VI-RADS score cutoff of ≥3, the pooled sensitivity and specificity for detecting muscle invasion were 88.2% and 80.6%, respectively. Interobserver agreement showed excellent consistency with a mean κ value of 0.82. Individual study sensitivities ranged from 74.1% to 94.6%, while specificities varied from 43.9% to 96.5%.

Conclusions

VI-RADS demonstrates high diagnostic accuracy and excellent interobserver agreement in BC staging, supporting its role as a reliable non-invasive diagnostic tool. However, it should be used as a complementary tool to, not a replacement for, histopathological confirmation. Moreover, the variability in specificity suggests the need for standardized training and interpretation protocols. Clinical correlation and adequate reader experience are essential for optimal implementation. Future integration with pathological data may further enhance its predictive value.