Abstract
OBJECTIVE: To determine the sensitivity of dual-energy (DE) virtual non-contrast computed tomography (vNCT), generated from the excretory phase of a CT urogram, compared to true non-contrast CT (tNCT) for the detection of urinary calculi. METHODS: A search of multiple medical literature databases was performed using predetermined search terms. Inclusion and exclusion criteria were applied, and bias risk was assessed by two independent reviewers using the quality assessment of diagnostic accuracy studies (QUADAS) tool. Collated estimates of sensitivity were generated, and sources of heterogeneity were identified and reviewed. RESULTS: Thirteen studies (1760 patients; 1740 urinary calculi) were included for sensitivity assessment. Pooled sensitivity for urinary calculi on vNCT was 78.1% (95% CI: 70.2 to 85.0%); however, heterogeneity between studies was very high (I(2) = 92.0%). Sources of heterogeneity between studies were explored through subgroup analysis by categorising studies according to slice thickness (≥ 2 mm and < 2 mm), use of oral hydration, and use of intravenous furosemide. Pooled sensitivity for detection of urinary calculi on vNCT for studies that used oral hydration and < 2 mm slice thickness was 92.2% (95% CI: 89.5 to 94.5%). Pooled specificity was not performed as true negatives were not reported in most studies. Potential sources of bias were identified in included studies. CONCLUSION: vNCT demonstrated a moderate pooled sensitivity compared to tNCT for the detection of urinary calculi in split bolus CT urogram protocols. However, subgroup analysis suggests higher sensitivity when employing oral hydration and < 2 mm slice thickness or increment. KEY POINTS: • vNCT demonstrated moderate pooled sensitivity for the detection of urinary calculi in split bolus CT urogram protocols. • Subgroup analysis suggested higher sensitivity with oral hydration and < 2 mm slice thickness or increment.