Grading of carotid artery stenosis with multidetector-row CT angiography: visual estimation or caliper measurements?

To assess the optimal method for grading carotid artery stenosis with computed tomographic angiography (CTA), we compared visual estimation to caliper measurements, and determined inter-observer variability and agreement relative to digital subtraction angiography (DSA). We included 46 patients with symptomatic carotid stenosis for whom CTA and DSA of 55 carotids was available. Stenosis quantification by CTA using visual estimation (CTA_VE) (method 1) was compared with caliper measurements using subjectively optimized wide window settings (method 2) or predefined contrast-dependent narrow window settings (method 3). Measurements were independently performed by two radiologists and two residents. To determine accuracy and inter-observer variability, we calculated linear weighted kappa, performed a Bland-Altman analysis and calculated mean difference (bias) and standard deviation of differences (SDD). For inter-observer variability, kappa analysis was “very good” (0.85) for expert observers using CTA_VE compared with “good” (0.61) for experts using DSA. Compared with DSA, method 1 led to overestimation (bias 5.8–8.0%, SDD 10.6–14.4), method 3 led to underestimation (bias −6.3 to −3.0%, SDD 13.0–18.1). Measurement variability between DSA and visual estimation on CTA (SDD 11.5) is close to the inter-observer variability of repeated measurements on DSA that we found in this study (SDD 11.6). For CTA of carotids, stenosis grading based on visual estimation provides better agreement to grading by DSA compared with stenosis grading based on caliper measurements.