Agreement between site-reported and ultrasound core laboratory results for duplex ultrasound velocity measurements in the Carotid Revascularization Endarterectomy versus Stenting Trial

Objective Patients in the Carotid Revascularization Endarterectomy vs Stenting Trial (CREST) had duplex ultrasound (DU) scans prior to treatment and during follow-up to document the severity of carotid disease and the anatomic outcome of carotid endarterectomy (CEA) or carotid artery stenting (CAS). An ultrasound core laboratory (UCL) reviewed DU data from the clinical sites. This analysis was done to determine the agreement between site-reported and UCL-verified DU velocity measurements. Methods Clinical site DU worksheets, B-mode images, and Doppler velocity waveforms for the treated carotid arteries were reviewed at the UCL. The highest internal carotid artery peak systolic velocity (PSV) and associated Doppler angle were verified. If the angle was misaligned by >3 degrees, it was remeasured at the UCL and the PSV was recalculated. Agreement for PSV was defined as site-reported PSV within ±5% of UCL-verified PSV. Transcription errors were corrected by the UCL but were not considered as disagreements. Follow-up analysis was limited to patients who received the assigned treatment. Results The UCL reviewed 1702 prior-to-treatment and 1743 12-month follow-up DU scans (873 CEA, 870 CAS) from 111 clinical sites. Site-reported and UCL-verified PSV agreed in 1124 (66%) of the prior-to-treatment scans and 1200 (69%) of the follow-up scans. In those cases with a disagreement, Doppler angle accounted for disagreement in 339 (59%) of the prior-to-treatment scans and 277 (51%) of the follow-up scans. Based on a threshold PSV for ≥70% stenosis of ≥230 cm/s on the prior-to-treatment scans and ≥300 cm/s on the follow-up scans, UCL review resulted in reclassification of stenosis severity in 75 (4.4%) of the prior-to-treatment scans and 13 (0.75%) of the follow-up scans. There is evidence that the proportion of reclassification at follow-up was greater for CAS (10 scans; 1.2%) than for CEA (three scans; 0.34%) ( P = .057). Conclusions There was a high rate of agreement between site-reported and UCL-verified DU results in CREST, and UCL review was associated with a low rate of stenosis reclassification. However, angle alignment errors were quite common and prompted recalculation of velocity in 20% of prior-to-treatment scans and 18% of follow-up scans. The use of a UCL provides a uniform process for DU interpretation and can identify sources of error and suggest technical improvements for future studies.