Middle cerebral artery vasospasm: Transcranial color-coded duplex sonography versus conventional nonimaging transcranial Doppler sonography*

Abstract

To prospectively compare accuracies of transcranial color-coded duplex sonography (TCCS) and transcranial Doppler sonography (TCD) in the diagnosis of middle cerebral artery (MCA) vasospasm. Prospective blinded head-to-head comparison TCD and TCCS methods using digital subtraction angiography (DSA) as the reference standard. Department of Radiology in a tertiary university health center in a metropolitan area. Eighty-one consecutive patients (mean age, 53.9 +/- 13.9 years; 48 women). The indication for DSA was subarachnoid hemorrhage in 71 patients (87.6%), stroke or transient ischemic attack in five patients (6.2%), and other reasons in five patients (6.2%). The MCA was graded as normal, narrowed <50%, and >50% using DSA. The accuracy of ultrasound methods was estimated by total area (Az) under receiver operator characteristic curve. To compare sensitivities of ultrasound methods, McNemar's test was used with mean velocity thresholds of 120 cm/sec for the detection of less advanced, and 200 cm/sec for the more advanced MCA narrowing. Angiographic MCA narrowing <or=50% was found in 21, and >50% in 10 of 135 arteries. Accuracy of TCCS was insignificantly higher than that of TCD in the detection of <or=50% and >50% narrowing, total Az for mean velocity being 0.83 +/- 0.05, 0.77 +/- 0.05, and 0.95 +/- 0.02, 0.86 +/- 0.08, respectively. Sensitivity of TCCS at commonly used threshold of 120 cm/sec for less advanced MCA spasm was significantly better than that of TCD at similar specificity, 55% vs. 39%, p = 0.038, whereas at a threshold of 200 cm/sec used for more advanced spasm, sensitivities and specificities of both methods were not different. The accuracy of TCCS and TCD is similar, but TCCS is more sensitive than TCD in the detection of MCA spasm. Sensitivity of both techniques in the detection of mild and more advanced spasm using 120 cm/sec and 200 cm/sec thresholds, respectively, is poor; however, a larger sample is required to increase precision of our sensitivity estimates.