Computed tomography angiography for quantification of cerebral vasospasm following aneurysmal subarachnoid hemorrhage

The purpose of this study was to assess the accuracy of computed tomography angiography (CTA) for quantification of cerebral vasospasm following aneurysmal subarachnoid hemorrhage in proximal and middle segments of intracranial arteries.

Twenty consecutive patients (7 men, 13 women; mean age, 47 ± 7 [SD] years; age range: 27-78 years) with aneurysmal subarachnoid hemorrhage who underwent CTA and digital subtracted angiography (DSA) with a 6-hour window at baseline and during vasospasm period were included. Twelve artery segments were analyzed in each patient. Vasospasm was blindly quantified on CTA and digital subtracted angiography (DSA) by two independent readers with discordance > 10% resolved by open data consensus. Inter-reader and inter-test correlations with DSA as reference, and causes of discordant readings were analyzed. The best sensitivity and specificity of CTA for determination of vasospasm ≥ 50% on DSA was determined using receiver operating curve analysis.

Two-hundred-and-ten arterial segments were analyzed after exclusion of 30 segments with missing data or metallic artifacts. An inter-reader discordance >10% was observed in 82 segments (82/210; 39% [95% CI: 32-46]). Inter-test discordances >10% were observed respectively in 115 segments (115/210; 55% [95% CI: 49-62]) with the junior reader and in 73 segments (73/210; 35% [95% CI: 29-42]) with the senior reader. They were related to reader error in 55 (55/210; 26% [95% CI: 20-32]) with the junior reader and 13 (13/210; 6% [95% CI: 3-9]) with the senior reader, as well systematic biases in 8 (8/210; 4% [95% CI: 1-6]), and intrinsic limitation in 52 (52/210; 25% [95% CI: 19-31]). Best sensitivity and specificity of CTA were observed for a threshold value of 30% (sensitivity = 88% [95% CI: 78-97%]; specificity = 84% [95% CI: 77-90%]; area under curve = 0.92 [95% CI: 0.86-0.97]). On a patient basis, sensitivity was 100% (specificity = 60% [95% CI: 38-81%]; area under curve = 0.97 [95% CI: 89-100%] for this same threshold.

Our study shows a moderate accuracy of CTA for the quantification of cerebral vasospasm, mostly related to challenging interpretation and intrinsic limitations. CTA may rule-out angiographic vasospasm ≥ 50% when no segment has vasospasm over than 30%.