Objective: Quantitative studies of cerebral blood flow (CBF) combined with a vasodilatory challenge have defined a subgroup of patients with symptomatic carotid occlusion who have an increased risk for stroke. These are patients whose CBF paradoxically decreases in response to a vasodilatory challenge. Recent reports suggest that qualitative CBF techniques, such as single photon emission tomography with 99m-hexamethylpropyleneamine oxime, can also define the same high-risk subgroup. To determine whether qualitative measures of CBF are sufficient for predicting the risk of stroke, we converted our quantitative CBF data, obtained with xenon-enhanced computed tomography (Xe/CT), to qualitative ratios in a manner similar to that used with single photon emission tomography data.
Methods: We analyzed CBF values within the territory of the middle cerebral artery for 94 patients with symptomatic carotid occlusion. Values obtained using Xe/CT before and after the intravenous administration of 1 g of acetazolamide were used to derive an asymmetry index: (Coccl - Cnon)/Cavg x 100. The difference between the postacetazolamide asymmetry index and the baseline asymmetry index was used to classify the patients into groups according to CBF values. The threshold for abnormal qualitative CBF reactivity was defined as a percent change in the asymmetry index of less than -10%. Quantitative (Xe/CT) CBF was considered abnormal ("steal" response) when the response to acetazolamide (percent change) on the occluded side was a decrease of 5% or greater.
Results: Of 34 patients whose cerebrovascular reserves were considered compromised based on qualitative criteria, 17 (50%) did not have a steal response as defined by quantitative Xe/CT CBF (i.e., false positive). Eleven of 62 (18%) who were not considered compromised by qualitative criteria had a steal response (i.e., false negative). Our data indicate that a qualitative approach has a 61% sensitivity and a 75% specificity for detecting patients with compromised reserves. Further, the positive predictive value of this method is only 50%. Therefore, the two methodologies do not predict the same patients as having compromised reserves.
Conclusion: Previous studies have shown that patients at high risk for stroke can be identified with quantitative CBF methods. This study shows that the important subgroup cannot be accurately defined with qualitative methodology. The implications of using the more reliable methodology are important for individual patient management and for designing clinical trials.