Diagnosis of Basilar Artery Vasospasm
To the Editor:
I read with interest the recent article in Stroke by Soustiel et al1 regarding improvement in the diagnosis of basilar artery vasospasm by transcranial Doppler (TCD) sonography. The article describes normative and patient-derived data pertaining to a BA/EVA ratio, defined as the ratio between the highest recorded basilar artery (BA) flow velocities (insonation depth >80 mm) and the average of the flow velocities from the 2 extracranial vertebral arteries (EVA) (insonation depth 45 to 55 mm). The purpose of the study was to learn if this ratio could help discriminate between BA vasospasm and BA hyperemia, given that BA flow velocities may be difficult to interpret in this setting.2,3⇓ In fact, a previous study2 showed that 5/6 false-positive TCD examinations for BA vasospasm were of unknown cause. It has recently been hypothesized that this observation may be due to dysautoregulation or hyperemia.3–5⇓⇓ The present authors showed (1) a strong linear correlation between BA diameter measured by computed tomographic angiography and the BA/EVA ratio, (2) that a BA/EVA ratio >2 has 100% sensitivity for BA vasospasm, and (3) that a BA/EVA ratio >3 identified severe angiographic BA vasospasm.
While these data are useful for the interpretation of BA flow velocities patients with subarachnoid hemorrhage, there are a number of methodologic concerns6 about the work of Soustiel et al.1 First, the definition of the BA/EVA ratio bears a striking resemblance to the Vba/Vva ratio described 5 years ago.7 In that article,7 normative data and patient-derived data using the same rules for insonating the vessels and calculation of the so-called posterior circulation flow index5 were presented. Using a Vba/Vva ratio >2.5, the specificity of TCD for BA vasospasm was substantially improved from 42.3% to 87.5%, suggesting that the Vba/Vva ratio can eliminate most false-positive TCD examinations attributable to BA hyperemia. As such, the present data1 confirm and extend the value of a posterior circulation flow index that may adjust for hyperemia or increased intracranial pressure when interpreting BA flow velocities. However, it is not clear why the present authors believe that an insonation depth of 45 to 55 mm corresponds to the extracranial portion instead of the proximal intracranial portion of the vertebral arteries, since the TCD probe is typically placed immediately inferior to the occipital protuberance to insonate the vertebrobasilar circulation. Second, it is not clear why the authors chose computed tomographic angiography as the “gold standard” against which TCD was compared, especially since reference to a study showing the comparability of computed tomographic angiography and catheter angiography was not provided. Were all the computed tomographic angiograms of high quality? Use of an imperfect gold standard often magnifies measurement error, leading to misclassification bias, with unpredictable effects on calculation of sensitivity and specificity. In addition, what was the time interval between performance of the TCD and computed tomography examinations? Given the dynamic nature of the vasospastic process over time, a 24-hour interval between studies has typically been used.2,6,7⇓⇓ Third, the study population was heterogeneous and included patients with spontaneous and traumatic subarachnoid hemorrhage and 1 patient with an arteriovenous malformation. Thus, the presence, location, and grade of subarachnoid clot, as well as resultant hemodynamic aberrations, may have varied between subgroups in the study population. As a result, it may be more appropriate to split the study population into homogeneous subgroups (ie, spontaneous and traumatic subarachnoid hemorrhage, excluding arteriovenous malformations), perform this type of analysis in each subgroup, and develop disease-specific diagnostic criteria using the appropriate gold standard.2,7–9⇓⇓⇓ Fourth, were the sonographers or clinicians blinded to the angiogram results when selecting which TCD result was used to compare with the angiogram? If not, then bias may have been introduced into all reported calculations. Fifth, selection of the computed tomographic angiogram with the smallest BA diameter for correlation purposes in patients who had multiple angiograms introduces a biased assessment of test accuracy, since only the “best” data are used. It is preferable to use all available data, ie, multiple correlations in patients with multiple gold standard examinations, when evaluating test accuracy. Sixth, it appears that a BA/EVA ratio >2 only slightly improves the specificity (85% to 95%) of TCD for the diagnosis of BA vasospasm in this data set. Whether this is a reflection of different study populations, differences in study methodology, variation in pathophysiologic derangements, differences in cerebral perfusion pressures, different concomitant treatments, or other reasons may be determined by further research.
- ↵Soustiel JF, Shik V, Shreiber R, Tavor Y, Goldsher D. Basilar vasospasm diagnosis: investigation of a modified “Lindegaard Index” based on imaging studies and blood velocity measurements of the basilar artery. Stroke. 2002; 33: 72–78.
- ↵Sloan MA, Burch CM, Wozniak MA, Rothman MI, Rigamonti D, Permutt T, Numaguchi Y. Transcranial Doppler detection of vertebrobasilar vasospasm following subarachnoid hemorrhage. Stroke. 1994; 25: 2187–2197.
- ↵Sloan MA. Transcranial Doppler monitoring of vasospasm after subarachnoid hemorrhage.In: Tegeler CH, Babikian VL, Gomez CR, eds. Neurosonolog. St. Louis, Mo: BC Decker/Mosby; 1995: 156–171.
- ↵Sloan MA, Wozniak MA, Macko RF. Transcranial Doppler and subarachnoid hemorrhage.In: Babikian VL, Wechsler LR, eds. Transcranial Doppler Ultrasonography. 2nd ed. Boston, Mass: Butterworth-Heinemann; 1999: 109–127.
- ↵Lysakowski C, Walder B, Costanza MC, Tramer MR. Transcranial Doppler versus angiography in patients with vasospasm due to a ruptured cerebral aneurysm: a systematic review. Stroke. 2001; 32: 2292–2298.
- ↵Sloan MA, Zagardo MT, Wozniak MA, Macko RF, Aldrich EF, Simard JM, Rigamonti D, Jones D, Deaver R, Mathis JM. Sensitivity and specificity of flow velocity ratios for the diagnosis of vasospasm after subarachnoid hemorrhage: preliminary report.In: Klingelhofer J, Bartels E, Ringelstein EB, eds. New Trends in Cerebral Hemodynamics and Neurosonology. Amsterdam, Netherlands: Elsevier Science; 1997: 221–227.
I read with interest the comments made by Dr Sloan in his letter to the editor. In general, I found in these comments some puzzling contradiction as they directly dispute the reliability and the accuracy of the findings presented in our article, while at the same time the author claimed to have obtained similar findings using a method bearing “a striking resemblance” to that described in our article. Although I was not aware of the existence of the quoted article since it was not published in a peer-reviewed journal and does not appear in medical publication listings, I think that this would only add to the clinical value of the presented BA/EVA velocity ratio, especially when similar findings based on a similar protocol were obtained by a different team using different imaging methods.
The first question raised by Sloan addressed the insonation depth chosen for the extracranial segment. Three different and simple arguments stood behind this. The first relates to depth measurements that were made in numerous CT scans and averaged, leading to a range of 45 to 50 mm in most instances. Second, most authors would agree on an average minimal insonation depth of 60 mm for the intracranial segment of the vertebral artery. Last but most important, we found that a progressive reduction of the insonation depth resulted in an abrupt decrease in the flow velocity that is most likely consistent with the larger diameter of the vertebral arteries in their extracranial segment. I would agree, however, that unlike the extracranial segment of the internal carotid artery, the distinction may be difficult, just as it is for the basilar and the vertebral artery.
The second question related to the CT-angiography technique that was used in our study. Although it is indisputable that the conventional digital substraction angiography remains the gold standard for vessels imaging, high-resolution CT angiography has gained a wide acceptance, and there are numerous studies that the reader may refer to easily. Furthermore, since the purpose of CT angiography in our study was not interventional or surgical in nature, the obtained resolution was high enough for the evaluation of vessel diameters. Unless papaverine or balloon dilation is contemplated, it would have been difficult to justify an arterial angiography for the mere diagnosis of vasospasm, especially in critically ill patients, as were some of our head-injured patients. Clearly, the improvement in the resolution of CT angiography offers a more flexible diagnostic attitude in such indications. TCD in all patients was performed immediately before the CT angiography in all instances.
The third concern addressed the heterogeneity of our patients’ population. Although it is a widely accepted convention to deal with a homogeneous population, this is not immediately relevant to the present study. On the contrary, it is most likely that different kinds of pathologies may raise different clinical situations. Hyperemia, for instance, is uncommon following aneurysmal rupture. Consequently, it is not surprising, as specifically discussed in our article, that the issue of false-positive TCD was mostly overlooked in the article by Sloan et al,1 as they evaluate patients suffering from spontaneous subarachnoid hemorrhage. Furthermore, should a velocity ratio be reliable enough, it should apply to any situation where hyperemia is to be differentiated from vasospasm. Similarly, the Lindegaard index is applied to posttraumatic as well as to aneurysmal subarachnoid hemorrhage and, as such, has gained wide acceptance.
The fourth concern has been actually addressed in detail in the methods section of our article.
The fifth comment relates to the choice of the narrowest location for the measurement of vessel diameter. The answer is dictated by simple hydrodynamic considerations. The flow in a stenotic tube would be rather determined by the point of higher resistance and smallest diameter (“bottle neck”) than the contrary. Furthermore, the increase in flow velocities is a direct consequence of the decrease in the vessel diameter. It is therefore only logical to correlate to highest recorded flow velocity to the narrowest stenosis. A correlate and good example of this would be the fundamental principles of ultrasonographic diagnostic of carotid stenosis. However, I agree, as most would, that any particular study should better rely on the “best data” than on the worst.
The sixth point is the essence of our article itself. The most important outcome of the use of such a ratio is an improved differential diagnosis between hyperemia and vasospasm, which may be potentially useful in head injury. In the present series, despite the selection of the “best data” (narrowest diameter), 3 patients had elevated flow velocities in the basilar artery without concurrent radiological evidence of vasospasm. Two had a BA/EVA ratio suggestive of hyperemia, hence the potential clinical value of such a parameter.
In conclusion, I think that there is no such clinical study that is not subject to some degree of criticism, and I hope that the methodological limitations of the present study will eventually trigger the design of improved research tools rather than skepticism.