Comparison of oxygen-15 PET and transcranial Doppler CO₂-reactivity measurements in identifying haemodynamic compromise in patients with symptomatic occlusion of the internal carotid artery

The risk of recurrent ischaemic stroke in patients who present with transient ischaemic attack (TIA) or ischaemic stroke, associated with an internal carotid artery (ICA) occlusion, may be as high as 12% per year in case of a demonstrated compromised flow to the brain [1‐3]. Because of this increased risk, revascularization surgery has been considered in these patients. ‘The Carotid Occlusion Surgery Study’ used PET to select patients with symptomatic ICA occlusion and high oxygen extraction fraction (OEF) for inclusion in their study that aimed to investigate whether extracranial-intracranial (EC/IC) bypass surgery prevents recurrent stroke [4]. The results showed that the 2-year risk of ipsilateral stroke did not differ between the surgical and non-surgical group (p = 0.78) despite an improvement in OEF ratio and a bypass patency of 96% at the last follow up [5]. Other types of revascularization, such as carotid endarterectomy of the contralateral ICA or of the ipsilateral external carotid artery, or surgery or stenting of the vertebral artery, for patients with symptomatic ICA occlusion may be considered as well, although firm evidence that these treatments reduce the risk of stroke is lacking [1, 6‐8].

The haemodynamic state of the brain can be subdivided into stage 0, the normal flow state of the brain; stage 1, the phase of vasodilatation to compensate for a decrease in blood flow towards the brain through cerebral autoregulation; and stage 2, the phase of autoregulation failure, with a compensatory rise in OEF [9]. Haemodynamic compromise stage 2 can only be demonstrated by positron emission tomography (PET) using oxygen-15-labelled tracers [10]. Previous oxygen-15 PET studies of patients with symptomatic ICA occlusion have shown a 2-year risk between 25% and 75% of recurrent ischaemic stroke in those with an increased OEF in comparison with a risk between 5% and 10% in patients without an increased OEF [2, 3]. A disadvantage of PET scans with O-15 tracers is that this technique is not widely available and has a failure rate between 20% and 40% for obtaining complete quantitative data, mostly due to technical difficulties [2, 9, 11, 12]. A widely available and cheap alternative for identification of patients with haemodynamic compromise is transcranial Doppler (TCD) with measurement of cerebrovascular reactivity [13, 14]. Although cerebrovascular reactivity cannot identify patients with haemodynamic stage 2, it allows distinction between normal and a compromised haemodynamic state. Little is known about the agreement between oxygen-15 PET and TCD CO₂-reactivity. If these two methods would identify the same patients as being at risk of future stroke, clinical trials may not need to be restricted to centres with PET facilities. The purpose of this study was to investigate whether, in patients with recent TIA or stroke associated with ICA occlusion, oxygen-15 PET parameters and TCD CO₂-reactivity identify the same patients as having an impaired flow state of the brain.

Clinical characteristics and angiogram findings of the 24 patients are shown in Table 1. The mean time between last symptoms and TCD was 33 ± 25 days. The median time between TCD and PET was 6 days (range 1 to 39). Seventeen patients had TCD before PET, and seven patients had TCD 7 days at the most after PET. Seventeen patients (71%) had impaired TCD CO₂-reactivity, of whom three had absent CO₂-reactivity and three showed a steal phenomenon. PET studies indicated that haemodynamic compromise was present in 15 patients (63%), of whom 13 patients were classified as haemodynamic stage 1 (with decreased CBF in 10, increased CBV in 3, and both decreased CBF and increased CBV in 2 patients) and 2 patients were classified as stage 2. As an example, Figure 1 shows parametric images of a patient with stage 2 haemodynamic failure. In 12 of the 24 patients (50%), TCD and PET showed agreement in haemodynamic status. In 10 of these 12 patients, haemodynamic compromise was demonstrated by both TCD and PET. In two patients, the flow state of the brain was normal in both TCD and PET. Of the 12 patients with different haemodynamic assessment in TCD and PET, 7 patients with impaired CO₂-reactivity (−4%, 0%, 0%, 5%, 11%, 15%, and 19%, respectively) had a normal haemodynamic state (stage 0) in PET. Of five patients with normal CO₂-reactivity, 4 were classified as stage 1 and one as stage 2 based on PET findings. The other patient with stage 2 haemodynamic failure based on PET showed steal on CO₂-reactivity (−12%). In patients with agreement between PET and TCD, the median change in systolic blood pressure during TCD was 0 mmHg (interquartile range (IQR) 25), which did not differ from the median change in systolic blood pressure of 5 mmHg (IQR 10) in those without agreement (p = 0.850). The median change in diastolic blood pressure in patients with agreement between PET and TCD (0 mmHg, IQR 5) did not differ from the median change in those without agreement (3.5 mmHg, IQR 10; p = 0.257).

In Table 2, PET findings are compared between patients with and without impaired CO₂-reactivity and between patients with and without absent reactivity or steal as measured by TCD. No significant differences were found in absolute CBF, OEF, and CBV/CBF values between patients with impaired and normal CO₂-reactivity and between patients with and without absent CO₂-reactivity or steal. A prolonged mean transit time in the symptomatic hemisphere in relation to the asymptomatic hemisphere (CBV/CBF hemispheric ratio) was found to be associated with an impaired CVR (p = 0.035).

Figure 2 illustrates that, on visual inspection, ipsilateral/contralateral ratios of CBF, OEF, and CBV/CBF corresponded better with CO₂-reactivity than absolute ipsilateral values of CBF, OEF, and CBV/CBF. Nevertheless, there was no significant correlation between CO₂-reactivity and CBF hemispheric ratio (Pearson r = 0.168, p = 0.432), OEF hemispheric ratio (Spearman r = −0.242, p = 0.255), and CBV/CBF hemispheric ratio (Pearson r = −0.368, p = 0.077). Remarkably, three patients with high CO₂-reactivity (indicated by grey dots in Figure 2) were the main reason for the poor correlations. Two of them were males aged 45 and 73 years, and the female patient was 76 years old. The number of days between TCD and PET was at most 7 days in these three patients. Only in one of the three patients, blood pressure during TCD increased with 50 mmHg systolic and 20 mmHg diastolic; in the other two, the systolic blood pressure remained stable or decreased with 15 mmHg, and the diastolic blood pressure remained stable. After excluding these three patients in a post hoc analysis, significant correlations were obtained between CO₂-reactivity and CBF ratio (Pearson r = 0.563, p = 0.008), OEF ratio (Spearman r = −0.542, p = 0.011), and CBV/CBF ratio (Pearson r = −0.677, p = 0.001).

The main finding of this study is that, in patients with symptomatic ICA occlusion, identification of the presence or absence of haemodynamic compromise by oxygen-15 PET and TCD CO₂-reactivity corresponds in only half of the patients. This is important as the presence of haemodynamic compromise has been shown to predict recurrent stroke [1‐3] and may play a role in the decision whether or not to advise revascularization surgery. We did not find a difference in the change in blood pressure during TCD between patients with and without agreement between PET and TCD. Therefore, it is unlikely that an increase in blood pressure during TCD resulting in an overestimated CO₂-response is responsible for the incongruent findings between PET and TCD. After excluding three patients with high CO₂-reactivity, a correlation with the various ipsilateral/contralateral PET ratios was found, albeit only at a moderate level. It is not clear why these three patients were outliers as age or time between TCD and PET were not different from the other patients.

Previous studies investigating the relationship between cerebrovascular reactivity and CBF or OEF measured by PET also included patients with other types of vascular obstruction other than ICA occlusion and reported conflicting results (Table 3). Only one previous study made the same comparison as we did between TCD CO₂-reactivity with oxygen-15 PET, but they selected 22 patients with hypertension and diabetes. In concordance with our findings, this study did not find an association between increased OEF and decreased TCD CO₂-reactivity [24]. Other studies have reported an association between impaired cerebrovascular reactivity to hypercapnia or acetazolamide measured with SPECT or stable-xenon CT studies and increased PET OEF [25‐29].

The results of the present study indicate that, in patients with recently symptomatic ICA occlusion, TCD measurements of the haemodynamic state of the brain cannot be replaced by oxygen-15 PET and vice versa for identification of patients with haemodynamic compromise. This may be explained by important differences between TCD and oxygen-15 PET. First, TCD measures blood flow velocity in the MCA itself, whereas PET measures perfusion at the level of the brain tissue. The method of TCD relies on the assumption that changes in flow velocity are directly proportional to changes in CBF. For that to be true, the cross-sectional area of the insonated artery needs to remain constant [30, 31]. In contrast, PET directly measures CBF, CBV, and OEF, which are components of the autoregulation itself. In addition, TCD measures changes in flow velocity in response to hypercapnia, whereas PET, according to the scan protocol in this study, measures haemodynamic parameters at rest, i.e., without vasodilatory stimuli. Of the previous studies that compared the CBF response after a vasodilatory stimulus measured by SPECT or stable-xenon CT with TCD reactivity [32‐35], one study of 38 patients with ICA stenosis or occlusion found only a weak relation [32], and others reported moderate to good correlations between CBF response after a challenge and TCD cerebrovascular reactivity [33‐35]. It is possible that the agreement between TCD and PET in our study would have been better if a vasodilatory stimulus had also been included in the PET studies. An advantage of PET is that it does not only provide information on perfusion in the MCA territory but also on the flow state of the other vascular territories of the brain. TCD and PET do not only measure different physiologic parameters but also use quite different techniques. As a gold standard for measuring the flow state of the brain is not yet available, the sensitivity and specificity of TCD and PET cannot be determined [1, 10].

A strength of the present study in comparison with previous reports is that a more homogenous population of patients was included as all had recent symptoms of the hemisphere ipsilateral to an ICA occlusion. In addition, 92% of patients had ischaemic symptoms after documented occlusion, which is an important clinical risk factor for recurrent ischaemic stroke [36]. Furthermore, in contrast to some previous studies [24, 25], the majority of patients in our cohort showed haemodynamic compromise measured by TCD or PET. It is in this subcategory of patients that accurate identification of haemodynamic compromise has implications for prognosis and possible treatment decisions.

This study has some limitations. First, the patients were not investigated by TCD and ¹⁵O-PET on the same day. Compromised cerebral perfusion can improve over time [37]. Although the majority of patients (88%) had TCD and PET within 2 weeks, the time interval between TCD and PET may have contributed to the poor agreement.

Second, the classification in haemodynamic stages may be a matter of debate. In the model of Derdeyn et al. [9], stage 1 was defined as a slight decrease of CBF and slight increase of OEF with or without increase of CBV, but a cutoff value for a slight difference was not provided. In another clinical PET study, the patients were only divided into two groups with a normal or increased OEF [3]. We defined stage 1 as a decreased CBF or increased CBV as both are signs of autoregulation. This study shows that both the comparison of haemodynamic stages by PET and TCD, as well as absolute values obtained by PET and TCD, did not show agreement.

Third, ten patients had an additional stenosis or occlusion in the contralateral ICA, which may have influenced the results in hemispheric ratios. Fourth, this study has a relatively small sample size. However, previous PET studies in relation to cerebrovascular reactivity were even smaller consisting of at most 19 patients with ICA occlusion [25].

The present study shows that, in patients with symptomatic ICA occlusion, identification of presence or absence of haemodynamic compromise by oxygen-15 PET and TCD CO₂-reactivity corresponds in only half of patients. In future trials, PET and TCD CO₂-reactivity measurements cannot be used according to local preference to identify patients with haemodynamic compromise for study inclusion.