Internal carotid artery occlusion (ICAO) can cause recurrent ipsilateral stroke due to hemodynamic failure of the cerebral collateral supply (1). In the case of ICAO, the primary collateral circulation based on the circle of Willis (CoW) is first activated (Supplementary Figure 1). However, if this primary compensatory route is insufficient, recruitment of secondary collaterals including leptomeningeal vessels occurs, indicating inadequate circulation through the CoW (2). Hence, patients with ICAO often require prolonged observation and intense blood pressure management. Some may even undergo revascularization procedures in cases of recurrent symptoms (3). However, identifying patients at particular risk of recurrent ischemia is challenging. Current methods to detect hemodynamic failure are MRI-based cerebrovascular reactivity (CVR) measurements or (15O)-H2O-positron emission tomography (PET) including a vasodilator stimulus (4–6). Collateral flow in patients with ICAO is accessible through transtemporal insonation using transcranial duplex sonography (TCD). Increased compensatory flow efforts of activated leptomeningeal collaterals result in higher flow velocities in the ipsilateral PCA-P2 segment (7), and have been implicated as indicators of hemodynamic failure (8, 9). Peak systolic flow values >85 cm/s in the ipsilateral PCA-P2 segment indicated hemodynamic failure stage 2 on the affected hemisphere in blood oxygen level dependent (BOLD)-CVR measurements, with an accuracy of 95% (8, 9). Our goal was to assess to what extent TCD-measurements of flow in the ipsilateral PCA-P2, alone or in combination with clinical parameters, predict risk of recurrent ischemic events in symptomatic ICAO patients.

Out of 870 patients with symptomatic ICAO, we included 111 (see Figure 1). Patients received a TCD at a median time of 1 day after admission [IQR 1], and were observed during a median follow-up of 109 days [IQR 37]. Thirteen (11.7%) patients suffered a recurrent ischemic event within the same vascular territory at a median time of 6 days [IQR 26]. Previous TIA as well as pre-stroke use of anticoagulants and statins were more common in the group with recurrent ischemic event (shown in Table 1). There was no difference in the type of acute stroke treatment between patient with or without recurrent ischemic event. There was a tendency for more atherosclerotic causes (TOAST I) of stroke in the group with recurrence, however, this difference was not significant (p = 0.209). In patients with ICAO classified as other determined etiology (TOAST IV), these were carotid artery dissection (n = 22), vasculitis (n = 1), myeloproliferative disease (n = 1), and fibromuscular dysplasia (n = 1).

Ipsilateral PCA-P2 peak systolic velocity (PSV) values were significantly higher in patients with recurrent ischemia (86 cm/s [IQR 41] vs. 60 cm/s [IQR 26], Figure 2). In addition, the NIHSS on admission was lower in the group with recurrence (1 [IQR 3] vs. 5 [IQR 12]). We found no difference regarding functional disability in both groups on admission or follow up (shown in Table 1).

In order to test the variables “previous TIA,” NIHSS on admission and PSV in the ipsilateral PCA-P2-segment for their ability to predict recurrent ischemia, we fit logistic regression models including these variables alone or in combination (shown in Supplementary Figure 2 and Table 2). PSV in the ipsilateral PCA-P2 segment was indeed a good predictor for recurrent ischemic events (for 86 cm/s: sensitivity of 46.5%, specificity of 91.8%, AUC of 0.79). However, the model including all three co-variables (NIHSS on admission, previous TIA and PSV of the ipsilateral PCA-P2 segment) performed better in predicting ischemia recurrence (accuracy of 89.2% and AUC of 0.91).

To exemplify use of this model, in order to predict the probability for ischemia recurrence in a new patient with ICAO, we use:

In a patient with a PCA-P2 Flow of 65 cm/s, no previous TIA and an NIHSS admission of 4 would have an estimated risk of ischemia recurrence of 2.5%:

A patient with PCA-P2 Flow of 90 cm/s, previous TIA and an NIHSS admission of 2 would have an estimated risk of ischemia recurrence of 82.1%:

Our data show that in patients with symptomatic ICAO, we can forecast the risk of an ipsilateral recurrent ischemic event by history of TIA, a low NIHSS on admission and high flow values in the ipsilateral P2-segment of the PCA with an AUC of 0.91. These results build on existing evidence that increased flow velocity in PCA-P2 after symptomatic ICAO indicates intensified compensatory efforts, associated with risk of hemodynamic failure (8, 9).

In patients with symptomatic ICAO, anterior cerebral artery (ACA) and PCA-P1 segment are considered primary routes of the collateral network (12). P2-supplied leptomeningeal collaterals are considered secondary routes, because they are less efficient (2) (shown in Figure 1). Therefore, increased flow within the P2 segment indicates exhaustion of primary collateral pathways via the CoW as an indicator of hemodynamic compromise.

The correlation of previous TIA with ischemia recurrence was shown previously, highlighting the importance of clinical symptoms in predictive models (8, 13–17). In our sample, patients with a low NIHSS score on admission had a higher risk of recurrence. This appears counterintuitive, since low NIHSS scores indicate less tissue damage (18). One explanation for this finding could be that in severely affected patients, minor recurrent ischemic events due to hemodynamic failure may not cause clinically apparent worsening and thus, may be missed. Since our study included stroke patients with rather minor deficits on admission (median NIHSS of 4) resulting in underrepresentation of a more severe stroke population.

Choosing the best preventive strategy after ischemia has recurred in patients with ICAO is challenging. In our patients, prevention was up to the preference of the treatment team and included addition of a second platelet inhibitor, anticoagulation, or interventional procedures, such as extra-intracranial (EC-IC) -bypass surgery. While our sample size is too small to derive meaningful conclusions about the best treatment concept, our data strongly support that more studies are needed to establish therapeutic recommendations.

Overall, our results support the hypothesis that TCD PCA-P2 measurements can help detecting patients at an increased risk of recurrent ischemic events. TCD is more readily available as a screening tool and can identify high risk ICAO patients that will need a more comprehensive hemodynamic imaging work-up with for instance, BOLD-CVR or (15O-)H2O-PET. However, if used in conjunction with these advanced imaging readouts of CVR, TCD and clinical variables are likely to further increase prediction accuracy, which should be validated in prospective studies. Several variants of the CoW exist in healthy subjects as well as in patients with stroke. While we did not detect an association of CoW variants with outcome in patients with stroke due to MCA-M1 occlusion previously (19), such variants may well affect flow in the PCA-P2 segment. However, even without further consideration of CoW variants in our study cohort, P2 flow was highly associated with recurrent hemodynamic stroke in our study. This facilitates use of this easily derived non-invasive TCD parameter for the detection of hemodynamic impairment in stroke patients. One limitation of our analysis is the small patient cohort. Furthermore, the strict exclusion of patients with other moderate or high-grade intra- or extracranial vessel pathology limits generalizability of our findings. Validation in a larger, multicenter cohort would improve generalizability of our findings. Standardization of TCD measurements is required to reduce examiner-dependent variability. However, if angle correction is omitted, interrater reliability of TCD flow assessment in the PCA-P2 is high.

To conclude, in patients with symptomatic ICAO, increased ipsilateral TCD PCA-P2 flow velocity together with history of TIA and low admission NIHSS indicate an increased risk of recurrent ischemia with high accuracy. TCD can aid in non-invasive assessment of hemodynamic impairment, and should be included into clinical algorithms to improve detection of patients at risk.

Anonymized raw data can be made available from the corresponding author by reasonable request.

The studies involving human participants were reviewed and approved by KEK-ZH (Ethics Committee of the Canton Zurich, Switzerland). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

SW conceived and designed the study. SJ and RK wrote the first draft of the manuscript and were responsible for data collection and data analysis. CN, MS, and JF were involved in concept development and image analysis. LH performed statistical data analysis. AL and JF were involved in interpretation of study data. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

This project was supported by the Swiss National Science Foundation (PP00P3_202663), the UZH CRPP Stroke, the Baugarten foundation, and the P&K Pühringer Foundation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.