We analyzed a subset of the Carotid Plaque Imaging in Acute Stroke (CAPIAS) study (ClinicalTrials.gov: NCT01284933) that prospectively recruited patients with acute ischemic stroke in the territory of one carotid artery and unilateral or bilateral carotid artery plaques with a plaque thickness of at least 2 mm at any side. Patients with carotid artery stenosis of ≥70%, according to NASCET criteria, were excluded. CAPIAS was conducted at four large tertiary stroke centers in Germany: Ludwig Maximilian University Munich, Technical University of Munich, Eberhard Karl University of Tuebingen, and Albert Ludwig University of Freiburg. The study design and inclusion criteria have been published previously (5, 20). The study was approved by the responsible ethics committees, and all participants provided written informed consent.

In our present analysis, we included only carotid arteries with plaques and MRI data with sufficient image quality to assess bifurcation geometry.

All participants were examined with a 3-Tesla MRI scanner (Magnetom Verio, Skyra, Tim Trio, Prisma or Biograph mMR, Siemens Healthineers, Erlangen, Germany) and a four-channel surface coil (Machnet, Eelde, Netherlands). We obtained time-of-flight (TOF) MR angiographs and axial pre- and postcontrast black-blood T1-, PD-, and T2-weighted sequences with fat suppression and an in-plane resolution of 0.5 mm² as described previously (5, 20, 21). Carotid plaque composition was classified according to the modified AHA-LT criteria (22) by two experienced radiologists (TS and AS) blinded to the patients' characteristics (5). Plaques were defined as AHA-LT III if they showed a small eccentric plaque without calcification or a diffuse intimal thickening; AHA-LT IV/V plaques were characterized by a lipid-rich necrotic core surrounded by fibrous tissue with possible calcifications; complicated AHA-LT VI plaques were defined by the presence of a surface defect, intraplaque hemorrhage, or thrombus; and AHA-LT VII plaques were calcified plaques. There were no AHA-LT VIII plaques without lipid core and with possible small calcifications in our study sample (22). Wall area measurements were performed with a custom-designed semiautomatic image analysis tool (CASCADE, University of Washington, Seattle, Washington) (23).

TOF data sets were imported into CaroTo, an extension of the MEVISFlow research software (Fraunhofer MEVIS, Bremen, Germany) (24). A centerline was created in each carotid bifurcation after the manual definition of CCA, ICA, and external carotid artery (ECA) as start- and endpoints for centerline computation. Consecutively landmarks including the flow diverter (FD) and ICA were marked (Figures 1A,B). From these landmarks, predefined points were automatically generated on the centerline, indicating the level of each analysis plane. Analysis planes were manually corrected to align perpendicular to the carotid lumen if necessary. This resulted in six cross-sectional planes (Figure 1C): plane 1 on the CCA centerline 1 cm below the flow diverter (CCA1); plane 2 within the carotid bulb between CCA1 and the flow diverter (CCA2); and planes 3–6 along the ICA with the starting point at the FD and oriented perpendicularly to the centerline with a spacing of 3 mm each (ICA1–ICA4). Geometric parameters have been previously described and successfully applied (7, 8, 25) and included (a) ICA/CCA diameter ratio; (b) bifurcation angle (measured by two tangential lines of the first 1 cm of the outer wall of the ECA and ICA starting at the FD); and (c) CCA-ICA tortuosity (ratio of the centerline connection and the direct line of plane CCA1 and plane ICA4; Figure 1C). All analyses of carotid geometry were performed by one experienced rater (LT). In cases of uncertainty, another experienced rater (CS) was consulted for the final decision.

Data are presented as mean and standard deviation or median (interquartile range) for continuous variables and as absolute and relative frequencies for categorical variables. Deviations from normality were determined using the Shapiro–Wilk statistic. Depending on data distribution, two-tailed t-tests or nonparametric tests were applied as appropriate for continuous variables. To detect statistically significant relations between categorical variables, a chi-square test was applied.

We used mixed-effects logistic regression models to study associations between geometric parameters and cCAPs and account for the clustering of carotid arteries per patient. For regression analysis, the geometric parameters were standardized. We used the following geometric parameters as predictors: ICA/CCA ratio, tortuosity, and bifurcation angle. Models were constructed with patient identifiers as random effects, geometric parameters as fixed-effects independent variables, and cCAP as a dependent variable. In the first step, we adjusted for age, sex, and wall area (model 1), and in the second step, we adjusted additionally for cardiovascular risk factors including hypertension, diabetes mellitus, hypercholesterolemia, and smoking (model 2). Model 3 included model 2 and all three geometric parameters. All models were adjusted for wall area because of the potential interaction of pre-existing atherosclerotic changes with lumen geometry. A two-tailed p-value < 0.05 was considered to indicate statistical significance. Odds ratios (ORs) and 95% confidence intervals were calculated. All analyses were performed with IBM SPSS Statistics 28.

From 392 carotid arteries ipsilateral and contralateral to the ischemic infarct in 196 patients in the CAPIAS study, 354 carotid arteries (182 patients) were included in the present analysis. Carotid arteries with insufficient image quality/incomplete TOF coverage (n = 28) or absence of plaques (n = 10) were excluded (Figure 2). Baseline characteristics are provided in Table 1. Complicated plaques were found in 70 carotid arteries (19.8%). Prevalence of noncomplicated plaques was AHA-LT III in 110 (31.3%), AHA-LT IV/V in 68 (19.2%), and AHA-LT VII in 106 (29.9%) carotid arteries. Table 2 compares the geometric parameters and wall areas between carotid arteries with and without complicated plaques.

The presence of cCAPs, i.e., AHA-LT VI plaques, was significantly associated with a low ICA/CCA ratio (OR per SD increase 0.59 [95%CI: 0.42–0.83]; p = 0.003) and a low bifurcation angle (OR per SD increase 0.57 [95%CI: 0.39–0.83]; p = 0.003) after adjusting for age, sex, and wall area (model 1). However, such an association was not existent for carotid tortuosity. These findings remained significant for the ICA/CCA ratio (OR per SD increase 0.60 [95% CI: 0.42–0.85]; p = 0.004) and bifurcation angle (OR per SD increase 0.61 [95%CI: 0.42–0.90]; p = 0.012) after additionally adjusting for cardiovascular risk factors (model 2), indicating that a carotid artery with a smaller lumen expansion of the carotid bulb (i.e., a lower ICA/CCA ratio and thus a steep vessel tapering) in relation to the CCA and a straighter course (lower bifurcation angle) is connected with a higher probability of cCAPs. By integrating all three geometric parameters (model 3), only the ICA/CCA ratio remained significant (OR per SD increase 0.65 [95%CI: 0.45–0.94]; p = 0.023), indicating that smaller lumen expansion may be the strongest factor associated with cCAPs. Results of mixed-effects logistic regression analysis regarding geometric parameters (i.e., ICA/CAA ratio, bifurcation angle, and carotid tortuosity) and their association with cCAPs are given in Table 3. The key findings for two carotid arteries with and without a cCAP are illustrated in Figure 3.

Jiang et al. (17) demonstrated that a smaller luminal expansion of the carotid bifurcation, i.e., a lower “flareA” showed the strongest association with “vulnerable plaques” in Chinese patients. Vulnerable plaques in their study corresponded to AHA-LT VI and IV/V plaques with a large lipid-rich/necrotic core. In our study on Caucasian stroke patients, we used the ICA/CCA ratio, while Jiang et al. used “flareA,” the diameter ratio between distal and more proximal segments in the CCA, in a Chinese cohort. Despite these differences, we revealed similar findings; namely, that a smaller luminal expansion in the more distal segments of the carotid artery in relation to the CCA is significantly associated with complicated carotid artery plaques. Thus, our study provides further evidence that individual carotid geometry is associated with plaque composition and plaque instability. In conclusion, there is now evidence from almost 700 stroke patients from Europe and Asia indicating that carotid geometry, especially smaller lumen expansion of the internal carotid artery bulb with steep tapering of the vessel into the distal ICA, is an independent and so far under-recognized predictor for rupture-prone atheroma in the ICA.

A steep tapering of the carotid bulb into the ICA, together with a straight vessel course due to a low carotid bifurcation angle, influences the bloodstream and is likely to result in a local acceleration of blood flow velocity and pressure. Accordingly, such geometry results in an environment with high local WSS and high pulse pressure acting on the proximal shoulder of the wall and especially on the plaque surface in the case of a significant plaque protrusion. Unfortunately, we cannot provide data on hemodynamics because 4D flow MRI was not part of the protocol, but the interaction of geometry and hemodynamics has been demonstrated in several previous studies (8, 25, 26).

Carotid bifurcations with larger area ratios, i.e., larger ICA/CCA ratios, are more susceptible to low and oscillatory WSS, the so-called disturbed flow, which can be mitigated by a more curved, respectively more tortuous, vessel course (16, 26). Following flow physiology, it is plausible that a smaller lumen expansion, i.e., a steeper tapering of the vessel, reduces disturbed flow and increases WSS accordingly (17). As pointed out, the association between geometry-derived low and oscillating shear stress (8, 9, 26) is well established for the early stages of atherosclerosis. Similarly, there is growing evidence that high WSS plays an important role in the advanced stages of atherosclerosis (12, 27, 28). Combining cellular responses (including plasmin-induced metalloproteinase activity) and high WSS may lead to the degradation of matrix components, smooth muscle cell apoptosis, and reduced matrix synthesis, possibly resulting in intraplaque hemorrhage, fibrous cap thinning, and thrombus formation (10, 27, 29), as shown by Tuenter et al. (12).

As outlined before, carotid geometry also influences mechanical forces of the pulsatile blood flow, especially at the proximal shoulder of plaques and vessel stenosis. These interactions can be measured as plaque stress and tensile stress using finite-element analysis (14, 29–31). It is highly plausible that the steep tapering of the ICA in patients with complicated plaques in our cohort (i.e., low ICA/CCA ratios and low bifurcation angles) leads to high velocities that impinge on the atheroma. Accordingly, high WSS and high pressure, which are the highest in the region of maximal stenosis, act on the wall of the proximal plaque shoulder. On this side, plaque surface and volume deform with each cardiac cycle, predisposing to endothelial leakage, thinning of the fibrous cap, and intraplaque hemorrhage due to shear forces followed by vessel breakage within the plaque (15, 32, 33). This cascade leads to plaque instability and increases the risk of plaque rupture and stroke. This is supported by the observation of Jiang et al. (17) and Lu et al. (34), who demonstrated that vulnerable plaques occurred more frequently distal to the flow diverter in the ICA, i.e., at the tip or the narrowest section of the funnel.

We must consider some limitations of our study. First, we included carotid arteries with mild-to-moderate stenosis. More advanced atherosclerotic changes may potentially alter the geometric parameters. Thus, we corrected for wall area in our models. Even after adjusting for wall area, the association of carotid geometry and complicated plaques remained significant, suggesting that there must be other factors explaining this association beyond wall area and stenosis. In addition to the pathophysiological perspective, including mild-to-moderate stenosis is interesting from a clinical point of view, given that the prevalence of complicated plaques in symptomatic arteries is threefold higher in patients with cryptogenic stroke (5). Second, we could not interrelate geometry with WSS and plaque type to confirm our pathophysiological explanations since we did not directly measure blood flow; this would have required 4D flow MRI or CFD analysis and should have been done in future studies. Third, our geometry measurements based on 3D TOF images might be hampered due to local irregular blood flow patterns such as flow separation and recirculating flow. Finally, due to the cross-sectional study design, we can only show an association but cannot prove causality. Here, further studies with a longitudinal design are needed.