Introduction

Moyamoya disease (MMD) is a rare neurological condition marked by the gradual narrowing or blockage of the distal internal carotid artery (ICA). This progressive constriction leads to the formation of collateral vessels at the base of the brain in an attempt to maintain adequate blood flow. Although MMD is an uncommon disease, it is notably the leading cause of stroke in children of East Asian heritage1,2 Clinical manifestations of MMD commonly involve cerebral ischemia and intracranial hemorrhage3.

At present, the pathogenesis of MMD is still unclear, some studies have suggested that it may be related to multiple factors such as genetic inheritance, immune inflammatory response and other related pathological changes3,4,5 In the circulation, permanent exposure of endothelial cells (ECs) to WSS caused by blood flow can significantly affect the expression of different factors such as vascular endothelial growth factor, endothelial colony-forming cells and various other angiogenic factors, which in turn affect vascular remodeling. However, there is a lack of exploration of the role of hemodynamics and morphological in the progression of moyamoya disease6,7.

The influence factors of hemodynamic and morphological changes on the long-term prognosis of patients with moyamoya disease still poorly understood8,9. A recent study revealed a significant correlation between high relative Wall Shear Stress (rWSS) and the development of considerable risk of recurrent stroke in intracranial atherosclerosis disease patients10. Another study indicated that elevated Wall Shear Stress (WSS) serves as an independent predictor of stenosis progression and the onset of vascular complications in patients with non-occlusive Moyamoya disease (MMD)11. However, it remains uncertain whether relative Wall Shear Stress (rWSS) can reliably predict long-term outcomes in MMD patients following revascularization. Computational fluid dynamics (CFD) has been used to characterize the local hemodynamic features that contribute to the development of cerebral vascular diseases. Despite its potential, current research has not extensively investigated the hemodynamic mechanisms of the ICA in MMD using CFD analysis.

Therefore, we hypothesize that changes in the WSS of the terminal ICA are associated with long-term unfavorable outcomes in patients with MMD. To investigate this, we initiated a CFD-based study to explore the hemodynamic characteristics of the terminal ICA in MMD and their relationship with clinical manifestations and long-term prognosis.

Methods

Study design and participants

In this cross-sectional study, we examined participants from a single center, comprising Moyamoya disease (MMD) patients treated between July 2020 and December 2021. We reviewed a total of 240 consecutive patients diagnosed with MMD at Beijing Tiantan Hospital, encompassing both their clinical records and radiological data. The diagnosis of MMD was confirmed by cerebrovascular experts at our center using digital subtraction angiography (DSA), in accordance with the 2012 Japanese MMD Guidelines12. The inclusion criteria for this study included (1) Suzuki stage ≤ 3, absence of occlusion in the distal ICA in each hemisphere; (2) availability of complete preoperative head Computed Tomography Angiography(CTA), CT perfusion(CTP) for computational fluid dynamics (CFD) analysis and evaluation of the hemodynamic status of patients with MMD; (3) comprehensive medical and follow-up records. (4) the side of the hemisphere that underwent revascularization was included. After rigorous review, 129 patients and 173 cerebral hemispheres were included in the study. (Fig. 1) All participants have signed informed consent. The study received approval from the Ethics Committee of Beijing Tiantan Hospital (KY2016-048-01) and all methods were carried out in accordance with relevant institutional guidelines and regulations.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Flowchart of study participants.

Data collection

The study collected comprehensive data on various patient characteristics and past medical history, this included information on the patient’s age, gender, primary symptoms, and modified Rankin Scale (mRS) score at admission. Additionally, data on past medical history were recorded, encompassing conditions such as hypertension, hyperhomocysteinemia, smoking, drinking, and diabetes. The surgical methods employed were also documented, including indirect bypass, direct bypass, and combined bypass procedures, and perioperative stroke and Transient Ischemic Attack (TIA) events (including instances of intracerebral infarction and hemorrhage) These events were confirmed through brain imaging, specifically using CT scans and/or diffusion-weighted MRI, to identify definite new infarctions and hemorrhages. We categorized the primary symptoms into 3 main types: stroke-type (infarction, various hemorrhage), TIA-type, and nonspecific-type (including headache, epilepsy, and asymptomatic cases). Angiographic features of the Suzuki stage (Suzuki’s Vascular Criteria)13, on the surgical side were blindly evaluated by two experienced neurosurgeon(P.G,C.Z) using DSA. In cases with discrepancies in imaging findings, a third experienced neurosurgeon (Q.Z) performed a re-evaluation. All patients had blood samples collected upon diagnosis of MMD, and RNF213 p.R4810K variant was detected using the previous research described primer designation14. Two independent neurosurgeons qualitatively analyzed the preoperative CTP data in the hospital’s imaging system, collecting information on cerebral blood volume (CBV), cerebral blood flow (CBF), mean transit time (MTT), and time to peak (TTP) in the surgical hemisphere. Compared with the contralateral hemisphere at different levels, CBV and CBF were categorized as increased, decreased, or unchanged based on the color differences in the pseudo-color map, while MTT and TTP were defined as prolonged or unchanged. When there was a dispute regarding the color differences, perfusion parameter changes within ± 10% of the average value were defined as unchanged.

Treatment and clinical outcomes

In our hospital’s routine operation scheme, direct revascularization and combined revascularization were the preferred methods. However, in cases where the superficial temporal artery or middle cerebral artery was too slender for anastomosis, indirect surgical revascularization was selected. Priority should be given to revascularization procedures in the symptomatic hemisphere. Long-term efficacy was assessed through in-person and telephone follow-up conducted 24.66 ± 5.88 months post-discharge. Throughout the follow-up period, the mRS score was utilized to assess neurological status, with scores < 2 indicating favorable outcome and scores ≥ 2 indicating unfavorable outcome. Follow-up incidents included TIA, ischemic events, hemorrhagic events. Stroke events were defined as a new neurological deficit lasting more than 24 h, confirmed by MRI or CT scans to be associated with a new infarction or hemorrhage. Two experienced neurosurgeons (P.G and C.Z), blinded to the side of revascularization surgery and CFD reconstruction status, evaluated mRS scores and rates of TIA and stroke during follow-up15.

CTA-Based CFD simulation and morphology/hemodynamic parameters measurement

CTA was performed following admission and the data was imported into Mimics 19.0 to create a 3D model of the ICA for morphological-hemodynamic analysis. The 3D model was segmented and smoothed using Geomagic Studio software, and then imported into ICEM CFD to generate a volumetric grid for fluid dynamics simulation. Following Bouthillier et al.'s 1996 7-segment ICA segmentation approach, the C6-7 Segment was identified as the distal vascular region and meshed accordingly16. The maximum cell size at the inlet and outlet was set to 0.1, and 0.2 for other parts of the mesh, with each case having at least 1 million cells. The blood flow simulation utilized the following settings and assumptions: (1) Rigid, non-deformable walls and no-slip boundaries; (2) Outlet pressure set at 0 Pa. With each patient assigned a specific value based on Suzuki stage, individualized inlet blood flow velocity was determined by measuring the average blood flow velocity of the ICA vasculature using ultrasonography17; (3) In CFX-pre software, blood flow simulation is performed by solving the Navier–Stokes equation (4) Blood is treated as an incompressible Newtonian fluid with a constant viscosity of 0.0035 kg/(m*s) and a density of 1060 kg/m3.

Morphological parameters were assessed by two separate researchers (L.M. and C.Z) using the reconstructed 3D ICA model. The inlet and outlet cross-sectional areas were measured from the proximal (C1-segment) and distal (C7-segment) ICA. To assess tortuosity, we introduced the tortuosity index, which was calculated as the ratio of the actual length within the 3D space to the straight-line distance between the entrance and exit points. The tortuosity index was determined for both the overall internal carotid artery (ICA) and the terminal ICA independently.

$${\text{Tortuosity index}} = {\text{Actual length}}/{\text{Absolute length}}$$

In order to offset the influence of vascular morphology on WSS between patients, this study used the relative WSS (rWSS) and relative Pressure(rP) value18,19,20. Hemodynamic parameters (WSS and Pressure) at the terminal ICA and the overall ICA were obtained through CFD-post (Fig. 5). To assess the impact of elevated rWSS on the progression of MMD, we defined the high rWSS group (> 1.71) as individuals with rWSS values in the highest tertile.

Statistical analysis

The statistical analyses were conducted using SPSS (version 26.0) and R software (version 4.2.3). A complete case analysis was performed for all participants included in the study. Categorical variables were reported as frequencies, while continuous data were presented as either mean (with standard deviation, SD) or median (with interquartile range, IQR). To compare continuous variables between two groups, either t-tests or Mann–Whitney U tests were employed. For comparisons involving multiple groups, Kruskal–Wallis tests or one-way ANOVA were applied. Categorical data comparisons were carried out using Pearson’s chi-squared tests, Fisher’s exact tests, and Kruskal–Wallis tests.

Logistic regression analysis was conducted to determine which variables were associated with the long-term prognosis of MMD. Variables that achieved a p-value of less than 0.10 in the univariate analysis were included in the multivariate analysis. Three logistic regression models were employed to examine the impact of rWSS on the long-term prognosis of MMD, considering both continuous and categorical variables. The crude model represented the unadjusted regression model for rWSS. Model 1 was adjusted for age, gender, RNF213 variant, hypertension, diabetes, hyperhomocysteinemia, and primary symptom. Model 2 included all the variables in Model 1, plus admission mRS score, surgical modalities, perioperative events, Suzuki stage, rP, CTP parameters and morphological parameters. To assess the predictive performance of each regression model in identifying unfavorable outcomes in MMD patients, a logistic regression model was used to construct a receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) was then calculated, with a significance level of p < 0.05 indicating statistical significance. Additionally, Kaplan–Meier survival analysis was conducted to compare TIA and stroke-free survival events between participants in the high rWSS and low rWSS groups.

Result

Study participants and baseline characteristics grouped by rWSS and long-term outcomes

A total of 123 patients, encompassing 173 hemispheres, were enrolled in this study. Among these, 83 underwent indirect bypass procedures, while 90 underwent direct and combined bypass procedures. The mean age at the time of operation was 34.09 ± 16.08 years, with females comprising 49.7% of the participants. Based on primary symptoms, the clinical presentation was categorized as follows: nonspecific-type in 8 cases (4.6%), TIA-type in 50 cases (28.9%), ischemic stroke-type in 72 cases (41.6%), and hemorrhagic-type in 43 cases (24.9%). In the perioperative events, stroke-type accounted for 6.9%, TIA-type accounted for 14.5%. In the follow-up events, stroke-type accounted for 4.6%, TIA-type accounted for 17.9%. No patients died as a result of these complications.

As shown in Table 1, patients with high rWSS tended to be older and higher Suzuki stage (all p < 0.05). Additionally, patients with high rWSS had worse mRS scores at follow-up (p = 0.014). As detailed in Supplementary Table 1, the participants grouped by rWSS were compared between morphological and hemodynamic parameters. Compared to the low rWSS group, the high rWSS group exhibited greater changes in CBF (p = 0.016), with a higher proportion of patients experiencing prolonged MTT (p = 0.05).

Table 1 Baseline and clinical characteristics of the participants grouped by rWSS.
Full size table

As presented in Table 2, those patients with unfavorable outcomes tended to be older, had more cases of ischemic stroke-type as the primary symptom and higher incidence of diabetes mellitus, and higher Suzuki stages (all p < 0.05). Moreover, patients with unfavorable outcomes exhibited decreased CBV and CBF (p < 0.05). (Supplementary Table2).

Table 2 Baseline and clinical characteristics of long-term outcomes in participants.
Full size table

Risk factors for unfavorable long-term outcomes

Univariate logistic regression analyses indicated that high rWSS at the terminal ICA: [OR = 4.000(1.765–9.063), p = 0.001]; adult: [OR = 4.887(1.110–21.522), p = 0.036]; diabetes mellitus: [OR = 4.941(2.031–12.018), p < 0.001]; TIA and stroke in primary symptom: [OR = 3.944(1.305,11.915). p = 0.015], Suzuki stage: [OR = 2.564(1.071–6.140). p = 0.035] and CBF: [OR = 0.428(0.190,0.965). p = 0.041] might be associated with unfavorable long-term outcomes (Table 3). After adjusting for all potential covariates, including age, gender, RNF213 variant, hypertension, diabetes, hyperhomocysteinemia, primary symptom, admission mRS score, surgical modalities, perioperative events, Suzuki stage, rP, CTP and morphological parameters, multivariate logistic regression analyses revealed that high rWSS at terminal of ICA: [OR = 3.039(1.191,7.754), p = 0.020] and diabetes mellitus: [OR = 3.164(1.141,8.773), p = 0.027] were significant predictors of unfavorable long-term outcomes.

Table 3 Analysis of variables associated with unfavorable outcome in the logistic regression models.
Full size table

ROC analysis was performed using the Crude model, Model 1, and Model 2 to differentiate between patients with unfavorable and favorable outcomes in MMD (Fig. 2A). The AUC values demonstrated an improvement across the models, with the Crude model showing an AUC of 0.661, Model 1 an AUC of 0.771, and Model 2 an AUC of 0.889. In the Kaplan–Meier analysis, high rWSS levels showed a significant ability to predict future TIA and stroke events (log-rank test: p = 0.032, Fig. 2B).

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

(A) The ROC curves of rWSS to distinguish unfavorable outcome patients from favorable outcome MMD patients: The area under the ROC curves was 0.661 (95% CI 0.557–0.765) for the Crude model, 0.771(95% CI 0.686–0.855) for the Model 1, 0.889(95% CI 0.889–0.942) for the Model 2. ROC, receiver operating characteristic. (B) The Kaplan–Meier cumulative hazard curve for TIA and stroke recurrence comparing high rWSS and low rWSS participants (Log Rank p = 0.032).

The association between morphological, hemodynamic parameters and Suzuki stage of MMD patients

The association between morphological, hemodynamic parameters and Suzuki stage of MMD patients are shown in Fig. 3. The rWSS increased as the Suzuki stage increased (p < 0.05, Fig. 3A), while the inlet area (C1), outlet area (C7), tortuosity index of ICA and terminal ICA decreased as the Suzuki stage advanced (all p < 0.001, Fig. 3C–F).

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

The association between Suzuki stage and morphological and hemodynamic parameters of MMD patients. (A–F) The rWSS, inlet area, outlet area, tortuosity index of terminal ICA and ICA between stage I-II, stage III groups were statistically significant (rWSS: p < 0.05, inlet area: p < 0.001, outlet area: p < 0.001, tortuosity index of terminal ICA: p < 0.001, tortuosity index of ICA: p < 0.001).

The association between morphological, hemodynamic parameters and perioperative, long-term prognosis information of MMD patients

The higher rWSS had more cases with male, adult, diabetes mellitus and ischemic stroke-type in primary symptom (gender: p < 0.05, Fig. 4A; age: p < 0.001, Fig. 4B; diabetes mellitus: p < 0.01, Fig. 4E; primary symptom: p < 0.01 Fig. 4G–H). The rWSS increased as the admission and follow-up mRS score increased (follow-up mRS: p < 0.01, Fig. 4L). While no statistical differences were seen in RNF213 type, hypertension, hyperhomocysteinemia, admission mRS score, perioperative and follow-up period events. (Fig. 4C–D, F, I–K) (Fig. 5).

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

The association between rWSS and perioperative/prognosis information of MMD patients. (A–L) The rWSS between gender, age, diabetes mellitus, follow-up and admission mRS score, primary symptoms were statistically significant (gender: p < 0.05, age: p < 0.001, diabetes mellitus: p < 0.01, follow-up mRS: p < 0.01, primary symptom p < 0.01) *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Quantification of the hemodynamic features of ICA in patients with MMD in a computational fluid dynamics model. (A) Adult patient with MMD, on the left hemisphere in Suzuki stage III. (A.1–2) The terminal region of the ICA was segmented into corresponding meshes, with the boundary defined at the origin of the ophthalmic artery. (A.3, B.3) 3-dimensional (3D) ICA structure for the following hemodynamic analysis. (A.4, B.4) Relative WSS = Average terminal ICA WSS/ Average ICA WSS, rWSS of this case was 46.3 Pa/23.19 Pa = 2. (A.5) Relative Pressure = Average terminal ICA Pressure/ Average ICA Pressure, rP of this case was 7598.4 Pa /10,439.5 Pa = 0.73.(A.6) (B) Pediatric patient with MMD, on the left hemisphere in Suzuki stage I-II. (B.1–2) rWSS of this case was 29.8 Pa/20.1 Pa = 1.48 (B.5) rP of this case was 5201.3 Pa /8527.5 Pa = 0.61(B.6).

The higher inlet area had more cases with male, adult, hypertension and worse admission mRS score (gender: p < 0.01, Supplementary Fig. 1A; age: p < 0.05, Supplementary Fig. 1B; hypertension: p < 0.01, Supplementary Fig. 1D; admission mRS: p < 0.01 Supplementary Fig. 1I). Higher vessel tortuosity of ICA was more common in adults (p < 0.001 Supplementary Fig. 2B), while lower tortuosity index was more common in the RNF213 variant group (p < 0.05 Supplementary Fig. 2C). There was no significant statistical difference between the inlet area, tortuosity index of ICA and other variables in the Supplementary Fig. 1–2. The association between outlet area (C7), rP, tortuosity index(C6-7) and perioperative /prognosis information of MMD patients showed no significant statistical difference. (Supplementary Figs. 3–5).

Discussion

This is the first study to date with the largest sample size to investigate the association between morphological, hemodynamic parameters and CTP parameters, critical information of perioperative, long-term prognosis and DSA results of Suzuki stage in MMD patients. Utilizing CTA-based CFD models, we quantified WSS at terminal of the ICA to evaluate its ability in predicting unfavorable long-term outcomes. Our findings demonstrated that both diabetes and higher rWSS were independent risk factors for unfavorable long-term outcomes in patients with MMD. Additionally, rWSS demonstrated a satisfactory ability to predict future TIA and stroke events in the Kaplan–Meier analysis. Our results implied that morphological and hemodynamic parameters were closely related to DSA results and the clinical information of patients with Moyamoya disease (MMD). These parameters might be useful in evaluating the progression of MMD at an early stage. Future research needs to further validate the role of hemodynamics in the pathogenesis and risk of stroke in patients with MMD.

Research on atherosclerosis generally suggested that low WSS was related to plaque growth and progression of stenosis, while both low and high WSS were linked to increased plaque vulnerability21,22. This contradicted our research findings, but it should be noted that MMD differed significantly from atherosclerosis in terms of its pathogenesis, vascular morphology, inflammatory response, and other factors23. High WSS might have played essential role in the progression of stenosis. Unlike atherosclerosis, the changes in stenosis in MMD were not characterized by lipid core, inflammatory cells aggregated and invaded the intima. Furthermore, as luminal stenosis progresses in atherosclerosis, external vascular remodeling occurred with an increase in external vessel diameter. whereas in MMD, there was a decrease in external vessel diameter. The prevailing view was that constrictive remodeling of affected arteries were the primary distinction between MMD and atherosclerotic stenosis. MMD patients also had a higher risk of ischemic stroke compared to patients with atherosclerosis24.

WSS had been widely used in studies of the natural history and progression mechanisms of ischemic stroke patients10,25,26. Vascular remodeling also contributed to the development of a variety of vascular diseases27. Our study found that high WSS could be an independent risk factor for adverse outcomes in MMD patients, and previous literature suggested that high WSS may promote the progression of stenosis in MMD and was associated with vascular complications11. There was a clear association and a substantial amount of research between vascular remodeling, subsequent distal cerebral hypoperfusion and poor prognosis in stroke patients, but the mechanism linking WSS to unfavorable long-term outcomes in MMD patients remained unclear. The second-hit hypothesis might help us understand this process2. The occurrence of moyamoya disease was related to genetic factors and influenced by infections, inflammatory disorders, hemodynamic changes, and other risk factors. Theoretically, specific vascular remodeling might have been the result of long-term hemodynamic effects. We hypothesized that high WSS could have increased the risk of adverse long-term outcomes by promoting vascular remodeling in MMD. In vessels without stenosis, physiological levels of shear stress might have maintained normal endothelial cell function by regulating gene expression and levels of inflammatory factors, thereby inhibiting cell proliferation, inflammation, and vascular stenosis. However, stenosis might have increased WSS, and high WSS could have resulted in changes in the levels of various factors, such as growth factors and ECPs, resulting in extracellular matrix degradation and promoting the migration and proliferation of vascular smooth muscle cells, abnormal angiogenesis, and ultimately vascular remodeling28,29. It was also shown that supra-physiological WSS could induce inflammation /EC erosion, leading to aggressive remodeling26,30,31. During vascular remodeling, the endothelial cells at terminal of ICA in MMD might be more susceptible to local wall shear stress, facilitating the invasion of EPCs into the intima, inducing the expression of factors such as hyaluronic acid, resulting in vascular stenosis. Stenosis then generates high WSS, leading to a vicious cycle32,33.

Our findings confirmed our hypothesis that patients with high WSS, through the interaction with vascular remodeling and poor perfusion, contributing to the long-term unfavorable outcomes in MMD. In the early stages of MMD, according to the Suzuki classification, abnormal moyamoya vessels were not yet apparent13. A higher Suzuki stage indicated the progression of ICA stenosis and intracranial anterior circulation contraction, which were risk factors for stroke and associated with the severity of clinical symptoms and treatment prognosis. Furthermore, previous studies had reported that stenosis of the ICA and PCA is closely associated with prolonged MTT due to reduced perfusion pressure34. Shi et al. thought early-stage CTP in MMD as a promising tool for predicting perioperative stroke. In our research, high rWSS, unfavorable outcome and changes in CBF, CBV, and MTT were correlated, indicating that the high-risk surgical hemispheres are more prone to both ischemic and compensatory states35. Previous studies had identified diabetes mellitus as a risk factor for adverse outcomes after revascularization in MMD patients8. Elevated blood glucose can stimulate the vascular wall, vascular endothelial dysfunction, platelet dysfunction and so on, which can lead to the occurrence and development of vascular stenosis. We observed that MMD patients with higher WSS had a lower survival rate free from TIA and stroke (p = 0.032). In previous studies, both adult and pediatric patients with moyamoya disease exhibited similar morphological expressions, characterized by lower tortuosity of ICA and smaller ICA cross-sectional area36,37. Building on previous studies, our research also found significant morphological differences in the ICA among patients at different Suzuki stages. Patients with advanced Suzuki stages exhibited smaller internal carotid artery cross-sectional areas and straighter ICAs. Since this study was designed as a cross-sectional study, these morphological features could either be a cause of MMD progression or a result of progressive carotid artery stenosis. The RNF213 variant was involved in the vascular remodeling process at the terminal ICA through high WSS induced by lower ICA tortuosity, consistent with our study38.

This study has some limitations. Firstly, as a single-center study, there is a potential for selection bias despite the large number of cases. Since CFD modeling cannot accurately simulate occluded vessels, this study only included patients with Suzuki stage 3 or below. therefore, the conclusions should be interpreted with caution for higher-grade MMD patients. Further, only a simplified CFD model including the ICA was reconstructed, and although relatively individualized inlet flow rates were used, uniform outlet conditions and standardized blood properties were applied across all simulations. Therefore, in this study, we used relative WSS values, and corrected for covariates in multivariable regression analysis to offset the potential influence of confounders. Hemodynamic changes in vivo are instantaneous, and cannot be fully represented in steady models. In the future CFD models incorporating fluid–structure interaction could yield more detailed insights into local hemodynamics in patients with MMD patients. To better understand the causal relationship between the morphology and hemodynamics of MMD, additional serial observational studies with larger patient cohorts.

Conclusion

Our results suggested that high rWSS and diabetes mellitus were independent risk factors for unfavorable long-term outcomes and WSS might have the potential to predict future TIA and stroke events in MMD patients. The DSA results of Suzuki stage were close related to morphology and hemodynamics. Local hemodynamics at terminal of ICA might play a critical role in the progression of anterior circulation stenosis in relative early MMD. The relationship between focal WSS, Morphology of vessel, blood flow, ICA stenosis, and stroke risk is intricate and dynamic. The pathogenesis and progression of MMD remains to be further elucidated.