2025 Volume 32 Issue 7 Pages 840-852
Aim: This study assessed the predictive value of pericarotid fat density (PFD) on carotid computed tomography angiography (CTA) for recurrent ischemic stroke or transient ischemic attack (TIA).
Methods: In total, 739 patients who underwent CTA between January 2014 and December 2021 were retrospectively included in this study. The PFD was evaluated using carotid CTA. The clinical endpoint was recurrent ischemic stroke or transient ischemic attack (TIA). The association between PFD and the endpoint was examined using Kaplan-Meier and Cox analyses. The combination model was established using significant clinical imaging risk factors and PFD. The predictive performance of the model was assessed using the receiver operating characteristic curve (ROC).
Results: A total of 739 patients (mean age: 64.28±9.44 years old, 496 males) completed a median of 3.31 years of follow-up (interquartile range, 2.11-4.05). During the follow-up period, 166 patients reached the clinical end point. The event-free survival (EFS) rate was lower in the high-PFD group than in the low-PFD group (log-rank P＜0.001). Multivariate Cox analyses showed that the PFD was associated with recurrent stroke or TIA (all P＜0.05). The combination model demonstrated excellent performance in predicting the clinical endpoint (area under the curve = 0.89). In addition, the endpoint event prognostic value was significantly improved by adding the PFD to the baseline model (C-statistic improvement: 0.61–0.84).
Conclusion: CTA-assessed PFD is an independent predictor of recurrent stroke or TIA.
Shuai Zhang and Ximing Wang as joint senior authors.
Cerebrovascular disease (CVD) is a leading cause of disability and death worldwide1). Recurrent strokes or transient ischemic attacks (TIAs) more frequently disable or kill individuals than incident ones2). According to statistics, the average incidence rate of stroke in China is 505.2/100,000 years3). Therefore, the prediction of early recurrent stroke or TIA is clinically useful.
A previous study showed that the severity of carotid stenosis is an important risk factor for ischemic stroke4). More recent studies have shown that carotid plaque composition, such as hemodynamics and inflammation, is associated with an increased risk of cerebrovascular disease5, 6). In particular, inflammation is considered a major contributor to atherosclerotic plaque formation, destabilization, and rupture7).
Perivascular adipose tissue (PVAT) is a specific type of adipose tissue deposit surrounding blood vessels. Recent studies have shown that PVAT plays an important role in maintaining metabolic and vascular homeostasis by producing paracrine factors8, 9). The inflammatory changes occurring in PVAT have been suggested to play an important role in the pathogenesis of cardiovascular diseases10). Recently, such changes have been shown to be detectable by computed tomography (CT)11).
Increased pericarotid fat density (PFD) is closely associated with PVAT inflammation12). Growing evidence suggests that PFD is associated with vascular diseases, such as coronary artery disease (CAD), small-vessel disease, and worsening cervical vascular stenosis13-15). To our knowledge, however, few studies have reported an association between PVAT and recurrent stroke or TIA.
The present study explored the association of attenuation in the carotid perivascular adipose tissue on CT angiography (CTA) with recurrent stroke or TIA and evaluated the prognostic value of PFD relative to clinical and CTA variables.
This study was approved by the institutional review board, and all patients provided their informed consent. We retrospectively evaluated consecutive symptomatic individuals who underwent carotid CTA between January 2014 and December 2021. The inclusion criterion was the presence of cerebrovascular symptoms within 14 days of admission. The exclusion criteria were as follows: (a) a history of cancer, (b) a history of carotid endarterectomy or carotid stenting, (c) a compromised image quality of carotid CTA not feasible for accurate analysis, and (d) participants lost to follow-up. A flow diagram of the subject selection process is presented in Supplementary Fig.1. Clinical data, including the age, sex, hypertension, hyperlipidemia, drinking history, smoking history, statin use, and aspirin use, were collected from medical records.
CTA, CT angiography.
CTA was performed using a third-generation dual-source CT scanner (SOMATOM Force; Siemens Healthineers, Erlangen, Germany). CTA studies were performed in helical scanning mode with coverage extending from the aortic arch to the skull vertex. A 60- to 80-mL volume of contrast medium (Omnipaque 350; GE Healthcare, Shanghai, China) was injected at a rate of 4 mL/s, followed by a 40 mL saline flush using a power injector. Bolus tracking was used to trigger the acquisition 5 s after an attenuation threshold of 100 Hounsfield units (HU) was reached in the aortic arch. The carotid CTA scanning parameters were as follows: tube voltage, 100 KV; pitch, 1.0; reconstructed slice thickness, 0.5 mm; reconstructed slice interval, 0.5 mm; rotation time, 350 ms.
Image AnalysesAll images were independently evaluated by two radiologists with more than 10 years of experience in vascular imaging, both of whom were blinded to patient information. Any disagreement in assessment was resolved by consensus.
The density of the perivascular fat surrounding the carotid plaque was analyzed using a dedicated software program (Perivascular Fat Analysis Tool; Shukun Technology, Beijing, China). Perivascular fat was defined as adipose tissue situated within a radial distance from the outer vessel wall, equivalent to the average diameter of the target vessel. We adopted an established approach used in the coronary arteries and measured the density of perivascular fat in a semi-automated manner by tracking the contour of vascular stenosis9). The PFD was determined by quantifying the weighed perivascular fat attenuation after adjusting for technical parameters on the basis of the attenuation histogram of perivascular fat within the range of −190 to −30 HU. All fat density measurements were reported in HU. An analysis of the PVAT of plaques is shown in Fig.1.
a-d depict carotid plaque and surrounding adipose tissue density in a 62-year-old male patient without recurrent ischemic stroke or TIA. a CTA image of the carotid plaque (red arrow) in the patient. b Visualization of the PVAT (white arrow) encircling the carotid plaque in the patient, as shown by red pixels on CTA. c The pixel diagram of PVAT density of plaque is shown. d Histogram of the PVAT density in the plaque with a mean PFD of -74.29 HU. e-h display carotid plaque and surrounding adipose tissue density in a 70-year-old male patient with recurrent ischemic stroke. e CTA image of the carotid plaque (red arrow). f PVAT (white arrow) surrounding the carotid plaque in the patient, visualized on CTA by red pixels. g The pixel diagram of the PVAT density of plaque is shown. h Histogram of PVAT density in the plaque, with a mean PFD of -57.83 HU. i For the patient in a-d (blue solid arrows), a nomogram yielded a total of 36.0 points and corresponding risk of IPH of ＜0.3. For patient in e-h (red dashed arrows), a nomogram yielded a total of 72.5 points and corresponding risk of IPH of ＜0.9. PVAT, perivascular adipose tissue; TIA, transient ischemic attack; CTA, computed tomography angiography; HU, Hounsfield Unit. PFD, pericarotid fat density.
The area of the PVAT was measured using the ImageJ software program [public domain software; National Institutes of Health (NIH), USA. Perivascular fat within the range of 190 to −30 HU was identified and calculated. The program was then used to measure the area of these tissues and determine the area of the PVAT.
Various CTA markers, including the degree of luminal stenosis, maximum plaque thickness, occurrence of calcification, and presence of ulceration were measured using a post-processing workstation (Syngovia, Siemens Force, Germany). The degree of stenosis was determined according to the North American Symptomatic Carotid Endarterectomy Trial (NASCET) criteria using CTA16). Measurements of maximal plaque thickness were performed on plaque maximum cross sections perpendicular to the long axis of the vessel. A focus of ≥ 4 contiguous pixels accompanied by a CT density of ≥ 130 HU was defined as calcification according to the Agatston score method17). Ulceration was defined as irregular depression with a minimum depth of 1 mm in any plane18).
Clinical OutcomePatients or their guardians were contacted via telephone at 3, 6, 9, and 12 months after discharge. The primary endpoint was a composite of clinical ipsilateral recurrent stroke and TIA. The terminating points were carotid endarterectomy (CEA), carotid artery stenting (CAS), and death. All recurrent ischemic events were confirmed according to clinical medical records. Two experienced neurologists reviewed the patients’ medical documents or telephone follow-ups to ensure a reliable diagnosis of recurrent ischemic stroke or TIA.
Statistical AnalysesThe Kolmogorov-Smirnov test was used to assess the normality of the data distribution. For descriptive statistics, continuous variables are expressed as the mean±standard deviation or median (interquartile range [IQR]), and categorical variables are expressed as percentages. The patients were divided into two groups according to the cutoff value estimated using the Youden index. The cumulative incidence of the study endpoints over time was analyzed using the Kaplan-Meier method, and log-rank tests were performed to study differences. Cox proportional hazards regression models were used to determine the association of clinical and imaging characteristics with clinical endpoints. Significant factors (P＜0.05) in the univariate Cox regression analysis were included in the Cox regression model for a multivariate analysis. The results are presented as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) for the Cox proportional hazards models.
A multivariable baseline model (multivariable Cox regression, P＜0.05) was used to evaluate the predictive value. The PFD was then added to the baseline model to assess its incremental prognostic value using C-statistics. Receiver operating characteristic (ROC) curves were generated, and the areas under the curves (AUCs) were calculated with sensitivity and specificity. Intrareader and inter-reader agreement in measuring the density of perivascular fat was calculated using the intraclass correlation coefficient.
All statistical analyses were performed using the SPSS software program (version 22.0; IBM, Chicago, IL, USA) and R statistical software program (version 4.3.2; R Foundation for Statistical Computing). A value of P＜0.05 was considered statistically significant.
Of the 995 included patients, 110 were excluded due to a history of cancer, a history of carotid endarterectomy or stenting (n = 29), poor quality images (n = 19), and loss to follow-up (n = 98).
A total of 739 patients (mean age: 64.28±9.44 years old, 496 males) completed a median of 3.31 years of follow-up (IQR, 2.11-4.05). During the follow-up period, 166 of the 739 patients met the clinical endpoint, which included recurrent cerebral ischemic stroke (n = 69) and TIA (n = 97). Among these, 213 (28.82%) and 206 (27.88%) patients had a history of smoking and alcohol consumption, respectively. A total of 462 (62.52%) and 213 (28.82%) patients had a history of hypertension and diabetes, respectively, and 426 (57.65%) and 432 (58.46%) patients had statin and aspirin use, respectively.
Patients were divided into event (n = 166) and non-event (n = 573) groups according to whether or not clinical outcomes occurred during follow-up. Compared with patients without events, the event group was more likely to have hypertension (69.28% vs. 60.56%; P = 0.041), diabetes (37.95% vs. 26.18%; P = 0.003), and a history of alcohol consumption (35.54% vs. 25.65%; P = 0.012), while there were no significant differences in other factors (all P＞0.05) (Table 1).
| Variables | Total (n = 739) | With event (n = 166) | Without event (n = 573) | P |
|---|---|---|---|---|
| Age, y | 64.28±9.44 | 64.75±9.44 | 64.14±9.45 | 0.470 |
| Male | 496 (67.12) | 113 (68.07) | 383 (66.84) | 0.088 |
| Hypertension | 462 (62.52) | 115 (69.28) | 347 (60.56) | 0.041 |
| Diabetes | 213 (28.82) | 63 (37.95) | 150 (26.18) | 0.003 |
| Alcohol history | 206 (27.88) | 59 (35.54) | 147 (25.65) | 0.012 |
| Smoking history | 213 (28.82) | 46 (27.71) | 167 (29.14) | 0.719 |
| Medication in use | ||||
| Statin use | 426 (57.65) | 104 (62.65) | 322 (56.20) | 0.138 |
| Aspirin use | 432 (58.46) | 101 (60.84) | 331 (57.77) | 0.479 |
| Laboratory findings | ||||
| AST, U/L | 23.41±11.48 | 22.06±8.02 | 23.80±12.28 | 0.085 |
| ALT, U/L | 23.26±16.33 | 23.08±16.73 | 23.32±16.23 | 0.870 |
| GGT, U/L | 31.90±36.77 | 29.48±21.70 | 32.61±40.08 | 0.334 |
| TG, mmol/L | 1.47±1.00 | 1.59±1.39 | 1.44±0.86 | 0.084 |
| TC, mg/dL | 4.54±1.22 | 4.39±1.11 | 4.58±1.25 | 0.088 |
Continuous variables were described as the mean±standard deviation, and categorical variables were presented as numbers (%).
AST, aspartate aminotransferase; ALT, alanine transaminase; GGT, gamma-glutamyl transferase; TG, triglycerides; TC, total cholesterol.
Compared with patients in the group without events, patients in the group with events showed significantly more severe stenosis (39.64%±20.32% vs. 34.15%±23.96%, P = 0.007), a higher prevalence of ulceration (33.16% vs. 22.16%, P＜0.001), and higher PFD (-59.31±7.45 HU vs. -69.78±5.22 HU, P＜0.001). In addition, compared with patients without events, patients with events had a lower prevalence of mild stenosis (32.53% vs. 52.18%, P＜0.001) and a higher prevalence of moderate stenosis (56.02% vs. 33.68%, P＜0.001). There were no significant differences in the prevalence of other factors between the 2 groups (P＞0.05) (Table 2).
| Variables | Total (n = 739) | With event (n = 166) | Without event (n = 573) | P |
|---|---|---|---|---|
| Degree of stenosis | 35.38±23.30 | 39.64±20.32 | 34.15±23.96 | 0.007 |
| Categories of stenotic degree | ||||
| ＜30% | 353 (47.77) | 54 (32.53) | 299 (52.18) | ＜0.001 |
| 30–69% | 286 (38.70) | 93 (56.02) | 193 (33.68) | ＜0.001 |
| ≥ 70% | 100 (13.53) | 19 (11.45) | 81 (14.14) | 0.372 |
| Maximum wall thickness, mm | 3.03±1.42 | 3.04±1.41 | 3.03±1.42 | 0.913 |
| Calcification | 553 (74.83) | 125 (75.30) | 428 (74.69) | 0.874 |
| Ulceration | 190 (25.71) | 63 (33.16) | 127 (22.16) | ＜0.001 |
| Area of PVAT, mm2 | 68.11±31.33 | 69.12±29.37 | 67.82±31.90 | 0.639 |
| PFD, HU | -67.42±7.25 | -59.31±7.45 | -69.78±5.22 | ＜0.001 |
Continuous variables were described as the mean±standard deviation, and categorical variables were presented as numbers (%). PVAT, perivascular adipose tissue; PFD, pericarotid fat density.
Using the Youden index, the cutoff value for the PFD was determined to be -66.15 HU in the whole cohort, -65.95 in the mild stenosis cohort, -63.96 in the moderate stenosis cohort, and -67.34 in the severe stenosis cohort. After dividing the patients into high- and low-PFD groups, a Kaplan-Meier analysis showed that patients in the high-PFD group had a lower 2.9-year event-free survival (EFS) rate than those in the low-PFD group across the entire cohort (23.50% vs. 88.97%, log-rank P＜0.001) (Fig.2).
Patients with a high PFD had a lower EFS rate than those with a low PFD (23.50% vs 88.97%, log-rank P＜0.001) across the whole cohort. EFS, event-free survival; TIA, transient ischemic attack; PFD, pericarotid fat density.
A Kaplan-Meier analysis of patients with mild-to-moderate and severe stenosis stratified by the optimal cutoff value of PFD showed that the EFS was significantly lower in patients in the high-PFD group, irrespective of the degree of stenosis, than those in the low-PFD group. In addition, the EFS was significantly lower and unaffected by the degree of stenosis in high PFD groups than in low PFD groups (log-rank P＞0.05) (Fig.3).
a. Patients with mild-moderate stenosis vs. those with severe stenosis are stratified by the optimal cutoff value of PFD. b. Patients with ulceration vs. those without ulceration are stratified by the optimal cutoff value of PFD.
Furthermore, a Kaplan-Meier analysis of patients with ulceration versus those without ulceration stratified by the optimal cutoff value of PFD showed that the EFS was lower for patients in the high-PFD group, irrespective of ulceration, than for those in the low-PFD group. In addition, ulceration had no significant effect on the EFS in the high PFD group (log-rank P = 0.208). However, patients in the low-PFD group with ulcerated plaque had a significantly lower EFS than those without ulcerated plaques (log-rank P = 0.030) (Fig.3).
Univariable and Multivariable Cox Regression Analyses Whole Cohort AnalysesA univariate Cox regression analysis showed that hypertension (HR, 1.45, 95% CI, 1.04-2.02; P = 0.027), diabetes (HR, 1.60, 95% CI, 1.17-2.18; P = 0.004), an alcohol-drinking history (HR, 1.62, 95% CI, 1.17-2.22; P = 0.003), statin use (HR, 1.64, 95% CI, 1.19- 2.26; P = 0.003), aspirin use (HR, 1.38, 95% CI, 1.01-1.90; P = 0.045), TC (HR, 0.85, 95% CI, 0.75-0.96; P = 0.010), degree of stenosis (HR, 1.01, 95% CI, 1.00-1.01; P = 0.009), ulceration (HR, 2.83, 95% CI, 1.97-4.08; P＜0.001), and the PFD (HR, 1.17, 95% CI, 1.15-1.20; P＜0.001) were associated with the clinical outcome (Table 3).
| Variable | Univariable HR | P Value | Multivariable HR | P Value |
|---|---|---|---|---|
| Age, y | 1.01 (0.99-1.03) | 0.469 | - | - |
| Male | 0.95 (0.65-1.37) | 0.766 | - | - |
| Hypertension | 1.47 (1.02-2.13) | 0.042 | 1.04 (0.64-1.71) | 0.870 |
| Diabetes | 1.73 (1.20-2.48) | 0.003 | 1.98 (1.21-3.24) | 0.006 |
| Alcohol history | 1.60 (1.11-2.31) | 0.013 | 1.68 (1.02-2.75) | 0.041 |
| Smoking history | 0.93 (0.63-1.37) | 0.719 | - | - |
| Medication in use | ||||
| Statin use | 1.31 (0.92-1.87) | 0.139 | - | - |
| Aspirin use | 1.14 (0.80 -1.62) | 0.479 | - | - |
| Laboratory findings | ||||
| AST, U/L | 0.98 (0.96-1.00) | 0.082 | - | - |
| ALT, U/L | 1.00 (0.99-1.01) | 0.869 | - | - |
| GGT, U/L | 1.00 (0.99-1.00) | 0.337 | - | - |
| TG, mmol/L | 1.14 (0.97-1.33) | 0.105 | - | - |
| TC, mg/dL | 0.88 (0.76-1.02) | 0.089 | - | - |
| Degree of stenosis | 1.01 (1.00-1.02) | 0.008 | 1.00 (0.99-1.01) | 0.523 |
| Maximum wall thickness, mm | 1.01 (0.89-1.14) | 0.913 | - | - |
| Calcification | 1.03 (0.69-1.54) | 0.874 | - | - |
| Ulceration | 2.14 (1.48-3.11) | ＜0.001 | 2.00 (1.21-3.32) | 0.007 |
| Area of PVAT, mm2 | 1.00 (1.00-1.01) | 0.638 | - | - |
| PFD, HU | 1.28 (1.24-1.34) | ＜0.001 | 1.29 (1.24-1.34) | ＜0.001 |
AST, aspartate aminotransferase; ALT, alanine transaminase; GGT, gamma-glutamyl transferase; TG, triglycerides; TC, total cholesterol; PFD, pericarotid fat density.
A multivariate Cox regression analysis showed that diabetes (HR, 1.98; 95% CI, 1.21-3.24; P = 0.006), an alcohol-drinking history (HR, 1.68; 95% CI, 1.02-2.75; P = 0.041), ulceration (HR, 2.00; 95% CI, 1.21-3.32; P = 0.007), and the PFD (HR, 1.29; 95% CI, 1.24-1.34; P＜0.001) were associated with clinical outcomes. (Table 3). A nomogram was developed based on the results of the multivariate analysis (Fig.2).
Subgroup AnalysesThe results of the subgroup analysis of the degree of stenosis showed that the association of PFD with the clinical endpoint remained robust in patients with mild stenosis (HR, 1.30; 95% CI, 1.21-1.39; P＜0.001), moderate stenosis (HR, 1.29; 95% CI, 1.21-1.38; P＜0.001), and severe stenosis (HR, 1.25; 95% CI, 1.14-1.38; P＜0.001) (Table 4). Similar results were observed in the subgroup divided according to ulceration (Supplementary Table 1).
| Variable | Patients with mild stenosis (n = 353 ) | Patients with moderate stenosis (n = 286) | ||||||
|---|---|---|---|---|---|---|---|---|
| Univariable HR | P Value | Multivariable HR | P Value | Univariable HR | P Value | Multivariable HR | P Value | |
| Age, y | 1.01 (0.98-1.04) | 0.493 | - | - | 0.99 (0.96-1.02) | 0.408 | - | - |
| Male | 0.87 (0.47-1.59) | 0.648 | - | - | 1.36 (0.79-2.32) | 0.265 | - | - |
| Hypertension | 1.22 (0.67-2.23) | 0.523 | - | - | 1.97 (1.16-3.34) | 0.013 | 1.48 (0.70-3.10) | 0.304 |
| Diabetes | 2.26 (1.24-4.11) | 0.008 | 2.11 (0.98-4.58) | 0.058 | 1.49 (0.88-2.54) | 0.140 | - | - |
| Alcohol history | 1.56 (0.83-2.95) | 0.169 | - | - | 1.54 (0.90-2.64) | 0.114 | - | - |
| Smoking history | 0.91 (0.47-1.80) | 0.795 | - | - | 0.92 (0.53-1.61) | 0.776 | - | - |
| Medication in use | ||||||||
| Statin use | 1.11 (0.62-1.99) | 0.728 | - | - | 1.64 (0.98-2.75) | 0.060 | - | - |
| Aspirin use | 0.66 (0.37-1.19) | 0.167 | - | - | 1.91 (1.13-3.22) | 0.016 | 1.75 (0.84-3.66) | 0.136 |
| Laboratory findings | ||||||||
| AST, U/L | 0.99 (0.96-1.02) | 0.396 | - | - | 0.99 (0.96-1.01) | 0.441 | - | - |
| ALT, U/L | 1.00 (0.98-1.02) | 0.914 | - | - | 1.00 (0.99-1.02) | 0.688 | - | - |
| GGT, U/L | 1.00 (0.98-1.01) | 0.541 | - | - | 1.00 (0.99-1.01) | 0.382 | - | - |
| TG, mmol/L | 1.15 (0.94-1.41) | 0.168 | - | - | 1.13 (0.86-1.48) | 0.398 | - | - |
| TC, mg/dL | 0.81 (0.64-1.04) | 0.105 | - | - | 0.88 (0.71-1.08) | 0.217 | - | - |
| Degree of stenosis | 1.10 (1.05-1.15) | ＜0.001 | 1.07 (1.00-1.13) | 0.047 | 0.99 (0.96-1.02) | 0.392 | - | - |
| Maximum wall thickness, mm | 1.07 (0.79-1.45) | 0.652 | - | - | 0.74 (0.60-0.91) | 0.004 | 0.62 (0.45-0.84) | 0.002 |
| Calcification | 1.11 (0.60-2.07) | 0.742 | - | - | 0.74 (0.41-1.35) | 0.325 | - | - |
| Ulceration | 2.35 (1.19-4.64) | 0.013 | 2.70 (1.11-6.55) | 0.028 | 1.88 (1.13-3.14) | 0.016 | 2.66 (1.24-5.69) | 0.012 |
| Area of PVAT, mm2 | 0.99 (0.98-1.00) | 0.187 | - | - | 1.01 (1.00-1.01) | 0.227 | - | - |
| PFD, HU | 1.30 (1.22-1.39) | ＜0.001 | 1.30 (1.21-1.39) | ＜0.001 | 1.28 (1.21-1.36) | ＜0.001 | 1.29 (1.21-1.38) | ＜0.001 |
| Variable | Patients with severe stenosis (n = 100) | ||||||
|---|---|---|---|---|---|---|---|
| Univariable HR | P Value | Multivariable HR | P Value | ||||
| Age, y | 0.99 (0.93-1.06) | 0.799 | - | - | |||
| Male | 0.47 (0.13-1.78) | 0.268 | - | - | |||
| Hypertension | 0.98 (0.32-3.05) | 0.972 | - | - | |||
| Diabetes | 1.39 (0.49-3.95) | 0.542 | - | - | |||
| Alcohol history | 1.89 (0.69-5.17) | 0.216 | - | - | |||
| Smoking history | 0.96 (0.35-2.63) | 0.930 | - | - | |||
| Medication in use | |||||||
| Statin use | 0.77 (0.28-2.12) | 0.609 | - | - | |||
| Aspirin use | 0.45 (0.15-1.40) | 0.168 | - | - | |||
| Laboratory findings | |||||||
| AST, U/L | 0.96 (0.90-1.02) | 0.160 | - | - | |||
| ALT, U/L | 0.99 (0.96-1.02) | 0.561 | - | - | |||
| GGT, U/L | 1.00 (0.99-1.01) | 0.996 | - | - | |||
| TG, mmol/L | 1.38 (0.77-2.47) | 0.285 | - | - | |||
| TC, mg/dL | 1.08 (0.71-1.64) | 0.723 | - | - | |||
| Degree of stenosis | 1.03 (0.96-1.09) | 0.454 | - | - | |||
| Maximum wall thickness, mm | 0.90 (0.64-1.26) | 0.534 | - | - | |||
| Calcification | 0.93 (0.18-4.79) | 0.932 | - | - | |||
| Ulceration | 1.10 (0.39-3.12) | 0.852 | - | - | |||
| Area of PVAT, mm2 | 0.99 (0.98-1.01) | 0.292 | - | - | |||
| PFD, HU | 1.25 (1.14-1.38) | ＜0.001 | 1.25 (1.14-1.38) | ＜0.001 | |||
AST, aspartate aminotransferase; ALT, alanine transaminase; GGT, gamma-glutamyl transferase; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglyceride; PFD, pericarotid fat density.
| Variable | Patients with ulceration (n = 190 ) | Patients without ulceration (n = 549) | ||||||
|---|---|---|---|---|---|---|---|---|
| Univariable HR | P Value | Multivariable HR | P Value | Univariable HR | P Value | Multivariable HR | P Value | |
| Age, y | 1.01 (0.98-1.05) | 0.561 | - | - | 1.00 (0.98-1.02) | 0.903 | - | - |
| Male | 1.74 (0.90-3.36) | 0.102 | - | - | 0.77 (0.48-1.22) | 0.268 | - | - |
| Hypertension | 1.61 (0.82-3.16) | 0.165 | - | - | 1.33 (0.85-2.08) | 0.217 | - | - |
| Diabetes | 2.23 (1.17-4.22) | 0.014 | 2.91 (1.22-6.91) | 0.016 | 1.49 (0.95-2.36) | 0.085 | - | - |
| Alcohol history | 1.16 (0.62-2.19) | 0.638 | - | - | 1.75 (1.10-2.77) | 0.018 | 1.88 (1.02-3.44) | 0.042 |
| Smoking history | 0.61 (0.31-1.18) | 0.141 | - | - | 1.08 (0.67-1.74) | 0.761 | - | - |
| Medication in use | ||||||||
| Statin use | 2.35 (1.22-4.53) | 0.011 | 1.85 (0.74-4.61) | 0.188 | 0.97 (0.63-1.50) | 0.906 | - | - |
| Aspirin use | 1.22 (0.64-2.29) | 0.547 | - | - | 1.03 (0.67-1.59) | 0.886 | - | - |
| Laboratory findings | ||||||||
| AST, U/L | 0.97 (0.94-1.01) | 0.144 | - | - | 0.98 (0.96-1.01) | 0.201 | - | - |
| ALT, U/L | 1.00 (0.98-1.02) | 0.957 | - | - | 1.00 (0.99-1.01) | 0.949 | - | - |
| GGT, U/L | 1.00 (0.98-1.01) | 0.342 | - | - | 1.00 (0.99-1.01) | 0.704 | - | - |
| TG, mmol/L | 0.91 (0.66-1.25) | 0.557 | - | - | 1.27 (1.01-1.61) | 0.042 | 1.33 (0.92-1.94) | 0.135 |
| TC, mg/dL | 0.74 (0.57-0.97) | 0.027 | 0.72 (0.50-1.04) | 0.080 | 0.94 (0.79-1.12) | 0.489 | - | - |
| Degree of stenosis | 1.00 (0.99-1.01) | 0.760 | - | - | 1.01 (1.00-1.02) | 0.012 | 1.01 (1.00-1.02) | 0.193 |
| Maximum wall thickness, mm | 0.82 (0.66-1.03) | 0.086 | - | - | 0.96 (0.82-1.14) | 0.649 | - | - |
| Calcification | 1.29 (0.53-3.10) | 0.577 | - | - | 0.83 (0.52-1.31) | 0.418 | - | - |
| Area of PVAT, mm2 | 1.00 (0.99-1.01) | 0.789 | - | - | 1.00 (0.99-1.01) | 0.708 | - | - |
| PFD, HU | 1.28 (1.19-1.38) | ＜0.001 | 1.30 (1.21-1.41) | < 0.001 | 1.29 (1.23-1.35) | ＜0.001 | 1.28 (1.22-1.35) | ＜0.001 |
AST, aspartate aminotransferase; ALT, alanine transaminase; GGT, gamma-glutamyl transferase; TG, triglycerides; TC, total cholesterol; PFD, pericarotid fat density.
Adding PFD on top of the multivariable baseline model significantly improved the performance and prognostic value of the model for recurrent stroke and TIA (C-statistic improvement: 0.61 [95% CI: 0.65-0.56] to 0.84 [95% CI: 0.87-0.81]) (Supplementary Table 2). An ROC curve analysis revealed that the AUC of the combination model was the highest (0.89, 95% CI: 0.86-0.91) for predicting recurrent stroke and TIA (Fig.4). Furthermore, the Delong test illustrated that the differences in AUC were statistically significant (P＜0.05) when clinical risk factors or imaging risk factors or the PFD were compared with the combination model (Supplementary Table 3).
| Model | C-Statistic (95% CI) |
|---|---|
| Baseline model | 0.61 (0.65-0.56) |
| Baseline model+PFD | 0.84 (0.87-0.81) |
CI, confidence interval; PFD, pericarotid fat density.
An ROC analysis of clinical risk factors (0.58, 95% CI: 0.54-0.61), imaging risk factors (0.58, 95% CI: 0.54-0.62), PFD (0.87, 95% CI: 0.84-0.89), and the combination model (0.89, 95% CI: 0.86-0.91) for predicting the clinical endpoint. Clinical risk factors included diabetes and the alcohol-drinking history, imaging risk factors included ulceration, and the combination model included diabetes, the alcohol-drinking history, ulceration, and PFD. AUC, area under the curve; ROC, receiver operating characteristics; CI, confidence interval; PFD, pericarotid fat density.
| Model | AUC (95% CI) | Sensitivity (%) | Specificity (%) | P | |||||
|---|---|---|---|---|---|---|---|---|---|
| 1 vs 2 | 1 vs 3 | 1 vs 4 | 2 vs 3 | 2 vs 4 | 3 vs 4 | ||||
| clinical risk factors (1) | 0.58 (0.54 - 0.61) | 17.47 | 92.32 | 0.978 | ＜0.001 | ＜0.001 | ＜0.001 | ＜0.001 | 0.011 |
| imaging risk factor (2) | 0.58 (0.54 - 0.62) | 37.95 | 77.84 | - | - | - | - | - | - |
| PFD (3) | 0.87 (0.84 - 0.89) | 75.90 | 81.15 | - | - | - | - | - | - |
| combination model (4) | 0.89 (0.86 - 0.91) | 75.30 | 87.26 | - | - | - | - | - | - |
AUC, area under the curve; CI, confidence interval; PFD, pericarotid fat density.
The intraclass correlation coefficient for PFD measurements was 0.873 (95% CI 0.811-0.914) for intra-reader agreement and 0.973 (95% CI 0.960-0.982) for inter-reader agreement.
Our study revealed that PFD is an independent marker for predicting recurrent stroke or TIA (HR, 1.29; 95% CI, 1.24-1.34; P＜0.001). Notably, the combination model showed an excellent predictive value (AUC = 0.89). Therefore, PFD may be a novel and non-invasive marker for predicting recurrent stroke or TIA.
In our study, 22.4% of the patients experienced stroke recurrence within 3.3 years, which is consistent with findings from previous large-scale cohort studies. In a population-based study of 0.5 million Chinese adults, the 3-year stroke recurrence rate was approximately 32%, and the 5-year recurrence rate was reported to be 41% for all stroke types, with specific rates of 41% for ischemic stroke and 44% for intracerebral hemorrhaging19). These figures indicate that stroke recurrence remains a significant global challenge.
Previous studies have confirmed that the degree of carotid stenosis and characteristics of carotid plaques, such as ulceration, can predict recurrent stroke or TIA with higher sensitivity than standard imaging evaluations that do not consider plaque characteristics20, 21). Vasileios et al. confirmed that plaque surface and ulceration are risk factors for stroke in patients with carotid disease22). Our findings, which support this notion, are consistent with those of these studies. Furthermore, the model that included TC, ulceration, and PFD showed excellent discrimination performance. This result confirmed the utility of PFD in predicting recurrent stroke or TIA.
The relationship between the vascular wall and surrounding adipose tissue is complex. Perivascular adipose tissue contains various bioactive molecules that can diffuse directly into the vascular wall via paracrine mechanisms, thereby affecting its biological characteristics, including inflammation and oxidative stress23). These factors are associated with the development of atherosclerosis and other adverse clinical events24). In advanced atherosclerotic disease, inflammation and oxidative stress can induce the secretion of vascular endothelial growth factor (VEGF), leading to an increase in immature angiogenesis25). Following the rupture of these neovessels, inflammatory factors are released into the vascular wall, and these cytokines diffuse into the perivascular space, causing inflammation in the PVAT. This results in a hypercoagulable state and recurrent ischemic stroke26).
The attenuation of PVAT is influenced by inflammatory signals from the vascular wall, which induce an opposing phenotype. Systemic markers of inflammation, including C-reactive protein, have been associated with cardiovascular risk prediction, but they lack specificity for vascular inflammation and cannot identify inflammatory changes in a specific vascular region. Conversely, measuring PFD using CTA is helpful for identifying localized perivascular inflammation. These biological changes create a balanced gradient between the lipid and water phases, with the gradient signal of CT attenuation in the perivascular space surpassing the systemic signal captured by the attenuation of non-perivascular adipose tissue27). CTA, owing to its high spatial resolution, can capture these subtle changes in tissue composition. The density of PVAT, measured in HU on CTA, reflects the balance between the lipid and water content within the tissue28). Antonopoulos et al. confirmed a link between PVAT inflammation and CT attenuation. PVAT attenuation negatively correlates with histological adipocyte size and degree of adipocyte differentiation, where higher pericoronary tissue attenuation reflects lower lipid content in smaller adipocytes27). Therefore, this change in PVAT composition triggered by inflammation leads to an increase in the attenuation values of the CTA-assessed PFD. Recent studies have reported that local coronary inflammation associated with vulnerable plaques can be detected using standard noninvasive CTA modalities, providing information similar to 18FNaF-positron emission tomography (PET)-CT29). In addition, CTA is more cost effective than PET.
The involvement of PVAT in the development of vascular diseases is now widely acknowledged and supported by evidence from both translational and clinical research. Baradaran et al. reported that the PFD in the internal carotid artery ipsilateral to a stroke or TIA was significantly higher than in the asymptomatic internal carotid artery30). However, their study was cross-sectional and retrospective. Zhang et al. showed that PFD is independently associated with IPH in the carotid artery31). Saba et al. demonstrated a positive association between PFD and contrast plaque enhancement on CTA, especially in symptomatic patients32). Thus, the present findings suggest that PFD could serve as a potentially valuable radiological marker for predicting ischemic stroke. Our study further demonstrated that PFD is an independent predictor of recurrent stroke or TIA. Although we analyzed the PVAT area, no significant correlation was found between recurrent stroke and TIA. This may be because the PVAT area mainly reflects the distribution of adipose tissue and cannot reveal its functional state, such as inflammation levels. In contrast, PFD can better reflect the physiological and pathological status of adipose tissue, particularly inflammation-related changes. Therefore, PFD as an indirect indicator of inflammation has a stronger predictive ability than the PVAT area.
LimitationsSeveral limitations associated with the present study warrant mention. First, it was a single-center study with a short follow-up period. Therefore, future prospective studies with larger sample sizes and longer follow-up periods are required. Second, inflammatory changes in the plaques and PVAT were not identified via histopathological validation. Third, the range of PFD measurements was uncertain, and our study adopted an established method for evaluating coronary arteries9).
In conclusion, our study demonstrated that CTA-derived PFD is an independent predictor of recurrent stroke or TIA. Incorporating PFD into clinical and imaging risk models may improve the risk prediction. These findings offer new insights into carotid risk stratification and early prevention of targeted diseases.
All the authors contributed to the conception and design of the study. The material preparation, data collection, and analysis were performed by Wu and Huang. The first draft of the manuscript was written by Tianqi Xu. Review and editing: Shuai Zhang and Ximing Wang. All authors commented on the previous versions of the manuscript. All authors have read and approved the final manuscript.
This study was supported by the National Science Foundation for Scientists of China (82271993 and 81871354).
The authors have no relevant financial or non-financial interests to disclose.
The authors of this manuscript declare no relationships with any companies whose products or services may be related to the subject matter of this article.
The datasets generated or analyzed during the study are available from the corresponding author upon reasonable request.
