A web-based dynamic nomogram for individualized risk prediction of magnetic resonance imaging–defined cerebral small-vessel disease: Model development and internal validation

Abstract

Background

To develop and internally validate a web-based dynamic nomogram to predict magnetic resonance imaging–defined cerebral small-vessel disease.

Methods

We retrospectively enrolled 164 hospitalized adults who underwent brain magnetic resonance imaging between January 2022 and March 2023. The binary outcome was the presence versus absence of magnetic resonance imaging–defined cerebral small-vessel disease according to standardized STandards for ReportIng Vascular changes on nEuroimaging markers. Candidate clinical and laboratory variables were screened using univariable logistic regression and entered into multivariable logistic regression with stepwise selection. Model performance was assessed by receiver operating characteristic analysis, bootstrap calibration (1000 resamples), and decision curve analysis.

Results

Cerebral small-vessel disease was present in 73/164 (44.5%) patients. The final model included hypertension, glycated hemoglobin, homocysteine, C-reactive protein, triglycerides, and total cholesterol. Discrimination, defined as the model’s ability to distinguish patients with magnetic resonance imaging–defined cerebral small-vessel disease from those without magnetic resonance imaging–defined cerebral small-vessel disease, was good (area under the receiver operating characteristic curve: 0.82, 95% confidence interval: 0.77–0.88), with acceptable calibration after bootstrapping. Decision curve analysis showed a net clinical benefit across threshold probabilities of 0.10–0.80. An online dynamic nomogram was created to facilitate point-of-care risk estimation.

Conclusion

A simple model based on routine clinical data can provide individualized cerebral small-vessel disease risk estimates and may help prioritize magnetic resonance imaging evaluation and intensify the management of modifiable risk factors.

Keywords

Cerebral small-vessel disease risk factors logistic model contribution prediction nomogram

Introduction

Cerebral small-vessel disease (CSVD) encompasses a spectrum of disorders characterized by varying etiologies and pathological processes that affect components of the brain's vascular system, including small arteries, arterioles, capillaries, and veins.

Importantly, acute ischemic lacunar strokes arising from CSVD differ from other acute ischemic cerebrovascular stroke subtypes in their pathophysiology, clinical features, and prognosis: they result from occlusion of a single deep perforating arteriole rather than from large-artery atherothrombosis or cardioembolism, are frequently preceded by a fluctuating or stuttering onset, and carry a comparatively favorable short-term but less favorable long-term cognitive prognosis owing to the progressive nature of the underlying microangiopathy.¹ Histopathological examinations demonstrate a reduction in the intimal lumen of affected arteries, thickening of the vessel walls, and obstruction of blood perfusion and gas exchange.² Neuroimaging features of CSVD include white matter hyperintensities (WMH), enlarged perivascular spaces (EPVS), lacunar infarcts (LI), cerebral microbleeds (CMBs), and brain atrophy, in accordance with the STandards for ReportIng Vascular changes on nEuroimaging (STRIVE) framework. This disease contributes to 20%–30% of ischemic strokes and represents a significant health burden in low- and middle-income countries.^3,4

The etiology of CSVD is multifaceted and involves several mechanisms. Homocysteine (HCY), a sulfur-containing amino acid, has been linked to conditions such as cognitive impairment and CSVD.^5–7 Hypertensive vasculopathy, primarily affecting the perforating arterioles of the deep gray nuclei and white matter, is another critical factor in CSVD.⁸ Recent research has identified strong positive correlations between CSVD and risk factors such as smoking, an unhealthy diet, arterial hypertension, and ventricular fibrillation. These risk factors damage blood vessels and trigger the expression of hypoxia-sensitive genes and molecular cascades during hypoxic phases. The subsequent release of cytokines, inflammatory matrix metalloproteinases, and cyclooxygenase-2 induces inflammation, which compromises the blood–brain barrier (BBB), promotes the expression of adhesion molecules in endothelial cells and leads to leukocyte and platelet adhesion and microvessel occlusion.⁹ Additionally, hypertension can increase vascular fibrosis and alter the distribution of type IV collagen and other extracellular matrix components, resulting in vascular wall stiffness and reduced cerebral blood flow.¹⁰ Given the multitude of factors influencing CSVD, exploring its risk factors is crucial. Therefore, this study was conducted to assess the clinical features and risk factors of CSVD and to develop a multivariable logistic regression model to provide a reference for clinical diagnosis

Methods

Study cohort and route

This retrospective cohort study was conducted at The Second Hospital of Lanzhou University, a 5000-bed, Grade 3A national-level hospital. The study was approved by the Institutional Review Board of Lanzhou University Second Hospital and conducted in accordance with the Declaration of Helsinki. It included patients who sought medical treatment for various neurological symptoms from 1 January 2022 to 31 March 2023. The inclusion criteria were as follows: (a) age ≥30 years; (b) completed imaging studies; and (c) provision of written informed consent. The exclusion criteria were as follows: (a) severe dementia; (b) a history of brain surgery; and (c) absence of magnetic resonance imaging (MRI). Severe dementia was operationalized as a documented clinical diagnosis of dementia with a Clinical Dementia Rating of 2–3 (or an equivalent Mini-Mental State Examination score <10) recorded in the electronic medical record. This criterion was applied because patients with severe dementia are frequently unable to cooperate with MRI acquisition, resulting in motion artifact that compromises the assessment of CSVD markers, and cannot reliably provide the clinical history required for accurate risk-factor ascertainment. Their inclusion would therefore have introduced differential misclassification of both the imaging outcome and the predictors. Patient demographics (sex, age, smoking, and drinking habits), underlying diseases (stroke, hypertension, and diabetes), and laboratory test results (coagulation profiles, renal function, lipid levels, blood glucose, and glycated hemoglobin (HbA1c)) were extracted from the hospital's electronic medical record system. The study protocol received approval from the Medical Ethics Committee of The Second Hospital Affiliated to Lanzhou University, Lanzhou, China. All participants provided written informed consent (Figure 1).

Figure 1.

Flowchart of the case selection process.

Patients definition

MRI scans were performed using a 1.5-T scanner from GE Healthcare (Signa, Milwaukee, Wisconsin) or Siemens (Magnetom SONATA, Munich, Germany). The imaging protocol included T2-weighted fast spin-echo (repetition time (TR)/echo time (TE) = 5000/127 ms), T1-weighted spin-echo (TR/TE = 500/11 ms), and fluid-attenuated inversion recovery (FLAIR) sequences (TR/TE = 8800/127 ms; inversion time = 2250 ms), each consisting of 26 axial slices with a slice thickness of 5 mm and an interslice spacing of 1 mm.

CSVD definition

Cerebral lacunar stroke (LS) is caused by the occlusion of small perforating arteries and typically presents as lesions with an axial diameter of <20 mm.¹¹ LS lesions are classified by shape (tubular or elliptical) and size, ranging from 0 to 14 mm or from 15 to 20 mm.¹² The term leukoaraiosis, introduced by Hachinski, Potter, and Merskey in 1987, describes bilateral periventricular white matter hypodensity observed on CT scans, primarily in older adults.¹³ This corresponds to WMH, which are areas of diffuse white matter alteration that appear hyperintense on T2-weighted MRI, predominantly in the periventricular and subcortical regions. WMH are also common in asymptomatic patients; however, they are more prevalent in those affected by ADD.^14,15 Another manifestation of CSVD is CMBs. On MRI, cerebral microbleeds are best identified on susceptibility-sensitive sequences—T2*-weighted gradient-recalled echo (GRE) or susceptibility-weighted imaging (SWI)—on which they appear as small, rounded or ovoid hypointense foci with associated blooming artifact, rather than on conventional T2-weighted imaging.¹⁶ The perivascular space, which surrounds the perforating arteries and veins as they pass from the subarachnoid space through the brain parenchyma, serves as an important drainage pathway for interstitial fluid and solutes from the brain. The EPVS, also known as Virchow–Robin spaces, are most readily detected on T2-weighted MRI and are characterized by punctate or linear signal intensity resembling cerebrospinal fluid.^17,18

Imaging manifestations of CSVD. Based on MRI findings, CSVD can be categorized into cerebral infarction, leukoaraiosis, microhemorrhage, atrophy, and EPVS. In the studies we reviewed, CSVD was defined as patients presenting with new LI or microhemorrhages with >10 lesions on SWI or lesions scoring ≥2 on the Fazekas scale in the parietal ventricles or deep white matter regions (Table 1). Additionally, patients with grade 3 or 4 perivascular spaces in the basal ganglia or subcortical regions were included (Table 2). EPVS were graded separately in two standard anatomical regions—the basal ganglia and the centrum semiovale—on axial T2-weighted images using the validated five-point scale detailed in Table 2 (0, none; 1, 1–10; 2, 11–20; 3, 21–40; 4, >40), with the slice showing the greatest burden used for scoring. The study outcome was defined as a single binary variable: each participant was classified as having CSVD (case) if any one of the above STRIVE marker thresholds was met or as not having CSVD (control) if none was met. All images were independently rated by two neurologists blinded to the clinical data, and disagreements were resolved by consensus.

Table 1.

Leukoaraiosis was scored according to the Fazeks scale.

Parietal ventricle	Deep brain parenchyma
0 points—no lesions	0 points—no lesions
1 point—pencil-like thin lesion	1 point—pitting lesion
2 points—smooth halo	2 points—lesions begin to merge
3 points—lesion fusion macroglossia	3 points—lesions are widely fused

Table 2.

Scoring the perivascular space according to guidelines.

Subcortical	Basal joints
1 level—10	1 level—5
2 Level—maximum number of sections <10	2 level—5–10
3 Level—maximum number of sections: 10–20	3 level—>10, still countable
4 Level—maximum number of sections >20	4 levels—uncountable

In this study, diabetes was confirmed by the use of antidiabetic therapy or by repeated abnormal blood test results indicating a fasting blood glucose level of ≥7 mmol/L (126 mg/dL) or a 2-h postprandial glucose level of ≥11.1 mmol/L. Hypertension was defined as a blood pressure reading of ≥140/90 mmHg on three separate occasions or a previous diagnosis with ongoing antihypertensive treatment. A history of stroke was defined as physical or speech impairment confirmed by relevant imaging, and a significant smoking history was defined as smoking an average of 10 cigarettes per day for at least 20 years.

Measurement of various clinical indicators

Blood samples were collected after an overnight fast, typically between 6:00 AM and 8:00 AM. Levels of blood urea nitrogen, uric acid, urinary microalbumin, urinary α1-microglobulin, and urinary β2-microglobulin were measured using standard laboratory techniques. A Hitachi 7600 autoanalyzer was used for these measurements, whereas C-reactive protein (CRP) levels were analyzed using a latex-enhanced immunoturbidimetric assay. Serum cystatin C and creatinine levels were measured using automated particle-enhanced immunoturbidimetric and Jaffe kinetic methods, respectively.

After blood collection, the tubes were filled, the samples were processed and stored, and quality control procedures were performed to ensure adherence to laboratory standards, thereby guaranteeing the accuracy and reliability of the results. Blood collection devices, centrifuges, biochemical analyzers, and the required reagents were used for testing biochemical markers, with procedures carried out in a sequential manner, and biochemical markers strictly defined in accordance with guideline indicators. The biochemical analyzer used was an Olympus model from Japan, and the indices measured included low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides (TG), total cholesterol (TC), and HCY.

Statistical analysis

Continuous variables were presented as the mean ± SD or median (interquartile range (IQR)) and categorical variables as percentages (n (%)). All continuous variables summarized using the median were reported with the accompanying IQR throughout the manuscript. Continuous variables were compared using the t test or Mann–Whitney U test, as appropriate. Categorical variables were compared using the chi-square test or Fisher's exact test. The strength of the association between each candidate predictor and the binary CSVD outcome was quantified using odds ratios (ORs) with 95% confidence intervals (CIs) derived from logistic regression. Univariable logistic regression was first performed for each candidate variable, and variables with p <0.05 were entered into a multivariable logistic regression model. Backward stepwise selection based on the Akaike information criterion (AIC) was then applied to obtain the final parsimonious model. Model discrimination was evaluated using the receiver operating characteristic (ROC) curve. Internal validation was performed using 1000 bootstrap resamples, from which an optimism-corrected estimate of discrimination was obtained. Multicollinearity was assessed using variance inflation factors, and overall calibration was evaluated with the Hosmer–Lemeshow test (Tables S1 and S2; Figure S8). Prediction accuracy was assessed using calibration plots, and clinical utility was evaluated using decision curve analysis (DCA). In addition, the relative percent contribution of determinants to CSVD morbidity were calculated using partial R-squared values (R 2 β), and marginal R-squared values (R 2 M) were used to express the variance explained (contribution = R 2 β/R 2 M × 100%) by fixed factors in the model. The classification performance metrics reported in section “Results from the backward-type stepwise regression model for CSVD” (sensitivity, specificity, positive and negative predictive value, precision, recall, F1-score, and balanced accuracy) were derived from the confusion matrix of the final model at the optimal probability threshold determined by the Youden index. Based on the final logistic regression model, we developed a web-based dynamic nomogram using the R Shiny package to provide an accessible tool for individualized CSVD risk prediction. The nomogram generates both graphical and numerical representations of risk estimates with corresponding 95% CIs. A two-sided p value <0.05 was considered statistically significant. Data analyses were performed using R (version 4.1.3).

Result

Patient demographics and baseline clinical characteristics

A total of 207 patients were enrolled, of whom 164 were included in the analysis (Figure 1); 43 were excluded because of inadequate demographic or baseline clinical data. The median age was 66 years (IQR, 60–71 years). Patient characteristics are summarized in Table 3.

Table 3.

Demographics and baseline clinical characteristics of patients.

Variable	The final analysis set (n = 164)
Sex
Female	82 (50.00%)
Male	82 (50.00%)
Stroke
No	150 (91.46%)
Yes	14 (8.54%)
Hypertension
No	113 (68.90%)
Yes	51 (31.10%)
Diabetes
No	136 (82.93%)
Yes	28 (17.07%)
Smoking
No	144 (87.80%)
Yes	20 (12.20%)
Drinking
No	149 (90.85%)
Yes	15 (9.15%)
Cerebral small-vessel disease (CSVD)
No	91 (55.49%)
Yes	73 (44.51%)
Age (years)	65.54 ± 8.11
Body mass index (BMI; kg/m²)	24.47 ± 3.01
Fasting glucose (FG; mmol/L)	5.61 ± 1.68
Glycated hemoglobin (HbA1c; mmol/L)	5.97 ± 1.15
D-dimer (D-D; ng/L)	128.80 ± 96.57
Fibrinogen (Fb; g/L)	3.05 ± 1.53
Urea nitrogen (UN; mmol/L)	5.39 ± 1.57
Uric acid (UA; μmol/L)	303.65 ± 92.38
Creatinine (Cr; μmol/L)	81.75 ± 34.45
Cystatin C (CyC; mg/L)	0.97 ± 0.31
Homocysteine (HCY; μmol/L)	11.51 ± 8.40
C-reactive protein (CRP; mmol/L)	1.18 ± 0.84
Low-density lipoprotein (LDL; mmol/L)	2.67 ± 1.07
Triglycerides (TG; mmol/L)	1.22 ± 0.76
Total cholesterol (TC; mmol/L)	4.04 ± 1.07

Baseline characteristics comparison in CSVD

Baseline characteristics for the CSVD and non-CSVD groups are presented in Table 4. The mean ages were 66.70 ± 9.02 and 64.62 ± 7.21 years, respectively, with no significant difference between groups. Similarly, sex, age, and baseline renal and liver function test results did not differ significantly between groups. Significant differences were observed in hypertension, diabetes, smoking, and alcohol consumption. Markers, including body mass index (BMI) (p = 0.041), fasting glucose (FG) (p < 0.001), HbA1c (p < 0.001), D-D (p < 0.001), cystatin C (CyC) (p < 0.001), HCY (p < 0.001), CRP (p < 0.001), LDL (p < 0.001), TG (p < 0.001), and TC (p < 0.001) were significantly higher in the CSVD group. The distributions of the principal continuous predictors by CSVD status are shown as violin plots in Supplementary Figure S1, and the corresponding correlation structure is shown in Supplementary Figure S2. Notably, although all six predictors in the final model were positively correlated with CSVD, the inter-predictor correlations were uniformly weak (all |ρ| < 0.35), and the corresponding variance inflation factors were low (all <1.1), indicating the absence of meaningful multicollinearity (Supplementary Table S1).

Table 4.

Comparison of demographics and baseline clinical characteristics.

Variable	CSVD (n = 73)	No. of CSVD (n = 91)	p value
Sex			0.530
Female	39 (53.42%)	43 (47.25%)
Male	34 (46.58%)	48 (52.75%)
Hypertension			<0.001
No	32 (43.84%)	81 (89.01%)
Yes	41 (56.16%)	10 (10.99%)
Diabetes			0.001
No	52 (71.23%)	84 (92.31%)
Yes	21 (28.77%)	7 (7.69%)
Smoking			<0.001
No	56 (76.71%)	88 (96.70%)
Yes	17 (23.29%)	3 (3.30%)
Drinking			0.037
No	62 (84.93%)	87 (95.60%)
Yes	11 (15.07%)	4 (4.40%)
Age (years)	66.70 ± 9.02	64.62 ± 7.21	0.111
BMI (kg/m²)	25.00 ± 2.96	24.04 ± 2.99	0.041
FG (mmol/L)	6.36 ± 2.26	5.00 ± 0.48	<0.001
HbA1c (mmol/L)	6.39 ± 1.55	5.64 ± 0.44	<0.001
D-D (ng/L)	162.62 ± 110.08	101.67 ± 74.33	<0.001
Fb (g/L)	3.12 ± 0.93	2.99 ± 1.89	0.541
UN (mmol/L)	5.63 ± 1.83	5.19 ± 1.31	0.086
UA (μmol/L)	317.73 ± 96.36	292.35 ± 87.97	0.084
Cr (μmol/L)	86.10 ± 31.73	78.26 ± 36.29	0.142
CyC (mg/L)	1.05 ± 0.41	0.90 ± 0.19	0.005
HCY (μmol/L)	14.84 ± 11.58	8.85 ± 2.11	<0.001
CRP (mmol/L)	1.49 ± 0.92	0.94 ± 0.69	<0.001
LDL (mmol/L)	2.91 ± 1.36	2.47 ± 0.71	0.013
TG (mmol/L)	1.49 ± 0.96	1.00 ± 0.44	<0.001
TC (mmol/L)	4.28 ± 1.19	3.85 ± 0.91	0.011

BMI: bod

y mass index; Cr: creatinine; CRP: C-reactive protein; CSVD: cerebral small-vessel disease; CyC: cystatin C; D-D: D-dimer; Fb: fibrinogen; FG: fasting glucose; HbA1c: glycated hemoglobin; HCY: homocysteine; LDL: low-density lipoprotein; TC: total cholesterol; TG: triglycerides; UA: uric acid; UN: urea nitrogen.

Univariate and multivariate model analysis of CSVD

Table 5 shows significant differences between the CSVD and non-CSVD groups analyzed using univariate regression. This table presents results from the univariate model of CSVD, including variables such as hypertension (95% CI: 4.65–23.17), diabetes (95% CI: 1.93–12.19), smoking (95% CI: 2.50–31.78), alcohol consumption (95% CI: 1.17–12.68), BMI (95% CI: 1.01–1.24), FG (95% CI: 1.69–4.06), HbA1c (95% CI: 1.63–5.06), D-D (95% CI: 1.01–1.01), CyC (95% CI: 1.63–5.06), HCY (95% CI: 1.31–1.83), CRP (95% CI: 1.18–1.46), LDL (95% CI: 1.10–2.17), TG (95% CI: 1.87–6.94), and TC (95% CI: 1.10–2.05). All of these variables had p <0.05 in the univariate logistic regression analysis and were subsequently entered into the multivariable logistic regression model.

Table 5.

Results from the univariate model of CSVD.

Variables	β	SE	Z	p	OR (95% CI)
Hypertension
No					1.00 (Reference)
Yes	2.34	0.41	5.71	<0.001	10.38 (4.65–23.17)
Diabetes
No					1.00 (Reference)
Yes	1.58	0.47	3.35	<0.001	4.85 (1.93–12.19)
Smoking
No					1.00 (Reference)
Yes	2.19	0.65	3.37	<0.001	8.90 (2.50–31.78)
Drinking
No					1.00 (Reference)
Yes	1.35	0.61	2.22	0.026	3.86 (1.17–12.68)
BMI (kg/m²)	0.11	0.05	2.02	0.043	1.11 (1.01–1.24)
FG (mmol/L)	0.96	0.22	4.29	<0.001	2.62 (1.69–4.06)
HbA1c (mmol/L)	1.06	0.29	3.66	<0.001	2.87 (1.63–5.06)
D-D (ng/L)	0.01	0.00	2.83	0.005	1.01 (1.01–1.01)
UA (μmol/L)	0.00	0.00	1.73	0.084	1.00 (1.00–1.01)
UN (mmol/L)	0.18	0.10	1.76	0.078	1.20 (0.98–1.47)
HCY (μmol/L)	0.43	0.09	5.11	<0.001	1.54 (1.31–1.83)
CyC (mg/L)	1.69	0.59	2.85	0.004	5.42 (1.69–17.36)
CRP (mmol/L)	0.18	0.10	1.75	0.003	1.19 (1.18–1.46)
LDL (mmol/L)	0.43	0.17	2.49	0.013	1.54 (1.10–2.17)
TG (mmol/L)	1.28	0.33	3.83	<0.001	3.60 (1.87–6.94)
TC (mmol/L)	0.41	0.16	2.55	0.011	1.50 (1.10–2.05)

BMI: body mass index; CI: confidence interval; CRP: C-reactive protein; CSVD: cerebral small-vessel disease; CyC: cystatin C; D-D: D-dimer; FG: fasting glucose; HbA1c: glycated hemoglobin; HCY: homocysteine; LDL: low-density lipoprotein; OR: odds ratio; TC: total cholesterol; TG: triglycerides; UA: uric acid; UN: urea nitrogen; SE: standard error.

The multivariable logistic model highlighted significant determinants such as hypertension, HbA1c, HCY, CRP, TG, and TC. Individuals with hypertension had an OR of 7.51 (95% CI: 2.13–29.77) (Table 6). A 1-mmol/L increase in HbA1c was associated with a 2.82-fold increase in the odds of CSVD (95% CI: 1.23–10.91). Similarly, each 1 -µmol/L increase in HCY and each 1 mmol/L increase in CRP, TG, and TC were associated with increases in the odds of CSVD of 1.40 (95% CI: 1.17–1.75), 2.12 (95% CI: 1.11–4.17), 4.35 (95% CI: 1.64–12.87), and 2.20 (95% CI: 1.22–4.37), respectively.

Table 6.

Results from the multivariate model of CSVD.

Variable	n	OR 95% CI	p value
Hypertension
No	113	Reference
Yes	51	7.51 (2.13–29.77)	0.002
Diabetes
No	136	Reference
Yes	28	0.23 (0.03–1.58)	0.154
Smoking
No	144	Reference
Yes	20	5.22 (0.47–82.36)	0.196
Drinking
No	149	Reference
Yes	15	1.98 (0.18–19.28)	0.549
BMI (kg/m²)	164	0.95 (0.77–1.16)	0.621
FG (mmol/L)	164	1.66 (0.81–4.14)	0.212
HbA1c (mmol/L)	164	2.82 (1.23–10.91)	0.048
D-D (ng/L)	164	1.00 (1.00–1.01)	0.169
CyC (mg/L)	164	3.71 (0.36–53.40)	0.309
HCY (μmol/L)	164	1.40 (1.17–1.75)	<0.001
CRP (mmol/L)	164	2.12 (1.11–4.17)	0.024
LDL (mmol/L)	164	0.91 (0.44–1.76)	0.786
TG (mmol/L)	164	4.35 (1.64–12.87)	0.004
TC (mmol/L)	164	2.20 (1.22–4.37)	0.014

Results from the backward-type stepwise regression model for CSVD

The results of the backward stepwise regression model are shown in Figure 2. Compared with individuals without hypertension, patients with hypertension had an OR of 10.94 (95% CI: 3.80–35.57, p < 0.001). Increases of 1 mmol/L in HbA1c, HCY, CRP, TG, and TC were associated with increases in the odds of CSVD of 3.05 (95% CI: 1.54–7.51, p < 0.01), 1.45 (95% CI: 1.22–1.82, p < 0.001), 2.21 (95% CI: 1.25–4.10, p < 0.01), 3.80 (95% CI: 1.62–10.00, p < 0.01), and 1.90 (95% CI: 1.17–3.25, p < 0.05), respectively. The relative contributions of these variables were highlighted in Figure 2, with HCY concentration being the dominant contributor to CSVD morbidity, accounting for 60.60%, followed by HbA1c (27.84%). Hypertension (9.21%), TG (7.92%), TC (2.57%), and CRP (1.50%) accounted for the remaining variance, and the complete ranking of relative contributions is presented in Supplementary Figure S3. Consistent with this ranking, the multivariable-adjusted odds ratios confirmed that all six variables were independently associated with CSVD, with hypertension showing the largest categorical effect (OR = 9.38, 95% CI: 3.27–26.88), followed by TG (OR = 3.96), HbA1c (OR = 3.07), TC (OR = 1.73), and HCY (OR = 1.45 per µmol/L) (Supplementary Figure S4). The R 2 M was 0.467 (Table 7), indicating good agreement between the predicted model and the observed data. The ROC curve for the training set demonstrated good discrimination (area under the receiver operating characteristic curve (AUC): 0.82; 95% CI: 0.77–0.88) (Figure 3). Notably, this combined performance substantially exceeded that of any individual predictor, none of which achieved an AUC >0.81 (HCY, 0.81; hypertension, 0.73; CRP, 0.70; TG, 0.69; HbA1c, 0.65; TC, 0.62), confirming that the six variables contributed complementary rather than redundant information (Supplementary Figure S5 and Supplementary Table S2). To further characterize these relationships, we examined the prevalence of CSVD across ascending quartiles of each continuous predictor and observed clear dose–response gradients, most strikingly for HCY, for which prevalence rose from 17.3% in the lowest quartile to 90.9% in the highest quartile (Supplementary Figure S6 and Supplementary Table S3). To minimize deviations in the results, a calibration curve based on 1000 bootstrap resamples was constructed (Figure 4), showing good agreement between predicted probabilities and observed outcomes. This concordance was confirmed by a nonsignificant Hosmer–Lemeshow test (χ² = 4.30, p = 0.829), and the bootstrap-derived 95% CI band for the calibration curve encompassed the line of identity throughout (Supplementary Figure S7 and Supplementary Table S1). Furthermore, internal validation using 1000 bootstrap resamples yielded a mean optimism of only 0.016, resulting in an optimism-corrected AUC of 0.903, which was closely matched by the 10-fold cross-validation AUC of 0.899, indicating minimal overfitting (Supplementary Figure S8 and Supplementary Table S4). This DCA shows the multivariate model excels between thresholds 0.1–0.8, particularly between 0.2 and 0.5, where its net benefit surpasses “All” and “None.” At a threshold probability of 0.2, it significantly reduces unnecessary interventions, and at threshold 0.5, its benefits remain high, indicating strong clinical utility for identifying patients with a high risk of CSVD (Figure 5). The logistic regression model for diagnosing CSVD exhibited strong performance, with a sensitivity of 0.77, specificity of 0.94, positive predictive value of 0.92, and negative predictive value of 0.83. Precision was 0.92 and recall was 0.77, resulting in an F1-score of 0.84. The model had a prevalence of 0.44, a detection rate of 0.34, and a detection prevalence of 0.37. The balanced accuracy was 0.86 (Figure 6).

Figure 2.

Results from the optimized multivariate model of CSVD morbidity.

Figure 3.

ROC curves for optimized multivariate models for predicting the probability of CSVD occurrence.

Figure 4.

Calibration curves of the optimized multivariate model for predicting predict the probability of CSVD.

Figure 5.

DCA curves for optimized multivariate models for predicting the probability of CSVD occurrence.

Figure 6.

Performance evaluation heatmap of logistic regression model.

Table 7.

The partial R-squared (R 2 β) of the optimized multivariate model.

Variable	R 2 β ^a	Contribution (%)
Hypertension
No	Reference	−
Yes	0.043	9.21%
HbA1c (mmol/L)	0.131	27.84%
HCY (μmol/L)	0.283	60.60%
CRP (mmol/L)	0.007	1.50%
TG (mmol/L)	0.037	7.92%
TC (mmol/L)	0.012	2.57%

^aThe R 2 M was 0.467.

CRP: C-reactive protein; HbA1c: glycated hemoglobin; HCY: homocysteine; TC: total cholesterol; TG: triglycerides.

To assess the robustness and generalizability of the model, we performed a series of prespecified subgroup and sensitivity analyses. First, the prevalence of CSVD increased progressively across age strata, from 35.1% in patients younger than 65 years to 51.6% in those aged 65–74 years and 53.6% in those aged 75 years or older (trend OR = 1.54 per stratum, 95% CI: 1.01–2.36, p = 0.045). This increase was accompanied by a parallel rise in mean homocysteine levels (Supplementary Figure S9 and Supplementary Table S5). Importantly, the model retained strong discrimination within every age stratum (AUC: 0.93, 0.91, and 0.90, respectively; Supplementary Figure S10), as well as in both sexes (male AUC: 0.955; female AUC: 0.876; Supplementary Figure S11), indicating that its performance was not driven by any single demographic group. The corresponding classification metrics and subgroup discrimination are summarized in Supplementary Table S6. Second, when the metabolic burden was summarized as the number of metabolic syndrome criteria met, CSVD prevalence increased monotonically from 18.2% in participants meeting no criteria to 100% in those meeting three or more criteria (OR = 3.59 per additional criterion, 95% CI: 2.35–5.49, p < 0.001; Supplementary Figure S12 and Supplementary Table S7), underscoring the synergistic effect of clustered vascular risk factors. Finally, a sensitivity analysis in which the only nonsignificant predictor (CRP) was removed produced a negligible change in discrimination (ΔAUC: 0.001; Supplementary Table S8), confirming that the model does not depend on any single weak predictor.

Development and validation of a web-based dynamic nomogram for predicting CSVD risk using logistic regression model

To facilitate clinical decision making and risk stratification for CSVD, we developed a web-based dynamic nomogram (available at https://18582894689wxj.shinyapps.io/dynnomapp/). This predictive tool integrates the six key risk factors identified through logistic regression analysis. Figure 7(a) illustrates the graphical user interface of the nomogram, featuring input fields for the six identified risk factors: hypertension (categorical: Yes/No), HbA1c, HCY, CRP, TG, and TC. The interface displays a graphical summary with 95% CIs for the predicted probabilities, allowing clinicians to visualize the uncertainty associated with individualized risk predictions. Figure 7(b) presents the numerical output of the model, demonstrating its performance across different parameter combinations. The tabulated results show that, for patients with similar biochemical parameters (HbA1c: 5.971, HCY: 11.2, CRP: 1.33, TG: 1.22, and TC: 4.2), the presence of hypertension substantially increases the predicted probability of CSVD, with estimated probabilities ranging from 0.374–0.528 without hypertension to 0.701–0.941 with hypertension. Figure 7(c) presents the underlying statistical framework, including the logistic regression coefficients and model specifications. Significant associations were observed for most predictors: hypertension (β = 2.23849, p < 0.001), HbA1c (β = 1.12281, p < 0.01), HCY (β = 0.37119, p < 0.001), TG (β = 1.37531, p < 0.01), and TC (β = 0.54738, p < 0.05). The model demonstrated a good fit, with an AIC of 126.17 and convergence after seven Fisher scoring iterations. The residual deviance of 112.17 (157 degrees of freedom) also indicated an appropriate model fit. This dynamic nomogram should be regarded as a user-facing implementation of the internally validated logistic regression model rather than as a distinct analytic advance or a validated point-of-care instrument. It is intended to illustrate how individualized risk estimates could be generated in clinical practice. External validation and prospective evaluation will be required before clinical application.

Figure 7.

Development and validation of a web-based dynamic nomogram for predicting CSVD risk. (a) Interactive interface of the dynamic nomogram; (b) numerical summary of the predictive model; and (c) statistical parameters of the logistic regression model.

Discussion

Clinical manifestations of LS include pure motor stroke, pure sensory stroke, mixed sensory-motor stroke, ataxic hemiparesis, and dysarthria coupled with hand clumsiness, often remaining asymptomatic in approximately 20%–50% of older adults. We acknowledge that a more detailed breakdown of the classical lacunar syndromes—pure motor hemiparesis, pure sensory stroke, sensorimotor stroke, ataxic hemiparesis, dysarthria–clumsy hand syndrome, and atypical or incomplete presentations—would have enriched the clinical characterization of our cohort. However, because the present analysis was anchored on MRI-defined CSVD markers rather than on the acute clinical syndrome at presentation, and because a substantial proportion of CSVD in our cohort was radiologically silent (only 14 of 164 participants (8.5%) had a documented prior clinical stroke), syndromic subtype was not systematically recorded for every participant and could not be reconstructed retrospectively. Prospective syndrome-level phenotyping is a planned extension of this work.¹⁹ Molecular oxygen, which is essential for cellular function and life, plays a pivotal role in cell biology, impacting tissues and the entire organism. It is involved in signal transduction, regulation of gene transcription, and other cellular activities.^20–22 Although essential for life and involved in various cellular functions such as signal transduction, gene transcription regulation, and other cellular activities, oxygen can also exert deleterious effects on biological macromolecules in the form of reactive oxygen species (ROS) and free radicals. These harmful effects are attributable to the monovalent reduced state of oxygen, which leads to ROS formation.^20,21,23 Excessive ROS formation underlies the pathology of CSVD.²³ Recent studies have demonstrated that oxidative stress is a major factor affecting different types of CSVD, resulting from an imbalance between oxidative and antioxidant processes. This imbalance occurs when free radicals increase or antioxidant processes become inefficient.^24,25 Oxygen-derived free radicals initiate inflammatory processes and result in the formation of oxidized LDL, impacting the endothelial portion of the vessel wall.²⁶ Numerous studies have confirmed significant effects of oxidative deficiencies on the endothelium, accelerating the degenerative effects of blood flow in the central nervous system of patients with CSVD. Monovalent reactive forms of free radicals are encouraged by arterial hypertension, oxidation of low-density lipoprotein (oxLDL), diabetes mellitus, high levels of HCY, systemic infection, and cigarette smoking.²⁷ The present study showed that high blood pressure, elevated HbA1c, and high homocysteine levels are associated with CSVD, although the results concerning oxLDL differ from previous findings.

In the present study, over 60% of CSVD cases were attributed to HCY in the multivariate optimized CSVD model. Consistent with prior research, an elevation in serum HCY correlated with the emergence of various CSVD markers.^27,28 Although the specific process linking serum HCY and CSVD is not fully understood, several plausible mechanisms have been proposed to explain the correlation. Evidence suggests that endothelial dysfunction significantly affects the relationship between elevated serum HCY and CSVD.²⁹ Elevated HCY levels may induce endothelial dysfunction through multiple mechanisms.³⁰ In addition, oxidative stress, inflammation, and the promotion of neurodegenerative processes contribute to the association between serum HCY and CSVD.³⁰ It should be noted that, although our study could conclude that HCY may increase the occurrence of CSVD, we did not explore it in various types of CSVD,³¹ necessitating further investigation.

Hypertension and diabetes have been identified as important risk factors.^32,33 In this study, hypertension and HbA1c remained elevated during the progression of CSVD. A crucial pathological mechanism involves endothelial dysfunction, leading to vasoconstriction, inflammation, and proliferation of the affected vessels. Cerebral microhemorrhages are perivascular deposits of degraded blood products within macrophages, viewed from a neuropathological standpoint, forming as a result of extravasation of blood due to the degenerative state of small vessels and the subsequent breakdown of hemoglobin released from red blood cells.³⁴ The formation of CMBs involves multiple pathologic processes. Hypertension and amyloid angiopathy are currently recognized as crucial facets of vascular pathology. The emergence of stroke is significantly influenced by the presence of CMBs. Pathologically, hypertension-induced CSVD narrows the vessel lumen through collagen hypertrophy of the vessel wall and exudation of serum proteins.³⁵ This microangiopathy mainly affects the perforating small arteries of the deep gray nuclei and deep white matter,³⁵ causing hemorrhages in deeper brain regions such as the basal ganglia, thalamus, and brainstem. The etiology of CSVD is intricate and involves numerous mechanisms. In atherosclerosis and age-related CSVD, there is a reduction in the density of smooth muscle in the intima-media, leading to narrowing of the arterial lumen because of fibrous-hyaline deposits. This disease is closely associated with systemic vascular disease and shares common risk factors, including aging, diabetes mellitus, and hypertension.³⁶ Hypertension also damages the small blood vessels of the brain and initiates the expression of hypoxia-sensitive genes (e.g. hypoxia-inducible factor 1-alpha (HIF-1α)) and related molecular cascades during the hypoxic phase. The subsequent release of cytokines, inflammatory matrix metalloproteinases, and cyclooxygenase-2 ultimately results in inflammation. This, in turn, compromises the BBB and prompts endothelial cells to express adhesion molecules, resulting in leukocyte and platelet adherence as well as microvessel occlusion.³⁷ Hypertension-induced small-vessel lesions are likely due to diffuse cerebral vascular endothelial failure, which in turn leads to BBB damage, localized inflammation, and loss of cerebral blood flow (CBF) due to loss of autoregulation and a subsequent reduction in CBF.³⁸ Several studies have demonstrated a correlation between diabetic retinopathy and LS in patients with diabetes.^38,39 Researchers hypothesize that alterations in the blood–retinal barrier and related changes may also disrupt the BBB, thereby damaging small-vessel walls and promoting perivascular edema in patients with diabetes.⁴⁰ Furthermore, other studies have demonstrated a strong association between retinal microangiopathy⁴¹ and LS.⁴² These findings underscore the significant association between retinal microangiopathy and LS, highlighting that elevated blood pressure and blood glucose levels may lead to CSVD.

Inflammation triggers pathological changes in the small cerebral vessels, resulting in lacunar cerebral infarction. Cavernous cerebral infarction involves thickening and fibrin invasion of the middle layer of deep cerebral arteries, with associated lipid deposition and fibrosis, also known as lipodystrophy. In addition to structural abnormalities, inflammatory endothelial activation and dysfunction⁴³ are considered key factors contributing to cerebrovascular disease. CRP levels are strongly associated with endothelial dysfunction,⁴⁴ which may serve as an epiphenomenon or marker. Recent findings further suggest that CRP directly contributes to endothelial dysfunction by inducing cytokine release and the surface expression of adhesion molecules, leading to endothelial dysfunction.^45,46 Accordingly, and consistent with the findings of the present study, CRP may be the most responsive indicator of inflammation to predict the occurrence of CSVD.^41,47

The pathogenesis of CSVD and TG remains uncertain. However, several potential explanations exist. First, proliferation and hypertrophy of smooth muscle cells in the arterial wall may contribute. Insulin may induce steatosis by enhancing sympathetic nerve activity or acting as a growth factor,^48,49 leading to CSVD through diffuse cerebral hypoperfusion or blockage of small perforating arterioles.³⁵ Second, endothelial dysfunction, frequently accompanied by subclinical inflammation and oxidative stress in patients with insulin resistance (IR), should also be considered.^50,51 Third, atherosclerosis may provide a mechanistic link between the TG index and CSVD, with CSVD potentially resulting from diffuse hypoperfusion and microembolization due to atherosclerotic lesions. Unlike this study, LDLs are known to contribute to atherosclerosis development through mechanisms that facilitate arterial wall penetration and oxidation or by binding to glycosaminoglycans, reducing LDL receptor binding capacity, and prolonging plasma residence time.⁵² However, the present study did not identify a positive correlation with LDLs, whereas TC was associated with CSVD. Although previous studies did not link TC directly to CSVD, it is significantly correlated with serum lipids, including TC, LDL-C, and TG,⁵³ which may indirectly contribute to the development of CSVD.

Limitations include the retrospective design, which precluded causal investigation. Additionally, the study sample, drawn from a single center in China, may not be generalizable to other populations. Other potential confounding factors, such as B-vitamin status, which may influence HCY levels, were not evaluated and therefore may have biased our findings.⁵⁴ Future prospective studies should address these issues.

Our findings also speak to the broader interplay among white matter alterations, cognitive impairment, and clustered vascular risk factors. Diffuse white matter changes associated with CSVD are well-established correlates of cognitive decline and adverse functional outcomes. This relationship is particularly pronounced in patients with metabolic syndrome, in whom a cluster of individually subclinical or “silent” vascular risk factors—including hypertension, dysglycemia, central adiposity, and atherogenic dyslipidemia—acts synergistically to accelerate cerebrovascular pathology during aging.⁵⁵ Consistent with this concept, the prevalence of CSVD in our cohort increased monotonically with the number of metabolic syndrome criteria met, reaching 100% among participants meeting three or more criteria. This dose–response pattern supports the view that the co-occurrence of metabolic risk factors, rather than any single factor in isolation, is a principal driver of the microangiopathic burden that underlies later cognitive impairment. A further dimension concerns the oldest patients. The demographic profile and risk-factor distribution of acute stroke differ substantially in the very old (≥85 years), among whom atrial fibrillation and cardioembolism become relatively more prominent and the classical metabolic risk factors comparatively less so.⁵⁶ Our cohort contained only one participant aged ≥85 years, precluding a dedicated very old subgroup analysis. Nevertheless, as shown in the Results, both CSVD prevalence and mean homocysteine levels increased across the age strata we could examine, whereas model discrimination remained high in every stratum. Dedicated characterization of the very old population is therefore an important direction for future research.

Several lines of future research arise naturally from this work. First, external and prospective multicenter validation, ideally across populations of differing ethnicity and including a substantial very old subgroup, is needed before the tool could be considered for clinical deployment. Second, incorporation of B-vitamin and folate status, ambulatory blood pressure variability, and quantitative imaging metrics may refine risk estimation. Third, linking the model to longitudinal cognitive and functional outcomes would clarify whether earlier CSVD risk stratification translates into measurable clinical benefit. Finally, syndrome-level phenotyping of lacunar presentations and analysis of individual CSVD marker subtypes (WMH, lacunes, CMBs, and EPVS) would enable more granular, mechanism-specific prediction. In keeping with these caveats, and as the reviewers rightly note, the web-based dynamic nomogram presented here should be regarded as a preliminary, user-facing implementation of an internally validated logistic regression model rather than as a validated point-of-care clinical instrument. Its scientific validity rests first on the reliability of the MRI-defined CSVD outcome and the robustness of the underlying model, both of which we have sought to strengthen during revision. In the absence of external validation, prospective evaluation, or demonstrated impact on clinical decision making, the nomogram is offered to illustrate how the model could ultimately be operationalized, and any clinical use should await confirmatory studies.

Conclusion

In this study, we examined the risk factors associated with CSVD and evaluated their relative contributions using a final multivariable model derived through stepwise regression. We first characterized the distribution of CSVD in the study population and then assessed the variability of its risk factors. Multivariable regression analysis subsequently identified differences between groups, and forest plots were used to visualize these findings. Hypertension, HbA1c, HCY, CRP, TG, and TC were identified as independently associated with CSVD, alongside numerous other risk factors and laboratory indicators. HCY was the dominant contributor, followed by HbA1c and hypertension, each significantly influencing CSVD. Other risk factors contributed to a lesser extent.

Supplemental Material

sj-docx-1-imr-10.1177_03000605261466575 - Supplemental material for A web-based dynamic nomogram for individualized risk prediction of magnetic resonance imaging–defined cerebral small-vessel disease: Model development and internal validation

Supplemental material, sj-docx-1-imr-10.1177_03000605261466575 for A web-based dynamic nomogram for individualized risk prediction of magnetic resonance imaging–defined cerebral small-vessel disease: Model development and internal validation by Xuhui Liu, Xujie Wang, Rongfei Xie, Minmin Li, Zhaohui Liu and Zheng Pan in Journal of International Medical Research

Supplemental Material

sj-docx-2-imr-10.1177_03000605261466575 - Supplemental material for A web-based dynamic nomogram for individualized risk prediction of magnetic resonance imaging–defined cerebral small-vessel disease: Model development and internal validation

Supplemental material, sj-docx-2-imr-10.1177_03000605261466575 for A web-based dynamic nomogram for individualized risk prediction of magnetic resonance imaging–defined cerebral small-vessel disease: Model development and internal validation by Xuhui Liu, Xujie Wang, Rongfei Xie, Minmin Li, Zhaohui Liu and Zheng Pan in Journal of International Medical Research

Footnotes

Acknowledgments

The authors would like to express their sincere gratitude to The Second Hospital of Lanzhou University and the The Affiliated Hospital of Qinghai University for their invaluable support throughout the conduct of this research. Their assistance and collaboration have been instrumental in the successful completion of this study.

ORCID iD

Xuhui Liu

Ethics approval statement

The authors confirm that the study has been approved by the Hospital Ethics Committee of The Second Hospital of Lanzhou University (Approval No: 2024A-1392) and informed consents were obtained from all included participants. This study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the 2025 Graduate Research and Practice Innovation Project of Qinghai University (No. 2025-GMKY-14).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to institutional data protection policies but are available from the corresponding author (panzh888@126.com)on reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Arboix

Massons

García-Eroles

, et al. Nineteen-year trends in risk factors, clinical characteristics and prognosis in lacunar infarcts. Neuroepidemiology 2010; 35: 231–236.

Markus

de Leeuw

. Cerebral small vessel disease: recent advances and future directions. Int J Stroke 2023; 18: 4–14.

Fisher

Miners

Love

. Pathological changes within the cerebral vasculature in Alzheimer’s disease: new perspectives. Brain Pathol (Zurich, Switzerland) 2022; 32: e13061.

Pantoni

. Cerebral small vessel disease: from pathogenesis and clinical characteristics to therapeutic challenges. Lancet Neurol 2010; 9: 689–701.

Jakubowski

. Homocysteine modification in protein structure/function and human disease. Physiol Rev 2019; 99: 555–604.

Jan

Cueto

Jiang

, et al. Molecular processes mediating hyperhomocysteinemia-induced metabolic reprogramming, redox regulation and growth inhibition in endothelial cells. Redox Biology 2021; 45: 102018.

Javierre-Petit

Schneider

Kapasi

, et al. Neuropathologic and cognitive correlates of enlarged perivascular spaces in a community-based cohort of older adults. Stroke 2020; 51: 2825–2833.

Yang

Kimura-Ohba

Thompson

, et al. Vascular tight junction disruption and angiogenesis in spontaneously hypertensive rat with neuroinflammatory white matter injury. Neurobiol Dis 2018; 114: 95–110.

Benveniste

Nedergaard

. Cerebral small vessel disease: a glymphopathy? Curr Opin Neurobiol 2022; 72: 15–21.

10.

Kumar

Yeo

Whittaker

, et al. Vascular collagen type-IV in hypertension and cerebral small vessel disease. Stroke 2022; 53: 3696–3705.

11.

Wardlaw

Smith

Biessels

, et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol 2013; 12: 822–838.

12.

Del Bene

Makin

Doubal

, et al. Variation in risk factors for recent small subcortical infarcts with infarct size, shape, and location. Stroke 2013; 44: 3000–3006.

13.

Prins

Scheltens

. White matter hyperintensities, cognitive impairment and dementia: an update. Nat Rev Neurol 2015; 11: 157–165.

14.

Ferrer

Andrés-Benito

. White matter alterations in Alzheimer’s disease without concomitant pathologies. Neuropathol Appl Neurobiol 2020; 46: 654–672.

15.

Sarabia-Cobo

Pérez

Hermosilla

, et al. Apathy and leukoaraiosis in mild cognitive impairment and Alzheimer’s disease: multicenter diagnostic criteria according to the latest studies. Dementia Geriatr Cogn Disord Extra 2014; 4: 228–235.

16.

Joshi

Morley

. Cognitive impairment. Med Clin N Am 2006; 90: 769–787.

17.

Potter

Doubal

Jackson

, et al. Enlarged perivascular spaces and cerebral small vessel disease. Int J Stroke 2015; 10: 376–381.

18.

Yang

Qin

Yang

, et al. The relationship between ambulatory blood pressure variability and enlarged perivascular spaces: a cross-sectional study. BMJ Open 2017; 7: e015719.

19.

Jimenez Diaz

Kapadia

Linke

, et al. Cerebral embolic protection during transcatheter heart interventions. EuroIntervention 2023; 19: 549–570.

20.

Kawamura

Muraoka

. Exercise-induced oxidative stress and the effects of antioxidant intake from a physiological viewpoint. Antioxidants (Basel, Switzerland) 2018; 7: 119.

21.

Höhn

Weber

Jung

, et al. Happily (n)ever after: aging in the context of oxidative stress, proteostasis loss and cellular senescence. Redox Biol 2017; 11: 482–501.

22.

Rahal

Kumar

Singh

, et al. Oxidative stress, prooxidants, and antioxidants: the interplay. BioMed Res Int 2014; 2014: 761264.

23.

Navarro-Yepes

Burns

Anandhan

, et al. Oxidative stress, redox signaling, and autophagy: cell death versus survival. Antioxi Redox Signal 2014; 21: 66–85.

24.

Liguori

Russo

Curcio

, et al. Oxidative stress, aging, and diseases. Clin Intervent Aging 2018; 13: 757–772.

25.

Zhang

Wang

Lei

, et al. Redox signaling: potential arbitrator of autophagy and apoptosis in therapeutic response. Free Rad Biol Med 2015; 89: 452–465.

26.

Yao

Wang

Zhang

, et al. Klotho ameliorates oxidized low density lipoprotein (ox-LDL)-induced oxidative stress via regulating LOX-1 and PI3 K/Akt/eNOS pathways. Lipids Health Dis 2017; 16: 77.

27.

Kloppenborg

Nederkoorn

van der Graaf

, et al. Homocysteine and cerebral small vessel disease in patients with symptomatic atherosclerotic disease. The SMART-MR study. Atherosclerosis 2011; 216: 461–466.

28.

Lau

Schulz

, et al. Total small vessel disease score and risk of recurrent stroke: validation in 2 large cohorts. Neurology 2017; 88: 2260–2267.

29.

Huijts

Duits

van Oostenbrugge

, et al. Accumulation of MRI markers of cerebral small vessel disease is associated with decreased cognitive function. A study in first-ever lacunar stroke and hypertensive patients. Front Aging Neurosci 2013; 5: 72.

30.

Ihara

Yamamoto

. Emerging evidence for pathogenesis of sporadic cerebral small vessel disease. Stroke 2016; 47: 554–560.

31.

Teng

Feng

Liu

, et al. Cerebral small vessel disease mediates the association between homocysteine and cognitive function. Front Aging Neurosci 2022; 14: 868777.

32.

Kase

Wolf

Chodosh

, et al. Prevalence of silent stroke in patients presenting with initial stroke: the Framingham study. Stroke 1989; 20: 850–852.

33.

Jackson

Hutchison

Dennis

, et al.

Differing risk factor profiles of ischemic stroke subtypes: evidence for a distinct lacunar arteriopathy?

Stroke 2010; 41: 624–629.

34.

Lawrence

Pretorius

Ezra

, et al. Early detection of cerebral microbleeds following traumatic brain injury using MRI in the hyper-acute phase. Neurosci Lett 2017; 655: 143–150.

35.

Charidimou

Pantoni

Love

. The concept of sporadic cerebral small vessel disease: a road map on key definitions and current concepts. Int J Stroke 2016; 11: 6–18.

36.

Wang

Lin

Zhao

, et al. Sex disparity of cerebral white matter hyperintensity in the hypertensive elderly: the Shanghai Changfeng study. Hum Brain Mapp 2023; 44: 2099–2108.

37.

Doubal

MacLullich

Ferguson

, et al. Enlarged perivascular spaces on MRI are a feature of cerebral small vessel disease. Stroke 2010; 41: 450–454.

38.

Wardlaw

. Blood-brain barrier and cerebral small vessel disease. J Neurol Sci 2010; 299: 66–71.

39.

Wardlaw

Doubal

Armitage

, et al. Lacunar stroke is associated with diffuse blood-brain barrier dysfunction. Annal Neurol 2009; 65: 194–202.

40.

Doubal

Hokke

Wardlaw

. Retinal microvascular abnormalities and stroke: a systematic review. J Neurol Neurosurg Psychiatr 2009; 80: 158–165.

41.

Mitaki

Nagai

Oguro

, et al. C-reactive protein levels are associated with cerebral small vessel-related lesions. Acta Neurol Scand 2016; 133: 68–74.

42.

Umemura

Kawamura

Hotta

. Pathogenesis and neuroimaging of cerebral large and small vessel disease in type 2 diabetes: a possible link between cerebral and retinal microvascular abnormalities. J Diabet Invest 2017; 8: 134–148.

43.

Knottnerus

Ten Cate

Lodder

, et al. Endothelial dysfunction in lacunar stroke: a systematic review. Cerebrovascul Dis (Basel, Switzerland) 2009; 27: 519–526.

44.

Khreiss

József

Potempa

, et al. Conformational rearrangement in C-reactive protein is required for proinflammatory actions on human endothelial cells. Circulation 2004; 109: 2016–2022.

45.

Verma

Badiwala

, et al. Endothelin antagonism and interleukin-6 inhibition attenuate the proatherogenic effects of C-reactive protein. Circulation 2002; 105: 1890–1896.

46.

Bakhashab

Megantara

Mahaputra

, et al. Decoding of miR-7-5p in colony forming unit-hill colonies as a biomarker of subclinical cardiovascular disease-A MERIT study. Int J Mol Sci 2023; 24; 11977.

47.

van Dijk

Prins

Vermeer

, et al. C-reactive protein and cerebral small-vessel disease: the Rotterdam scan study. Circulation 2005; 112: 900–905.

48.

Jian

Su-Mei

Xue

, et al. Association and interaction between triglyceride-glucose index and obesity on risk of hypertension in middle-aged and elderly adults. Clin Exp Hyperten (New York, N.Y. : 1993) 2017; 39: 732–739.

49.

Bornfeldt

Tabas

. Insulin resistance, hyperglycemia, and atherosclerosis. Cell Metab 2011; 14: 575–585.

50.

Mao

Zhou

, et al. The triglyceride-glucose Index predicts coronary artery disease severity and cardiovascular outcomes in patients with non-ST-segment elevation acute coronary syndrome. Dis Markers 2019; 2019: 6891537.

51.

Lee

Kang

Warach

. Silent new brain lesions: innocent bystander or guilty party? J Stroke 2016; 18: 38–49.

52.

Djekic

Vekic

Zeljkovic

, et al. HDL subclasses and the distribution of paraoxonase-1 activity in patients with ST-segment elevation acute myocardial infarction. Int J Mol Sci 2023; 24: 9384.

53.

Wei

, et al. Association between small dense low-density lipoprotein cholesterol and neuroimaging markers of cerebral small vessel disease in middle-aged and elderly Chinese populations. BMC Neurol 2021; 21: 436.

54.

Staals

Booth

Morris

, et al. Total MRI load of cerebral small vessel disease and cognitive ability in older people. Neurobiol Aging 2015; 36: 2806–2811.

55.

Segura

Jurado

Freixenet

, et al. White matter fractional anisotropy is related to processing speed in metabolic syndrome patients: a case-control study. BMC Neurol 2010; 10: 64.

56.

Torres-Riera

Arboix

Parra

, et al. Predictive clinical factors of in-hospital mortality in women aged 85 years or more with acute ischemic stroke. Cerebrovascul Dis (Basel, Switzerland) 2025; 54: 11–19.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.02 MB

1.83 MB