Baseline clinical assessment
Baseline subject demographics, risk factors (blood pressure, body mass index, drinking and smoking status), medical history (hypertension, diabetes, dyslipidemia and heart disease) and medication use (antihypertensive, lipid-lowing, antidiabetic, antiplatelet, anticoagulant) were collected. Fasting venous blood samples were collected at baseline. Several laboratory tests were conducted, including routine blood examination, total cholesterol, triglycerides, high-density lipoprotein, low-density lipoprotein, fasting blood glucose, and homocysteine levels. All of the data were evaluated through face-to-face interviews by centralized trained personnel at Lishui Hospital.
MRI acquisition and assessment
Participants underwent brain MRI scans by a 3.0T MRI scanner (Ingenia 3.0T, Philips, Best, The Netherlands) based on a standardized protocol. The MRI sequences included three-dimensional T1-weighted magnetization prepared rapid acquisition gradient-echo (3D T1w MPRAGE), axial T2-weighted, fluid-attenuated inversion recovery (FLAIR), and axial susceptibility-weighted imaging. The detailed scanner parameters are listed in Additional file
1: Table S1. Imaging data were collected in digital imaging and communications in medicine (DICOM) format on discs and further analyzed by the imaging research centre at Beijing Tiantan Hospital. WMH was defined as increased brightness on T2 images in the brain white matter. The periventricular and deep WMH were evaluated according to the Fazekas rating scale [
15]. Lacune was defined as a rounded or ovoid lesion of CSF signal measuring 3–20 mm in diameter. Cerebral microbleeds (CMBs) were rounded, hypodense lesions with sizes of 2–10 mm in a gradient-recalled echo image or susceptibility-weighted image. The total number of lacunes and CMBs were also recorded. Enlarged perivascular spaces (EPVS) was defined as small (< 3 mm) punctate or linear hyperintensities on T2 images, and EPVS in the basal ganglia was graded with the semi-quantitative rating scale developed by the Edinburg group [
16]. Imaging assessment of each CSVD marker was rated by two well-trained raters (M Zhou, Y Chen, J Pi, and M Zhao, one rater was responsible for two markers) who were blinded to the participants’ clinical data, according to the standards for reporting vascular changes on neuroimaging (STRIVE) [
17]. Images with inconsistent results were finally assessed by another senior neurologist (Y Yang) who was blinded to the initial results. The kappa coefficients of CSVD markers on brain MRI between raters were as follows: 0.82 for the Fazekas scale of WMH, 0.80 for the presence of lacune, 0.80 for the presence of CMB and 0.90 for the severity of BG-EPVS.
Based on a previously described and validated total CSVD score designed by Wardlaw’s group, we rated the total CSVD burden on an ordinal scale from 0 to 4. One point was allocated for WMH burden (PV-WMH Fazekas 3 or deep-WMH Fazekas 2–3), presence of lacune, presence of CMB, and moderate-to-severe basal ganglia EPVS (BG-EPVS) (
N > 10) [
18]. Furthermore, we also evaluated the modified total CSVD burden using a recently validated ordinal score designed by Rothwell’s group, which ranges from 0 to 6 [
19]. One point was allocated for the presence of lacunes, CMB burden (
N 1–4), severe BG-EPVS (
N > 20), modified WMH burden (total periventricular + subcortical WMH grade 3–4), two points were allocated for CMB burden (
N ≥ 5) and modified WMH burden (total periventricular + subcortical WMH grade 5–6).
Statistical analysis
Categorical variables are presented as frequencies and percentages, while continuous variables are presented as the mean with standard deviation or median with interquartile range. The baseline characteristics of the participants were compared between the absence of CSVD (total CSVD burden = 0) and the presence of CSVD (total CSVD burden ≥ 1) according to the total CSVD score designed by Wardlaw et al. and Rothwell et al., respectively. The Chi-square test was used for categorical variables (such as sex and medical history), and ANOVA or the Kruskal–Wallis test was used for continuous variables (such as age and neutrophil count).
The relationships of the neutrophils, NLR or SII with the presence and severity of CSVD or CSVD image markers were evaluated. For the total CSVD burden, modified total CSVD burden, modified WMH burden and CMB burden, ordinal logistic regression models were conducted and common odds ratios (cORs) with their 95% confidence intervals (CIs) were calculated. For the presence of CSVD and other CSVD image markers, logistic regression models were performed and odds ratios (ORs) with their 95% CIs are presented. Two models were conducted for each outcome. In Model 1, covariates including age and sex were adjusted. In Model 2, age, sex, body mass index, diabetes, hypertension, total cholesterol, high-density lipoprotein, low-density lipoprotein, fasting blood glucose, homocysteine, previous dyslipidemia, previous heart disease, current smoking, current drinking, previous antiplatelet, anticoagulant, antihypertensive, antidiabetic, and lipid-lowering drug use were adjusted. The hypertension adjusted in Model 2 was defined as either self-reported hypertension previously diagnosed by a physician or current use of antihypertensive drugs or systolic blood pressure ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg. The diabetes adjusted in Model 2 was defined as a self-reported diabetes previously diagnosed by a physician or current use of antidiabetic drugs of fasting plasma glucose ≥ 7.0 mmol/L or 2-h postload glucose ≥ 11.1 mmol/L or hemoglobin A1c ≥ 6.5%.
For all analyses, correction for multiple comparisons was done by using the false discovery rate (FDR) approach, and the result was presented as P_FDR. Statistical significance was set at a P_FDR < 0.05. Association not reaching this threshold, but showing a p < 0.05, was considered suggestive of an association.
Moreover, two-sample MR approaches were used to evaluate the association of NC with CSVD using summary-level data of the SNP-NC and SNP-CSVD associations. The random-effect inverse-variance weighted (IVW) is the most widely applied method for MR analysis because it provides robust causal estimates under absence of directional pleiotropy [
25]. Here, IVW was performed to estimate the effect by generalized weighted linear regression of SNP-CSVD against SNP-NC estimates with the inverse-variance of SNP-CSVD estimate as weights and the intercept set to zero. Based on summary statistics for each trait, we extracted variants with
r2 < 0.001 (linkage disequilibrium),
p < 5 × 10
–8 (genome-wide significance) and minor allele frequency > 0.05. The correction for multiple testing was done for each outcome with FDR. The corrected P_FDR < 0.05 was considered as statistically significant. For heterogeneity assessment of each instrument in the IVW analysis, Cochran’s Q statistic was used (
p < 0.10 indicates the presence of nominal heterogeneity) [
26]. In sensitivity analysis, we further applied four alternative MR methods that were more robust to the inclusion of pleiotropic and/or invalid instruments, including MR-Egger regression, weighted median, Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR PRESSO) and weighted mode methods. As a measure of pleiotropy of included genetic variants (
p < 0.05 indicates statistical significance), MR-Egger allows for the estimation of an intercept term which used to represent the average pleiotropic effects of all SNPs [
27]. The weighted median method allows the use of invalid instruments under the assumption that at least 50% of the instruments used in the MR analysis are valid [
28]. MR PRESSO allows to detect and correct for horizontal pleiotropic outlier SNPs in multi-instrument summary-level MR testing [
29]. The weighted mode method assumes a plurality of genetic variants are valid instruments, and has low bias [
30]. The effect estimates of genetically predicted NC on CSVD phenotypes were presented as odds ratios (ORs) with their 95% CIs per 1-standard deviation (SD) increment of NC.
Additionally, the net reclassification index (NRI) and absolute integrated discrimination improvement (IDI) were calculated to establish the performance of the addition of NC, NLR or SII to the basic model. The covariates involved in the basic model are the traditional vascular risk factors associated with cerebral small vessel disease, including age, sex, body mass index, diabetes, hypertension, total cholesterol, high-density lipoprotein, low-density lipoprotein, fasting blood glucose, homocysteine, previous dyslipidemia, previous heart disease, current smoking, current drinking, previous antiplatelet, anticoagulant, antihypertensive, antidiabetic, and lipid-lowering drug use. The NRI is used to quantify the amount of correct reclassification introduced by using a model with added variables, while the IDI is used to quantify the increase in the separation of events and nonevents. NRI or IDI > 0 reflects improvements in performance between new and old models [
31,
32].
All analyses were performed using R 4.0.3 (R Development Core Team) and SAS software version 9.4 (SAS Institute Inc, Cary, NC).