A total of 2351 acute ischemic stroke patients consecutively enrolled in the Department of Neurology, the First Affiliated Hospital of Xinxiang Medical University, Henan, China, between June 2013 and September 2018 were screened. The main inclusion criteria were: (1) ischemic stroke confirmed by imaging (computed tomography (CT) or magnetic resonance imaging (MRI)) within 72 hours of symptom onset, according to the criteria of the 10th Edition of International Classification of Diseases (ICD-10); (2) unrelated Han Chinese people and a Mini-Mental State Examination (MMSE) score of illiterate $>$ 17, primary $>$ 20, and secondary or higher $>$ 24 out of 30. MMSE scores were based on the cultural level of the subjects. Patients with brain tumors, brain trauma, thyroid dysfunctions, alcoholism, severe medical conditions, or neurological condition with consciousness were excluded. Also, those with severe stroke with NIHSS (National Institute of Health stroke scale) $>$ 25 or patients unable to complete assessment were excluded from this study. Exclusion criteria and the exact number of patients for each study stage are shown in Fig. 1.

Fig. 1.

Flow chart of participants’ selection.

After the onset of 3 months [28], 583 patients underwent a follow-up evaluation, and 486 patients were assessed with global cognitive functions by Changsha version of the Montreal Cognitive Assessment (MoCA-CS) [29] and according to the flow chart [30, 31]. Yu K.H’s protocol was referred (Fig. 1) [32]. Age- and gender-matched cognitively normal patients after ischemic stroke (MoCA $\ge$ 26) were used as a control. In addition, age- and gender-matched 150 healthy control subjects unrelated Han people with no ischemic stroke and cognitive dysfunction were randomly selected from the outpatients for health check-ups.

All subjects were evaluated based on MMSE and MoCA (Fig. 1). MoCA was scored based on a 30-point scale with 7 cognitive subtests: visual-spatial executive, naming, attention, language, abstraction, delayed recall, and orientation. The MoCA test was conducted to rectify the educational level of the bias correction, and subjects with education 12 years were given an additional 1 score on their test results. A higher score indicates the better cognitive function, and $\ge$ 26 were divided into normal [33]. MoCA-CS in China was used in this study [29].

Unified and standardized survey questionnaire terminologies were used in a quiet environment without interferences. The neuropsychological assessments for each patient were completed by one psychological surveyor on the same day.

The blood samples were collected during cognitive assessment between three months and half a year after stroke. And the blood sample were collected in EDTA-containing tubes and separated the serum immediately and frozen at –80 ${}^{\circ }$ C until further use. Genomic DNA was isolated using the TIANamp Blood DNA Kit (Tiangen, Beijing, China) according to the kit’s protocol. The lab technician performing the test was blinded to all patients and clinical information, and all lab work was carried out in the First Affiliated Hospital of Xinxiang Medical University.

The MMP-9 rs3918242 polymorphisms were performed by polymerase chain reaction (PCR) coupled with restriction fragment length polymorphism, as described in our previous studies [19].

The collection, processing and storage of blood sample are same as DNA extraction. The process of measuring MMP-9 is according to the MMP-9 ELISA Kit’s (Wuhan Fine Biotech Co., Ltd.; Wuhan; China) protocol.

SPSS 24.0 (SPSS Inc., Chicago, IL, USA) was used for statistical processing. The measurement data in the results were expressed as mean $\pm$ SD, and the count data were expressed as a percentage. Analysis of variance and two independent samples t-test were used for normally distributed data; rank-sum test was used for non-normally distributed data, and the $\chi$ ${}^2$ test was used for the count data to compare the statistical differences of demographic and clinical data between groups. Spearman correlation analysis was used to analyze the correlation between MMP-9 and MoCA score. Logistic regression analysis was used to analyze the risk factors of PSCID by estimating the odds ratio (OR) and 95% confidence interval (95% CI). The Hardy-Weinberg equilibrium of MMP-9 rs3918242 genotype was tested by Chi-square ( $\chi$ ${}^2$ ) goodness-of-fit test. The associations between the MMP-9 rs3918242 polymorphisms and PSCID risk were determined by logistic regression. The major homozygous genotype of MMP-9 rs3918242 was used as a reference, two-tailed, and p 0.05 was considered statistically significant.

There were significant differences in diabetes, hyperhomocysteinemia, serum MMP-9 levels, and MoCA scores between the three groups (all p 0.05). Compared with the NC group, the serum MMP-9 expression level in the PSCI group and the NCI group was significantly increased; yet, the MMP-9 expression level in the PSCI group was significantly higher than that in the NCI group (all p 0.05). Moreover, the MoCA scores of the PSCI group and the NCI group were significantly reduced compared with the NC group, and the MoCA scores of the PSCI group were significantly lower than that of the NCI group (p 0.05) (Table 1).

With diabetes mellitus, when hyperhomocysteinemia and serum MMP-9 level were used as independent variables, and PSCID as dependent variables, logistic regression analysis showed that diabetes mellitus, hyperhomocysteinemia, and elevated serum MMP-9 level were the main risk factors of PSCID (OR = 1.77, 95% CI: 1.07–2.93; OR = 1.88, 95% CI: 1.18–3.56; OR = 1.01, 95% CI: 1.01–1.02) (Table 2).

The Spearman correlation analysis showed that the serum MMP-9 level was negatively correlated with the total score of MoCA scale in patients with PSCID (p 0.05), and the serum MMP-9 level was negatively correlated with visual-spatial executive, naming, attention, language and delayed recall (p 0.05), but not with abstraction and orientation (p $\mathrm{>}$ 0.05) (Table 3).

The genotype frequency of MMP-9 rs3918242 in the NC group conformed to Hardy-Weinberg equilibrium (p $\mathrm{>}$ 0.05), while the genotype frequency of MMP-9 rs3918242 in the PSCID group and the NCI group did not (p 0.05). The $\chi$ ${}^2$ test showed that the genotypes of the three groups were significantly different ( $\chi$ ${}^2$ = 14.57, p 0.05) (Table 4).

The logistic regression analysis showed TC genotype carrying MMP-9 rs3918242 (OR = 3.91, 95% CI: 1.63–9.38) and CC genotype (OR = 2.79, 95% CI: 1.12–7.00) were associated with a significantly increased risk of PSCID (p 0.05). Also, in PSCID and NCI groups, TC genotype carrying MMP-9 rs3918242 (OR = 2.03, 95% CI: 1.02–4.05) significantly increased the risk of PSCID using TT genotype as a control (p 0.05) (Table 5).

The relationship between MMP-9 rs3918242 and PSCID risk of was further analyzed and stratified by demographic characteristics such as diabetes mellitus and hyperhomocysteinemia (Table 6). The logistic regression analysis showed that diabetes mellitus and hyperhomocysteinemia in the TCCC genotype of MMP-9 rs3918242 were associated with increased risk of PSCID (OR = 1.26, 95% CI: 1.14–2.66; OR = 1.24, 95% CI: 1.02–2.69). In addition, hyperhomocysteinemia increased the risk of PSCID (OR = 1.19, 95% CI: 1.10–2.38).