Cerebral vascular disorder is a major cause of various types of dementia, and cerebral small vessel disease (cSVD) is the most common vascular contributor to cognitive dysfunction (1, 2). Signatures (or markers) of cSVD on MRI include enlarged perivascular spaces (EPVS), white matter hyperintensities (WMHs), recent small subcortical infarct, lacune, cerebral microbleed, and brain atrophy (3). The cross-sectional and longitudinal investigations have revealed that cognitive dysfunction can be associated with each of these cSVD markers, especially WMHs and lacune or lacunar infarcts (4–8). The effect of these markers may be additive. A proposed scoring system combines all the four MRI markers into a total cSVD score, to represent the total damage burden of cSVD in the brain (9). Several studies have identified an association between the total cSVD score and cognitive decline in the patients with hypertension or lacunar infarction (10, 11), and community-dwelling elders (12, 13). However, the patients with cognitive impairment in memory clinic differ from these cohorts in some aspects, in whom there are fewer cardiovascular risk factors, and more other coexisting neuropathological changes. The present study enrolled the patients from a memory clinic, most of whom suffered from amnestic complaints, and were with dementia, or at high risk of dementia. Within this population, an association between total cSVD burden and cognitive performance was explored.

Among the initial 337 patients enrolled, 25 patients were excluded clinically for brain injury secondary to trauma (two patients), chemotherapy (two patients), hypothyroidism (five patients), vitamin B deficiency (two patients), alcohol (four patients), severe depression (three patients) and large vessel infarctions or severe cerebral hemorrhage (seven patients). Additionally, 23 patients were excluded for unfinished or invalid MoCA assessment. Among them, two patients did not finish MoCA due to the very poor hearing and one patient due to the severe visual impairment (caused by cataract). Twenty patients did not get valid MoCA scale for advanced dementia [16 ones did not finish MoCA, the median MMSE score for these 16 patients was 10.5 (9.0–13.8); and four patients got extremely low and unreliable MoCA scale, the scales for them were 3, 3, 3, and 5] (Figure 1). Finally, 289 subjects were included in this analysis (Table 1). There were no significant differences on demographic characteristics and common vascular risk factors between the subjects included and those excluded except for a higher MMSE score [27.0 (23.0–29.0) vs. 18.0 (10.5–26.0), P < 0.001] (Supplementary Table S2 in the supplementary data).

Inter-rater reliability for all the cSVD markers was quite acceptable. Cohen κ for periventricular WMH Fazekas score was 0.820 [95% CI (0.767, 0.873), P = 0.027], and for deep WMH Fazekas score was 0.815 [95% CI (0.760, 0.870), P = 0.028]. Cohen κ for presence of ≥ 10 EPVS was 0.816 [95% CI (0.747, 0.885), P < 0.001], presence of ≥ 1 lacune: 0.834 [95% CI (0.771, 0.897), P < 0.001] and presence of ≥ 1 microbleed: 0.824 [95% CI (0.732, 0.916), P < 0.001]. The number of discordant scores of individual cSVD markers was showed in Supplementary Table S3 of the supplementary data. The MRI scans showed that 30.4% (88/289) of these patients had no markers, and 27.0% (78/289), 24.2% (70/289), 15.6% (45/289), and 2.8% (8/289) had 1, 2, 3, and 4 markers, respectively. The distribution and combination of cSVD markers in different cSVD score was showed in Table 2 and Figure 2. Because there were only eight patients displaying all four markers, the overall population was apportioned to four groups based on the cSVD scores of 0, 1, 2, and ≥ 3 markers.

The patients with higher cSVD scores were significantly older, with higher percentage of previous stroke, higher levels of MTA, and had lower MoCA scores (Table 3). Years of education, the duration of amnestic disorder before visiting, and other clinical factors did not differ among the four cSVD score groups, and there were also no significant differences in the serum glucose or lipid levels.

Spearman's correlation analysis revealed a significant negative association between cSVD score and MoCA score (r = −0.377, P < 0.001). A linear regression found that high cSVD score was a significant risk factor for low MoCA score [β = −1.667, 95% CI (−2.171, −1.163), P < 0.001]. After adjusting for gender, age, education, MTA levels, and classical cardiovascular risk factors, this significance remained [β = −0.729, 95% CI (−1.244, −0.213), P = 0.006] (Table 4). When the individual cSVD markers were included in the regression model rather than total cSVD score, the independent risk factors were: lacune, microbleed, and WMH. After adjusting for the same confounding factors mentioned above, only microbleed was a significant independent risk factor for low MoCA score [β = −3.007, 95% CI (−4.533, −1.480), P < 0.001].

This study investigated the influence of cSVD on cognitive status in a group of patients from a memory clinic. It was found that the cSVD burden, as reflected by the total cSVD score on MRI, was associated with decreased cognitive function, measured by the MoCA scale. Moreover, among the four cSVD markers, only microbleed was a significant independent risk factor for cognitive impairment in this cohort.

There have been a good number of studies of a link between cSVD and cognitive impairment, but most have focused on the influence of individual cSVD markers (5–8), or the combination of two markers (21). However, a total cSVD score has been proposed to reflect the total burden of cSVD on MRI (9, 15). Several studies showed an inverse association between total cSVD score and cognitive performance in the elder patients (12, 13) or in the patients with lacunar stroke or hypertension (10, 22). In contrast to these studies, we recruited a group of patients from a memory clinic whose main complaint was amnestic disorders; only 8.4% had a history of previous stroke. The results were similar to that of study populations with a higher incidence of cerebrovascular disease.

The total cSVD score of the present cohort [1.0 (0–2.0)] suggested a lower burden of cSVD compared with that of another memory clinic [2.0 (0–3.0)] (23), and smaller percentage of lacune, microbleed, and WMH, but the total cSVD score and cognitive dysfunction were similar. Rather than testing the different cognitive domains, we chose the MoCA scale to measure overall cognitive function, which is more practical in daily clinical work and screening. Considering another study about cSVD and cognition that used the MoCA scale in a population of community-dwelling elders (12), the participants in the present study were older (72.4 ± 9.3 cf. 70.1 ± 8.1), but better educated [15.0 (9.0–15.0) years of education cf. 79% primary school only], and the MoCA score was higher (21.2 ± 5.3 cf. 19.2 ± 4.8). Both studies showed a significantly negative association between total cSVD score and MoCA score.

The present research adds to the evidence of a link between cSVD and cognitive impairment, as well as verification for using the total cSVD score to study cSVD and cognition. Only a small portion of our patients had suffered stroke, and most of them had mild-to-moderate cognitive impairment [median MoCA score: 22.0 (18.0–25.0)]. Although the causal relationship between cSVD burden and cognitive impairment could be not identified in retrospective studies, the results indicate that cSVD might play an important role in cognitive dysfunction, even in people without obvious symptom of stroke, with a lower cSVD burden, and in the early stage of cognitive impairment. And the role of cSVD in the development of cognitive impairment need to be clarified in more longitudinal studies.

Several mechanisms have been proposed for cognitive damage caused by cSVD. One explanation is that subcortical damage caused by WMH and lacunae disrupts the integrity of the brain network, which impairs cognitive function that relies on cortical integration (23–25), and subsequently, structural disconnection leads to secondary cortical atrophy (26). Another underlying mechanism is the hypoperfusion and blood brain barrier leakage caused by neuro-glio-vascular damage in cSVD (27). Decreased regional cerebral blood flow was observed in cSVD, and local increase of cerebral blood flow in specific areas in response to neuronal activity was impaired. Hypoperfusion and ischemic damage occurs in the tissues vulnerable to ischemia, such as the hippocampus and the neocortex. On the other hand, dysfunction of the blood brain barrier leads to the leakage of fluids, proteins, and other plasma components into the perivascular tissues and further impairs oxygen and nutrient transport. Moreover, these components might activate inflammation or block amyloid β clearance and then cause Aβ plaque deposit and neuronal damage (28, 29). A third explanation is that cSVD lesions, such as EPVS and WMH, disrupt the interstitial fluid drainage of the glymphatic system, causing impaired clearance of metabolites (β amyloid and other neurotoxic proteins) and abnormal protein elimination (30, 31).

In the present study, when each marker was analyzed individually rather than the total cSVD score, only microbleed was a significant risk factor of lower MoCA performance. Recent studies identified microbleed as a risk factor for cognitive decline (5, 32), and our research provides further evidence of this. Other cSVD markers, especially WMH and lacune, have been significantly associated with cognitive impairment (4, 6–8), but this was not found in the present study. This may be because of the difference in cohorts; our participants showed a higher cSVD score, were much older, and had a remarkably higher incidence of microbleed. This may explain the significant association between microbleed and MoCA score and its dominant contribution in cSVD impact on cognition. In addition, the dichotomization of every marker in this scoring system led to a great loss of information about the severity of the MRI features (13), and thus the relative contribution of the individual markers in the regression is not recognized. Finally, and most importantly, an accumulative effect of individual cSVD features has been implied on global cognitive ability (13), so it is quite reasonable that the addition of several insignificant contributors lead to the final significant impact. It indicates from another side that it is quite necessary to use total cSVD burden instead of individual markers for studying cSVD as a whole.

The strengths of our study include the focus on a cognitively impaired population in a memory clinic, in which a significant association between cSVD burden and cognitive performance was revealed. Few studies of a similar cohort are reported. We chose the total cSVD score on MRI to represent the cSVD burden instead of individual markers, which treated the cSVD more globally and took the accumulative effect of individual markers into consideration. It is a useful method to represent the global cSVD burden in research and practice, even there is some information loss. The MoCA score was chosen for cognitive function. Both the total cSVD score and MoCA score are quite available and reliable in daily clinical work. This study confirms the viability of routinely applying cSVD evaluation in the memory clinic. Furthermore, this study confirms a significant link between microbleed and lowered cognitive performance that echoes the previous studies.

This research may be limited by subjectivity in the visual rating of the MRI, and the dichotomized scoring of each individual marker may overlook information. For improvement, a modified cSVD score was used to account for the severity of the individual cSVD markers, and this improved the identification of patients with vascular cognitive impairment (11, 33). Moreover, a recent study used artificial intelligence to evaluate the global cSVD burden quantitatively (the volumetric data of WMH, lacunes, EPVS, and chronic cortical infarcts) and found it a powerful predictor of long-term cognitive decline (34). The more objective and quantitative evaluation of cSVD burden is necessary for in-depth study on cSVD and it would be used more widely in the future. On the other hand, the dichotomization scale has its advantages, for example, it is relatively simple to use, less time-consuming and need no specific software. It would still be helpful for the primary evaluation of cSVD, especially for screening. In our opinion, the combination of the two methods for cSVD evaluation would be a better choice in clinical practice in the future. Also of note, as we have mentioned, the present study was retrospective cross-sectional research, with data from a single center with limited samples. More longitudinal multicenter studies are needed to investigate the long-term effects of cSVD on cognition.

This study showed a significant association between total cSVD score and cognitive impairment in a group of patients from a memory clinic. This is further evidence of the value of the total cSVD score for evaluating cSVD. These results warrant further work to explore the role of cSVD in cognition impairment, with larger cohorts, longer follow-up time, and a more accurate assessing system for cSVD burden.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of Peking University People's Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

YF collected, analyzed and interpreted the patient data, and was a major contributor in writing the manuscript. YX helped to analyze the data and was another major contributor in writing the manuscript. MS made contributions to the acquisition of data. HG revised the manuscript and helped to interpret the data. ZZ designed the present study and provided funding support. All the authors have read and approved the final version of this manuscript.

This study was supported by the Shandong Provincial Natural Science Foundation, China (Grants No. 2017G006021).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.