Lacune and Large Perivascular Space: Two Kinds of Cavities Are of Different Risk Factors and Stroke Risk

Lacunes are small, fluid-filled cavities frequently seen on brain imaging in elderly individuals, indicating a healed stage of small deep brain infarcts or hemorrhage. Perivascular spaces (PVSs) are another kind of fluid-filled cavity commonly seen in elderly people with disparate pathological features [1]. According to brain MRI and pathological study, PVSs are found to be usually small in size, and 3 mm has been widely accepted as the diameter cutoff in differentiating lacunes and PVSs on MRI imaging [2‒4]. However, accumulating evidence suggests that large PVSs (L-PVSs), with a diameter >3 mm, of similar size and shape to lacunes, are highly prevalent in the normal aging population [3, 5]. In a 3C-Dijon MRI study, L-PVSs were detected in 33.2% of 1,826 community participants older than 65 years [3]. With similar morphological features and signal intensity seen in brain MRI, L-PVSs are frequently misdiagnosed as lacunes in clinical practice.

The lacune has long been regarded as the original feature of cerebral small vessel disease (CSVD) and is found to be associated with an increased risk of stroke, motor dysfunction, cognitive impairment, and dementia [6‒9]. More recently, PVSs have also been suggested as MRI markers of CSVD [1, 4, 10]. The increased burden of PVSs is found to be associated with increased lacunes and white matter (WM) hyperintensities [1, 11]. However, most previous studies focused on detecting small PVSs at a maximum diameter less than 3 mm and evaluating the global burden of PVS enlargement.

Given the histopathological features of an L-PVS, a regular cavity filled with interstitial fluid containing a patent artery, it is significantly different from lacune, and these 2 kinds of intracerebral cavities are more likely to have diverse risk factors and clinical significance [12]. However, few clinical studies on PVSs >3 mm have been reported so far, leaving the necessity of differentiating these 2 cavities in clinical practice undetermined. In the present study, we aimed to examine large PVSs with diameters >3 mm on 3T MRI in a community-dwelling cohort study, in order to compare the risk factors and stroke risk of lacunes and L-PVSs in the general population.

The Shunyi study is an ongoing population-based cohort study conducted in Beijing, China. All inhabitants aged ≥35 years and independently living in 5 villages of Shunyi, a suburb district of Beijing, were invited to participate in this cohort study. From June 2013 to April 2016, a total of 1,586 participants responded to participate. All participants underwent standard baseline assessments, including structured questionnaires, physical examination, and laboratory tests. All subjects were invited to enter the brain MRI study, and 329 participates refused or had potential contradictions to MRI (cardiac peacemaker, coronary and peripheral artery stents, mental foreign body, or claustrophobia), resulting in a final total of 1,257 brain MRIs. Of these, 5 cases were excluded because of poor image quality, and 48 were excluded because of self-reported stroke history. Thus, the final sample for analysis was based on 1,204 subjects.

The final analysis included a lower proportion of male participants and current smokers than those excluded. Participants were followed every year, and 2 cases were lost to follow-up because they had moved out of the area by the 4-year visit, making the rate of loss to follow-up 0.42%. The flowchart of participation is presented in Figure 1.

Cardiovascular risk factors were defined as follows: hypertension was defined as blood pressure ≥140/90 mm Hg or self-reported history of hypertension or use of antihypertensive drug; diabetes mellitus was defined as a fasting plasma glucose level ≥7.0 mmol/L, self-reported history of diabetes mellitus, or taking antidiabetic treatment; hyperlipidemia was defined as total cholesterol >5.2 mmol/L, low-density lipoprotein >3.36 mmol/L, taking lipid-lowering drugs, or a reported history of hyperlipidemia; and smoking status was classified into current smoking (at least within the past 1 month) and noncurrent smoking. Participants were invited to a follow-up interview every year. Participants were interviewed annually and stroke event was screened using standardized questionnaire by asking symptoms, diagnosis, imaging findings, and treatment. When stroke events were suspected, we obtained detailed information on all subjects by checking the medical records in hospitals they attended. An experienced vascular neurologist verified the stroke diagnosis blinded to baseline MRI evaluation. Stroke was defined according to World Health Organization criteria and was further classified into ischemic stroke or spontaneous intracerebral hemorrhage based on neuroimaging results [13]. When it was difficult to make definite diagnosis, we resorted to an expert panel to form a consensus.

MRI was performed from July 2014 to April 2016 using one single 3-Tesla Siemens Skyra scanner (Siemens; Erlangen, Germany). Three-dimensional T1-weighted images were acquired using magnetization-prepared rapid gradient-echo sequence in sagittal planes (repetition time [TR] = 2,530 ms, echo time [TE] = 3.43 ms, inversion time = 1,100 ms, field of view [FOV] = 256 × 256 mm², voxel size = 1 × 1 × 1.3 mm³, flip angle = 8°, 144 sagittal slices). T2-weighted images (TR = 6,000 ms, TE = 125 ms, FOV = 230 × 230 mm², slice thickness = 5 mm, gap = 1 mm, flip angle = 90°, 20 axial slices), fluid-attenuated inversion recovery (FLAIR) images (TR = 8,500 ms, TE = 81 ms, FOV = 230 × 230 mm², slice thickness = 5 mm, gap = 1 mm, flip angle = 150°, 20 axial slices), and susceptibility-weighted images (TR = 20 ms, TE = 27 ms, FOV = 200 × 220 mm², slice thickness = 1.5 mm, flip angle = 15°, 80 axial slices) were acquired in axial planes.

Lacunes and L-PVSs were identified and differentiated according to their signal intensities on MRI sequences, size, location, shape, and border [3, 5, 14]. For lacunes, they were defined as focal fluid-filled cavities of 3–15 mm in diameter; with a hyperintensity ring seen on FLAIR; located in the basal ganglia (BG), subcortical WM, or brain stem; of irregular or wedge shape; and with irregular margin. For L-PVSs, they were defined as fluid-filled cavities of 3 mm or larger in diameter; without a hyperintensity ring seen on FLAIR; of well-defined oval, round, or tubular shape; and with smooth margin. Although in theory no location is exempt, L-PVSs are usually located in several conservative areas, including around the anterior commissure and inferior one-third of the BG (along the proximal part of lenticulostriate arteries), in WM along the path of perforating arteries, in midbrain, and hippocampus (Fig. 2). For cavities that were difficult to identify, multiplanar reformatting on three-dimensional (3D) T1-weighted images was further used. Those with a typical vascular shape and following the orientation of perforating vessels (including cystic lesions with an extension of vascular shape) were more likely to be PVS.

All other MRI markers of CSVD were defined according to the Standards for Reporting Vascular Changes on Neuroimaging [4]. The degree of severity of the PVS in BG and WM was rated using a previously established 4-level severity score on 3D T1-weighted images. The automatic segmentation of gray matter, WM, and cerebrospinal fluid volume on structure T1-weighted images was performed by Statistical Parametric Mapping 12 (http://www.fil.ion.ucl.ac.uk/spm/) and CAT12 toolbox (http://www.neuro.uni-jena.de/vbm/). The automatic segmentation of WM hyperintensity (WMH) was assessed using the lesion growth algorithm as implemented in the lesion segmentation tool toolbox (http://www.statistical-modelling.de/lst.html) for statistical parametric mapping at κ = 0.15. The analysis procedures of total intracranial volume, WMH volume, and brain parenchymal fraction, as a surrogate index of brain atrophy, were based on a previously validated method [15].

Arterial stenosis was assessed at the site of the most severe degree of stenosis using time-of-flight magnetic resonance angiography using established criteria [16]. Intracranial atherosclerotic stenosis (ICAS) was defined as any degree of stenosis in at least one of the following arteries: internal carotid artery, middle cerebral artery, anterior cerebral artery, intracranial segment of the vertebral artery, basilar artery, and posterior cerebral artery.

Well-trained neurologists who were blinded to all clinical data-rated lacunes, L-PVS, severity degree of PVS in BG and WM, cerebral microbleeds (CMBs), and ICAS independently. The intra- and inter-rater agreements were assessed in a random sample of 50 individuals with an interval of longer than 1 month between the first and second readings. The kappa coefficients for the intrarater agreements were 0.73 for lacunes, 0.83 for L-PVSs, 0.90 for CMBs, and 0.94 for ICAS. The kappa coefficients for the inter-rater agreements were 0.82 for lacunes, 0.62 for L-PVSs, 0.86 for CMBs, and 0.90 for ICAS.

The original proton density-weighted images were registered to the Montreal Neurological Institute standard space image using affine registration with FLIRT (FMRIB’s Linear Image Registration Tool). Volume of interest images of each patient were traced using MRIcron (http://www.mricro.com/mricron/). The tracings were normalized to the Montreal Neurological Institute standard space by using T1 images for best structural corresponding. Finally, individual lacune/L-PVS maps were added up across all subjects to create a group-level lacune/L-PVS frequency map, which is presented in Figure 3.

Continuous variables are expressed as mean and standardized deviation or median and range as appropriate, and categorical variables are expressed as frequencies and proportions. Because of the similarity between L-PVSs and lacunes, we also classified the subjects into groups based on the presence of L-PVSs or lacunes. We compared the clinical and neuroimaging characteristics between groups using the t test (for means), Wilcoxon rank-sum test (for medians), and χ² test (for percentage). Furthermore, we applied logistic multivariable regression models to investigate potential independent risk factors of lacunes and L-PVSs. The candidate risk factors included age, sex, and conventional vascular risk factors (current smoking, BMI, hypertension, systolic blood pressure, and diabetes), and ApoE ε4 carrier.

The associations between MRI changes (independent variables, including ICAS, CMB, WMH, and BPF) and lacunes or L-PVSs (dependent variables) were also investigated using binary logistic regression. The WMH volume was natural log transformed to normalize the skewness.

The predicting roles of lacunes and L-PVSs on incident stroke were investigated, and hazard ratio (HR) and its 95% confidence interval (CI) were calculated using COX proportional hazard models. The participants were divided into 4 groups, one is none of lacunes or L-PVSs (reference), one is with lacunes only, one is with L-PVSs only, and one is with both lacunes and L-PVSs. All analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC, USA), and 2-sided p values of <0.05 were taken to indicate statistical significance.

Baseline characteristics and MRI measurements of the study sample are displayed in Table 1. Of the 1,204 participants, 446 (37.0%) were male, and the mean age was 55.7 years (standard deviation = 9.3). A total of 347 L-PVSs were detected in 235 subjects, 31.2% (73/235) of which had >1 L-PVS. A total of 219 lacunes were detected in 183 subjects, 21.9% (40/183) of which had >1 lacune. Sixty-two (5.1%) subjects had both lesions simultaneously. In the participants with L-PVS, most (63.8%) had L-PVS in the BG, 24.7% in the subcortical WM, and 15.0% in the hippocampus. Among the subjects with lacune, 47.5% had lacune in BG, 41% in the WM, and 26.8% in other brain areas. The detail distribution is presented in Figure 3, showing that the most prevalent location for L-PVS is near anterior perforated substance, where the lenticulostriate arteries enter the basal ganglia, while the most prevalent location for lacunes is in the corpus striatum and thalamus.

The baseline characteristics of participants without L-PVS or lacune are listed in Table 1. After adjusting for age and gender, the presence of L-PVS was significantly increased with age (odds ratio [OR] per standard deviation, 1.40; 95% CI, 1.20–1.63) and was more prevalent in ApoE allele ε4 carriers (OR, 1.50; 95% CI, 1.02–2.22), but it was not associated with other cardiovascular risk factors (BMI, smoking, hypertension, diabetes, and hypercholesterolemia). However, the presence of lacune was significantly associated with age, male gender, hypertension, and diabetes, which were typical cardiovascular risk factors, but not ApoEε4 carrier genotype (Table 2).

As is shown in Table 3, there were significant associations between lacunes and other imaging markers of CSVD, including the mean volume of WMH, CMB, and global atrophy, as well as ICAS. All associations persisted when adjusted for cardiovascular risk factors (including age, gender, and hypertension). The presence of L-PVSs was associated with a higher frequency of lacunes (OR, 1.76; 95% CI, 1.20–2.59; p < 0.01) but not with other imaging markers of cerebral small vascular disease, including WMH, CMB, or global atrophy.

During 3,933 person-years of follow-up (median of 3.2 years), 22 cases of incident stroke were diagnosed (crude incident rate: 5.59/1,000 person-years). The presence of lacune, without L-PVS, on MRI at baseline was strongly associated with an increased risk of incident stroke after adjustment for age, gender, and other vascular risk factors (HR, 4.68; 95% CI, 1.15–19.07). However, there was no significant relationship detected between the presence of L-PVS and incident stroke (HR, 3.84; 95% CI, 0.82–18.04) (Table 4). That is, compared to those without L-PVS or lacune, subjects with lacune suffered at least 4.68 times the stroke risk, whereas subjects with L-PVS alone have no more risk of stroke at the 3-year follow-up. Patients with a presence of both lacune and L-PVS had an even higher risk of stroke, but the CI was wide (HR, 10.25; 95% CI, 2.31–45.44).

We found significant differences between lacune and L-PVS in terms of the location, spectrum of risk factors, association with other MRI markers of CSVD, and 3-year stroke risk. The presence of lacunes was significantly related to age, male gender, hypertension, diabetes, and ICAS, whereas the presence of L-PVSs was associated only with age and ApoEε4 carrier genotype. Other MRI markers (e.g., WMH, microbleeds) were more prevalent in those who had lacunes but not in those who had L-PVS. Finally, those who had lacunes seen on MRI at baseline had much higher stroke risk at the 3-year follow-up but those who had L-PVS did not.

Our results regarding the risk factors for lacune, including age, male gender, hypertension, diabetes, and ICAS, are in line with previous studies. We found a significant discrepancy in the risk factor spectrum between lacune and L-PVS, suggesting different pathogenetic mechanisms related to these 2 kinds of cavities. Gutierrez et al. [10] found in the Northern Manhattan Study that L-PVS was associated with only age and hypertension [11, 14], while Ding et al. [5] found in the Reykjavik Study that L-PVS was related to age and gender. Both studies indicated the reduced relevance of vascular risk factors and large PVSs [5]. In addition, unlike the small PVSs commonly located in the BG and WM, most of the L-PVSs are conservatively distributed in the anterior commissure and in subcortical WM surrounding the posterior horn of the lateral ventricle [17‒19]. This regional predisposition probably suggests that the underlying mechanisms of L-PVS might be also different from that of small PVSs. Although not fully understood, different mechanisms related to large cystiform enlargement of the PVS were suggested. First, the L-PVS located adjacent to the anterior commissure might relate to regional anatomic characteristics, including the angular trajectory of the vascular tree and the presence of multiple branches crossing the surface of the cerebrum in this area [20]. Second, the L-PVSs in subcortical WM, which often appear to be cystiform cavities, might be associated with the disturbance of the drainage of interstitial fluid [3] when amyloid accumulates along the cortical perforating arteries [18, 21]. This hypothesis was supported by the marginally significant association between ApoE genotype and the presence of L-PVS detected in the present study. Further studies are needed to confirm this finding in other samples, including in patients with cerebral amyloid angiopathy or Alzheimer disease.

In the present study, we found a significant association between lacunes and the other MRI markers of CSVD; in contrast, L-PVS was not associated with WMH, CMBs, or atrophy. Consistent with previous reports, in our cohort, we actually found prominent associations between the severity of PVS, which was mainly taking small PVSs into account, and other markers of CSVD (data unpublished). According to this discrepancy, L-PVS may not be a reliable MRI marker indicating CSVD. This result, along with the risk factors and distribution of L-PVS found in the present study, implied that L-PVS was more likely related to certain anatomical characteristics, especially those located around the anterior commissure and inferior one-third of the BG. Similar results were also reported by the Northern Manhattan Study [10, 14].

The association between lacune and increased risk of stroke was consistently detected in previous research [7, 22] and was confirmed again in our present study. However, no correlation was observed between L-PVS and stroke risk in this sample. The discrepancy in the predictive value of stroke further hinted at the difference in potential mechanisms related to lacune and L-PVS. Most lacunes are probably due to occlusion of arteries, which has a similar pathophysiological mechanism to stroke, whereas L-PVSs are more likely related to the accumulation of interstitial fluid around small vessels, which is distinct from stroke. Our findings highlighted the importance of distinguishing between lacune and L-PVS in the clinical setting to avoid overestimation of stroke risk and unnecessary antithrombosis treatment as consequence.

We observed a strong correlation between L-PVSs and lacunes, which is consistent with previous studies [1]. This may be mainly due to the inevitable misreading for each other during imaging analysis, which is a potential limitation of our study. The diagnostic criteria of PVS often differ among studies, either using FLAIR sequence (in the Northern Manhattan Study [10, 14], AGES-Reykjavík [5]), according to morphological features (mainly according to shape, size, and margin, as in the Rotterdam Scan Study [23]) or using 3D-structure analysis of the cavity (in the 3C study [3]). We actually combined all of the above-mentioned methods to improve the reliability of distinguishing between lacunes and L-PVS. However, without pathological investigation, identifying these 2 kinds of cavities is still an empirical and less precise task. However, we believe that, to a certain extent, we defined 2 groups of cavities with significant differences in risk factors and clinical significance using these criteria, which is feasible in clinical practice. This study was conducted in a Chinese rural population, which had much higher risk of stroke because of the high prevalence of vascular risk factors and low control rate, and the racial difference may also be a cause. Thus, our results should be extrapolated to other populations with caution.

In this population-based cohort, we found that L-PVSs are significantly different from lacunes, in terms of risk factors, association with MRI markers of CSVD, and the risk of incident stroke, which serves to distinguish these 2 kinds of cavities in clinical practice. Further studies on the distribution of L-PVS and its association with β-amyloid are needed to further understand the underlying mechanisms of L-PVS.

The authors would like to thank all the contributors of the Shunyi study group.

The study was approved by the Medical Review Ethics Committee of Peking Union Medical College Hospital (reference number: B-160), and all participants provided written informed consent.

The authors have no conflicts of interest to disclose regarding this manuscript.

This study was supported by the National Key Research and Development Program of China (No. 2016YFC0901004), National Natural Science Foundation of China (No. 81671173), CAMS Innovation Fund for Medical Sciences (CIFMS #2017-I2M-3-008), the Strategic Priority Research Program (pilot study) “Biological basis of aging and therapeutic strategies” of the Chinese Academy of Sciences (grant XDPB10), the 2016 PUMCH Science Fund for Junior Faculty (pumch-2016-1.3), the Fundamental Research Funds for the Central Universities (3332018034), China Guanghua Technology Foundation (CDCH #201801), and Chengde Science and Technology Research and Development Plan Project (201601A019 & 201804A011).

Zhu Y.C. and Zhang J.T.: study concept and design. Zhang J.T., Han F., Zhang D.D., and Zhai F.F.: acquisition, analysis, and interpretation of data. Liang X.Y., Zhang D.D., and Zhang J.T.: statistical analysis, tables, and figures. Zhang J.T., Zhu Y.C., Zhou L.X., Ni J., Yao M., Zhang S.Y., Cui L.Y., and Jin Z.Y.: drafting and critical revision of the manuscript for important intellectual content and final approval of the version to be published.