Admission Brain Cortical Volume
An Independent Determinant of Poststroke Cognitive Vulnerability
Background and Purpose—Several markers of poststroke cognitive impairment have been reported. The role of brain cortical volume remains uncertain. The aim of this study was to evaluate the influence of brain cortical volume on cognitive outcomes using a voxel-based morphometry approach in subjects without prestroke dementia.
Methods—Ischemic stroke patients were prospectively recruited 24 to 72 hours post stroke (M0). Cognition was evaluated at M0, 3 months, and 1 year (M12) using the Montreal Cognitive Assessment, the Isaacs set test, and the Zazzo’s cancellation task. A 3-T brain magnetic resonance imaging was performed at M0. Grey matter (GM) was segmented using Statistical Parametric Mapping 12 software. Association between global GM volume and cognitive score slopes between M0 and M12 was evaluated using a linear mixed model. Correlations between focal GM volumes and changes in cognitive performance were evaluated using Statistical Parametric Mapping 12.
Results—Two-hundred forty-eight patients were included (mean age 65±SD 14 years old, 66% men). Global GM volume was significantly associated with changes in Montreal Cognitive Assessment scores (β=0.01; P=0.04) and in the number of errors on the Zazzo’s cancellation task (β=−0.02; P=0.04) independently of other clinical/radiological confounders. Subjects with lower GM volumes in the left fronto-temporo-insular cortex were more vulnerable to transient Montreal Cognitive Assessment and Isaacs set test impairment. Subjects with lower GM volumes in right temporo-insular cortex, together with basal ganglia, were more vulnerable to transient cognitive impairment on the Zazzo’s cancellation task.
Conclusions—Smaller cortical volumes in fronto-temporo-insular areas measured 24 to 72 hours post stroke are associated with cognitive vulnerability in the subacute stroke phase.
Stroke is a major source of cognitive impairment with a prevalence of poststroke dementia estimated around 12% at 1 year.1 In some patients, cognitive disturbances can be observed as soon as the acute phase of stroke and could result from prestroke cognitive impairment, transient cognitive impairment2,3 or be the first step toward a more permanent cognitive decline. The risk to develop poststroke cognitive impairment has been associated with brain structural changes located outside the ischemic lesion, such as the extent of white matter hyperintensities (WMHs)4,5 and brain volume, often considered as a marker of brain atrophy. Cortical volume is of special interest because it is a well-known marker of underlying neurodegenerative processes that could increase the susceptibility to develop poststroke cognitive impairment.6 Previous studies have reported an association between global7,8 and focal cerebral atrophy, mainly in the medial temporal lobe9 and poststroke cognitive impairment, but most of them has several methodological issues, including the type of scanner (tomodensitometry or magnetic resonance imaging) used and the tools used to assess the atrophy (ie, visual inspection).10,11
The aim of this study was to investigate the association between cortical volume measured on a brain magnetic resonance imaging performed early after stroke and cognitive vulnerability in a population of patients without prestroke dementia.
Materials and Methods
Inclusion and Exclusion Criteria
Patients were prospectively recruited at the Bordeaux University Hospital. This study was part of the BBS study (Brain Before Stroke), a biomedical research protocol. The protocol was accepted by the regional ethical board (CPP 2012/19 2012-A00190-43), and all patients or their legal representative provided a written informed consent to participate. Inclusion criteria were an acute supratentorial ischemic stroke in men or women aged ≥18 years old, with a National Institute of Health Stroke Score ranged from 1 to 25 between 24 and 72 hours after stroke onset. The main exclusion criteria were prestroke disability related to history of neurological disorder with a modified Rankin Scale (mRS) score of ≥1 and inability to perform cognitive tests or to complete the magnetic resonance imaging protocol (all exclusion criteria are in the online-only Data Supplement).
Demographic data, including educational level and cardiovascular risk factors, were recorded. A clinical assessment was performed at baseline in the first 24 to 72 hours (M0), at 3 months (M3), and at 1 year (M12) post stroke. Stroke severity was evaluated using the National Institute of Health Stroke Score. A prestroke cognitive impairment was evaluated using the Informant Questionnaire in Cognitive Decline in the Elderly.12 Cognitive assessment included the Montreal Cognitive Assessment (MoCA) for the evaluation of global cognition,13 with a different version for each time point to limit learning effect; the Isaacs set test (IST) for the evaluation of executive functions, consisting in naming the maximum of words in 4 different semantic categories in the first 15 s of the task for each category14; and the Zazzo’s cancellation task,15 consisting in cancelling target signs randomly distributed among distractors on 8 lines. Time to perform the test and the number of errors were recorded for the evaluation of processing speed and attention. Functional outcome was assessed at M3 and M12 using the mRS. Mood changes were assessed at M0, M3, and M12 using the Hospital Anxiety and Depression scale.
Imaging protocol is described in the online-only Data Supplement.
First, a mask of strokes was created for each patient by segmenting the lesions on diffusion weighted imaging in a semiautomatic way using 3-dimensional Slicer 4.3.1 software (http://www.slicer.org).16
Then, a correction of magnetic field inhomogeneities was applied on the T1-weighted imaging (T1-wi), fluid attenuated inversion recovery (FLAIR), and diffusion weighted imaging using Advanced Normalization Tools software,17 and FLAIR and diffusion weighted imaging were coregistered on T1-wi. The transformation matrixes obtained for the diffusion weighted imaging were applied to coregister the stroke masks on the T1-wi.
WMHs were segmented using the Lesion Segmentation Tool, version 2.0.14 for Statistical Parametric Mapping (SPM) 1218 from T1-wi and FLAIR sequences, which provided the extraction of WMH volume.
Subsequently, T1-wi and FLAIR sequences were used to segment the different tissue classes with a voxel-based morphometry approach using SPM 12 software, available on MATLAB (R2012b).19 Grey matter (GM), white matter, and cerebral spinal fluid were segmented based on a voxel-by-voxel intensity analysis, resulting in a mask of these 3 tissue classes. Strokes and WMH were segmented in an additional fourth tissue class using a new tissue probability map created from the coregistration of WMH and stroke masks. The volumes of these 4 tissue classes were summed to get the total intracranial volume.
The normalized modulated GM partitions were smoothed with a 8 mm full width half height Gaussian kernel. After statistical analyses, the regions of interest were labeled using the SPM toolbox Automated Anatomical Labeling.20
Cognitive Changes Over Time
To explore the longitudinal evolution of cognitive scores over the year of follow-up, we first calculated a slope of evolution for each cognitive score using a generalized linear mixed model with random slopes and the restricted maximum likelihood method.21 The models were constructed using the lmerTest package available on the R software. They were validated by visual inspection of residuals and random slopes distributions on histograms which approximated a normal distribution. Random slopes were extracted using the ranef function. To analyze the evolving cognitive profile of the patients, we distinguished 2 groups according to the positive or negative direction of their slopes for each cognitive score. Demographic, clinical, and radiological data were compared between groups using a χ2 test for categorical variables, or an unpaired 2-sample Wilcoxon test or an independent Student t test after verification of normality with a Shapiro–Wilk test. Changes in cognitive and mood scores were compared using a Friedman rank-sum test for repeated measures.
GM Volume on Cognitive Outcome
To evaluate the impact of global GM volume on changes in cognitive performances between M0 and M12 in the whole population, we performed bivariate analyses using a generalized linear mixed model with each cognitive score as dependent variable (MoCA, IST, Zazzo’s cancellation task: time to perform the task and number of errors), and radiological parameters (GM volume, WMH volume and stroke volume corrected for total intracranial volume, and cortical location), or demographic and clinical confounders (age, sex, educational level, cardiovascular risk factors, mRS at M12, and changes in Hospital Anxiety and Depression scale scores) as independent variables. In addition, we performed bivariate cross-sectional analyses at M0 and M12 using linear regressions with the same variables. All these intermediate analyses are presented in Tables I and II in the online-only Data Supplement. We then performed multivariate analyses using mixed models, including all variables being significant (P<0.05) in the simple model with MoCA as the dependant variable. To evaluate the respective impact of global GM, WMH, and stroke volumes on changes in cognitive performances, we built a radiological model, including only those radiological variables as independent predictors. We then built a clinical/radiological model, including the same covariates as the radiological model, adjusting for demographic and clinical confounders. The models were validated by verifying the residuals and random slopes normality on histograms. Statistical analyses were performed on the R software 3.2.4. Statistical significance was set at 0.05 for all tests.
To investigate the association between focal GM volumes and the evolution of cognitive scores over the year of follow-up in the whole population, we performed multiple regression analyses using SPM 12, with the slopes of cognitive scores as covariates, corrected for age, sex, educational level, mRS at M12, and total intracranial volume. Stroke volume and location, together with WMH volume, were segmented in a separate tissue class and were not added as covariate. P<0.05 after multiple comparisons (family-wise error rate) for 100 contiguous voxels was considered statistically significant.
Over Time Changes of Cognitive Performances
Two-hundred forty-eight patients were included in the analysis (mean age 65±SD 14 years old, 66% men; Figure 1). Demographic, clinical, and radiological data are presented in Table 1 and Table III in the online-only Data Supplement. MoCA was achieved at the 3 time points in 199 patients, IST in 202 patients, and Zazzo’s cancellation task in 171 patients.
Patients with positive slopes for MoCA (104 patients, 52%) and IST (107 patients, 53%) and negative slopes for Zazzo’s cancellation task (71 patients, 42%) improved significantly their performances between the 3 time points (Figure 2; Table IV in the online-only Data Supplement). They constituted the Improvement groups. Patients with negative slopes for MoCA and IST and positive slopes for Zazzo’s cancellation task had nonsignificant change in their performances between the 3 time points. Therefore, they constituted the Stable groups.
Relationship Between Global GM Volumes and Cognitive Performances
In the radiological model (Table 3), total GM volume was the only variable predictive of changes in all cognitive scores over the year of follow-up (P<0.001 for MoCA, P=0.03 for IST, P=0.002 for time to perform Zazzo’s cancellation task and P<0.001 for the number of errors).
In the clinical/radiological model (Table 3), total GM volume was independently associated with changes in MoCA (P=0.04) and number of errors during Zazzo’s cancellation task (P=0.04) over the year of follow-up.
Relationship Between Focal GM Volumes and Cognitive Performances
The SPM 12 multiple regression analyses evaluating the relationship between GM volumes and changes in cognition assessed by the slopes of evolution showed negative correlations with the slopes of MoCA and IST (Figure 3), meaning that patients with cognitive improvement had lower focal GM volumes, regardless of age, sex, educational level, mRS at M12, and total intracranial volume. The main significant volumes of interest were located in left fronto-temporo-insular cortex (Table V in the online-only Data Supplement). No positive correlation was observed for MoCA and IST.
Positive correlations were observed between GM volumes and the slopes of Zazzo’s cancellation task (Figure 3), meaning that patients with cognitive improvement had lower focal GM volumes, regardless of the confounding covariates. The main clusters were right temporo-insular cortex, together with right caudate nucleus (Table V in the online-only Data Supplement). There was no negative correlation between GM volumes and the slopes of Zazzo’s cancellation task.
The main results of this study are that (1) in a selected stroke population with low-to-moderate stroke severity and without preexisting cognitive disability, transient cognitive impairment is observed in almost half of the patients; (2) patients with lower GM volumes before stroke have a greater cognitive vulnerability at the acute phase of stroke; and (3) specific areas, including left fronto-temporo-insular regions, right temporo-insular cortex, and basal ganglia, are predominantly involved in this poststroke cognitive vulnerability.
Despite a selection of patients without prestroke dementia, we observed worse cognitive performances in 52% of the patients tested with the MoCA at the acute phase of stroke, before improving at 3 months and 1 year (group improvement), suggesting a progressive cognitive recovery during the year after the cerebrovascular insult. This result is in accordance with the reversibility of cognitive dysfunction previously reported by Snaphaan and de Leeuw3 who identified that 50% of patients have memory difficulties in the weeks after a stroke, whereas only 11% had persistent impairment at 1 year. This transient alteration in cognition could share some pathophysiological similarities with acute poststroke delirium.22 Indeed, poststroke delirium is thought to result from patient’s cognitive frailty related to prestroke cognitive impairment which is exacerbated by the physical disability and biochemical changes related to the acute brain insult. Del Brutto et al23 recently suggested that brain cortical volume could be an magnetic resonance imaging marker of this cognitive frailty. In line with this hypothesis, Oldenbeuving et al24 reported that global cortical atrophy was an independent predictor of acute delirium. Similarly, in this study, we found that patients with worse cognitive performances at M0 and progressive recovery had the lowest GM volumes.
Importantly, the influence of GM volume was independent of other imaging markers such as the extent of white matter lesions and stroke volume, 2 parameters usually associated with poststroke cognitive outcome.25,26 Moreover, GM volume remained significantly associated with changes in some poststroke cognitive measures even after adjustment on usual major clinical determinants of poststroke cognitive outcome such as age, educational level, and general recovery measured by mRS at 1 year. This result strongly suggests that low global GM volumes represent a biomarker of cognitive vulnerability in patients with acute ischemic stroke. However, the lower strength of the association between global GM volume and changes in cognitive measures after addition of the clinical parameters suggests that either the role of GM volume is limited or that only focal GM structures contribute to this association. Supporting this last hypothesis, SPM 12 analysis demonstrated that some focal areas, such as the left fronto-temporo-insular areas, right temporo-insular cortex, and right caudate nucleus, were associated with cognitive changes, even after controlling for demographic factors and mRS at 1 year. These structures are well known to be involved in various cognitive neuronal networks, such as the fronto-temporal network,27–29 the salience network,30 the limbic network,31 the central executive network,32 and the visual ventral stream.33
The underlying mechanism of these focal smaller cortical volumes is still a matter of debate, but several hypotheses can be evoked. First, the observation of better cognitive performances and fairly stable evolution among patients who had the highest educational level could support the hypotheses of brain and cognitive reserve, suggesting that people with bigger brain and sustained cerebral activity better cope against brain lesions.32,34,35 Second, smaller cortical volumes might be a biomarker of preclinical neurodegenerative processes that might have contributed to an increased cognitive vulnerability after the acute cerebral insult.6,10 Moreover, the role of baseline GM volume in the prediction of MoCA and Zazzo’s cancellation task scores, independently of the extent of WMH, does not support the hypothesis of the sole consequence of the severity of white matter vascular injuries or degeneration.
While performed on a large sample of patients and using a quantitative and validated method for cortical volume assessment, the results of this study should be interpreted in light of several limitations. First, the neuropsychological evaluation included only 3 tests, which does not allow for extrapolation to all cognitive domains. Second, patients included in the study had mild-to-moderate disability to make possible the assessment of clinical and radiological tests, thereby explaining the large number of patients excluded from the analysis and preventing the generalization of the results to all stroke patients. Third, the definition of groups based on the slope of cognitive tests is a potential methodological limitation because some patients with negative slopes on the MoCA could have experienced significant cognitive decline and correspond to patients with lower cortical volumes at M0. However, no larger cortical volumes were observed among patients with positive slopes which make unlikely the hypothesis that a large number of patients with severe cortical atrophy were part of the group with negative slopes.
This study supports the role of GM volume in specific focal areas, such as the fronto-temporo-insular regions, as radiological biomarkers of poststroke cognitive vulnerability. A longer follow-up would be necessary to investigate the role of this brain vulnerability in long-term poststroke cognitive outcomes.
We thank Sylvain Ledure and Nathalie Heyvang for their contribution in data collection, and Dr Joël Swendsen for his revision.
Sources of Funding
This study was supported by the French government and the French Agence Nationale de la Recherche.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.017646/-/DC1.
- Received January 23, 2017.
- Revision received May 11, 2017.
- Accepted May 23, 2017.
- © 2017 American Heart Association, Inc.
- Nijsse B,
- Visser-Meily JM,
- van Mierlo ML,
- Post MW,
- de Kort PL,
- van Heugten CM
- Snaphaan L,
- de Leeuw FE
- Dufouil C,
- Godin O,
- Chalmers J,
- Coskun O,
- MacMahon S,
- Tzourio-Mazoyer N,
- et al
- Debette S,
- Beiser A,
- DeCarli C,
- Au R,
- Himali JJ,
- Kelly-Hayes M,
- et al
- Tuladhar AM,
- Reid AT,
- Shumskaya E,
- de Laat KF,
- van Norden AG,
- van Dijk EJ,
- et al
- Tang WK,
- Chan SS,
- Chiu HF,
- Ungvari GS,
- Wong KS,
- Kwok TC,
- et al
- Isaacs B,
- Kennie AT
- Zazzo R
- Del Brutto OH,
- Mera RM,
- Cagino K,
- Fanning KD,
- Milla-Martinez MF,
- Nieves JL,
- et al
- Chang LJ,
- Yarkoni T,
- Khaw MW,
- Sanfey AG
- Ruscheweyh R,
- Deppe M,
- Lohmann H,
- Wersching H,
- Korsukewitz C,
- Duning T,
- et al
- McIntosh AR,
- Grady CL,
- Ungerleider LG,
- Haxby JV,
- Rapoport SI,
- Horwitz B
- Amieva H,
- Mokri H,
- Le Goff M,
- Meillon C,
- Jacqmin-Gadda H,
- Foubert-Samier A,
- et al