Remote Lower White Matter Integrity Increases the Risk of Long-Term Cognitive Impairment After Ischemic Stroke in Young Adults
Background and Purpose—Poststroke cognitive impairment occurs frequently in young patients with ischemic stroke (18 through 50 years of age). Accumulating data suggest that stroke is associated with lower white matter integrity remote from the stroke impact area, which might explain why some patients have good long-term cognitive outcome and others do not. Given the life expectancy of decades in young patients, we therefore investigated remote white matter in relation to long-term cognitive function.
Methods—We included all consecutive first-ever ischemic stroke patients, left/right hemisphere, without recurrent stroke or transient ischemic attack during follow-up, aged 18 through 50 years, admitted to our university medical center between 1980 and 2010. One hundred seventeen patients underwent magnetic resonance imaging scanning including a T1-weighted scan, a diffusion tensor imaging scan, and completed a neuropsychological assessment. Patients were compared with a matched stroke-free control group (age, sex, and education matched). Cognitive impairment was defined as >1.5 SD below the mean cognitive index score of controls and no cognitive impairment as ≤1 SD. Tract-Based Spatial Statistics was used to assess the white matter integrity (fractional anisotropy and mean diffusivity).
Results—About 11 years after ischemic stroke, lower remote white matter integrity was associated with a worse long-term cognitive performance. A lower remote white matter integrity, even in the contralesional hemisphere, was observed in cognitively impaired patients (n=25) compared with cognitively unimpaired patients (n=71).
Conclusions—These findings indicate that although stroke has an acute onset, it might have long lasting effects on remote white matter integrity and thereby increases the risk of long-term cognitive impairment.
Cognitive impairments are common after stroke in young adults (18 through 50 years of age).1–3 Stroke severity at admission,4 stroke volume, and stroke location5 are related to poststroke cognitive impairment. These factors, however, do not fully explain why some patients show good cognitive outcome after stroke and others do not.6
Accumulating data suggest that, apart from temporary remote neurophysiological changes after stroke (ie, diaschisis),7 a focal (ischemic) lesion could potentially affect long-term remote structural integrity (eg, of the hippocampus and thalamus) in the ipsilesional hemisphere after stroke,3,8–11 which has been associated with a worse poststroke memory performance.11,12 Besides remote structural changes in the ipsilesional hemisphere, a few small neuroimaging studies (6–16 patients) in older patients with stroke (≥60 years) found loss of white matter integrity as remote as the contralesional hemisphere,13 and this was associated with motor impairment ≤6 months after stroke14,15 and poorer cognitive recovery 3 months after right middle cerebral artery stroke.16 It is unknown whether lower microstructural integrity in areas remote from the initial stroke, including as remote as the contralesional hemisphere, is related to long-term (ie, years) cognitive dysfunction after stroke. In addition, the presence of preexisting (subclinical) cerebral small vessel disease, such as diffuse white-matter hyperintensities or subclinical infarcts, may also contribute to the poststroke spectrum of cognitive impairment6; however, this has never been investigated in young patients with stroke. A better understanding of the cause of poststroke cognitive impairment and its possible recovery is especially important in young stroke survivors, as they are in a demanding phase of their lives with respect to educational, vocational, and family-related functioning.
We hypothesized that patients with cognitive impairment years after their index stroke show a lower white matter structural integrity remote from the initial stroke and more coexisting cerebrovascular disease compared with cognitively unimpaired patients. Conversely, patients with a higher structural integrity in remote brain structures were expected to have an average poststroke cognitive performance compared with stroke-free controls.
Patients and Methods
This study is part of the FUTURE study, a prospective cohort study of prognosis after transient ischemic attack (TIA), ischemic stroke, or hemorrhagic stroke in adults aged 18 through 50 years admitted to the Radboud university medical center, The Netherlands, between January 1, 1980 and November 1, 2010.3,17 The Medical Review Ethics Committee region Arnhem-Nijmegen approved the study, and written informed consent was obtained from all participants.
Patients were identified through a prospective registry of all consecutive young stroke/patients with TIA that has been kept at the department since the 1970s with a standardized collection of baseline and clinical characteristics. Ischemic stroke was defined as focal neurological deficit persisting >24 hours. Patients had the opportunity to undergo an extensive neuropsychological examination together with subsequent magnetic resonance imaging scanning during the follow-up between November 2009 and December 2011.3
This study comprises all consecutive patients with a first-ever ischemic stroke in one of the hemispheres. Exclusion criteria in this study were a previous stroke or TIA, cerebral venous sinus thrombosis, retinal infarction,17 recurrent stroke/TIA during the follow-up period, and severe aphasia. Lesion location was based on neuroimaging findings at follow-up, which was related to the stroke location at the time of the event described in medical records and radiological findings.
Controls were recruited among the patients’ spouses, relatives, or social environment. Inclusion criteria were ≥18, without a history of stroke or TIA. The control group and patient group were matched for age (P=0.5), sex (P=0.6), and education (P=0.1). Controls were used to define cognitive impairment and no impairment in patients with stroke. Controls were all living independently, none fulfilling the clinical criteria of dementia.
The neuropsychological tests used in the FUTURE study covered the main cognitive domains, and detailed information on these cognitive tests has been described extensively elsewhere.3,17 A z score per test was calculated for each participant based on the mean and the SD of the controls (n=84). Next, averaging z scores of cognitive tests that mainly reflected the same cognitive domain resulted in a composite z score per cognitive domain. The 7 cognitive domains were processing speed (Symbol-Digit Modalities Test, Abbreviated Stroop Color Word Test parts I and II), visuoconstruction (Rey-Osterrieth Complex Figure [ROCF]-copy trial), working memory (Paper and Pencil Memory Scanning Test [PPMST]), immediate memory (ROCF-immediate recall and the total number of words immediately recalled in the 3-trial version of the Rey Auditory Verbal Learning Test [RAVLT]), delayed memory (delayed recall of the ROCF and the RAVLT), attention (Verbal Series Attention Test [VSAT]), and executive functioning (Verbal fluency and Stroop interference).
The cognitive index score is a composite score defined as the average of the 7 cognitive domains. Cognitive impairment was defined as a cognitive index score >1.5 SD below the mean of controls, which has often been used as a cutoff for vascular cognitive impairment in patients with stroke.18 No cognitive impairment was defined as a cognitive index score ≤1 SD below the mean of controls.3,19
Neuroimaging Data Acquisition
Participants underwent a 1.5-T magnetic resonance imaging scanning on the Siemens, Magnetom Avanto. A T1-weighted whole-brain MPRAGE scan, as well as a set of whole-brain diffusion weighted images, a FLAIR pulse sequence, and a gradient echo susceptibility-weighted imaging were collected. Detailed information on the neuroimaging data acquisition can be found in the online-only Data Supplement.
Neuroimaging Data Processing
Ischemic Stroke Volume and Lesion Probability Map
Ischemic stroke was defined as a hypointense area on a T1-weighted MPRAGE whole-brain scan with corresponding gliotic rim on the FLAIR image. One experienced investigator, blinded for baseline characteristics and outcome measures, traced all ischemic strokes manually using a T1-weighted image (in the coronal plane). Stroke volumes (mL) were normalized using the following formula: average intracranial volume of the total population×(stroke volume of the participant/intracranial volume of the participant).20 Intracranial volume was calculated as the sum of gray matter, white matter, and cerebrospinal fluid using VBM8.12 Next, brain-extracted images21 were registered, along with the lesion mask, to the Montreal Neurological Institute standard space by an affine transformation (12 degrees of freedom) using FSL-FLIRT (FMRIB Software Library - FMRIB's Linear Image Registration Tool),21 followed by nonlinear registration using FNIRT (FMRIB’s Nonlinear Image Registration Tool).21 Next, all stroke masks were merged and averaged, which resulted in a lesion probability map.
Diffusion Tensor Imaging Preprocessing
The raw diffusion tensor imaging (DTI) data were denoised using a local PCA filter, which reduces random noise by locally shrinking less significant principal components using an overcomplete approach.22 Next, misalignments from eddy currents and subject motion were corrected by a mutual information-based coregistration technique (SPM [Statistical Parametric Mapping]; affine transformation). An average b0-image mask was constructed and was used to mask the results. Magnetic susceptibility–induced EPI distortions in the diffusion tensor images were unwarped along the phase-encode direction by mapping the mean unweighted image onto the T1 reference image.23 The diffusion tensors and their indices (fractional anisotropy [FA] and mean diffusivity [MD]) were robustly estimated using the in-house developed iteratively reweighted least squares algorithm named “PATCH.”24
Tract-Based Spatial Statistics
In short, tract-based spatial statistics (TBSS) projects all FA and MD images onto a mean FA tract skeleton, before applying voxel-wise cross-subject statistics.25 The TBSS nonlinear registration of a large (cortical) stroke failed in some cases and, therefore, a FNIRT configuration file optimized for FA data was used21 where the stroke lesions were masked out. The FA skeleton was thresholded at 0.3 to include the major white matter tracts and exclude brain parts with low intersubject reliability. For all TBSS analyses, the FA and MD white matter skeleton were first symmetrized for all subjects25 and flipped (right to left) for patients with right hemispheric stroke.
Vascular Risk Factors and Treatment at Follow-Up Assessment
Blood pressure readings in the left and right arm were performed 3× in supine position. Next, the mean blood pressure was calculated for the left and right arm separately. The highest mean blood pressure was used to identify hypertension. Hypertension was defined as a systolic blood pressure of ≥135 mm Hg, a diastolic blood pressure of ≥85 mm Hg, or the use of antihypertensive medication. Dyslipidemia was defined as total cholesterol ≥5.0 mmol/L, low-density lipoprotein ≥2.5 mmol/L, triglycerides ≥2.0 mmol/L,26 or the use of statins. Diabetes mellitus was defined as random blood glucose level of >11.1 mmol/L, 2 consecutive fasting venous plasma glucose levels of ≥7.0 mmol/L,27 the use of antidiabetics (oral or insulin). Dyslipidemia was defined as total cholesterol ≥5.0 mmol/L, low-density lipoprotein ≥2.5 mmol/L, triglycerides ≥2.0 mmol/L,26 or the use of statins. Information on smoking was collected by a structured questionnaire. Current smoking was defined as smoking ≥1 cigarette per day in the year before follow-up. The body mass index was calculated as the weight in kilograms divided by the square of the height in meters.
Cerebral Small Vessel Disease
White matter hyperintensities (WMHs) of presumed vascular origin were defined as hyperintense signal on FLAIR, without cavitation.28 We used an in-house developed validated semiautomatic approach to define WMH. All scans were inspected for segmentation errors and gliosis surrounding lacunes or surrounding the ischemic stroke were excluded. WMH-volumes were normalized to intracranial volume.20 Microbleeds were defined as small areas of signal void with associated blooming seen on susceptibility-weighted imaging with a diameter of <10 mm, excluding signal voids in the area of the ischemic stroke.28,29 Lacunes of presumed vascular origin were defined as round or ovoid, subcortical, fluid-filled cavities in the territory of 1 perforating arteriole on the FLAIR image with a diameter of 3 mm to about 15 mm.28 One experienced researcher analyzed all magnetic resonance imaging data and was blinded for baseline characteristics and outcome measures. Inter-rater and intrarater reliability for the presence of microbleeds yielded a κ value of 1.0 and 0.92, respectively. In a random sample of 10%, the inter-rater reliability for the presence of lacunes yielded a κ value of 0.76 and intrarater reliability a κ value of 0.80.
Group differences in baseline characteristics between those who participated and those who did participate in the FUTURE study, but not in the present DTI study, were tested with a Mann–Whitney U test, ANOVA, or Pearson χ2 test, when appropriate. Two-tailed P values <0.05 were considered statistically significant.
Normalized WMH-volume and normalized stroke volume showed a right skewed distribution, and therefore a base-10 logarithm transformation was used. Because the WMH-volume data contained zeros', a constant number of 0.001 was added to all data points before log transformation.
For each TBSS analysis, 5000 permutations were used and significant associations were determined using the threshold-free cluster enhancement with a threshold of P<0.05 corrected for multiple comparisons.25,30,31 All TBSS analyses were adjusted for age, sex, follow-up duration, education, lesion location (left/right hemisphere), normalized WMH-volume, depressive symptoms, and fatigue unless otherwise stated. In case of missing values for these covariates (always <1.7%), the mean of the whole group was taken.
First, voxel-wise statistical analysis using TBSS was carried out to investigate the relationship between remote white matter integrity (FA and MD of the ipsilesional and contralesional hemisphere) in patients with stroke (n=117) and the cognitive index score, additionally adjusted for normalized stroke volume. Subsequently, this TBSS analysis was repeated for the performance on each of the 7 cognitive domains. In addition, to investigate the strength of the association between white matter integrity in the contralesional hemisphere (skeletal DTI parameters FA and MD) and the cognitive index score, a linear regression model was used with the cognitive index score as dependent variable and the mean FA/MD as independent variable, adjusted for age, sex, education, follow-up duration, depressive symptoms, fatigue, lesion location, normalized lesion volume, and normalized WMH-volume.
Next, we investigated whether low remote white matter integrity increased the risk of long-term cognitive impairment and whether a higher remote white matter integrity was associated with no poststroke cognitive impairment. Therefore, FA and MD white matter of the ipsilesional and contralesional hemisphere of patients with cognitive impairment were compared with cognitively unimpaired patients, additionally adjusted for normalized stroke volume.
Finally, we expected stroke volume and preexisting cerebral small vessel disease to be associated with remote lower white matter integrity and consequently account for poststroke cognitive performance itself. Therefore, the relationship between stroke volume and ipsilesional and contralesional FA and MD of the white matter was assessed, adjusting for age, sex, follow-up duration, lesion location (left/right), WMH-volume, depressive symptoms, and fatigue. The prevalence of microbleeds, lacunes, vascular risk factors, and treatment were compared between cognitively impaired and unimpaired patients using a χ2 test. Log-transformed WMH-volume and stroke volume were compared between these 2 groups with AN(C)OVA, adjusted for age in case of WMH-volume.
T1-weighted whole-brain imaging, DTI, FLAIR, susceptibility-weighted imaging, and overall cognitive performance were available from 117 patients with ischemic stroke and 84 stroke-free controls (Figure I in the online-only Data Supplement).
Differences in characteristics between those who participated in this study and nonparticipants are reported in Table I in the online-only Data Supplement. Mean follow-up duration of the population with stroke was 10.7 years (SD 8.1), mean age at follow-up was 49.8 years (SD 9.4), and the most frequent lesion location was in the middle cerebral artery territory (Table 1; Figure 1).
About 11 years after stroke, the FA of the white matter both in the ipsilesional and contralesional hemisphere was positively related with the cognitive index score, whereas white matter MD of the ipsilesional and contralesional hemisphere was negatively associated with the cognitive index score (n=117; Figure 1). These associations were independent of WMH-volume, stroke volume, and other confounding variables. Subsequent analyses showed that a lower remote white matter integrity was associated with a lower performance on the domains of processing speed, attention, working memory, and executive functioning (Figure II in the online-only Data Supplement).
A higher mean FA value of the white matter skeleton in the contralesional hemisphere was related to a higher cognitive index score (β=0.24, P=0.008). A higher mean MD value of the white matter in the contralesional hemisphere was related to a worse cognitive index score (β=−0.18, P=0.03).
Cognitively impaired patients more often had a stroke in the frontal (P=0.02) lobe and parietal lobe (P=0.004) compared with cognitively unimpaired patients (Table 2).
Even almost 11 years after stroke, a significant lower ipsilesional and contralesional white matter FA and a higher MD were observed in cognitively impaired patients compared with cognitively unimpaired patients (Figure 2).
Stroke volume was associated with lower white matter FA and higher MD values in the ipsilesional and contralesional hemisphere, especially involving the corpus callosum (Figure 3). Cognitively impaired patients showed a significant higher median WMH-volume, a larger stroke volume, had a higher prevalence of ≥1 microbleed(s), not of lacunes compared with cognitively unimpaired patients (Table 2). Cognitively impaired patients did not have a higher proportion of patients with vascular risk factors and treatment (except for statins) compared with unimpaired patients (Table 2).
This study showed that almost 11 years after ischemic stroke in young adults, a lower white matter integrity remote from the index stroke area, which could be as remote as the contralesional hemisphere, was associated with an increased risk of long-term cognitive impairment. Conversely, higher remote white matter integrity was associated with a low risk of long-term cognitive impairment. A larger stroke volume was associated with lower remote white matter integrity, even in the contralesional hemisphere. The presence of concomitant cerebral small vessel disease was more frequently observed in cognitive impaired patients compared with unimpaired patients; however, after controlling for WMH-volume in the TBSS analysis, this did not change the results on lower remote white matter integrity after stroke.
Although stroke has been associated with remote neurophysiological effects, termed as diaschisis, which tends to normalize over time,7 we now showed that about 11 years after stroke, a lower structural integrity of remote white matter is associated with long-term cognitive impairment.
There are some potential underlying mechanisms for lower remote white matter integrity and worse concomitant cognitive performance. As we adjusted for stroke volume in the analyses, stroke volume is not the only predictor for difference in remote white matter integrity and concomitant differences in long-term cognitive outcome. An alternative explanation could be the higher prevalence of cerebral small vessel disease in cognitively impaired patients compared with cognitively unimpaired patients. Lower white matter integrity, as measured by DTI, could be an early marker of the pathogenic cerebral small vessel disease–related mechanism before WMH, lacunes, and microbleeds occur on conventional magnetic resonance imaging.32 However, the absolute number of microbleeds, lacunes, or WMH-volume was relatively low in both groups, as expected in these relatively young patients, and after adjustment for WMH-volume, we still observed a lower remote white matter integrity in cognitively impaired patients compared with unimpaired patients. Therefore, it seems less likely that cerebral small vessel disease largely explained our results on remote white matter integrity and the concomitant worse cognitive performance.
Another explanation might be that lower remote microstructural integrity is caused by vascular damage caused by being exposed to the same vascular risk factors that caused the initial stroke event. Furthermore, based on studies in rats, stroke itself might cause spreading depression in the ipsilesional hemisphere,11,33 which may allow the onset of secondary (Wallerian) degeneration of remote white matter after ischemic stroke.34 However, the extent of this phenomenon should be further investigated to whether it also could explain the lower microstructural integrity in contralesional white matter as we observed in cognitively impaired patients. For instance, a lower white matter integrity of the corpus callosum in the contralesional hemisphere was associated with cognitive impairment. It seems that beside cognitive outcome, microstructural integrity of the corpus callosum has been found to predict the degree of motor function after stroke,35 and therefore seems to be an important white matter structure for outcome after stroke.
Another important clinical finding is the higher remote white matter integrity in patients with preserved cognitive performance, which might be due to a different underlying mechanism compared with lower remote white matter integrity. Plasticity and vicariation (intact brain areas take over the functions of the stroke area) might have occurred in cognitively unimpaired patients.7 This idea is possibly supported by our findings that FA correlated better with remote white matter integrity compared with MD, as lower FA may not necessarily reflect lower underlying structural integrity because FA reflects the directionality of molecular displacement by diffusion and is influenced by crossing fibers.36 MD reflects the magnitude of water diffusion, which is less influenced by the direction of fibers and therefore MD remains relatively constant.36,37 Therefore, lower FA, but not higher MD in patients with stroke may reflect plasticity, resulting into a reorganization of white matter and subsequently into more (new) crossing fibers.38 Caution must be taken in interpreting these findings as different pathologies, different stages of disease, or the rate of degeneration can lead to different tensor behaviors (FA and MD).37 Another possible explanation might be that no secondary neurodegeneration occurred in cognitively unimpaired patients as opposed to cognitively impaired patients.
Strengths of our study include its large sample size and the long follow-up duration. We collected baseline and follow-up information according to the identical procedures in all patients, used strict protocols for cognitive assessment, and researchers were trained, to reduce the risk of information bias.
However, some methodological issues of this study need to be considered. Although the FUTURE study has a prospective design, the current analysis is cross-sectional and we, therefore, can only report on lower microstructural integrity after stroke in cognitively impaired patients. Although this lower integrity may be caused by poststroke microstructural degradation of white matter, a longitudinal design is required to further support causality. However, our data clearly demonstrate a relationship between stroke volume and reduced white matter integrity.
Although we have tried to statistically correct for differences in stroke characteristics (volume and location) as thoroughly as possible between cognitively impaired patients and unimpaired patients, there still might be some residual confounding between these 2 groups. For instance, we observed a higher proportion of cognitively impaired patients with a stroke in the dorsolateral prefrontal cortex and intraparietal sulcus compared with unimpaired patients. These areas have been found to be connector hubs,39 which play a central role in integrating information from multiple networks.40 A stroke in these areas is considered as a strategic lesion, as it hampers the communication between different networks important for multiple cognitive components. Consequently, this may result into deterioration of multiple remote white matter fibers, which in turn more severely affects poststroke cognitive functioning.
Also, selection might have occurred, as patients who participated in the FUTURE study, but could not participate in the present DTI study, had a poorer outcome compared with participants. However, selection bias seems unlikely because both the direction and magnitude of the association between white matter integrity and cognitive performance will not be selectively different in participants and nonparticipants.
Another limitation of this study is the exclusion of patients with severe aphasia from cognitive testing. This might have potentially limited the generalizability of our present results on cognitive tests to a young first-ever ischemic population with stroke in general. However, our previous work on long-term cognitive outcome in these young patients with ischemic stroke showed a low number of patients with aphasia.12 Therefore, it seems unlikely that excluding these small number of patients largely influenced the generalizability of present results to young patients with ischemic stroke.
In conclusion, our results implicate that remote structural integrity of the white matter is associated with cognitive performance 11 years after stroke. Longitudinal studies are needed to investigate the course of remote lower white matter integrity and concomitant cognitive performance after the onset of stroke, which might reveal early individual treatment opportunities after stroke.
Sources of Funding
This study was funded by the Dutch Epilepsy Fund (grant No. 10–18).
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.116.014356/-/DC1.
- Received June 9, 2016.
- Revision received July 24, 2016.
- Accepted August 5, 2016.
- © 2016 American Heart Association, Inc.
- Cao M,
- Ferrari M,
- Patella R,
- Marra C,
- Rasura M
- Schaapsmeerders P,
- Maaijwee NA,
- van Dijk EJ,
- Rutten-Jacobs LC,
- Arntz RM,
- Schoonderwaldt HC,
- et al
- Carrera E,
- Tononi G
- Hervé D,
- Molko N,
- Pappata S,
- Buffon F,
- LeBihan D,
- Bousser MG,
- et al
- Reijmer YD,
- Freeze WM,
- Leemans A,
- Biessels GJ
- Visser E,
- Qin S,
- Zwiers MP
- Alberti KG,
- Zimmet PZ
- Wardlaw JM,
- Smith EE,
- Biessels GJ,
- Cordonnier C,
- Fazekas F,
- Frayne R,
- et al
- Nitkunan A,
- Barrick TR,
- Charlton RA,
- Clark CA,
- Markus HS
- Ayata C
- Thomalla G,
- Glauche V,
- Weiller C,
- Röther J
- Jiang Q,
- Zhang ZG,
- Chopp M
- Bertolero MA,
- Yeo BT,
- D’Esposito M