Association of White Matter Hyperintensities With Short-Term Outcomes in Patients With Minor Cerebrovascular Events
Background and Purpose—White matter lesions (WML) are associated with cognitive decline, increased stroke risk, and disability in old age. We hypothesized that superimposed acute cerebrovascular occlusion on chronic preexisting injury (leukoaraiosis) leads to worse outcome after minor cerebrovascular event, both using quantitative (volumetric) and qualitative (Fazekas scale) assessment, as well as relative total brain volume.
Methods—WML volume assessment was performed in 425 patients with high-risk transient ischemic attack (TIA; motor/speech deficits >5 minutes) or minor strokes from the CATCH study (CT and MRI in the Triage of TIA and Minor Cerebrovascular Events to Identify High Risk Patients). Complete baseline characteristics and outcome assessment were available in 412 patients. Primary outcome was disability at 90 days, defined as modified Rankin Scale score of >1. Secondary outcomes were stroke progression, TIA recurrence, and stroke recurrence. Analysis was performed using descriptive statistics and regression models including interaction terms.
Results—Median age was 69 years, 39.8% were female. Sixty-two patients (15%) had unfavorable outcome with disability at 90 days (modified Rankin Scale score >1). Higher Fazekas scores were strongly correlated with higher WML volume (r=0.79). Both higher Fazekas score and higher WMH volume were associated with disability at 90 days in univariate regression (odds ratio 1.22; 95% confidence interval, 1.04–1.43 and odds ratio, 1.25 per milliliter increase; 95% confidence interval, 1.02–1.54, respectively) but not with stroke progression, TIA recurrence, or stroke recurrence. In multivariable-adjusted analyses, additive interaction terms were associated with unfavorable outcome (adjusted odds ratio 3.99, 95% confidence interval, 1.87–8.49).
Conclusions—Our data suggest that quantitative and qualitative WML assessments are highly correlated and comparable in TIA/minor stroke patients. WML burden is associated with short-term outcome of patients with good prestroke function in the presence of intracranial stenosis/occlusion.
White matter lesions (WML) or leukoaraiosis are associated with cognitive decline, dementia, increased stroke risk, and disability in old age. This might be explained by axonal changes that interfere with recruitment and reorganization of ipsi- and contralesional brain regions during poststroke recovery.1 In patients with acute ischemic stroke, severity of leukoaraiosis measured on magnetic resonance imaging (MRI) is associated with progression of infarct independent of initial insult size, age, admission blood glucose, admission blood pressure, and stroke subtype.2 WML are likely a marker of chronic cerebrovascular injury, and its burden may signify a diminished capacity of cerebral tissue to withstand ischemia.3
We examined the association between one of the most widely used and well-validated visual rating scales (the Fazekas scale) and volumetric assessment using a previously well-described image-processing method in a large cohort.4 We also examined the association between WML burden alone or relative to total brain volume and rate of progression, recurrence or disability and whether these associations differ according to whether quantitative or qualitative assessments are being used. We hypothesized that superimposed acute cerebrovascular occlusion on chronic preexisting injury (leukoaraiosis) leads to worse outcome after minor cerebrovascular event, both using quantitative and qualitative assessment, as well as relative to total brain volume.
The data that support the findings of this study are available from Dr Shelagh Coutts () on reasonable request.
We analyzed patients of the previously published CATCH study (CT and MRI in the Triage of TIA and Minor Cerebrovascular Events to Identify High Risk Patients).5,6 CATCH was a prospective study in which consecutive patients aged at least 18 years presenting with a high-risk transient ischemic attack (TIA) with focal weakness or speech disturbance (lasting ≥5 minutes) or minor ischemic stroke (National Institutes of Health Stroke Scale score ≤3) who were referred to the stroke team at Foothills Medical Center, Calgary, were considered for enrollment. Patients were examined by a stroke neurologist and had a computed tomography/computed tomographic angiography completed within 24 hours of onset. A majority of patients (83% of the CATCH cohort) had an MRI brain scan completed and are the focus of this analysis. Comparative baseline characteristics and outcomes for patients who underwent MRI and those who did not are provided in Table I in the online-only Data Supplement. Exclusion criteria were premorbid modified Rankin Scale (mRS) score of ≥2 or a serious comorbid illness that would likely result in death within 3 months and thrombolysis as offered according to Canadian Best Practice guidelines for disabling strokes within 4.5 hours of symptom onset. The local institutional ethics committee approved this protocol, and patients provided written informed consent. Detailed baseline clinical and outcome information were prospectively collected for each patient. Baseline clinical information included age, sex, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, smoking, pre-event antiplatelet or anticoagulant drug use, nature of presenting symptoms, whether symptoms were ongoing in the emergency department, and National Institutes of Health Stroke Scale score. Premorbid conditions were based on medical history obtained through self-reported detailed clinical history by the patient. Patients whose symptoms resolved before assessment in the emergency department were classified as TIA; those with ongoing symptoms at the time of assessment, even if minor, were classified as ischemic stroke.
Computed Tomographic Imaging and Vasculature Assessment
Standard whole-brain axial computed tomography was performed and immediately followed by computed tomographic angiography from the aortic arch to skull vertex with a helical scan technique. Computed tomographic angiography source images were reformatted into thin 3-mm sagittal, coronal, and axial images and thick 24-mm axial maximum intensity projection slabs for the intracranial circulation and 3-mm oblique sagittal section through the carotid bifurcations.5 Computed tomographic angiographies were assessed for intracranial or extracranial occlusion or stenosis ≥50%.
Magnetic Resonance Imaging
MRIs were completed on either a GE 3-T scanner or a Siemens 1.5-T MR scanner. Sequences included diffusion-weighted imaging, apparent diffusion coefficient, and fluid-attenuated inversion recovery (FLAIR). Sequence parameters are provided in Table II in the online-only Data Supplement.
Ischemic Infarct Measurement
MRI was assessed for acute or hyperacute lesions using axial diffusion-weighted imaging and apparent diffusion coefficient by a rater blinded to clinical information.
Fazekas Visual Rating Scale
A neuroradiology fellow (J.M.) assessed FLAIR sequences of available MRI scans and scored WML on Fazekas scale, blinded to clinical information. In the Fazekas rating scale, periventricular and deep WMLs are rated separately.7 Periventricular hyperintensities are scored as follows: 0=absence, 1=caps or pencil-thin lining, 2=smooth halo, and 3=irregular periventricular hyperintensities extending into the deep white matter. Deep white matter hyperintensities are scored as follows: 0=absence, 1=punctuate foci, 2=beginning confluence of foci, and 3=large confluent areas. For statistical comparison with WML volumes, a total Fazekas WML score, ranging from 0 to 6, was obtained by summing the periventricular and deep white matter scores.
WML Volume and Total Brain Volume Measurement
These methods were previously described.8 In brief, WML were defined according to consensus standards and measured quantitatively using a computer software tool Quantomo (version 1.0; Cybertrials, Inc, Calgary, Canada).9,10 In this method, WMLs are segmented by placing a single seed within each visible WML, with identification of the WML borders by the automated computer algorithm. The WML borders can be adjusted as necessary by the user to correct errors such as misidentification of skull as WML. WML volumes were quantified using the spatial dimensions of the voxels in each MRI slice. Acute diffusion-weighted imaging lesions that were also visible on FLAIR were not included in these measurements. Quantitative WML measurements on FLAIR sequences were performed by 2 trained readers (1 with an undergraduate and 1 with a graduate degree). Our MRI core laboratory has a protocol for training readers to measure WML, specified in standard operating procedures that are designed to comply with Food and Drug Administration guidance for imaging end points. A neurologist (E.E.S.) with 14-year experience reviews at least 10 scans with the readers before they are allowed to make independent measurements. E.E.S. was one of the leaders of the STRIVE (Standards for Reporting Vascular Changes on Neuroimaging) consensus group that provided definitions for WML and other radiological signs of cerebral small vessel disease.9 We require all readers to show good reliability (intraclass correlation coefficient >0.90) with a training set before making independent measurements.
A blinded study coordinator rated the 90-day modified Rankin Scale outcome. Progression was defined as a worsening deficit during hospitalization compared with the baseline assessment with or without a change in the total National Institutes of Health Stroke Scale score. A recurrent event was defined as a new sudden focal neurological deficit of vascular origin lasting <24 hours (recurrent TIA) or >24 hours (recurrent stroke) occurring at any time between the initial assessment and 90-day follow-up mandating repeat imaging. These outcome events were reviewed in detail by 2 stroke neurologists and a neuroradiologist, and a consensus decision was made on progression or recurrence of stroke.5
Measures of central tendency and measures of variability are reported using standard descriptive statistics. WML volume and WML volume ratio (WML volume divided by total brain volume) were transformed using a logarithm function to normalize the distribution. Frequencies of baseline characteristics were compared using Fisher exact test. We tested whether there was a difference in mean WML volume between mRS score 0 to 1 and mRS score 2 to 6 using Student t test. Agreement between Fazekas scale score and WML volume was assessed using Spearman correlation.
The primary outcome was the presence of disability, defined as mRS score >1. Secondary outcomes were stroke progression, TIA recurrence, and stroke recurrence. In a multivariable logistic regression model, we included variables that were statistically significant in the univariable analysis and then used backwards elimination to develop parsimonious models that best explained the data using main effects only. Then, in 3 separate models, we assessed the multiplicative interaction term—intracranial occlusion/stenosis by white matter disease measure (Fazekas scale score, WML volume, or WML volume ratio). We also assessed for an additive interaction and constructed a 4-level dummy variable coded as 0 (neither WML nor an intracranial stenosis/occlusion is present), 1 (WML is present but no intracranial stenosis/occlusion), 2 (no WML is present but intracranial stenosis/occlusion is), and 4 (both WML and intracranial stenosis/occlusion are present). Backward elimination was again used to find the final parsimonious model. For this dummy variable, WML was defined as absent (Fazekas scale score 0–2) versus present (Fazekas scale score 3–6).
All reported P values are 2 sided.
Statistical analysis was performed using STATA software (Stata 14; Stata Corp, College Station, TX).
White matter and brain volume assessment were available from 425 CATCH patients. Six were excluded for poor quality of FLAIR sequences and 7 were lost to follow-up, leaving 412 out of 425 (96.9%) patients with complete baseline characteristics and outcome assessments for analysis. Seventy-three patients were scanned on a 1.5-T scanner and 346 on a 3-T scanner. Median age of the patients was 67.3 years, and 39.8% were female. Median symptom onset time to MRI was 17.3 hours (25%–75%; confidence interval, 10.3–22.4). The mean WML volume across all patients was 9.42 mL. Twenty-six (6.3%) patients had a large vessel occlusion/stenosis (either intracranial internal carotid artery or middle cerebral artery main branch). Other baseline characteristics are shown in Table 1.
Higher Fazekas scale scores were highly correlated with higher WML volume (r=0.79; Figure).
Higher Fazekas scale scores, higher WML volume, and higher WML volume ratio at baseline were associated with unfavorable outcome at 90 days (defined as mRS score >1) using univariable regression analysis, but they were not associated with stroke progression, TIA recurrence, or stroke recurrence.
In the final parsimonious models of the multivariable regression analysis, high Fazekas score, high WML volume, and WML volume ratio were not associated with disability at 90 days (mRS score 2–6; Table 2). Only the presence of an intracranial stenosis/occlusion was significantly associated with disability at 90 days (adjusted odds ratio, 2.70; 95% confidence interval, 1.67–4.40; adjusted odds ratio, 2.59; 95% confidence interval, 1.61–4.17; adjusted odds ratio, 2.58, 95% confidence interval, 1.60–4.16, respectively). There was no evidence of multiplicative interaction (P=0.40, P=0.92, and P=0.95, respectively).
However, there was evidence of an additive interaction between WMLs and intracranial occlusion/stenosis. The percentage of disability (mRS score 2–6) at 90 days in patients with neither WML (using dichotomized Fazekas scale score) or an intracranial stenosis/occlusion was 12.2%, in patients with WML only was 12.4%, and in patients with an intracranial stenosis/occlusion was 14.6%. However, if patients had presence of both WML and an intracranial stenosis/occlusion, the percentage of disability at 90 days increased to 35.6%. This effect was maintained in the final parsimonious multivariable regression model (Table 3).
The main finding of our prospective cohort study is that WML burden does impact short-term outcome of TIA and minor stroke patients with good prestroke function in the presence of intracranial stenosis/occlusion on an additive scale. Patients with a high WML burden have a poorer outcome if they also present with an intracranial occlusion/stenosis. Additionally, we found that qualitative and quantitative assessments of WML burden assessment were highly correlated and generally showed similar associations with outcomes, suggesting that a quick visual rating may be sufficient to identify cross-sectional associations with other characteristics of minor stroke and TIA patients.
Qualitative and quantitative assessment of WML burden in our prospective cohort was highly correlated. High correlation between Fazekas scale scores and volumetric WML assessment has previously been demonstrated in the LADIS study (Leukoaraiosis and Disability) that included patients aged 65 to 85 years with no or mild disability in everyday life and some degree of WML on MRI that presented for evaluation in various settings (stroke unit, memory clinic, neurological or geriatric units, population studies on aging, and controls in other studies).11,12 A study of community-dwelling individuals with a mean age of 72.7 years of whom 27% had a history of cerebrovascular disease also found that visual and volumetric WML assessments are highly correlated.13 Our results complement these prior studies by also demonstrating high correlation for the first time in a more diverse patient population presenting with TIA and minor stroke. The high correlation with gold-standard quantitative measurements supports the use of the Fazekas scale, which is faster and easier to apply than quantitative methods, for WML measurement in stroke patients. However, quantitative WML methods may have other advantages over visual rating scales, including increased sensitivity to change over time and provide more detail on WML location for studies correlating WML location with clinical and cognitive deficits.
Over the past years, longitudinal population-based and hospital-based studies in stroke patients showed a dose-dependent relationship between WML and outcome, including the finding that large confluent WML are associated with dementia and disability.14 A recent study of >5000 ischemic stroke patients found that higher white matter hyperintensity quintiles were independently associated with worse 3-month mRS scores and that leukoaraiosis may affect stroke subtypes differently.15 A Japanese study demonstrated that a higher degree of WML was associated with worse inpatient convalescent rehabilitation outcomes measured by functional independence measure motor and cognitive scores in patients with ischemic stroke.16 Onteddu et al17 retrospectively analyzed 185 patients with minor stroke (National Institutes of Health Stroke Scale score ≤5) and found that leukoaraiosis was a more sensitive predictor for neurological deficit recovery at 90 days compared with age. Our study adds to the field that the combination of qualitative and quantitative WML assessment in TIA and minor stroke patients plus the presence of an intracranial stenosis/occlusion predicts functional disability.
Strengths of our study are the prospective data acquisition and enrollment of a large typical patient cohort presenting emergently with minor ischemic stroke and TIA symptoms. Limitations include the follow-up period of only 90 days and that cognitive assessment was not taken into consideration as part of the follow-up assessment at the time the study was designed.
When modeling postevent disability using independent terms for WML and intracranial occlusion, we found no evidence of multiplicative interaction, with a significant effect of intracranial occlusion as a main effect. However, when using an additive interaction approach where the group without either WML or intracranial occlusion serves as the reference, it was only the patients with both high WML and intracranial occlusion, who were at higher risk for postevent disability. Although the appropriate scale on which to statistically assess interaction in such a way that it reflects biological interaction is still debated, our data could be interpreted as evidence that patients with the combination of high WML and intracranial occlusion are at the highest risk of postevent disability.18 The mechanism of this association, and whether it can be modified to improve outcomes, cannot be determined from our observational study.
As a limitation, the CATCH study excluded patients with premorbid disability, and, thus, results do not apply to these patients. Furthermore, the mRS score might not be an appropriate outcome measurement, with a limited number of scores being less responsive to small changes that could be caused by WML (cognitive impairment, walking speed, etc).19 We do not have information on prior strokes or TIAs, but if any stroke had occurred previously, it must have been nondisabling because premorbid disability was an exclusion criterion. Additional reasons for disability after a minor cerebrovascular event are still not fully understood and should be important areas of future work.
The main finding of our prospective cohort study is that WML burden was associated with short-term outcomes in TIA and minor stroke patients who had good prestroke function in the presence of intracranial stenosis/occlusion. Qualitative and quantitative WML burden assessments are highly correlated in TIA/minor stroke patients, suggesting that quick visual rating may be sufficient.
We thank Michael D. Hill for his help with the statistical analysis.
Sources of Funding
The CATCH study was supported by the Canadian Institute of Health Research and a Pfizer Cardiovascular research award.
Guest Editor for this article was Jeffrey L. Saver, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.017429/-/DC1.
- Received March 21, 2017.
- Revision received January 31, 2018.
- Accepted February 16, 2018.
- © 2018 American Heart Association, Inc.
- Ay H,
- Arsava EM,
- Rosand J,
- Furie KL,
- Singhal AB,
- Schaefer PW,
- et al
- Rost NS,
- Fitzpatrick K,
- Biffi A,
- Kanakis A,
- Devan W,
- Anderson CD,
- et al
- Coutts SB,
- Modi J,
- Patel SK,
- Demchuk AM,
- Goyal M,
- Hill MD
- Zerna C,
- Modi J,
- Bilston L,
- Shoamanesh A,
- Coutts SB,
- Smith EE
- Fazekas F,
- Niederkorn K,
- Schmidt R,
- Offenbacher H,
- Horner S,
- Bertha G,
- et al
- Patel SK,
- Hanly PJ,
- Smith EE,
- Chan W,
- Coutts SB
- Wardlaw JM,
- Smith EE,
- Biessels GJ,
- Cordonnier C,
- Fazekas F,
- Frayne R,
- et al
- Kosior JC,
- Idris S,
- Dowlatshahi D,
- Alzawahmah M,
- Eesa M,
- Sharma P,
- et al
- van Straaten EC,
- Fazekas F,
- Rostrup E,
- Scheltens P,
- Schmidt R,
- Pantoni L,
- et al
- Gouw AA,
- Van der Flier WM,
- van Straaten EC,
- Barkhof F,
- Ferro JM,
- Baezner H,
- et al
- Senda J,
- Ito K,
- Kotake T,
- Kanamori M,
- Kishimoto H,
- Kadono I,
- et al