Impact of White Matter Hyperintensity Volume on Prognosis After Mechanical Thrombectomy in Ischemic Stroke Patients

Abstract

Background

Uncertainties remain about the impact of white matter hyperintensity (WMH) on prognosis after mechanical thrombectomy.

Methods

In this retrospective study from a national prospective registry of patients with ischemic stroke due to large vessel occlusion treated with mechanical thrombectomy, we assessed WMH volume using a quantitative semi‐automatized segmentation on baseline magnetic resonance imaging. We determined the association between WMH volume and the prognosis of patients with ischemic stroke presenting between 2019 and 2022.

Results

Among 902 patients (mean age 70.4 years, 50% women) with a baseline magnetic resonance imaging, the median WMH volume was 2.79 (0.75–9.14) mL. In multivariate analyses, increasing WMH volume was associated with a reduced probability of favorable outcome (modified Rankin Scale score 0–2) (adjusted odds ratio per 1 log+1 increase = 0.66, 95% confidence interval 0.54–0.82) and an increased risk of mortality at 90 days (adjusted odds ratio per 1 log+1 increase = 1.53, 95% confidence interval 1.23–1.90), with the greatest risk in patients with the highest WMH volume (>11 mL) compared to those with the lowest WMH volume (<2 mL) (adjusted odds ratio = 0.38, 95% confidence interval 0.21–0.67 and adjusted odds ratio = 3.04, 95% confidence interval 1.66–5.59, respectively). There was no association between WMH volume and recanalization success and risks of any post treatment intracranial hemorrhage, symptomatic intracranial hemorrhage, and parenchymal hemorrhage.

Conclusion

WMH volume is associated with increased risks of poor functional outcome and death at 90 days post mechanical thrombectomy but not with the probability of recanalization success and posttreatment intracranial hemorrhage. The use of semi‐automatized tool to assess WMH volume may help better identify patients who would benefit the most from mechanical thrombectomy and predict their prognosis.

Nonstandard Abbreviations and Acronyms

ASPECT

Alberta Stroke Program Early Computed Tomography

mRS

modified Rankin scale

NIHSS

National Institutes of Health Stroke Scale

TICI

Thrombolysis in Cerebral Infarction

WMH

white matter hyperintensity

Clinical Perspective

What Is New?

• This study provides new insights on the impact of white matter hyperintensity (WMH) on prognosis of ischemic stroke patients treated with mechanical thrombectomy.

What Are the Clinical Implications?

• WMH volume was significantly associated with increased risks of poor functional outcome (modified Rankin Scale score 0–2) and death at 90 days postmechanical thrombectomy.

• There was no association between WMH volume and the probability of recanalization success or intracranial hemorrhage postmechanical thrombectomy.

• The use of semi‐automatized tool to assess WMH volume may help better identify patients who would benefit the most from mechanical thrombectomy and predict their prognosis.

Mechanical thrombectomy is effective and recommended in most ischemic stroke patients with large vessel occlusion. ¹ However, half of patients remain disabled at 3 months after mechanical thrombectomy despite high rates of successful recanalization. ² , ³ The main complication after mechanical thrombectomy is the occurrence of symptomatic intracranial hemorrhage, with a prevalence ranging between 3% and 5%. ² , ³ Although this complication has been identified as the leading cause of neurological deterioration after stroke, this does not entirely explain the absence of functional recovery after successful reperfusion. ⁴

Lesions of cerebral small vessel disease, particularly white matter hyperintensity (WMH), might be a major contributor to patients’ outcome after ischemic stroke. ⁵ Indeed, WMH is associated with covert cognitive decline and physical disabilities that can be exacerbated after a stroke. ⁵ Authors have reported differing findings on the association between WMH and the prognosis after mechanical thrombectomy in patients with ischemic stroke; some studies found WMH severity to be a risk factor for symptomatic intracerebral hemorrhage and poor clinical outcome after mechanical thrombectomy ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ whereas others did not find evidence for an association. ¹² , ¹³ A recent systematic review totalizing approximately 1000 patients with ischemic stroke from 5 studies showed no significant association between WMH and symptomatic intracerebral hemorrhage post mechanical thrombectomy. ¹⁴ Nevertheless, the findings are unadjusted for important outcome predictors such as age, baseline National Institutes of Health Stroke Scale (NIHSS) score, use of intravenous (IV) thrombolysis and time from symptom onset to groin puncture, and the included studies used mixed methods for the detection of WMH volume. These discrepancies among the studies are likely to be due to the heterogeneous methods used to evaluate WMH severity and studies ‘small sample size. Also, most studies relied on computed tomography (CT) scan, which is less sensitive than magnetic resonance imaging (MRI) for the detection of WMH, and used a visual grading system, usually the Fazekas scale, rather than a semi‐automated method. ⁵ Moreover, some studies did not take into account potential effect modifiers, such as success of recanalization or delay from stroke onset to groin puncture. ⁸ , ¹¹ Furthermore, some studies involved selective populations recruited from tertiary stroke centers with short treatment delays. Also, the types of devices used for mechanical thrombectomy have improved over time. Further understanding of the association between WMH and prognosis after mechanical thrombectomy would certainly help to better identify patients who would benefit from mechanical thrombectomy and those in whom with mechanical thrombectomy would be futile. An objective, reliable, and reproductible tool for measurement of WMH volume is therefore necessary to interpret the findings. Also, such a tool should be rapidly and easily available to be useful in routine clinical practice.

We aimed to determine the association between WMH volume, using a semi‐automatic segmentation on baseline MRI and prognosis in patients with an ischemic stroke due to a large vessel occlusion treated with mechanical thrombectomy.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request.

We used data from the Endovascular Treatment in Ischemic Stroke (ETIS) registry, an ongoing, national prospective multicenter observational study since January 1, 2015 evaluating patients with an acute ischemic stroke due to an anterior or posterior large vessel occlusion treated with mechanical thrombectomy in 30 of the 35 comprehensive stroke centers in metropolitan France (https://www.clinicaltrials.gov; unique identifier: NCT03776877).

The study was approved by an independent ethics committee. In accordance with the French legislation, non‐opposition was sought and no written consent was required to collect data from the medical records up to 3 years after the inclusion event.

For this retrospective study using prospectively collected data, we included consecutive adults (age ≥18 years) who had a baseline MRI and consecutively underwent mechanical thrombectomy in the ETIS registry between January 2019 and January 2022. Patients were excluded if the quality of the T2‐fluid attenuated inversion recovery (T2‐FLAIR), T2 gradient recall echo (GRE), or susceptibility weighted imaging (SWI) sequences interfered with images interpretation. All GRE or SWI sequences on baseline MRI were assessed to exclude patients with acute intra or extra‐axial hemorrhage prior to mechanical thrombectomy.

The T2‐fluid attenuated inversion recovery sequences were segmented to calculate the WMH volume in milliliters using the lesion prediction algorithm (Chapter 6.1), ¹⁵ by a probability algorithm implemented in the Lesion Segmentation Toolbox for Statistical Parametric Mapping 12 toolbox, operating under MATLAB (R2021a). The lesion prediction algorithm consists of a binary classifier in the form of a logistic regression model with spatial covariate. It builds a lesion belief map to which a probability threshold is then used to classify voxels with a high probability of being WMH. The software enables the automatized segmentation of both hemispheres in less than 3 minutes. The segmented maps were all confirmed and manually edited if necessary (Figure 1). Sequelae of previous ischemic or hemorrhagic stroke and lacunes were not counted as WMH as recommended by the experts on small vessel disease. ⁵ The number of microbleeds was assessed on a GRE or SWI sequence. Any discrepancies in radiological assessments were discussed with a second investigator (M.B.).

Figure 1.

White matter hyperintensity segmentation in fluid‐attenuated inversion recovery (T2‐FLAIR) sequence axial magnetic resonance imaging (MRI) images using the lesion prediction algorithm (LPA) in patients with ischemic stroke. (A), Native T2‐FLAIR sequence axial MRI image from one patient in the registry. (B), Final segmentation masks of white matter hyperintensity volume (red dots) using the LPA from image (A), which was classified as between 2 and 11 mL. (C), Native T2‐FLAIR sequence axial MRI image from another patient in the registry. (D), Final segmentation masks of white matter hyperintensity volume (red dots) using the LPA from image (C), which was classified as >11 mL.

Data on demographic characteristics, past medical history, pre‐stroke modified Rankin scale (mRS) score, NIHSS score, Alberta Stroke Program Early Computed Tomography (ASPECT) score, blood pressure at admission, location of arterial intracranial occlusion, use of IV thrombolysis, the Thrombolysis in Cerebral Infarction (TICI) score after mechanical thrombectomy, any hemorrhagic transformation after mechanical thrombectomy were collected from medical records during hospitalization. The mRS score and presence of a recurrent ischemic stroke or transient ischemic attack at 90 days post‐stroke was assessed via face‐to‐face or telephone interviews. A favorable outcome was defined as a 90‐day post‐stroke mRS score of 0–2 or equal to the pre‐stroke value and an excellent outcome as 90‐day post‐stroke mRS score of 0–1 or equal to the pre‐stroke value. The mRS scale is a 7‐level scale that measures functional outcome with scores ranging from 0 (no symptoms) to 6 (death). ¹⁶ Other efficacy outcomes were the 24 hour‐change in the NIHSS score, the 24 hour‐change in the ASPECT score, successful (defined as TICI score ≥2b), and excellent recanalization (TICI score ≥2c). The NIHSS score measures the neurologic deficit due to stroke and ranges from 0 to 42, with higher scores indicating more severe deficits. ¹⁷ The ASPECT score measures the early ischemic changes on baseline brain imaging in acute anterior‐circulation ischemic stroke. The ASPECT score ranges from 0 to 10, with 1 point subtracted for any evidence of early ischemic change in each defined region. ¹⁸ , ¹⁹ Because all patients had baseline MRI, we used the diffusion weighted imaging ASPECT (DWI‐ASPECT) score. ¹⁸ , ²⁰ For posterior circulation stroke, we used the posterior circulation‐ASPECT score. ²¹ The TICI score is a 6‐level score that measures flow restoration after mechanical thrombectomy. ²² , ²³ Safety outcomes included 90‐day all‐cause mortality, presence of intracranial hemorrhage, symptomatic intracranial hemorrhage, and parenchymal hematoma on follow‐up brain imaging performed within 24 hours post‐stroke according to the European Cooperative Acute Stroke Study 2 classification. ²⁴

Statistical Analysis

Quantitative variables were expressed as means (SD) in the case of normal distribution or medians (interquartile range) otherwise. Categorical variables were expressed as numbers (percentage). Normality of distributions was assessed using histograms and the Shapiro–Wilk test. Outcomes rates were reported with 95% confidence intervals (CIs) calculated using Clopper–Pearson method.

We assessed the associations between WMH volume and binary outcomes using mixed logistic regression models including the center as a random effect. We also investigated the associations between WMH volume and difference in the NIHSS and the ASPECT scores from baseline to 24 hours in the subset of patients with excellent recanalization (TICI ≥2c) using linear mixed regression models including the center as a random effect and baseline values as fixed covariate.

First, WMH volume was analyzed as a 3‐level categorical variable (<2, 2–11, >11 mL). We did not use tercile ranges to classify WMH volume into 3 groups to avoid some sample specific findings regarding the associations between the distribution of the WMH volumes and the prognosis after mechanical thrombectomy. Because there are no established cutoffs for WMH volume categories, we empirically chose a 3‐category scale to define low, intermediate, and high WMH volume (<2, 2–11, >11 mL) as per existing literature. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ We considered a 3‐category scale to be a simpler and more reproducible measurement scale for WMH in routine clinical practice than a scale with a larger number of categories of classification. Second, WMH volume was analyzed as a continuous variable after applying a log+1‐transformation. Effect sizes (odds ratios [ORs] for binary outcomes or linear regression coefficients [slope] for quantitative outcomes) with their 95% CIs were calculated using the lowest WMH volume category as reference, or per log+1 increase in WMH volume.

Associations were further adjusted for prespecified variables (age, baseline NIHSS score, use of IV thrombolysis, admission systolic blood pressure, presence of at least 1 cerebral microbleed, time from symptom onset to groin puncture [the last known well time was used if symptom onset was unknown]) and location of arterial occlusion (M1 segment of the middle cerebral artery versus M2 middle cerebral artery versus intracranial internal carotid artery versus tandem occlusion versus vertebrobasilar artery) in mixed logistic or linear regression models. We also investigated the same associations in sensitivity analyses in the subset of patients with an intracranial occlusion of the anterior circulation (M1 or M2 middle cerebral artery, intracranial internal carotid artery, and tandem occlusion).

Finally, we assessed the potential heterogeneity in the association between WMH volume and favorable outcome (prespecified as the primary outcome measure) in prespecified subgroups: excellent recanalization (TICI <2c versus TICI ≥2c), sex, age (<60 versus 60–80 versus >80 years), use of IV thrombolysis, admission systolic blood pressure (<185 versus ≥185 mm Hg; we chose this threshold in reference to the target of blood pressure levels recommended before IV thrombolysis and mechanical thrombectomy by the international guidelines ¹ ), previous history of atrial fibrillation, time from symptom onset to groin puncture (<3 versus ≥3 hours); presence of at least 1 cerebral microbleed, baseline ASPECT score (0–5 versus >5), and prestroke mRS score (0–1 versus >1), by including the corresponding interaction term into multivariable mixed logistic regression models.

To avoid case deletion in analyses due to missing baseline characteristics or outcomes, missing data were imputed by multiple imputations using regression‐switching approach (chained equations with m = 10). ²⁵ Imputation procedure was performed under the missing at random assumption using all baseline characteristics and study outcomes with a predictive mean matching method for continuous variables and multinomial or binary logistic regression model for categorical variables. Estimates obtained in the different imputed data sets were combined using Rubin's rules. ²⁶

Statistical testing was performed at the 2‐tailed α level of 0.05. Data were analyzed using the SAS software package, release 9.4 (SAS Institute, Cary, NC) by the statistician of the ETIS registry (M.K.).

Results

Over the study period, the ETIS registry included 1017 ischemic stroke patients that had an intracranial occlusion treated with mechanical thrombectomy and a baseline MRI. We excluded 115 patients (45 for presence of an acute intracranial or extracranial hemorrhage on baseline MRI, 57 for artifacts on T2‐fluid attenuated inversion recovery or GRE/SWI and 13 for missing GRE/SWI), leaving 902 patients in the study (Figure 2 ). Of 902 patients (mean age 70 years, 49.8% women, 12.3% prestroke mRS score >1), 394 (43.7%) had a WMH volume <2 mL, 309 (34.2%) between 2 and 11 mL, and 199 (22.1%) >11 mL (Table 1). The overall median (interquartile range) WMH volume was 2.79 mL (0.75–9.14).

Figure 2.

Flow chart of the study. ETIS indicates Endovascular Treatment in Ischemic Stroke; FLAIR, fluid attenuated inversion recovery; GRE, gradient recall echo; MRI, magnetic resonance imaging; and SWI, susceptibility weighted imaging.

Table 1.

Baseline Characteristics of Patients

	Values
Demographic characteristics
Mean (SD) age, y	70.4 (15.0)
Women	449/902 (49.8)
Past medical history
Arterial hypertension	506/874 (57.9)
Diabetes mellitus	134/864 (15.5)
Hypercholesterolemia	266/858 (31.0)
Current smoking	175/821 (21.3)
Previous ischemic or hemorrhagic stroke	139/866 (16.1)
Coronary artery disease	128/855 (15.0)
Atrial fibrillation	176/857 (20.5)
Prestroke modified Rankin Scale score >1	104/844 (12.3)
Baseline characteristics
Mean (SD) systolic blood pressure (mm Hg) *	148.1 (26.3)
Mean (SD) diastolic blood pressure (mm Hg) †	82.7 (18.0)
Median (IQR) NIHSS score ‡	15 (10 to 20)
Median (IQR) ASPECT score §	8 (6 to 9)
Intracranial occlusion location
M1 segment middle cerebral artery	475/902 (52.7)
M2 segment middle cerebral artery	117/902 (13.0)
Tandem	135/902 (15.0)
Internal carotid artery	123/902 (13.6)
Vertebro‐basilar artery	52/902 (5.8)
Treatment and median (IQR) procedural times (minute)
Use of IV thrombolysis	468/899 (52.1)
Alteplase	328/446 (73.5)
Tenecteplase	118/446 (26.5)
Symptom onset to imaging ||	132 (98 ‐184)
Symptom onset to groin puncture **	262 (197‐345)
Stroke etiology
Atheroma	149/864 (17.3)
Cardio‐embolism	388/864 (44.9)
Other	327/864 (37.8)

Values are no./total no. (%) unless otherwise indicated. ASPECTS indicates Alberta Stroke Program Early Computed Tomography; IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale.

^*Data missing for 109 patients.

^† Data missing for 111 patients.

^‡ Data missing for 28 patients.

^§ Data missing for 102 patients.

^|| Data missing for 50 patients.

^**Data missing for 49 patients.

Follow‐up brain imaging within 24 hours after stroke symptoms was conducted by MRI in 445 patients (49.3%) and CT scan in 457 patients (50.7%). Any intracranial hemorrhage within 24 hours of stroke treatment occurred in 413 patients (45.8%; 95% CI, 42.4%–49.2%), including 84 (10.3%) parenchymal hematomas, and 76 (8.7%) symptomatic intracranial hemorrhages. At 90 days post‐stroke, 407 patients (45.1%, 95% CI 41.5%–48.6%) had a favorable outcome, 288 (31.9%, 95% CI 28.6%–35.3%) had an excellent outcome, and 181 (20.1%, 95% CI 17.4%–23.0%) had died.

In univariate and multivariate analyses, increasing WMH volume was associated with a reduced probability of favorable outcome at 90 days (adjusted OR [aOR] per 1 log+1 increase = 0.66, 95% CI, 0.54–0.82, Table 2, Figure 3) and excellent outcome (aOR per 1 log+1 increase: 0.71, 95% CI 0.57–0.88). Compared to patients with the lowest WMH volume (<2 mL), those with the highest WMH volume (>11 mL) had a significantly reduced chance of favorable outcome (aOR = 0.38, 95% CI 0.21–0.67, Table 2) and excellent outcome (aOR = 0.38, 95% CI 0.21–0.71) at 90 days. Patients with an intermediate WMH volume (2–11 mL) had also a significantly reduced probability of favorable outcome (aOR = 0.66, 95% I 0.45–0.98) and a nonsignificantly reduced chance of excellent outcome (aOR = 0.76, 95% CI 0.51–1.13) at 90 days than those with the lowest WMH volume (<2 mL).

Table 2.

Association Between White Matter Hyperintensity Volume and Efficacy and Safety Outcomes in Ischemic Stroke Patients with an Intracranial Occlusion of the Anterior or Posterior Circulation Treated with Mechanical Thrombectomy

	WMH volume (mL)			P value *	Effect size (95% CI)	P value †
	<2 (n=394)	2–11 (n=309)	>11 (n=199)
Efficacy outcomes
Favorable outcome (90‐day mRS score of 0–2 or equal to the prestroke value)
N (%)	233 (59.2)	119 (38.5)	44 (22.4)
OR (95% CI)	–	0.43 (0.32–0.59)	0.20 (0.13–0.31)	<0.001	0.52 (0.45–0.61)	<0.001
Adjusted OR (95% CI) ‡	–	0.66 (0.45–0.98)	0.38 (0.21–0.67)	0.004	0.66 (0.54–0.82)	<0.001
Excellent outcome (90‐day mRS score of 0–1 or equal to the prestroke value)
N(%)	160 (40.7)	92 (29.6)	33 (16.3)
OR (95% CI)	–	0.61 (0.44–0.85)	0.28 (0.17–0.46)	<0.001	0.63 (0.54–0.74)	<0.001
Adjusted OR (95% CI) ‡	–	0.76 (0.51–1.13)	0.38 (0.21–0.71)	0.008	0.71 (0.57–0.88)	0.002
90‐day all‐cause mortality
N(%)	37 (9.4)	71 (23.0)	69 (34.7)
OR (95% CI)	–	2.88 (1.86–4.47)	5.06 (3.21–8.00)	<0.001	1.82 (1.56–2.13)	<0.001
Adjusted OR (95% CI) ‡	–	2.10 (1.25–3.52)	3.04 (1.66–5.59)	0.001	1.53 (1.23–1.90)	<0.001
Difference in NIHSS score from baseline to 24 h §
Mean difference (95% CI)	4.8 (3.8–5.9)	5.6 (4.5 to 6.9)	2.6 (1.1 to 4.0)
Mean difference (95% CI)	–	0.85 (−0.72 to 2.43)	−2.26 (−4.08 to 0.77)	0.006	−0.80 (−1.42 to −0.17)	0.013
Adjusted mean difference (95% CI) ‡	–	1.17 (−0.48 to 2.81)	−2.15 (−4.17 to 0.14)	0.002	−0.90 (−1.64 to −0.17)	0.016
Difference in ASPECT score from baseline to 24 h §
Mean difference (95% CI)	−0.1 (−0.5–0.2)	−0.1 (–0.5 to 0.3)	0.0 (−0.5 to 0.4)
Mean difference (95% CI)	–	0.02 (−0.35 to 0.40)	0.11 (−0.30 to 0.52)	0.79	0.06 (−0.09 to 0.21)	0.45
Adjusted mean difference (95% CI) ‡	–	−0.10 (−0.50 to 0.30)	−0.12 (−0.60 to 0.36)	0.91	−0.00 (−0.18 to 0.18)	0.98
Successful recanalization (TICI ≥2b)
N(%)	352 (89.4)	275 (89.0)	175 (88.0)
OR (95% CI)	–	0.97 (0.59 to 1.58)	0.57 (0.51 to 1.51)	0.91	0.94 (0.78 to 1.14)	0.53
Adjusted OR (95% CI) ‡	–	1.12 (0.63 to 1.99)	1.23 (0.61 to 2.49)	0.85	1.08 (0.83 to 1.39)	0.58
Excellent recanalization (TICI ≥2c)
N(%)	267 (68.4)	209 (67.7)	129 (65.0)
OR (95% CI)	–	0.96 (0.69 to 1.33)	0.86 (0.59 to 1.25)	0.75	0.95 (0.84 to 1.08)	0.46
Adjusted OR (95% CI) ‡	–	1.05 (0.71 to 1.55)	1.06 (0.66 to 1.72)	0.97	1.03 (0.87 to 1.24)	0.70
Safety outcomes:
Any intracranial hemorrhage
N(%)	180 (45.6)	148 (47.8)	87 (43.8)
OR (95% CI)	–	1.11 (0.82 to 1.50)	0.94 (0.66 to 1.34)	0.67	1.01 (0.90 to 1.14)	0.86
Adjusted OR (95% CI) ‡	–	1.07 (0.74 to 1.55)	0.85 (0.54 to 1.35)	0.55	0.98 (0.83 to 1.15)	0.79
Symptomatic intracranial hemorrhage
N(%)	34 (8.7)	27 (8.8)	21 (10.5)
OR (95% CI)	–	1.03 (0.60 to 1.77)	1.28 (0.70 to 2.33)	0.72	1.13 (0.92 to 1.39)	0.25
Adjusted OR (95% CI) ‡	–	0.95 (0.49 to 1.85)	1.10 (0.49 to 2.43)	0.92	1.12 (0.84 to 1.48)	0.45
Parenchymal hemorrhage
N(%)	38 (9.6)	41 (13.4)	17 (8.7)
OR (95% CI)	–	1.50 (0.92 to 2.45)	0.91 (0.47 to 1.76)	0.15	0.99 (0.81 to 1.22)	0.93
Adjusted OR (95% CI) ‡	–	1.43 (0.77 to 2.65)	0.76 (0.33 to 1.77)	0.13	0.88 (0.67 to 1.17)	0.39

Descriptive parameters, P values, and effect size (95% CI) were calculated after handling missing values using multiple imputation procedure (m = 10). ASPECT indicates Alberta Stroke Program Early Computed Tomography; 95% CI, 95% confidence intervals; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; TICI, Treatment in Cerebral Infarction; and WMH, white matter hyperintensity.

^* Odds ratio and P value obtained using a mixed logistic regression model including WMH volume as categorical variable with lower level as reference.

^† Odds ratio (or regression coefficient) per 1 log increase in WMH volume and P values obtained using a mixed regression model (logistic or linear according the outcomes) including white matter hyperintensity volume (log‐transformed) as a continuous variable.

^‡ Adjusted for center, age, baseline NIHSS score, use of intravenous thrombolysis, time from symptom onset to groin puncture, admission systolic blood pressure, presence of at least 1 cerebral microbleed, and location of arterial occlusion.

^§ Difference in ASPECT and NIHSS scores from baseline to 24 hours were calculated in the subgroup of patients with excellent recanalization (TICI ≥2c).

Figure 3.

Distribution of the modified Rankin Scale score at 90 days after mechanical thrombectomy stratified by white matter hyperintensity (WMH) volume: >11, 2–11, and <2 mL. Numbers in bars represent percentages.

Increasing WMH volume was also associated with a higher risk of all‐cause mortality at 90 days (aOR per 1 log+1 increase = 1.53, 95% CI 1.23–1.90), with the greatest risk being in patients with the highest WMH volume compared to those with the lowest WMH volume (aOR = 3.04, 95% CI 1.66–5.59, Table 2).

Among patients with an excellent recanalization (TICI ≥2c), increasing WMH volume was associated with a reduced chance of improvement in the NIHSS score between baseline and 24 hours after treatment (aOR per 1 log+1 increase = −0.90, 95% CI −1.64 to −0.17, Table 2). Compared to patients with the lowest WMH volume, those with the highest WMH volume had less neurological improvement between baseline and 24 hours after treatment (mean difference in 24‐hour change in NIHSS score = −2.15 point, 95% CI −4.17 to −0.14), whereas there was no significant difference in patients with intermediate WMH volume (1.17, 95% CI −0.48 to 2.81, Table 2). There was no association between WMH volume and 24‐hour change in ASPECTS score in patients with an excellent recanalization (Table 2).

After adjustment for the prespecified variables, there was no association between WMH volume (both in continuous and categorical analyses) and recanalization success as well as the risk of any post‐treatment intracranial hemorrhage, symptomatic intracranial hemorrhage, and parenchymal hemorrhage (Table 2).

The sensitivity analyses in patients with an intracranial occlusion of the anterior circulation showed similar findings (Table 3).

Table 3.

Association Between White Matter Hyperintensity Volume and Efficacy and Safety Outcomes in Patients With Ischemic Stroke an Intracranial Occlusion of the Anterior Circulation Treated with Mechanical Thrombectomy

	WMH volume (ml)			P values *	Effect size (95% CI)	P value †
	<2 (n=372)	2–11 (n=289)	>11 (n=189)
Efficacy outcomes
Favorable outcome (90‐day mRS score of 0–2 or equal to the prestroke value)
N(%)	222 (59.6)	109 (37.5)	43 (22.5)
OR (95% CI)	–	0.41 (0.30 to 0.56)	0.20 (0.13 to 0.31)	<0.001	0.51 (0.44 to 0.60)	<0.001
Adjusted OR (95% CI) ‡	–	0.64 (0.43 to 0.97)	0.38 (0.21 to 0.71)	0.009	0.65 (0.52 to 0.82)	<0.001
Excellent outcome (90‐day mRS score of 0–1 or equal to the prestroke value)
N(%)	151 (40.6)	81 (28.0)	32 (16.7)
OR (95% CI)	–	0.57 (0.41 to 0.80)	0.29 (0.17 to 0.48)	<0.001	0.62 (0.52 to 0.74)	<0.001
Adjusted OR (95% CI) ‡	–	0.73 (0.48 to 1.12)	0.41 (0.22 to 0.78)	0.023	0.70 (0.56 to 0.89)	0.003
90‐day all‐cause mortality
N(%)	31 (8.3)	68 (23.5)	62 (32.8)
OR (95% CI)	–	3.40 (2.13 to 5.42)	5.35 (3.27 to 8.75)	<0.001	1.85 (1.57 to 2.17)	<0.001
Adjusted OR (95% CI) ‡	–	2.50 (1.44 to 4.33)	3.22 (1.69 to 6.15)	0.001	1.57 (1.25 to 1.97)	<0.001
Difference in NIHSS score from baseline to 24 h §
Mean difference (95% CI)	5.2 (4.0–6.2)	5.8 (4.6 to 7.0)	2.6 (1.2 to 4.1)
Mean difference (95% CI)	–	0.68 (−0.90 to 2.25)	−2.48 (−4.31 to −0.65)	0.005	−0.88 (−1.51 to −0.25)	0.006
Adjusted mean difference (95% CI) ‡	–	1.13 (−0.53 to 2.79)	−2.03 (−4.07 to 0.02)	0.003	−0.92 (−1.67 to −0.18)	0.015
Difference in ASPECT score from baseline to 24 h §
Mean difference (95% CI)	−0.1 (−0.4–0.2)	−0.0 (−0.4 to 0.3)	0.0 (−0.4 to 0.4)
Mean difference (95% CI)	–	0.04 (−0.30 to 0.39)	0.13 (−0.27 to 0.53)	0.76	0.07 (−0.07 to 0.21)	0.32
Adjusted mean difference (95% CI) ‡	–	−0.12 (−0.51 to 0.27)	−0.11 (−0.61 to 0.36)	0.79	0.00 (−0.18 to 0.18)	0.99
Successful recanalization (TICI ≥2b)
N(%)	332 (89.3)	258 (89.1)	166 (88.0)
OR (95% CI)	–	0.97 (0.59 to 1.61)	0.87 (0.50 to 1.54)	0.92	0.93 (0.76 to 1.13)	0.45
Adjusted OR (95% CI) ‡	–	1.12 (0.62 to 2.03)	1.22 (0.59 to 2.53)	0.88	1.04 (0.80 to 1.36)	0.76
Excellent recanalization (TICI ≥2c)
N(%)	245 (67.1)	191 (66.2)	122 (64.4)
OR (95% CI)	–	0.95 (0.68 to 1.32)	0.89 (0.61 to 1.30)	0.84	0.95 (0.84 to 1.09)	0.47
Adjusted OR (95% CI) ‡	–	1.05 (0.70 to 1.56)	1.09 (0.67 to 1.78)	0.95	1.03 (0.87 to 1.23)	0.72
Safety outcomes
Any intracranial hemorrhage
N(%)	178 (47.8)	144 (49.7)	84 (44.6)
OR (95% CI)	–	1.10 (0.80 to 1.51)	0.88 (0.61 to 1.27)	0.55	0.99 (0.88 to 1.13)	0.94
Adjusted OR (95% CI) ‡	–	1.03 (0.71 to 1.51)	0.80 (0.50 to 1.28)	0.47	0.96 (0.81 to 1.13)	0.61
Symptomatic intracranial hemorrhage
N(%)	34 (9.2)	27 (9.4)	19 (10.0)
OR (95% CI)	–	1.03 (0.60 to 1.78)	1.14 (0.61 to 2.11)	0.93	1.10 (0.89 to 1.35)	0.39
Adjusted OR (95% CI) ‡	–	0.95 (0.49 to 1.86)	1.00 (0.44 to 2.24)	0.98	1.09 (0.82 to 1.46)	0.56
Parenchymal hemorrhage
N(%)	38 (10.2)	41 (14.3)	16 (8.4)
OR (95% CI)	–	1.54 (0.94 to 2.52)	0.83 (0.42 to 1.62)	0.090	0.98 (0.80 to 1.20)	0.82
Adjusted OR (95% CI) ‡	–	1.44 (0.78 to 2.69)	0.70 (0.30 to 1.66)	0.086	0.87 (0.65 to 1.16)	0.35

Descriptive parameters, P values and effect size (95% CI) were calculated after handling missing values using multiple imputation procedure (m = 10). ASPECTS indicates Alberta Stroke Program Early Computed Tomography; 95% CI, 95% confidence intervals; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; TICI, Treatment in Cerebral Infarction; and WMH, white matter hyperintensity.

^* Odds ratio and P value obtained using a mixed logistic regression model including WMH volume as categorical variable with lower level as reference.

^† Odds ratio (or regression coefficient) per 1 log increase in WMH volume and P values obtained using a mixed regression model (logistic or linear according the outcomes) including WMH volume (log‐transformed) as a continuous variable.

^‡ Adjusted for center, age, baseline NIHSS score, use of IV thrombolysis, time from symptom onset to groin puncture, admission systolic blood pressure, presence of at least 1 cerebral microbleed, and location of arterial occlusion.

^§ Difference in ASPECT and NIHSS scores from baseline to 24 hours were calculated in the subgroup of patients with excellent recanalization (TICI ≥2c).

There was no significant heterogeneity in the association between WMH volume and favorable outcome at 90 days across the prespecified subgroups (Figure 4).

Figure 4.

Forest plot showing adjusted association between white matter hyperintensity (WMH) volume and favorable outcome (defined as a 90‐day mRS score of 0–2 or equal to the pre‐stroke value) after mechanical thrombectomy in prespecified subgroups. Adjusted odds ratios (aORs) were calculated per 1 log+1 increase of WMH volume after adjustment for center, age, NIHSS score at baseline, use of IV thrombolysis, time from symptom onset to groin puncture, admission systolic blood pressure, presence of at least 1 cerebral microbleed, and site of arterial occlusion. P values obtained using a mixed logistic regression model including WMH volume as continuous variable. P‐het: P alue for heterogeneity in the association between WMH volume and favorable outcome in subgroups calculated by including the corresponding interaction term in multivariable mixed logistic regression models. 95% CI indicates 95% confidence intervals; ASPECT, Alberta Stroke Program Early Computed Tomography; HR, hazard ratio; IV, intravenous; mRS, modified Rankin scale; NIHSS, National Institutes of Health Stroke Scale; and TICI: Treatment in Cerebral Infarction.

At 90 days post‐stroke, the rates of recurrent ischemic stroke or transient ischemic attack were 2.5% (9/358) in patients with the lowest WMH volume (<2 mL), 3.7% (10/273) in patients with an intermediate WMH volume (2–11 mL) and 3.5% (6/174) in those with the highest WMH volume (>11 mL).

Discussion

In this large retrospective study from a national prospective registry of 902 ischemic stroke patients treated with mechanical thrombectomy, we found that WMH volume was an independent risk factor for 90‐day poor functional outcome and all‐cause mortality, after adjustment for important confounders such as use of IV thrombolysis, time from stroke onset to groin puncture, age, and NIHSS score at baseline. We also found no heterogeneity in the association between WMH volume and 90‐day favorable outcome across subgroups.

Our study brings new insights on the association between MWH volume and prognosis after mechanical thrombectomy. We used a quantitative semi‐automatized method for assessment of WMH in a large and unselected population of patients with ischemic stroke. We also adjusted our analyses for several confounders. Furthermore, we evaluated various outcomes, both in the early and later phases of ischemic stroke. The worse functional outcome and increased mortality rate at 90 days in patients with higher WMH volume is not explained by higher risks of intracranial hemorrhage post‐treatment. Indeed, we showed no association between WMH volume and any post‐treatment intracranial hemorrhage, symptomatic intracranial hemorrhage and parenchymal hematoma after adjustment for the prespecified variables. This updates the unadjusted findings of a systematic review assessing the association between WMH and symptomatic intracranial hemorrhage posttreatment, which includes studies that used heterogenous methods to detect WMH volume. ¹⁴ Moreover, our data show that the association is not explained by a lower probability of successful recanalization in patients with higher WMH volume. Interestingly, among patients with an excellent recanalization (TICI ≥2c), those with the highest WMH volume had less 24 hour‐NIHSS improvement than those with the lowest WMH volume.

The mechanisms underlying the impact of WMH burden on prognosis in patients with ischemic stroke is not entirely understood. WMH burden has been hypothesized to enhance susceptibility to acute ischemia, microvascular rarefaction, and reduced blood flow. ²¹ , ²⁷ , ²⁸ , ²⁹ , ³⁰ Poorer outcomes with WMH burden are also thought to be related to disruption of neuronal networks, with reduction of cognitive reserve and therefore less potential for recovery with rehabilitation. ³¹ The poorer longer‐term prognosis with WMH burden might also be due to increased risk of stroke recurrence, although the risk was similar in this study. ³²

Our findings have several clinical implications. The use of WMH volume assessment in baseline imaging, in addition to clinical variables such as age and NIHSS score, may help clinicians to better select patients with ischemic stroke for mechanical thrombectomy. Detection of WMH burden could also help to inform on prognostic expectations post‐treatment for patients and caregivers. Additionally, several trials aim to further extend indications of mechanical thrombectomy, both in time window and in population selection, such as those with minor clinical deficits or large infarct volume. Assessment of WMH burden would therefore add to interpretation of such trials’ findings. WMH volume should be included, among other baseline prognostic factors, in future trials of mechanical thrombectomy.

The key strength of the study is the use of a quantitative semi‐automatized method to assess WMH volume, rather than a visual method that is prone to degrees of inter‐rater variability, and the use of MRI, rather than CT scan, which tends to be less sensitive to detect WMH. ²¹ These have minimized the risk of information bias. Furthermore, imaging findings were assessed by a core laboratory that reviewed all imaging data. Another strength of this study includes the use of a large multicenter national prospective cohort of patients treated with mechanical thrombectomy, which is one of the largest registries worldwide. Also, we adjusted the analyses for several prespecified confounding factors, including the presence of other markers of small vessel disease such as cerebral microbleeds. Moreover, we analyzed WMH volume as a categorical and a quantitative variable and were able to confirm a log linear dose‐effect relationship with functional outcomes and mortality. We also showed a dose‐response effect in the association between WMH volume and functional outcome and mortality in the analyses using WMH as a 3‐level categorical variable. Finally, we investigated various outcomes, both in the early and later phases of ischemic stroke, to better elucidate the association between WMH volume and prognosis.

Our study has nevertheless some limitations. First, although the national multicenter ETIS registry exhibits a good representation of the population treated with mechanical thrombectomy in France, patients are predominantly White, which may limit the generalizability of the study. Second, we did not take into account the distribution of WMH, that is, periventricular or deep WMH. However, WMH volume in our patients was low (median volume = 2.79 mL) and categorizing by the distribution of WMH would have created small subgroups. A smaller study found no association between deep or periventricular WMH and symptomatic intracranial hemorrhage, 90‐day functional outcome, and mortality postmechanical thrombectomy. ⁹ Prevalence of WMH increases with higher age and disability. ⁵ However, the vast majority of our population had no or minimal prestroke disability and we adjusted analyses for age; therefore, this should not have significantly affected our findings.

Conclusion

In patients with ischemic stroke due to large vessel occlusion, WMH volume is significantly associated with increased risks of poor functional outcome and death at 90 days post mechanical thrombectomy. However, there is no association between WMH volume and the probability of recanalization success or intracranial hemorrhage postmechanical thrombectomy. The use of semi‐automatized tool to assess WMH volume may help better identify patients who would benefit the most from mechanical thrombectomy and predict their prognosis.

The complete list of the ETIS registry collaborators is available in the Supplemental Material.

Sources of Funding

This work is supported by a government grant managed by the French National Research Agency (ANR) as part of the future investment program integrated into France 2030, under grant agreement No. ANR‐18‐RHUS‐0001.

Disclosures

None.

Supporting information

List of the ETIS registry centers and co‐investigators