Fluid-Attenuated Inversion Recovery Vascular Hyperintensities–Diffusion-Weighted Imaging Mismatch Identifies Acute Stroke Patients Most Likely to Benefit From Recanalization

Mismatch between perfusion-weighted imaging (PWI) and diffusion-weighted imaging (DWI) magnetic resonance imaging (MRI) has been proposed to select patients with acute stroke for recanalization therapy.¹ However, this approach requires gadolinium injection and dedicated software and increases scan duration. We have recently proposed an alternative model based on the mismatch between fluid-attenuated inversion recovery (FLAIR) vascular hyperintensities (FVH) and DWI to assess the presence of large penumbra.² FVH have been reported in 45–100% of stroke patients with intracranial occlusion.^3–16 They represent slow retrograde flow in leptomeningeal collaterals,⁴ which maintain some perfusion distal to the occlusion until revascularization.¹⁷ The FVH-DWI mismatch focuses on FVH beyond the boundaries of the cortical DWI lesion, ignoring FVH adjacent to the DWI lesion. Therefore, only the FVH facing tissue at risk of infarct expansion are considered in this definition. Rapidly assessable with the naked eye, this model is simple and reproducible and predicts large PWI-DWI mismatch with excellent sensitivity.²

We hypothesized that complete recanalization after intravenous tissue-type plasminogen activator would be associated with better outcome in patients with, as compared to those without, FVH-DWI mismatch.

This retrospective analysis was based on a prospectively collected monocenter registry of patients exclusively treated by intravenous tissue-type plasminogen activator (0.9 mg/kg) for acute ischemic stroke between 2004 and 2014. Inclusion criteria were (1) proximal (M1) middle cerebral artery occlusion, (2) pretreatment and 24-hour follow-up MRI to assess recanalization, and (3) 3-month clinical follow-up, including modified Rankin Scale score. The study was approved by the local Ethics Committee.

Pretreatment and follow-up magnetic resonance were performed on 1.5-T units (General Electric Healthcare) using a standardized protocol.^2,18 PWI was included in the pretreament magnetic resonance whenever feasible with no delay, but was not part of follow-up magnetic resonance. FVH were defined as focal, tubular, or serpentine hyperintensities in the subarachnoid space relative to cerebrospinal fluid³ and corresponding to typical arterial course. Axial FLAIR and DWI images were reviewed together, blinded to PWI and clinical data. FVH-DWI mismatch was considered present when FVH extended beyond the boundaries of the cortical DWI lesion, that is, when at least one FVH was facing isointense cortex on DWI (Figure). Pretreatment (DWI₁) and follow-up (DWI₂) lesions were segmented to compute DWI volumes as detailed elsewhere.¹⁸ Hypoperfused tissue defined as time to maximum value of the residue function (T_max)>6 s was segmented from PWI maps (BrainStat AIF, READY View, GE Healthcare), and PWI-DWI mismatch was defined as Volume_{hypoperfusion}/Volume_DWI1>1.8.¹ Pretreatment occlusion and complete recanalization were assessed on magnetic resonance angiography.

FVH-DWI and PWI-DWI mismatch were coded as binary variables. Continuous variables were compared with Student’s t-test or Mann–Whitney U test. Categorical variables were compared using χ² or Fisher’s exact test, as appropriate. Odds ratios (OR) for 3-month favorable outcome (modified Rankin Scale ≤2) were calculated from 2×2 tables. Analyses were stratified on 24-hours recanalization. Logistic regression analyses were done in the FVH-DWI mismatch group and in the no FVH-DWI mismatch group, with 3-month favorable outcome as the dependent variable and recanalization as the explanatory variable. These analyses were secondarily adjusted on admission National Institutes of Health Stroke Scale score in multivariable logistic models using Firth correction¹⁹ to avoid potential overfitting, given the small number of patients without FVH-DWI mismatch but with favorable outcome.²⁰ We tested the potential interaction between FVH-DWI mismatch and recanalization to predict favorable outcome using the Breslow-day test.

One hundred and sixty-four patients (85 men) met the inclusion criteria (median [interquartile range] age 72 years [59–82], admission National Institutes of Health Stroke Scale score 17 [12–21], and 3-month modified Rankin Scale 3 [1–5]). PWI was available in 104/164 (63%) patients. Median time-to-initial MRI, time-to-treatment, and time-to-follow-up MRI were 117 mins [88–148], 153 mins [122–189], and 23 hours [20–27].

121/164 (74%) patients had a FVH-DWI mismatch. Median time-to-initial MRI was slightly shorter in patients with FVH-DWI mismatch than in patients without FVH-DWI mismatch (115 mins [86–142] versus 122 mins [107–149] respectively; P=0.043). PWI-DWI mismatch was present in 72/104 (69%) patients. Agreement between FVH-DWI and PWI-DWI for mismatch classification was 86/104 (83%), with κ=0.58. Seven patients had a PWI-DWI mismatch but no FVH-DWI mismatch, and 11 had no PWI-DWI mismatch but a FVH-DWI mismatch.

In the whole population, complete recanalization occurred in 50/164 (30%) patients and was significantly associated with favorable outcome (OR=8.3; 95% confidence interval [CI], 3.9–17.8; P<0.001).

After stratification for FVH-DWI mismatch status (Table), the association between recanalization and favorable outcome was significant in patients with FVH-DWI mismatch (OR=16.2; 95% CI, 5.7–46.5; P<0.0001), but not in patients with no FVH-DWI mismatch (OR=2.6; 95% CI, 0.6–12.1; P=0.22), P=0.048 for interaction. After adjustment on National Institutes of Health Stroke Scale, the odds ratios for favorable outcome with recanalization were 16.6 (95% CI, 5.2–52.7; P<0.0001) in patients with FVH-DWI mismatch and 6.5 (95% CI, 0.95–44.4; P=0.06) in patients without FVH-DWI mismatch.

After stratification for PWI-DWI mismatch status, the association between complete recanalization and favorable outcome remained significant in patients with PWI-DWI mismatch (OR=9.9; 95% CI, 3.1–31.3; P=0.0001) and was marginally significant in patients without PWI-DWI mismatch (OR=7.0; 95% CI, 1.1–44.1; P=0.047), P=0.76 for interaction.

The present study shows that a simple combination of DWI and FLAIR findings can identify a subgroup of patients with M1 occlusion who are likely to benefit from recanalization. Indeed, the effect size for favorable outcome in recanalizers was much stronger (OR=16.2) in patients with than in those without FVH-DWI mismatch (OR=2.6), with a significant interaction between these variables. These results strengthen the notion that FVH located beyond the DWI cortical lesion boundaries reflect impaired yet viable tissue. When recanalization is achieved, this tissue restores its function, explaining why patients with FVH-DWI mismatch are more likely to reach favorable clinical outcome. Similar results have been shown using the PWI-DWI mismatch, with more favorable clinical outcomes in target mismatch patients who had early reperfusion after endovascular stroke treatment.¹ We have previously reported an excellent (92%) sensitivity of the FVH-DWI mismatch to identify patients with large PWI-DWI mismatch, indicating that the former only rarely missed the latter.²

The benefit from recanalization (OR=7.0; 95% CI, 1.1–44.1; P=0.047) in patients without PWI-DWI mismatch was somewhat unexpected. Owing to a stringent and binary definition for the mismatch (here, as in DEFUSE 2,¹ PWI_Tmax>6sec/DWI=1.8), some patients in the no PWI-DWI mismatch group likely had a clinically meaningful amount of salvageable tissue, even if not large enough to be classified as mismatch positive.

There are several advantages of using the FLAIR-based mismatch model: it obviates the need for gadolinium contrast; it is easily assessable, reproducible,² does not require any dedicated software, and does not expand the acquisition time because FLAIR is commonly used to rule-out stroke mimics and estimate stroke age.²¹ Other mismatch alternatives that do not require contrast injection include the clinical-DWI mismatch²² and the magnetic resonance angiography-DWI mismatch.^23,24 However, both imply a volumetric analysis of the DWI lesion, which requires dedicated automated software or manual segmentation. Although untested to date, the FVH-DWI mismatch could be transposed to CT by replacing the FVH by a collateral CT score²⁵ and use the CT-cerebral blood flow threshold <30% instead of DWI for the core.²⁶ Finally, the FVH-DWI mismatch might be a predictor of smaller final infarct volume, even in patients who do not recanalize. However, this point is beyond the scope of the present study and should be investigated in larger cohorts, which would allow adjusting for initial DWI volume and other potential confounders.

Our study is retrospective and based on selected patients with M1 occlusion from a single center, with a sample size that limits the direct comparison of PWI- and FVH-DWI mismatch models after stratification on recanalization. We, however, previously compared the 2 methods using different T_max thresholds² and found that FVH-DWI was a reasonable surrogate to identify patients with large PWI-DWI mismatch. Finally, results might have been different had the stratification been based on early recanalization or reperfusion instead of 24-hours magnetic resonance angiography.

Because the FVH-DWI mismatch seems to identify patients most likely to benefit from recanalization, it may provide an alternative to PWI-DWI mismatch to select patients who are candidates for recanalization and enrich trial cohorts. Our results need to be validated in an external data set and tested in patients treated by thrombectomy or bridging therapy.