Fluid-Attenuated Inversion Recovery Hyperintense Ischemic Stroke Predicts Less Favorable 90-Day Outcome after Intravenous Thrombolysis

The duration of cerebral ischemia prior to reperfusion is one of the strongest predictors of functional outcome [1, 2]. The duration of ischemia and the appearance of the stroke on MRI fluid-attenuated inversion recovery (FLAIR) are related: longer time from onset to imaging increases the likelihood that the lesion will appear on FLAIR [3-7]. However, 23–44% of acute ischemic strokes (AIS) within 4.5 h from known onset was found to be FLAIR positive [3-6]. These data suggest that either the time from onset is imprecise or lesion visibility on FLAIR is mediated by factors other than time that may influence the benefit of thrombolysis.

We hypothesized that in AIS patients treated with intravenous (IV) thrombolysis within 4.5 h of known onset, those with a visible lesion on pretreatment FLAIR would be associated with less favorable 90-day outcome. Furthermore, given the apparent dissociation of time and FLAIR positivity of the lesion in some patients, we hypothesized that lesion visibility on FLAIR would predict outcome independent of time.

This is an analysis of patients enrolled in the Institutional Review Board-approved NIH Natural History of Stroke study (NCT00009243); written informed consent was obtained for all patients. Patients screened and enrolled by the NIH stroke team at MedStar Washington Hospital Center in Washington, DC, USA, and Suburban Hospital in Bethesda, MD, USA, between January 1, 2009, through December 13, 2019, were included in this study if (1) admission diagnosis was AIS, (2) last known normal (LKN) time and symptom discovery time were known, and (3) standard IV thrombolysis within 4.5 h of LKN was administered. Patients were excluded if (1) not screened with MRI prior to thrombolysis; (2) thrombectomy, intra-arterial thrombolysis, or emergent hemicraniectomy was performed; or (3) enrolled in other clinical trials. Patients were excluded based on their initial MRI if they had imaging evidence of (1) infratentorial stroke [8, 9], (2) diffusion-weighted imaging (DWI) negative, or (3) missing FLAIR data. Figure 1 presents the study population profile.

Patients were scanned before and 24 h after thrombolysis on one of the 3 commercial systems (1.5T Signa; GE Medical Systems, Milwaukee, WI, USA; 3T Skyra; Siemens AG, Munich, Germany; and 3T Achieva; Philips Healthcare, Best, the Netherlands). The specific MRI sequences acquired at both time points included DWI, gradient recalled echo (GRE), and pre- and postcontrast FLAIR after a single weight-dependent dosage of gadolinium. The MRI series parameters were comparable across systems. The imaging protocol parameters for the FLAIR varied slightly across the study period but in general consisted of the following: repetition time/echo time 9,000/120–140 ms, inversion time 2,200 ms, acquisition matrix of 256 × 128 or 256 × 256, forty 3.5-mm thick contiguous axial oblique slices, and 24-cm field of view.

Pre- and postthrombolysis MRIs were evaluated for abnormalities on DWI, apparent diffusion coefficient (ADC), GRE, and pre- and postcontrast FLAIR sequences. Three readers (Y.K., M.L., and N.B.) performed imaging assessments. Each reader, blinded to the LKN time and outcome, reviewed 2/3 of the MRI studies independently of the other readers, providing 2 complete sets of imaging reads for each patient. Any discrepancies in imaging reads were determined by consensus of 3 readers.

Each index ischemic lesion was identified as DWI bright and ADC dark on prethrombolysis MRI and confirmed to be consistent with the localization provided by the stroke clinician at the time of screening. The patients evaluated as DWI negative were confirmed by all readers and then were excluded from the analyses.

The DWI lesion volume was calculated by a semiautomated planimetric measurement method, using the Medical Image Processing, Analysis, and Visualization application (v10; Center for Information Technology, NIH). Lesion visibility on precontrast FLAIR was characterized as a parenchymal hyperintensity concordant with the DWI and ADC findings. Visible parenchymal hyperintensity was assessed as FLAIR positive or FLAIR negative if completely absent.

GRE at 24 h after thrombolysis was assessed for hemorrhagic transformation (HT) by applying the Heidelberg Bleeding Classification (HBC) [10]. The highest grade of HBC was reported when a mixture of grades was present. Symptomatic intracerebral hemorrhage was defined as a local or remote parenchymal hematoma combined with a neurological deterioration by ≥4 points on the 24-h NIH stroke scale (NIHSS) or death.

Postcontrast FLAIR of the 24-h postthrombolysis MRI was assessed for the hyperintense acute reperfusion injury marker (HARM) [11]. HARM presented in ≥10 slices was rated as severe. If the severe HARM was restricted to the lesion territory, then it was graded as severe focal; otherwise, it was graded as severe diffuse.

Demographic and clinical data including age, sex, race, ethnicity, lesion laterality, and risk factors including hypertension, diabetes mellitus, and tobacco use were reported. Any risk factor data with <2% prevalence in the study population were excluded from the analyses. Clinical severity assessments included preadmission modified Rankin Scale (mRS), admission NIHSS, 24-h NIHSS, 30-day mRS, and 90-day mRS. Time intervals calculated in minutes were LKN-to-MRI, MRI-to-needle, and LKN-to-needle times.

Patients with FLAIR-negative and FLAIR-positive stroke were compared according to outcomes. The primary outcome was a favorable functional outcome at 90 days defined as mRS equal to 0–1 or equal to preadmission mRS. Secondary clinical outcomes were 90-day functional independence defined as a 90-day mRS of 0–1, return to baseline defined as 90-day mRS equal to preadmission mRS, all-cause mortality at 90 days, early neurologic improvement (ENI) defined as a reduction of ≥4 points or 24-h NIHSS equal to 0–1 [12], major ENI defined as a reduction of ≥8 points or 24-h NIHSS equal to 0–1 [13], and the rates of discharge to home from the hospital.

Unadjusted analyses were performed on the ordinal or continuous variables using the Mann-Whitney U test. As appropriate based on sample size, either the χ² test or Fisher’s exact test was used for categorical variables. Spearman’s Rho test was used to assess the correlation between 2 continuous variables. Cohen’s Kappa test was used to assess interrater agreement. Odds ratios (OR) and 95% confidence intervals (CI) for each outcome were estimated adjusting for the following covariates in a logistic regression model: LKN-to-needle time, DWI lesion volume, age, female sex, admission NIHSS, and unfavorable baseline defined as preadmission mRS >1. Descriptive data were reported as median (25th quartile–75th quartile) or number (percentage) unless otherwise noted. Statistical analyses were performed with IBM SPSS software (version 1.0.0.1406).

From January 1, 2009, through December 13, 2019, 1,003 patients were treated with IV thrombolysis at the 2 study centers (shown in Fig. 1). A total of 297 patients were eligible and included in this study. Patients were 48% female, 51% White, 44% Black, and 4% Hispanic. Median age was 70 years (59–83), admission NIHSS was 6 (3–14), and preadmission mRS was 0 (0–1).

In 62% of patients, LKN time equaled symptom discovery time. Median difference of LKN and symptom discovery time for the remaining 38% was 30 min (15–46). LKN-to-MRI time was 109 min (79–156), MRI-to-needle time was 32 min (22–44), and LKN-to-needle time was 150 min (112–190). The MRI-to-needle time was not different between the FLAIR-negative and FLAIR-positive groups. LKN-to-MRI time and LKN-to-needle time were strongly related (ρ = 0.939, p < 0.001). The interrater agreement across the 3 readers was 84% for the FLAIR-positive assessment (κ = 0.604, 95% CI: 0.557–0.652), 96% for the presence of HT (κ = 0.889, 95% CI: 0.860–0.917), 91% for HT classification (κ = 0.795, 95% CI: 0.760–0.829), and 97% for the presence of any HARM (κ = 0.932, 95% CI: 0.908–0.956).

Seventy-four patients (25%) had visible lesions on prethrombolysis FLAIR. Figure 2 compares the clinical presentation, treatment timing, and lesion on the initial MRI for 2 patients with FLAIR-negative versus FLAIR-positive lesion. Baseline demographic and clinical characteristics are shown in Table 1. FLAIR-positive lesion was associated with right hemispheric lesion (OR 1.005, 95% CI: 1.000–1.010, p = 0.045) and longer LKN-to-MRI time (OR 1.786, 95% CI: 1.039–3.071, p = 0.036) after the regression analysis of the baseline characteristics.

Overall, 48% of patients in the study had a favorable clinical outcome. The median score on the mRS at 90 days was 2 (1–4) in the FLAIR-negative group and 3 (1–4) in the FLAIR-positive group. After adjustment for the covariates, FLAIR-positive lesion was significantly associated with less favorable 90-day outcome (OR 0.428, 95% CI: 0.201–0.912, p = 0.028). Among covariates, older age (OR 0.934, 95% CI: 0.908–0.960, p < 0.001), greater admission NIHSS (OR 0.907, 95% CI: 0.859–0.957, p < 0.001), and longer LKN-to-needle time (OR 0.993, 95% CI: 0.987–0.999, p = 0.018) predicted less favorable 90-day outcome.

Table 2 presents clinical and radiologic outcomes. The FLAIR-positive lesion was associated with less-frequent ENIs and more frequent development of severe HARM at 24 h after thrombolysis.

Our study is the first to report an association between the lesion visibility on prethrombolysis FLAIR and less favorable 90-day outcome in AIS patients who receive IV thrombolysis within 4.5 h of LKN. The difference is clinically meaningful with approximately a 20% greater chance of favorable 90-day outcome when the ischemic lesion is FLAIR negative versus FLAIR positive. The association between FLAIR-positive lesion and 90-day outcome in thrombectomy patients was shown in another study [14].

We found that AIS patients with FLAIR-negative lesion had a 47% greater chance of ENI, compared to the patients with FLAIR-positive lesion. This finding is consistent with a prior study [12]. FLAIR-positive stroke had a stronger association with ENIs than 90-day outcome. The short-term outcomes are less affected by compliance with rehabilitation or social support; thus, they may be better predicted by early imaging features.

Visible stroke on prethrombolysis FLAIR was associated with greater odds of severe HARM at 24 h. HARM is an MRI marker of blood-brain barrier (BBB) disruption associated with reperfusion and is associated with worse clinical outcomes [11, 15]. One proposed mechanism for FLAIR lesion hyperintensity is vasogenic edema [16]. As ischemia progresses, the BBB becomes more permeable, and there is influx of water from the vasculature to the parenchyma [17]. An early FLAIR hyperintensity may represent early loss of BBB integrity that may predispose to reperfusion injury, ultimately compromising the clinical outcome.

Therefore, the FLAIR-positive stroke patients within 4.5-h window may benefit from neuroprotective agents. Many neuroprotective agents have been studied in AIS, but so far none have shown benefit, even in combination with recanalization [18]. Targeting AIS patients with early FLAIR hyperintensity could yield more successful neuroprotection trials.

While a prior study reported a higher rate of postthrombolysis HT in the FLAIR-positive group [19], the rate of HT was not different between groups in our study. The development of HT may require additional factors with BBB disruption, such as hypertension. In our study, patients with severe HT had higher admission mean arterial pressure than patients with mild or no HT (122 vs. 106 mm Hg, p = 0.002). FLAIR-positive stroke patients may benefit from a lower postthrombolysis blood pressure goal than the conventional 185/105 mm Hg to prevent severe HT.

Our study has several limitations. It was a retrospective study and is hypothesis generating, and therefore statistical significance particularly in secondary outcomes should be weighed appropriately. Despite approaching all treated AIS patients consecutively, not every patient was enrolled. The proportion of FLAIR-positive patients was small, which is consistent with the early time window but decreases our statistical power. Not everyone in the study population had matched LKN and symptom discovery time, decreasing the predictive power of time on stroke outcome. Across a 10-year period, various imaging protocols were used. However, this adds to the generalizability of our findings, and future studies should determine the optimal FLAIR sequence parameters to select treatment candidates while minimizing scan time. Finally, most stroke centers in the USA screen patients with CT, so our findings are not easily adopted. As more stroke centers adopt acute MRI protocols, predictive early imaging markers will become more valuable.

Even when the LKN was well constrained by the treatment window, patients with FLAIR-positive stroke had less favorable functional outcome. A future clinical trial of thrombolysis combined with neuroprotection or aggressive blood pressure management in patients with FLAIR-positive stroke within 4.5 h of LKN may be warranted.

The authors are grateful to their patients and their families for being involved in the research. They acknowledge the contribution of the stroke team at the National Institute of Neurological Diseases and Stroke and the clinical care teams at both Suburban Hospital and MedStar Washington Hospital Center. They also would like to acknowledge Saman Mojibi for her support and management of the program.

This study is an analysis of the Institutional Review Board-approved NIH Natural History of Stroke study (NCT00009243) data; written informed consent was obtained for all patients.

The authors have no financial conflicts of interest.

This research was supported by the Intramural Research Program of the National Institute of Neurological Disorders and Stroke, NIH.

Y.K. and M.L.: design and conceptualization of the study, major role in the acquisition, analysis, and interpretation of data, drafting and revising the manuscript for intellectual content. N.B.: major role in the acquisition of data and revising the manuscript for intellectual content. G.N.: analysis and interpretation of data. R.L., C.B.W., K.C.K., A.W.H., J.K.L., and M.M.A.: revising the manuscript for intellectual content. L.L.L.: design of the study and revising the manuscript for intellectual content.