Routine Use of FLAIR Negative MRI in the Treatment of Unknown Onset Stroke

Abstract

Background:

Treatment of FLAIR-negative stroke in patients presenting in an unknown time window has been shown to be safe and effective. However, implementation can be challenging due to the need for hyper-acute MRI screening. The purpose of this study was to review the routine application of this practice outside of a clinical trial.

Methods:

Patients presenting from 3/1/16 to 8/22/18 in a time window <4.5 hours from symptom discovery but >4.5 hours from last known normal were included if they had a hyper-acute MRI performed. Quantitative assessment based on the MR WITNESS trial and qualitative assessment based on the WAKE-UP trial were used to grade the FLAIR images. The MR WITNESS trial used a quantitative assessment of FLAIR change where the fractional increase in signal change had to be <1.15, whereas the WAKE-UP trial used a visual assessment requiring the absence of marked FLAIR signal changes.

Results:

During the study period, 136 stroke patients presented and were imaged in the specified time window. Of these, 17 (12.5%) received IV tPA. Three patients had hemorrhage on 24-hour MRI follow up; none had an increase in NIHSS ≥4. Of the 119 patients who were screened but not treated, 18 (15%) were eligible based on FLAIR quantitative assessment and 55 (46%) were eligible based on qualitative assessment. In all cases where patients were not treated, there was an identifiable exclusion based on trial criteria. During the study period, IV tPA utilization was increased by 5.6% due to screening and treating patients with unknown onset stroke.

Conclusions:

Screening stroke patients in an unknown time window with MRI is practical in a real-world setting and increases IV tPA utilization.

Introduction

Two recent clinical trials, MR WITNESS and WAKE-UP, tested the practice of MRI-guided thrombolysis for patients presenting in an unknown time window and found it to be safe¹ and effective.² Specifically, these trials used fluid attenuated inversion recovery (FLAIR) MRI to identify patients who were likely to be in an early time window based on the lack of T2 signal change within the acute stroke which was identified on diffusion weighted imaging (DWI). The MR WITNESS trial used a quantitative assessment of FLAIR signal change by measuring a signal intensity ratio (SIR) relative to normal tissue.¹ The WAKE-UP trial used a qualitative FLAIR assessment in which there had to be an absence of marked FLAIR change on visual assessment.² Based on these trials, the updated guidelines now recommend the use of MRI guided thrombolysis in patients presenting in an unknown time window.³

The NIH stroke team has used MRI as its first-line imaging modality to triage stroke patients since the program was created 20 years ago. The MR WITNESS trial, in part, grew out of this environment based on observations about the dynamic changes on FLAIR imaging during the acute phase of ischemic stroke.⁴ When the MR WITNESS trial ended, revealing treatment of FLAIR-negative stroke to be safe, the NIH stroke team began treating this population off-label based on the criteria of the trial. Two years later, the WAKE-UP trial reported the practice to also be efficacious, and the NIH stroke team adopted this practice as standard of care.

The purpose of this study is to review the experience of the NIH stroke team in screening and treating unknown onset, FLAIR-negative stroke outside of a clinical trial to assess the feasibility and implications of this practice when in routine use.

Methods

Patient Population

This research was conducted as a retrospective analysis of de-identified registry data for which we obtained a determination of Not Human Subjects Research from the NIH Office of Human Subjects Research Protections (OHSRP).

The NIH stroke team, which runs the stroke services at MedStar Washington Hospital Center in Washington, DC, and Suburban Hospital in Bethesda, MD, uses MRI as the first-line imaging modality for triaging patients with suspected stroke. On 2/29/16, the NIH stroke team stopped enrolling in the MR WITNESS trial, which had closed, and began screening patients for treatment outside of the clinical trial based on the trial criteria. On 5/16/18, the results of the WAKE UP trial² were published, and the NIH stroke team adopted the practice of treating unknown onset stroke based on FLAIR-negative MRI as standard of care. This review of that practice was initiated on 8/31/19 and includes all patients screened for acute stroke from 2/29/16 up to that date. The use of FLAIR negative stroke to treat patients in an unknown time window was part of our standard treatment pathway for the duration of the study and did not use consent prior to treatment with thrombolysis, just as consent was not used prior to receiving thrombolysis in a known time window.

Imaging Protocol

Patients were scanned with one of two clinical 3T scanners at the hospital sites (Siemens Skyra, Siemens AG, Munich, Germany and Philips Achieva, Philips Healthcare, Best, the Netherlands) using established acute stroke imaging protocols.⁵ Typical scan time for an NIH hyper-acute MRI scan is 15 minutes.

Image Review

All images were independently reviewed by two raters (MA and ML). Discrepancies were adjudicated by a third author (RL). Using the DWI to identify the acute ischemic stroke, the corresponding region on the FLAIR image was categorized as either 1) definitely FLAIR negative, 2) equivocal, or 3) definitely FLAIR positive. For lesions classified as equivocal on qualitative assessment, a quantitative signal intensity ratio (SIR) was calculated on the FLAIR.¹

Patients were considered to have met the FLAIR criteria for the WAKE-UP trial if they were classified as definitely FLAIR negative or equivocal. Patients were considered to have met the imaging criteria for the MR WITNESS trial if they were definitely FLAIR negative or, in the case of patients who were equivocal, if their SIR was < 1.15 (the threshold used in the trial).⁴

In addition to classifying the FLAIR changes, the baseline MRI was also reviewed for the following based on available sequences: stroke location, vascular territory involved, presence of a PWI lesion, presence of a mismatch between the DWI and the PWI lesions, the presence of hyperintense vessels on FLAIR, the presence of a thrombus on GRE, the number and location of microbleeds on and the presence of a vessel cutoff on MRA. All patients who were treated with IV tPA had follow-up imaging at 24 hours after treatment. The follow-up was reviewed for the presence of hemorrhagic transformation (HT) according to the ECASS criteria.⁶ In patients where follow-up MRI was not available HCT was used.

Outcome measures and statistical analysis

Outcome measures were compiled for treated patients. Symptomatic intracranial hemorrhage (sICH) was defined based on the repeat imaging as a parenchymal hematoma type 2 (PH2) with an increase in the NIHSS of 4 or more points (SITS-MOST criteria).⁷ Good functional outcome was defined as a modified Rankin score (mRS) of 0 or 1 on follow up at either 1 or 3 months. If mRS was available at both 1 and 3 months, then the 3-month value was used. Patients who had a premorbid mRS>=2, prior to treatment, were excluded from functional outcome analysis but not from the assessment of sICH.

Univariate analyses were performed with Chi-square tests for categorical variables and t-tests for continuous variables to identify differences in study variables by using statistical software SAS 9.4 (SAS Institute, Cary, NC).

Results:

During the 2.5-year study period, 1,767 patients were evaluated and given a diagnosis of ischemic stroke. Of those, 201 did not have last known normal at the time of acute evaluation. Of the remaining 1566 patients, 674 (43%) presented within standard IV tPA window, and 605 (39%) were known to be out of the treatment window (>4.5 hour from symptom discovery). The remaining 287 (18%) patients presented <4.5 hours from symptom discovery but >4.5 hours from last seen normal. Of these, 136 patients had an MRI performed <4.5 hours from symptom discovery and were included in this study. None of the patients screened with MRI were taken for thrombectomy.

The demographics of the population are summarized in Table 1. The median age was 75 years, and 59% were female. The median pre-morbid mRS was 1, and the median NIHSS was 6. Of the 136 patients, 17 (12.5%) were treated with IV tPA.

Table 1:

Patient characteristics stratified according to treatment with intravenous tPA (n=136)

Characteristics ^	Treated with IV tPA (n= 17)	Not treated with IV tPA (n=119)	p-value
Age (yrs.)	77 [65-81]	74 [64-84]	0.3
Sex (Female)	9 (53%)	71 (60%)	0.6
Race (Black or African-American)	6 (35%)	63 (53%)	0.3
Admit NIHSS	8 [5-14]	6 [3-14]	0.5
Pre admit mRS	0 [0-1]	1 [0-3]	0.1
LSN-to-SD time, min	405 [165-510]	510 [360-720]	0.001
Triage-to-LSN time, min	523 [312-613]	642 [415-869]	0.001
Triage-to-MRI time, min	37 [30-46]	45 [34-88]	0.3
Triage-to-IV tPA time, min	72 [61-94]	NA
LSN-to-IV tPA time, min	576 [395-687]	NA
MRI-to-IV tPA time, min	34 [26-45]	NA
MR Imaging Variables
DWI lesion	17 (100%)	113 (95%)	0.6
PWI lesion	12 (70%)	56 (47%)	0.002
FLAIR changes
Negative	11 (64%)	10 (8.4%)	0.001
Equivocal	5 (29%)	45 (38%)
Positive	0 (0%)	57 (48%)
Hyperintense vessels on FLAIR	9 (53%)	47 (39.5%)	0.5
SIR for minimal FLAIR hyperintensity	1.10 [1.04-1.17]	1.24 [1.18-1.33]	0.02
GRE thrombus	8 (47%)	28 (23.5%)	0.08

^{^}Values reported as: median IQR [25-75] or n (%)

Abbreviations used: IQR: interquartile range, n: number, NIHSS: National Institutes of Health Stroke Scale, mRS: modified Rankin scale, LSN: last seen normal, SD: symptom discovery, MRI: magnetic resonance imaging, tPA: tissue plasminogen activator, DWI: diffusion-weighed imaging, PWI: perfusion-weighted imaging, FLAIR: Fluid-attenuated inversion recovery, SIR: Signal Intensity Ratio, GRE: Gradient recalled echo.

Of the 17 patients who were treated, 14 were rated as definitely FLAIR negative, two were rated as equivocal, and one patient was rated as not having adequate quality FLAIR imaging. None of the 16 patients were rated as being definitely FLAIR positive, and thus all of these treatment cases would have met the imaging criteria for the WAKE-UP study. One of the two equivocal patients was found to have an SIR>1.15 and thus would not have been eligible for the MR WITNESS trial. Additionally, six patients would not have been eligible for WAKE-UP due to age >80 years and/or NIHSS >25, whereas three patients would not have been eligible for MR WITNESS due to age >85 years, NIHSS >25, and/or SIR >1.15.

The median (IQR) pretreatment NIHSS was 8 (5-14) and the post treatment NIHSS at 24-hours was 4 (2-9). There were 11 patients whose premorbid mRS was 0-1 and had available mRS scores at 1-3 months. Of these, five (45%) had a good functional outcome (mRS 0-1). There were three ICH events detected on the 24-hour follow up imaging: one HI-2 and two PH-2, none of which were symptomatic according to SITS-MOST criteria.⁷ However one patient with a PH-2 did have a worsening of NIHSS which increased by three points and thus met the criteria for sICH according to the WAKE-UP trial which considered any clinical worsening in the setting of PH-2 as sICH (complication rate 1/17=5.8%).

During the same 2.5-year time period that this study reviewed, 301 patients were treated with standard IV tPA (<4.5 hours from last known normal). Thus, by using MRI to screen and treat FLAIR-negative stroke, IV tPA utilization was increased by 5.6%. Treatment based on FLAIR-negative stroke occurred approximately once every other month.

Of the 119 patients who were screened but not treated, 19 were FLAIR-negative according to MR WITNESS criteria, and 55 were FLAIR-negative according to WAKE-UP criteria. Two patients did not have FLAIR imaging to rule out FLAIR positivity by either SIR or qualitative assessment. Of the 19 patients who were eligible based on MR WITNESS, 18 patients had an identifiable reason not to be treated which included age, NIHSS >25, non-disabling deficit, subacute infarct, and/or >10 microbleeds. There was only one patient who did not have an identifiable reason not to be treated based on the MR WITNESS trial. Of the 55 patients who were eligible based on WAKE-UP imaging criteria, 54 patients had an identifiable reason not to be treated which included age, NIHSS >25, non-disabling deficit, subacute infarct, and/or >10 microbleeds. There was one patient who did not have an identifiable reason not to be treated based on the WAKE-UP trial. The reasons for not receiving tPA in the untreated group are listed in Table 2.

Table 2:

Reasons for not receiving tPA in the untreated group (n=119)

Age > 80 years	47 (39%)
Admit NIHSS >25	5 (4%)
Non-disabling deficit	5 (4%)
Subacute infarct on FLAIR	7 (6%)
Microbleeds >10	7 (6%)
FLAIR SIR >1.15	40 (35%)
Bright FLAIR positive (SIR not measured)	57 (48%)
Recent surgery or serious trauma	6 (5%)
INR > 1.7	1 (0.8%)
Hemorrhage on initial imaging	2 (1.6%)
Other	10 (8%)

Values reported as: n (%)

Abbreviations used: NIHSS: National Institutes of Health Stroke Scale, FLAIR: Fluid-attenuated inversion recovery, SIR: Signal Intensity Ratio, INR: International normalization ratio.

Discussion

This study found that routine screening and treatment of patients presenting greater than 4.5 hours from last seen normal but less than 4.5 hours from symptom discovery based on FLAIR-negative MRI was feasible and resulted in an increase in tPA utilization of 5.6%. On average, over a two-month period, nine patients were screened, and one patient was treated. While this study was not powered to, or intended to, determine if treating FLAIR-negative stroke is safe and effective, the outcomes in this study were similar to those reported in the two studies that were designed to test these outcomes.^{1, 2}

When comparing the imaging criteria from the prior studies, the more inclusive criteria used to simulate the WAKE-UP study identified half of all patients screened to be FLAIR negative. This is in line with the published screening logs of the WAKE-UP trial which found the same ratio.⁸ However this increased inclusiveness did not result in much of an increase in eligibility in our study due to non-imaging exclusions, which could be due to chance given our small sample size. The WAKE-UP trial resulted in a larger increase in tPA utilization of 9%⁹ which may in part be due to less strict inclusion criteria. For the majority of the time period reviewed for this study, the WAKE-UP trial was ongoing and thus MR WITNESS criteria were being applied. Post-hoc, quantitative analysis of the WAKE-UP trial data demonstrated that the benefit of tPA decreased with increasing FLAIR signal change.¹⁰

During the study period, 18% of patients presenting with stroke had an unwitnessed onset but presented within 4.5 hours of symptom discovery. Half of these patients were taken to MRI to screen for treatment. While we did not determine why the other half were not screened, it is likely that they had non-imaging exclusions since it has been the policy of the NIH stroke team to proceed to MRI on any patient who might be eligible for MR WITNESS. These findings are in line with the published prevalence of WAKE-UP stroke of 14-27%.^{11, 12}

One barrier to the implementation of FLAIR-negative stroke treatment is the need for hyper-acute MRI. While hyper-acute MRI as first line for screening stroke patients has been in use by the NIH stroke team for almost two decades, this practice is not common in the US. However it can be performed within a time frame that adheres to the guidelines both in the US¹³ and in Europe.¹⁴ In our study, the median time from door-to-needle (DTN) and MRI-to-needle was 72 and 34 minutes respectively. Although patients were not consented prior to treatment, it was standard procedure to explain to the patients what the evidence was for treatment since this practice was outside of the published guidelines at the time. This may account for the slightly elevated DTN times. Additionally, the calculation of an SIR may also introduce a time delay. Given the results of the WAKE-UP trial, this may not be necessary for future cases. Extrapolation of our results to other centers is dependent on the availability of hyper-acute MRI.

Conclusion

In conclusion, adopting the validated practice of treating unknown onset stroke based on FLAIR negative MRI is possible and will increase tPA utilization.