Association between pre-treatment perfusion profile and cerebral edema after reperfusion therapies in ischemic stroke

Abstract

The relationship between reperfusion and edema is unclear, with experimental and clinical data yielding conflicting results. We investigated whether the extent of salvageable and irreversibly-injured tissue at baseline influenced the effect of therapeutic reperfusion on cerebral edema. In a pooled analysis of 415 patients with anterior circulation large vessel occlusion from the Tenecteplase-versus-Alteplase-before-Endovascular-Therapy-for-Ischemic-Stroke (EXTEND-IA TNK) part 1 and 2 trials, associations between core and mismatch volume on pre-treatment CT-Perfusion with cerebral edema at 24-hours, and their interactions with reperfusion were tested. Core volume was associated with increased edema (p < 0.001) with no significant interaction with reperfusion (p = 0.82). In comparison, a significant interaction between reperfusion and mismatch volume (p = 0.03) was observed: Mismatch volume was associated with increased edema in the absence of reperfusion (p = 0.009) but not with reperfusion (p = 0.27). When mismatch volume was dichotomized at the median (102 ml), reperfusion was associated with reduced edema in patients with large mismatch volume (p < 0.001) but not with smaller mismatch volume (p = 0.35). The effect of reperfusion on edema may be variable and dependent on the physiological state of the cerebral tissue. In patients with small to moderate ischemic core volume, the benefit of reperfusion in reducing edema is related to penumbral salvage.

Introduction

Cerebral edema is a devastating complication of large hemispheric infarction, associated with up to 80% mortality in untreated patients. ¹ Among patients receiving thrombolysis or endovascular thrombectomy (EVT), edema is consistently associated with poorer outcomes.^2–4 With the advent of EVT as the standard of care for large vessel occlusions (LVO), there is increasing interest in the pathophysiology of edema and potential treatments to mitigate secondary injury after reperfusion.

Recent clinical studies suggest therapeutic reperfusion reduces edema and the risk of malignant infarction. ² , ⁵ , ⁶ In contrast, preclinical studies demonstrated exacerbation of edema after reperfusion.^7–9 The conflicting results suggest there is a complex relationship between reperfusion and edema that may be subject to additional influences. Better understanding of the biology of edema development after reperfusion is crucial to design therapies to attenuate this complication.

The volume of hypoperfused but viable ischemic penumbra is strongly associated with the clinical and imaging response to reperfusion. ¹⁰ , ¹¹ How salvageable and irreversibly-injured tissue influence edema development after therapeutic reperfusion is unclear. This study aimed to evaluate the association of pre-treatment estimates of ischemic core volume and penumbral mismatch volume, and their interaction with reperfusion success. We hypothesized that the extent of irreversible injury and the availability of penumbral tissue to be salvaged would modulate the effect of reperfusion on cerebral edema development after treatment. Specifically, our first hypothesis was that core volume is associated with edema development, regardless of reperfusion status. Secondly, we hypothesized that penumbral mismatch volume is associated with edema in the absence of reperfusion, and that the effect of reperfusion to reduce edema is dependent on the volume of salvageable penumbral mismatch.

Material and methods

Study design and inclusion criteria

We performed a post-hoc pooled exploratory analysis of the Tenecteplase versus Alteplase before Endovascular Therapy for Ischemic Stroke (EXTEND-IA TNK) part 1 (2015–2017) and part 2 (2017-2019) randomized trials. ¹² , ¹³ We restricted the analysis to the cohort of patients with anterior circulation strokes who had baseline CT-perfusion imaging and 24-hour follow-up CT or MRI. Patients with parenchymal hematoma, pre-existing cerebral structural pathology, bilateral infarcts or those who had neurosurgical procedures prior to 24-hours follow-up scan were excluded.

The EXTEND-IA TNK studies were prospective, randomized, multicenter clinical trials comparing intravenous thrombolytic agent before EVT (alteplase 0.9 mg/kg vs tenecteplase 0.25 mg/kg in Part 1; tenecteplase 0.25 mg/kg vs. tenecteplase 0.40 mg/kg in Part 2) in 502 adult patients with acute proximal LVO (intracranial internal carotid artery, middle cerebral artery first or second segments, or basilar artery) presenting within 4.5 hours after symptom onset. ¹² , ¹³ There was no restriction on clinical severity assessed using National Institutes of Health Stroke Scale (NIHSS) scores (range, 0 [no deficit] to 42 [death]). Participants with severe premorbid disability, defined as a modified Rankin Scale (mRS) score (range, 0 [normal] to 6 [death]) greater than 3, were excluded. Patients with extensive unenhanced CT hypodensity (>one-third of the middle cerebral artery) were excluded as per standard practice. Detailed inclusion and exclusion criteria have been reported previously. ¹⁴ , ¹⁵ CT-Perfusion was performed as part of a routine stroke multimodal imaging protocol. The EXTEND-IA TNK part 1 trial originally required the presence of CT-Perfusion mismatch for enrolment but this inclusion criteria was removed after enrolling 80 patients, following the analysis of data from other trials that showed benefit of thrombectomy in patients with larger ischemic-core volumes. ¹⁶

Standard protocol approvals, registrations, and patient consents

The studies were conducted in accordance with the Helsinki declaration, the National Health and Medical Research Council National Statement on Ethical Conduct in Research Involving Humans (2007) and the Notes for Guidance on Good Clinical Practice and was approved by respective ethics committees of 13 recruiting sites (Royal Melbourne Hospital, Royal Adelaide Hospital, Box Hill Hospital, Austin Hospital, Western Hospital, Gold Coast university Hospital, John Hunter Hospital, Princess Alexandra Hospital, Royal Brisbane and Women’s Hospital, Monash Medical Centre, Lyell McEwin Hospital, Royal North Shore Hospital, Christchurch Hospital). Written informed consent was obtained from the participant or a legal representative before enrollment, although emergency treatment followed by consent for continued participation was allowed in some jurisdictions. Both studies were prospectively registered with ClinicalTrials.gov (EXTEND-IA TNK part 1, NCT02388061; part 2, NCT03340493).

Imaging analysis and outcomes

Baseline CT-Perfusion was processed using commercially available fully-automated software (RAPID, iSchemaView) with post-processing removal of artifacts performed by two stroke neurologists (BCVC, FN) to obtain core and mismatch volumes. Core volume was expressed in milliliters (ml) and defined as relative cerebral blood flow less than 30% of that in normal tissue. ¹⁷ Mismatch volume was defined as the region with time to maximum (Tmax) delay more than 6 seconds not included in the core region. ¹⁸ Reperfusion was defined as centrally adjudicated extended Treatment in Cerebral Ischemia (eTICI) score of 2b to 3 on an angiogram obtained at the conclusion of the EVT procedure. ¹⁹ If intracranial angiography was not obtained, reperfusion was assessed as the restoration of blood flow to greater than 50% of the involved territory (ie. reduction of >50% of the pre-treatment baseline Tmax > 6-seconds lesion volume) on a post-thrombolysis CT-Perfusion 1 to 2 hours after thrombolysis. ¹² , ¹³ Follow-up ischemic lesion volume was defined by manual segmentation on DWI and CT at 24-hours.

Cerebral edema was assessed as anatomical distortion from tissue expansion (ie. swelling) measured in midline shift (MLS) in millimeters, defined as the maximal deviation of midline cerebral structures in the axial plane on follow-up CT or MRI performed at 24-hours ²⁰ and trichotomized as either negligible (<1mm), mild (≥1 to <5mm) or severe (≥5mm) for analysis as MLS was not normally distributed (Shapiro-Wilk test of normality, p < 0.001). (Figure 1) MLS measurements were measured by a neurologist (FN) blinded to clinical data and independently assessed in 10% of the study cohort (n = 42 consecutive subjects) by a second neurologist (BCVC). Interrater agreement was excellent (intraclass absolute correlation = 0.97, 95% CI 0.95–0.98). Hemorrhagic transformation was scored and adjudicated centrally according to the European Cooperative Acute Stroke Study II classification. ²¹ The primary analysis included patients with Hemorrhagic Infarction Type 1 and 2 as these do not contribute to mass effect but excluded parenchymal hematoma. ²¹

Figure 1.

Cerebral edema assessment.

Cerebral edema was measured as midline shift in millimetres at 24hours and analysed on a trichotomized ordinal scale of negligible (<1mm), mild (1–5mm) or severe (>5mm) as midline shift measurements were not normally distributed.

Statistical analysis

To investigate the associations between core volume and cerebral edema, multivariable ordinal logistic regression models were created with (1) cerebral edema as the dependent variable, core volume (in 10 ml units) as the independent variable, and (2) age, reperfusion status (with or without reperfusion), site of vessel occlusion (Internal Carotid Artery or Middle Cerebral Artery) and mismatch volume as adjustment covariates. The proportional odds assumption was tested using the Brandt test.

To investigate whether there was a core volume-by-reperfusion status interaction with cerebral edema, multivariable ordinal logistic regression models were created with (1) cerebral edema as the dependent variable, (2) age, reperfusion, site of occlusion and mismatch volume as adjustment covariates, and (3) core volume-by-reperfusion status multiplicative interaction term. We report the p-values of the interaction term, and the effect sizes (common odds ratios, with 95% confidence intervals) of core volume on edema in patients with and without reperfusion.

Similar analyses were performed for mismatch volume with adjustment for core volume. In addition, we tested the interaction between reperfusion and mismatch volume as a dichotomised variable (>median vs. ≤median) and reported the effect sizes of reperfusion on edema in patients with large or small mismatch volume.

Pre-specified sensitivity analyses were performed by adjusting for follow-up imaging modality (CT vs. MRI), intravenous thrombolysis agent (alteplase vs tenecteplase), additional inclusion of patients with parenchymal hematoma, modality of reperfusion assessment (angiographic eTICI vs post-thrombolysis CT-Perfusion), the use of thrombectomy (patients who underwent thrombectomy after thrombolysis vs patients who had thrombolysis only) time from onset-to-thrombolysis, time from thrombolysis-to-groin puncture and history of diabetes mellitus in ordinal regression models.

We investigated the robustness of our analyses using a dichotomized measure of cerebral edema (Present = MLS ≥1mm vs Absent = MLS <1mm) in binary logistic regression modelling. ²

Statistical analyses were performed with IBM SPSS Statistics, version 26 software (SPSS Inc), and 2-sided p-values less than 0.05 were considered statistically significant.

Data availability policy

Anonymized study data will be shared on request to the corresponding author by a qualified investigator.

Results

Among 502 patients enrolled in EXTEND-IA TNK Part 1 and Part 2 studies, 415 patients were included in the analysis after excluding 87 patients (20 with basilar occlusion, 21 with parenchymal hematoma on follow-up imaging, 45 without baseline CT-Perfusion and 4 without 24-hour imaging due to 2 early deaths and 2 neurosurgical procedures, 1 pre-existing arachnoid cyst and 1 with bilateral acute strokes). Decompressive craniectomy was performed in two patients after 24-hours. Majority of patients presented with small to moderate core volume (median core volume = 12 ml; IQR 0–33) (Table 1). Twenty-four (5.8%) patients presented with a core volume exceeding 80 ml. The median mismatch volume was 102 ml (IQR 67–141). Reperfusion status was assessed by post-treatment CT-perfusion in 22 patients (5.3%, median time from thrombolysis = 182 minutes [IQR 12–219 minutes]) who did not proceed to invasive angiography and thrombectomy because of substantial neurological improvement or distal migration of retrievable thrombus after thrombolysis. Overall, >50% reperfusion was achieved in 347 patients (83.6%). Follow-up ischemic lesion volume (median 17.6 ml, IQR 4.2–55.6) significantly correlated with cerebral edema (Pearson correlation 0.638, p < 0.001). Cerebral edema was independently associated with unfavorable functional outcome at 3-months (modified Rankin Scale 3–6; aOR 1.35 per mm of MLS, 95%CI 1.44–3.07, p < 0.001) after adjusting for age, clinical severity (NIHSS), core volume, and reperfusion.

Table 1.

Patient Characteristics of study cohort stratified by cerebral edema severity grade.

Characteristics	Edema assessed by midline shift			P value
	Negligible (<1mm)(n = 222)	Mild (1–5mm) (n = 170)	Severe (>5mm) (n = 23)
Age, median (IQR), yr	74 (64.25–82)	72 (64–80)	71 (54–82)	0.40
Male, No. (%)	116 (52.2)	90 (52.9)	14 (60.9)	0.74
Admission NIHSS, median (IQR)	14 (9–19)	18 (14–22)	20 (14–22)	<0.001
Time from onset to presentation, median (IQR), min	69.5 (45.5–83.5)	75 (44–106)	65 (39–91)	0.24
Baseline core volume, median, (IQR), ml	6 (0–23.75)	17.5 (6–38)	47 (7–90)	<0.001
Baseline mismatch volume, median, (IQR), ml	101 (61–140)	102 (73.5–142.75)	107 (78–139)	0.30
Occlusion location, No. (%)				0.18
Internal Carotid Artery	41 (18.5)	34 (20)	8 (34.8)
Middle Cerebral Artery (M1 or M2)	181 (81.5)	136 (80)	15 (65.2)
Time from onset to thrombolysis, median (IQR), min	128 (101–170)	136 (108–169)	142 (113–173)	0.33
Endovascular treatment, No. (%)	180 (81.1)	151 (88.8)	17 (73.9)	0.05
Time from lysis to groin puncture, median (IQR), min	46 (24–66)	45 (28–72)	57 (38–127)	0.17
Successful reperfusion, No. (%)	197 (88.7%)	134 (78.8)	16 (69.6)	0.005
MRI as follow-up neuroimaging modality, No. (%)	150 (67.6%)	93 (57.1)	14 (60.9)	0.03
Midline Shift, median, (IQR), mm	0 (0–0)	2.3 (1.6–3.0)	6.6 (5.5–8.8)	<0.001
Infarct Growth volume, median, (IQR), ml	1.3 (0–12.3)	14.1 (1.2–48.5)	130 (66.7–198.7)	<0.001
Follow-up ischemic lesion volume, median, (IQR), ml	8.8 (1.3–26.2)	33.8 (13.6–87.2)	195.5 (107.7–292.6)	<0.001
Functional outcome (mRS), median, (IQR)	1 (0–3)	2 (1–4)	5 (3–6)	<0.001
Favorable outcome (mRS 0–2), No. (%)	162 (73.0)	86 (50.6)	3 (13.0)	<0.001

NIHSS: National Institutes of Health Stroke Scale; M1: middle cerebral artery segment 1; M2: middle cerebral artery segment 2; mRS: Modified Rankin Scale.

In multivariable analysis adjusting for age, reperfusion status, site of occlusion and mismatch volume, larger core volume was significantly associated with increased cerebral edema (adjusted cOR = 1.27 per 10 mL, 95%CI 1.18–1.37; p < 0.001). There was no significant interaction between reperfusion status and core volume on cerebral edema (p = 0.82). (Table 2)

Table 2.

Association of core and mismatch volume with increased cerebral edema stratified by reperfusion status.

Variable	Patient with reperfusion (n = 347)		Patient without reperfusion(n = 68)		p-value for interaction
	Adjusted cOR (95% CI)	P value	Adjusted cOR (95% CI)	P value
Core volume (per 10 ml)a	1.26 (1.17–1.37)	<0.001	1.30 (1.09–1.55)	0.003	0.82
Mismatch volume (per 10 ml)b	1.02 (0.98–1.07)	0.27	1.13 (1.03–1.25)	0.009	0.03

Note: The associations of core volume and mismatch volume with cerebral edema were tested in patients with and without successful reperfusion (>50% reperfusion with eTICI 2 b-3) using multivariable ordinal regression modelling with cerebral edema severity as the outcome. In patients with successful reperfusion, Core Volume (cOR = 1.26, p < 0.001) but not Mismatch Volume (cOR = 1.02, p = 0.27) was associated with cerebral edema. In patients without reperfusion, both Core (cOR = 1.30, p = 0.003) and Mismatch Volume (cOR = 1.13, p = 0.009) were associated with cerebral edema. The results were concordant when defining reperfusion as eTICI 2c-3 (Supplemental material, Table 1).

^aCore volume wad adjusted for mismatch volume, age and occlusion location.

#Mismatch volume was adjusted for core volume, age and occlusion location.

ml: milliliters; ICA: internal carotid artery; MCA: middle cerebral artery.

Increased mismatch volume was also associated with more cerebral edema overall (cOR = 1.04, 95% CI 1.00–1.08; p = 0.03). However, a significant interaction between reperfusion and mismatch volume (p = 0.03) was observed. Mismatch volume was significantly associated with increased cerebral edema only in patients without reperfusion (n = 68; cOR = 1.13 per 10 mL, 95% CI 1.03-1.25; p = 0.009), and not in patients with reperfusion (n = 347; cOR = 1.02 per 10 mL, 95%CI 0.98–1.07; p = 0.27). (Table 2) (Figure 2)

Figure 2.

Example of patients with and without reperfusion.

Mismatch volume was associated with increased cerebral edema only in the absence of reperfusion. Mismatch volume at baseline was determined on baseline CT-Perfusion before treatment and defined as volume of the perfusion lesion (cerebral region with time to maximum [Tmax] delay more than 6 second, shown as green, middle two columns) that was not included in the core region (Cerebral region with relative cerebral blood flow less than 30% of that in normal tissue, shown as purple, left two columns). The top role (a) depicts a patient with a right middle cerebral artery (M1) occlusion with large mismatch volume (core volume = 0ml, Tmax>6 perfusion lesion = 228ml, mismatch volume = 228ml) without reperfusion who developed severe edema on follow-up. The bottom row (b) depicts a patient with a left middle cerebral artery (M1) occlusion with large mismatch volume (core volume = 11ml, Tmax>6 perfusion lesion = 160ml, mismatch volume = 149ml) with reperfusion who had negligible edema on follow-up .

This interaction remained significant (p = 0.010) using a dichotomized mismatch volume term (larger defined as mismatch volume > than median [102 ml]). In those with large mismatch volume, reperfusion was associated with reduced cerebral edema (cOR = 0.19, 95%CI 0.08–0.43; p < 0.001). This association was not significant in those with small mismatch volume (cOR = 0.72, 95% CI 0.36–1.45; p = 0.35). (Figure 3)

Figure 3.

Example of reperfused patients who had large or small mismatch volume at baseline.

Reperfusion was only associated with less cerebral edema when there was a large mismatch volume at baseline. Mismatch volume at baseline was determined on baseline CT-Perfusion before treatment as previously described (ie. volume of the perfusion lesion (green, middle two columns) that was not included in the core region (purple, left two columns). The top role (a) depicts a patient described with a right middle cerebral artery (M1) occlusion with large mismatch volume (core volume = 5ml, Tmax>6 perfusion lesion = 149ml, mismatch volume = 144ml) with reperfusion who had negligible edema. The bottom row (b) depicts a patient with a right middle cerebral artery (M1) occlusion with small mismatch volume (core volume = 124ml, Tmax>6 perfusion lesion = 151ml, mismatch volume = 27ml) with reperfusion who developed severe edema.

Pre-specified sensitivity and robustness analyses yielded similar results (Tables 3 and 4).

Table 3.

Sensitivity analyses for parenchymal hematoma, imaging modalities, thrombolysis agent and treatment time metrics.

Variable	Common Odds Ratio	95% confidence interval	P value
Mismatch volume (per 10 ml)	1.04	(1.002–1.08)	0.039
Core volume (per 10 ml)	1.28	(1.19–1.38)	<0.001
Reperfusion	0.34	(0.19–0.59)	<0.001

Note: Pre-specified sensitivity analyses (n = 430) were performed by adjusting for parenchymal hematoma, follow-up imaging modality (CT vs. MRI), intravenous thrombolysis agent and dosage (Alteplase vs Tenecteplase 0.25 mg/kg vs Tenecteplase 0.4 mg/kg), modality of reperfusion assessment (angiographic eTICI vs post-thrombolysis CT-Perfusion), the use of thrombectomy (patients who underwent thrombectomy after thrombolysis vs patients who had thrombolysis only) time from onset to thrombolysis, time from thrombolysis to groin puncture and history of diabetes mellitus in addition to adjustment covariates of the primary analysis (ie. age and occlusion location) in ordinal regression models with midline shift as the outcome. Mismatch (p = 0.04), core volume (p < 0.001) and reperfusion (p = 0.001) were associated with cerebral edema in concordance with the primary analysis. A significant interaction effect between reperfusion and mismatch volume (p = 0.018) but not core volume (p = 0.58) was also observed in the sensitivity analysis multivariable model. An additional sensitivity analysis adjusting for onset-to-puncture time also yielded similar results (Supplemental materials, Table 2).

Table 4.

Robustness analysis: cerebral edema as a dichotomized variable.

Variable	Adjusted odds ratio	95% confidence interval	P value
Mismatch volume (per 10 ml)	1.05	(1.01–1.09)	0.02
Core volume (per 10 ml)	1.26	(1.16–1.37)	<0.001
Reperfusion	0.40	(0.23–0.71)	0.002

The primary analysis was repeated with cerebral edema converted to a dichotomized outcome variable (Present = MLS ≥1mm vs Absent = MLS <1mm) adjusting for age and occlusion location in logistic regression modelling. Core volume (p < 0.001), reperfusion (p = 0.002) and mismatch (p = 0.02) were associated with cerebral in concordance with the primary analysis. Similarly, a significant interaction effect between reperfusion and mismatch volume (p = 0.003) but not core volume (p = 0.28) was observed.

Discussion

This study has identified several important relationships between pre-treatment brain tissue viability, reperfusion and the development of cerebral edema after ischemic stroke. First, we confirmed that baseline ischemic core volume is consistently associated with edema development, regardless of reperfusion status. More importantly, we found an interaction between mismatch volume and reperfusion in its association with cerebral edema. Mismatch volume was associated with increased cerebral edema only in the absence of reperfusion. Reperfusion was only associated with less cerebral edema when there was a large mismatch volume at baseline. Specifically, reperfusion was not associated with less cerebral edema when there was minimal mismatch volume to salvage. Our findings suggest reperfusion does not universally reduce edema. Instead, the presence of sufficient salvageable penumbral tissue is required in order for reperfusion to reduce edema and prevent secondary injury from tissue swelling.^2–5 These results also have potential clinical implications. Measurement of ischemic core and mismatch volume could provide additional prognostic information regarding likely edema development and help stratify high risk patients who may require secondary intervention for edema, especially when reperfusion cannot be achieved.

Previous studies have showed younger age,^22–24 higher NIHSS,^23–25 more proximal and extensive arterial occlusion, ²⁶ , ²⁷ hyperglycemia, ²⁸ , ²⁹ and larger admission infarct volume assessed on DWI^24,27,30 or non-enhanced CT ³¹ , ³² to be risk factors for cerebral edema after ischemic stroke. More severe hypoperfusion within the ischemic lesion and a higher core volume defined on CT-Perfusion Cerebral Blood Volume-based thresholds have also been associated with increased edema, ²³ , ³¹ , ³³ consistent with the general principle that more severe stroke is associated with greater edema. ⁶ However, to our knowledge, the interplay between core and mismatch volume with reperfusion in edema development has not been investigated previously.

Our data provide novel insights into how salvageable penumbral tissue and irreversibly-injured tissue may respond differently to reperfusion in terms of cerebral edema development. In the absence of reperfusion, penumbral tissue becomes unsalvageable leading to a larger infarct with increased edema. Conversely, penumbral salvage led to reduced edema. These findings are consistent with recent clinical data that showed reduced edema measured by midline shift in patients with reperfusion. ² Our data further suggest the benefit of reperfusion in reducing cerebral edema is mediated through penumbral salvage and dependent on the presence of a large mismatch at baseline.

In comparison, we did not observe reperfusion to attenuate swelling in irreversibly-injured tissue. In other words, irreversibly-injured tissue was associated with increased edema, regardless whether or not the tissue was reperfused. This is in contrast to a recent retrospective study that suggested endovascular reperfusion may reduce edema within the core lesion when using Net Water Uptake as an alternative measure of cerebral edema. ⁵ Net Water Uptake, a recently proposed radiological marker of edema at the cellular level, is based on the voxel parameters of the ischemic tissue on unenhanced CT in which a lower Hounsfield Unit (ie. darker) indicates more tissue edema. In this study, patients with successful angiographic reperfusion had higher Hounsfield Unit within the infarct lesion, which implies patients who had reperfusion also had less cerebral edema. Of note, the study did not account for the confounding effects of post-angiographic iodine contrast staining on post-treatment unenhanced CT which can occur in nearly 50% of patients. ³⁴

Overall, our findings are consistent with earlier observations from animal studies that showed that reperfusion reduced edema if the ischemic insult was shorter and less severe (likely with ischemic tissue being salvageable) but did not reduce edema if the initial ischemia was prolonged and more severe (likely with cerebral tissue becoming irreversibly-injured by the time of reperfusion).^35,36 We did not observe an interaction between reperfusion and core volume in our analysis but our study population predominantly consisted of patients presenting with small to moderate core volume. As such, our study only assessed the impact of large versus small mismatch volume on the effect of reperfusion on edema development, but was under-powered to evaluate how the amount of core volume modulates the effect of reperfusion. Experimental data have suggested that reperfusion may be paradoxically detrimental by increasing edema through futile reperfusion into established infarcts.^7–9 This notion is supported by clinical data from a recent meta-analysis of the HERMES thrombectomy studies showing that reperfusion exacerbates midline shift in patients with very high pre-treatment baseline core volume (>130ml). ³⁷

The concept of reperfusion having a variable effect on edema depending on the physiological state of the parenchyma being reperfused provides a plausible biological rationale for the unexplained discrepancy described between preclinical and clinical data. The earliest data on cerebral edema after reperfusion conducted in animal models have repeatedly shown reperfusion to increase edema. In a rat model of transient and permanent MCA occlusion by Yang and Betz, animals that underwent reperfusion had significantly greater water content per volume of tissue and larger infarcts compared to animals without reperfusion. ⁸ This was also observed in primate models using baboons in which reperfusion exacerbated water accumulation in cerebral tissue in proportion to the extent of the reperfusion after temporary MCA surgical clipping. ³⁶ In contrast, clinical studies in humans have generally demonstrated that therapeutic reperfusion by thrombolysis and EVT reduces edema. In a post-hoc analysis of the MR CLEAN prospective randomized controlled trial, Kimberly et al showed successful reperfusion was associated with a reduced likelihood of developing MLS at 24 hours and 5–7 days. ² Similarly, Thorén et al. showed recanalization by intravenous thrombolysis with or without EVT was associated with a lower risk of parenchymal swelling in the SITS-International Stroke Treatment Registry. ³ The reason for this discrepancy may relate to the amount of salvageable penumbral tissue in experimental versus human studies. By study design, prospective clinical studies are orientated to identifying treatment benefits among selected cohorts of patients who are likely to respond to intervention. Similarly, retrospective clinical data are predominantly derived from patients who are eligible for reperfusion therapies based on clinical guidelines. Human clinical studies may therefore be inadvertently biased towards including patients with earlier presentation, more favorable perfusion status and greater volume of penumbral tissue to mediate the benefit of reperfusion in reducing cerebral edema. In comparison, animal MCA occlusion models allow evaluation of the cerebral edema response to reperfusion in pathophysiological scenario which would otherwise be excluded from clinical trials or deemed ineligible for reperfusion treatment in routine clinical practice. In models where the induced transient ischemia is sufficiently severe to generate infarction of the majority of the hemisphere before restoration of cerebral blood flow, penumbral salvage is minimal and reperfusion occurs predominantly into irreversibly-injured tissue. In such circumstances, reperfusion does not attenuate cerebral edema development, analogous to patients with small mismatch volume at baseline who had reperfusion in our study cohort. Accordingly, the varying findings from clinical and preclinical studies may be considered as being complementary, rather than contradictory, in illustrating how cerebral edema develops in response to reperfusion across the entire spectrum of pathophysiological scenarios. Given the importance of baseline tissue viability in modulating the effect of reperfusion, future studies on cerebral edema should incorporate penumbra and core volume evaluation as important determinants and treatment modifiers.

Limitations

The study has several limitations. All patients presented within 4.5 hrs of symptoms onset and our findings may not be generalizable to later-presenting patients with more established ischemic changes. The early time frame may have contributed to the relatively low baseline core volume in our cohort. This restricted our ability to analyze the impact of core volume across its entire range or the effect of mismatch to core ratio. Dedicated studies in patients with large core will be needed to examine the relationship between reperfusion and cerebral edema when substantial established ischemic change is already present. Second, MLS was used as an indirect marker of cerebral edema and may not reflect the influence of more localized edema, particularly in smaller cortical regions, and the effect of pre-existing cerebral atrophy. In a further sensitivity analysis testing whether core volume modulates the relationship between mismatch volume, reperfusion, and cerebral edema, we did not find core volume to interact with these associations (Supplementary material, Tables 3 and 4). Age was included in all analyses as a surrogate marker of cerebral atrophy. Measuring MLS at 24-hours may not capture the maximum extent of cerebral edema which generally peaks at 72 hours, especially in patients with persistent occlusion who may have ongoing infarct growth beyond 24-hours. Nonetheless, we demonstrated MLS assessment at 24-hours to be an independent prognostic marker of poor clinical outcome at 3 months in addition to age, baseline clinical and radiological severity and reperfusion. The overall distribution of MLS at 24-hours in our study population is consistent with a recent thrombectomy randomized controlled trial of patients presenting within 6 hours of onset where approximately half of patients did not have measurable MLS on follow-up. ² In this study, analysis using MLS measured at 24-hours and at 5-7 days yielded the same results. Third, we did not assess the role of follow-up ischemic lesion volume in cerebral edema in our analysis as the diffusion MRI lesion and hypodensity on CT incorporate both infarction and edema volumes. Advanced imaging analysis by coregistration using high-resolution structural T1 imaging and pre-treatment DWI to quantify edematous anatomical distortion as distinct from infarction is required to further examine this complex relationship in future studies. ³⁸ , ³⁹ Finally, the thrombolytic agent and dose varied due to the design of the 2 EXTEND-IA TNK studies. However, sensitivity analysis adjusting for reperfusion modalities and treatment time metrics yielded similar findings.

Conclusion

The volume of irreversibly injured ischemic core was consistently associated with increased cerebral edema, regardless of reperfusion status. In comparison, the volume of salvageable tissue was only associated with increased edema in the absence of reperfusion. Reperfusion was associated with reduced edema only if there was substantial penumbral tissue to salvage. Collectively, our findings suggest the effect of reperfusion on edema development is not universal but may be variable and dependent on the physiological state of the cerebral tissue. The variable effect of reperfusion on cerebral edema development may explain the apparent discrepancy between preclinical and clinical data. Among patients with small to moderate core, the benefit of reperfusion in reducing edema is related to penumbral salvage. Future studies on cerebral edema should incorporate penumbra and core volume evaluation as important determinants and treatment modifiers.