Symmetric CTA Collaterals Identify Patients with Slow-progressing Stroke Likely to Benefit from Late Thrombectomy

Abstract

Background

Understanding ischemic core growth rate (IGR) is key in identifying patients with slow-progressing large vessel occlusion (LVO) stroke who may benefit from delayed endovascular thrombectomy (EVT).

Purpose

To evaluate whether symmetric collateral pattern at CT angiography (CTA) can help to identify patients with low IGR and small 24-hour diffusion-weighted MRI ischemic core volume in patients with LVO not treated with reperfusion therapies.

Materials and Methods

In this secondary analysis of clinical trial data from before EVT became standard of care from January 2007 to June 2009, patients with anterior proximal LVO not treated with reperfusion therapies were evaluated. All patients underwent admission CTA and at least three MRI examinations at four time points over 48 hours. Arterial phase CTA collaterals at presentation were categorized as symmetric, malignant, or other. Diffusion-weighted MRI ischemic core volume and IGR at multiple time points were determined. The IGR at presentation was defined as follows: (ischemic core volume in cubic centimeters)/(time since stroke symptom onset in hours). Multivariable analyses and receiver operator characteristic analyses were used.

Results

This study evaluated 31 patients (median age, 71 years; interquartile range, 61–81 years; 19 men) with median National Institutes of Health Stroke Scale (NIHSS) score of 13. Collaterals were symmetric (45%; 14 of 31), malignant (13%; four of 31), or other (42%; 13 of 31). Median ischemic core volume was different between collateral patterns at all time points. Presentation was as follows: symmetric, 16 cm³; other, 69 cm³; and malignant, 104 cm³ (P < .001). At 24 hours, median ischemic core volumes were as follows: symmetric, 28 cm³; other, 156 cm³; and malignant, 176 cm³ (P < .001). Median IGR was also different, and most pronounced at presentation: symmetric, 4 cm³ per hour; other, 17 cm³ per hour; and malignant, 20 cm³ per hour (P < .001). After multivariable adjustment, independent determinants of higher presentation IGR included only higher NIHSS (parameter estimate [β = 0.20; 95% CI: 0.05, 0.36; P = .008) and worse collaterals (β = –2.90; 95% CI: –4.31, –1.50; P < .001). The only independent determinant of 24-hour IGR was worse collaterals (β = –2.03; 95% CI: –3.28, –0.78; P = .001). Symmetric collaterals had sensitivity of 87% (13 of 15) and specificity of 94% (15 of 16) for 24-hour ischemic core volume less than 50 cm³ (area under the receiver operating characteristic curve, 0.92; 95% CI: 0.81, 1.00; P < .001).

Conclusion

In patients with large vessel occlusion not treated with reperfusion therapies, symmetric collateral pattern at CT angiography was common and highly specific for low ischemic core growth rate and small 24-hour ischemic core volume as assessed at diffusion-weighted MRI. After further outcome studies, collateral status at presentation may prove useful in triage for endovascular thrombectomy, especially when MRI and CT perfusion are unavailable.

Clinical trial registration no. NCT00414726.

© RSNA, 2021

Online supplemental material is available for this article.

See also the editorial by Messina in this issue.

Summary

Presentation collateral status at CT angiography was associated with ischemic core growth over 48 hours and may be useful in triage of patients with stroke for endovascular thrombectomy, especially when MRI and CT perfusion are unavailable.

Key Results

• ■ In this secondary analysis of 31 patients with large vessel occlusion stroke not treated with reperfusion therapies with at least three MRI examinations over 48 hours, median ischemic core volume and growth rate were significantly different in comparing collateral patterns at CT angiography at nearly all time points.

• ■ At diffusion-weighted MRI, median 24-hour ischemic core volumes were 28 cm³ (symmetric collaterals), 156 cm³ (other), and 176 cm³ (malignant).

• ■ Symmetric collaterals were common (45%) with high sensitivity (87%) and specificity (94%) for 24-hour ischemic core volume less than 50 cm³ (area under the receiver operating characteristic curve, 0.92; P < .001).

Introduction

The largest proportion of stroke-related death and disability is related to large vessel occlusion (LVO) ischemic stroke, which occurs because of embolization from a proximal source or in situ large vessel disease (1). Endovascular thrombectomy (EVT) has revolutionized the care of these patients (2–4), but treatment selection can be challenging especially when CT perfusion imaging and MRI are unavailable. This is a common scenario in community hospitals and underserved regions (5,6). Moreover, there are often delays in the transfer of patients to thrombectomy-capable centers, which can reduce the likelihood of EVT (7). Understanding the natural history of ischemic core growth is key to identification of patients with slow-progressing stroke that may benefit from EVT, even if delayed. Ischemic core growth rate (IGR) has recently been described as a determinant of clinical outcomes after EVT (8). Indeed, the presentation IGR is accurate in predicting 24-hour ischemic core volumes (9).

The IGR is likely highly dependent on the degree of collateral circulation, which is variable among patients with stroke (8). These alternative vessels, consisting of primary circle of Willis and secondary pial-pial leptomeningeal anastomoses, compensate for reduced blood flow in LVO (1). Where perfusion imaging and MRI are unavailable, the collateral pattern assessed with CT angiography (CTA) at presentation may be an appropriate proxy for ischemic core volume and IGR (10). We sought to use a unique data set of patients with LVO not treated with reperfusion therapies who were enrolled in a trial before modern EVT was available and who underwent admission CTA and serial MRIs at four time points during the 48 hours after presentation to characterize the natural history of ischemic core growth. Our aim was to evaluate whether a symmetric collateral pattern at CTA helped to identify patients with smaller ischemic core volumes, slower IGR, and 24-hour ischemic core volume less than 50 cm³ among patients with LVO not treated with reperfusion therapies.

Materials and Methods

Study Sample

This secondary analysis of clinical trial data was compliant with the Health Insurance Portability and Accountability Act and was approved by the local institutional review board. There was no industry support for this study. The data are from the phase 2 Normobaric Oxygen Therapy in Acute Ischemic Stroke Trial from 2007–2009 (see http://clinicaltrials.gov/show/NCT00414726 for the complete trial inclusion and exclusion criteria). Briefly, written informed consent was obtained for enrollment. The investigational treatment of normobaric oxygen had no significant effect on ischemic core growth, justifying data assimilation from active and placebo arms for our analysis. The original trial was terminated by the funding agency (National Institute of Neurological Disorders and Stroke) because of an imbalance in deaths noted by the data safety and monitoring board on the third interim analysis (P = .03, unadjusted). After the final statistical analysis and in-depth blinded chart review by the internal and external medical monitors, deaths were deemed unrelated to normobaric oxygen. The trial results have been presented as abstracts at the American Academy of Neurology and American Heart Association International Stroke Conference meetings (11) but remain unpublished because of the neutral results attributable to the small sample size from premature trial termination. There is no overlap between this study and previous work. Patients were included in this secondary analysis on the basis of the following additional criteria: MRI with diffusion-weighted imaging (DWI) that showed acute ischemic injury, presentation CTA of the head that showed a proximal anterior circulation LVO, and three or more MRI examinations performed within 48 hours. Patients were excluded if they did not have a presentation CTA available for review or three or more MRI examinations performed within 48 hours (Fig 1).

Figure 1:

Flowchart shows reasons patients were excluded from this study. CTA = CT angiography, MRA = MR angiography.

Age and sex were recorded at the time of enrollment (12). The presentation National Institutes of Health Stroke Scale (NIHSS) was determined at clinical examination by certified neurologists to measure clinical stroke severity (13). The time from last seen well and stroke onset was documented. For patients in whom stroke onset was not witnessed, onset time was estimated as midway between last seen well and first seen with stroke symptoms (14).

CTA Protocol

CTA was performed by using multidetector scanners (GE Medical Systems) from the vertex to the aortic arch following injection of 65–140 mL of a nonionic contrast agent (Isovue, Bracco Diagnostics) at a rate of 3–4 mL per second. The median parameters were 1.25-mm section thickness, 220-mm reconstruction diameter, 120 kV, and 657 mA. Volume CT dose index ranged from 65–95 mGy, and dose-length product ranged from 2593–3784 mGy · cm.

MRI and DWI

MRI scans were obtained at presentation and imaging was repeated approximately 4 hours, 24 hours, and 48 hours after onset by using a clinical 1.5-T system (GE Healthcare). DWI scans were acquired by using the following median values: field of view, 220 mm; 25 sections; section thickness, 5 mm; 1-mm gap; repetition time seconds/echo time msec, 5/85.3; 128 × 128 acquisition matrix; and b values, 0 sec/mm² and 1000 sec/mm² in at least six diffusion-gradient directions. Isotropic DWI and apparent diffusion coefficient maps were calculated by using techniques that were described previously (15). Fluid-attenuated inversion recovery imaging was performed with a fast-spin echo sequence, with the following median values: 10/145; inversion time msec, 2200; 256 × 192 matrix; field of view, 220 mm; and 25 5-mm sections with a 1-mm gap. Perfusion MRI data were acquired by using the following median values: field of view, 220 mm; 15 sections; section thickness, 5 mm; 1-mm gap; 1.5/40; flip angle, 60°; and 128 × 128 acquisition matrix. Mean transit time perfusion maps were calculated by using automated oscillation index regularized deconvolution (16).

Image Analysis

All imaging analyses were blinded to treatment assignment, time, and clinical information. Specifically, assessors of CTA collaterals were blinded to DWI. CTA image interpretations were performed independently by Certificate of Added Qualification–certified neuroradiologists (R.G.G. and M.H.L., each with more than 25 years of experience interpreting acute stroke studies). Vessel occlusion site on the CTA image was documented as internal carotid artery terminus, first middle cerebral artery segment (hereafter, referred to as M1), and second middle cerebral artery segment (hereafter, referred to as M2) (17). Collateral patterns were determined by visual review of the maximum intensity projection arterial phase CTA images, which were classified as symmetric (Fig 2A), malignant (Fig 2B), or other, consistent with previously published three-level categorizations (18). A symmetric pattern was defined as contrast enhancement viewed with similar or near-similar conspicuity (ie, no or minimally detectable reduction in opacification) of the ischemic compared with the contralateral nonischemic middle cerebral artery territory at risk, which was similar to definitions of several published CTA collateral scoring systems (10,19–21). A malignant pattern was defined as no contrast enhancement viewed over at least 50% of the middle cerebral artery territory at risk, also similar to earlier published definitions. (10,21) Other was defined as any additional pattern, rated as intermediate between symmetric and malignant. The collateral pattern was classified as symmetric or malignant if at least one reader gave one of these ratings and the second reader gave a rating of other (in cases of disagreement, occurred in only two cases). Ischemic cores with reduced diffusivity were outlined visually by using both the DWI and apparent diffusion coefficient maps from the same time point. Mean transit time abnormalities were outlined visually with knowledge of DWI, apparent diffusion coefficient, and CT perfusion time-to-maximum maps. All outlines were performed by experienced readers with semiautomated open-source software (Display, version 2.0; Montreal Neurologic Institute). The IGR was defined as the difference in DWI ischemic core volume divided by the time in hours from last seen well or from the previous measurement (9). Accordingly, IGR at presentation was calculated as follows: (presentation DWI ischemic core volume in cubic centimeters)/(time in hours since stroke symptom onset).

Figure 2:

Collateral patterns at presentation CT angiography (CTA) and corresponding ischemic cores at presentation and 24-hour diffusion-weighted imaging (DWI). (A) Examples of patients with symmetric collaterals. (B) Examples of patients with malignant collaterals. Patient 1 is an 84-year-old woman with National Institutes of Health Stroke Scale (NIHSS) score of 7 and symptom onset 4.1 hours prior. Patient 2 is a 70-year-old woman with NIHSS score of 7 and symptom onset 5.6 hours prior. Patient 3 is a 71-year-old man with NIHSS score of 21 and symptom onset 10.8 hours prior. Patient 4 is an 85-year-old man with NIHSS score of 6 and symptom onset 2.5 hours prior.

Statistical Analysis

Median and interquartile range were reported for continuous nonparametric variables. Percentage and count were reported for categorical variables. Differences among three groups of nonparametric continuous variables were assessed by using the Kruskal-Wallis test. Associations with nonparametric continuous variables were assessed with ordinal regression. Variables of interest were selected a priori for their possible relevance to each of the three outcomes modeled. Those with prespecified significance of P value less than .10 at univariable analyses were subsequently included in multivariable models. Each regression model focused on a single point. Receiver operator characteristic curve analyses were performed to evaluate sensitivity and specificity. The distribution was assumed nonparametric on the basis of the Kolmogorov-Smirnov and Shapiro-Wilk tests. Two-tailed P values less than .05 were considered to indicate statistical significance. Analyses were performed with software (Prism, version 6.01, GraphPad; and SPSS, version 23.0, IBM).

Results

Characteristics of Study Sample

Among the 60 participants who underwent serial MRI in the clinical trial, 31 patients (median age, 71 years; interquartile range, 61–81 years; 19 men) met inclusion criteria for this analysis (Fig 1). Normobaric oxygen was administered to 45% (14 of 31) of patients in the study sample as part of the previously mentioned clinical trial, which was not associated with presentation collaterals or ischemic core volume at any time point. The median presentation NIHSS score was 13 and median onset-to-MRI time was 5.0 hours. Vessel occlusion sites included internal carotid artery (23%), M1 (42%), and M2 (36%). Collaterals were malignant (13%), other (42%), or symmetric (45%) (Table 1). The only classification disagreement between raters was for symmetric versus other, which occured in two cases; in both cases, a consensus was reached for symmetric. At presentation, the median DWI ischemic core volume was 46 cm³, mean transit time volume was 145 cm³, and mean transit time DWI “mismatch” volume was 60 cm³. The median presentation IGR was 8 cm³ per hour. At 4 hours, the median ischemic core volume was 40 cm³ and IGR was 0 cm³ per hour. At 24 hours, the median ischemic core volume was 52 cm³ and IGR was 1 cm³ per hour. At 48 hours, the median ischemic core volume was 80 cm³ and IGR was 1 cm³ per hour. The median 90-day fluid-attenuated inversion recovery lesion volume was 52 cm³ (Table 1).

Table 1:

Study Sample Characteristics

Ischemic Core Volumes

In comparing collateral patterns at each time point, at least one ischemic core volume was found to differ significantly from the others. At presentation, median ischemic core volume was 16 cm³ in patients with symmetric collaterals, 69 cm³ in patients with other collaterals, and 104 cm³ in patients with malignant collaterals (P < .001). At 4 hours, median ischemic core volume was 19 cm³ in patients with symmetric collaterals, 95 cm³ in patients with other collaterals, and 127 cm³ in patients with malignant collaterals (P < .001). At 24 hours, median ischemic core volume was 28 cm³ in patients with symmetric collaterals, 156 cm³ in patients with other collaterals, and 176 cm³ in patients with malignant collaterals (P < .001). At 48 hours, median ischemic core volume was 36 cm³ in patients with symmetric collaterals, 183 cm³ in patients with other collaterals, and 281 cm³ in patients with malignant collaterals (P = .002) (Fig 3A).

Figure 3:

Collateral patterns at presentation CT angiography (CTA) have different ischemic core volumes and ischemic core growth rates at multiple time points. (A) Box-and-whisker plots with Kruskal Wallis test show that at presentation, median ischemic core volume was 16 cm³ (interquartile range, 3–27 cm³) for symmetric collaterals, 69 cm³ (interquartile range, 62–110) for other collaterals, and 104 cm³ (interquartile range, 60–145) for malignant collaterals. At 4 hours, median ischemic core volume was 19 cm³ (interquartile range, 5–32 cm³) for symmetric collaterals, 95 cm³ (interquartile range, 75–129 cm³) for other collaterals, and 127 cm³ (interquartile range, 61–196 cm³) for malignant collaterals. At 24 hours, median ischemic core volume was 28 cm³ (interquartile range, 11–44 cm³) for symmetric collaterals, 156 cm³ (interquartile range, 84–172 cm³) for other collaterals, and 176 cm³ (interquartile range, 111–272 cm³) for malignant collaterals. At 48 hours, median ischemic core volume was 36 cm³ (interquartile range, 24–54 cm³) for symmetric collaterals, 183 cm³ (interquartile range, 86–208 cm³) for other collaterals, and 281 cm³ (interquartile range, 194–369 cm³) for malignant collaterals. (B) Box-and-whisker plots with Kruskal Wallis test show that at presentation, median ischemic core growth rate (IGR) was 4 cm³ per hour (interquartile range, 1–5 cm³ per hour) for symmetric collaterals, 17 cm³ per hour (interquartile range, 12–27 cm³ per hour) for other collaterals, and 20 cm³ per hour (interquartile range, 15–28 cm³ per hour) for malignant collaterals. At 4 hours, median IGR was 0 cm³ per hour (interquartile range, 0–1 cm³ per hour) for symmetric collaterals, 0 cm³ per hour (interquartile range, –2 to 3 cm³ per hour) for other collaterals, and 3 cm³ per hour (interquartile range, 0–8 cm³ per hour) for malignant collaterals. At 24 hours, median IGR was 1 cm³ per hour (interquartile range, 0–1 cm³ per hour) for symmetric collaterals, 2 cm³ per hour (interquartile range, 1–4 cm³ per hour) for other collaterals, and 4 cm³ per hour (interquartile range, 4–6 cm³ per hour) for malignant collaterals. At 48 hours, median IGR was 0 cm³ per hour (interquartile range, 0–0 cm³ per hour) for symmetric collaterals, 1 cm³ per hour (interquartile range, 0–2 cm³ per hour) for other collaterals, and 2 cm³ per hour (interquartile range, 1–3 cm³ per hour) for malignant collaterals.

Ischemic Core Growth Rates

The IGR was also significantly different between collateral patterns at three of the four time points and was most pronounced at presentation. At presentation, median IGR was 4 cm³ per hour in patients with symmetric collaterals, 17 cm³ per hour in patients with other collaterals, and 20 cm³ per hour in patients with malignant collaterals (P < .001). At 4 hours, median IGR was 0 cm³ per hour in patients with symmetric collaterals, 0 cm³ per hour in patients with other collaterals, and 3 cm³ per hour in patients with malignant collaterals (P = .46). At 24 hours, median IGR was 1 cm³ per hour in patients with symmetric collaterals, 2 cm³ per hour in patients with other collaterals, and 4 cm³ per hour in patients with malignant collaterals (P = .001). At 48 hours, median IGR was 0 cm³ per hour in patients with symmetric collaterals, 1 cm³ per hour in patients with other collaterals, and 2 cm³ per hour in patients with malignant collaterals (P = .01) (Fig 3B).

Understanding Presentation Ischemic Core Volume and Growth Rate

At univariable analyses, determinants of presentation ischemic core volume included age, presentation NIHSS score, occlusion site, and collateral pattern (Table E1 [online]). In a multivariable model including these variables, more severe presentation NIHSS (parameter estimate [β] = 0.25; 95% CI: 0.08, 0.41; P = .003) and worse collateral pattern (β = –2.19; 95% CI: –3.47, –0.91; P = .001) were independently associated with greater presentation ischemic core volume (Table 2). At univariable analyses, determinants of presentation IGR included age, presentation NIHSS score, and collateral pattern (Table E1 [online]). In a multivariable model including these variables, more severe presentation NIHSS score (β = 0.20; 95% CI: 0.05, 0.36; P = .008) and worse collateral pattern (β = –2.90; 95% CI: –4.31, –1.50; P < .001) were independently associated with greater presentation IGR (Table 2).

Table 2:

Collateral Pattern as a Determinant of Presentation Ischemic Core Volume, Presentation Ischemic Core Growth Rate, and 24-hour Ischemic Core Growth Rate in Multivariable Models

Understanding 24-hour IGR

At univariable analyses, determinants of 24-hour IGR included age, presentation NIHSS score, occlusion site, and collateral pattern (Table E1 [online]). In a multivariable model including these variables, only worse collateral pattern (β = –2.03; 95% CI: –3.28, –0.78; P = .001) was independently associated with 24-hour IGR (Table 2). The receiver operating characteristic curve for collateral pattern to predict 24-hour ischemic core volume less than 50 cm³ is shown (area under the receiver operating characteristic curve, 0.92; 95% CI: 0.81, 1.00; P < .001) (Fig 4). Symmetric collaterals had a sensitivity of 87% (13 of 15) and a specificity of 94% (15 of 16) for 24-hour ischemic core volume less than 50 cm³. Nonmalignant collaterals had a sensitivity of 100% (15 of 15) and a specificity of 25% (four of 16) for 24-hour ischemic core volume less than 50 cm³.

Figure 4:

Collateral pattern can predict 24-hour ischemic core volume less than 50 cm³. Receiver operator characteristic curve (blue line; area under the receiver operating characteristic curve, 0.92; 95% CI: 0.81, 1.00; P < .001) showed symmetric collaterals had a sensitivity of 87% (13 of 15) and a specificity of 94% (15 of 16); nonmalignant collaterals had a sensitivity of 100% (15 of 15) and a specificity of 25% (four of 16). Green line indicates the line of no-discrimination.

Discussion

In patients with large vessel occlusion stroke not treated with reperfusion therapies, we characterized the natural history of ischemic core growth over 48 hours, which is key to understanding and identifying patients with slow-progressing stroke that may benefit from endovascular thrombectomy even in delayed fashion.

At presentation, median ischemic core volumes were 16 cm³ for symmetric, 69 cm³ for other, and 104 cm³ for malignant collateral patterns. Median IGR was 4 cm³ per hour for symmetric collaterals, 17 cm³ per hour for other, and 20 cm³ per hour for malignant. Worse collaterals were an independent determinant of larger ischemic core volume and faster IGR at presentation. At 24 hours, ischemic cores were 28 cm³ for symmetric, 156 cm³ for other, and 176 cm³ for malignant. Median IGR was 1 cm³ per hour for symmetric, 2 cm³ per hour for other, and 4 cm³ per hour for malignant. The only independent determinant of 24-hour IGR was collateral pattern. At 48 hours, ischemic cores were 36 cm³ for symmetric, 183 cm³ for other, and 281 cm³ for malignant. Median IGR was 0 cm³ per hour for symmetric, 1 cm³ per hour for other, and 2 cm³ per hour for malignant.

We used a three-category approach to classify collaterals, similar to that of other recent published studies (18). Although CT perfusion can be used to estimate collaterals (8,18), advantages of CTA highlighted in recent studies include its widespread availability, lack of threshold dependence, and ability to directly view vascular anatomy. Moreover, visual classification as symmetric or malignant is robust and immediately translatable. Collateral patterns are important because of their close relationship to clinical outcomes after stroke, which are often difficult to predict (22,23). In a cohort of patients with middle cerebral artery occlusive stroke, those with diminished CTA-defined collaterals had greater risk of clinical worsening, as measured by change in NIHSS score, during hospital admission (19). Furthermore, CTA-defined collaterals were an independent determinant of 6-month functional outcome after stroke, defined as modified Rankin Scale score of 2 or less, among a cohort with internal carotid artery and proximal middle cerebral artery occlusions (20). In another study that supports these findings, patients with malignant collaterals had higher median NIHSS score (21 vs 15) and greater functional dependence at 90 days (96% vs 64%) (10).

The relationship between quality of collateral flow and cerebral infarction has undergone significant study. In a cohort of patients with clinical middle cerebral artery stroke syndromes, CTA was more sensitive in depicting presentation ischemic core and more accurate for final ischemic core volume than was noncontrast-enhanced CT (24). This may not be surprising because of the low sensitivity of CT for early infarction and the physiologic relationship between collaterals and infarction. Ischemic core volume is one of the strongest determinants of 90-day outcomes, even among patients who undergo EVT (25). In our study, worse collaterals were an independent determinant of larger ischemic core volume at presentation, even when controlling for other variables including occlusion site, NIHSS score, and age.

Our findings are well supported in the literature. One study showed better CTA-defined collateral pattern was negatively associated with presentation DWI-defined ischemic core volume among a cohort of internal carotid artery and proximal middle cerebral artery occlusions in which 42% of occlusions were treated with EVT. A malignant collateral pattern had 98% specificity and 55% sensitivity for presentation ischemic core volume greater than 100 cm³; patients with malignant collaterals had larger mean presentation ischemic core volumes (166 cm³ vs 33 cm³, respectively) (10). We described additional time points in the present work. At each, median ischemic core volume was different between collateral patterns.

Our data also described the relationship of collaterals and IGR; worse collaterals were an independent determinant of faster presentation IGR, even when controlling for NIHSS score and age. Presentation IGR gained attention recently in the literature. In a cohort of internal carotid artery and middle cerebral artery occlusions, CT perfusion–defined collateral index was associated with CT perfusion–defined presentation IGR (18). Whereas Lin et al (18) did not use MRI, they found similar results to our data. Median presentation IGR was 3 cm³ per hour for the first collateral tertile, 9 cm³ per hour for the second collateral tertile, and 25 cm³ per hour for the third collateral tertile. Another study, in which 80% underwent EVT, suggested presentation IGR may predict outcomes (8). A CT perfusion–defined presentation IGR cutoff of less than 10 cm³ per hour had maximal sensitivity and specificity in selecting a favorable outcome. The authors defined slow progressors on the basis of this cutoff, showing they had lower NIHSS score, higher last known well–to–groin puncture time, and better CT perfusion–defined collaterals.

Few studies have explored ischemic core growth and IGR beyond presentation. Interestingly, ischemic core growth may even continue to some extent despite EVT (26). In a previous study, we showed logarithmic growth of DWI-defined ischemic core volume for 2 days in 38 patients with LVO not treated with reperfusion therapies (9). Understanding this natural history is important because patients with LVO can present 24 hours after onset and beyond. We showed that median IGR differences between collateral patterns are most pronounced at presentation but do persist at multiple time points over 48 hours, which has not previously been well established in the literature. Our current results support that although the overall rates are slower, collateral patterns still make a difference. The fairly steady IGR over time suggests that there is no so-called collapse of collaterals at least over the first 48 hours in humans (27).

Finally, previously published data in 24 patients with anterior LVO stroke demonstrated that presentation IGR less than 4 cm³ per hour and presentation ischemic core volume less than 20 cm³ had accuracies greater than 89% in identifying patients who would still have core less than 50 cm³ at 24 hours (9). This has implications, especially in patients who first present to a hospital that does not have EVT capabilities and who require transfer for treatment (28). However, these markers require MRI or CT perfusion, which are not readily available at many centers (5,29). Furthermore, recent data have suggested extended window EVT is beneficial even without advanced imaging (30). Our data support that single arterial phase CTA–defined collateral patterns may be equally predictive. When controlling for age, NIHSS score, and occlusion site, the only independent determinant of 24-hour IGR was collateral pattern. Symmetric collaterals had a sensitivity of 87% and a specificity of 94%, whereas nonmalignant collaterals had a sensitivity of 100% and a specificity of 25% for predicting 24-hour ischemic core volume less than 50 cm³. For these analyses, 24 hours and 50 cm³ were chosen a priori as conservative targets that have been used previously in clinical trials (2,3,31).

Our study had limitations. First, its relatively small sample size and single center design could limit generalizability. Our analysis, however, leveraged the benefits of a data set derived from patients who underwent multiple, serial MRI examinations in a protocolized fashion, which would not have been possible with a retrospective study of routine clinically acquired data. The use of this data set allowed assessment of the natural history of ischemic core growth in LVO stroke without EVT as a confounder, which is uncommon in modern imaging literature (32). Second, the investigational treatment of normobaric oxygen used in the original clinical trial was a potential confounder, although it was not associated with presentation collaterals or ischemic core volume or growth rate at any point in our study, and to our knowledge none were reported previously for either imaging or clinical outcome variables (11). Third, although there are several potential advantages of using data from a previously conducted randomized controlled trial, we performed our assessments and analyses of collateral status in patients with acute stroke retrospectively. Finally, we cannot exclude the possibility of selection bias, because although only patients who underwent CTA and serial MRI at presentation were included in our analyses, exclusions were primarily attributable to lack of acute MRI availability at one center in the trial. Furthermore, the collateral pattern category other was heterogeneous, with some patients rated as closer to symmetric and others rated as closer to malignant. Despite these potential limitations, this simple three-category classification system is easy to learn, easy to use, robust, and immediately translatable.

In conclusion, our data set characterized the natural history of ischemic core growth in the setting of large vessel occlusion over 48 hours. Comparison of collateral patterns showed that there were different ischemic core volumes and ischemic core growth rates (IGRs) at multiple points. Symmetric collateral patterns at CT angiography were associated with smaller ischemic core volumes and slower ischemic lesion growth, which was determined at diffusion-weighted MRI. As an independent determinant of both presentation and 24-hour IGR, collateral pattern may be useful in the triage of patients for endovascular thrombectomy, especially where MRI and CT perfusion are unavailable. Further study of outcomes is warranted.