Article Text
Abstract
Background and purpose Collateral assessment using CT angiography is a promising modality for selecting patients for endovascular thrombectomy (EVT) in the late window (6–24 hours). The outcome of these patients compared with those selected using perfusion imaging is not clear.
Methods We pooled data from seven trials and registries of EVT-treated patients in the late-time window. Patients were classified according to the baseline imaging into collateral imaging alone (collateral cohort) and perfusion plus collateral imaging (perfusion cohort). The primary outcome was the proportion of patients achieving independent 90-day functional outcome (modified Rankin Scale ‘mRS’ 0–2). We used the propensity score–weighting method to balance important predictors between the cohorts.
Results In 608 patients, the median onset/last-known-well to emergency arrival time was 8.8 hours and 53.2% had wake-up strokes. Both cohorts had collateral imaging and 379 (62.3%) had perfusion imaging. Independent functional outcome was achieved in 43.1% overall: 168/379 patients (45.5%) in the perfusion cohort versus 94/214 (43.9%) in the collateral cohort (p=0.71). A logistic regression model adjusting for inverse-probability-weighting showed no difference in 90-day mRS score of 0–2 among the perfusion versus collateral cohorts (adjusted OR 1.05, 95% CI 0.69 to 1.59, p=0.83) or in a favourable shift in 90-day mRS (common adjusted OR 1.01, 95% CI 0.69 to 1.47, p=0.97).
Conclusion This pooled analysis of late window EVT showed comparable functional outcomes in patients selected for EVT using collateral imaging alone compared with patients selected using perfusion and collateral imaging.
PROSPERO registration number CRD42020222003.
- stroke
- image analysis
- interventional
- cerebrovascular
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Statistics from Altmetric.com
Introduction
A significant proportion of patients who had a stroke with large vessel occlusion (LVO) present in the late window (6–24 hours from onset).1 Those patients have poor prognosis if reperfusion is not achieved, with estimated mortality or severe morbidity rates of about 40%.2 Two randomised trials showed that endovascular thrombectomy (EVT) significantly improves the outcome of late-window patients with LVO and clinical-core mismatch (DAWN) or target mismatch (DEFUSE-3) on perfusion imaging, or MRI-diffusion imaging (in a subset of DAWN patients).2 3 Other trials such as ESCAPE used CT angiography (CTA) collateral-based imaging paradigms when testing EVT in time windows up to 12 hours from the last-known-well time.4 Consensus on the imaging paradigm to be used for patient with EVT selection in the late-time windows is therefore lacking. A key reason for this lack of consensus is that the different available imaging paradigms for patient selection have not been subject to randomised trials. Furthermore, the economic impact of one imaging paradigm versus another, given variability in ease of availability and cost, have not been studied thoroughly.5–7
Since the publication of DAWN and DEFUSE-3 trials, hospitals across the world have accumulated ample data to allow for reasonable assessment of clinical outcomes in late-window patients selected for EVT using various imaging paradigms. We sought to perform an individual-patient level analysis of such data gathered from comprehensive stroke centres worldwide and describe the imaging selection and outcomes of EVT-treated patients presenting in the late-time window.
Materials and methods
The Selection Of Late-window Stroke for Thrombectomy by Imaging Collateral Extent Consortium comprises an international group of investigators who aim to study the selection and outcome of patients who had a stroke treated with EVT in the late-time window.
Data from seven trials and registries were pooled for this analysis. These are the Acute STroke Registry and Analysis of Lausanne,8 Lausanne, Switzerland; the Beaumont Hospital Registry,9 Dublin, Ireland; the efficacy and safety of nerinetide for the treatment of acute ischaemic stroke (ESCAPE-NA1) trial,10 the Endovascular Treatment for Small Core and Proximal Occlusion Ischemic Stroke (ESCAPE) trial,4 the Italian Registry of Endovascular Thrombectomy (IRETAS),11 Italy; the Precise and Rapid Assessment of Collaterals Using Multiphase CTA in the Triage of Patients with Acute Ischemic Stroke for IV or IA Therapy (PRove-IT) study12; and the Seoul National University Bundang Hospital stroke registry.13 All included studies and registries were approved by local ethics review committees or analysed only anonymised data as permitted by local legislation. The principal investigators of these studies were approached to contribute their data into the consortium. This current study is an individual-patient level analysis of these data.
All patients underwent EVT in the late window (>6 hours from onset to imaging time). Details of the imaging method used for patient selection are shown in online supplemental table 1. Further, the pooled data contain the following patients’ details:
Supplemental material
Patients’ demographics and clinical characteristics including age, sex and baselineNational Institutes of Health Stroke Scale (NIHSS).
Imaging parameters including site of occlusion, collateral and perfusion profiles and core/penumbral tissue volumes.
Procedural details including final reperfusion status and time.
Outcome parameters such as symptomatic intracranial haemorrhage after treatment and 90-day modified Rankin Scale (mRS).
Imaging cohorts
We defined two cohorts according to the imaging modality done for patients’ selection:
Perfusion cohort: Patients who underwent perfusion imaging at baseline, regardless of whether they had CTA collateral imaging.
Collateral cohort: Patients who only underwent CTA collateral imaging. This second cohort included two subcohorts:
Single-phase CTA: Patients in whom collateral assessment was done using the conventional, single-phase CTA.
Multiphase CTA: Patients in whom collateral imaging was done using time-resolved images of pial arteries.14
For this analysis, if a patient had both collateral imaging and perfusion imaging, perfusion imaging was considered the imaging modality used for EVT selection. The details of the imaging modalities and definitions, the collateral assessment grading and definitions of core/penumbra used for each study are provided in online supplemental table 1. A favourable collateral profile was defined as ≥50% filling of the middle cerebral artery (MCA) pial arterial circulation distal to the intracranial occlusion.14 The definition of favourable perfusion profile was what was defined by the contributing trials and registries. A mismatch ratio ≥1.8 was used to define favourable perfusion profile by most trials and registries (n=195 of 379, 52%). The exception was the IRETAS that used ischaemic core ≤50% of hypoperfusion extent or <33% of the MCA territory according to Turk et al to define a favourable perfusion profile (n=184 of 379, 48%).15
Outcomes
Successful reperfusion was defined as modified treatment in cerebral infarction (mTICI) score of 2b or 3. We collected 90-day mRS data from the included studies. The primary outcome was the proportion of patients with independent functional outcome (mRS 0–2) at 90 days. In addition, the incidence of symptomatic intracranial haemorrhage (sICH) using the European–Australian Cooperative Acute Stroke Study 2 definition was collected.16
Statistical analysis
We summarised baseline clinical and imaging characteristics using appropriate statistics. We used a propensity score weighting approach to minimise the effects of baseline imbalance between the perfusion versus collateral cohorts. Propensity score weighting has the advantage of using the entire study sample in contrast to the propensity score matching approach which only uses matched subjects. We calculated the propensity scores as the predicted probability of a patient being in the perfusion cohort versus collateral cohort using hierarchical mixed-effects logistic regression after adjusting for prespecified covariables namely age, witnessed onset versus wake-up stroke, time from last-known-well to emergency department (ED), site of occlusion, tandem cervical occlusion, successful reperfusion and time from last-known-well to reperfusion; ‘study ID’ was used as a random effects variable to account for clustering of data within studies. Each patient was then assigned a weight based on their propensity score odds, that is, propensity score/(1 − propensity score). We applied the inverse probability of treatment weighting method to balance the distribution of these covariates between the perfusion and collateral cohort and checked its validity using the absolute standardised differences as a balance diagnostic with a margin <0.1.
For the primary outcome, we ran a mixed-effects logistic regression model using mRS 0–2 as the dependent variable and the imaging paradigm (perfusion vs collateral cohort) as the independent variable while adjusting for the inverse probability weighting. As sensitivity analyses, we ran a mixed-effects logistic regression model without propensity-weighting, as well as a mixed-effects ordinal logistic regression model (after testing the proportional odds assumption using the Brant’s test); study ID was included as random effects variable in these analyses too. The adjusted OR and 95% CIs are reported.
All analyses were performed using Stata Statistical Software (Release V.15., StataCorp) with a significant level set at <0.05 (two-sided).
Data availability
The data that support the findings of this study are available in the results and supplement of this paper.
Results
Overall cohort characteristics
A total of 608 patients from seven studies were included in this analysis. The median (IQR) age was 70 (21) years and 50.5% were women. Table 1 summarises the clinical and imaging characteristics of these patients. The median (IQR) times from onset/last-known-well to ED arrival was 530.5 (290) min, and from onset/last-known-well to baseline imaging (first image time) 559 (273) min. A total of 310 patients (53.2%) had wake-up strokes while 273 patients (46.8%) had witnessed onset of stroke.
The median baseline NCCT Alberta Stroke Program Early CT Score (ASPECTS) score was 8 (IQR 2), and 187 patients (30.8%) had ASPECTS of 5–7 while 29 patients (4.8%) had scores of 2–5. The most common occlusion location was the M1-MCA segment (60.4%); a tandem cervical internal carotid artery occlusion was present in 112 (18.4%) patients. A total of 55 patients (9.1%) received intravenous alteplase prior to EVT.
Imaging characteristics
Among 608 patients, 379 patients (62.3%) underwent CTA and CT perfusion (CTP) (perfusion cohort) and 229 had CTA alone (collateral cohort). Multiphase CTA was performed in 202 patients in the perfusion paradigm cohort (53.3%) and in 192 patients (83.8%) in the collateral cohort. Patients in the perfusion cohort were younger compared with the collateral cohort (mean age 66.5 vs 68.9 years, table 1) with fewer women (48.8% vs 53.3%). The median time from last-known-well to emergency arrival was shorter in the perfusion versus collateral cohort (515 vs 555 min). The median NCCT ASPECTS was similar in the perfusion and collateral cohorts (median (IQR) of 8 (3) vs 8 (2), p=0.22). Ninety-four per cent of patients in the perfusion cohort had NCCT ASPECTS of six or more compared with 97% in the collateral cohort. The most common occlusion site was M1-MCA in both cohorts (62.9% vs 58.8%, p=0.32). The proportion of patients with M2-MCA occlusion in the CTP cohort were almost double those in the collateral cohort (16.6% vs 8.3%, p=0.004). Tandem cervical and intracranial occlusions were noted in 20.1% of patients in the perfusion cohort versus 15.7% of the collateral cohort (p=0.18).
Perfusion imaging cohort
Of the 379 patients in this cohort, 350 patients (92.4%) had a favourable perfusion profile while 350 patients (92.4%) had a favourable collateral profile. Three hundred and twenty-six patients (86%) had favourable perfusion and collateral imaging profile while 10 (2.6%) patients had unfavourable perfusion and collateral imaging profiles (concordant imaging profiles). A favourable perfusion but not favourable collateral profile was noted in 23 patients (6.1%). A favourable collateral but not perfusion profile was noted in 23 patients (6.1%).
Collateral imaging cohort
Among the 229 patients in the collateral cohort, 192 (83.8%) had multiphase CTA. A favourable collateral profile was seen in 86.3% of this cohort (89.5% with multiphase CTA vs 69.5% with single-phase CTA). The characteristics of the single-phase CTA versus multiphase CTA subcohorts are outlined in online supplemental table 2.
EVT procedural times and outcomes
The median imaging to puncture time was 65 (IQR 70) min overall. This was significantly shorter in the perfusion versus collateral cohorts (table 1). A total of 493 (81.2%) patients achieved successful reperfusion (mTICI 2b–3) with a median puncture to reperfusion time of 35 (IQR 40) min. Successful reperfusion was achieved in 77.8% in the perfusion cohort versus 86.9% in the collateral cohort (p=0.005) with a median (IQR) puncture to reperfusion time of 45 (43) min. This was longer than that of the collateral cohort (median 24, IQR 23 min, p=0.0001). However, the median time from last-known-well to reperfusion in the perfusion cohort was shorter than the collateral cohort: 648 (302) min versus 723 (281) min (p=0.016).
Outcomes
sICH occurred in 70 patients (12.4%) overall. This was higher in patients in the collateral versus perfusion cohort (18.8% vs 9.2%, p=0.0011). Ninety-day mortality was noted in 102 patients (17.5%) which was marginally higher in the collateral (20.6%) cohort compared with the perfusion (15.7%) cohort (p=0.138).
A total of 583 patients (96%) had 90-day mRS scores available. Among the 25 patients with no follow-up data, 10 (40%) were in the perfusion cohort and 15 (60%) in the collateral cohort. Of 583 patients, 262 (44.9%) achieved independent functional outcome (mRS 0–2) at 90 days (figure 1). This proportion was slightly higher among patients in the perfusion cohort (45.5%) than the collateral cohort (43.9%; p=0.71).
Outcomes in the perfusion cohort
With concordant perfusion and collateral imaging profiles
One hundred and forty-four of 309 patients (46.6%) achieved 90-day mRS 0–2 when both perfusion and collateral imaging profiles were favourable. In the 10 patients with unfavourable perfusion and unfavourable collateral profiles, 3 patients (30%) achieved favourable outcome (p=0.3 for difference in outcomes) (table 2).
With discordant perfusion and collaterals profiles
Among 19 patients who had a favourable perfusion profile but unfavourable collaterals, 6 patients (31.6%) achieved mRS 0–2. On the other hand, among 30 patients with unfavourable perfusion but favourable collateral profiles, 15 patients (50%) achieved mRS 0–2 (p=0.21 for difference in outcome).
Outcomes in the collateral cohort
Among patients selected for EVT using collateral imaging alone, 94 patients (43.9%) achieved 90-day mRS 0–2 (online supplemental table 3).
Among 34 patients who had single-phase CTA assessment, 34.8% achieved mRS 0–2 when the single-phase CTA-collateral profile was favourable. On the other hand, of 159 patients with a favourable multiphase CTA-collateral profile, 48.4% achieved mRS 0–2 (p=0.20 for difference in outcome).
When the collateral profile was unfavourable on single-phase CTA, 18.2% achieved mRS 0–2 compared with 30% of patients with unfavourable multiphase CTA-collateral profile (p=0.24 for difference in outcome).
Outcome modelling and propensity weighted analyses
The clinical characteristics of the overall cohort before and after propensity weighting is shown in table 3. A logistic regression model adjusting for inverse-probability-weighting showed no difference in the odds of 90-day mRS score of 0–2 outcomes according to the imaging modality used (perfusion vs collateral cohorts OR 1.05, 95% CI 0.69 to 1.59, p=0.83), or a favourable shift in 90-day mRS (common adjusted OR 1.01, 95% CI 0.69 to 1.47, p=0.97).
We also conducted sensitivity analyses using conventional mixed-effects binary and ordinal logistic regression models accounting for clustering of data within sites and adjusting for age, ASPECTS, sex, wake-up versus witnessed onset, occlusion location, time from last-known-well to emergency arrival, time from last-known-well to reperfusion, tandem occlusion and alteplase treatment. These models showed no significant association between the imaging paradigm used and the likelihood of achieving 90-day mRS score of 0–2 (adjusted OR 1.10, 95% CI 0.91 to 1.25, p=0.44) or a favourable shift in 90-day mRS (common adjusted OR 1.02, 95% CI 0.87 to 1.19, p=0.24).
The 90-day outcomes were missing for 25 patients (4%). We imputed values for missing data using multiple imputation via multivariate normal distribution. The approach imputes missing data from a conditional distribution and provides reasonable estimates even with a small sample size. We also ran the analyses with all missing values replaced, and the results were identical (results not shown). A total of 583 patients (96%) had 90-day mRS scores available.
Discussion
In this individual-patient level multinational data of late-window patients with LVO from seven studies conducted across different healthcare settings and practice patterns, 90-day functional outcomes were comparable among patients selected for EVT in the perfusion versus collateral imaging cohorts. Patients in the collateral cohort receiving EVT had similar baseline characteristics as those included in the DAWN and DEFUSE-3 trials.2 3
Perfusion imaging is not used routinely for selection of patients treated with EVT in the early time window (≤6 hours).17 One possible explanation is that in the early time epoch, it is highly likely that most patients have salvageable brain tissue, even if these patients represent a mix of fast and slow stroke progressors.18 This is in contrast to patients presenting in the late-time window where differentiating slow progressors (with salvageable brain) from fast progressors may be more useful. The DAWN and DEFUSE-3 trials therefore used imaging selection criteria to maximise the benefit from reperfusion up to 24 hours from onset/last-known-well time. This however does not mean that perfusion or diffusion imaging is the only way to select patients. It is highly probable that patients with favourable perfusion profiles also have favourable collateral profiles that can be identified using CTA-based collateral imaging.14 Our analyses shows a high concordance (84%) between favourable or unfavourable perfusion and collateral profiles. Even when one imaging paradigm was not favourable, the likelihood of functional independence at 90 days after EVT was reasonable. These observations and others support the use of collateral imaging to select patients for EVT in the late-time window.14
A critique of collateral imaging on CTA is that assessments can be inaccurate due to delay in contrast bolus timing.19 The use of multiphase CTA for collateral assessment avoids this shortcoming of collateral assessment on single-phase CTA.14 In addition, multiphase CTA improves reliability of collateral assessments without added contrast and minimal added radiation exposure when compared with single-phase CTA.20–22 These advantages of multiphase CTA led to its use in clinical trials like ESCAPE and ESCAPE-NA1 that included patients presenting up until 12 hours from last-known-well time.23 24 Simpler than even multiphase CTA assessment would be using CT ASPECTS along with single-phase CTA LVO detection for imaging selection in the late window. Prior studies that used ASPECTS-based paradigms for EVT selection in the late window showed similar outcomes to patients selected with perfusion imaging.25–27 This supports the usefulness of NCCT in identifying patients with LVO beyond the hyperacute period who may benefit from EVT.28 29
These analyses highlight the need for a better understanding of the reasons for and limitations of various imaging paradigms in informing patient selection in the late-time window. The key question to ask is whether imaging paradigms that rely on excluding extensive ischaemic changes on NCCT (eg, ASPECTS), while confirming the presence of LVO (on CT/CTA) is all that is needed. More and more data show that CT/CTA-based imaging paradigms will increase the number of patients being offered EVT given their wide availability.1 Visual assessment of collaterals was recently described to be independently associated with functional outcome at 90 days, similar to automated and quantitative perfusion assessments.30 This shows that the feasibility and ease of collateral assessment do not jeopardise its prognostic value. Prior studies suggest that less than 25% of all late-window patients who had a stroke stroke meet DAWN or DEFUSE-3 imaging criteria versus 46.6% that meet CT/CTA-based criteria for EVT eligibility.8 A crucial question is whether treating more patients with EVT would result in better or worse outcomes overall? Of note, our results do not show worse clinical outcomes in patients treated with EVT in the collateral cohort versus the perfusion cohort. The key issue then is to estimate how patients who were not offered EVT based on perfusion or collateral imaging would have fared at 90 days. We do not have this data for analysis. Based on results from the recent EVT trials, we can however conjecture that such patients with LVO, if not offered EVT, would have not achieved good outcomes overall. Our results therefore suggest that the use of one imaging paradigm versus another, that is, (perfusion or collateral imaging) may not result in significant differences in functional outcomes in patients with LVO presenting late.
To compare the utility of different imaging paradigms in patient selection in the late window, we will also need to understand their relative availability and costs. Imaging paradigms used in DAWN or DEFUSE-3 trials aimed to select patients who are most likely to benefit from EVT. The perfusion cohort in our study had lower incidence of symptomatic haemorrhages and death compared with the collateral cohort. However, these advantages, of perfusion imaging are limited by the lack of high quality data on the outcomes of patients without favourable perfusion mismatch profile who are treated with EVT. Our findings provide insights into the trade-offs associated with the use of different imaging paradigms. Estimating the consequences of using each imaging approach will be challenging given the ethical, economical and societal implications of undertreating versus overtreating, and the high likelihood of poor outcome without reperfusion.15
Available body of evidence suggests that CT/CTA-based imaging paradigms are associated with similar outcomes to perfusion-based paradigms when used to select patients for EVT in the late-time window. The DAWN and DEFUSE-3 trials demonstrated the superiority of EVT over medical treatment in the late-time window. The use of perfusion imaging in these trials was to help identify the population that is most likely to benefit from EVT given its high specificity for identifying patients who can benefit from treatment. Our report and other studies show that a favourable collateral profile is often concordant with that of perfusion. Moreover, there is evidence to suggest that perfusion imaging can overestimate the extent of irreversibly damaged brain tissue.31 Results from ongoing randomised trials that select late-window patients for EVT versus medical treatment using CT/CTA may provide further evidence on this issue. (https://clinicaltrials.gov/ct2/show/NCT04256096, https://doi.org/10.1186/ISRCTN19922220).
Our study has limitations. First, we combined data from different sites and studies that used different imaging protocols and parameters for perfusion and collateral imaging. Nonetheless, these studies prospectively defined these parameters, and used protocols that are practiced in their usual patient care, thus making our results applicable to routine care settings. Second, the imaging interpretation was not centralised, and some centres did not use automated software for perfusion post processing. Despite this non-centralised imaging interpretation, our findings were similar to those from randomised trials signifying the robustness of the results. Third, we did not have data on the patients who were screened but not offered EVT. Such data would have provided a better understanding of the patient population presenting in the late-time window. Fourth, the incidence of sICH was low overall to assess their effect on long-term outcomes. Finally, differences in workflow speed and clinical outcomes could be due to variability in local practice patterns rather than because of the use of a particular imaging paradigm. Statistical methods like mixed-effects modelling and propensity weighting that we used only mitigate this limitation to an extent.
Conclusion
In this large multinational cohort, patients with LVO who present greater than 6 hours from onset and were selected for EVT using CTA-based collateral assessment achieved comparable functional outcomes to patients who were selected for EVT using perfusion imaging. While collateral imaging and perfusion imaging appear comparable as selection tools, a diagnostic randomised controlled trial is needed to decide the imaging modality of choice for selecting patients for EVT in the late-time window.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
This study does not involve human participants.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Correction notice This article has been corrected since it first published. A colon has been added to the title for clarity.
Contributors MAA and BKM conceived the manuscript idea and organised the data collection. MAA, BJK, AA and BKM developed the statistical analysis plan and performed the analyses. The rest of the coauthors participated in the data collection from their respective registries and trials. MAA drafted the manuscript. All authors discussed the results and contributed to the final manuscript. MAA acts as guarantor and accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests MAA reports being a member of the scientific advisory board of Palmera Medical. AD reports personal fees from Medtronic. MG reports grants or personal fees from Medtronic, Stryker, Microvention, Cerenovus, and has a patent Systems of Acute Stroke Diagnosis issued to GE Healthcare. MDH has received grant support from Medtronic, consultant fees from Boehringer Ingelheim and speaker’s fees from Amgen. SM acted as a consultant for Cerenovus. BKM reports shares in Circle NVI; patent for systems of triage in acute stroke. Dr Patrik reports research grants from the Swiss Heart Foundation and the Swiss National Science Foundation. DT received honoraria as a member of advisory board of Abbott, Boehringer Ingelheim, Bayer, Pfizer-BMS, Medtronic and Daiichi Sankyo. The other authors report no conflicts.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.