WHAT IS ALREADY KNOWN ON THIS TOPICWHAT THIS STUDY ADDS

This study performs a detailed analysis of how procedure time affects functional outcome after thrombectomy in different subpopulations of patients traditionally excluded from clinical trials. We demonstrate that the ‘golden hour’ effect—that is, successful recanalization within an hour of groin puncture, is universal across age groups, Alberta Stroke Program Early CT scores, location of occlusion, or tissue plasminogen activator therapy. However, patients with posterior circulation strokes and elderly patients are most sensitive to the effect of procedure time on thrombectomy outcomes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

Despite the expanded indications for endovascular thrombectomy to include more distal occlusions, delayed presentation, older patients, and those with larger cores, the procedure time remains a major predictor of clinical outcomes. The effect of procedure time in all these subpopulations was reproducible, indicating increased risk of worsening outcomes and hemorrhage with longer procedure times. At the same time, the impact of procedure time is most significant in elderly patients and those with posterior circulation strokes, where extending procedure time to achieve perfect recanalization might have diminishing returns. Yet, decision to abort futile intervention remains at the discretion of the operator.

Introduction

Endovascular thrombectomy (EVT) is the current standard of care for treatment of acute ischemic stroke secondary to large vessel occlusion (LVO). Despite the relatively large effect size of intervention, nearly 50% of patients who achieve successful recanalization after EVT still do not achieve functional independence at 90 days.1–7 Procedure time, defined as time from groin access to vessel recanalization, is among recently recognized predictors of poor outcome after EVT.7–10 Despite the clear advantage of successful recanalization, a major question in neurointerventional procedures remains when to stop or abort a failed thrombectomy. Previous studies have demonstrated that procedure times longer than 1 hour are correlated with higher complication rates, higher rates of hemorrhage, increased healthcare burden, and worse functional outcomes.7 11 These studies suggested that after three attempts or 1 hour of procedure time, neurointerventionalists should carefully consider the futility of intervention. However, these studies included all patients undergoing EVT without taking into account subpopulations of patients who might be at higher or lower sensitivity to the effect of procedure time. This is specifically relevant in the subpopulations that have become more recently included in real-world EVT practice, including elderly patients,12 those with posterior circulation or distal occlusions,9 13–17 and those with low Alberta Stroke Program Early CT (ASPECT) score or delayed presentation.10 18–22

In this study, we carry out a large, retrospective, multicenter investigation of how the impact of procedure time on thrombectomy outcomes is influenced by admission variables, including age, location of thrombus, use of tissue plasminogen activator (tPA), time from onset, and ASPECT scores. We also aim to investigate whether the ‘golden hour’ for intervention applies across patient stratification based on these variables.

MethodsPatient population

The study population was drawn from prospectively maintained databases of patients who had undergone EVT for acute ischemic stroke between January 2013 and March 2022 at 39 comprehensive stroke centers in the United States and globally as part of the Stroke Thrombectomy and Aneurysm Registry (STAR).9 13 23 24 No patients under 18 years of age were included in this study. Patient were included if they underwent EVT using modern thrombectomy devices used in the 2015 randomized controlled trials, as detailed before.7 13 23 24 Patients with anterior and posterior circulation stroke were included. Patients with tandem occlusions were not included in this subpopulation analysis. Intravenous tPA was administered to patients according to current guidelines, and independently of any decision to undergo EVT. The EVT approaches used included aspiration, stent retriever or combined approaches. Data were collected from each site independently using a comprehensive shared database and then were curated and analyzed by the statistical team at the registry. The institutional review board at each individual site approved this study.

Data collection

Demographic variables, National Institutes of Health Stroke Scale (NIHSS) scores on admission, use of IV tPA, and pre-stroke modified Rankin Scale (mRS) scoreswere retrieved by review of patient charts. CT and CT angiography scans, obtained preoperatively and read by trained radiologists, were used to compute ASPECT scores and assess location of the occlusion. Technical variables, technical outcomes, and procedural complications were obtained from interventional procedure notes. Procedure time was defined as the duration of time beginning with groin access and ending either with first recanalization (Thrombolysis In Cerebral Infarction (TICI) score 2b or more) or with abortion of the procedure if the procedure was aborted before TICI 2b was achieved.

Clinical outcomes

The primary clinical outcome was defined as the 90-day mRS score recorded during a follow-up visit to a stroke neurologist at 90±14 days after the stroke or by contacting rehabilitation facilities. Secondary clinical outcomes included the presence of postprocedural symptomatic intracranial hemorrhage (sICH) defined as a parenchymal hematoma type 2 hemorrhage by European Cooperative Acute Stroke Study II, or any type of hemorrhage with associated four-point decline in NIHSS score on surveillance imaging within 24 hours of EVT, and mortality.

Statistical analysis

SPSS version 24 (IBM Corporation, New York, New York, USA), GraphPad Prism 6 (GraphPad, La Jolla, California, USA), and Microsoft Office Excel (Microsoft, Redmond, Washington, USA) were used to sort data and perform various statistical analyses. Univariate analysis was first compared with the Student’s t-test or analysis of variance (for parametric variables), Mann-Whitney test, or Kruskal-Wallis test (for non-parametric variables), or Χ2 test (for categorical variables). Bonferroni test was used for multiple comparisons, and likelihood ratios were used for multiple comparisons in categorical measures. The significance level was set at a two-sided α <0.05.

When demographic or other baseline variables were not available for patients, including race, admission NIHSS score, ASPECT score, or comorbidities, imputations were performed to avoid the potential bias of excluding these patients completely from a multivariate analysis. Multiple imputations were used for patients missing multiple variables. Multivariate mixed regression models were used to analyze the relationship between procedure time and good outcome (measured by 90-day mRS score), presence of intracerebral hemorrhage, and mortality.

All variables were included in the model and selected in the final models using a backpropagation algorithm in SPSS to avoid bias in variable selection. Variables included in models included those shown in sections I and II of table 1 in addition to frontline technique used, internal carotid artery angioplasty, and successful recanalization, where appropriate. Procedure time was used as a continuous variable, or as bins of 10 min intervals for the analyses performed in this work. Independent models were generated for each subpopulation of patients, and adjusted odds ratios (aORs) for procedure time variables were computed and reported. Cumulative percentage curves and change in ORs of outcomes as a function of procedure time were visualized, and metrics were computed based on best fit curve, linear or exponential; curve metrics were computed in GraphPad prism for the best fitting model. For exponential curves, time constant (τ) was defined as the procedure time at which the dependent variable reaches 67% of maximal value. For each curve the τ estimate was then computed with its associated error, and Z-scoring was used to compare the different time constants between curves.

Table 1

Description of study cohort

ResultsPatient population

A total of 9954 patients were included in the registry. Of these, 993 were excluded because procedure time or outcome data were not available. A total of 8961 patients were included in this study, of whom 49% were female and average age was 69 years (table 1). Posterior circulation stroke occurred in 11% of subjects and 19% experienced distal occlusions. Pre-EVT thrombolysis was administered in 45%. The median ASPECT score was 9, and the rate of successful recanalization, defined as TICI 2b or more, was 85%. The average procedure time was 51±44 min, and the average number of EVT passes was 2.2±1.6. The rate of sICH was 6%, good functional outcome (mRS score 0–2) was 38%, and 90-day mortality was 21% (table 1). Distribution of mRS scores by procedure time group is shown in online supplemental figure 1.

Impact of procedure time on functional outcome

We first investigated the impact of increasing procedure time (by 10 min increments) on functional outcomes assessed by 90-day mRS scores using multivariate regression models that controlled for patient and procedural covariates (figure 1). In the full cohort, longer procedure times were associated with higher odds of poor outcomes (aOR=1.1, P<0.001). Patients were then stratified by stroke location (anterior vs posterior), depth of occlusion (distal vs proximal), administration of IV tPA, age group (0–40, 41–60, 61–80, ≥81), ASPECT scores (0–5, 6–10), and time between stroke onset and groin puncture (0–6 hours, 6–12 hours, 12–24 hours, >24 hours). Longer procedure times were associated with higher odds of poor functional outcome at 90 days in all studied subpopulations except for patients presenting after 24 hours of onset (figure 1). Of note, the most prominent modulator of the effect of procedure time on EVT outcomes was posterior location of stroke (aOR=1.17, 95% CI 1.1 to 1.23, figure 1A).

Figure 1

The impact of procedure time on functional outcomes after endovascular thrombectomy stratified by patient variables. Reported are adjusted odds ratios (aOR) obtained from multivariate logistic models as described in Methods. (A) Procedure time was used as a continuous variable with 10 min time increments. (B) Procedure time used as a categorical variable dichotomized into within ‘golden hour’ (<1 hour) or >1 hour. All individual models were controlled for the following variables where applicable: age, gender, race, diabetes, hypertension, hyperlipidemia, atrial fibrillation, prior stroke, baseline modified Rankin Scale score pre-stroke, admission National Institutes of Health Stroke Scale score, Alberta Stroke Program Early CT (ASPECT) score, use of tissue plasminogen activator (tPA), and onset-to-groin time. Reported are aORs with 95% CI between error bars. Dashed line represents aOR=1.0. (C) Analyses similar to those in (B) were performed, but only in patients with successful recanalization (TICI 2b or more). OTG, onset-to-groin time; PT, procedure time; TICI, Thrombolysis in Cerebral Infarction.

The golden hour of procedure time by subpopulation

Previous studies on stroke thrombectomy outcomes have described the golden hour for stroke thrombectomy in trial-eligible patients and real-world registries, suggesting higher rates of futile intervention and complications in patients undergoing EVT with procedure time exceeding 60 min.7 11 25 Therefore, we investigated whether the prognostic effect of the golden hour also applies to subpopulations of patients undergoing EVT (figure 1B). In the full cohort of patients, procedure time less than 60 min was associated with higher odds of good functional outcomes at 90 days (aOR=2.0, 95% CI 1.9 to 2.3, P=0.001). This golden hour effect was consistent in patients with anterior and posterior circulation strokes (aOR=1.9 and 3.1, respectively), proximal or distal occlusions (aOR=1.1 and 1.1, respectively), across age groups (figure 1B), in groups with both high and low ASPECT scores (figure 1B), and in patients presenting less than 24 hours from onset of symptoms(figure 1B). When selecting only patients with successful recanalization, a similar effect for the golden hour was observed (figure 1C).

Differential impact of procedure time on outcome

To compare the impact of procedure time on functional outcomes after thrombectomy within different subpopulations of EVT patients, we analyzed the cumulative rate of good functional outcomes as a function of procedure time for each subpopulation of patients (figure 2A). A one-phase association least-squares fit was used to interpolate a standard curve for each subpopulation of patients with R2 between 0.97 and 0.99 (figure 2A). Each curve plateaus at the overall rate of good outcome for that group irrespective of procedure time. The rate of curve growth was used to compare the sensitivity with procedure time across different subpopulations. We computed the time constant (τ) which reflects the procedure time at which 67% of maximal cumulative rate of good outcome is attained. A smaller time constant (τ) indicates a faster rising growth curve and thus higher sensitivity to procedure time.

τ was then compared across the strata of patients in this study (figure 2B). Compared with patients with anterior circulation stroke, those with a posterior circulation stroke had significantly lower time constant—that is, higher sensitivity to procedure time (figure 2). In comparison with patients with a distal occlusion, patients experiencing a proximal occlusion stroke had a significantly lower time constant, indicating higher sensitivity to procedure time. Patients who did not receive tPA and patients presenting at longer time from onset followed the same pattern of increased sensitivity to procedure time (lower τ) compared with those who did receive tPA or presented in the earlier window, respectively (figure 2). When the cohort was stratified into age groups, there was an age-dependent increase in sensitivity to procedure time, with the octogenarian group having the highest sensitivity to procedure time in their clinical outcomes (figure 2B). There was no significant difference in sensitivity to procedure time in patients with different ASPECT scores despite procedure time still being a predictor of good outcome in this group.

Figure 2

Differential effect of procedure time on functional outcome by different patient subgroups. (A) Curves demonstrate the cumulative rate of good functional outcome (mRS score 0–2) as function of procedure time for different subgroups. A one-phase association least squares fit was used to interpolate a standard curve for each subpopulation of patients with R2 between 0.97 and 0.99. (B) For each curve in (A), the time constant (τ) was computed demonstrating the procedure time at which 67% of good outcome rate is achieved. The τ serves as surrogate for procedure time sensitivity. Smaller τ indicating higher sensitivity of subgroup to procedure time compared with higher τ. Reported are τ and 95% CI interval. The τ estimates were then compared using t-test (comparing dichotomized variables) or analysis of variance for overall effect (comparing variables with more than two strata). *P<0.05, **P<0.01, n.s, not significant. ASPECT, Alberta Stroke Program Early CT; mRS, modified Rankin Scale; OTG, onset-to-groin; tPA, tissue plasminogen activator.

Impact of procedure time on hemorrhage

We then investigated the impact of procedure time on rates of postprocedural sICH in the full cohort and specific subpopulations studied in this paper. In the full cohort, procedure time (10 min increments) was associated with higher aOR of sICH (aOR=1.03, 95% CI 1.02 to 1.04) on logistic regression model controlled for the covariates in figure 1. Procedures exceeding 1 hour of procedure time were associated with 40% more odds of sICH (aOR=1.4, 95% CI 1.3 to 1.5). We then computed the interplay between the aOR for good clinical outcome (mRS score 0–2) and postprocedural sICH as function of procedure time. Procedure time was split into 10 min intervals and aORs were computed for each interval compared with those with a longer procedure time than the studied interval. Results for each subpopulation are shown in figure 3. The aOR for good outcome and sICH appear to intersect within 60 min of procedure time in all groups studied (figure 3). Beyond 90 min of procedure time, for each additional 10 min of procedure time the aOR for sICH exceeds that of good outcome at 90 days. These findings support the fact that the golden hour remains the trigger for evaluation of futility of further interventions.

Figure 3

Impact of different procedure time intervals on functional outcome at 90 days and postprocedural sICH. For each subgroup of patients, independent multivariate regression models were computed for impact of procedure time interval (10 min intervals) on good functional outcome at 90 days (mRS score 0–2) and sICH. Covariates controlled for are similar to those in figure 1. For each data point, an independent regression model is computed for the OR of the specific procedure time interval compared with those with higher procedure times. A series of models were computed for each procedure time interval. Reported are adjusted odds ratio (aOR) with 95% confidence intervals as error bars. A one-phase associated curve is fitted for each subgroup. Dashed line represents OR=1.0. ASPECT, Alberta Stroke Program Early CT; PT, procedure time; sICH,symptomatic intracranial hemorrhage; tPA, tissue plasminogen activator.

Discussion

Endovascular thrombectomy for acute ischemic stroke remains the standard of care for patients presenting up to 24 hours from onset. More recent randomized controlled trials (RCTs) have demonstrated safety and efficacy of EVT in previously excluded patient populations such as those with posterior circulation stroke15 26 and those with large core infarcts.10 20 Data from early RCTs and real-world experience have studied the impact of procedure time on EVT outcomes and demonstrated that procedures taking longer than 1 hour or requiring more than three attempts are associated with worse outcome and should alert neurointerventionalists to re-evaluate the futility of further attempts.7–10 20 Procedure time is defined as the time from groin puncture to successful recanalization or abortion of a failed procedure. In this work, we performed a comprehensive analysis of the effect of EVT procedure time on clinical outcomes in different subpopulations of patients with a stroke in the real world using a large-volume multicenter international registry. We showed that the golden hour of procedure time is applicable in all studied populations, including those with anterior and posterior stroke, proximal or distal occlusions, and small or large core infarcts, different age groups, bridging thrombolysis or EVT alone, and those presenting in the extended window from onset. We also demonstrate that the effect size of the impact of procedure time on outcomes is influenced by these covariates. with the most sensitive group being those with posterior circulation stroke.

When considering the location of thrombus, early EVT trials strictly included patients with internal carotid artery and proximal middle cerebral artery occlusions that encompassed the early definition of LVOs. After the success of the trials, real-world practice expanded those indications to include patients with posterior circulation stroke9 25 medium vessel and distal occlusions. The safety and efficacy profile of EVT was comparable in those patients to the early RCT results on LVOs.

In this work, we studied the sensitivity of outcome to procedure time in patients with different locations of thrombus. Even if the golden hour applies to these subpopulations, the impact of procedure time on outcome was the highest in patients with posterior circulation strokes in comparison with those with anterior circulation strokes. Patients with a posterior circulation stroke undergoing thrombectomy had nearly triple the odds of good outcome when the procedure was completed within 1 hour compared with 1.5 times higher odds for anterior circulation stroke. There are several possible reasons for this differential effect. Longer manipulation of the vertebrobasilar circulation during EVT is associated with higher risk of injury to perforators, and longer duration of ischemia to the brainstem.9 In addition, postprocedural hemorrhage, which could be a complication of longer procedure times, is likely to have a more devastating impact on outcome for the patient with posterior fossa stroke compared with the anterior middle cranial fossa.9 25 This is supported by the data from this work showing that a longer procedure time is associated with higher rates of symptomatic hemorrhage in this cohort. In contrast, patients with distal occlusions were less sensitive to longer procedures than those with proximal occlusions. This could be explained by the fact that distal occlusions require longer time to access the target vessels compared with proximal occlusions, and this time does not involve manipulation of the thrombus. In addition, the collateral pattern for distal occlusions is different from that of proximal occlusions and may be less sensitive to differences in reperfusion time.

Recent data supports EVT in select patients with large core infarcts.10 20 In this work, patients with different ASPECT scors did not showed significant difference in sensitivity to procedure time for both their clinical outcomes and the risk of sICH. Patients with larger core infarcts are often those with poor collateral circulation, advanced intracranial vascular disease, or those presenting at an extended time from onset. In this subset of patients, longer procedure time is likely to be secondary to advanced atherosclerotic disease. Notably, in the most recent RCT by Sarraj et al, the median procedure time in the low ASPECT group was 38 min with an IQR of 25–61 min19.

When patients were stratified based on whether or not bridging thrombolysis was administered, those who did not receive IV tPA prior to thrombectomy were more sensitive to longer procedure times with respect to their clinical outcomes. This might be explained by the effect of tPA on the microcirculation as well as its effectiveness at lysing and potentially preventing intraprocedure distal emboli. This finding was similar to that for patients presenting at longer duration from onset, where the sensitivity procedure time increased depending on the duration from onset. A possible explanation for this differential effect of procedure time in this setting is that the group with a longer presentation from onset has reached a steady-state collateral status, which allows their outcome to be less affected by faster reperfusion.

We also investigated the interplay between procedure time and different age groups with respect to clinical outcomes. The highest sensitivity to procedure time was noted in the elderly population (above 80). In elderly subjects, significantly longer procedure times are likely to involve higher rates of vascular injury, including sICH, as well as underlying atherosclerotic disease, which is not favorable for successful thrombectomy. In addition, prolonged procedures in this population could entail longer duration of anesthesia and hemodynamic stress on subjects with poor hemodynamic reserves.

Finally, we studied the interplay between procedure time and both functional outcomes and sICH in the different subgroups. When different covariates were accounted for in regression models, the OR of good functional outcome as well as those of sICH appear to equilibrate between 60 and 90 min of procedure time across all the studied populations. We also demonstrate that the golden hour is a universal cut-off point after which the odds of sICH start to match the odds of good outcome with successful completion of the procedure. Beyond 90 min, the OR for sICH starts to exceed that for good functional outcome. These findings have direct implications on neurointerventional practice in the real world where it can be unclear on how aggressive to be in recalcitrant clots or tortuous anatomy.

Limitations

Data from this work are curated from multicenter international registries without uniform selection criteria for intervention between centers that reflects real-world data. There is no randomization, but data are prospectively maintained at each center. Radiographic outcomes were self-reported and not independently adjudicated. Different thrombectomy techniques were not investigated independently in this study. Additionally, despite our effort to control for different variables that might influence the effect of procedure time on outcomes, additional variables could still confound these studies, including difficulty in navigating the aortic arch, need to perform femoral-to-radial conversion or the opposite, pursuing thrombectomy for distal embolization, and others. In addition, we did not collect data on tortuosity or clot composition in our registry, which could be investigated in future studies.

Conclusions

In this work, we performed a comprehensive analysis of the impact of procedure time on thrombectomy outcomes in different subpopulations of patients undergoing EVT. Given the extended indication for EVT with more recent trials, we demonstrate the universality of the golden hour effect where procedures lasting more than 1 hour are associated with worse clinical outcomes and higher rates of symptomatic hemorrhage. We also demonstrate that, specifically, patients with posterior circulation strokes, elderly patients, and those with large core infarcts are most sensitive to this effect. The results of this study should not deter from pursuing successful recanalization after multiple attempts, but will assist clinicians in determining when to be more aggressive or conservative in their approach when aiming for perfect angiographic outcomes in different subpopulations of patients with a stroke. However, the decision to abort the procedure remains at the discretion of the operator weighing all radiographic, clinical, and patient-specific factors.

Data availability statement

Data are available upon reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Work by the STAR collaboration is granted approval by the institutional review board at the Medical University of South Carolina under protocol Pro00090704. Informed consent was waived for this collaboration.