Trends in mechanical thrombectomy and decompressive hemicraniectomy for stroke: A multicenter study

Abstract

Background and purpose

Acute ischemic stroke has increasingly become a procedural disease following the demonstrated benefit of mechanical thrombectomy (MT) for emergent large vessel occlusion (ELVO) on clinical outcomes and tissue salvage in randomized trials. Given these data and anecdotal experience of decreased numbers of decompressive hemicraniectomies (DHCs) performed for malignant cerebral edema, we sought to correlate the numbers of strokes, thrombectomies, and DHCs performed over the timeline of the 2013 failed thrombolysis/thrombectomy trials, to the 2015 modern randomized MT trials, to post-DAWN and DEFUSE 3.

Materials and methods

This is a multicenter retrospective compilation of patients who presented with ELVO in 11 US high-volume comprehensive stroke centers. Rates of tissue plasminogen activator (tPA), thrombectomy, and DHC were determined by current procedural terminology code, and specificity to acute ischemic stroke confirmed by each institution. Endpoints included the incidence of stroke, thrombectomy, and DHC and rates of change over time.

Results

Between 2013 and 2018, there were 55,247 stroke admissions across 11 participating centers. Of these, 6145 received tPA, 4122 underwent thrombectomy, and 662 patients underwent hemicraniectomy. The trajectories of procedure rates over time were modeled and there was a significant change in MT rate (p = 0.002) without a concomitant change in the total number of stroke admissions, tPA administration rate, or rate of DHC.

Conclusions

This real-world study confirms an increase in thrombectomy performed for ELVO while demonstrating stable rates of stroke admission, tPA administration and DHC. Unlike prior studies, increasing thrombectomy rates were not associated with decreased utilization of hemicraniectomy.

Introduction

Acute ischemic stroke (AIS) is a leading cause of death and disability, with increasing global stroke burden. ¹ In 1995, the National Institute of Neurological Disorders and Stroke Study Group trial showed a 30% relative risk reduction with intravenous (IV) tissue plasminogen activator (tPA) within 3 h of AIS, becoming the standard of care. ² In 2008, IV-tPA within 4.5 h for most patients was incorporated into standard practice.^2,3 In 2013, three randomized trials showed no difference in outcomes between mechanical thrombectomy (MT) and IV-tPA.^4–6 However, in 2015, seven randomized trials established MT as the standard of care treatment for patients presenting with emergent large vessel occlusion (ELVO).^2,7–11 In early 2018, the benefit of thrombectomy was shown even in patients presenting up to 24 h.^12,13

Cerebral infarctions due to ELVO often are large volume and may cause malignant cerebral edema with intracranial hypertension. Mortality has been observed in up to 78% of patients with completed middle cerebral artery infarct.^14,15 In cases of malignant intracranial hypertension as a result of a completed large volume stroke, decompressive hemicraniectomy (DHC) can be performed as a life saving measure, although there is controversy over its utility for salvaging functional outcomes. ¹⁶ With increasing use of MT, many centers have anecdotally noted decreasing rates of DHCs at their centers, presumably secondary to smaller infarct volumes after successful revascularization. ¹⁷ Several nationwide database studies have supported this phenomenon;^18–21 however, this methodology has significant limitations and does not include data since the extended time windows seen after the 2018 trials.

This study was undertaken to quantify the real-world rates of stroke, MT, and DHC in comprehensive stroke centers across the country, and to compare trends spanning the landmark MT trials.

Material and methods

A retrospective cohort study was designed to evaluate the incidence of AIS and the rates of MT and DHC from 1 January 2013 to 31 December 2018. This time period was chosen to evaluate trends before and after the publication of landmark trials supporting the use of thrombectomy, even in extended windows of stroke onset. Eleven tertiary medical centers across the United States specializing in the delivery of stroke care provided stroke admission data during this review period. These centers were invited based on their participation in endovascular stroke registries and voluntarily agreed to participate. They represented centers that are actively engaged in tracking stroke cases and implementing endovascular treatments across diverse geographic locations. Wake Forest University School of Medicine served as the coordinating site with data contributions from all additional sites. Institutional review board approval was obtained at all participating centers. Individual patient consents were not obtained as only aggregate data was used.

The following data were collected from each participating institution: annual number of patients admitted for treatment of AIS, annual number of patients who received IV-tPA, annual number of patients who underwent MT, and annual number of patients who required DHC for malignant cerebral edema due to stroke. These numbers were identified by international classification of disease version 9 (ICD-9) and current procedural terminology (CPT) code from each institution. In order to avoid inclusion of thrombectomy or hemicraniectomy for indications other than acute large vessel occlusion and malignant cerebral edema, respectively, individual MT and DHC cases were then reviewed by each data center to ensure inclusion criteria were met. Total numbers for each category were sent electronically to the primary coordinating center where they were compiled into a master database. In this data collection, each annual incidence of diagnosis or procedure at each center was considered a data point. With four diagnosis/procedure categories across 6 years from 11 centers, a total of 264 data points were sought. Across sites there were 17 (6.4%) missing data points.

Statistical analysis

To examine change over time in the total number of stroke admissions, IV-tPA administration rate, MT rate, and rate of DHC, mixed models were performed. These models are preferable to repeated measures analysis of variance when seeking to control for within-site correlation among time points and missing data. An entire site’s data need not be removed from the analysis for not being complete. Both linear and quadratic model forms were considered. Fits were compared using Bayesian information criterion with linear proving to have the best fit for all four models. The intercept and slope estimates were allowed to vary across sites (random effects) with time as the only fixed effect. A p-value < 0.05 was considered significant.

Results

Total stroke admissions

Between 2013 and 2018, there were 55,247 stroke admissions across the 11 participating centers (Table 1). The annual median number of stroke admissions was 778.5 (SD 518.4; range 181-2667). The mean number of stroke admissions in 2013 was 681.9, compared with 1071.3 in 2018. Over the 6-year study period, there was no significant change in total stroke rates across the 11 centers (p = 0.11). We observed a trend at two centers toward increasing stroke admissions, likely due to institutional growth from expanded number of sites (Figure 1). Of note, there were missing data points at early time points which were accounted for with the utilization of a linear mixed model (Figure 1).

Table 1.

Total number of patients (percentage of total stroke admissions).

	2013	2014	2015	2016	2017	2018	Total
Sites contributing	9	11	11	11	11	11
Stroke admissions	6137	7604	9326	10287	10109	11784	55247
Sites contributing	8	9	9	10	10	10
Received IV-tPA	649 (10.6%)	772 (10.2%)	954 (10.2%)	1177 (11.9%)	1207 (11.8%)	1386 (11.8%)	6145 (11.1%)
Sites contributing	9	11	11	11	11	11
Thrombectomy	297 (4.8%)	327 (4.3%)	593 (6.4%)	759 (7.4%)	910 (9.0%)	1236 (10.5%)	4122 (7.5%)
Sites contributing	9	10	11	11	11	11
Decompressive hemicraniectomy	95 (1.5%)	72 (0.9%)	107 (1.1%)	130 (1.3%)	127 (1.3%)	131 (1.1%)	662 (1.2%)

IV: intravenous; tPA: tissue plasminogen activator.

Figure 1.

Total number of stroke admissions per site per year, showing no statistical increase in total stroke. Sites 2 and 9 increased number of sites over the study period.

IV-tPA

During the study period, a total of 6145 patients (11.1%) received IV-tPA across all 11 centers. The annual median number of patients receiving IV-tPA was 1065 (SD 281.77; range 649–1386). In 2013, the rate of IV-tPA administration was 10.6% of the total annual stroke admissions, compared with 11.8% of the total annual stroke admissions in 2018. During the 6-year study period, no significant change in the rate of IV-tPA administration was noted across all 11 institutions (p = 0.20) (Figure 2).

Figure 2.

Predicted intervention rates per year as percentage of total stroke admissions (all sites included). Only thrombectomy was found to have a significant change over time.

MT

During the study period, a total of 4122 patients (7.5%) underwent MT. The annual median number of patients undergoing MT was 676 (SD 359.68; range 297–1236). Over the 6-year study period, there was a significant increase in MT procedures (p = 0.002), increasing from 4.8% of all stroke admissions or a mean of 33 MTs annually per institution in 2013, to 10.5% of all stroke admissions or a mean of 112 MTs annually per institution in 2018 (Figure 2).

DHC

During the study period, a total of 662 patients (1.2%) underwent DHC for malignant cerebral edema due to AIS. The annual median number of patients who underwent DHC was 117 (SD 23.7; range 72–131). No significant change in the rate of DHC was observed across the 6-year study period (p = 0.860), stable from 1.5% of all stroke admissions or a mean of 10.6 annual DHCs per institution in 2013, to 1.1% of all stroke admissions or a mean of 11.9 annual DHCs per institution in 2018 (Figure 3).

Figure 3.

Decompressive hemicraniectomy administration rates per site per year as percentage of total number of stroke admissions.

Discussion

This study evaluates real-world stroke practice trends at 11 academic comprehensive stroke centers across the nation, representing over 55,000 acute stroke presentations over the 6-year period encompassing the renaissance of interventional stroke management. From 2013 to 2018, 11.1% AIS patients received IV-tPA, 7.5% received MT, and 1.2% underwent DHC. There has been a steady rise in the incidence of MT rates which mirrors the progression of interventional stroke trials during this time period (10.5% in 2018 from 4.8% in 2013), whereas rates of the other interventions saw no significant change (Figure 3). Most notably DHC rates have not changed significantly across these centers during this time period, contrary to prior nationwide database studies that have suggested a decreasing rate of DHC.

This observation is in contradistinction to recent studies using data from the Nationwide Inpatient Sample (NIS), as well as data from a single center in Germany.^19–21 Analysis of NIS data found similar rates of IV-tPA and MT (11.2% and 5.8%, respectively), but a decrease in DHC from 11.4% in 2012 to 4.8% in 2016. ¹⁹ Another analysis of NIS data showed a trend toward decreased DHC, from 7.07% in 2006 to 6.43% in 2016. ²⁰ The German study also showed a modest reduction in DHC rates: 2.8% in 2009 to 1.9% in 2017. One theory explaining the changes in rates observed by these studies is a progressive shift towards centralized consolidation of stroke care at tertiary stroke centers. Our comprehensive stroke study centers may have a practice pattern that differs from that of the conglomerate of US centers submitting data to the NIS, which may explain discrepancies in reported DHC rates. ¹⁹ The present study may better reflect this centralization of care as it includes data only from tertiary US comprehensive stroke centers. Furthermore, the present study is not subject to the same data limitations as national database studies, which include errors in billing and coding and an inability to audit rates that may be inconsistent with other data sources. Finally, due to a delay in accessing NIS data, the aforementioned studies were unable to study or capture changes in practice following DAWN and DEFUSE 3.^12,22

Initial trials indicated the benefit of DHC to reduce mortality in large volume stroke patients. ²³ DESTINY, DECIMAL, and HAMLET all concurred that DHC within 48 h significantly reduced mortality rates (absolute risk reduction ranging from 38% (HAMLET) to 52.8% (DECIMAL)).^24–26 However, improved functional outcomes were not demonstrated.^24–26 DESTINY II (DHC for age >60 years) ²⁷ and HeADDFIRST (DHC for patients ≤75 years) ²⁸ showed significantly higher survival rates in the surgical group with increased rates of modified Rankin Scale (mRS) scores 0–4, although there were also concomitant increases in mRS score 5 in the surgical group (28% versus 13%). ²⁷

There is a logical but unproven indirect relationship between MT and the development of malignant edema and need for DHC. Although no model has been shown to reliably predict malignant edema, factors known to be associated with its development are generally those reflecting a larger volume infarct. ²⁹ Involvement of more than one-third of the middle cerebral artery territory, frank hypodensity on computed tomography scan within 6 h of symptoms onset, early midline shift, and diffusion restriction volume of >80 mL on magnetic resonance imaging are all indicators of malignant cerebral edema. ² The major aim of MT is to salvage penumbra and decrease both the rate and volume of completed infarcts. It follows, therefore, that preventing a larger infarct volume would obviate the need for DHC.

In our anecdotal experience, the indication for DHC in the modern thrombectomy era is threefold: (a) completed infarction that precludes the opportunity for MT; (b) failed or incomplete revascularization; or (c) successful revascularization with pre-existing core infarct. Data from our real-world study show that even with an expanded time window to 24 h, only 10.5% of stroke patients underwent MT despite an estimated 40% of stroke due to ELVO. ³⁰ Failed revascularization, which can lead to large infarct volume, occurs in up to 10–12% of cases^7,9–11,13 Additionally, certain circumstances of incomplete recanalization, such as patchy, nonconfluent infarction, or residual occlusion of a nondominant inferior M2 division, can result in large volume infarct with significant herniation risk while not posing significant functional outcome impairment. These are both scenarios where concomitant MT and DHC may be indicated.

Over the last 5 years (2015–2020), following landmark randomized trials, there has been a renewed surge of intervention for AIS. In a study of US centers, MT incidence increased from one MT every 5 days in 2016 to more frequently than one MT every 2 days in 2018.^31,32 Neurointerventionalists report a significant increase in their tendency to intervene in acute stroke. ³³ Despite the stable incidence of stroke and utilization of IV-tPA, there is growing data supporting the use of MT in more patients, regardless of previous exclusionary factors such as age, stroke severity at presentation, ELVO location, and time from symptom onset.^34–39 These trends all point toward a culture of more aggressive treatment with MT for stroke, a culture which would also support an aggressive approach to performing DHC. However, a stable rate of DHC may suggest that although we are applying the same level of aggressive treatment to both MT and DHC, there are simply fewer patients requiring DHC due to the success of MT in salvaging tissue.

The present study strikes a balance between the small data limitations of a single center dataset and the confounders inherent in a national dataset based exclusively on billing codes. However, there are limitations to the current methodology, most notably its retrospective nature and limited number of participating centers. The numbers were initially collected from billing codes, which often initially overestimated stroke presentations or included DHCs performed for indications other than malignant cerebral edema due to AIS. Indications for DHC were reviewed, and those performed for reasons other than malignant edema were excluded. There were also several missing data points in the earlier years of the study, but these were accounted for statistically. An additional limitation of data collection based on billing codes is the inability to discriminate patients presenting with ELVO versus all ischemic stroke. In the absence of evidence of increased incidence of ELVO, increased MT incidence could be due to either higher rate of overall use or reorganization of stroke referral networks to funnel patients with ELVO to tertiary centers for the purpose of undergoing MT. Finally, there was some variability with CPT numbers as sites evolved and grew over time alongside increasing overall stroke rates—although this did not result in statistical significance. Because the data included in the present study are purely annual incidences in isolated categories, patients could not be tracked from diagnosis to intervention (or even which intervention a patient was eligible for). In a more granular analysis it would be beneficial to also examine the treatment experience of which patients undergo which combination of interventions. A prospective trial would be able to improve upon these limitations, but the current data is consistent across all participating centers. The next step in evaluating DHC in the era of MT will be to first perform randomized trials to understand the additive benefit of surgical intervention following MT on functional outcomes and then evaluate the current rate at which patients still do not arrive in time to salvage the penumbra, which leads to the development of malignant cerebral edema.

Conclusions

In this retrospective study analyzing data from multiple high-volume US comprehensive stroke centers, we were able to ascertain trends in MT from the failed MT trials in 2013, to the positive MT trials of 2015, to the extended window MT trials of 2018. In analysis of over 55,000 AIS patients, we found no significant change in rates of stroke admission, IV-tPA administration, or DHC for malignant cerebral edema. However, there has been a significant increase in rates of MT, which may be an indicator of overall more aggressive treatment of stroke at US tertiary academic centers. Given that only 10.5% of all AIS patients underwent MT in 2018, there is clearly an opportunity to continue to improve the interventional care of AIS. Although it remains unclear whether or not MT directly decreases the need for DHC, this study offers findings that diverge from those previously published on trends relating MT and DHC, thereby contributing a new perspective on the application of these procedures.