Article Text

Download PDFPDF

Original research
Impact of landmark endovascular stroke trials on logistical performance measures: a before-and-after evaluation of real-world data from a regional stroke system of care
  1. Alexander Marco Kollikowski1,
  2. Franco Amaya1,
  3. Guido Stoll2,
  4. Wolfgang Müllges2,
  5. Michael Klaus Schuhmann2,
  6. Mirko Pham1
  1. 1 Department of Neuroradiology, University Hospital of Würzburg, Würzburg, Germany
  2. 2 Department of Neurology, University Hospital of Würzburg, Würzburg, Germany
  1. Correspondence to Dr Alexander Marco Kollikowski, Department of Neuroradiology, University Hospital of Würzburg, Würzburg 97080, Germany; kollikowsk_a{at}


Background and purpose Endovascular treatment of large-vessel occlusion stroke often necessitates patient transfer by a twin-track approach: endovascular thrombectomy (ET) in endovascular-capable facilities preceded by intravenous thrombolysis in primary stroke centers. We tested the open hypothesis that recent landmark trials on ET had any significant effect on logistical performance measures among different modes of admission.

Methods We retrospectively categorized 250 patients who presented at our institution as: (A) primarily admitted or transferred from (B) inner-city and (C) regional hospitals. The period from May 2015 to June 2017 was compared with the preceding period of August 2009 to April 2015 with respect to real-life transfer distances and sectional time metrics from symptom onset to angiographic recanalization.

Results Onset-to-recanalization time decreased in the primary admission path, whereas delays persisted for inter-hospital transfer: (A: 261 min (210–315) vs 198 (167–264) P<0.0001; B: 257 (214–306) vs 265 (199– 360) P=0.566; and C: 371 (322–415) vs 346 (307–405) P=0.559). Onset-to-recanalization time was negatively correlated with recanalization success (mTICI; r=-0.4195 P<0.0001). The rate of secondarily referred patients (26% vs 48% P=0.0004) and off-hour presentation (36% vs 44% P=0.004) increased, as did the catchment area (C: 52.2 km (30,4–64,5) vs 64.4 (43,2–78,9) P=0.032). Improvement in door-in-door-out time at the referring hospitals (C: 113 min (30) vs 86 (29) P=0.0236) did not translate into reduced total referral times or the accelerated initiation of ET.

Conclusion Recent landmark trials already led to a considerable streamlining of ET workflow if patients were directly admitted. Prehospital time management and triage seem to be the major determinants of optimization.

  • intervention
  • stroke
  • thrombectomy

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Recanalization by endovascular thrombectomy (ET) in conjunction with intravenous (IV) recombinant tissue plasminogen activator (rt-PA) if possible, has evolved to represent the standard of care for patients with embolic large-vessel occlusions (LVO) of the anterior cerebral circulation.1 The treatment effect of successful recanalization of LVO is potentially very strong, however this effect is time-dependent: an estimated number of 1.9 million neurons and 14 billion synapsis are lost each minute without restoration of physiological blood flow.2 3 The empirical estimate demonstrating this time dependency is convincing, with greatly reduced functional independence from 64.1% to 46.1% (symptom onset-to-reperfusion time of 180 min and 480 min, respectively) as assessed by the modified Rankin Scale (mRS 0–2).4

While rt-PA for intravenous thrombolysis in acute ischemic stroke (AIS) is widely available in community or primary stroke center hospitals, its recanalization success is limited by the occlusion location in that LVOs are particularly associated with low rates of complete recanalization when treated only by intravenous rt-PA: It could already be shown during the early Phase II evaluation of IV thrombolysis that recanalization rates are as low as, e.g., internal carotid artery 8%, middle cerebral artery section M1 26% and M2 35%, respectively.5 By contrast, the average rate of full recanalization for ET is 71%, as demonstrated in the HERMES (Highly Effective Reperfusion evaluated in Multiple Endovascular Stroke Trials collaboration) metaanalysis based on individual patient data from the largest five randomized controlled trials (RCTs).6

To facilitate broad access to ET, efforts are increasing to implement a concept of triage and transfer into existing stroke systems of care, some of which were initially designed for the decentralized and facilitated provision of rt-PA to improve onset-to-needle times as well as early availability of rt-PA treatment. Despite existing clinical stroke scale cut-off points and neurological assessment tests to triage patients with LVO to the nearest comprehensive stroke center more quickly,7 there is a pressing need for greater effectiveness. Although current data on ET suggest good functional outcome both for patients along the primary and secondary referral paths, delays in symptom-to-imaging time and time to recanalization remain highly problematic factors as compared with directly admitted patients.8 9 Although smaller observational studies have suggested that these delays do not translate into significant differences in outcome (when only the ET-treated patients are compared), the largest observational study based on prospective registry data could clearly demonstrate that delays in the secondary referral paths were significantly associated with adverse clinical effects.10

Here, we investigated whether key measures of logistical performance in the primary or secondary referral path for ET have changed over recent years. In particular, we tested the open hypothesis of whether any significant changes occurred between the period of May 2015 to June 2017 (after the publication of landmark RCTs) and the preceding period of August 2009 to April 2015 (before landmark trials).


This study was approved by the ethics committee of the Medical Faculty of the University of Würzburg. We retrospectively categorized the following three groups of acute anterior circulation ischemic stroke patients who underwent ET at our institution between August 2009 and June 2017: patients who were either: (A) primarily admitted (direct admission) or transferred from (B) inner-city hospitals (inner-city transfer – within the city limits) and (C) regional facilities (long-distance referral – beyond the city limits).

Covering a population of approximately 1.4 million people,11 the regional stroke system of care comprises three tertiary medical centers, four primary stroke centers, and seven community hospitals for which our institution serves as a hub in terms of teleneurological and teleradiological diagnosis, assessment of eligibility for ET, coordination of referral, and subsequent mechanical thrombectomy with 24 hours/7 day access to the angiography suite. In practice, associated hospitals within the stroke system of care use an emergency hotline to contact the CSC stroke fellow and neuroradiologist in service, thereby discussing essential clinical data such as medical history, symptom onset, degree of severity assessed by National Institute of Health Stroke Scale (NIHSS), and previous imaging findings (CT/CTA/MRI). Access to images is guaranteed via a teleradiological communication system with the associated hospitals.

If indicated, patients were transferred with previous rt-PA administration. Depending on primary admittance (A) or secondary referral (B and C), patients with suspected stroke received a standardized imaging protocol prior to ET to exclude hemorrhage or extensive infarction equivalent to ASPECTS<5.12–14 Protocol (A) comprised a cranial non-contrast CT and CT-angiography, which could be complemented by a CT-perfusion scan (depending on the time passed after symptom onset and CT angiography findings). In the case of secondary referral (B and C) either: non-contrast CT was performed if external CT angiography images were available and/or IV thrombolysis was administered; or protocol (A) was applied analogously.

In the period reviewed, patients with proven occlusion of the ICA, ICA bifurcation, M1, dominant M2 branches, and subsequent treatment by thrombectomy in our center were included. To homogenize the cohort of patients, patients with unclear time of symptom onset (eg, wake-up stroke), air transfer, and prolonged time interval from symptom onset to groin puncture (>6 h) were excluded. The collected data comprised age at presentation, sex, mode of admission (A/B/C), time of symptom onset, time of presentation (CSC), occlusion location, previous IV rt-PA administration, NIHSS (CSC), and ASPECTS (CSC). External and CSC imaging times were stored by the respective picture archiving and communication system. Recanalization success was assessed by the modified treatment in cerebral ischemia scale (mTICI)15 and determined by finding a consensus between two authors of this study experienced in interventional stroke treatment and diagnostic cerebral angiography (AMK/MP). The overall treatment process from symptom onset to recanalization was assessed by using real-world transfer distances and sectional time metrics of which the latter were further subdivided for calculational handling and clarity. According to the available medical records, we performed geocoding, calculation of street-level travel distances as well as traffic- and daytime-dependent estimation of transfer times between site of symptom onset, transferring hospital, and CSC using Google’s Distance Matrix Application Programming Interface (Mountain View, CA). For this retrospective study, a prospective observation and documentation of the time of initial hospital presentation was not possible. Given the assumption that AIS patients undergo neurological examination immediately after hospital presentation and immediately prior to the CT scan, the time of hospital presentation and the documented time of image acquisition were considered equal in order to be able to obtain a reliable estimate of early delays within the respective facility. Based on these estimates, we defined the calculated time interval from arrival to discharge at the referring hospital as the door-in-door-out (dido) time. The time interval from image acquisition at the referring hospital to image acquisition at the CSC was defined as the total referral time (figure 1).

Figure 1

Time and distance components of secondary referral (left) and primary admission (right) paths. CT, computed tomography; CTA CT angiography; MRI, magnetic resonance imaging; PSC, primary stroke center; dido, door-in-door-out time; CSC, comprehensive stroke center; mTICI, modified treatment in cerebral ischemia scale. The graphic content of this figure was freely available and free to be adapted under the creative commons attribution license (

Statistical analysis

We used standard descriptive statistical measures including mean values±SD deviation or median with IQR for numerical data or absolute and relative frequency distribution for categorial variables. Depending on the parametric or nonparametric distribution of data, inter-group comparison was performed using Student’s t test or Mann– Whitney U test for continuous variables, χ2 test or Fisher’s exact test for categorical variables, and the Kruskal– Wallis test for continuous measures followed by Dunn’s post hoc test. A two-sided P-value<0.05 was considered statistically significant. Correlation between mTICI and time to recanalization was analyzed with the Spearman coefficient. Statistical analysis was performed using GraphPad Prism (GraphPad Prism 6.0; GraphPad Software, La Jolla, CA).


In the overall time period between August 1 2009 and June 30 2017, 313 patients with proven LVO received ET at our institution. Of those, 250 (80%) were included because of the aforementioned criteria.

The details of relevant baseline demographic, clinical, and radiological parameters are given separately for May 2015 to June 2017 and the preceding period of August 2009 to April 2015 in table 1.

Table 1

Baseline characteristics, CSC

It is noteworthy that patient age and the rate of secondary referrals has significantly increased over time in the overall cohort. Mean age at intervention increased from 70 (±14) to 73 (±13) years (P=0.041). The relative frequency of secondarily referred patients rose from 26% to 48% (P=0.0004) of which the largest share was accounted for by long-distance referrals in each period (15% vs 36%). Consistent observations were made with regard to off-hour presentation which collectively rose from 36% to 44% (P=0.004). While the relative frequency of directly admitted patients decreased (A: 80% vs 44%), the share of patients being referred either from inner-city hospitals (B: 3% vs 12%) or regional facilities (C: 17% vs 42%) increased. A clear but non-significant trend toward decreased levels of stroke severity was found (NIHSS 24 (±14) vs 18 (±6) P=0.315). The proportion of patients receiving IV rt-PA before ET remained without significant change, as remained also the principal radiological parameters: the proportions of LVO target sites (ICA, M1, M2, tandem occlusion), median ASPECTS score (8 (7–10)) vs 8 (7–9) P=0.122), and recanalization success (mTICI 2b/3) (65% vs 72% P=0.646).

A step-by-step approach to the entire logistic chain from symptom onset to recanalization in terms of distances and time metrics is provided in table 2.

Table 2

Distances and time metrics

Inter-group comparisons for each period are given separately (online supplementary table 3). According to the respective transfer process, results are specified for each group (A/B/C) and time period (August 2009 to April 2015 vs May 2015 to June 2017). Median time (min) from symptom onset to initial imaging revealed significant improvement over time for patients being directly admitted to the CSC (A: 85 (66.5–107.5) vs 74 (58–93.5); P=0.013), whereas a non-significant trend toward improved onset-to-primary imaging times was seen for groups B and C. There were no significant inter-group differences when each period was regarded alone. Consistent for all groups, we observed a significant period-dependent reduction of transportation distances (km) between site of symptom onset and initial hospital (A: 17.8 (11.1–29.5) vs 13.1 (9.9–21.6) P=0.026; B: 19.9 (11.9–32.2) vs 8.5 (6.4–16.3) P=0.076; C: 17 (13-79) vs 15.8 (7.9–23.3) P=0.007). While there was no baseline difference between inner-city-transfers and long-distance referrals on dido-time and total referral time (min) for each time period, a significant period-dependent reduction of dido-time was shown for remote facilities (C: 113 (±30) vs 86 (±29) P=0.024). CSC catchment area (km) in terms of long-distance referrals significantly increased (C: 52.2 (30.4–64.5) vs 64.4 (43.2–78.9) P=0.032) in conjunction with a nearly significant reduction of onset-to-(secondary) imaging times (CSC) (C: 250 (208–286) vs 216 (185–238) P=0.066). Inter-group comparison of each period collectively showed significantly reduced picture-to-puncture times when the secondary referral paths were compared with primary admission. A statistically relevant period-dependent reduction of onset-to-puncture times was only seen if patients were directly admitted (A: 158 (132–200) vs 133 (113–156) P=0.0002). In this regard, primary admittance to the CSC was associated with a considerable period-dependent reduction of onset-to-recanalization times (A: 261 (210–315) vs 198 (167–264) P<0.0001), whereas median times from symptom onset to puncture and recanalization were generally significantly increased in the case of secondary referral. With respect to the overall treatment process, regression analysis for single data points showed a significant correlation between the time span from symptom onset to recanalization, and recanalization success as assessed by mTICI (r=-.4195 p<0.0001) (figure 2).

Figure 2

Time-dependence of recanalization grade. Onset-to-recanalization time was negatively correlated with recanalization success as assessed by mTICI (r=-0.4195 P<0.0001). The 95% confidence interval is given as short dotted lines.


The present study provides relevant real-life information regarding the as yet undetermined logistic implementation of up-to-date evidence-based endovascular stroke care into existing stroke networks. Based on the dependence of good clinical outcome on time to recanalization and a potentially very strong treatment effect,3 the timely initiation of ET as the central treatment step in LVO patients is one major procedural goal of acute stroke care.1 As a comparative workflow analysis of successive landmark RCTs illustrates, enormous time savings concerning onset-to-reperfusion times resulted from technical improvements (ET) or center-related efficiency gains, whereas inefficiencies in triaging and external workflow remain highly problematic factors.16–18 Therefore, we conducted a comprehensive analysis of key measures representing the complex logistic chain of endovascular treatment of AIS. In particular, it was analyzed to what extent the publication of the five landmark RCTs in this field in 2015/2016 have influenced the evolution of real-life time and distance measures within a regional stroke network.

With regard to the temporal components of the logistic chain, we observed a significant period-dependent reduction across the sectional time metrics which were most pronounced for primary admitted patients. While there were no relevant inter-hospital time differences between median time from symptom onset to imaging-based initial diagnosis in the post-RCT period (min, A: 74; B: 77; C: 77), considerable delays between symptom onset and final recanalization resulted from inter-hospital patient transfer and aggravated with increasing distance to the CSC (min, A: 198; B: 265; C: 346). In spite of an at least a partly significant post-RCT reduction in procedural treatment delays as assessed by dido-time (min, C:113 vs 86), the overall transfer process including ground transfer remained largely unaffected and resulted in mean treatment delays of B: 102 vs 95 min and C: 154 vs 140 min, respectively. Likewise, there were no significant changes concerning the time interval from initial imaging (PSC) to angiographic recanalization. Together, these data support persistent shortcomings in the speed of stroke recognition, the decision-making process concerning the initiation of patient transfer and, potentially to a lesser extent, the transportation path itself. When comparing the cohorts of primary admitted to referred patients, we observed significantly reduced picture-to-puncture times (CSC), suggesting center-related work-flow optimization, probably due to effective prenotification.

Regional stroke networks are facing tremendous challenges in accelerating or re-structuring the referral process to the nearest CSC since time savings along the early phase of stroke recognition and logistics chain may greatly influence stroke outcome.19 In line with the HERMES meta-analysis, a recent analysis of the STRATIS registry led to a deeper understanding of the intertwining of reduced functional outcome with increasing time to treatment.6 10 Similar to our findings, inter-hospital transfer prior to ET was identified as the major source of treatment delay, leading to a comparable lag time of 116 min (mean) while simultaneously lowering the chances of excellent clinical outcome (47.4% vs 38.0%) and functional independence (60.0% vs 52.2%).10 Comparing both periods, our results indicate that the observed efficiency gains concerning facilitated direct access to endovascular care aggravate the discrepancies between the modes of admission in terms of resulting recanalization delays (picture-to-recanalization time CSC (A) vs PSC (B/C)).

The complexity of this yet insufficiently solved logistical problem is due to the fact that ET not just simply replaces IV thrombolysis as stated in the current guidelines for the early management of AIS, according to which ET as well as rt-PA treatment should be initiated as quickly as possible (class of recommendation I, level of evidence A).1 Therefore, assuming emergency medical services prenotification and involvement of both PSC and CSC, optimized stroke process time metrics were suggested by the Society of Neurointerventional Sergery (SNIS), including door to CTA interpretation of <20 min, door to IV rt-PA of <30 min, CSC door to puncture of <60 min, CSC door to recanalization of <90 min, and PSC picture to CSC puncture of <90 min.20 In our analysis, only CSC-related metrics complied with these time requirements. However, it must be noted that such time metrics predominantly focus on easily assessable inner-facility process optimization while excluding the overall outcome-relevant no-flow time (time interval from onset to recanalization),3 actual decision-making process between initial presentation at a non-ET-capable facility and discharge (dido-time)21 as well as the regional infrastructure. As indicated, picture-to-puncture time was initially defined as the time interval between external baseline CT scan and groin puncture at the CSC. Capturing a broad continuum of stroke care, delays of >90 min significantly reduce the likelihood of good outcome among patients receiving ET.22 Since we tried to cover a wider time span including sectional time metrics starting from symptom onset, picture-to-puncture time was solely used to characterize the CSC-related workflow. Onset-to-puncture time most closely matches the picture-to-puncture time interval in its wider sense, again showing no relevant changes for group B and C in the post-RCT era. By contrast, onset-to-puncture time significantly improved if patients were directly admitted, leading to a substantial time saving of 133 min compared with long-distance referrals.

Interestingly, these delays were clearly associated with poorer recanalization: r=-0.4195 P<0.0001. This observation is consistent with the current understanding of declining collateral flow before recanalization where poorer collateral status itself is also a strong predictor of poorer recanalization rates (ie, higher rates of mTICI <2b/3).23 24

Since recovery in acute ischemic stroke is associated with arterial recanalization grade and speed, shortcomings in recanalization affect outcome.25 In our study, the delay which was associated with mTICI-deterioration could, at least to a great extent, be attributed to the local PSC and community hospital density. Geocoded probability models for outcome-based optimized transportation in proximity to a CSC suggest that external thrombolysis prior to the transfer for ET (drip and ship), in contrast to primary admission (mothership), is beneficial only when door-to-needle times of <30 min are achievable.26 If, however, all intervals are optimized or a door-to-groin puncture time of <60 min is achieved (while holding other time metrics constant), shifting takes place toward superiority of the mothership concept.27 Using the Rapid Arterial oCclusion Evaluation (RACE) scale prehospital screening tool, the ongoing randomized RACECAT trial (Direct Transfer to an Endovascular Center Compared with Transfer to the Closest Stroke Center in Acute Stroke Patients With Suspected Large Vessel Occlusion) is currently evaluating which of the opposing approaches offers a better outcome (mRS at 90 days) in AIS patients with clinically suspected LVO.28

Another major finding of our study was the substantial increase in the proportion of secondary referrals from 26% to 48%, of which the largest share was accounted for by long-distance referrals. Off-hour presentation affected 44% of all LVO-patients, showing an increase by 12% as compared with the pre-RCT era. In parallel, the catchment area of all primary hospitals tended to decrease, whereas the CSC catchment area regarding long-distance referrals considerably expanded by 12.2 km. Although some studies rule out adverse off-hour effects on timely initiation of ET, rt-PA administration and outcome,29 30 the particular trends since the publication of the landmark RCT have not been studied yet. Since imaging-based patient selection recently extended the therapeutic window of ET up to 16 to 24 hours after stroke onset,31 32 our findings raise the general question whether the operational actionability of current stroke networks meet the future needs of a further increasing number of AIS patients with suspected LVO or if restructuring is needed due to a considerable absorption of personnel resources and potential daytime-dependent performance shortfalls. Finally, it is noteworthy that besides structural and logistical efforts also novel supportive pharmacological strategies are highly desirable which are capable of sustaining collateral flow to the penumbra to avoid, or at least to delay, progressive infarction especially along the paths of secondary referral.

There are several limitations to this study. It was intended and carried out as a retrospective single center study. The participating hospitals were heterogeneously located in peri-urban, urban, and rural areas, which limits the external generalizability of our findings to other regions with different landscapes and infrastructure.


Our results deliver strong indicators showing that the recent landmark RCTs in the field of ET already led to a considerable shortening of the overall time interval between onset and recanalization, as well as particular subintervals if patients were directly admitted to the CSC. The rate of secondary referrals and off-hour presentation significantly rose in parallel, as did the CSC catchment area. By contrast, the time metrics of the secondary referral paths remained largely unaffected. Therefore, prehospital management and triage seem to be the major determinants of AIS-workflow optimization in areas in close proximity to CSCs.


We thank Dr Guenthner-Lengsfeld and Dr Bison for their support in data acquisition and patient handling.



  • Contributors AMK and MP conceptualized the study. AMK performed data acquisition, analysis, and interpretation. AMK and MP co-wrote and edited the first draft of the manuscript. FA, GS, WM, and MKS provided support at all stages and critically reviewed and/or revised the manuscript for significant intellectual content. All authors have contributed to production of the final version of this manuscript.

  • Funding AK/MP/GS were supported by a research grant from the Deutsche Forschungsgemeinschaft (DFG CRC/TR B02).

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval Ethics Committee of the University Hospital of Würzburg.

  • Provenance and peer review Not commissioned; externally peer reviewed.