Article Text
Abstract
Background Debate continues as to whether patients with acute ischemic stroke with (suspected) large vessel occlusion benefit from direct referral versus secondary transportation.
Aims To analyze the change in early infarct signs, collaterals, and acute ischemia volume and their association with transfer time and functional outcome.
Methods We retrospectively analyzed consecutive transfers between 2013 and 2016 for patients with anterior circulation stroke transported from referring hospitals to our center as potential candidates for thrombectomy. Alberta Stroke Programme Early CT Scores (ASPECTS) were automatically calculated on external and in-house CT using the Brainomix e-ASPECTS software, and collaterals were assessed using the e-CTA tool. Functional status after stroke using the modified Rankin scale (mRS) was obtained.
Results 102 patients with CT scans both at the referring hospital and our center were identified. During patient transfer, e-ASPECTS declined by a median of 1 point (0–2). Functional outcome correlated with the change in e-ASPECTS (decline, n=54) (Spearman rs =0.322, 95% CI 0.131 to 0.482, p=0.001). The median image-to-image time was 149 min (IQR 113–190), but did not correlate with change in e-ASPECTS (p=0.754) and mRS score at 3 months (p=0.25). Preserved good collateral status assessed at the comprehensive stroke center was associated with better functional outcome (rs =−0.271, 95% CI −0.485 to −0.037, p=0.02).
Conclusions Patient transfer in a drip-and-ship network was associated with declines in e-ASPECTS associated with worse functional outcome. Image-to-image time did not influence this association, but worsening collateral status did.
- stroke
- ct angiography
- thrombolysis
- thrombectomy
Statistics from Altmetric.com
Introduction
There is an ongoing debate as to whether patients with an acute ischemic stroke and assumed large vessel occlusion should be transferred directly to an endovascular therapy (EVT)-capable center (‘mothership’), or if they benefit from being admitted first to an intravenous thrombolysis (IVT)-ready community hospital (‘drip and ship’), or if the interventionalist should drive to the remote hospital (‘drip-and-drive’).1 2 Even when applying prehospital triage scores, about 30% of patients with large vessel occlusion (LVO) would be missed and might require secondary transfers.3 4 The Alberta Stroke Programme Early CT Score (ASPECTS) has been established as an important predictor of EVT outcome.5–7 ASPECTS may thus be used as a selection criterion before patient transfer in order to reduce the number of futile secondary transfers. However, the threshold that would be applicable to withheld transfers is not known, as it is largely unknown whether ASPECTS changes to a clinically relevant degree during transport for evaluation of EVT in a drip-and-ship model.8 9 The current American Heart Association/American Stroke Association guideline recommends EVT in patients with ASPECTS ≥6,10 but the benefit of EVT when ASPECTS is lower (3–5) is currently being investigated in a randomized trial.11 Electronically calculated ASPECTS (e-ASPECTS) has been shown to be not inferior to scores obtained by neuroradiologists,12 and to be user independent and available for sites where experienced neurologists or radiologists are not available 24/7.
We aimed to analyze the change in early infarct signs according to e-ASPECTS in a real-world, drip-and-ship network of patients with acute anterior circulation stroke and its association with image-to-image time and functional outcome. Furthermore, acute ischemic stroke volume and collateral blood flow were assessed before and after transport.
Methods
Study design, setting, and patients
By retrospective screening of all adult patients with acute ischemic stroke (ICD-10, I63.x) referred to our Heidelberg University Hospital stroke center between 2013 and 2016, we identified all patients with acute anterior circulation stroke referred for evaluation of EVT. After 2016, a formal stroke allocation algorithm including prehospital triage was implemented. Patients with clinically suspected LVO according to the local physicians, telestroke consultation, or with proven LVO according to CT angiography (CTA) were transferred from 18 local community hospitals located within a maximum distance of 81 km around our center. We excluded patients without available non-contrast CT (NCCT) acquired both before transport and at our stroke center before thrombectomy, and patients with anterior cerebral artery stroke. From August 2012 to November 2014, MRI was our in-house standard for suspected LVO. Approval was obtained from the ethics committee of the Medical Faculty of Heidelberg.
Data acquisition
All diagnostic and treatment decisions were left at the discretion of the treating physicians. Decision to perform thrombectomy was based on local standard operating procedures (published previously13). Briefly, during the study inclusion period, it was recommended that thrombectomy should be withheld when ASPECTS was <6 or the measured infarct core >100 mL, but the final decision was left to the attending physicians. Information about medical history, stroke severity, and clinical course and the time of stroke onset (or last seen well in cases of unknown onset) were extracted from the digital hospital archive. Grade of reperfusion after EVT according to the modified Thrombolysis in Cerebral Infarction (mTICI) score14 was extracted from the final radiological reports; good reperfusion was defined as mTICI 2b/3. Functional status before stroke and after stroke was assessed using the modified Rankin scale (mRS). The follow-up outcome was obtained through rehabilitation reports, outpatient assessments, or a standardized interview by an unblinded investigator. If outcome data at 3 months were missing, the discharge mRS was carried forward.
Imaging analyses
Baseline NCCT before transport and the first NCCT at our comprehensive stroke center (CSC) were analyzed using the e-ASPECTS software (version 7.0/7.1; Brainomix Ltd, Oxford, UK). Briefly, e-ASPECTS uses machine-learning algorithms in combination with advanced imaging processing and image enhancement filters to reduce differences between scanners and imaging artifacts.5 Automatic 3D reconstruction is applied to correct for transformations, including rotation and tilt.5 6 All 10 ASPECTS regions are categorized as ‘abnormal’ (presence of early ischemic changes) or ‘normal’ using rigorous, statistical weighting tests, and one point is subtracted from a maximum of 10 for each abnormal region. Reflecting the fact that physicians are aware of the clinical syndrome, e-ASPECTS was adjusted for the clinically affected hemisphere.
Automated estimates of the acute ischemia volume on NCCT scans were also provided by the e-ASPECTS software for the affected hemisphere.
Presence of collateral blood vessels was quantified using the e-CTA tool (version 7.1; Brainomix Ltd).15 For craniocervical CTA analyses, contrast-enhanced images were automatically analyzed, and for each scan, a 4-point ordinal collateral status (CS) score was calculated (with 0=absence of collaterals, 1=collaterals filling >0% and<50%, 2= >50% and <100%, and 3=100% of the occluded territory).16
Statistical analyses
Categorical variables are presented using medians and interquartile ranges, and absolute and relative frequencies are reported. Continuous variables are presented using means and SD. We used image-to-image time, defined as the time difference between performance of the NCCT at the referring hospital and the NCCT at the CSC, to obtain a standardized estimate of the transfer time between the hospitals. The Shapiro-Wilk test was used to ascertain the distribution of data. To quantify correlations between the image-to-image time and differences in e-ASPECTS, acute ischemia volume, and functional outcome as well as collateral status, Spearman’s non-parametric correlation was calculated (rS ). To compare continuous independent variables, the Mann-Whitney test was used, and for dependent samples, the Wilcoxon test. The Fisher exact test was used to compare the proportions of demographics and clinical characteristics between patients with, versus without, decline in e-ASPECTS. For exploratory analysis of factors influencing poor outcome (defined as mRS score 3–6 at 3 months), we performed a multivariable binary logistic regression analysis (method: enter). Adjusted models include e-ASPECTS decline, performance of thrombectomy, and acute ischemia volume (at the CSC). All statistical tests were two-sided, and p values of <0.05 were considered statistically significant. If not indicated otherwise, analyses were conducted using IBM SPSS Statistics, version 25 (IBM SPSS, Armonk, New York, USA). This study was performed according to the Strengthening of the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies.
Results
During the 3-year study period, 102 patients with anterior circulation stroke shipped for evaluation of thrombectomy were identified, who received both NCCT scans at the referring hospital and our CSC (figure 1). The mean age of these patients was 74.3 years (SD 11.3), and 50% were women (table 1). The median image-to-image time was 149 min (IQR 113–190), with no difference between patients with (n=54) and without decline (n=48) in e-ASPECTS (table 1). During patient transfer, e-ASPECTS declined by a median of 1 point (0–2), resulting in a median e-ASPECTS at the CSC of 8 (IQR 6–9) compared with 9 (IQR 8–10) at the referring hospital (see figure 2 for an exemplary case). Patients without e-ASPECTS decline had lower baseline e-ASPECTS (9, IQR 6–10) compared with those with decline (9, IQR 8–10, p=0.039). Further clinical and radiological characteristics of the patients with and without e-ASPECTS decline did not differ, except for a higher rate of diabetes in the group of patients without decline (table 1). Thrombectomy was performed in 60 patients (58.8%). Patients in whom thrombectomy was performed had lower acute ischemia volumes both before transfer and at the CSC (online supplementary table) compared with patients not treated. The main reason reported for not performing thrombectomy was established infarct (61.9%).
Supplementary file 1
The outcome of patients with decline in e-ASPECTS during transport was worse than in those without decline (median mRS score 5 (IQR 3–6) vs 4 (IQR 3–5; p=0.014; table 1)). A decrease to <6 points at the CSC was observed in 18.6% of the cases and a decrease to <3 points in 1% of cases with initial e-ASPECTS ≥6. Functional outcome at 3 months was associated with a decrease in e-ASPECTS during transport (rS =0.322, 95% CI 0.131 to 0.482, p=0.001). A difference in e-ASPECTS also correlated strongly with the acute ischemia volume change (rS =0.813, 95% CI 0.704 to 0.893, p<0.001). During patient transfer, the acute ischemia volume increased by a median of 7.1 mL (IQR −4.9–26.6). The observed volume growth correlated with the functional outcome at 3 months (rS =0.377, 95% CI 0.195 to 0.532, p<0.001). In binary logistic regression analysis, acute ischemia volume at the CSC was predictive of poor outcome (OR=1.06, 95% CI 1.02 to 1.1, p=0.003), but not e-ASPECTS decline or performance of thrombectomy itself.
For the transfer time, we found no correlation between image-to-image time and e-ASPECTS at the CSC (rS =0.017, 95% CI −0.188 to 0.214, p=0.866) or any difference in the e-ASPECTS between the images taken at the different service levels (rS =0.031, 95% CI −0.157 to 0.243, p=0.754). Nor did the image-to-image time correlate with functional outcome at 3 months (rS =0.116, 95% CI −0.083 to 0.318, p=0.25) or with the change in the acute ischemia volume.
A subgroup analysis of patients who obtained electronic collateral status (n=49 at the referring hospital and n=74 patients at the CSC) showed that, after transfer, patients had a worse collateral status than before (median CS at referring hospital 3 (IQR 2–3) vs 2 (IQR 2–3), p=0.033). All patients with improved or stable CS (n=15) received intravenous thrombolysis before/during transport, in contrast to 66.7% (n=8) with collateral worsening (p=0.028). Collateral status obtained at the referring hospital did not correlate with the functional outcome at 3 months (rS =−0.149, 95% CI −0.415 to 0.131, p=0.316), but better collateral status assessed at the CSC was associated with better functional outcome (rS =−0.271, 95% CI −0.485 to −0.037, p=0.02).
Minor variance in e-ASPECTS was expected owing to different scanner calibrations, local image preprocessing, and patient positioning, but an increase of >3 points was seen in three patients. Individual patient analysis showed that patient positioning during the CT scan (lateral head deviation and rotation) and differences in gantry tilt/head holder angulation were the crucial factors (online supplementary figure 1). We found no difference in slice thickness of NCCT scans between the referring hospitals and our CSC (median 4 mm; p=0.913); analyses were repeated using thin reconstructions with no differences in e-ASPECTS calculation. In addition, administration of radiocontrast during CTA at the referring hospital was not associated with later increases in e-ASPECTS (p=0.28). A sensitivity analysis restricted to patients without any increase in e-ASPECTS during transfer showed no major differences in reported associations as compared with the entire group of patients (online supplementary figure 1).
Discussion
Our main findings are that (1) in half of the patients shipped for evaluation of EVT, the e-ASPECTS declined and (2) a clinically relevant increase in acute ischemia volume was found, which were both associated with worse functional outcome. Collateral status also worsened, but only in patients not treated with IVT before transfer.
Baseline e-ASPECTS were lower in patients without subsequent decline than in those with decline. At first sight this is counterintuitive, but this observation might be explained by dynamically changing infarct growth kinetics, potentially reflecting a previously described lack of correlation between time and acute infarct volume explained by variations of perfusion.17 Likewise, the image-to-image time did not correlate with a decline in e-ASPECTS across all e-ASPECTS levels or with ischemia volume growth, independent of the transportation time, a finding similar to that described by Boulouis et al.18 From a pathophysiological point of view, individual collateral status rather than transfer time alone influences the infarct growth dynamics. Collateral flow is highly individual, and it cannot be predicted when collateral blood flow will fail. Collateral status before transport was numerically worse in patients with subsequent e-ASPECTS decline (median CS 2 vs 3), although performance of CTA at both the referring hospital and the receiving center was not mandatory and differences did not reach statistical significance. However, preserved good collateral status at the CSC was associated with better functional outcome, consistent with findings from randomized trials of mechanical thrombectomy confirming the importance of preprocedure collateral status on treatment effect and functional outcome.19 Collateral worsening was seen only in patients not treated with IVT before transport, while all patients with stable or improved collaterals had received IVT. IVT might have resolved small distal thrombi and improved the microvasculature, but these findings need to be replicated in a larger sample.
With a median of 2.5 hours, the image-to-image time was long, but within the range reported in previous studies (1.4–3.2 hours).18 20 21 Generalizing our results to systems with even longer transfer times might be inappropriate. Recently, however, Boulouis and colleagues found that the image-to-image time could not predict a decrease in ASPECTS to <6 in patients with initial ASPECTS ≥6 at the referring hospital, which was observed in 19.6% of transferred patients.18 In our study we similarly found that 18.6% had a decrease to <6 points, but only one patient dropped to <3 points. Currently, an ASPECTS of 3 seems to be considered as the lower bound of the range in which EVT is more effective than IVT alone or no therapy.22 Nevertheless, data for the range of 3 to 5 are scarce, and inclusion of patients in randomized trials such as TENSION is necessary.11 Importantly, according to our study, the number of futile transfers—that is, in which e-ASPECTS declines to levels not allowing for EVT, is dramatically decreased if the lower threshold is modified from 6 to 3.
In univariate analysis, the presence of diabetes was higher in patients with no decrease in e-ASPECTS—a finding that needs to be replicated in a larger sample, as a history of diabetes is generally associated with worse collaterals and thus more rapid infarct progression.23
Our analysis showed that the ASPECTS is affected by correct head positioning and correct scanner calibration for brain examination. While minor differences in scoring with both e-ASPECTS and clinician-defined ASPECTS cannot be avoided owing to differences in image features,12 this did not explain the improvement in ASPECTS >3 points seen in a minority of cases after transfer. Despite software-based corrections, extensive lateral head positioning towards one side of the scanner, incorrect tilt of the gantry/head holder, and rotation of the head within the axis, resulting in poor image quality, were crucial factors.
Limitations of our study include the smaller number of CTA scans performed at the referring hospitals in comparison to our CSC, thus limiting analyses of the initial collateral status. Furthermore, the non-standardized performance of CTA scans introduces a potential selection bias. Although the decision to transfer patients was not determined by ASPECTS within our network and only very few patients were not evaluated for EVT in the drip-and-ship service, we cannot exclude the possibility that some patients were not transferred, which would have introduced a selection bias. The strength of our study is that we collected real-world hospital data in an acute stroke care setting with standardized, reader-independent imaging analyses. Inclusion of referring hospitals at different service levels enhanced the generalizability of the results. Increasing use of telestroke concepts might enhance the preselection of patients to avoid futile transfers.24 25
In conclusion, patient transfer in a drip-and-ship network was associated with declines in both e-ASPECTS and infarct growth, associated with worse functional outcome. The transfer time alone could not explain the increase in ischemia volume and decline in e-ASPECTS, but during transfer, collateral status worsens, and preserved collateral status at the CSC was associated with better functional outcome. In addition to the demand for further investigation of maintaining collaterals, our data underscore the need to further evaluate prehospital triage concepts and to avoid secondary transports by direct referrals to stroke centers.
References
Footnotes
SN and CG contributed equally.
Contributors Study concept and design: JCP, CG, SN. Acquisition, analysis or interpretation of data: all authors. JCP wrote the first draft of the manuscript. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: JCP, NM, CG.
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 JCP has received consultation fees and travel expenses from Akcea, Boehringer Ingelheim, and Pfizer. CH received consultation fees and travel expenses from Brainomix Ltd. MM reports consultation fees from Medtronic, MicroVention, Stryker, and grants/grants pending from Balt, MicroVention (money paid to the institution), and payment for lectures including service on speaker bureaus from Medtronic, MicroVention, Stryker. PAR received travel support and lecture fee from Boeheringer Ingelheim, Daiichi Sankyo, Pfizer, Bayer. SN has received consulting fees from Brainomix and Boehringer Ingelheim, and lecture fees and travel expenses from Medtronic and Pfizer.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data that support the findings of this study are available from the corresponding author upon reasonable request.
Patient consent for publication Not required.