Browse the corpus

Introduction Stroke is expensive in terms of its personal, societal and financial impact. The clinical benefit of stroke treatments is usually evaluated according to the functional outcome measures assessed at least three months after stroke, when most of the acute recovery has occurred. The spectrum of stroke outcomes can be assessed using the mRS,1,2 which is the most prevalent outcome measure in published trials across recent decades. The 90-day mRS is also the recommended primary outcome measure in acute stroke trials by the European Stroke Organisation (ESO) Outcomes Working Party.1,2 The treatment of stroke is complex and costly with effective treatments including stroke unit care, intravenous thrombolysis with recombinant tissue plasminogen activator and most recently, thrombectomy using stent retriever devices.3,4 Implementation of new treatments requires the assessment of both cost and outcomes in relation to alternative available interventions or current practice using cost-effectiveness analyses. Reliable cost data relating the mRS by category would be valuable to the wider stroke community when undertaking these forms of health economic evaluations.

atments requires the assessment of both cost and outcomes in relation to alternative available interventions or current practice using cost-effectiveness analyses. Reliable cost data relating the mRS by category would be valuable to the wider stroke community when undertaking these forms of health economic evaluations. The collection of robust data for economic evaluations may be complex and time-consuming, increasing the expense of trials. Therefore, to include cost-effectiveness evaluations as part of stroke trials can be challenging, and add to responder burden through the need for additional questionnaires. Quantifying the cost of a chronic, disabling condition such as stroke is complicated since, to provide a full picture of the cost impact to society, it is important to capture the direct costs of hospital care, as well as the direct and indirect costs over the longer term, including lost productivity. Having reliable estimates of costs by functional outcome that could be applied in cost-effectiveness studies would facilitate the ability of investigators to perform these important evaluations more often.

osts of hospital care, as well as the direct and indirect costs over the longer term, including lost productivity. Having reliable estimates of costs by functional outcome that could be applied in cost-effectiveness studies would facilitate the ability of investigators to perform these important evaluations more often. We have undertaken a systematic review of the current literature investigating the relationship between costs of stroke and functional outcome as measured by the mRS as a basis for informing the field and understanding the evidence base that may be available for cost-effectiveness evaluations where mRS data have been captured. Through the assessment of the literature, the ESO aims to eventually develop practical guidance for the integration of health economic data collection in future studies. By identifying and reporting current information on the costs for each mRS category, these could then be applied in decision-analytic simulations or estimations of the potential cost-effectiveness of new interventions in stroke, where primary collection of cast data has not been possible. Methods We performed a systematic review of the published literature on studies where the costs of stroke by mRS category were reported. To guide the systematic review, we applied the principles of the PRISMA statement (Appendix 2). We reviewed publications from 1 January 2004 to 13 February 2015 in the following electronic databases: MEDLINE (Ovid); EMBASE (Ovid); PsychINFO (EBSCO); CINAHL (EBSCO) and National Health Service Economic Evaluation Database (NHS EED).

systematic review, we applied the principles of the PRISMA statement (Appendix 2). We reviewed publications from 1 January 2004 to 13 February 2015 in the following electronic databases: MEDLINE (Ovid); EMBASE (Ovid); PsychINFO (EBSCO); CINAHL (EBSCO) and National Health Service Economic Evaluation Database (NHS EED). A sensitive search strategy was designed to incorporate two concepts, (1) Stroke and (2) Health economics, which were linked using the Boolean operator ‘AND’. We developed the Concept 1 strategy using guidance from the Cochrane Stroke Group and the strategy for Concept 2 using NHS Centre for Review and Dissemination (CRD) economic study search guidelines. Terms were tailored to each database taking into account unique topic headings and syntax. We also applied a Concept 3 utilising pre-coordination of information retrieval. This permits direct access to topic results using Emtree or MeSH subheadings e.g. Stroke/ec [Economics] for MeSH in Medline and cerebrovascular accident/dm [Disease Management] for Emtree in EMBASE. The results of our Concept 1 and Concept 2 searches were linked to Concept 3 by search operator ‘OR’. Appendix 1 shows the detail of the search strategies for all databases, and any limitations that applied to the results by author AW. Duplicate results were filtered out using EndNote reference manager (version X7.2.1, Thomson Reuters, USA) and citations were screened by title for relevance. We also filtered out citations that referred only to conference proceedings or abstracts before screening citations by title for their relevance.

A sensitive search strategy was designed to incorporate two concepts, (1) Stroke and (2) Health economics, which were linked using the Boolean operator ‘AND’. We developed the Concept 1 strategy using guidance from the Cochrane Stroke Group and the strategy for Concept 2 using NHS Centre for Review and Dissemination (CRD) economic study search guidelines. Terms were tailored to each database taking into account unique topic headings and syntax. We also applied a Concept 3 utilising pre-coordination of information retrieval. This permits direct access to topic results using Emtree or MeSH subheadings e.g. Stroke/ec [Economics] for MeSH in Medline and cerebrovascular accident/dm [Disease Management] for Emtree in EMBASE. The results of our Concept 1 and Concept 2 searches were linked to Concept 3 by search operator ‘OR’. Appendix 1 shows the detail of the search strategies for all databases, and any limitations that applied to the results by author AW. Duplicate results were filtered out using EndNote reference manager (version X7.2.1, Thomson Reuters, USA) and citations were screened by title for relevance. We also filtered out citations that referred only to conference proceedings or abstracts before screening citations by title for their relevance. The following inclusion and exclusion criteria were applied to title/abstract review of relevant search results: Inclusion Adult (18+). Includes costs data (indirect and/or direct costs reported i.e. hospital stay, carer, medications, loss of workplace earnings, etc. were all eligible). Acute stroke. mRS reported as the health outcome. Exclusion

The following inclusion and exclusion criteria were applied to title/abstract review of relevant search results: Inclusion Adult (18+). Includes costs data (indirect and/or direct costs reported i.e. hospital stay, carer, medications, loss of workplace earnings, etc. were all eligible). Acute stroke. mRS reported as the health outcome. Exclusion Subarachnoid haemorrhage or traumatic brain injury. Protocols or methodologies for randomised controlled trials (RCT). Cost-effectiveness studies comparing one or more intervention. We assessed the included studies for reporting and methodological quality. Currently, there is no consensus on the best instrument for assessing the methodological and reporting quality of cost-of-illness studies. In this review, we followed the recommendation of Cochrane handbook and utilised the checklist developed by Drummond and Jefferson,5 as relevant to cost-of-illness studies This focuses on three domains: study design; data collection and analysis; and interpretation of results. This checklist can be applied to range of health economic designs encompassing both full cost-effectiveness studies and cost-of-illness studies.

ed by Drummond and Jefferson,5 as relevant to cost-of-illness studies This focuses on three domains: study design; data collection and analysis; and interpretation of results. This checklist can be applied to range of health economic designs encompassing both full cost-effectiveness studies and cost-of-illness studies. The costs from the included studies were abstracted and then converted to relative 2015 costs in Euros accounting for inflation to allow for direct comparison of the results. Purchasing power parity (PPP) was used to calculate the relative value to each currency. Germany was chosen as having the most representative healthcare system and economy, and provided the ‘baseline’ Euro currency from which to calculate the PPP. The calculations were performed using a web-based calculator developed by Campbell and the Cochrane Economics Methods Group in conjunction with the Evidence for Policy and Practice Information and Coordinating Centre.6,7 Our aims were to present an estimate of cost of illness relative to stroke severity as measured by the mRS. However, given the recognised heterogeneity in methods used in health economic studies such as cost-of-illness studies,8 where we were unable to make any meaningful comparison among studies, we have presented a narrative review of the findings.

mate of cost of illness relative to stroke severity as measured by the mRS. However, given the recognised heterogeneity in methods used in health economic studies such as cost-of-illness studies,8 where we were unable to make any meaningful comparison among studies, we have presented a narrative review of the findings. Results The literature search yielded 8486 unique full text articles that were screened for inclusion in the study (Figure 1). From these, we identified and selected 61 relevant studies for full text review. Of these, only 13 met the inclusion criteria and have been included for reporting in this review. The characteristics of the selected studies are shown in Table 1. We included one study9 that had reported costs by individual mRS categories as part of a nested cohort study, whereby these cost-of-illness estimates were then later applied in a cost-effectiveness analysis of thrombolysis treatment. Figure 1. Result of systematic search strategy. Table 1. Characteristics of included studies. Studies Number of Patients Country Time frame Currency Follow up Additional info.

Results The literature search yielded 8486 unique full text articles that were screened for inclusion in the study (Figure 1). From these, we identified and selected 61 relevant studies for full text review. Of these, only 13 met the inclusion criteria and have been included for reporting in this review. The characteristics of the selected studies are shown in Table 1. We included one study9 that had reported costs by individual mRS categories as part of a nested cohort study, whereby these cost-of-illness estimates were then later applied in a cost-effectiveness analysis of thrombolysis treatment. Figure 1. Result of systematic search strategy. Table 1. Characteristics of included studies. Studies Number of Patients Country Time frame Currency Follow up Additional info. Average total cost per mRS (SD) 0 1 2 3 4 5 6 Asil et al.10 328 Turkey January 2007– December 2007 USD Discharge 1000 (729) 1838 (1895) 2777 (5792) Baeten et al.11 338 Netherlands January 1999– January 2000 Euro 6 months Stroke service 8400 11,080 29,644 27,371 6 months Usual Care 9856 14,868 37,682 46,089 12 months Usual Care 1761 4196 17,824 22,515 Christensen et al.12 820 Worldwide May 2005– February 2007 USD 3 months 9466 (4614) 15,547 (7873) 18,742 (9978) 27,387 (11,621) 27,281 (8,919) 27,330 (11,495) 8136 (3719) (95% CI) 7130– 11,902 13,336– 17,757 15,987– 21,496 24,372– 30,402 25,198– 29,364 22,182– 32,479 7421–9032 Christensen et al.13 167 Argentina January 2004– August 2006 USD Discharge ICH 1475 6370 Discharge IS 1622 3188 Christensen et al.14 316 Brazil January 2006– May 2007 USD Discharge ICH 6307 30,693 Discharge IS 1800 6771 Dawson et al.15 1717 Worldwide March 1998– May 1999 GBP 3 months 2493 to 3412 3369 to 4479 5784 to 7008 7300 to 8515 10,095 to 11,141 11,772 to 13,560 Dodel et al.16 340 Germany January 2000– June 2000 Euro Discharge 3160 (1300) 4030 (2780) Epifanov et al.17 253 Germany January 1998– June 1998 Euro Discharge 3,210 (2170) 3350 (3050) Fattore et al.18 411 Italy August 2005– March 2007 Euro 0–3 months 1225 3370 9185 3–6 months 343 1688 2335 6–12 months 513 939 2182 12 month total 1964 5239 13,381 Hayes et al.19 172 USA May 2001– September 2002 USD 3 months 7044 6159 7885 11,723 22,156 18,670 3–12 months 4321 5727 3994 6811 5154 9958 12 month total 14,901 12,637 12,751 23,218 30,971 28,628 Luengo-Fernandez et al.20 153 United Kingdom April 2002– March 2007 GBP 3 months 3945 (7558) 17,406 (18,417) 25,279 (16,396) Annual costs 2135 (3675) 4165 (7668) 6234 (14,898) Spieler et al.21 435 France Unavailable Euro 18 months 10,255 17,457 31,728 (95% CI) 9679–10,831 14,46– 020,453 28,811–34,645 Tanny et al.9 378 Australia January 2003– December 2011 USD 3 months 29,406 32,214 36,205 37,878 39,522 41,780 16,727 Note: Tan Tanny et al, calculated hospital costs from 378 patients based on actual expenditure sourced from the Clinical costing unit of Royal Melbourne Hospital and inputted these costs in the cost effectiveness model of thrombolysis

ary 2003– December 2011 USD 3 months 29,406 32,214 36,205 37,878 39,522 41,780 16,727 Note: Tan Tanny et al, calculated hospital costs from 378 patients based on actual expenditure sourced from the Clinical costing unit of Royal Melbourne Hospital and inputted these costs in the cost effectiveness model of thrombolysis treatment. Description of included studies Among the articles that we identified, the authors had investigated populations from diverse locations. Six studies were European (46%) and two were worldwide multicentre trials.13,15 Costs were quoted in three currencies: US dollars, Euros and Pounds Sterling. A broad range of methods had been used to determine costs in these currencies, but most had applied PPP to establish a common value to each currency worldwide. Patient data collection for the included studies was conducted from March 199815 through December 2011.9 Eleven studies reported costs up to 90 days (84%); and in five studies, the longer term costs of stroke of between 6 to 18 months were reported.11,18–21 Quality assessment The application of the Drummond et al.5 checklist to the studies shows the overall quality of the study was high (Table 4). However, presentation of results in both aggregate and disaggregate forms was handled poorly by the authors of these studies. Only 30% presented results as full ordinal mRS in relation to costs.

t The application of the Drummond et al.5 checklist to the studies shows the overall quality of the study was high (Table 4). However, presentation of results in both aggregate and disaggregate forms was handled poorly by the authors of these studies. Only 30% presented results as full ordinal mRS in relation to costs. Cost of stroke by mRS category Table 1 shows the total cost of stroke by mRS grade alongside any measures of uncertainty. The data collected from the studies are heterogeneous, with diverse resources recorded and included in the overall total cost per mRS grade (Table 2). Table 2. Perspective and resources collected in identified studies. Studies Perspective Index hospitalisation costs Post-acute resources Thrombolysis included in costs Length of staya Direct medical costsb Professional appointmentsc Rehabilitation appointmentsd Home health caree Productivity lossf Length of stayg Asil, et al.10 Hospital ✓ ✓ 8 patients in cohort Baeten11 Healthcare ✓ ✓ ✓ ✓ ✓ Not included Christensen and Morris12 Healthcare ✓ ✓ Not included Christensen et al.13 Healthcare ✓ ✓ Not included Christensen et al.14 Healthcare ✓ ✓ ✓ ✓ ✓ Not included Dawson et al.15 Hospital ✓ Not included Dodel et al.16 Hospital ✓ ✓ Included in cohort Epifanov et al.17 Hospital ✓ ✓ Not stated Fattore et al.18 Societal ✓ ✓ ✓ ✓ ✓ ✓ Not stated Hayes et al.19 Hospital ✓ ✓ ✓ ✓ ✓ Not stated Luengo-Fernandez et al.20 Healthcare ✓ ✓ ✓ Not stated Spieler et al.21 Societal ✓ ✓ ✓ ✓ ✓ ✓ Not stated Tanny et al.9 Hospital ✓ ✓ ✓ All patients in cohort Note:

✓ ✓ Included in cohort Epifanov et al.17 Hospital ✓ ✓ Not stated Fattore et al.18 Societal ✓ ✓ ✓ ✓ ✓ ✓ Not stated Hayes et al.19 Hospital ✓ ✓ ✓ ✓ ✓ Not stated Luengo-Fernandez et al.20 Healthcare ✓ ✓ ✓ Not stated Spieler et al.21 Societal ✓ ✓ ✓ ✓ ✓ ✓ Not stated Tanny et al.9 Hospital ✓ ✓ ✓ All patients in cohort Note: aStroke unit, ER, ICU, General ward, intermediate care facility, rehabilitation facility, nursing/convalescence home. bImaging, diagnostic tests, laboratory tests, surgical interventions and drug costs. cGeneral practitioner visits, emergency care, outpatient visits. dPhysiotherapy, speech therapy, ergo therapy. ePaid home healthcare, informal care, home adaptation, ortheses. fLoss of working days. gRehabilitation facility, nursing/convalescence home. Table 3 shows all mRS scores aggregated with associated costs in a common currency (Euro) adjusted for inflation and PPP and presented according to the time of assessment. The range of costs reported for mRS 1 is €1614 to €26,079 and for mRS 4: €4,754 to €35,050. The evaluated studies represent a range of follow-up time points at which the costs were recorded. The majority present costs and mRS scores at discharge or at 90 days; but some studies only recorded costs until 10 days or after the initial stroke event or until hospital discharge. In contrast, the studies of Fattore et al.18 and Spieler et al.21 focussed on longer term time points, reporting mean costs for mRS 3 of €5722 and €21,324 over 12 and 18 months, respectively. Table 3. Costs of Stroke by the mRS scores.

d costs until 10 days or after the initial stroke event or until hospital discharge. In contrast, the studies of Fattore et al.18 and Spieler et al.21 focussed on longer term time points, reporting mean costs for mRS 3 of €5722 and €21,324 over 12 and 18 months, respectively. Table 3. Costs of Stroke by the mRS scores. Note: Costs displayed in Euro adjusted to 2015 using purchasing power parity with Germany as the target currency. All calculations done using CCEMG – EPPI-Centre Cost Converter.10, http://eppi.ioe.ac.uk/costconversion/Default.aspx) Table 1 presents costs in original currency at time of study. Table 4. Quality of Included studies assessed by Dummond et al Checklist. Note: Colours indicate the level to which the study fulfils criteria; Green – Complete, Yellow – Not clear and Red – Does not fulfil criteria and Blank cells – Category not appropriate to study. Discussion The primary aim of this review was to collate the available data describing the relationship between costs and outcomes based on the mRS scale categories. Establishing a reliable estimation of costs by mRS categories is highly relevant since it may provide an indirect method for undertaking cost-effectiveness analyses of novel interventions to be compared against usual care. This review, however, found that it was not possible to effectively undertake any meaningful analyses due to the heterogeneous nature of the identified studies and lack of long-term follow-up data.

e an indirect method for undertaking cost-effectiveness analyses of novel interventions to be compared against usual care. This review, however, found that it was not possible to effectively undertake any meaningful analyses due to the heterogeneous nature of the identified studies and lack of long-term follow-up data. We identified 13 studies incorporating cost of stroke relating to an mRS score; only three studies provided an estimate of people who later died from stroke (mRS 6). However, there was significant methodological heterogeneity which precluded the ability to make any meaningful comparison between the stated costs either at a single mRS category or across the scale. Tables 1 and 2 highlight this heterogeneity, showing the diversity in time horizon (30–540 days), included resources and study perspectives.

ignificant methodological heterogeneity which precluded the ability to make any meaningful comparison between the stated costs either at a single mRS category or across the scale. Tables 1 and 2 highlight this heterogeneity, showing the diversity in time horizon (30–540 days), included resources and study perspectives. The time horizon for the collection of costs in these studies will have a large influence on the overall costs associated with stroke. Among the 13 studies identified, five recorded costs up until discharge, five at 90 days and five included costs for longer time frames (6 to 18 months) post stroke. Baeten et al.11 and Hayes et al.19 included costs at multiple time points. Costs in stroke are highly dependent on the time of collection with the intervention, rehabilitation and associated hospital costs concentrated in the acute phase (up to 90 days), while longer term costs including home health care, social services assistance, as well as productivity loss are more significant across a broader time period. This is highlighted in the two studies that considered longer term costs. Fattore et al.18 and Spieler et al.21 provided evidence that direct medical costs were initially high, but quickly plateaued and remained steady after the first 90 days.21 However, indirect costs such as productivity losses and paid care increased over time18 highlighting the importance of including of capturing costs across a broader time horizon when considering the health economic impact of stroke.

re initially high, but quickly plateaued and remained steady after the first 90 days.21 However, indirect costs such as productivity losses and paid care increased over time18 highlighting the importance of including of capturing costs across a broader time horizon when considering the health economic impact of stroke. Even when considering the studies that focussed on collecting data from comparable time horizons, there remained a high level of variability between costs reported at each category. This can be accounted for by the heterogeneity in reported resources (Table 2). Of the four studies looking at costs at 90 days using the full ordinal mRS Dawson et al.15 and Christensen and Morris12 focused on length of stay as their primary cost metric.22,23 Additionally, Christensen and Morris11also included coverage of rehabilitation and home healthcare costs. Hayes et al.19 and Tanny et al.9 calculated costs related directly from a patient cohort and extrapolated out of hospital information from relevant local cost-of-illness studies applied to their cohort based on discharge destinations. There was also a high level of heterogeneity in the reporting of outcomes with only 4 of 13 studies using the mRS as a complete ordinal scale. In other studies, the information on costs by mRS was dichotomised or trichotomised. This latter approach discards valuable information and undermined the ability to undertake meaningful comparisons between the included studies.

reporting of outcomes with only 4 of 13 studies using the mRS as a complete ordinal scale. In other studies, the information on costs by mRS was dichotomised or trichotomised. This latter approach discards valuable information and undermined the ability to undertake meaningful comparisons between the included studies. To be useful in cost-effectiveness evaluations, the mRS as a measure of functional ability beyond the acute phase of the disease needs to be costed from the perspective of society whereby the direct and indirect costs to the health sector, patients and other sectors, e.g. workforce are captured and summarised. Consistently, the identified studies provided evidence that increasing severity of mRS was associated with increasing direct medical costs. All studies but one15 included direct medical costs such as treatment, diagnostic costs and imaging in the estimation of costs at each mRS category. Hospital stay15 alone was used as the cost metric in the final study and highlighted the correlation between increased length of stay, mRS severity and increased costs. Capturing finer grained direct medical costs in hospitals is important since a patient who has achieved an mRS of 0 through costly treatment such as thrombectomy24 will incur little or no out-of-hospital costs but high direct medical costs. This review has shown that the estimate of costs includes some, if not all of the direct medical costs for the patients care associated with mRS category. However, to allow for comparison and generic estimates to be generated, future studies require more consistency in their methods.

osts but high direct medical costs. This review has shown that the estimate of costs includes some, if not all of the direct medical costs for the patients care associated with mRS category. However, to allow for comparison and generic estimates to be generated, future studies require more consistency in their methods. Strengths and limitations We employed a comprehensive search strategy utilising validated search strings designed to capture the broadest range possible of available literature investigating both stroke and cost-of-illness studies before combining these themes. The strategy was employed on the four major scientific databases, as well as the NHS EED. This review was carried out using a defined methodological approach to data extraction and critical evaluation of included studies. Our methods still have limitations. The systematic search and data extraction was carried out by a single author (AW) under the supervision of TQ. The data collected in the review yielded a highly heterogeneous sample based on what was available in the published literature: individual study authors were not approached. Additionally, the scope of this review was focussed on the use of the mRS and did not look at the systematic comparison of trials investigating the Health economics of Stroke using alternative outcome measures.

sample based on what was available in the published literature: individual study authors were not approached. Additionally, the scope of this review was focussed on the use of the mRS and did not look at the systematic comparison of trials investigating the Health economics of Stroke using alternative outcome measures. ESO Health economics working group meeting 2015 The results of the analysis were presented at the 2015 Health economics workshop at the European Stroke Organisation (ESO) meeting in Glasgow attended by participants from industry and academia. The attendees agreed in principle that standardisation of health economic data collected through clinical trials is required and suggested an international collaboration to develop guidelines for future trials. Attendees at the workshop also noted limited comparability across studies identified within this review, lending further credence to the suggestion that standardisation of resources collected by trialists is required to reduce heterogeneity. The importance of including out-of-hospital direct and indirect costs alongside direct medical costs that are incurred in hospital in future studies was emphasised by workshop members due to the long-term disabling nature of stroke. Attendees also noted that the WHO Research Agenda for Health Economic Evaluation (RAHEE) project in Stroke is working towards similar aims and could be approached for collaboration.25 Working groups to develop these guidelines have been assembled and the development of a prospective study investigating resource use in stroke trials is being undertaken.

ESO Health economics working group meeting 2015 The results of the analysis were presented at the 2015 Health economics workshop at the European Stroke Organisation (ESO) meeting in Glasgow attended by participants from industry and academia. The attendees agreed in principle that standardisation of health economic data collected through clinical trials is required and suggested an international collaboration to develop guidelines for future trials. Attendees at the workshop also noted limited comparability across studies identified within this review, lending further credence to the suggestion that standardisation of resources collected by trialists is required to reduce heterogeneity. The importance of including out-of-hospital direct and indirect costs alongside direct medical costs that are incurred in hospital in future studies was emphasised by workshop members due to the long-term disabling nature of stroke. Attendees also noted that the WHO Research Agenda for Health Economic Evaluation (RAHEE) project in Stroke is working towards similar aims and could be approached for collaboration.25 Working groups to develop these guidelines have been assembled and the development of a prospective study investigating resource use in stroke trials is being undertaken. Summary of suggested guidelines for future trials The result of the systematic review has yielded a four-point list of suggested guidelines for stroke researchers to optimise the collection of health economic information in future trials which are summarised below. Resource use and mRS to be collected at 90 days post stroke*.

of suggested guidelines for future trials The result of the systematic review has yielded a four-point list of suggested guidelines for stroke researchers to optimise the collection of health economic information in future trials which are summarised below. Resource use and mRS to be collected at 90 days post stroke*. mRS to be presented as a complete ordinal scale to preserve information relating to costs including those for patients who later died (mRS 6)*. Collection of resources used to be standardised. To this end, it is proposed that a group such as the ESO Economics Working Group develop a template and recommended costing methods as a resource to support this activity. Presentation of cost analyses to include measures of variability allowing for meta-analysis of aggregate data. *As recommended by the European Stroke Organisation (ESO) Outcomes Working Party.

Collection of resources used to be standardised. To this end, it is proposed that a group such as the ESO Economics Working Group develop a template and recommended costing methods as a resource to support this activity. Presentation of cost analyses to include measures of variability allowing for meta-analysis of aggregate data. *As recommended by the European Stroke Organisation (ESO) Outcomes Working Party. Conclusion Our findings have provided a valuable insight into the heterogeneity seen in health economic reporting in the field of stroke, in particular for the most commonly collected stroke outcome measure used in trials, the mRS. This heterogeneity undermined the meaningful comparison of the included studies and until further data are available for systematic analysis, we recommend readers refer to the original source data when assessing critical quality and relevance to ongoing research. It has also outlined a need for more real world and trial data investigating health economic outcomes in stroke looking at both short and long-term costs related to the mRS as an ordinal scale. This work has provided a foundation from which to address the need for the development of guidelines for health economic data and promotion of its importance amongst current and future trialists in the area of stroke.

c outcomes in stroke looking at both short and long-term costs related to the mRS as an ordinal scale. This work has provided a foundation from which to address the need for the development of guidelines for health economic data and promotion of its importance amongst current and future trialists in the area of stroke. Supplementary Material Supplementary material Supplementary material Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: KRL was President of the European Stroke Organisation (ESO) and chairman of VISTA, which held a joint ESO-VISTA workshop to collate and harmonise health economic data in stroke. ESO receives funding from numerous industry sponsors but none had influence over the analysis or reporting of the material in this manuscript. DAC was supported by a fellowship from the National Health and Medical Research Council (1063761 co-funded Heart Foundation [Australia]), Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the European Stroke Organisation (ESO) and Virtual International Stroke Trials Archive (VISTA). VISTA is a not-for-profit collaboration of researchers from academia and commercial organisations. PMB is Stroke Association Professor of Stroke Medicine. TQ is funded by a joint Stroke Association / Chief Scientist Office Senior Clinical Lecturer Fellowship. Informed consent Not applicable.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the European Stroke Organisation (ESO) and Virtual International Stroke Trials Archive (VISTA). VISTA is a not-for-profit collaboration of researchers from academia and commercial organisations. PMB is Stroke Association Professor of Stroke Medicine. TQ is funded by a joint Stroke Association / Chief Scientist Office Senior Clinical Lecturer Fellowship. Informed consent Not applicable. Ethical approval Not applicable. Contributorship AW carried out the systematic review, data extraction and analysis under the supervision of TQ. AW wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the European Stroke Organisation (ESO) and Virtual International Stroke Trials Archive (VISTA). VISTA is a not-for-profit collaboration of researchers from academia and commercial organisations. PMB is Stroke Association Professor of Stroke Medicine. TQ is funded by a joint Stroke Association / Chief Scientist Office Senior Clinical Lecturer Fellowship. Informed consent Not applicable. Ethical approval Not applicable. Contributorship AW carried out the systematic review, data extraction and analysis under the supervision of TQ. AW wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript ESO-VISTA Working Group Dr. Myzoon Ali – University of Glasgow, Prof. Philip MW Bath – University of Nottingham, Dr. Eivind Berge – University of Oslo, Anne-Laure Bocquet – Stryker, Prof. Natan Bornstein – Tel Aviv Sourasky Medical center, Prof. Marian Brady – Glasgow Caledonian University, Chloe Brown – Neuravi, Assoc. Prof. Dominiqe Cadilhac – Monash University, Dr. Bruce Campbell – Royal Melbourne Hospital, Assoc. Prof. Hanne Christensen – University of Copenhagen, Alain Cornil – Penumbra, Matthieu Cuche – Covidien, Daniel d Atillio – Penumbra, Avinoam Dayan – Brainsgate, Edith Doppler – Ever NeuroPharma, Prof. Gary Ford – University of Oxford, Dr. Rachael Fulton – University of Glasgow, Prof. Markku Kaste – University of Helsinki, Dr. Matthew Leathley – University of Central Lancashire, Prof Kennedy R Lees – University of Glasgow UK, Noam Levy – Brainsgate, Assoc. Prof. Atte Meretoja – University of Melbourne, Dr. Patrik Michel – University of Lausanne, Natalie Mühlemann – Nestle, Marine Provoyeur – Penumbra, Stacey Pugh – Covidien, Dr. Terry Quinn – University of Glasgow, Prof. Jeffrey Saver – UCLA, Dr Jan-Friedrich Scheitz – Charité – Universitätsmedizin Berlin, Dr. Peter Schellinger – John Wesling Medical Center Minden, Lauren Sheppard – Deakin University, Yoram Solberg – Brainsgate, Assoc. Prof. Nikola Sprigg – University of Nottingham, Dr Matthew Taylor – York Health Economics Consortium, Dr. Götz Thomalla – University of Hamburg-Eppendorf, Prof. Matthew Walters – University of Glasgow, Prof. Steven Warach – University of Texas Southwestern Medical Center, Prof. Joanna Wardlaw – University of Edinburgh, Prof. Christian Weimer – University of Essen, Dr. Alastair Wilson – University of Glasgow, Claudia Wolff – Medtronic

Introduction The endpoints commonly used in trials of treatments for moderate and severe stroke,1 such as death or dependency (often measured on the modified Rankin Scale (mRS)),2 may occur less frequently in patients with minor stroke and, therefore, inflate the sample size required in a randomised controlled trial (RCT). Such trials might include testing treatments for lacunar stroke, an important but neglected subtype of ischaemic stroke for which currently there is no specific treatment, but where trials are planned.3 Although death or dependence2 is important, other individual outcomes may also be of concern to patients with minor stroke, such as cognitive decline. Combining outcome measures into a composite outcome has the potential to increase trial efficiency by increasing the proportion of patients with the endpoint, improving power at smaller sample sizes and reducing costs and trial duration. Combined outcomes may also provide an overall outcome which captures several factors of relevance to patients. We used data from a longitudinal observational study of patients with a lacunar or minor non-lacunar ischaemic stroke to test the effect of several possible single and composite outcomes, assessed at one year after index stroke, on sample size estimates for RCTs.

Although death or dependence2 is important, other individual outcomes may also be of concern to patients with minor stroke, such as cognitive decline. Combining outcome measures into a composite outcome has the potential to increase trial efficiency by increasing the proportion of patients with the endpoint, improving power at smaller sample sizes and reducing costs and trial duration. Combined outcomes may also provide an overall outcome which captures several factors of relevance to patients. We used data from a longitudinal observational study of patients with a lacunar or minor non-lacunar ischaemic stroke to test the effect of several possible single and composite outcomes, assessed at one year after index stroke, on sample size estimates for RCTs. Methods We recruited consecutive inpatients and outpatients who presented to our Regional Stroke Service with a lacunar or minor non-lacunar ischaemic stroke. ‘Minor’ stroke was defined as NIHSS ≤ 7 and expected to be non-disabling at the point of assessment, i.e. recovery to no disability in basic activities of daily living (ADLs)4 like washing, dressing walking, bathing, but which might cause some reduction in instrumental ADLs.5 We recorded patient characteristics and medical history including vascular risk factors at recruitment, as reported previously.5 The study was approved by Lothian Research Ethics committee (REC 09/81,101/54) and NHS Lothian R + D Office (2009/W/NEU/14), and all patients gave written informed consent.

Methods We recruited consecutive inpatients and outpatients who presented to our Regional Stroke Service with a lacunar or minor non-lacunar ischaemic stroke. ‘Minor’ stroke was defined as NIHSS ≤ 7 and expected to be non-disabling at the point of assessment, i.e. recovery to no disability in basic activities of daily living (ADLs)4 like washing, dressing walking, bathing, but which might cause some reduction in instrumental ADLs.5 We recorded patient characteristics and medical history including vascular risk factors at recruitment, as reported previously.5 The study was approved by Lothian Research Ethics committee (REC 09/81,101/54) and NHS Lothian R + D Office (2009/W/NEU/14), and all patients gave written informed consent. We introduced cognitive testing with the Addenbrooke’s cognitive examination-revised version (ACE-R)6 at one month and one year after stroke; cognitive testing did not start until after the first 56 patients had been recruited due to delays in obtaining ethics approval. We considered a score of ≤82 to indicate cognitive impairment as it was the cut-off recommended in a validation paper as having a high specificity for dementia.6 The ACE-R is a multi-domain cognitive screening tool, similar to the Montreal Cognitive Assessment (MoCA) in many respects including its sensitivity and specificity for dementia and multi-domain cognitive impairment in the post-stroke setting.7

commended in a validation paper as having a high specificity for dementia.6 The ACE-R is a multi-domain cognitive screening tool, similar to the Montreal Cognitive Assessment (MoCA) in many respects including its sensitivity and specificity for dementia and multi-domain cognitive impairment in the post-stroke setting.7 We followed-up all patients face-to-face at one year post-stroke to identify any history suggestive of recurrent stroke, TIA, ischaemic heart disease (IHD) whether new episode of angina, or myocardial infarction during follow-up, performed physical examination including NIHSS and blood pressure, and measured the modified Rankin scale (mRS) using the structured method.8 If patients were unable to attend we performed telephone assessment, and if that was impossible we obtained relevant information from carers or the family doctor. Statistical analysis We used R statistical software (R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/) to run Fisher’s exact test (dichotomous variables) and the Mann–Whitney U test (continuous non-parametric variables) in univariate analyses to compare the characteristics in patients with lacunar and non-lacunar stroke.

stical software (R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/) to run Fisher’s exact test (dichotomous variables) and the Mann–Whitney U test (continuous non-parametric variables) in univariate analyses to compare the characteristics in patients with lacunar and non-lacunar stroke. We calculated the sample size required to detect a 10% relative risk reduction in the outcome of interest, at 80% and 90% power, these effect sizes being similar to that of several commonly used medical interventions, e.g. antiplatelets for secondary stroke prevention.9 For example, if an outcome occurred in 40% of participants, we calculated the sample size required to detect a reduction of 4%, from 40% to 36%. If an outcome occurred in 5% of participants, we calculated the sample required to detect a reduction of 0.5%, from 5% to 4.5%. We performed all sample size calculations for powers of 80% and 90%, as these are two conventional values.10 We calculated the sample size for lacunar and non-lacunar stroke separately, and then for all stroke combined.

We calculated the sample size required to detect a 10% relative risk reduction in the outcome of interest, at 80% and 90% power, these effect sizes being similar to that of several commonly used medical interventions, e.g. antiplatelets for secondary stroke prevention.9 For example, if an outcome occurred in 40% of participants, we calculated the sample size required to detect a reduction of 4%, from 40% to 36%. If an outcome occurred in 5% of participants, we calculated the sample required to detect a reduction of 0.5%, from 5% to 4.5%. We performed all sample size calculations for powers of 80% and 90%, as these are two conventional values.10 We calculated the sample size for lacunar and non-lacunar stroke separately, and then for all stroke combined. We first tested single outcomes, e.g. ‘recurrent stroke’ or, ‘ACE-R ≤ 82,’ and then tested combinations of vascular events, e.g. ‘recurrent stroke or TIA.’ We then incorporated dependence into the outcomes (testing both mRS ≥ 2 and mRS ≥ 3), and finally we incorporated cognition e.g. ‘Stroke, TIA, IHD, ACE-R ≤ 82, dementia, death or mRS ≥ 2.’ We then tested outcomes that included cognition, dependency and death but not recurrent vascular events, e.g. ‘ACE-R ≤ 82, dementia, death or mRS ≥ 3’ to allow for RCTs of differing objectives and agents. We included dementia as well as ACE-R ≤ 82 since dementia, a clinical diagnosis, might be available in a patient who was not able to undergo trial-based cognitive test like the ACE-R. A composite endpoint is a binary outcome measure: it is considered to have occurred if a patient had one or more of the component endpoints: for example if a patient had either a recurrent stroke, or a TIA, or both a stroke and TIA we would consider that they had experienced the endpoint ‘Stroke or TIA.’

ike the ACE-R. A composite endpoint is a binary outcome measure: it is considered to have occurred if a patient had one or more of the component endpoints: for example if a patient had either a recurrent stroke, or a TIA, or both a stroke and TIA we would consider that they had experienced the endpoint ‘Stroke or TIA.’ Results We screened 471 patients with a potential diagnosis of minor ischaemic stroke and recruited 264 (details, Figure 1).5 About 208 patients had cognitive testing at baseline since cognitive testing was introduced after the first 56 patients were recruited. Figure 1. Recruitment and follow-up. Patient characteristics and rates of individual outcomes At baseline (Table 1) the median age was 67 (range 36–98), 110/264 (42%) were female, the median NIHSS was 2 (interquartile range, IQR 1–3) and 118/264 (45%) patients had a lacunar stroke. The median mRS at the time of initial cognitive assessment was 1 (IQR 0-2). Table 1. Characteristics of patients at baseline and at one year.

e (Table 1) the median age was 67 (range 36–98), 110/264 (42%) were female, the median NIHSS was 2 (interquartile range, IQR 1–3) and 118/264 (45%) patients had a lacunar stroke. The median mRS at the time of initial cognitive assessment was 1 (IQR 0-2). Table 1. Characteristics of patients at baseline and at one year. Lacunar  n = 118 Non-lacunar  n = 146 p All ischaemic stroke Median age years (IQR) 64(55–65) 69(61–77) 0.0063 67(59–67) Female gender (%) 51(43%) 59(41%) 0.71 110(42%) Previous TIA 11(9%) 17(12%) 0.68 28(11%) Previous stroke 16(14%) 16(11%) 0.57 32(12%) Ischaemic heart disease (IHD) 19(16%) 34(23%) 0.17 53(20%) Diagnosis of peripheral vascular disease (PVD) 3(3%) 12(8%) 0.032 15(6%) Diabetes 12(10%) 18(12%) 0.70 30(11%) Hypertension 81(69%) 106(72%) 0.50 187(71%) Atrial fibrillation (AF) 7(6%) 18(12%) 0.09 25(9%) Diagnosis of hyperlipidaemia prior to index stroke, or at presentation 73(62%) 88(60%) 0.80 161(62%) Current smoker 46(39%) 44(30%) 0.15 90(34%) Median NIHSS (IQR) 2(2–4) 2(1–3) 0.0092 2(1–3) Median systolic BP (IQR) 147(130–158.5) 138(125.5–159) 0.16 142.5(130–159) Median mRS (IQR) at baseline assessment 1(1–2) 1(1–2) 0.67 1(1–2) Characteristics at 1 year Diagnosis of dementia 1(1%) 2(1%) 1.00 3(1%) IHD in the year following the stroke  (e.g. ongoing angina, or new myocardial infarction) 14(12%) 22(15%) 0.48 36(14%) NIHSS at 1 year (IQR) 0(0–1) 0(0–0) 0.72 0(0–0.25) NIHSS at 1 year ≥1 23(26%) 27(24%) 0.75 50(19%) mRS (IQR) 1(0–2) 1(1–2) 0.12 1(1–2) mRS = 0 (No symptoms) 30(21%) 32(27%) 0.24 62(23%) mRS ≥ 1 (Some symptoms) 86(73%) 116(79%) 0.24 202(77%) mRS ≥ 2 47(40%) 71(49%) 0.17 118(45%) mRS ≥ 3 20(17%) 33(23%) 0.28 53(20%) New TIA 3(3%) 4(3%) 1 7(3%) New stroke 10(8%) 15(10%) 0.68 25(9%) Either new stroke or TIA 12(10%) 18(12%) 0.70 30(11%) ACE-R at 1 year median (Interquartile range)  in n = 151 tested at 1 year 92(71–96) 90(59–94) 0.54 91(59–95) ACE-R ≤ 82 in n = 151 tested at 1 year 14(22%) 15(17%) 0.53 29(19%) Bold = p values that indicate significant differences at p < 0.01 between lacunar and cortical stroke subgroups.

A 12(10%) 18(12%) 0.70 30(11%) ACE-R at 1 year median (Interquartile range)  in n = 151 tested at 1 year 92(71–96) 90(59–94) 0.54 91(59–95) ACE-R ≤ 82 in n = 151 tested at 1 year 14(22%) 15(17%) 0.53 29(19%) Bold = p values that indicate significant differences at p < 0.01 between lacunar and cortical stroke subgroups. At one year, we followed up all 264 patients to ascertain if they were alive or dead, had had a recurrent vascular event, and their functional status; patients seen in person underwent repeat cognitive assessment. We assessed 204 in person, 47 by telephone, five via carers or relatives, and eight via their GP (Figure 1). At one year, 30 (11%) patients had had a recurrent stroke or TIA, 5 (2%) had died, and 3 (1%) had been diagnosed clinically with dementia. Many patients 118/264 (45%) still had some symptoms of stroke and 53/264 (20%) required assistance from family or carers with activities of daily living at least once per week. More patients with non-lacunar stroke had new diagnosis of peripheral vascular disease during follow-up(10/146 patients with non-lacunar stroke v 0/118 patients with lacunar stroke), there were no other statistically significant differences in outcomes between patients with lacunar or non-lacunar stroke. Of the 208 patients recruited after cognitive testing was introduced, 151/208 were tested at one year of whom 29/151 (19%) had an ACE-R ≤ 82. Of the 57/208 patients not having one year cognitive testing, 3 had died, 32 declined further testing, 19 were too unwell, and 3 had visual or language disabilities precluding testing.

208 patients recruited after cognitive testing was introduced, 151/208 were tested at one year of whom 29/151 (19%) had an ACE-R ≤ 82. Of the 57/208 patients not having one year cognitive testing, 3 had died, 32 declined further testing, 19 were too unwell, and 3 had visual or language disabilities precluding testing. Sample size estimations The effect of several single and composite endpoints on sample size, at 80% and 90% power, is shown in Table 2. For example, 10% of patients had a recurrent stroke (26/264), so to detect a 10% relative reduction in recurrent stroke at one year (from 26/264 to 23/264) would require 29,818 patients at 80% power. A larger proportion of patients, 29/151, 19% of those cognitively tested, had an ACE-R ≤ 82 at one year, which would require a sample size of 12,570 to detect a 10% reduction. However, 118/264 (45%) of patients had a mRS ≥ 2, therefore a sample size of 3864 would be required to detect a 10% reduction in mRS ≥ 2. The sample size estimations were similar for patients with lacunar and non-lacunar stroke since the proportion of most outcomes (Table 1) was similar in these two stroke subtypes (supplementary information). Table 2. Estimated sample size required to detect a 10% reduction in event rate for various combined outcomes at 80% power. Full details of individual and different combinations of outcomes at 80% and 90% power for lacunar and non-lacunar stroke at two mRS cut points are given in Supplementary Table 1.

ry information). Table 2. Estimated sample size required to detect a 10% reduction in event rate for various combined outcomes at 80% power. Full details of individual and different combinations of outcomes at 80% and 90% power for lacunar and non-lacunar stroke at two mRS cut points are given in Supplementary Table 1. Sample size required Sample size if combined with mRS ≥ 3 Sample size if combined with mRS ≥ 2 AND ACE ≤ 82a Recurrent stroke or TIA 23,600 7958 3090 Recurrent stroke or TIA or IHD 9908 4398 2224 Recurrent stroke, TIA, IHD, death Similar to the major adverse cardiovascular events (MACE) endpoint used in cardiovascular trials. 9144 4398 2224 a Includes clinical diagnosis of dementia. For composite outcomes at one year, such as ‘recurrent stroke or TIA, new IHD, or death,’ 25% would have the outcome, requiring a sample size of 9144 to be followed up to one year at 80% power (Table 2; 12,240 patients at 90% power, Supplementary Table 1). At 80% power, adding mRS ≥ 3 to this composite reduced the sample size to approximately half (4398) and replacing mRS ≥ 3 with mRS ≥ 2 reduced the sample size to approximately a third (3126) of 9144. Then, adding ACE-R ≤ 82 to the composite outcome of ‘stroke, TIA, new IHD, dementia diagnosis, death or an mRS ≥ 2,’ which occurred in 56% of patients without any double counting, reduced the sample size to 2224 patients.

nd replacing mRS ≥ 3 with mRS ≥ 2 reduced the sample size to approximately a third (3126) of 9144. Then, adding ACE-R ≤ 82 to the composite outcome of ‘stroke, TIA, new IHD, dementia diagnosis, death or an mRS ≥ 2,’ which occurred in 56% of patients without any double counting, reduced the sample size to 2224 patients. Although including ACE-R in the composite outcome reduced the sample size, it introduced missing data: 57/208 (27%) patients recruited with baseline cognitive testing could not have cognitive testing at one year mainly for medical reasons. There were fewer missing outcomes when considered as part of a composite endpoint: although 27% of patients had a missing outcome for ‘ACE-R ≤ 82,’ only 11% had missing data for ‘stroke, TIA, IHD, ACE-R ≤ 82, dementia, death or mRS ≥ 2.’ Figure 2 illustrates that while composite endpoints that include ACE-R do help to reduce sample size, up to 30% of the cognitive data may be missing, whereas composites that do not include cognition have very little missing data but need larger sample sizes. Figure 2. The effect of adding variables to a combined outcome on overall sample size required to detect a 10% reduction at 80% power. One way to compensate for missing data is to increase the sample size recruited (Supplementary Table 2). If retention progresses in a similar way to MSS-2 then to detect a 10% reduction in ‘stoke, TIA, dementia, death, mRS < /=2 or ACE-R > /=82,’ 2499 patients would need to be recruited in order for 2106 to be followed up at one year.

nsate for missing data is to increase the sample size recruited (Supplementary Table 2). If retention progresses in a similar way to MSS-2 then to detect a 10% reduction in ‘stoke, TIA, dementia, death, mRS < /=2 or ACE-R > /=82,’ 2499 patients would need to be recruited in order for 2106 to be followed up at one year. Another method is to use the ‘last observation carried forward’ (LOCF) method. About 157 patients had cognitive testing at 1–3 months post stroke of which 36 had ACE-R ≤ 82. By one year, 19/36 had ACE-R ≥ 82, 8 had an ACE-R < 82 and 9 were not tested; in contrast, of the 121 patients who had ACE-R < 82 at 1–3 months 4 had an ACE-R ≤ 82, at one year and 13 were not tested. If the LOCF method is used for the missing cognitive data, then 126/208 patients have the outcome measure of stroke, TIA, dementia, death, mRS>/=2 or ACE-R ≤ 82 at one year, with only 9% of patients lost to follow-up at one year. This suggests that one year data on 1748 patients would need to be recorded to detect a 10% reduction in outcome rate, at 80% power and 1880 would need to be recruited in order to follow-up 1748 patients at one year.

ntia, death, mRS>/=2 or ACE-R ≤ 82 at one year, with only 9% of patients lost to follow-up at one year. This suggests that one year data on 1748 patients would need to be recorded to detect a 10% reduction in outcome rate, at 80% power and 1880 would need to be recruited in order to follow-up 1748 patients at one year. Discussion We demonstrate that in patients with minor ischaemic stroke, in whom some important individual outcome events are infrequent, that a composite outcome such as ‘recurrent stroke, TIA, new IHD, ACE-R ≤ 82, new diagnosis of dementia or an mRS ≥ 2’ produced an outcome event rate of 56% and hence could substantially reduce sample sizes required to detect modest but worthwhile treatment effects while retaining conventional power. The net effect would be smaller, less expensive and more rapidly completed RCTs in subtypes of stroke that are less common therefore provide a smaller pool from which to recruit, and have been less studied to date.

mple sizes required to detect modest but worthwhile treatment effects while retaining conventional power. The net effect would be smaller, less expensive and more rapidly completed RCTs in subtypes of stroke that are less common therefore provide a smaller pool from which to recruit, and have been less studied to date. Adding mRS ≥ 3 halved the sample size from that based on recurrent stroke/TIA/IHD; using mRS ≥ 2 reduced it to a third; and adding in ACE-R ≤ 82 reduced it to an eighth of the starting sample. There was little difference between samples calculated for lacunar and non-lacunar stroke because the proportions of outcomes in these minor stroke patients were similar. These calculations were based on patients with similar characteristics to patients currently being recruited to a RCT testing interventions to prevent progression of small vessel disease in patients with lacunar stroke (LACI-1, NCT02481323) and to those who were recruited in the Secondary Prevention of Small Subcortical Stroke (SPS3) trial in lacunar stroke.11 However, the data came from a single population at a single centre that may limit generalisation to other settings. A similar exercise should be undertaken in other populations since outcome rates may differ.

se who were recruited in the Secondary Prevention of Small Subcortical Stroke (SPS3) trial in lacunar stroke.11 However, the data came from a single population at a single centre that may limit generalisation to other settings. A similar exercise should be undertaken in other populations since outcome rates may differ. Despite these benefits, composite outcomes may also have drawbacks that should be considered carefully. Interpretation may be more difficult since analyses based on composite outcomes generally emphasise the first event so a minor initial outcome can mask a subsequent major one.12 Additionally, it is theoretically possible for a treatment to have a positive effect on one outcome, and a negative effect on another, so a neutral trial result may mask a clinically significant outcome. This can be partly mitigated by careful presentation of results so that both individual and composite outcomes are easily visible to the reader. It is also important to choose components that are both relevant to patients and are biologically plausible (e.g. cognitive rehabilitation may reduce dependence but not recurrent stroke).

partly mitigated by careful presentation of results so that both individual and composite outcomes are easily visible to the reader. It is also important to choose components that are both relevant to patients and are biologically plausible (e.g. cognitive rehabilitation may reduce dependence but not recurrent stroke). The combined outcome approach suggested here assumes that all outcomes are equally significant and of equal weight. Further data are needed to suggest, for instance, whether patients consider myocardial infarction to be as severe as a diagnosis of dementia. Other approaches, such as weighting the different outcome measures based on their relevance to patients, could be tested in future work. Example approaches include ordinalising recurrent events (as used in the TARDIS trial13,14), using global tests that integrate individual outcomes statistically (e.g. using the Wald or Wei–Lachin tests15,16) or using Pocock’s Win Ratio.17

me measures based on their relevance to patients, could be tested in future work. Example approaches include ordinalising recurrent events (as used in the TARDIS trial13,14), using global tests that integrate individual outcomes statistically (e.g. using the Wald or Wei–Lachin tests15,16) or using Pocock’s Win Ratio.17 Including cognitive testing in the outcome of any stroke RCT introduces attrition bias, as some patients are unable to have cognitive testing at follow-up for various reasons, even when a relatively simple screening tool is used, as here (Figure 2). This leads to missing data and significant underestimation of post-stroke cognitive impairment.18 Missing data is a challenge for any clinical trial; whilst increasing the number of patients recruited and using the LOCF method can ensure that the necessary number reach one year follow-up, both methods may increase bias. Patients that do return for follow-up are unrepresentative of those that do not return, and the use of LOCF assumes that cognition is static from one to three months to one year post-stroke. How missing data is handled can have a substantial impact on results, the method used should be explicit in the protocol and statistical analysis plan, and not decided post-hoc.19 The SPS3 trial20 managed to achieve more complete follow-up – with only 11% of patients having missing cognitive tests at five-year follow-up. However, they only recruited patients who were able to have baseline cognitive tests in the sub-acute phase, and our patients were recruited a median of four days post-stroke.

9 The SPS3 trial20 managed to achieve more complete follow-up – with only 11% of patients having missing cognitive tests at five-year follow-up. However, they only recruited patients who were able to have baseline cognitive tests in the sub-acute phase, and our patients were recruited a median of four days post-stroke. Telephone cognitive assessment could reduce attrition bias, although is only applicable to some aspects of cognition. Several phone tests, at various stages of validation, are available.21 However, these do not allow for multi-domain screening of visuospatial and certain aspects of executive function.22 Even with a telephone assessment of cognition, there would still have been more missing outcomes than for other endpoints, underlining the problem of assessing cognition after stroke. Whilst the sample size could be increased to account for the expected proportion of patients with missing outcomes, such a sample may still be biased towards the healthier patients. Further research is needed to estimate required sample size if telephone cognitive testing is used. Self-reported outcome measures are available for several stroke-related outcomes23 and more are needed.24,25

proportion of patients with missing outcomes, such a sample may still be biased towards the healthier patients. Further research is needed to estimate required sample size if telephone cognitive testing is used. Self-reported outcome measures are available for several stroke-related outcomes23 and more are needed.24,25 For a common condition such as a stroke, a relatively small reduction in adverse outcomes would benefit a large number of people and therefore small effect sizes are worth trying to detect reliably with the smallest sample and shortest duration of follow-up possible, to reduce trial costs and minimise participant and researcher trial fatigue. Composite outcomes have the potential to do this and so accelerate trials of potential treatments; however, they can be more challenging to interpret, and care needs to be given when considering how to handle missing data. Supplementary Material Supplementary material Acknowledgements We thank the patients and their families for participating in the study. We thank the radiographers at the Brain Research Imaging Centre who performed all scanning and the Stroke Research Network who helped identify patients.

For a common condition such as a stroke, a relatively small reduction in adverse outcomes would benefit a large number of people and therefore small effect sizes are worth trying to detect reliably with the smallest sample and shortest duration of follow-up possible, to reduce trial costs and minimise participant and researcher trial fatigue. Composite outcomes have the potential to do this and so accelerate trials of potential treatments; however, they can be more challenging to interpret, and care needs to be given when considering how to handle missing data. Supplementary Material Supplementary material Acknowledgements We thank the patients and their families for participating in the study. We thank the radiographers at the Brain Research Imaging Centre who performed all scanning and the Stroke Research Network who helped identify patients. Declaration of conflicting interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: TQ has received grant funding from CSO, TSA, EH FP7, CH&SS NIH and NHS GG&C for work to improving outcome assessment in stroke. PMB has received funding from Alzheimer’s Society, BHF, NIHR HTA, MRC and Stroke Association. JMW has received funding from the Wellcome Trust, CHSS, MRC, TSA, CSO, SFC, Row Fogo Trust, EPSRC, HTA Panel, EME, BHF, Alzheimer Society, European Union, Age UK, Fondation Leducq for research on stroke. FD has received funding from the Wellcome Trust, TSA, CSO, EPSRC, BHF, Alzheimer Society for research on stroke. M Dennis received funding from the Wellcome Trust, CHSS, MRC, TSA, CSO, NIHR for research on stroke. The study was performed independently of the funders.

on, Age UK, Fondation Leducq for research on stroke. FD has received funding from the Wellcome Trust, TSA, CSO, EPSRC, BHF, Alzheimer Society for research on stroke. M Dennis received funding from the Wellcome Trust, CHSS, MRC, TSA, CSO, NIHR for research on stroke. The study was performed independently of the funders. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was funded by the Wellcome Trust (grant 088134/Z/09/A), the Scottish Funding Council and the Chief Scientist Office, Scotland, through the Scottish Imaging Network: A Platform for Scientific Excellence (‘SINAPSE’) and the European Union Horizon 2020 research and innovation programme SVDs@Target under grant agreement 666881. FD and TQ are funded by the Stroke Association/Garfield Weston Foundation and Stroke Association/Chief Scientist Office Senior Lectureships respectively. PMB is Stroke Association Professor of Stroke Medicine and is a NIHR Senior Investigator. The work was supported by the Fondation Leducq Transatlantic Network of Excellence in Small Vessel Disease ref no. 16 CVD 05, and the Horizon 2020 Programme PHC-03-15, project No 666881, ‘SVDs@Target.’ The work was conducted independently of the funders. Informed consent All patients gave written informed consent prior to enrolment in the study. Ethical approval It was granted by the Lothian Research Ethics committee (REC 09/81,101/54) and NHS Lothian R + D Office (2009/W/NEU/14). Guarantor JM Wardlaw

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was funded by the Wellcome Trust (grant 088134/Z/09/A), the Scottish Funding Council and the Chief Scientist Office, Scotland, through the Scottish Imaging Network: A Platform for Scientific Excellence (‘SINAPSE’) and the European Union Horizon 2020 research and innovation programme SVDs@Target under grant agreement 666881. FD and TQ are funded by the Stroke Association/Garfield Weston Foundation and Stroke Association/Chief Scientist Office Senior Lectureships respectively. PMB is Stroke Association Professor of Stroke Medicine and is a NIHR Senior Investigator. The work was supported by the Fondation Leducq Transatlantic Network of Excellence in Small Vessel Disease ref no. 16 CVD 05, and the Horizon 2020 Programme PHC-03-15, project No 666881, ‘SVDs@Target.’ The work was conducted independently of the funders. Informed consent All patients gave written informed consent prior to enrolment in the study. Ethical approval It was granted by the Lothian Research Ethics committee (REC 09/81,101/54) and NHS Lothian R + D Office (2009/W/NEU/14). Guarantor JM Wardlaw Contributorship SDM designed and assessed cognition analysed the data, and drafted the manuscript. FND advised on study design, identified patients and edited the manuscript. TQ advised on sample size calculations and PMB advised on additional approaches to analysis and both critically appraised the manuscript. MSD advised on study design, cognitive tests, adjudicated stroke diagnosis. JMW conceived, obtained funding for and oversaw the study, data management, MRI assessment, devised the data analysis and interpretation, made many revisions of the manuscript and takes full responsibility for the study.

Introduction Endovascular thrombectomy (EVT) is an effective treatment for acute ischemic stroke with or without intravenous alteplase.1–8 The HERMES9 individual patient meta-analysis found that for every five patients treated with EVT, two would have reduced disability by at least one level on the modified Rankin Scale (mRS). However, providing EVT presents major challenges in many health care systems. The procedure is typically carried out by neuro-interventionists with anaesthetic support, and requires an infrastructure capable of rapidly performing computed tomography angiography (CTA), with or without advanced imaging (AI) by perfusion-computed tomography (CTP), magnetic resonance imaging (MRI) techniques or CTA collateral scoring (CTA-CS). In clinical trials, CTA alone was generally used to select patients within 6 h of onset, whereas AI techniques were used beyond and sometimes before a 6-h window. The additional infrastructure demands for EVT create the need for a more centralised model of hyperacute stroke care, and robust activity estimates are required for accurate planning to inform service reconfiguration. In seeking to estimate the anticipated annual demand for this treatment in the UK, we developed a decision tree to estimate the proportion of all stroke patients eligible for EVT, regardless of geographic or service constraints such as non-existent care pathways or a lack of imaging and EVT facilities.

Introduction Endovascular thrombectomy (EVT) is an effective treatment for acute ischemic stroke with or without intravenous alteplase.1–8 The HERMES9 individual patient meta-analysis found that for every five patients treated with EVT, two would have reduced disability by at least one level on the modified Rankin Scale (mRS). However, providing EVT presents major challenges in many health care systems. The procedure is typically carried out by neuro-interventionists with anaesthetic support, and requires an infrastructure capable of rapidly performing computed tomography angiography (CTA), with or without advanced imaging (AI) by perfusion-computed tomography (CTP), magnetic resonance imaging (MRI) techniques or CTA collateral scoring (CTA-CS). In clinical trials, CTA alone was generally used to select patients within 6 h of onset, whereas AI techniques were used beyond and sometimes before a 6-h window. The additional infrastructure demands for EVT create the need for a more centralised model of hyperacute stroke care, and robust activity estimates are required for accurate planning to inform service reconfiguration. In seeking to estimate the anticipated annual demand for this treatment in the UK, we developed a decision tree to estimate the proportion of all stroke patients eligible for EVT, regardless of geographic or service constraints such as non-existent care pathways or a lack of imaging and EVT facilities. Patients and methods Using national registry data from the prospective Sentinel Stroke National Audit Programme (SSNAP) for England, Wales and Northern Ireland,10 and adjusted for Scotland using data from the Scottish Stroke Care Audit (SSCA),11 we estimated the number of patients hospitalised annually with acute stroke. A decision tree was constructed based upon key inclusion and exclusion criteria from published trials: stroke type, severity, presence of anterior or posterior large artery occlusion (LAO), onset time, pre-stroke disability, the extent of ischemia on CT (or MRI), pre-EVT recanalisation and optional AI. These criteria were applied consistently irrespective of eligibility for intravenous thrombolysis (IVT). The distributions for stroke severity and onset time were extracted from two large UK stroke services. The final decision tree has 12 steps and includes pathways using AI within and beyond 6 h after stroke onset. We did not include basilar artery occlusions presenting after 12 h, as quantifying these at a national level is imprecise. We undertook sensitivity analyses of key decision points to determine the effect upon estimates (proportion of LAO cases, clinical severity, onset time to presentation and core volume).

roke onset. We did not include basilar artery occlusions presenting after 12 h, as quantifying these at a national level is imprecise. We undertook sensitivity analyses of key decision points to determine the effect upon estimates (proportion of LAO cases, clinical severity, onset time to presentation and core volume). Results Estimating annual stroke admissions in the UK The decision tree is presented in Figure 1. It begins with an estimate of annual UK stroke admissions derived from SSNAP and SSCA. SNNAP coverage is comprehensive, with over 80,000 admissions recorded in 2015 from 100% of acutely-admitting hospitals. Case ascertainment in SNNAP is over 98% in England when verified against Hospital Episode Statistics, with the majority of cases omitted being sub-acute or otherwise ineligible for acute intervention. Scaling up this figure by the populations of Wales and Northern Ireland added 4480 and 2240 admissions, respectively. With 8700 admissions from the SSCA, total stroke admissions (excluding subarachnoid haemorrhage) for the UK are 95,500. Figure 1. Eligible population ((a) Total UK population including those deemed to be geographically inaccessible. (b) Confirmed infarcts, excluding ∼2% of patients whose status is unconfirmed. (c) Includes basilar artery occlusions eligible for treatment if presenting within 12 h. Others are assumed eligible unless they meet any subsequent exclusion. (d) ‘Early presenters’ – those presenting within 4 h.) Note: Patients within the large lower grey shaded box are all dealt with by AI (9400 + 10,130) those who are early presenters (10,130 on the left-hand side) can bypass that step.

ng within 12 h. Others are assumed eligible unless they meet any subsequent exclusion. (d) ‘Early presenters’ – those presenting within 4 h.) Note: Patients within the large lower grey shaded box are all dealt with by AI (9400 + 10,130) those who are early presenters (10,130 on the left-hand side) can bypass that step. Eligibility by stroke type, location and severity SSNAP10 and SSCA11 data report that 13% and 12% stroke admissions respectively are due to intracerebral haematoma. The proportion of ischemic strokes caused by LAO was observed at approximately 41% by the Screening Technology and Outcome Project in Stroke Study (STOP-Stroke), a prospective imaging-based study of stroke outcomes,12 and in the trials contributing to the HERMES meta-analysis.9 This is supported by a recent UK study of 263 patients reporting a 39% LAO rate.13

by LAO was observed at approximately 41% by the Screening Technology and Outcome Project in Stroke Study (STOP-Stroke), a prospective imaging-based study of stroke outcomes,12 and in the trials contributing to the HERMES meta-analysis.9 This is supported by a recent UK study of 263 patients reporting a 39% LAO rate.13 ‘Minor strokes’ (a National Institutes of Health Stroke Scale [NIHSS] score below 6) are not conclusively proven to benefit from EVT and were therefore not included in the eligible population.9 Whilst the HERMES meta-analysis applied a cut-off of NIHSS ≤ 10 (showing a strong trend towards benefit but without statistical significance), there was no evidence of heterogeneity in treatment effect by NIHSS. However, individual trials have shown benefit from EVT with an NIHSS of 6 or more: ESCAPE (Endovascular Treatment for Small Core and Anterior Circulation Proximal Occlusion with Emphasis on Minimizing CT to Recanalization Times)3 and SWIFT PRIME (Solitaire FR With the Intention for Thrombectomy as Primary Endovascular Treatment for Acute Ischemic Stroke)4; and NIHSS of 8 or more: REVASCAT (Randomized Trial of Revascularization With Solitaire FR Device Versus Best Medical Therapy in the Treatment of Acute Stroke Due to Anterior Circulation Large Vessel Occlusion Presenting Within 8 Hours of Symptom Onset).5 Only MR CLEAN (Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands)1 specifically enrolled patients with NIHSS below 6 and failed to show statistically significant benefit from EVT in the subgroup with NIHSS 2–15. Taking account of these data, we applied an NIHSS cut-off of 6 aligning with the three trials that included the largest numbers of patients in the NIHSS range 6–10.

herlands)1 specifically enrolled patients with NIHSS below 6 and failed to show statistically significant benefit from EVT in the subgroup with NIHSS 2–15. Taking account of these data, we applied an NIHSS cut-off of 6 aligning with the three trials that included the largest numbers of patients in the NIHSS range 6–10. The STOP-Stroke study12 reported that 20% of LAO strokes had an NIHSS of less than 6 (decision-point C). This was reinforced by El Tawil et al.13 These proportions give an estimate of 26,590 moderate/severe stroke patients (NIHSS 6 or more) with LAO in the UK annually. Time of onset and eligibility Eligible stroke patients were defined as those with a known stroke onset time of less than 12 h before presentation or were Stroke with Unknown Time of Onset (SUTO) with a Last Seen Well (LSW) time within 12 h. No recent published thrombectomy trial has included patients beyond this time period.

f onset and eligibility Eligible stroke patients were defined as those with a known stroke onset time of less than 12 h before presentation or were Stroke with Unknown Time of Onset (SUTO) with a Last Seen Well (LSW) time within 12 h. No recent published thrombectomy trial has included patients beyond this time period. A distribution of presentation times was derived from SNNAP10 but this was not reported by stroke severity. Stratification by severity was performed using service level SSNAP data for the calendar year 2015 from a single large UK acute stroke unit (Northumbria Healthcare NHS Foundation Trust: 900 admissions annually) and for three years from a second unit (Royal Devon and Exeter NHS Foundation Trust: 700 admissions annually), which showed that 78% of stroke patients with NIHSS of 6 or more presented within 12 h of onset. The SSNAP figure for all stroke cases presenting within 12 h was lower at 55% which is consistent with Northumbria and Devon and Exeter data if later presentation of milder cases is accounted for. For the remaining 22% of patients with NIHSS of 6 or more, SNNAP data enabled estimation of the relative proportions presenting with (a) SUTO but LSW within 12 hours (68.5%) and (b) a known onset time greater than 12 h (31.5%; Figure 1, decision-point E).

ter data if later presentation of milder cases is accounted for. For the remaining 22% of patients with NIHSS of 6 or more, SNNAP data enabled estimation of the relative proportions presenting with (a) SUTO but LSW within 12 hours (68.5%) and (b) a known onset time greater than 12 h (31.5%; Figure 1, decision-point E). According to SSNAP10 data, 81% patients present with a known time of onset, of whom 60% are within 4 h and 21.1% between 4 and 12 h (with 18.9% after 12 h). Therefore, the split between those presenting within 4 h and those between 4 and 12 h is 74% and 26%, respectively (Figure 1, decision-point F). After exclusions for onset time, stroke type, severity and location, the decision tree contains two cohorts of patients potentially eligible for EVT: ‘early presenters’ – i.e. those presenting within 4 h (mostly eligible for IVT within 4.5 h) and ‘late presenters’ – those ineligible for IVT because either their stroke onset was 4–12 h ago, or they were SUTO but LSW within 12 h. At this point in the decision tree, approximately 24,750 (25%) of stroke admissions are potentially eligible for EVT (9400 + 15,350). It was assumed that only ‘early presenters’ would be able to receive EVT treatment within 6 h of onset.9 Trial data indicate that from arrival at thrombectomy centre to arterial puncture it will take >60 min on average to groin puncture and at least another 45 min for recanalisation to be achieved.14 In addition, the majority of UK patients will require secondary transfer for EVT after initial local assessment. For late presenting patients (arrive beyond 4 h post onset), it was assumed that IVT would not be used. From this point in our decision tree, the two groups (Figure 1, decision-points G and H) are differentially influenced by application of AI.

nts will require secondary transfer for EVT after initial local assessment. For late presenting patients (arrive beyond 4 h post onset), it was assumed that IVT would not be used. From this point in our decision tree, the two groups (Figure 1, decision-points G and H) are differentially influenced by application of AI. Clinical and radiological exclusions amongst the IVT eligible population The largest group eligible for EVT were those early presenters i.e. 13,770 (14% of all stroke admissions). Further EVT exclusions associated with little prospect of successful reperfusion were a CT ASPECTS (Alberta Stroke Programme Early CT Score)15 of less than 6 or visible infarction of more than one-third of the middle cerebral artery (MCA) territory, and a pre-stroke mRS of 3 or more. As only 1.6% of the HERMES patients had an mRS of 3 or more, this group are excluded as EVT benefit is unproven. The STOP-Stroke study12 identified 8.7% of LAO stroke patients with a pre-stroke mRS of 3 or more, which is not dissimilar to reports from the study logs of trials included in HERMES.9

roke mRS of 3 or more. As only 1.6% of the HERMES patients had an mRS of 3 or more, this group are excluded as EVT benefit is unproven. The STOP-Stroke study12 identified 8.7% of LAO stroke patients with a pre-stroke mRS of 3 or more, which is not dissimilar to reports from the study logs of trials included in HERMES.9 The HERMES meta-analysis reported that an ASPECTS of 0–5 did not demonstrate a statistically significant treatment benefit (odds ratio [OR] 1.24, 0.62–2.49)9 possibly because numbers in this category were small (9%). In contrast, clear benefit for EVT was demonstrated with a presentation ASPECTS score of 6–8 and 9–10. To estimate the differential impact on outcome of early radiological changes, we applied a post hoc analysis of the Interventional Management of Stroke (IMS)-3 trial CTA positive subgroup data,14 which reported LAO on CTA in 40/282 participants (14%) with ASPECTS 0–4 and 88/282 (31%) with ASPECTS 5–7. We allocated these proportions equally to each ASPECTS score, yielding an estimated proportion of almost 25% for ASPECTS 0–5 in proven LAO. A pre-stroke mRS of 3 or more, and/or ASPECTS of 0–5 would, therefore, exclude approximately 34%. It was assumed that no overlap exists between these two criteria as we were unable to identify any reports of an association between pre-stroke disability and the severity of early ischemic changes assessed by ASPECTS or any other method. Therefore, amongst the early presenting IVT eligible population, we estimated that 10% of total stroke admissions were eligible for EVT, before any AI exclusions. This equates to 10,130 patients per year (Figure 1, decision-point G).

ty and the severity of early ischemic changes assessed by ASPECTS or any other method. Therefore, amongst the early presenting IVT eligible population, we estimated that 10% of total stroke admissions were eligible for EVT, before any AI exclusions. This equates to 10,130 patients per year (Figure 1, decision-point G). Various modes of AI (CT-CTP, CTA-CS combined with ASPECTS, or MRI) have been proposed for the exclusion of patients with a large core infarct. Data from the EXTEND-IA (Extending the Time for Thrombolysis in Emergency Neurological Deficits–Intra-Arterial)2 trial and the Sistema Online d'Informació de l'Ictus Agut (SONIIA)16 Registry suggest that AI excludes a further 5% of those early presenters with moderate/severe LAO stroke and pre-stroke mRS below 3 because they have a large volume core and small penumbra. If optional AIs were used in the early presenting group, the decision tree shows that a further 500 patients would be excluded, leaving an EVT eligible population of 9620 patients, before any recanalisation (Figure 1, decision-point I).

nd pre-stroke mRS below 3 because they have a large volume core and small penumbra. If optional AIs were used in the early presenting group, the decision tree shows that a further 500 patients would be excluded, leaving an EVT eligible population of 9620 patients, before any recanalisation (Figure 1, decision-point I). Clinical and radiological exclusions amongst the late presenting/SUTO population ineligible for IVT In the group presenting with SUTO but LSW within 12 h, or with a known onset time between 4 and 12 h, information about EVT eligibility is less robust and reliant upon variable AI protocols. Within our population of moderate-to-severe ischemic strokes with LAO, we estimate 5390 would have a known time of onset between 4 h to 12 h. We also estimated from SSNAP that a population of 4010 would be LSW within 12 h, giving a population of 9400 in whom AI might identify salvageable brain tissue, the majority of whom would also have a pre-stroke mRS below 3 (Figure 1, decision-point H).

estimate 5390 would have a known time of onset between 4 h to 12 h. We also estimated from SSNAP that a population of 4010 would be LSW within 12 h, giving a population of 9400 in whom AI might identify salvageable brain tissue, the majority of whom would also have a pre-stroke mRS below 3 (Figure 1, decision-point H). To identify the proportion of this group excluded by imaging, data from SWIFT17 and IMS-318 trials were used. At baseline, 25% had an ASPECTS below 6. Furthermore, by comparing ASPECTS at baseline to follow-up (mostly at 24 h), 48% deteriorated from good to poor ASPECTS.16 It is assumed that this deterioration represented core infarct extension occurring within 12 h. Therefore, in total 73% of ‘late-presenting’ patients are excluded by an ASPECTS below 6 on initial CT. Clinical mRS exclusions (as in the early-presenting group) would exclude another 8%12 or 203 (of the remaining 2538 late-presenting patients with an ASPECTS score indicative of limited acute ischaemic damage), leaving a total of 2340 of 9400 eligible for AI (Figure 1, decision-point H). That is 75% of 9400 are excluded. Data from the CTP group in MR CLEAN19 indicate that 43% had a large core of greater than 70 mL (using the definition applied in EXTEND2 and SWIFT-PRIME4 trials). Applying this proportion means that 1330 of the group remained definitely eligible for EVT (i.e. they had a smaller core and a larger volume of salvageable penumbral tissue; Figure 1, decision-point K).

te that 43% had a large core of greater than 70 mL (using the definition applied in EXTEND2 and SWIFT-PRIME4 trials). Applying this proportion means that 1330 of the group remained definitely eligible for EVT (i.e. they had a smaller core and a larger volume of salvageable penumbral tissue; Figure 1, decision-point K). Recanalisation prior to EVT Our estimates identify 9620 or 10,920 patients eligible for EVT, depending upon whether AI is used to identify salvageable brain tissue in those early presenters. A small proportion of these patients will recanalise spontaneously or in response to IVT before EVT is performed. The HERMES trials indicate that this occurred in 5% of those receiving IVT. Spontaneous recanalisation among patients not receiving IVT is estimated at 2% based on expert consensus (PW, GF, MJ), and the finding from the PROACT-II trial (Prolyse in Acute Cerebral Thromboembolism II),20 in which 2% of patients in the placebo arm had TIMI 3 (Thrombolysis in Myocardial Infarction rating scale [in which three represents complete recanalisation]); in this context, any recanalisation that is less than complete would not exclude EVT. Thus recanalisation prior to EVT excludes 510 patients (480 if the AI pathway followed) from the early presenting population presenting (Figure 1, decision-points J). Spontaneous recanaliation would exclude 30 patients from the late presenting/SUTO group (Figure 1, decision-point L).

han complete would not exclude EVT. Thus recanalisation prior to EVT excludes 510 patients (480 if the AI pathway followed) from the early presenting population presenting (Figure 1, decision-points J). Spontaneous recanaliation would exclude 30 patients from the late presenting/SUTO group (Figure 1, decision-point L). Sensitivity analyses Results of sensitivity analyses are shown in Table 1. For LAO, we identified retrospective study extremes between 13%21 and 88%,22 which were regarded as unreliable for modelling. More robust data from a prospective cohort reported a lower LAO estimate of 33%,23 and the EXTEND-IA2 screening log-identified LAO in 53% of IVT-eligible patients, so these data were used as the basis for a 30–50% range of LAO incidence. In the absence of other credible data sources, a pragmatic 10% range was also used for exclusion by onset time, ASPECTS, mRS and the proportion excluded due to a large core. The LAO proportion and the numbers of patients presenting with a known onset time within 12 h, had the greatest impact on the estimates of eligibility. Table 1. Univariate sensitivity analyses.

a pragmatic 10% range was also used for exclusion by onset time, ASPECTS, mRS and the proportion excluded due to a large core. The LAO proportion and the numbers of patients presenting with a known onset time within 12 h, had the greatest impact on the estimates of eligibility. Table 1. Univariate sensitivity analyses. Decision point Value (%) Eligible population Proportion of LAO strokes High value 50 12,030–13,670 Low value 30 7220–8200 Proportion of moderate/severe strokes presenting early High value 88 10,860–12,100 Low value 68 8390–9860 Proportion of late-presenting patients excluded by ASPECTS and mRS of 3 or more High value 65 9620–11,460 Low value 85 9620–10,410 Proportion of late-presenting patients with large core High value 33 9620–11,170 Low value 53 9620–10,710 ASPECTS: Alberta Stroke Programme Early CT Score; LAO: large artery occlusion; mRS: modified Rankin Scale.

enting patients excluded by ASPECTS and mRS of 3 or more High value 65 9620–11,460 Low value 85 9620–10,410 Proportion of late-presenting patients with large core High value 33 9620–11,170 Low value 53 9620–10,710 ASPECTS: Alberta Stroke Programme Early CT Score; LAO: large artery occlusion; mRS: modified Rankin Scale. Discussion Based on the available evidence from intervention trials and prospective registries in EVT, we estimate 9140–10,920 patients in the UK with acute ischemic stroke are eligible for EVT annually i.e. approximately 10% of strokes admitted to hospital. This is consistent with other reports. Chia et al.22 estimated a range of 7–13% for EVT eligibility presenting to two of three Australian hyper-acute stroke sites serving a population of approximately 150,000. The lower bound of our estimate is defined by restricting EVT only to those early presenters (9620/year). The upper bound is defined by the inclusion of all early-presenting patients without the use of AI (9620/year) to which are added those late-presenting patients with a favourable imaging profile (1310/year). AI would exclude around 5% (500/10,130) of early-presenting and otherwise eligible patients from EVT but would include around 56% (1310/2350) of late-presenting (IVT-ineligible) patients as eligible for EVT. Thus, although the overall requirement (eligibility) for EVT is relatively unchanged by AI, its use would affect EVT treatment decisions in approximately 15% (1810/12,470) of otherwise eligible patients.

from EVT but would include around 56% (1310/2350) of late-presenting (IVT-ineligible) patients as eligible for EVT. Thus, although the overall requirement (eligibility) for EVT is relatively unchanged by AI, its use would affect EVT treatment decisions in approximately 15% (1810/12,470) of otherwise eligible patients. Where possible our decision points are based upon the large prospective SSNAP registry, which covers the UK excluding Scotland. Case ascertainment by SSNAP in England (population 55 million) exceeds 98%. SSNAP or randamised controlled trials (RCTs) data provide the main evidence criterion for 9 of the 12 decision points. The main uncertainties are in the smaller group of late-presenting patients with LAO and NIHSS greater than 6, for whom limited high-quality data are available around eligibility for EVT (decision-point H) since this population was the least represented in the trials. However, this group is small and sensitivity analyses show that changing assumptions have little impact upon model outcomes.

ts with LAO and NIHSS greater than 6, for whom limited high-quality data are available around eligibility for EVT (decision-point H) since this population was the least represented in the trials. However, this group is small and sensitivity analyses show that changing assumptions have little impact upon model outcomes. The proportion of patients considered appropriate for EVT is dependent upon the frequency of LAO, but previous reports vary. Amongst the recent thrombectomy trials which reported screening and eligibility data, the rate of LAO was 53% in EXTEND-IA2 and 48% in SWIFT PRIME.4 Rai et al.21 estimated the incidence of LAO from a retrospective sample of nearly 3000 patients referred to a tertiary-level academic hospital in West Virginia, over 90% of whom had CTA, with LAO demonstrated in only 12%. However, complete case ascertainment is uncertain as many patients were secondary transfers, and over 70% of LAO were M1 occlusions. Smith et al.12 identified, after expert review, an LAO rate of 46% in patients with confirmed stroke referred to two large academic US centres, using a broader definition which included the anterior and posterior cerebral arteries, and second-order branches (so M2). A recent prospective study in the UK identified an LAO rate of 39%.13 Our rate of LAO at 40% may be a small overestimate, but we consider this to be based on the most reliable information available.

sing a broader definition which included the anterior and posterior cerebral arteries, and second-order branches (so M2). A recent prospective study in the UK identified an LAO rate of 39%.13 Our rate of LAO at 40% may be a small overestimate, but we consider this to be based on the most reliable information available. The selection of patients by AI based upon current best evidence had relatively little effect on the overall numbers eligible for treatment but altered the eligibility decision in 15% of cases. The impression that a relatively small proportion of early-presenting patients with LAO on CTA would be subsequently ruled out by AI (5% in our model) is corroborated by EXTEND-IA2 trial. The results from the DAWN trial (NCT02142283) will be valuable for clarifying the proportion of patients with an unknown symptom onset time who should be offered EVT according to AI.

of early-presenting patients with LAO on CTA would be subsequently ruled out by AI (5% in our model) is corroborated by EXTEND-IA2 trial. The results from the DAWN trial (NCT02142283) will be valuable for clarifying the proportion of patients with an unknown symptom onset time who should be offered EVT according to AI. With no formally commissioned services, the UK is starting from a low baseline; in 2017, NHS England anticipates funding treatment of 1000 patients in the first year of formal commissioning. The midpoint of our estimate for a UK population suitable for EVT (10.8% of all stroke admissions) combined with the absolute benefits estimated in a recent individual patient data meta-analysis24 suggest that EVT with national coverage could achieve an additional 2420 patients with independent functional outcomes, or as many as 4280 patients (4% all stroke admissions) with a reduced level of disability compared to IVT alone. Implicit in this estimates is the assumption that outcomes for posterior circulation EVT (which are included in our estimates of eligible population) are the same as those for anterior circulation EVT. There is an absence of evidence about posterior circulation EVT, but in light of outcomes for basilar artery occlusions treated with IVT, we judge this assumption reasonable at this time. Based on a range of estimates, the mean monthly cost to the UK National Health Service and social care providers of caring for people who lose their independence because of stroke (an mRS of 3, 4 or 5) was estimated at £790 (US$1,300/€980 at 2014 exchange rates) at 2013–2014 prices.25 Assuming 2420 people would maintain independence because of EVT, the savings (before costs for EVT are included) over 12 months post-stroke are greater than £22 million (US $36 million/€27 million at 2014 exchange rates). A cost-effectiveness analysis from the US26 reported that EVT is a highly cost-effective intervention in the prevention of stroke-related disability with an incremental cost-effectiveness ratio of $3,000 per quality-adjusted life year (QALY).

greater than £22 million (US $36 million/€27 million at 2014 exchange rates). A cost-effectiveness analysis from the US26 reported that EVT is a highly cost-effective intervention in the prevention of stroke-related disability with an incremental cost-effectiveness ratio of $3,000 per quality-adjusted life year (QALY). A more recent study projected that EVT dominated thrombolysis alone when future savings from reduced social care need were included, and despite the higher costs of providing EVT, there was a saving of £30,000 over a patient’s lifetime to health and social care providers and before the consequences of lost productivity in the working age stroke population were accounted for.27 This equates each year in the UK, to a net realisable saving of £73 million each year over patient’s lifetimes. Conclusion Between 9620 and 10,920 stroke patients per year in the UK could be eligible for EVT based on current level-1 evidence, which approximates to 10% of stroke admissions. Given the magnitude of the potential clinical and wider economic benefits from EVT, it should now be a key priority to address the substantial infrastructure and workforce obstacles impeding rapid and widespread implementation.

e for EVT based on current level-1 evidence, which approximates to 10% of stroke admissions. Given the magnitude of the potential clinical and wider economic benefits from EVT, it should now be a key priority to address the substantial infrastructure and workforce obstacles impeding rapid and widespread implementation. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: PW is co-PI for two randomised trials (PISTE and STABILISE) investigating different aspects of thrombectomy in acute stroke. Start-up phase of PISTE was mainly funded by Stroke Association but was also part funded by unrestricted institutional educational grants from Covidien and Codman who both manufacture devices used for stroke thrombectomy. STABILISE is part funded by Microvention. PW has also undertaken educational consultancy work within last three years for Codman and Microvention who both manufacture devices used for stroke thrombectomy. GAFs institution has received an educational grant from Medtronic to model thrombectomy need, and honoraria from Pulse Therapeutics for medical advisory work. GAF has received personal remuneration from Medtronic, AstraZeneca, Boehringer Ingelheim, Lundbeck, Cervast, Daiichi Sankyo and Pfizer. MAJ has received personal fees and non-financial support from Boehringer Ingelheim, Bayer, Bristol-Myers-Squibb and Daiichi-Sankyo outside the submitted work.

or medical advisory work. GAF has received personal remuneration from Medtronic, AstraZeneca, Boehringer Ingelheim, Lundbeck, Cervast, Daiichi Sankyo and Pfizer. MAJ has received personal fees and non-financial support from Boehringer Ingelheim, Bayer, Bristol-Myers-Squibb and Daiichi-Sankyo outside the submitted work. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article presents independent research funded by the National Institute for Health Research (NIHR) under its Programme Grant for Applied Research Programme (RP-PG-1211-20012). MAJ is supported by the NIHR Collaboration for Leadership in Applied Health Research and Care for the South West Peninsula. GAF was supported by an NIHR Senior Investigator award. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the English Department of Health. The work was supported by an unrestricted grant from Medtronic. Ethical approval Not applicable Informed consent Not applicable Guarantor Not applicable Contributorship GF, PW, MJ and PM researched literature and conceived the study. CP and DF were involved in analysis and interpretation. PM wrote the first draft of the manuscript and all authors reviewed and edited the manuscript and approved the final version of the manuscript.

Introduction It is recognised that adherence to secondary preventative medications after stroke is variable; in some studies more than half of participants stopped taking their prescribed drugs 1–2 years after the stroke incident.1–3 Use of the secondary prevention strategies has been reported to result in 80% reduction in the risk of stroke recurrence, vascular events or death4,5 and poor adherence is related to adverse outcomes.6–8 Many factors interfere with the ability of stroke patients to regularly take their medications. Stroke survivors may have disability or cognitive issues which make them unable to self-administer medication.9–11 Personal beliefs and preferences may also impact adherence.10 Medication factors also affect adherence. Drugs such as anti-coagulants typically have less adherence than anti-platelets11 and cost of medications is also of potential importance.9 Health care system failure exists through lack of access to health care and inadequate communication with health care providers.12 Several studies have attempted to identify barriers to adherence to medication after stroke. Patients with stroke expressed that concerns about prescribed medication and unawareness of the rationale of treatment as primary reasons for non-adherence.13 We performed a systematic review and meta-analysis of studies that assessed predictive factors for adherence to preventative medications in patients with stroke or transient ischaemic attack (TIA).

concerns about prescribed medication and unawareness of the rationale of treatment as primary reasons for non-adherence.13 We performed a systematic review and meta-analysis of studies that assessed predictive factors for adherence to preventative medications in patients with stroke or transient ischaemic attack (TIA). Methodology We performed a systematic review and meta-analysis following Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines14 for design, conduct and reporting. The review protocol was registered in PROSPERO (registration number: CRD42015027531). Search strategy and study selection We generated search strings based on concepts of ‘Stroke’ and ‘Medication Adherence.’ We focussed on MeSH terms and other controlled vocabulary (available in the supplementary appendix, which can be found online with this review). Two independent reviewers (SA and WD) searched Web of Knowledge, EMBASE, MEDLINE (both using Ovid), CINAHL, PsycINFO (both in EBSCOhost) and CENTRAL (Cochrane Library). Initially, titles were reviewed and possibly eligible articles were listed for abstract review. These were then retrieved for entire text review by SA. We also reviewed reference lists of included studies and related reviews to detect additional reports.

AHL, PsycINFO (both in EBSCOhost) and CENTRAL (Cochrane Library). Initially, titles were reviewed and possibly eligible articles were listed for abstract review. These were then retrieved for entire text review by SA. We also reviewed reference lists of included studies and related reviews to detect additional reports. Eligibility criteria We only included studies published in English. Studies had to include adults (aged ≥ 18 years) who had suffered stroke or TIA and were prescribed medication for the prevention of recurrent cardiovascular events. Studies had to assess factor(s) that influenced medication adherence. Where disagreement arose regarding study eligibility, a consensus meeting was arranged with an arbitrator (JD). We excluded from this review studies that did not include a measure of medication adherence, studies that assessed non-pharmacological preventative strategies only or did not include stroke or TIA patients. Data extraction We designed a data extraction form that summarised information on study characteristics, inclusion criteria, sample size, secondary preventative medications, method used to measure adherence and predictive factors. We did not contact the study authors for missing information or for clarification.

Eligibility criteria We only included studies published in English. Studies had to include adults (aged ≥ 18 years) who had suffered stroke or TIA and were prescribed medication for the prevention of recurrent cardiovascular events. Studies had to assess factor(s) that influenced medication adherence. Where disagreement arose regarding study eligibility, a consensus meeting was arranged with an arbitrator (JD). We excluded from this review studies that did not include a measure of medication adherence, studies that assessed non-pharmacological preventative strategies only or did not include stroke or TIA patients. Data extraction We designed a data extraction form that summarised information on study characteristics, inclusion criteria, sample size, secondary preventative medications, method used to measure adherence and predictive factors. We did not contact the study authors for missing information or for clarification. Assessment of risk of bias in included studies We assessed risk of bias in included studies using a pre-specified tool generated using Cochrane Library tool for assessing risk of bias15 and the Newcastle–Ottawa scales.16 Two independent reviewers (SA and JD) assessed risk of bias and met to finalise the assessment. Disagreement was resolved via discussion until reaching a mutual agreement. We considered studies as of high quality if they met the criteria for all the assessment domains (selection, performance, attrition, reporting and confounders).

ent reviewers (SA and JD) assessed risk of bias and met to finalise the assessment. Disagreement was resolved via discussion until reaching a mutual agreement. We considered studies as of high quality if they met the criteria for all the assessment domains (selection, performance, attrition, reporting and confounders). Data synthesis and analysis We categorised preventative medications as anti-coagulants, anti-platelet, blood pressure or lipid lowering drugs. Some studies also reported adherence to the overall medication regimen without specification of medication classes. We listed predictive factors, significance (odds or hazard ratios and 95% confidence intervals) and the type of analysis used. We used the World Health Organization (WHO) classification of predictive factors of non-adherence, which categorised these into five domains:17 – Patient related factors – Social and economic related factors – Therapy-related factors – Health system or health care team related factors and – Condition (stroke)-related factors We described included studies and factors reported to be significant using a narrative review. Where a factor was assessed in more than three studies we described a summary value using random-effects models meta-analyses. We also described summary measures of medication non-adherence across non-case control studies. These analyses used Comprehensive Meta-Analysis software (CMA, version 2.0, Biostat Inc).

review. Where a factor was assessed in more than three studies we described a summary value using random-effects models meta-analyses. We also described summary measures of medication non-adherence across non-case control studies. These analyses used Comprehensive Meta-Analysis software (CMA, version 2.0, Biostat Inc). Results The search was completed in April 2014 and identified a total of 12,237 titles. Title review identified 143 papers for abstract review. Of these 57 were retrieved for full-text review. We identified 29 of these as meeting our eligibility criteria (Figure 1).1,2,9–12,18–40 Figure 1. PRISMA flow diagram. Risk of bias across included studies Studies included in this review were all of high risk of bias (except two34,36) mainly because details on performance bias, represented by blinding of outcome assessor, were not reported. It was also unclear whether there was a selective reporting of the outcomes in a study.23 Twelve studies were non-controlled.2,9,10,18–20,22,28,32,38–40 In addition, most studies used a subjective method to monitor adherence which has been reported to overestimate patients’ adherence.41,42 More details on other sources of bias in included studies are available in the supplementary appendix.

es in a study.23 Twelve studies were non-controlled.2,9,10,18–20,22,28,32,38–40 In addition, most studies used a subjective method to monitor adherence which has been reported to overestimate patients’ adherence.41,42 More details on other sources of bias in included studies are available in the supplementary appendix. Narrative review Description of eligible studies The 29 included studies were observational studies of which 14 were prospective cohorts,1,2,9,10,18,20,24,26,32,35,36,38–40 4 were retrospective cohorts,22,28,33,34 9 used a cross-sectional design11,12,21,25,27,29–31,37 and two performed a case-control analysis.19,23 Details of study characteristics can be found in Table 1. The total number of participants in the included studies was 69,137. Reported non-adherence rate ranged between 11.3%39 and 45.2%.30 Table 1. Characteristics of included studies.

ional design11,12,21,25,27,29–31,37 and two performed a case-control analysis.19,23 Details of study characteristics can be found in Table 1. The total number of participants in the included studies was 69,137. Reported non-adherence rate ranged between 11.3%39 and 45.2%.30 Table 1. Characteristics of included studies. Study Design Inclusion criteria Exclusion criteria Sample size Medication classes Adherence assessment measure Arif et al.21 Cross-sectional First-time stroke MI Non-ischaemic or non-haemorrhagic TIA 298 AP AH LLD Telephone interview Burke et al.22 Retrospective cohort First-time IS Previous cardiac condition Previous AT 1413 AP Prescription refill Bushnell et al.18 Observational cohort, 3 months IS or TIA – 2598 AP AC AH LLD Telephone interview Bushnell et al.18 Longitudinal study, 1 year IS or TIA – 2457 AP AC AH LLD Telephone interview Chambers et al.23 Case-control study First- time IS Institutional living 26 Not specified MARS and BMQ Choi-Kwon et al.24 Observational cohort, 1–5 years Early-onset stroke patients (onset between ages of 15–45 years) HS TIA Severe medical conditions Previous stroke 256 AH Patient interview Coetzee et al.25 Cross-sectional at 6 weeks Completed rehabilitation program – 26 (compared to 29 amputee patients) All classes Patient interview and pill count De Schryver et al.26 Cohort study, 1–2 years Patients in the Dutch TIA Trial and the Stroke Prevention In Reversible Ischaemia Trial – 3796 (aspirin) and 651 (AC) Aspirin AC Patient interview and pill count Edmondson et al.27 Cross-sectional Age > 40 years Stroke or TIA Institutional living Pregnant Aphasia Cognitive impairment 535 AT AH LLD MMAS and BMQ Glader et al.2 Prospective observational study, 2 year Patients in the Swedish Stroke Register – 24,024 AP AC AH LLD Prescription refill Huang et al.28 Retrospective cohort, 1 year IS or TIA In-hospital stroke 11,050 AT AH LLD Prescription refill Ji et al.29 Cross-sectional, at 3 months IS or TIA – 9998 AP AC AH LLD Telephone interview Ke et al.30 Cross-sectional Cerebral infarction TIA – 1240 Aspirin Telephone interview Kronish et al.31 Cross-sectional Stroke or TIA in the past 5 years Institutional living Pregnant Aphasia Cognitive impairment 535 Not specified MMAS Kronish et al.12 Cross-sectional study Stroke or TIA Age ≥ 40 years Aphasia Cognitive impairment Pregnant Institutional living 600 Not specified MMAS Levine et al.19 Case-control study Stroke Age ≥ 45 years Noninstitutionalized – 8673 Not specified Questionnaire Lopes et

asia Cognitive impairment 535 Not specified MMAS Kronish et al.12 Cross-sectional study Stroke or TIA Age ≥ 40 years Aphasia Cognitive impairment Pregnant Institutional living 600 Not specified MMAS Levine et al.19 Case-control study Stroke Age ≥ 45 years Noninstitutionalized – 8673 Not specified Questionnaire Lopes et  al.32 Longitudinal study, 1 year IS or TIA with AF in Get With The Guidelines (GWTG)–Stroke registry & Adherence eValuation After Ischemic Stroke Longitudinal (AVAIL) registry Bleeding Palliative-care Death or transfer from hospital 291 AC Patient interview Lummis et al.9 Cohort study, 1 year Stroke patients in the Stroke Outcome Study – 420 AT AH LLD Self-reported adherence O’Carroll et al.10 Longitudinal study, 1 year First-time IS Responsible for own medication Institutional living 180 AH Aspirin LLD MARS, BMQ and urinary- salicylate level Østergaard et al.33 Retrospective cohort Suspected stroke HS 503 AP Prescription refill Østergaard et al.34 Retrospective cohort, 1.7 years TIA Prior TIA or stroke & previous AC 594 AP Prescription refill Rodriguez et al.35 Longitudinal study, 1 year IS or TIA GWTG-Stroke program – 2720 AP AC AH LLD Telephone interview Sappok et al.36 Prospective observational study, 1 year IS or TIA Haemorrhage Migraine Epilepsy 470 AT Telephone interview Sjölander et al.38 Prospective observational study Ischemic stroke in the Swedish Stroke Register – 18,349 AH Medication refill Sjölander et al.37 Cross-sectional Stroke Institutional-living 578 Not specified MARS Thrift et al.20 Prospective cohort, 10 years Stroke Subarachnoid haemorrhage 1241 AT AH LLD Self-reported adherence Wang et al.11 Cross-sectional, at 1 year TIA or a cerebral infarction Haemorrhage Migraine Epilepsy 722 AT Telephone interview Weimar et al.39 Observational cohort, 1–2 years Cerebrovascular disease with AF Intracerebral haemorrhage 293 AC Patient interview Xu et al.40 Prospective cohort, 1-year Stroke Hypertension – 7880 AH Telephone interview AC: anti-coagulants; AF: atrial fibrillation; AH: anti-hypertensives; AP: anti-platelets; AT: anti-thrombotics; BMQ: beliefs about medicines questionnaire; HS: haemorrhagic stroke; IS: ischaemic stroke; LLD: lipid-lowering drugs; MARS: medication adherence report scale; MMAS: Morisky-medication adherence scale.

terview AC: anti-coagulants; AF: atrial fibrillation; AH: anti-hypertensives; AP: anti-platelets; AT: anti-thrombotics; BMQ: beliefs about medicines questionnaire; HS: haemorrhagic stroke; IS: ischaemic stroke; LLD: lipid-lowering drugs; MARS: medication adherence report scale; MMAS: Morisky-medication adherence scale. Description of predictive factors for non-adherence Two studies showed no difference in predictors within groups. One compared factors between rural and urban residence35 and the other compared patients living in different income quintiles.28 Factors related to non-adherence in the other 27 studies are classified below and detailed in the supplementary appendix. Patient-related factors Younger age at time of stroke was associated with reduced medication adherence in seven studies9,10,18,24,26,33,34 whereas younger age reported to associate with better adherence in five studies.2,29,36,39,40 Three studies reported that female sex predicted decreased adherence2,29,32 whereas one reported the opposite.37 Other patient-related factors included having concerns about medication, which associated with decreased adherence in four studies,10,12,27,30 or when patients perceived no benefit of treatment as reported in one study.10 On the other hand, when patients had positive beliefs about medication23,25,37 and indicated they were aware of the consequence of not taking prescribed medication,23 these factors were associated with enhanced adherence to medication.

27,30 or when patients perceived no benefit of treatment as reported in one study.10 On the other hand, when patients had positive beliefs about medication23,25,37 and indicated they were aware of the consequence of not taking prescribed medication,23 these factors were associated with enhanced adherence to medication. Socioeconomic factors Three studies indicated that having some sort of education21,40 or settled work status18 were associated with improved adherence. Four studies reported that the presence of patient carer or supporter also predicted better adherence.2,23,25,29 Two studies reported that living at care institution other than home was associated with worsened adherence.2,39 Therapy-related factors Disease- or health-related factors that predicted non-adherence included disability,1,9,18,29,37,39 reduced cognition function,10,23,25,37 poor quality of life2,11,18 and low mood.2,25 Smoking9,34 and alcohol consumption34,40 were also predictors of medication non-adherence. Existence of co-morbidities at the time of stroke associated with improved adherence to treatment. These included history of hypertension,18,29,34 diabetes,2,18 dyslipidaemia,18,21,40 coronary artery disease18,40 or myocardial infarction.18,33 Conversely, the absent history of atrial fibrillation was associated with better adherence.2,18,29,36,40

at the time of stroke associated with improved adherence to treatment. These included history of hypertension,18,29,34 diabetes,2,18 dyslipidaemia,18,21,40 coronary artery disease18,40 or myocardial infarction.18,33 Conversely, the absent history of atrial fibrillation was associated with better adherence.2,18,29,36,40 Prescribed regimen factors that predicted enhanced adherence included understanding of medication rationale,1,18,23,30 awareness of duration of treatment,30 knowledge of how to refill prescription,18 previous treatment by the same medication class,2,38,40 prescription and education at hospital discharge after the incident.20 Also, development of medication routine23 and use of compliance aid by patient.1 Medication regimen factors which associated with reduced adherence included cost of medication9,19,22 and number and frequency of prescribed drugs.1,9,18,29 Health system or caregiver-related factors Caregiver-related factors included prescriber speciality (e.g. neurologist).1 Patient–caregiver relationship factors included language barrier, low trust, perceived discrimination, inadequate continuity of care1 and inadequate communication of information regarding prescribed regimen.30 Institution factors associated with better adherence included treating facility i.e. treated in stroke unit,2,37 treated in academic hospital29 and hospital size.18 Additionally, arrangement of medical insurance11,24 and accessible health care facility2,12 predicted enhanced adherence.

Health system or caregiver-related factors Caregiver-related factors included prescriber speciality (e.g. neurologist).1 Patient–caregiver relationship factors included language barrier, low trust, perceived discrimination, inadequate continuity of care1 and inadequate communication of information regarding prescribed regimen.30 Institution factors associated with better adherence included treating facility i.e. treated in stroke unit,2,37 treated in academic hospital29 and hospital size.18 Additionally, arrangement of medical insurance11,24 and accessible health care facility2,12 predicted enhanced adherence. Stroke-related factors Stroke-related factors that predicted non-adherence included delay from onset of symptoms to evaluation,34 symptoms of post-traumatic stress disorder (PTSD),27,31 more severe stroke,33,36,39,40 previous stroke incidence2,9,37 and time from stroke onset.27 Stroke subtype was another predictor of non-adherence e.g. ischaemic stroke versus Tia,29 cardio-embolic36 and haemorrhagic stroke.2 Nevertheless, factors like reduced cognition, disability and poor quality of life could also be stroke-related. Meta-analysis Sixteen studies were eligible for the meta-analysis of prevalence of non-adherence as they provided a measure of medication non-adherence rate.1,11,20–22,26,27,29–31,33–35,37,39,40 The rate of non-adherence was 30.9% (95% CI 26.8–35.3%) (Figure 2). Figure 2. Meta-analysis of prevalence of non-adherence within included studies.

studies were eligible for the meta-analysis of prevalence of non-adherence as they provided a measure of medication non-adherence rate.1,11,20–22,26,27,29–31,33–35,37,39,40 The rate of non-adherence was 30.9% (95% CI 26.8–35.3%) (Figure 2). Figure 2. Meta-analysis of prevalence of non-adherence within included studies. For the meta-analysis of effect of factors on medication adherence, four factors were eligible which were: absent history of AF (4 studies2,18,29,36), disability (5 studies1,9,18,29,39), polypharmacy (4 studies1,9,18,29) and age of the patient (7 studies2,9,18,29,36,39,40). Meta-analyses of these factors showed that these factors did not significantly associate with medication adherence (no AF OR 1.02, 95% CI 0.72–1.5 (p = 0.9); disability OR 1.27, 95% CI 0.93–1.72 (p = 0.13); polypharmacy OR 1.29, 95% CI 0.9–1.9 (p = 0.17); age OR 1.04, 95% CI 0.96–1.14 (p = 0.34)). Forest plots for each factor analysis are available in Figure 3. There was considerable heterogeneity across all studies included in the meta-analyses (all I2 > 88%). Figure 3. Meta-analyses of predictive factors. Discussion In this review, we identified factors associated with adherence behaviour to secondary preventative medication after stroke or TIA. As stated by the WHO, patients alone used to be held responsible for non-adherence; however, it has been identified that other factors including the health care system or providers can also impact on non-adherence.17

d factors associated with adherence behaviour to secondary preventative medication after stroke or TIA. As stated by the WHO, patients alone used to be held responsible for non-adherence; however, it has been identified that other factors including the health care system or providers can also impact on non-adherence.17 Many factors associated with enhanced adherence to secondary preventative medication including positive beliefs about medication.23,25,37 This also included patients who encountered lower cost of medications9,19,22 or had medical insurance.11,24 Most of the published work focusses on patient and drug specific factors as determinants of adherence. The importance of institution or health care factors should not be neglected. Prescribing and educating patients on medication for secondary prevention before hospital discharge was linked to improved adherence.20 Numerous studies showed that in-hospital initiation of secondary preventative medication resulted in higher rates of adherence.20,43,44 This should include details on the purpose of treatment and regimen dosage.1,18,23,30 Also, patients should be ensured adequate continuity of care1 and access to health care after stroke.2,12 These simple measures could improve clinical outcomes. Nonetheless, stroke patients with disability,1,9,18,29,37,39 reduced cognitive function,10,23,25,37 increased number of prescribed medication,1,9,18,29 concerns about treatment,10,12,27,30 history of stroke2,9,37 or more severe stroke event33,36,39,40 commonly showed reduced adherence to treatment.

Most of the published work focusses on patient and drug specific factors as determinants of adherence. The importance of institution or health care factors should not be neglected. Prescribing and educating patients on medication for secondary prevention before hospital discharge was linked to improved adherence.20 Numerous studies showed that in-hospital initiation of secondary preventative medication resulted in higher rates of adherence.20,43,44 This should include details on the purpose of treatment and regimen dosage.1,18,23,30 Also, patients should be ensured adequate continuity of care1 and access to health care after stroke.2,12 These simple measures could improve clinical outcomes. Nonetheless, stroke patients with disability,1,9,18,29,37,39 reduced cognitive function,10,23,25,37 increased number of prescribed medication,1,9,18,29 concerns about treatment,10,12,27,30 history of stroke2,9,37 or more severe stroke event33,36,39,40 commonly showed reduced adherence to treatment. Factors reported in this review were similar to those reported to correlate with adherence to medication in cardiovascular disease including coronary heart disease and acute coronary syndrome45–48 and to medications in general.49,50

Nonetheless, stroke patients with disability,1,9,18,29,37,39 reduced cognitive function,10,23,25,37 increased number of prescribed medication,1,9,18,29 concerns about treatment,10,12,27,30 history of stroke2,9,37 or more severe stroke event33,36,39,40 commonly showed reduced adherence to treatment. Factors reported in this review were similar to those reported to correlate with adherence to medication in cardiovascular disease including coronary heart disease and acute coronary syndrome45–48 and to medications in general.49,50 Two patient-related factors were controversial in predicting adherence to secondary preventative medication, age at the time of stroke incident2,9,10,18,24,29,33,34,36,39,40 and sex of the patient.2,29,32,37 A study that assessed differences in prescribing secondary preventative drugs to stroke patients found significant differences where women were less likely to receive all recommended secondary preventative medication classes than men. However, younger patients were less likely to receive anti-platelet treatment.51 These factors are, however, non-reversible or amendable thus health care practitioners need to not hesitate with secondary prevention therapy if prescribing does not contrast with evidence-based recommendations. In the meta-analysis of prevalence of non-adherence, we found non-adherence to be high with almost a third of stroke patients not receiving adequate secondary prevention. This clearly indicates importance for applying interventions that would improve adherence especially in the group vulnerable for non-adherence.

Two patient-related factors were controversial in predicting adherence to secondary preventative medication, age at the time of stroke incident2,9,10,18,24,29,33,34,36,39,40 and sex of the patient.2,29,32,37 A study that assessed differences in prescribing secondary preventative drugs to stroke patients found significant differences where women were less likely to receive all recommended secondary preventative medication classes than men. However, younger patients were less likely to receive anti-platelet treatment.51 These factors are, however, non-reversible or amendable thus health care practitioners need to not hesitate with secondary prevention therapy if prescribing does not contrast with evidence-based recommendations. In the meta-analysis of prevalence of non-adherence, we found non-adherence to be high with almost a third of stroke patients not receiving adequate secondary prevention. This clearly indicates importance for applying interventions that would improve adherence especially in the group vulnerable for non-adherence. Despite the fact that none of the factors meta-analysed in this review showed significant association with medication adherence, caution should be taken not to interpret that association does not exist. This is explainable by the heterogeneity within included studies which was due to the considerable variation in subjects’ inclusion criteria, factors reported, medication classes, definition of adherence or compliance and the analysis used.

ence, caution should be taken not to interpret that association does not exist. This is explainable by the heterogeneity within included studies which was due to the considerable variation in subjects’ inclusion criteria, factors reported, medication classes, definition of adherence or compliance and the analysis used. Limitations There were several limitations of this review. Available data are heterogeneous as a result of lack of universal reporting of medication adherence. In addition, there was no standardised scale to critically appraise type of included studies. Also, inclusion and exclusion specification could have influenced reporting predictors e.g. if a study excluded participants of specific age or population who are known to have a high risk of non-adherence.

dication adherence. In addition, there was no standardised scale to critically appraise type of included studies. Also, inclusion and exclusion specification could have influenced reporting predictors e.g. if a study excluded participants of specific age or population who are known to have a high risk of non-adherence. Implication for practice and future research In this review, we aimed to identify factors correlated with adherence to secondary preventative medication after stroke. When clinicians are able to discuss barriers of adherence with their patients, they could ensure reducing the burden of treatment on their patients. It is also essential to identify reversible factors, e.g. misbeliefs or complex regimens, as these can be addressed. On the other hand, knowing factors that encourage stroke patients to adhere, clinicians would also be able to support stroke patients who are already adhering to maintain a good level of adherence. Researchers need to identify which interventions work best in supporting stroke patients to safely continue treatment with secondary preventative medication. Also, measures for detecting and tackling difficulties for medication administration after stroke need to be tested and implemented.

in a good level of adherence. Researchers need to identify which interventions work best in supporting stroke patients to safely continue treatment with secondary preventative medication. Also, measures for detecting and tackling difficulties for medication administration after stroke need to be tested and implemented. Conclusion Potential stroke patients with identified factors that predicted non-adherence require further attention, continuous encouragement and support with medication intake. Factors frequently reported to affect adherence included concerns about treatment regimen, increased disability, suffering severe stroke, polypharmacy and complex medication regimen. Focus should be more on reversible factors such as correcting misbeliefs about medication and providing convenient regimen. Stroke patients with disability or reduced cognition should be given additional care. Supplementary Material Supplementary material Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr Quinn is supported by a Joint Stroke Association and Chief Scientist Office Senior Clinical Lectureship. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Informed consent N/A Ethical approval N/A, not required Guarantor JD

Supplementary Material Supplementary material Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr Quinn is supported by a Joint Stroke Association and Chief Scientist Office Senior Clinical Lectureship. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Informed consent N/A Ethical approval N/A, not required Guarantor JD Contributorship SA– review protocol, literature research, data collection, data analysis and interpretation; figures, tables and quality assessment of papers; writing of the manuscript. TQ – review protocol, data interpretation and revision of manuscript. WD – review protocol and literature research. MW – revision of manuscript. JD – supervision, review protocol, data interpretation, quality assessment of papers and revision of manuscript.

Introduction Changing population demographics will increase stroke prevalence and healthcare burden.1 With technological advances such as mechanical thrombectomy, and finite healthcare budgets, it is increasingly important to consider not just efficacy of new interventions but also the cost-effectiveness.2 Cost-effectiveness analyses are often based on the number of quality-adjusted life years (QALYs) that are gained from implementing a treatment. QALYs conveniently provide a combined estimate of both length and quality of life and can be used across a broad range of conditions, treatments and settings. Calculation of QALYs is dependent on (a) reliable measurements of patients’ health-related quality of life on at least two occasions and (b) the availability of accurate health utility (HU) estimates, which define and assign preference weights to each possible health state. HU are represented on a scale of <0 to 1, with 0 indicating equivalence with death, 1 representing perfect health, and negative values indicating states considered worse than death. HU can be derived using diverse health state measures (e.g. the European Quality of Life Scale (EQ-5D-3L), the Health Utilities Index (HUI)3 and the Assessment of Quality of Life (AQoL)4); by various elicitation methods (Standard Gamble, Time Trade Off (TTO) and Visual Analogue Scale (VAS)5); and from various elicitation sources or populations. Value sets are usually collected from the general population. They exist for a range of different countries and describe preference weights for a particular health state.

us elicitation methods (Standard Gamble, Time Trade Off (TTO) and Visual Analogue Scale (VAS)5); and from various elicitation sources or populations. Value sets are usually collected from the general population. They exist for a range of different countries and describe preference weights for a particular health state. For studies such as decision modelling that rely on existing sources of HU and include stroke as a possible health state, accurate HU must be generated.6 Currently for stroke, variation exists in the choice of elicitation method, and the generated HU show diversity within stroke as a condition.2 Stroke is characterised by a spectrum of functional outcomes; it is unfortunate that some calculated stroke HU have described only limited functional outcome states.6 Existing studies have described population characteristics for patients with HU < 0,7 have described methods to translate functional states into EQ-5D-3L utility values,8 have examined diversity in quality of life responses from participants from various countries, and have examined proxy respondents compared with self-reported outcomes.9 For international stroke trials, HU estimates derived from a single country may not be applicable to all available trial data. There are limited international data to describe the range of expected HU across all possible levels of function, generated using a range of value sets. We sought to better inform future cost-effectiveness analyses that require HU estimates for stroke, by generating international HU based on European Quality of Life Scale (EQ-5D-3L) scores at a common acute stroke trial endpoint (3 months following stroke), and mapped across a spectrum of functional outcomes, assessed using the modified Rankin Scale (mRS).

-effectiveness analyses that require HU estimates for stroke, by generating international HU based on European Quality of Life Scale (EQ-5D-3L) scores at a common acute stroke trial endpoint (3 months following stroke), and mapped across a spectrum of functional outcomes, assessed using the modified Rankin Scale (mRS). Methods Data We conducted retrospective analyses of pooled, anonymised, patient-level data from the Virtual International Stroke Trials Archive (VISTA)10 on demography, (age, sex, medical history), neurological impairment (National Institutes of Health Stroke Scale score (NIHSS)), functional outcome (mRS), Quality of Life (EQ-5D-3L) and country of enrolment. The mRS is a 7-point observational scale that describes level of dependency, and ranges from 0 (no symptoms at all) to 6 (dead). The EQ-5D-3L is a standardised measurement tool for health-related quality of life and includes domains of mobility, self-care, usual activities, pain/discomfort and anxiety/depression. It can also be completed by proxy for people unable to complete the questionnaire themselves. HU generation We utilised published country-specific preference weights (value sets)11–23 to calculate HU. Each published value set was elicited from general population samples from the respective countries, using the Time Trade off (TTO) method. These value sets were applied in turn to individual-level EQ-5D-3L health state descriptions based on the five domains of mobility, self-care, usual activities, pain/discomfort and anxiety/depression for patients in our dataset (online Supplement I).

from the respective countries, using the Time Trade off (TTO) method. These value sets were applied in turn to individual-level EQ-5D-3L health state descriptions based on the five domains of mobility, self-care, usual activities, pain/discomfort and anxiety/depression for patients in our dataset (online Supplement I). We applied each published value set to our data, stratifying by mRS score at 3 months to illustrate expected variation when applying any single value set to an international trial population, as commonly occurs in cost effectiveness analysis. We examined potential differences in the distributions of HU according to the value set applied, with the Wilcoxon signed-rank test, and using HU generated from the UK value set as a reference population. Supplementary analyses applied each published value set to the country-specific population from which it was derived; if populations existed where no country-specific value set was available, we applied the value set of the nearest neighbouring country.

sing HU generated from the UK value set as a reference population. Supplementary analyses applied each published value set to the country-specific population from which it was derived; if populations existed where no country-specific value set was available, we applied the value set of the nearest neighbouring country. Validation Ordinary least squares (OLS) regression24,25 is recommended as a method of estimating unknown parameters (such as HU) from existing data (e.g. mRS).26 OLS Regression examines error: the differences between predicted outcomes and reality, and attempts to fit a line through the data that minimises the sum of the squared errors. This method was also previously described by Rivero-Arias et al.8 We used OLS regression to generate an equation to estimate HU based on mRS scores from our international population. We examined the proportion of variation in HU that was explained by mRS, adjusting for patients’ age and baseline NIHSS. The National Institute for Health and Clinical Excellence (NICE) caution against over-fitting covariates in an OLS regression; age and NIHSS were selected due to the strength of their association with post-stroke outcomes in our dataset (p < 0.0001).

was explained by mRS, adjusting for patients’ age and baseline NIHSS. The National Institute for Health and Clinical Excellence (NICE) caution against over-fitting covariates in an OLS regression; age and NIHSS were selected due to the strength of their association with post-stroke outcomes in our dataset (p < 0.0001). For this regression analysis, we applied published value sets from the USA, UK, Spain, Germany, China and Poland. These value sets were selected as they were generated using the most robust sample sizes,11 were published in the EQ-5D-3L inventory and user guide, represented countries that were typically included in international multicentre RCTs, and/or represented areas where emerging stroke research datasets warranted the generation of robust HU estimates. Performance was assessed using goodness of fit (adjusted R-squared values). We described the clinical and demographic characteristics of our population to inform generalisability for application to other clinical stroke populations.

here emerging stroke research datasets warranted the generation of robust HU estimates. Performance was assessed using goodness of fit (adjusted R-squared values). We described the clinical and demographic characteristics of our population to inform generalisability for application to other clinical stroke populations. Results We identified and extracted eligible data on 4946 patients (mean age: 68.8 ± 12.6 years, 2231 (45%) female, baseline NIHSS: 12 ± 9; Table 1) for whom assessment of EQ-5D-3L and mRS had been performed. Our analysis dataset comprised patients who were alive and had complete mRS and EQ-5D-3L scores at 3 months following stroke; by 3 months 817 (17.0%) patients had died; complete data on EQ-5D-3L (76.4% subject and 21.8% proxy respondents) and mRS were available for 3858 patients (mean age: 67.5 ± 12.5, baseline NIHSS: 12 ± 5) and missing for 271. Thirty-six countries were represented in our analysis dataset (online Supplement II). Table 1. Baseline demography (n = 4946 participants). Variable Value Age (years); mean (SD) 68.8 (12.6) Baseline NIHSS; median (IQR) 12 (9) Gender (female); n (%) 2231 (45.1%) RTPA (yes); n (%) 1915 (38.7%) Diabetes (yes); n (%) 1135 (22.9%) Hypertension (yes); n (%) 3665 (74.1%) Atrial fibrillation (yes); n (%) 1271 (25.7%) Previous stroke (yes); n (%) 979 (19.8%) Transient ischaemic attack (yes); n (%) 409 (8.3%) Myocardial infarction (yes); n (%) 641 (13.0%) Congestive heart failure (yes); n (%) 467 (9.4%)

1915 (38.7%) Diabetes (yes); n (%) 1135 (22.9%) Hypertension (yes); n (%) 3665 (74.1%) Atrial fibrillation (yes); n (%) 1271 (25.7%) Previous stroke (yes); n (%) 979 (19.8%) Transient ischaemic attack (yes); n (%) 409 (8.3%) Myocardial infarction (yes); n (%) 641 (13.0%) Congestive heart failure (yes); n (%) 467 (9.4%) Age and initial stroke severity by NIHSS were largely comparable across countries having a sample size of more than 50 patients. Medical history and use of thrombolytics varied by country of enrolment particularly in those countries that enrolled fewer patients (Argentina, Brazil, Chile, Finland, Greece, Hong Kong, Italy, Malaysia, Mexico, Netherlands, Norway, South Africa, South Korea and Sweden; online Supplement II). HU estimates After applying each published value set to our international dataset in turn, for mRS = 0, mean HU ranged from 0.88 to 0.95; for mRS = 5, mean HU ranged from −0.48 to 0.22 (Table 2). HU for mRS = 5 were perceived as corresponding to a health state that was worse than death when applying value sets from Singapore (−0.48), Spain (−0.34), and the UK (−0.15). Similar HU ranges were observed when excluding proxy responses on EQ-5D-3L (online Supplement III). The Wilcoxon signed-rank test revealed significant differences between the HU distributions generated using each country’s value set, when compared with those generated using the UK value set (p < 0.0001 for each country; Table 2). Table 2. Mean HU derived using EQ-5D-3L, stratified by mRS.

-3L (online Supplement III). The Wilcoxon signed-rank test revealed significant differences between the HU distributions generated using each country’s value set, when compared with those generated using the UK value set (p < 0.0001 for each country; Table 2). Table 2. Mean HU derived using EQ-5D-3L, stratified by mRS. Value set applied to entire international dataset Modified Rankin Scale score at 3 months 0 (n = 529) 1 (n = 866) 2 (n = 633) 3 (n = 669) 4 (n = 825) 5 (n = 336) Wilcoxon signed-rank test for differences in distributions of HU, relative to UK distribution (p values) Australia 0.93 (0.13) 0.86 (0.16) 0.76 (0.17) 0.61 (0.21) 0.35 (0.27) 0.02 (0.18) <0.0001 China 0.92 (0.12) 0.84 (0.15) 0.73 (0.16) 0.58 (0.17) 0.37 (0.20) 0.15 (0.16) <0.0001 Denmark 0.91 (0.15) 0.83 (0.16) 0.73 (0.16) 0.61 (0.19) 0.37 (0.19) −0.02 (0.27) <0.0001 Germany 0.95 (0.12) 0.90 (0.14) 0.83 (0.18) 0.68 (0.23) 0.38 (0.27) 0.09 (0.18) <0.0001 Netherlands 0.91 (0.16) 0.83 (0.18) 0.73 (0.19) 0.59 (0.23) 0.35 (0.25) 0.12 (0.21) <0.0001 Poland 0.94 (0.11) 0.89 (0.12) 0.81 (0.14) 0.70 (0.20) 0.43 (0.29) 0.06 (0.27) <0.0001 Singapore 0.88 (0.21) 0.74 (0.28) 0.51 (0.30) 0.23 (0.32) −0.16 (0.33) −0.48 (0.22) <0.0001 South Korea 0.94 (0.10) 0.88 (0.11) 0.80 (0.12) 0.69 (0.15) 0.42 (0.25) 0.09 (0.18) <0.0001 Spain 0.93 (0.14) 0.85 (0.18) 0.72 (0.21) 0.51 (0.28) 0.09 (0.36) −0.34 (0.23) <0.0001 UK 0.90 (0.17) 0.82 (0.19) 0.70 (0.21) 0.53 (0.26) 0.20 (0.31) −0.15 (0.23) – USA 0.92 (0.12) 0.85 (0.14) 0.77 (0.14) 0.64 (0.17) 0.41 (0.22) 0.14 (0.15) <0.0001 Zimbabwe 0.92 (0.12) 0.85 (0.13) 0.75 (0.13) 0.63 (0.15) 0.45 (0.19) 0.22 (0.17) <0.0001 Note: HU displayed as mean (SD) and generated using available, published value sets.

0.82 (0.19) 0.70 (0.21) 0.53 (0.26) 0.20 (0.31) −0.15 (0.23) – USA 0.92 (0.12) 0.85 (0.14) 0.77 (0.14) 0.64 (0.17) 0.41 (0.22) 0.14 (0.15) <0.0001 Zimbabwe 0.92 (0.12) 0.85 (0.13) 0.75 (0.13) 0.63 (0.15) 0.45 (0.19) 0.22 (0.17) <0.0001 Note: HU displayed as mean (SD) and generated using available, published value sets. Online Supplement IV describes HU generated by applying country-specific value sets to appropriate sub-populations. We observed that for mRS = 0, the mean HU estimates ranged between 0.81 and 0.98. For mRS = 5, mean HU ranged between −0.48 and 0.27. Validation After applying OLS regression, mRS scores at 3 months accounted for 65–71% of the variation in the generated HU estimates. The HU generated using OLS regression were consistent with those generated by applying each value set to the analysis dataset (Table 3). For mRS = 0 mean HU ranged between 0.9 and 0.95; for mRS = 1, mean HU ranged between 0.81 and 0.9, and for mRS = 5, mean HU ranged between −0.33 and 0.15. Table 3. Mean HU calculated using OLS regression, stratified by mRS. Value set applied to analysis dataset Mean HU, by modified Rankin Scale score at 3 months Adjusted R2 0 1 2 3 4 5 China 68.8 0.92 0.84 0.73 0.58 0.37 0.15 Germany 65.8 0.95 0.90 0.83 0.69 0.38 0.09 Poland 63.7 0.94 0.89 0.81 0.70 0.44 0.07 Spain 71.2 0.92 0.84 0.72 0.51 0.09 −0.33 UK 65.0 0.90 0.81 0.70 0.53 0.20 −0.14 USA 67.4 0.92 0.84 0.72 0.51 0.09 −0.33

an HU, by modified Rankin Scale score at 3 months Adjusted R2 0 1 2 3 4 5 China 68.8 0.92 0.84 0.73 0.58 0.37 0.15 Germany 65.8 0.95 0.90 0.83 0.69 0.38 0.09 Poland 63.7 0.94 0.89 0.81 0.70 0.44 0.07 Spain 71.2 0.92 0.84 0.72 0.51 0.09 −0.33 UK 65.0 0.90 0.81 0.70 0.53 0.20 −0.14 USA 67.4 0.92 0.84 0.72 0.51 0.09 −0.33 Discussion We generated exemplar international acute stroke HU based on published value sets, describing case mix and stratifying by mRS at 3 months to better inform future cost-effectiveness analyses. The range of observed HU generated by applying each published value set to our international population was similar to those generated when using OLS regression, and when excluding proxy responses. For mRS of 0, the mean HU ranged between 0.88 and 0.95, indicating that even though these patients were by definition asymptomatic, there were extraneous influences on the individual that affected perception of their health state. mRS states can be assigned on the basis of physical disability, cognitive impairment or a combination of both. Furthermore, the EQ-5D-3L has five domains and within any mRS level, patients can exhibit variation in which EQ-5D-3L domains have been affected. Therefore, it is possible for considerable variation to exist in HU estimates within a single mRS level. It is also possible that scoring errors or inconsistencies on mRS and on EQ-5D-3L contribute to this variation.

and within any mRS level, patients can exhibit variation in which EQ-5D-3L domains have been affected. Therefore, it is possible for considerable variation to exist in HU estimates within a single mRS level. It is also possible that scoring errors or inconsistencies on mRS and on EQ-5D-3L contribute to this variation. We observed that application of different value sets resulted in significantly different distributions of HU (compared with UK values). Since value sets vary according to country, heterogeneity in HU is expected when these diverse value sets are applied to a single multi-centred international trial population. The variation in value sets could arise from differing access to, and levels of health care services available, as well as differing cultural perceptions of disability across participating countries. This issue applies to both health state measurement and health state valuation, and should be taken into consideration when selecting appropriate value sets to inform cost-effectiveness of an intervention.

h care services available, as well as differing cultural perceptions of disability across participating countries. This issue applies to both health state measurement and health state valuation, and should be taken into consideration when selecting appropriate value sets to inform cost-effectiveness of an intervention. Guidance is needed on the application of appropriate value sets for pooled analyses of international populations. The application of one value set to an international population is commonly practiced, often as a matter of convenience or because this approach is applied to the corresponding cost data. However, this approach has some limitations; between country differences exist in health-related QoL, costs of healthcare, the degree of social support available and cultural perceptions of disability. These differences are not captured when applying a single value set to an international population. Application of country-specific value sets increases the relevance of the generated HU to each country, but creates problems for pooling of data for analyses (which is often necessary to preserve sample size). Our supplementary analysis still necessitated the application of a single value set to multiple neighbouring countries (online Supplement IV). For example data from Germany, Switzerland, Italy, and Greece were analysed using the German value set. The latter countries have strong family support for stroke survivors, and the application of German preference weights to these participants may not fully capture subtle differences in health perceptions within the same mRS level. Similar issues arise with the application of the USA value set to Central and South American countries. Our application of country-specific value sets to appropriate populations (online Supplement IV) highlights a challenge when dealing with smaller subgroups. We observed that when applying a Nordic value set to Nordic countries, HU were greater for mRS = 2 (0.92), than for mRS = 0 (0.9). Similarly, applying the UK value set to UK participants, HU for mRS = 1 (0.9) was greater than HU for mRS = 0 (0.81). This difference could be attributed to participant heterogeneity. Pooling data hides the country-level issues, and results are often not specifically relevant to any participating counties, while sub-group analyses carry analytical deficiencies.

participants, HU for mRS = 1 (0.9) was greater than HU for mRS = 0 (0.81). This difference could be attributed to participant heterogeneity. Pooling data hides the country-level issues, and results are often not specifically relevant to any participating counties, while sub-group analyses carry analytical deficiencies. There is often a trade-off between the availability and appropriateness of value sets for use in an international population, and preservation of a large enough sample size on which inferences can be made on health perception and cost-effectiveness. Debate also exists over the appropriate participant population from which to derive HU estimates.6 Those at risk of stroke are traditionally seen to be more suited to inform decisions from a patient’s perspective.27 However preference values derived from hypothetical scenarios may not be valid predictors of the preferences associated with actual experienced health states28; stroke survivors typically assign higher values to health states than those at risk of stroke, or healthy participants.6 Nevertheless, preference weight estimates from the general population are recommended when assessing cost-effectiveness from a societal perspective.6,27

iated with actual experienced health states28; stroke survivors typically assign higher values to health states than those at risk of stroke, or healthy participants.6 Nevertheless, preference weight estimates from the general population are recommended when assessing cost-effectiveness from a societal perspective.6,27 We described HU generated from application of both single and country-specific value sets to an international population. Previous studies have utilised a single country’s value set,7 described HU generated from a range of stroke and non-stroke populations, or stratified by broad categories of disability (minor stroke = mRS 2–3, major stroke = mRS 4–5).6 Previous estimates elicited from stroke survivors using the EQ-5D-3L described utilities of 0.71 and 0.32 for minor and major stroke, respectively.6 This contrasts with our findings where we observed a much wider HU range for the transition from mRS 2 to 5 (from 0.83 to −0.48). Although our data give HU values that differ from previously published estimates, our results are still within a range that would seem credible based on previous work.6 Furthermore, our generation of HU based on mapping approaches (Table 3) are consistent overall with HU generated from a prior study by Rivero-Arias et al.,8 though it should be noted that their study generated HU at different time points post-stroke.

till within a range that would seem credible based on previous work.6 Furthermore, our generation of HU based on mapping approaches (Table 3) are consistent overall with HU generated from a prior study by Rivero-Arias et al.,8 though it should be noted that their study generated HU at different time points post-stroke. Our approach to HU had a number of strengths. Our data are representative of the range of respondents that are typical in acute stroke RCTs. We employed OLS regression24 to validate our estimates. We used a generic patient reported outcome measure (EQ-5D-3L) that has been specified as a preferred method of utility measurement in clinical trials.24 Our analysis includes a much larger and more geographically diverse patient population than examined in previous studies. Baseline data suggest that included patients are broadly representative of acute stroke trial cohorts.

-3L) that has been specified as a preferred method of utility measurement in clinical trials.24 Our analysis includes a much larger and more geographically diverse patient population than examined in previous studies. Baseline data suggest that included patients are broadly representative of acute stroke trial cohorts. A limitation of our study is that perspectives on health states may change according to the time since stroke, and the values elicited based on EQ-5D-3L may not fully capture information from some patient subgroups such as those with communication problems. Those with cognitive or visual problems may rely on proxies to complete the EQ-5D-3L and thus their views may not be accurately represented. However, in our analysis dataset, 76.4% of EQ-5D-3L responses were elicited from stroke survivors themselves. Additionally, we analysed data only from those who had complete scores on all domains of EQ-5D-3L at 3 months; this may have biased the sample sizes available at higher levels of dependence. Furthermore, our data are based on an acute stroke clinical trial population. The HU generated for each stratum of mRS are therefore based on the experiences of a subgroup of the general stroke population. Future work could examine the generalisability of the HU generated in our population to general stroke population, and additional work is needed to examine the minimum sample size required for reliable country-specific HU generation.

are therefore based on the experiences of a subgroup of the general stroke population. Future work could examine the generalisability of the HU generated in our population to general stroke population, and additional work is needed to examine the minimum sample size required for reliable country-specific HU generation. Our study is based on acute stroke clinical trial data including information on dependency at a common endpoint, and involving patients from countries typically represented in acute stroke trials. Our findings can inform cost-effectiveness analyses of interventions in the acute stroke setting by providing conservative estimates of HU across a range of dependency levels; this may be of particular use to study designs reliant on secondary data sources, for example decision models. HU could feasibly be calculated in future studies through the collection of EQ-5D-3L data in parallel with common trial outcomes such as mRS. As more people survive stroke with long term disability,1 cost-effectiveness analyses should take into consideration whether an intervention has longer-term benefits for stroke survivors. Generation of HU for various levels of dependency at longer time points post-stroke is desirable. Future research could also involve calculation of the adjustment factors needed to convert known mRS distributions to HU according to age and sex, to refine our current estimates.

longer-term benefits for stroke survivors. Generation of HU for various levels of dependency at longer time points post-stroke is desirable. Future research could also involve calculation of the adjustment factors needed to convert known mRS distributions to HU according to age and sex, to refine our current estimates. Supplementary Material Supplementary material Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: TJQ has created investigator training materials for commonly used stroke outcomes and has received honoraria from Training Campus for this work. KRL is Former President of the European Stroke Organisation (ESO) and chairman of VISTA, which held a joint ESO-VISTA workshop to collate and harmonise health economic data in stroke. KRL and PB were invited speakers, and MA and MB were participants at this workshop. ESO receives funding from numerous industry sponsors but none had influence over the analysis or reporting of the material in this manuscript.

hich held a joint ESO-VISTA workshop to collate and harmonise health economic data in stroke. KRL and PB were invited speakers, and MA and MB were participants at this workshop. ESO receives funding from numerous industry sponsors but none had influence over the analysis or reporting of the material in this manuscript. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by a grant from Genentech. MB and the NMAHP Research Unit are funded by the Chief Scientist Office, (CSO) Scottish Government’s Health and Social Care Directorate, Scotland. The views expressed here are those of the authors and not necessarily those of the CSO. TJQ is supported by a Chief Scientist Office/Stroke Association Senior Lecturer Fellowship. PB is Stroke Association Professor of Stroke Medicine. VISTA is a not-for-profit collaboration of researchers from academia and commercial organisations. The VISTA Steering Committee members have each contributed to the organisation of contributing trials, and where these have involved industry support, they have acknowledged that within the original publications.

edicine. VISTA is a not-for-profit collaboration of researchers from academia and commercial organisations. The VISTA Steering Committee members have each contributed to the organisation of contributing trials, and where these have involved industry support, they have acknowledged that within the original publications. Ethical approval and informed consent We did not collect prospective data from patients for this project. Ethical committee permission and informed consent was not required as we used existing, anonymised clinical trial data, which had already been contributed to VISTA. The original trials have each sought ethical approval to conduct research and share these data, as well as obtaining informed consent. Guarantor MA. Contributorship MA, TQ, MB, KRL, YX and DLV researched literature and conceived the study. All authors were involved in protocol development, RI and MA were involved in data analysis. All authors were actively involved in drafting, reviewing and editing the manuscript and approved the final version of the manuscript.

Background People who suffer an ischaemic stroke are at risk of cognitive decline and recurrent vascular events.1–3 Guidelines recommend use of anti-thrombotic, blood pressure (BP) lowering and lipid lowering agents to reduce subsequent risk. These treatments reduce, but do not eliminate, risk of recurrent stroke and other vascular events and the effect of these treatments on cognitive decline is unclear.4,5 Since white matter hyperintensities (WMHs) on brain magnetic resonance imaging (MRI) are associated with higher risk of recurrent stroke and cognitive decline,6 it is hypothesised that these represent a biomarker of brain ischaemic injury. WMHs are seen in as many as 90% of people with ischaemic stroke,7,8 and the highest degrees of WMH burden are associated with higher rates of stroke, death and cognitive and physical decline.7,9 The burden of WMH often increases during longitudinal follow-up and such increases are also associated with increased incident stroke, dementia and cognitive decline.6 Thus, treatments that reduce WMH progression could improve several outcomes after stroke including cognition, functional outcome and recurrent stroke.

9 The burden of WMH often increases during longitudinal follow-up and such increases are also associated with increased incident stroke, dementia and cognitive decline.6 Thus, treatments that reduce WMH progression could improve several outcomes after stroke including cognition, functional outcome and recurrent stroke. Allopurinol, a drug commonly prescribed for the prophylaxis of gout, inhibits activity of xanthine oxidase leading to reduction in both serum uric acid (UA) and oxidative stress via reduced superoxide anion production. Higher serum UA is associated with increased risk of stroke,10 with adverse outcomes after ischaemic stroke11,12 and with vascular cognitive impairment. Allopurinol may deliver benefits independent of UA reduction. In people with history of stroke, allopurinol has been shown to reduce markers of inflammation,13 reduce augmentation index, reduce progression of carotid intima-media thickness and lower BP14 and has been shown to improve cerebral nitric oxide bioavailability.15 Thus, allopurinol has a number of effects on the vasculature, on both small and large cerebral arteries, which may make it an effective drug for stroke prevention. However, trials in people with stroke have typically been small, of short duration, and have not assessed progression of biomarkers of cerebrovascular disease or clinical outcomes.

number of effects on the vasculature, on both small and large cerebral arteries, which may make it an effective drug for stroke prevention. However, trials in people with stroke have typically been small, of short duration, and have not assessed progression of biomarkers of cerebrovascular disease or clinical outcomes. The xanthine oxidase inhibition for improvement of long-term outcomes following ischaemic stroke and transient ischaemic attack (XILO-FIST) trial is designed to test whether allopurinol reduces the rate of WMH progression and BP in people with recent ischaemic stroke. Methods/design XILO-FIST is a randomised, double-blind, placebo-controlled clinical trial of allopurinol 300 mg twice daily vs. placebo in 464 people with recent ischaemic stroke. The study is being conducted in stroke units in the UK Clinical Research Networks. The trial includes a sub-study of additional cardiac MRI in participants with left ventricular hypertrophy (LVH) and some participants also undergo carotid MRI. Trial status The first participant visit was in May 2015. As of 27 February 2018, the trial was open in 23 sites in the UK. A total of 418 participants were enrolled and 354 participants were randomised (with most of the remainder in the trial run in phase). On average, 12 participants per month are being randomised. One hundred and eighty-eight participants have completed one-year follow up; 57 participants have completed the two-year follow-up. We aim to finish recruitment by May 2018.

4 participants were randomised (with most of the remainder in the trial run in phase). On average, 12 participants per month are being randomised. One hundred and eighty-eight participants have completed one-year follow up; 57 participants have completed the two-year follow-up. We aim to finish recruitment by May 2018. Ethical and regulatory approval Ethics committee and regulatory approval has been obtained for all participating sites (REC number 14/WS/0113). The trial is being conducted in accordance with local regulations and UK law. Inclusion and exclusion criteria Participants are aged greater than 50 years and have suffered an ischaemic stroke or transient ischaemic attack (TIA) with positive imaging within the past four weeks. Ischaemic stroke and TIA are diagnosed by a stroke specialist. Symptoms must last more than 24 h or for symptoms lasting less than 24 h (TIA), there must be either a relevant diffusion-weighted imaging (DWI) lesion on MRI or a corresponding lesion on CT. The corresponding lesion on CT can include evidence of cerebral small vessel disease. Full inclusion and exclusion criteria are given in Table 1. Participants with evidence of LVH on electrocardiography (ECG, according to the Sokolow-Lyon or Cornell voltage criteria) or on screening echocardiography (defined as posterior or septal wall thickness of >11 mm, or increased left ventricular mass (LVM) defined as baseline LVM index of > 115 g/m2 (men) or > 95 g/m2 (women), or LVM >162 g (men) or >224 g (women)) and no current atrial fibrillation are eligible for the cardiac sub-study.

a) or on screening echocardiography (defined as posterior or septal wall thickness of >11 mm, or increased left ventricular mass (LVM) defined as baseline LVM index of > 115 g/m2 (men) or > 95 g/m2 (women), or LVM >162 g (men) or >224 g (women)) and no current atrial fibrillation are eligible for the cardiac sub-study. Table 1. Full inclusion and exclusion criteria. Inclusion criteria Exclusion criteria Ischaemic stroke/ischaemic lesion on brain imaging in relevant anatomical territory in patients with transient ischaemic attack Modified Rankin scale score of 5 Age greater than 50 Diagnosis of dementia Consent within one month of stroke Cognitive impairment deemed sufficient to compromise capacity or comply with the protocol Dependent on daily help from others for basic activities prior to stroke Significant co-morbidity or frailty likely to cause death within 24 months Contra-indication to or indication for administration of allopurinol Concurrent azathioprine, 6-mercaptopurine therapy, other cytotoxic therapies, cyclosporin, theophylline and didanosine Significant hepatic impairment eGFR < 30 ml/min Contraindication to MRI scanning Women of childbearing potential Prisoners Active participation in another CTIMP or device trial or participation within the past month eGFR < 60 and of Korean, Han Chinese or Thai descent MRI: magnetic resonance imaging; eGFR: estimated glomerular filtration rate; CTIMP: Clinical trial of investigational medicinal product.

n of childbearing potential Prisoners Active participation in another CTIMP or device trial or participation within the past month eGFR < 60 and of Korean, Han Chinese or Thai descent MRI: magnetic resonance imaging; eGFR: estimated glomerular filtration rate; CTIMP: Clinical trial of investigational medicinal product. Participant identification and recruitment Potential participants are identified during in-patient stay in an acute stroke unit or in a cerebrovascular out-patient clinic. Eligibility is confirmed by a medically qualified investigator and participants must give their own informed consent. Study schedule The study comprises a four-week run-in phase and a 104-week treatment phase. The detailed participant schedule is shown in Table 2. Table 2. Participant schedule. Activity Run-in phase Treatment phase Day 0 Week 4 Week 0 Week 4 Week 13 Week 26 Week 52 Week 78 Week 104 Review eligibility ✓ ✓ Informed consent ✓ Optimise preventative therapy ✓ Clinical evaluationa ✓ ✓ ✓ ✓ ✓ ✓ ✓ Safety blood testsb ✓ ✓ ✓ ✓ ✓ ✓ ✓ Blood for uric acid level ✓ ✓ Blood/urine for biobanking ✓ ✓ ECG ✓ ✓ Echocardiography ✓ Determine cardiac sub-study eligibility ✓ MRI brain (± carotid MRI) ✓ ✓ ABPM ✓ ✓ ✓ Cardiac MRI,c n=100 ✓ ✓ Detailed cognitive function evaluation ✓ ✓ Assessment of run in completion ✓ Randomisation ✓ Dispense ✓ ✓ ✓ ✓ ✓ ✓ Return/count IMP ✓ ✓ ✓ ✓ ✓ ✓ Adverse event review ✓ ✓ ✓ ✓ ✓ ✓ ✓ Note that the run in week 4 and treatment phase week 0 visits can take place concurrently. Participants will also be contacted by telephone at week 105, which marks the end of the study.

ssessment of run in completion ✓ Randomisation ✓ Dispense ✓ ✓ ✓ ✓ ✓ ✓ Return/count IMP ✓ ✓ ✓ ✓ ✓ ✓ Adverse event review ✓ ✓ ✓ ✓ ✓ ✓ ✓ Note that the run in week 4 and treatment phase week 0 visits can take place concurrently. Participants will also be contacted by telephone at week 105, which marks the end of the study. IMP: investigational medicinal product; ABPM: ambulatory blood pressure monitoring; ECG: electrocardiography; MRI: magnetic resonance imaging. aMeasures of stroke severity at week 4, modified Rankin scale score and MoCA at week 52 and blood pressure at all visits except week 78 and weight at week 104). bFBC, U + E, LFTs. cSub-study eligible participants only. Run-in phase The run-in phase comprises an enrolment visit on day 0 and a baseline assessment visit at four weeks. In order to successfully complete the run-in phase, participants must have had a medication review, completed baseline data collection and completed brain MRI. Those who do not successfully complete this phase are classed as screen failures. The baseline assessment visit includes a clinical evaluation (measurement of brachial sphygmomanometer BP), blood tests for safety and biobanking (including a blood test for serum UA to be centrally analysed), a urine sample for biobanking, ambulatory blood pressure monitoring (ABPM), ECG, brain MRI, assessment of cognitive function and assessment of eligibility for the cardiac sub-study (in participating sites only). In sites with experience of carotid plaque MRI, carotid MRI will be performed after the brain MRI scan.

ed), a urine sample for biobanking, ambulatory blood pressure monitoring (ABPM), ECG, brain MRI, assessment of cognitive function and assessment of eligibility for the cardiac sub-study (in participating sites only). In sites with experience of carotid plaque MRI, carotid MRI will be performed after the brain MRI scan. Randomisation is performed following successful completion of the run-in phase. Study medication then starts. The treatment phase Week 4 visit: Brachial sphygmomanometer BP is measured, safety bloods are taken, participants are asked regarding adverse events (AEs) and an ABPM is performed. Dose of study drug is increased at this visit if the criteria are met for dose escalation. Week 13/26/52/78 visit: Brachial BP is measured, safety bloods are taken (except at week 78), participants are asked regarding AEs and an ABPM is performed. A Montreal Cognitive Assessment (MoCA) is also performed at week 52. Week 104 visit: Brachial BP is measured, safety bloods and bloods for biobanking are taken, participants are asked regarding AEs and an ABPM is performed. Brain MRI and cognitive testing are repeated. Participants in the cardiac sub-study undergo an additional cardiac MRI. A further assessment for serious AEs (SAEs) is made at week 105 by telephone.

measured, safety bloods and bloods for biobanking are taken, participants are asked regarding AEs and an ABPM is performed. Brain MRI and cognitive testing are repeated. Participants in the cardiac sub-study undergo an additional cardiac MRI. A further assessment for serious AEs (SAEs) is made at week 105 by telephone. Randomisation Participants are randomised (1:1) to receive either allopurinol or placebo orally for 104 weeks. Randomisation codes are stored securely on the Robertson Centre for Biostatistics (RCB, University of Glasgow) network, with restricted access. Eighty per cent of participants are allocated to treatments by a minimisation algorithm which includes presence of WMH at baseline and sub-study eligibility as minimisation factors, Randomisation carried out via the study web portal. A telephone interactive voice response system is available as a backup.

ricted access. Eighty per cent of participants are allocated to treatments by a minimisation algorithm which includes presence of WMH at baseline and sub-study eligibility as minimisation factors, Randomisation carried out via the study web portal. A telephone interactive voice response system is available as a backup. Intervention Participants receive either allopurinol 300 mg or placebo twice daily for two years. During the first four weeks, a single 300 mg daily dose of allopurinol or placebo is taken. All participants then undergo dose titration to allopurinol 300 mg twice daily or placebo unless creatinine clearance (estimated via estimated glomerular filtration rate (eGFR)) is < 60 mL/min where once daily dosing is continued. Dose modification (a reduction from 300 mg twice daily to 300 mg once daily) occurs if renal function declines (to an eGFR of < 50 mL/min) or in the event of side effects. Dosing is stopped if renal function declines to an eGFR of < 30 mL/min. After the 104-week assessment, treatment with study medication stops. Blinding The study is double blind. An identical placebo is used. Changes in UA levels could compromise allocation concealment and these are not to be checked during the study. Unblinding should only occur in emergency situations where knowledge of the investigational product assignment is essential for the care of the participant. All investigators have received training in unblinding procedures.

levels could compromise allocation concealment and these are not to be checked during the study. Unblinding should only occur in emergency situations where knowledge of the investigational product assignment is essential for the care of the participant. All investigators have received training in unblinding procedures. Brain and carotid MRI Brain MRI is performed using 1.5 or 3T MRI, and in each individual, the same scanner should be used for baseline and follow-up. Study sequences include T1-weighted imaging, T2-weighted imaging and T2 fluid attenuation inversion recovery (FLAIR), DWI and susceptibility weighted imaging. Isotropic T1, T2 and FLAIR imaging will be performed where possible (Table 3). Carotid MRI imaging includes time of flight carotid angiography, black blood T1, black blood T2 and black blood proton density imaging of the carotid arteries. Table 3. Sequence parameters used for brain and carotid imaging in co-ordinating centre and on Siemens PRISMA 3T system.

Brain and carotid MRI Brain MRI is performed using 1.5 or 3T MRI, and in each individual, the same scanner should be used for baseline and follow-up. Study sequences include T1-weighted imaging, T2-weighted imaging and T2 fluid attenuation inversion recovery (FLAIR), DWI and susceptibility weighted imaging. Isotropic T1, T2 and FLAIR imaging will be performed where possible (Table 3). Carotid MRI imaging includes time of flight carotid angiography, black blood T1, black blood T2 and black blood proton density imaging of the carotid arteries. Table 3. Sequence parameters used for brain and carotid imaging in co-ordinating centre and on Siemens PRISMA 3T system. Scan Sequence Orientation TE TR TI Slice Thickness Slice gap Matrix FOV Slice number Total time Brain T1 TFL SAG 1.85 2000 900 1.0 50% 256 × 100 255 176 4.4 T2 SPC TRA 404 3000 – 0.9 – 256 × 100 230 176 5.32 FLAIR SPCIR SAG 397 5000 1800 1.0 – 256 × 100 255 160 4.02 DWI RESOLVE TRA 62 4100 – 4 30% 224 × 100 220 27 3.55 SWI SWI_r TRA 20 24 – 1.5 20% 256 × 95 230 96 4.45 DTI EPSE TRA 95 3600 – 4 30% 128 × 100 230 30 2.51 ASL EPFID TRA 11 3500 – 6 16% 64 × 100 255 20 6.06 Carotids TOF FL_r TRA 3.11 20 – 1.0 – 384 × 75 200 32/3 slabs 2.57 T1 tse TRA 17 740 – 2.0 50–200% 256 × 100% 140 5 to 11 2.37 T2 tse TRA 79 740 – 2.0 50–200% 192 × 100% 140 5–11 1.51 PD tse TRA 16 740 – 2.0 50–200% 192 × 100% 140 5–11 1.51 FLAIR: fluid attenuation inversion recovery; DWI: diffusion weighted imaging; SWI: susceptibility weighted imaging; DTI: diffusion tensor imaging. ASL: arterial spin labelling. FOV: field of view, TRA: transverse; SAG: sagittal; TOF: time of flight; PD: proton density; TE: echo time; TR: repitition time; TI: inversion time.

: fluid attenuation inversion recovery; DWI: diffusion weighted imaging; SWI: susceptibility weighted imaging; DTI: diffusion tensor imaging. ASL: arterial spin labelling. FOV: field of view, TRA: transverse; SAG: sagittal; TOF: time of flight; PD: proton density; TE: echo time; TR: repitition time; TI: inversion time. Assessment of WMH The STRIVE recommendations are followed during image review.16 WMH of presumed vascular origin are defined as hyperintense lesions on T2-FLAIR and can appear as isointense or hypointense (although often not as hypointense as CSF) on T1-weighted sequences. All scans are reviewed blinded to treatment allocation. A Fazekas and Scheltens scales is assigned.17,18 The Rotterdam progression score (RPS) and Schmidt’s progression score are calculated by simultaneous review of the baseline and two-year scans.19,20 All visual rating scales are assessed independently by two trained observers. Where there is any level of disagreement on a score, that score will be reviewed by at least two raters and a consensus score applied.

Schmidt’s progression score are calculated by simultaneous review of the baseline and two-year scans.19,20 All visual rating scales are assessed independently by two trained observers. Where there is any level of disagreement on a score, that score will be reviewed by at least two raters and a consensus score applied. Volumetric assessment of WMH volume will also be performed. The first step in automated extraction of WMH volumes is to estimate the white matter area in each subject using atlas-based segmentation.21 A probability map of white matter created from 313 volunteers aged 18–96 years is used,22 and this map is registered to each subject using non-linear (diffeomorphic) registration to provide an initial estimate of white matter in each subject.21 Hyperintense outliers are identified on T2 FLAIR by transforming each voxel to a standard (z) score. Voxels with z ≥ 1.5 and within the estimated white matter area are initially defined as WMH. Final WMH estimates are defined by 3D Gaussian smoothing to reduce noise and account for partial volumes around WMH edges. Automatic WMH estimates are visually checked and infarcts masked by a trained image analyst following STRIVE guidelines.16 Normal-appearing tissues including cortical grey matter, sub cortical grey matter, cerebral white matter and supratentorial cerebrospinal fluid are segmented using population-specific tissue probability maps, within-patient T1 intensity data, and adjoining voxel data.23,24 Normal-appearing tissue segmentations are checked and edited in the same manner as WMH.

al grey matter, sub cortical grey matter, cerebral white matter and supratentorial cerebrospinal fluid are segmented using population-specific tissue probability maps, within-patient T1 intensity data, and adjoining voxel data.23,24 Normal-appearing tissue segmentations are checked and edited in the same manner as WMH. Cardiac sub-study Participants in the sub-study undergo additional cardiac 3T MRI at baseline and two years. ABPM Twenty four-hour ABPM is performed at baseline, week 4 and week 104 unless contraindicated. A Spacelabs Ultralight Ambulatory Blood Pressure Monitor is used. This is set to take readings every 30 min during daytime (08:00 h–21:59 h) and every 60 min during night-time (22:00 h–07:59 h). ABPM data will not be performed in some participants with significant arm weakness due to safety concerns. Cognitive and quality of life measures The assessment of pre-stroke cognitive impairment uses the 16-item IQCODE.25 A score of 3.6 or greater is used as threshold to define probable pre-stroke dementia. A comprehensive cognitive examination will be performed at baseline and at the two-year follow-up. The battery comprises: the MoCA; Animal Naming test of semantic fluency; Controlled Oral Word Association Test; Letter Digit Coding Test; Hopkins Verbal Learning Test; Centre for Epidemiological Studies – Depression Scale (CES-D); neuropsychiatric inventory questionnaire version (final follow-up visit only) and a trail making test. The EQ-5D and the Stroke Impact Scale Short Form26 are also measured.

Oral Word Association Test; Letter Digit Coding Test; Hopkins Verbal Learning Test; Centre for Epidemiological Studies – Depression Scale (CES-D); neuropsychiatric inventory questionnaire version (final follow-up visit only) and a trail making test. The EQ-5D and the Stroke Impact Scale Short Form26 are also measured. The battery is administered by a trained assessor, scored to pre-specified marking sheets and are conducted in a standardised fashion. Participants are free to take breaks as needed. If participants are unable to complete the full battery, the assessor prioritises the MoCA and CES-D. Details of each of these scales are given in a detailed instruction booklet, which includes instructions for administering each assessment. Safety blood tests and pharmacovigilance Safety blood tests including a full blood count, urea and electrolytes and liver function tests are checked at all study visits with the exception of the week 78 visit. Blood for serum UA levels is obtained at the end of the run-in phase (baseline visit) and at week 104 during the treatment phase. Serum UA levels will be measured centrally.

luding a full blood count, urea and electrolytes and liver function tests are checked at all study visits with the exception of the week 78 visit. Blood for serum UA levels is obtained at the end of the run-in phase (baseline visit) and at week 104 during the treatment phase. Serum UA levels will be measured centrally. Predictable side effects of the investigation medicinal product (IMP) used in this trial (allopurinol) are referred to in the summary of medicinal product characteristics (SmPC). All SAEs occurring within the first 13 weeks of the treatment phase will be recorded and reported to the Sponsor. Thereafter and up to 30 days after completing the study, unexpected SAEs, suspected unexpected serious adverse reactions and events of special interest that meet criteria for an SAE will be reported. Criteria for stopping study medication and study withdrawal Participants developing a rash, fever, liver dysfunction (defined as bilirubin or transaminase levels increasing to three times the ULN), renal dysfunction (defined as a drop in eGFR to below 30 mL/min), eosinophilia (defined as an eosinophil count of > 0.45 × 109 or a fall in haemoglobin below 10 g/dL or neutrophil count of <1.5 × 109 on any blood sample of a platelet count of <50 × 109 that is not due to clumping cease taking study medication immediately and permanently if no alternative cause is found. We will not recruit additional participants to replace those who stop treatment and participants who withdraw from treatment will be asked to remain under follow-up.

e of a platelet count of <50 × 109 that is not due to clumping cease taking study medication immediately and permanently if no alternative cause is found. We will not recruit additional participants to replace those who stop treatment and participants who withdraw from treatment will be asked to remain under follow-up. Study outcomes The primary outcome is WMH progression measured using the RPS. Secondary outcomes and exploratory outcomes are shown in Table 4. The primary outcome in the cardiac sub-study is LVM index. Table 4. Primary, secondary and exploratory outcomes.

e of a platelet count of <50 × 109 that is not due to clumping cease taking study medication immediately and permanently if no alternative cause is found. We will not recruit additional participants to replace those who stop treatment and participants who withdraw from treatment will be asked to remain under follow-up. Study outcomes The primary outcome is WMH progression measured using the RPS. Secondary outcomes and exploratory outcomes are shown in Table 4. The primary outcome in the cardiac sub-study is LVM index. Table 4. Primary, secondary and exploratory outcomes. Primary outcome Secondary outcomes Exploratory outcomes Rotterdam progression scale Change in mean day-time systolic BP at one month Fazekas’ score Change in mean day-time diastolic BP at one month Schelten’s scale score Schmidt’s progression score Measures of BP variability WMH volume at two years One month mean day-time diastolic blood pressure Fazekas’ score at two years Two-year mean day-time diastolic BP Schelten’s scale score at two years Clinic brachial BP New brain infarction on MRI Incident atrial fibrillation Rotterdam progression score with those who die/become too frail to undergo repeat imaging assigned worst score Recurrent stroke MoCA score Recurrent myocardial infarction, stroke or cardiac death Change in mean day-time systolic BP at two years Hospitalisation for, or incident heart failure Change in mean day-time diastolic BP at two years Incident dementia Mortality Animal naming test Controlled word association test Hopkins verbal learning test Trail making test Quality of life (EQ-5D, SS-QOL) Modified Rankin scale score BP: blood pressure; MoCA: Montreal cognitive assessment.

or incident heart failure Change in mean day-time diastolic BP at two years Incident dementia Mortality Animal naming test Controlled word association test Hopkins verbal learning test Trail making test Quality of life (EQ-5D, SS-QOL) Modified Rankin scale score BP: blood pressure; MoCA: Montreal cognitive assessment. Statistical analysis Full details of all statistical issues and planned statistical analyses will be specified in a separate statistical analysis plan which will be agreed before database lock and unblinding of treatment codes to the study statisticians. All data will be summarised overall and by treatment group. Efficacy analyses will be carried out according to the intention-to-treat (ITT) principle, that is, in relation to randomised treatment allocation, regardless of treatments actually received. The population for ITT analyses will be all validly randomised participants, who do not have any major protocol violations. Additional analyses will be carried out using a per-protocol (PP) population, consisting of those members of the ITT population who remain on treatment as randomised and do not have any minor protocol violations. All protocol violations will be assessed and classified as major or minor, prior to database lock.

lations. Additional analyses will be carried out using a per-protocol (PP) population, consisting of those members of the ITT population who remain on treatment as randomised and do not have any minor protocol violations. All protocol violations will be assessed and classified as major or minor, prior to database lock. Unadjusted between-group comparisons of continuous and ordinal outcomes will be made using t-tests or Wilcoxon-Mann-Whitney tests as appropriate; categorical outcomes will be compared using Fisher’s exact tests. Adjusted analyses will use regression models, adjusted for variables used in the minimisation algorithm. Distributional assumptions will be assessed visually and where necessary, outcomes will be transformed, or a generalised linear model will be used with appropriate link and variance function. The model to be used for each outcome will be decided and documented prior to database lock. Treatment effect estimates will be reported with 95% confidence intervals and p-values. Analysis of primary outcome – WMH progression is not expected to follow a normal distribution. Unadjusted comparison of treatment groups will use a Wilcoxon-Mann-Whitney test. A generalised linear regression model, with appropriate link and variance function, will be used to model WMH progression in relation to treatment and variables used in the minimisation algorithm. A further model will be fitted adjusting for other baseline characteristics found to be associated with WMH progression during blinded analyses, prior to database lock.

ppropriate link and variance function, will be used to model WMH progression in relation to treatment and variables used in the minimisation algorithm. A further model will be fitted adjusting for other baseline characteristics found to be associated with WMH progression during blinded analyses, prior to database lock. This model will be extended to assess the mediating effects of other study outcomes, such as changes in BP and UA post-randomisation. Analysis of secondary outcomes – secondary efficacy outcomes will be analysed using appropriate two-sample tests followed by regression analyses to estimate between-group differences adjusted for variables used in the minimisation algorithm. For outcome measures recorded at baseline as well as at follow-up, regression models will be adjusted for the baseline value. For endpoints measured at several time points, each time point will first be analysed separately, and then a repeated measures model will be applied to model measurements at all time points simultaneously. Planned subgroup analyses – the moderating effects of baseline UA level in the primary analysis will be assessed through use of interaction terms in the model. These methods will also be used to investigate the moderating effects of other baseline characteristics in an exploratory manner.

and BW used Microsoft Excel to extract information from each RCT on pre-specified covariates that we hypothesised might be associated with completeness of reporting. Any uncertainties or disagreements about eligibility, completeness of reporting, or covariates were resolved by discussion with another reviewer (RA-SS). Data items Our primary outcome was a truncated version of the CONSORT checklist comprising the 10 most important CONSORT checklist items, identified by a group of experts from within the CONSORT group, based on their professional opinion and supported by empirical evidence where available (Table 2).30 We tackled the problem of partial reporting by adapting the wording of some criteria so that each item could be scored 1 if it was reported or 0 if it was not reported, for a total score ranging from 0 to 10. Table 2. Checklist items used for the truncated CONSORT score.

Analysis of secondary outcomes – secondary efficacy outcomes will be analysed using appropriate two-sample tests followed by regression analyses to estimate between-group differences adjusted for variables used in the minimisation algorithm. For outcome measures recorded at baseline as well as at follow-up, regression models will be adjusted for the baseline value. For endpoints measured at several time points, each time point will first be analysed separately, and then a repeated measures model will be applied to model measurements at all time points simultaneously. Planned subgroup analyses – the moderating effects of baseline UA level in the primary analysis will be assessed through use of interaction terms in the model. These methods will also be used to investigate the moderating effects of other baseline characteristics in an exploratory manner. Sample size We assumed that 90% of participants would have evidence of WMH at baseline and that, over two years, approximately 64% would progress by one point or more on the RPS and that the mean progression score in the placebo group will be 1.293. This and the progression rate seen with a 30% reduction in WMH progression score are shown in Table 3. Based on a Wilcoxon-Mann-Whitney test, a sample size of 192 participants per group would give 80% power to detect this difference at a 5% significance level (nQuery Advisor® v7.0). This treatment effect is substantially less than the 80% relative reduction seen in the PROGRESS MRI study.17

n score are shown in Table 3. Based on a Wilcoxon-Mann-Whitney test, a sample size of 192 participants per group would give 80% power to detect this difference at a 5% significance level (nQuery Advisor® v7.0). This treatment effect is substantially less than the 80% relative reduction seen in the PROGRESS MRI study.17 Although we will minimise loss to clinical follow-up (completed for all except one participant in our pilot study), we have increased sample size to account for a 10% drop out rate seen in our pilot study and by an additional small amount to account for the fact that those with no WMH at baseline may progress at a lower rate. We will thus randomise 232 participants per group (10% drop outs will give data on 209 participants giving a further 17 participants (8%) per group to ensure sufficient power). Other sample size calculations were based on two-sample t-tests. For ABPM, 101 participants per group will be required to give 80% power at a 5% significance level to verify the 3.3 mmHg reduction in systolic BP (SBP) seen in the recent meta-analysis18 (assumed SD of change in SBP 8.3). For the cardiac sub study, 25 participants per group would give 80% power to detect a 3.7 g difference in LV mass (assumed standard deviation 4.6). A 3.86 g difference was seen in a recent trial of allopurinol use in patients with type 2 diabetes.

P) seen in the recent meta-analysis18 (assumed SD of change in SBP 8.3). For the cardiac sub study, 25 participants per group would give 80% power to detect a 3.7 g difference in LV mass (assumed standard deviation 4.6). A 3.86 g difference was seen in a recent trial of allopurinol use in patients with type 2 diabetes. Trial and data management All study data will be held in the study Robertson Centre for Biostatistics, University of Glasgow, part of the UKCRC-registered Glasgow Clinical Trial Unit (number 16). A secure and restricted electronic data capture system (electronic case report form) will be used. A Trial Management Group (TMG) meets regularly during the study. Trial steering and data monitoring committees A Trial Steering Committee meets at least annually and comprises an independent chair, three other independent members, a participant and/or a carer representative, the chief investigator and statistician. An independent data monitoring committee (IDMC) meets at least annually. This comprises of four independent members. Only the IDMC has access to unblinded outcome data before the trial ends.

ent chair, three other independent members, a participant and/or a carer representative, the chief investigator and statistician. An independent data monitoring committee (IDMC) meets at least annually. This comprises of four independent members. Only the IDMC has access to unblinded outcome data before the trial ends. Summary and conclusion XILO-FIST is designed to test whether allopurinol reduces the rate of WMH progression and BP in participants with recent ischaemic stroke. The hypothesis underlying XILO-FIST is that UA reduction and xanthine oxidase inhibition will improve vascular health and lower BP leading to a reduction in progression of WMH progression on MRI. If WMH progression is reduced, allopurinol could be an effective preventative treatment for people with ischaemic stroke and clinical endpoint studies would be needed. If allopurinol reduces BP after stroke, then it could be used to help people reach BP targets.

to a reduction in progression of WMH progression on MRI. If WMH progression is reduced, allopurinol could be an effective preventative treatment for people with ischaemic stroke and clinical endpoint studies would be needed. If allopurinol reduces BP after stroke, then it could be used to help people reach BP targets. Acknowledgements The following people have contributed to recruitment, Lesley MacDonald, Dr Christine McAlpine, Ruth Graham, Dr Mark Barber, Derek Esson, Dr Alex Doney, Louise Cabrelli, Dr MJ Macleod, Vicky Taylor, Ganesh Subramanian, Gwendoline Wilkes, Dr Christopher Price, Vicky Riddell, Dr Alexander Dyker, Carole Hays, Dr Ajay Bhalla, Sagal Kullane, Dr Ahamad Hassan, Dean Waugh, Dr Sarim Omar, Sherma Turner, Dr Naweed Sattar, Andrew Smith, Dr Ramanathan Kirthivasan, Joanne Topliffe, Dr Ranjan Sanyal, Holly Maguire, Dr Louise Shaw, Suzanne Lucas, Dr Paul Guyler, Raji Prabakaran, Prof David Werring, Nina Francia, Dr Adam Webber, Vinodh Krishnamurthy, Lakshmanan Sekaran, Margaret Louise Tate, Dr John Corrigan, Caroline Blair, Dr Breffni Keegan, Sageet Amlani, Laura Howaniec. The following people have contributed to project management Pamela Mackenzie, Mairi Warren, Pamela Surtees, Dr Elizabeth Douglas, Dr Maureen Travers, Dr Marc Jones, Eileen Smillie, Dr Ewan Dougall, Louise Haggerty.

Margaret Louise Tate, Dr John Corrigan, Caroline Blair, Dr Breffni Keegan, Sageet Amlani, Laura Howaniec. The following people have contributed to project management Pamela Mackenzie, Mairi Warren, Pamela Surtees, Dr Elizabeth Douglas, Dr Maureen Travers, Dr Marc Jones, Eileen Smillie, Dr Ewan Dougall, Louise Haggerty. Authors’ note A writing committee will be convened and be responsible for writing all abstracts and manuscripts for publication. The writing committee will consist of the applicants and TSC chair and will be responsible for approving content and dissemination. Data will be shared after the trial is complete, all planned analysis have been completed and subject to data sharing agreements with the University of Glasgow. TSC members: Prof Philip Bath (Chair), Ms Shannon Amolis, Prof Jesse Dawson, Dr Kirsten Forbes, Dr Richard Francis, Prof Mark Kearney, Prof Kennedy Lees, Dr Alex McConnachie, Dr Marc Randall, Mr Charles Ross (Lay Member), and Prof Allan Struthers. IDMC members: Prof Peter Sandercock (Chair), Prof Gary Ford, Dr Chris Sutton, and Dr William Whiteley.

TSC members: Prof Philip Bath (Chair), Ms Shannon Amolis, Prof Jesse Dawson, Dr Kirsten Forbes, Dr Richard Francis, Prof Mark Kearney, Prof Kennedy Lees, Dr Alex McConnachie, Dr Marc Randall, Mr Charles Ross (Lay Member), and Prof Allan Struthers. IDMC members: Prof Peter Sandercock (Chair), Prof Gary Ford, Dr Chris Sutton, and Dr William Whiteley. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Jesse Dawson: Substantial academic grant was obtained by Prof Dawson for the execution of this trial. Niall Broomfield: Has nothing to disclose. Krishna Dani: Reports grants from Stroke Association, during the conduct of the study. David Alexander Dickie: Reports personal fees from DD Analytics Cubed Ltd, during the conduct of the study; personal fees from DD Analytics Cubed Ltd, outside the submitted work. Alex Doney: Has nothing to disclose. Kirsten Forbes: Has nothing to disclose. Graeme Houston: Has nothing to disclose. Sharon Kean: Has nothing to disclose. Kennedy Lees: Has nothing to disclose. Alex McConnachie: Reports grants from British Heart Foundation/Stroke Association, during the conduct of the study. Keith W Muir: Reports personal fees from Bayer plc, other from Boehringer Ingelheim, outside the submitted work. Terry Quinn: Has nothing to disclose. Allan Struthers: Reports that his institute will receive £800,000 from the BHF/Stroke Association. Struthers and his institute have a patent for the use of Allopurinol to treat chest pain in Angina Pectoris. Matthew Walters: Has nothing to disclose.

eim, outside the submitted work. Terry Quinn: Has nothing to disclose. Allan Struthers: Reports that his institute will receive £800,000 from the BHF/Stroke Association. Struthers and his institute have a patent for the use of Allopurinol to treat chest pain in Angina Pectoris. Matthew Walters: Has nothing to disclose. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This trial is financially supported by a joint Stroke Association and British Heart Foundation Programme Grant (ref: TSA BHF 2013/01). It is an investigator-led study. The work of David Dickie and Terry Quinn is funded by the Stroke Association. Ethics approval Ethics approval was granted by West of Scotland Research Ethics Committee 1 on 24 June 2014. All participants give informed consent. Informed consent All participants give informed consent. All authors consent to publication of this manuscript. Guarantor The Chief Investigator, Jesse Dawson, guarantees this work. Contributorship JD drafted the manuscript. All authors provided critical comment and contributed to the design of the study.

Introduction Strokes in atrial fibrillation (AF) patients are common and frequently devastating (70–80% of patients die or become disabled,1,2 yet these strokes are preventable with anticoagulant therapy: 64% reduction in the risk of stroke and 25% reduction in mortality.3 Recent evidence shows that women are more frequently affected by AF compared to men, and have a higher associated risk for thromboembolic events.4–6 Therefore, female sex has been added as an independent risk-factor when calculating the CHA2DS2-VASc-score.7 Despite of this increased risk, women with AF still tend to be less treated with anticoagulants.8,9 The “Early Recurrence and Cerebral Bleeding in Patients with Acute Ischemic Stroke and Atrial Fibrillation” (RAF) Study investigated for (1) the risk of recurrent ischemic event and severe bleeding; (2) the risk factors for recurrence and bleeding; and (3) the risks of recurrence and bleeding associated with anticoagulant therapy and its starting time after the acute stroke.10 The results of this study have been recently published.10 The aim of this study was to evaluate the sex-differences in patients with acute stroke and AF, regarding risk factors, treatments received and outcomes. Methods The methods and results of the RAF-study have been published recently.10 Briefly, RAF-study was performed between January 2012 and March 2014 and included 29 Stroke Units across Europe and Asia. All of the participating 29 Stroke Units provided standard stroke unit care and monitoring.10

Data items Our primary outcome was a truncated version of the CONSORT checklist comprising the 10 most important CONSORT checklist items, identified by a group of experts from within the CONSORT group, based on their professional opinion and supported by empirical evidence where available (Table 2).30 We tackled the problem of partial reporting by adapting the wording of some criteria so that each item could be scored 1 if it was reported or 0 if it was not reported, for a total score ranging from 0 to 10. Table 2. Checklist items used for the truncated CONSORT score. Criterion Description Outcomes Explicitly defined, pre-specified, primary outcome measure, including how and when they were assessed Sample size Justification for sample size Sequence generation Methods used to generate random allocation sequence Allocation concealment Explicitly state mechanism used to implement random allocation sequence (such as sealed envelopes or electronic sequence generation, and block sizes) and describe any steps taken to conceal the sequence until interventions were assigned (such as opaque nature of envelopes or a central telephone/web allocation centre). A description of both of these aspects was required to score a point. Where electronic sequence generation was used it had to be clear that this was concealed from researchers, and not predictable, either with a statement or example describing the use of a central allocation centre Blinding Clear statement about whether or not anyone (for example, participants’ care providers or those assessing outcomes) was blinded to interventions after assignment Outcome estimation For the primary outcome (identified as above), results for each group, the estimated effect size and its precision (i.e. 95% CI) Harms Mentions any harms or unintended effects in each group, or statement of no adverse effects Registration Registration number and the trial registry Protocol Where the trial protocol can be accessed, if available Funding Sources of funding and other support (such as supply of drugs) and role of funders CONSORT: Consolidated Standards Of Reporting Trials.

The aim of this study was to evaluate the sex-differences in patients with acute stroke and AF, regarding risk factors, treatments received and outcomes. Methods The methods and results of the RAF-study have been published recently.10 Briefly, RAF-study was performed between January 2012 and March 2014 and included 29 Stroke Units across Europe and Asia. All of the participating 29 Stroke Units provided standard stroke unit care and monitoring.10 On admission, stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS). All patients underwent cerebral computed tomography examination without contrast or cerebral magnetic resonance to exclude intracranial hemorrhage. Thrombolysis treatment was administered as per local standard protocol, when appropriate. All patients were monitored for blood pressure, temperature, glucose level, heart rate, and blood gases in the first days after stroke. The choice of anticoagulant treatment (low molecular weight heparin [LMWH] or oral anticoagulants), as well as the day of its initiation, was left to the discretion of the treating physicians. AF was classified as paroxysmal, persistent, or permanent. A second brain computed tomography scan or magnetic resonance was performed 24–72 h from stroke onset in all patients. Hemorrhagic transformation was defined as any degree of hyperdensity within the area of low attenuation and was classified as either hemorrhagic infarction or parenchymal hematoma.11,12 The sites and sizes of the qualifying infarcts were determined based on standard templates as small, medium, large anterior or large posterior infarctions.11,13,14 Data on known stroke risk factors and treatment were collected and reported in the main paper.10 The CHA2DS2-VASc score before the index event was also calculated. The standard protocol also included a transthoracic echocardiography (TTE) during the hospital stay. Patients were followed up prospectively through face-to-face or telephone interviews. Study outcomes were (1) recurrent ischemic cerebrovascular events (stroke or TIA) and symptomatic systemic embolism; (2) symptomatic cerebral bleedings and major extracerebral bleeding at 90 days. The primary outcome was the composite of stroke, TIA, symptomatic systemic embolism, symptomatic cerebral bleeding, and major extracerebral bleeding. Disability and mortality at 90 days were also assessed using the modified Rankin Scale (mRS). Functional outcome was defined as either favorable (mRS 0–2) or unfavorable (mRS 3–6).

y outcome was the composite of stroke, TIA, symptomatic systemic embolism, symptomatic cerebral bleeding, and major extracerebral bleeding. Disability and mortality at 90 days were also assessed using the modified Rankin Scale (mRS). Functional outcome was defined as either favorable (mRS 0–2) or unfavorable (mRS 3–6). Statistical analysis Continuous variables, as well as NIHSS score, were reported as mean ± SD, and categorical variables were reported as percentages. Pearson’s chi-square test was used to compare categorized proportions. A comparison of discrete variables was conducted using a non-parametric test (Mann–Whitney). Multivariate logistic regression was performed in order to investigate sex differences for dichotomous outcomes. Included variables were: age, vascular risk factors, NIHSS on admission, the type of AF, lesion size, antithrombotic treatment before and after stroke, previous use of statins. The decision concerning which variables to include and to adjust in the multivariable analysis was guided by either the presence of an a priori theoretical or biological relationship among the examined patient characteristics as well as primary endpoints.15 A two-sided p < 0.05 was considered significant for all statistical tests. All statistical analyses were performed using software SPSS/PC Win package 20.0.

s was guided by either the presence of an a priori theoretical or biological relationship among the examined patient characteristics as well as primary endpoints.15 A two-sided p < 0.05 was considered significant for all statistical tests. All statistical analyses were performed using software SPSS/PC Win package 20.0. Results Overall, 1037 consecutive patients were enrolled in the study (59 from Asia) and 1029 were included in the analysis (eight excluded for incomplete data). Overall, 561 women were included (54.5%, p < 0.001), and they were on average younger (p < 0.001) compared to men. There was no observed difference in AF subtypes (paroxysmal, persistent or permanent) between the sexes (Table 1). History of myocardial infarction was more common in men (p = 0.002), as was history of peripheral arterial disease (p = 0.003) and aortic atherosclerosis (p = 0.016). Also, men more often had a pacemaker (p = 0.023) and were more often taking statins at the time of stroke onset (p < 0.001), while smoking and alcohol abuse were more common in men (p < 0.001). Previous use of antiplatelet agents was not significantly different between the two groups, while the use of anticoagulants was less frequent in women (25.5% versus 31.9%, respectively, p = 0.026). Table 1. Baseline characteristics of patients.

.001), while smoking and alcohol abuse were more common in men (p < 0.001). Previous use of antiplatelet agents was not significantly different between the two groups, while the use of anticoagulants was less frequent in women (25.5% versus 31.9%, respectively, p = 0.026). Table 1. Baseline characteristics of patients. Variables Women Men p-Value Age (years) (mean ± SD) 75.1 ± 9.04 77.2 ± 9.73 <0.001 Type of AF, n (%): Paroxysmal 198 (35.4) 166 (35.6) 0.948 Permanent 248 (44.4) 224 (48.1) 0.232 Persistent 115 (20.2) 76 (16.3) 0.091 Concomitant medical history, n (%): Hypertension 457 (82.3) 364 (78.3) 0.113 Hyperlipidemia 161 (29.2) 171 (36.9) 0.011 Diabetes mellitus 144 (25.9) 120 (25.8) 1.0 History of stroke/TIA 138 (25.0) 127 (27.4) 0.39 Chronic heart failure 99 (17.7) 94 (20.2) 0.336 History of myocardial infarction 72 (13.0) 94 (20.2) 0.002 History of peripheral artery disease 36 (6.49) 56 (12.1) 0.003 Aortic atherosclerosis 44 (7.86) 58 (12.5) 0.016 Atherosclerosis in other sitesa 115 (21.3) 116 (25.2) 0.153 Pacemaker 36 (6.45) 49 (10.5) 0.023 Smoking 67 (12.1) 197(42.4) <0.001 Alcoholism 7 (1.25) 61 (13.1) <0.001 Previous use of oral anticoagulants, n (%): 141 (25.5) 148 (31.9) 0.026 Previous use of antiplatelets, n (%) 248 (44.8) 217 (46.9) 0.528 Previous use of statins, n (%) 115 (20.8) 145 (31.2) <0.001 Lesion site and size, n (%): Small lesion 187 (33.4) 192 (41.1) 0.011 Medium lesion 202 (36.1) 166 (35.5) 0.896 Large anterior lesion 137 (24.5) 83 (17.8) 0.012 Large posterior lesion 34 (6.1) 25 (5.4) 0.687 Leukoaraiosis, n (%) 261 (47.1) 165 (35.9) <0.001 NIHSS, mean ± SD 8.1 ± 7.5 9.2 ± 6.9 <0.001 Systolic AP, mean ± SD 148.4 ± 25.2 150.1 ± 25.3 0.048 Diastolic AP, mean ± SD 82.8 ± 14.3 82.9 ± 14.1 0.923 Laboratory data on admission, mean ± SD: Hemoglobin (g/dL), mean ± SD 14.3 ± 14.2 14.7 ± 12.3 0.65 Glycemia (mg/dL), mean ± SD 130.5 ± 39.8 223.5 ± 48.2 0.362 Total cholesterol (mg/dL), mean ± SD 171.8 ± 41.3 180.6 ± 42.8 <0.001 INR, mean ± SD 2.2 ± 3.3 3.5 ± 13.6 0.152 a Presence of internal carotid/vertebral artery stenosis ≥50%. AF: atrial fibrillation; AP: arterial pressure; INR: International Normalized Ratios at admission for all patients; TIA: transient ischemic attack.

tal cholesterol (mg/dL), mean ± SD 171.8 ± 41.3 180.6 ± 42.8 <0.001 INR, mean ± SD 2.2 ± 3.3 3.5 ± 13.6 0.152 a Presence of internal carotid/vertebral artery stenosis ≥50%. AF: atrial fibrillation; AP: arterial pressure; INR: International Normalized Ratios at admission for all patients; TIA: transient ischemic attack. Both large anterior lesions (p = 0.012) and leukoaraiosis (p < 0.001) were more frequent in women (Table 1). On transthoracic echocardiography, performed on 853 patients, mitral disease, aortic disease and severe atrial enlargement were more frequent in women (p = 0.025, p = 0.03 and p = 0.007, respectively), while cardiomyopathy was more frequent in men (p = 0.001). Tricuspid disease, and presence of aortic or mitral prostheses were not significantly different between the sexes (p = ns) (Table 2). Table 2. Findings on transthoracic echocardiography, performed in 843 patients. Variables Women Men p-value Atrial enlargement, n (%): 304 (68.5) 244 (62.7) 0.092 Mild 75 (28.3) 70 (26.8) 0.77 Moderate 113 (39.6) 90 (35.2) 0.287 Severe 116 (43.8) 81 (32.1) 0.007 Intracardiac thrombus, n (%) 4 (0.9) 7 (1.8) 0.383 Cardiomyopathy, n (%) 37 (8.31) 62 (15.9) 0.001 Mitral disease, n (%) 209 (46.8) 152 (39.0) 0.025 Aortic disease, n (%) 137 (30.6) 93 (23.9) 0.03 Tricuspid disease, n (%) 118 (26.3) 95 (24.5) 0.578 Biologic aortic prosthesis 4 (0.8) 6 (1.5) 0.526 Mechanical aortic prosthesis, n (%) 8 (1.7) 9 (2.2) 0.627 Biologic mitral prosthesis, n (%) 4 (0.8) 4 (0.7) 1.0 Mechanical mitral prosthesis, n (%) 15(3.2) 8 (2.0) 0.298

5 Aortic disease, n (%) 137 (30.6) 93 (23.9) 0.03 Tricuspid disease, n (%) 118 (26.3) 95 (24.5) 0.578 Biologic aortic prosthesis 4 (0.8) 6 (1.5) 0.526 Mechanical aortic prosthesis, n (%) 8 (1.7) 9 (2.2) 0.627 Biologic mitral prosthesis, n (%) 4 (0.8) 4 (0.7) 1.0 Mechanical mitral prosthesis, n (%) 15(3.2) 8 (2.0) 0.298 Types of revascularization therapy administered after ischemic stroke did not differ between the two groups. Anticoagulants were less often prescribed in women than in men after index stroke (71.3% versus 78.4%, respectively, p = 0.01). There were no sex differences regarding the time of initiating anticoagulant therapy between the two groups: (6.4 ± 11.7 days for men versus 6.5 ± 12.4 days for women, p = 0.902) (Table 3). A CHA2DS2-VASc score of 3 was found in 5.7% and 27.8% for women and men, respectively (p < 0.001), while a score between 7 and 9 was recorded more commonly in women (12.2% and 4.4% for women and men, respectively, p < 0.001). Men had more severe strokes than women on NIHSS (mean 9.2 ± 6.9 versus 8.1 ± 7.5, respectively, p < 0.001). Within 90 days, women had 46 (8.2%) recurrent ischemic events (stroke/TIA/systemic embolism) and 19 (3.4%) symptomatic cerebral bleedings compared to 30 (6.4%) and 18 (3.8%), respectively, in men (p = 0.28 and p = 0.74). At 90 days, 57.7% women were disabled or deceased compared to 41.1% of men (p < 0.001) (Table 3). In multivariate analysis, this significance was not confirmed (for unfavorable outcome – odds ratio (OR), 0.783, 95% confidence interval (CI), 0.536–1.143, p = 0.205 and for mortality – OR, 1.287, 95% CI, 0.726–2.284, p = 0.388) (Table 4). Table 3. Treatment of patients in the acute period of stroke and outcome effects at 90 days.

ate analysis, this significance was not confirmed (for unfavorable outcome – odds ratio (OR), 0.783, 95% confidence interval (CI), 0.536–1.143, p = 0.205 and for mortality – OR, 1.287, 95% CI, 0.726–2.284, p = 0.388) (Table 4). Table 3. Treatment of patients in the acute period of stroke and outcome effects at 90 days. Variables Women Men p-Value Revascularization therapy (IV and/or IA), n (%) 115 (20.6) 96 (20.6) 1.0 HT on neuroimaginga, n (%): 79 (14.1) 55 (11.9) 0.307 hemorrhagic infarction, n (%) 55 (11.0) 37 (9.1) 0.437 parenchymal hematoma, n (%) 24 (4.9) 18 (4.4) 0.874 Therapy with anticoagulants after index stroke, n (%) 399 (71.3) 367 (78.4) 0.01 Type of anticoagulation, n (%): LMWH 68 (12.1) 45 (9.6) 0.197 Oral anticoagulants (warfarin/DOA) 196 (35.0) 181 (38.7) 0.22 Bridging therapy (LMWH, followed by oral anticoagulants) 135 (24.1) 141 (30.1) 0.03 No anticoagulants at all 161 (28.8) 101 (21.6) 0.009 Time when anticoagulant therapy was initiated, mean ± SD, days 6.5 ± 12.4 6.4 ± 11.7 0.902 Outcome effects, n (%): Outcome ischemic events 46 (8.2) 30 (6.4) 0.284 Symptomatic HT 19 (3.4) 18 (3.8) 0.738 Mortality at 90 days 64 (11.6) 47 (10.1) 0.481 Unfavorable functional outcome (mRS 3-6) at 90 days 319 (57.7) 191 (41.1) <0.001 a Neuroimaging performed after 24–72 h from stroke onset. DOA: direct oral anticoagulants; HT: hemorrhagic transformation (either hemorrhagic infarction or parenchymal hematoma); IA: intra-arterial revascularization therapy; IV: intravenous revascularization therapy; LMWH: low molecular weight heparin.

91 (41.1) <0.001 a Neuroimaging performed after 24–72 h from stroke onset. DOA: direct oral anticoagulants; HT: hemorrhagic transformation (either hemorrhagic infarction or parenchymal hematoma); IA: intra-arterial revascularization therapy; IV: intravenous revascularization therapy; LMWH: low molecular weight heparin. Table 4. Multivariate logistic regression model for dichotomous outcomes.

91 (41.1) <0.001 a Neuroimaging performed after 24–72 h from stroke onset. DOA: direct oral anticoagulants; HT: hemorrhagic transformation (either hemorrhagic infarction or parenchymal hematoma); IA: intra-arterial revascularization therapy; IV: intravenous revascularization therapy; LMWH: low molecular weight heparin. Table 4. Multivariate logistic regression model for dichotomous outcomes. Variables Unfavorable outcome Mortality OR 95% CI p OR 95% CI P Age 1.039 1.018–1.062 <0.001 1.044 1.009–1.079 0.014 NIHSS on admission 1.178 1.136–1.222 <0.001 1.076 1.037–1.118 <0.001 Diabetes mellitus 1.133 0.734–1.747 0.574 1.277 0.695–2.345 0.431 Previous use of antiplatelets 0.999 0.669–1.492 0.996 0.907 0.496–1.661 0.752 Previous use of oral anticoagulants 1.139 0.708–1.834 0.591 1.484 0.720–3.058 0.285 Previous use of statins 0.629 0.360–1.099 0.104 1.080 0.475–2.456 0.854 Hypertension 0.830 0.513–1.343 0.448 0.874 0.421–1.814 0.717 Hyperlipidemia 0.890 0.537–1.476 0.653 0.814 0.376–1.763 0.602 Paroxysmal AF 0.848 0.524–1.374 0.503 1.037 0.501–2.147 0.922 History of stroke/TIA 1.734 1.148–2.620 0.009 0.741 0.405–1.357 0.332 Smoking 0.939 0.673–1.308 0.709 1.318 0.833–2.086 0.238 Alcoholism 1.528 0.720–3.243 0.269 1.928 0.690–5.387 0.211 Chronic heart failure 1.235 0.767–1.989 0.385 0.962 0.501–1.845 0.907 History of myocardial infarction 1.194 0.737–1.934 0.470 1.240 0.638–2.411 0.526 Pacemaker 1.048 0.562–1.955 0.883 2.130 0.959–4.733 0.063 Small ischemic lesion 0.408 0.195–0.853 0.017 0.090 0.027–0.295 <0.001 Leukoaraiosis 0.991 0.688–1.427 0.961 1.394 0.816–2.379 0.224 Therapy with anticoagulants after index stroke 0.389 0.249–0.605 <0.001 0.237 0.137–0.411 <0.001 AF: atrial fibrillation; AP: arterial pressure; INR: International Normalized Ratios at admission for all patients; TIA: transient ischemic attack.

–0.295 <0.001 Leukoaraiosis 0.991 0.688–1.427 0.961 1.394 0.816–2.379 0.224 Therapy with anticoagulants after index stroke 0.389 0.249–0.605 <0.001 0.237 0.137–0.411 <0.001 AF: atrial fibrillation; AP: arterial pressure; INR: International Normalized Ratios at admission for all patients; TIA: transient ischemic attack. Regarding sex differences, East European Stroke Units had 71 patients: 33 women (46.5%) and 38 men (53.5%)) and other Stroke Units in Europe (900 patients: 492 women (54.7%) and 408 men (45.3%)). Comparing women and men from Eastern Europe, the former were older (mean age 79.9 ± 8.3 versus 71.2 ±11.8, respectively, p < 0.001) and had more severe strokes than men (mean NIHSS 10.0 ± 7.2 versus 5.8 ± 4.96, respectively, p = 0.006) (Supplementary Table 1). No differences in the rates of prescribing anticoagulants between women and men in East Europe before and after index stroke were observed. While, after index stroke, LMWHs were prescribed more often to women than men (66.7% versus 31.6%, p = 0.004) and much more often than oral anticoagulants in East Europe. There were no observed sex differences regarding the time of initiating anticoagulant therapy in East Europe: (3.95 ± 15.9 days for men versus 3.67 ± 7.5 days for women, p = 0.923). (Supplementary Table 2).

en to women than men (66.7% versus 31.6%, p = 0.004) and much more often than oral anticoagulants in East Europe. There were no observed sex differences regarding the time of initiating anticoagulant therapy in East Europe: (3.95 ± 15.9 days for men versus 3.67 ± 7.5 days for women, p = 0.923). (Supplementary Table 2). Discussion This study found that women overall in Europe with AF were less likely to receive oral anticoagulants prior to and after stroke compared to men and regardless of the fact that women were significantly younger and with less severe stroke at onset; outcomes did not differ between the sexes. The RAF-study included more women (54.5%) than men, who were on average younger and had less severe stroke at onset compared to previous studies.16,17 Moreover, while previous studies had included patients with all types of stroke and with and without AF,16–18 the RAF-study included only ischemic stroke patients with AF. Furthermore, women more frequently had mitral disease and severe atrial enlargement. The latter fact has been reported by Gómez-Doblas, who has stating that women have more rheumatic aetiologies, while men tend to be more affected by ischemic or congenital aetiologies.19 Rheumatic aetiologies have been correlated with a higher embolic risk of AF and an earlier onset of stroke.20

ere atrial enlargement. The latter fact has been reported by Gómez-Doblas, who has stating that women have more rheumatic aetiologies, while men tend to be more affected by ischemic or congenital aetiologies.19 Rheumatic aetiologies have been correlated with a higher embolic risk of AF and an earlier onset of stroke.20 In this study, women were less likely to receive oral anticoagulants before the index stroke, a finding in line with past population studies.8,9 Likewise, women were less likely to receive anticoagulation therapy for secondary prevention. This clear disparity in treatment delivery has also been documented by the Austrian Stroke Unit Registry.18 This under-treatment for women has been hypothesized as being due to a lack of social support, as well as other concomitant diseases afflicting these patients, including cognitive decline, a higher burden of vascular brain disease, epilepsy and an increased risk of falls.18 Furthermore, we also observed that women had a lower rate of statin use, despite an equal rate of atherosclerosis between the sexes.18,21 Even though men were older and had more severe stroke, the mortality and disability rates between the sexes were similar. The selection of patient cohorts could have influenced this, as only cardioembolic strokes were included, which could also explain the younger age of women at stroke onset. Study limitations This hospital-based clinical study was not randomized but based upon consecutively admitted patients fulfilling inclusion criteria.

Even though men were older and had more severe stroke, the mortality and disability rates between the sexes were similar. The selection of patient cohorts could have influenced this, as only cardioembolic strokes were included, which could also explain the younger age of women at stroke onset. Study limitations This hospital-based clinical study was not randomized but based upon consecutively admitted patients fulfilling inclusion criteria. Conclusions The RAF-study observed that women regardless of lower NIHSS-score at admission and younger age, experienced the same outcomes as men with higher NIHSS at admission and older age. Moreover, this study also observed that women were less likely, compared to men, to have been prescribed anticoagulants before and after stroke.

RAF-study observed that women regardless of lower NIHSS-score at admission and younger age, experienced the same outcomes as men with higher NIHSS at admission and older age. Moreover, this study also observed that women were less likely, compared to men, to have been prescribed anticoagulants before and after stroke. Supplementary Material Supplementary material Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: M Paciaroni has received honoraria as a member of the speaker bureaus of Sanofi-Aventis, Boehringer Ingelheim, Bayer and Pfizer. G Agnelli has received honoraria as a member of the speaker bureaus of Boehringer Ingelheim and Bayer. C Becattini has received honoraria as a member of the speaker bureaus of Bristol Meyer Squibb and Bayer P Michel has received a Research Grant from the Swiss National Science Foundation and Swiss Heart Foundation; he has also received speaker fees from Bayer, Boehringer Ingelheim, Covidien, St. Jude Medical as well as received honoraria for an advisory relationship from Pierre-Fabre, Bayer, Bristol Meyer Squibb, Amgen, and Boehringer Ingelheim. J Putaala has received honoraria for lectures related to atrial fibrillation and anticoagulants for Orion Pharma, Bristol Meyer Squibb, Pfizer, Bayer, and Boehringer Ingelheim. T Tatlisumak received honoraria as consultant or advisory relationship by Lundbeck and Boehringer Ingelheim. G Tsivgoulis had research support by European Regional Development Fund, Project St. Anne's University Hospital, Brno, International Clinical Research Center (FNUSA-ICRC) (No. CZ.1.05/1.1.00/02.0123). D Toni has received honoraria as a member of speaker bureaus and as advisory board member for Boehringer Ingelheim and Bayer. The other authors report no conflicts.

Regional Development Fund, Project St. Anne's University Hospital, Brno, International Clinical Research Center (FNUSA-ICRC) (No. CZ.1.05/1.1.00/02.0123). D Toni has received honoraria as a member of speaker bureaus and as advisory board member for Boehringer Ingelheim and Bayer. The other authors report no conflicts. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Ethical approval The study was approved by the local hospital board. Informed consent not necessary as the study report on clinical practice. Guarantor KA. Contributorship All authors contributed equally.

Introduction Randomised controlled trials (RCTs) are the fairest tests of treatment, but judgements about their value are dependent on transparent and complete reporting. Inadequate reporting prevents a complete assessment of an RCT’s risk of bias and description of their results, which can also preclude the re-use of data in meta-analyses.1,2 In order to combat this, the Consolidated Standards Of Reporting Trials (CONSORT) Statement was developed in 1996,3 updated in 20014 and revised in 20105 to improve the reporting of RCTs. The CONSORT guidelines seem to have been endorsed by over 600 journals.6 In general, studies of RCT reporting have not only demonstrated incomplete reporting in numerous specialties but also modest improvements over time that were often associated with journal endorsement and uptake of CONSORT.7–20 Concerns about completeness of reporting remain, particularly in journals with low impact factors.7,21 Despite advances in the prevention and treatment of stroke supported by RCTs, stroke remains the leading cause of disability and second leading cause of death worldwide and stroke burden is projected to increase with changes in lifestyle and longevity.22,23 The limited funding available for stroke research to lessen this burden should not be wasted.24,25 But concerns remain about waste in stroke research, including the poor reporting of research.26,27

leading cause of death worldwide and stroke burden is projected to increase with changes in lifestyle and longevity.22,23 The limited funding available for stroke research to lessen this burden should not be wasted.24,25 But concerns remain about waste in stroke research, including the poor reporting of research.26,27 Other than an investigation of the reporting of a specific intervention for stroke rehabilitation,28 the last systematic assessment of stroke RCT reporting pre-dated CONSORT.9 That review found that the standard of reporting was poor, but it improved over time alongside an increase in RCT sample size. Reporting did not appear to be associated with journal impact factor but trials with a positive outcome tended to be less well reported than those with neutral or negative outcomes.9 This differs from the findings in other specialties.14,15 However, it is unclear whether stroke RCT reporting has improved since CONSORT guidelines were released and updated, what factors are associated with better reporting, and whether these associations differ from other diseases.27 Therefore, we aimed to assess: the extent to which the published reports of the final primary results of RCTs involving participants with transient ischaemic attack (TIA) or stroke have adhered to the CONSORT reporting guidelines between 1997 and 2016; whether adherence has changed over time, in particular following each revision of the CONSORT guidelines; and which factors are associated with better reporting.

results of RCTs involving participants with transient ischaemic attack (TIA) or stroke have adhered to the CONSORT reporting guidelines between 1997 and 2016; whether adherence has changed over time, in particular following each revision of the CONSORT guidelines; and which factors are associated with better reporting. Methods Protocol and registration All authors developed and approved the protocol, which we registered with PROSPERO before embarking on data collection (CRD42017072193). Eligibility criteria We included published reports of the final primary results of RCTs including patients with TIA or stroke, published in 1997–2016 inclusive. We applied eligibility criteria designed to obtain a representative sample of RCTs (Table 1). Table 1. Eligibility criteria for included studies. Inclusion criteria Exclusion criteria Published report of the final primary results of an RCT, published in 1997–2016 inclusive English language publication Participants included after TIA or any type of stroke Any type of therapeutic intervention (drug, surgery, device, rehabilitation, etc.) Reports of interim analyses that preceded the report of the final primary results of an RCT Reports of secondary or long-term follow-up analyses of an RCT Duplicate reports of the final primary results of an RCT Cost-effectiveness and economic studies of an RCT RCTs in which stroke was not the qualifying condition, but the outcome RCT: randomised controlled trial.

report of the final primary results of an RCT Reports of secondary or long-term follow-up analyses of an RCT Duplicate reports of the final primary results of an RCT Cost-effectiveness and economic studies of an RCT RCTs in which stroke was not the qualifying condition, but the outcome RCT: randomised controlled trial. Information sources On 30 August 2017, the Managing Editor of the Cochrane Stroke Group searched the Cochrane Stroke Group’s Trial Register for all publications of RCTs including patients with TIA or any type of stroke published in 1997–2016 inclusive. Study selection We sub-divided the results of the search into three epochs of publication (1997–2000, 2001–2009 or 2010–2016), each corresponding to the timing of a published revision of the CONSORT guidelines.3,5,29 We used a random number generator in Microsoft Excel to take a random sample of equal size from each of these three groups of RCTs, removed duplicate records of the same RCT in order to include only the report of the final primary results of each RCT, and took further random samples as required to achieve our target sample size of 180 RCTs. We chose this sample size so that it would adequately power a multiple regression analysis including 10 covariates based on the likely distribution of CONSORT reporting scores and be feasible for two reviewers to assess in the time available to the research team.

required to achieve our target sample size of 180 RCTs. We chose this sample size so that it would adequately power a multiple regression analysis including 10 covariates based on the likely distribution of CONSORT reporting scores and be feasible for two reviewers to assess in the time available to the research team. Data collection We imported the results of the search into Covidence (www.covidence.org). One reviewer (BW) screened the titles and abstracts of all RCTs to exclude any ineligible RCTs. Two reviewers (PB and BW) reviewed the full text of potentially eligible RCTs, independent of each other. PB and BW used the data collection tool within Covidence to independently assess the completeness of reporting of included RCTs. Any protocols that were referenced within the included papers were checked and included in the scoring. PB and BW used Microsoft Excel to extract information from each RCT on pre-specified covariates that we hypothesised might be associated with completeness of reporting. Any uncertainties or disagreements about eligibility, completeness of reporting, or covariates were resolved by discussion with another reviewer (RA-SS).

ffects in each group, or statement of no adverse effects Registration Registration number and the trial registry Protocol Where the trial protocol can be accessed, if available Funding Sources of funding and other support (such as supply of drugs) and role of funders CONSORT: Consolidated Standards Of Reporting Trials. In our protocol, we specified the following covariates to investigate associations with completeness of reporting based on prior evidence of their association with completeness of reporting in other diseases: (1) CONSORT endorsement by the journal preceding the publication of the RCT (we established this by searching the CONSORT database online, contacting the journal or searching the journal’s archived guidelines);11–13,20 (2) year of publication;7–10 (3) sample size of the RCT;8,9,14,21 (4) number of recruiting sites (single vs. multicentre);8,9,14,21 (5) direction and statistical significance of results with reference to aims/hypothesis (positive, neutral or negative);9,14,15 (6) type of intervention (drug, surgical or other);15,21 (7) funding source (academic/governmental/charitable vs. commercial vs. other)14,21 and (8) journal impact factor.7,8,21 However, we did not use journal impact factor because it is widely acknowledged to be a flawed metric as it is the arithmetic mean of a highly skewed distribution of citations and it is quoted to a higher level of precision (three decimal places) than is warranted by the underlying data,31 and hence we used a ‘modified journal impact factor’ that uses the median – rather than the mean – number of citations.32 We also pre-specified (9) TIA/stroke type and (10) intervention type (acute, prevention, rehabilitation), which we hypothesised might influence the completeness of reporting of stroke RCTs.

data,31 and hence we used a ‘modified journal impact factor’ that uses the median – rather than the mean – number of citations.32 We also pre-specified (9) TIA/stroke type and (10) intervention type (acute, prevention, rehabilitation), which we hypothesised might influence the completeness of reporting of stroke RCTs. Risk of bias We reduced bias in our assessments of completeness of reporting and covariates of interest by two reviewers assessing them independently. We looked for evidence of differences between reviewers, which might be systematic, by calculating the kappa statistic to assess the inter-reviewer variability for each CONSORT checklist item (Table 2). Summary measure We used the truncated CONSORT score, which was the sum of the score of each of the 10 fields in the truncated list (Table 2) as our summary measure.

Risk of bias We reduced bias in our assessments of completeness of reporting and covariates of interest by two reviewers assessing them independently. We looked for evidence of differences between reviewers, which might be systematic, by calculating the kappa statistic to assess the inter-reviewer variability for each CONSORT checklist item (Table 2). Summary measure We used the truncated CONSORT score, which was the sum of the score of each of the 10 fields in the truncated list (Table 2) as our summary measure. Statistical analysis We assessed trends in trial characteristics with time using the Cochran–Armitage test for trend. We quantified the mean and SD of the truncated CONSORT score. We assessed associations of categorical covariates with truncated CONSORT score using ANOVA and univariable associations with continuous variables using Spearman’s rank-order correlation. We assessed trends in reporting of individual CONSORT items with time using the Cochran–Armitage test for trend. We entered all 10 covariates into a single multiple linear regression model to investigate the association of each of our covariates with total truncated CONSORT score, after checking linearity of relationships, multivariable normality, multi-collinearity, the absence of auto-correlation and homoscedasticity. We used IBM SPSS 24, with α = 0.05. Results Study selection From 7813 studies in the Cochrane Stroke Group Trial Register published in 1997–2016 inclusive, we randomly sampled 180 that appeared eligible and included 177 (Figure 1). Figure 1. PRISMA flow chart.

Statistical analysis We assessed trends in trial characteristics with time using the Cochran–Armitage test for trend. We quantified the mean and SD of the truncated CONSORT score. We assessed associations of categorical covariates with truncated CONSORT score using ANOVA and univariable associations with continuous variables using Spearman’s rank-order correlation. We assessed trends in reporting of individual CONSORT items with time using the Cochran–Armitage test for trend. We entered all 10 covariates into a single multiple linear regression model to investigate the association of each of our covariates with total truncated CONSORT score, after checking linearity of relationships, multivariable normality, multi-collinearity, the absence of auto-correlation and homoscedasticity. We used IBM SPSS 24, with α = 0.05. Results Study selection From 7813 studies in the Cochrane Stroke Group Trial Register published in 1997–2016 inclusive, we randomly sampled 180 that appeared eligible and included 177 (Figure 1). Figure 1. PRISMA flow chart. Characteristics of the included RCTs The RCTs included a variety of combinations of TIA or stroke sub-types, the most frequent being ischaemic stroke alone (Table 3). Three-quarters of RCTs evaluated acute interventions and almost two-thirds of the interventions were drugs. Sample sizes ranged from 8 to 21,106 patients, median 99 (inter-quartile range 41–367). Roughly half of included RCTs reported statistically significant beneficial (i.e. ‘positive’) effects on their primary outcomes. Roughly one-fifth of included RCTs received commercial funding. More than one-third of the journals had not explicitly endorsed the CONSORT statement to require complete RCT reporting to the CONSORT standard. There was a statistically significant downward trend in the proportion of acute trials (p < 0.001), those investigating a pharmacological intervention (p = 0.002) and journals endorsing CONSORT (p = 0.014) over time (the latter due to an increase in the number of open access journals in recent times).

CONSORT standard. There was a statistically significant downward trend in the proportion of acute trials (p < 0.001), those investigating a pharmacological intervention (p = 0.002) and journals endorsing CONSORT (p = 0.014) over time (the latter due to an increase in the number of open access journals in recent times). Table 3. Descriptive characteristics of included RCTs.

CONSORT standard. There was a statistically significant downward trend in the proportion of acute trials (p < 0.001), those investigating a pharmacological intervention (p = 0.002) and journals endorsing CONSORT (p = 0.014) over time (the latter due to an increase in the number of open access journals in recent times). Table 3. Descriptive characteristics of included RCTs. All 177 RCTs (%) 1997–2000,n = 59 (%) 2001–2009,n = 59 (%) 2010–2016,n = 59 (%) Type of TIA/stroke included Any type 3 (2) 3 (5) 0 (0) 0 (0) Intracerebral haemorrhage 20 (11) 5 (8) 6 (10) 7 (12) Ischaemic stroke 84 (48) 30 (51) 29 (49) 25 (42) Sub-arachnoid haemorrhage 46 (26) 16 (27) 15 (25) 15 (25) TIA 6 (3) 0 (0) 1 (2) 6 (10) TIA or ischaemic stroke 12 (7) 3 (5) 6 (10) 3 (5) Unknown 5 (3) 2 (3) 1 (2) 2 (3) Type of RCT Acute 131 (74) 53 (90) 45 (76) 35 (59) Prevention 10 (6) 0 (0) 2 (3) 8 (14) Rehabilitation 32 (18) 6 (10) 10 (17) 16 (27) Other 4 (2) 0 (0) 2 (3) 0 (0) Type of intervention Drug 112 (63) 47 (80) 34 (58) 31 (53) Surgical 44 (25) 6 (10) 10 (17) 5 (8) Other 21 (12) 6 (10) 15 (25) 23 (39) Number of recruiting sites Multicentre 87 (49) 32 (54) 26 (44) 29 (49) Single centre 82 (46) 25 (42) 30 (51) 27 (46) Unknown 8 (5) 2 (3) 3 (5) 3 (51) Sample sizea Median (IQR) 99 (41–367) 142 (32–407) 90 (40–365) 94 (50–233) RCT outcome Positive (p < 0.05) 92 (52) 26 (44) 28 (47) 27 (46) Neutral 78 (44) 30 (51) 27 (46) 30 (51) Negative 7 (4) 2 (3) 4 (7) 2 (3) Funding source Not specified 66 (37) 24 (41) 26 (44) 16 (27) Commercial 35 (20) 16 (27) 11(19) 8 (14) Other 76 (43) 19 (32) 22 (37) 35 (59) Journal endorsed CONSORT Endorsed 108 (61) 43 (73) 35 (59) 30 (51) Impact factora Median (IQR) 4.2 (2.0–6.0) 4.8 (1.4–6.0) 5.2 (2.1–5.9) 3.0 (2.0–6.2) Modified impact factora,b Median (IQR) 2 (1–5) 2 (1–5) 3 (1–5) 2 (1–5) TIA: transient ischaemic attack; IQR: inter-quartile range; RCT: randomised controlled trial; CONSORT: Consolidated Standards Of Reporting Trials.

3) 35 (59) 30 (51) Impact factora Median (IQR) 4.2 (2.0–6.0) 4.8 (1.4–6.0) 5.2 (2.1–5.9) 3.0 (2.0–6.2) Modified impact factora,b Median (IQR) 2 (1–5) 2 (1–5) 3 (1–5) 2 (1–5) TIA: transient ischaemic attack; IQR: inter-quartile range; RCT: randomised controlled trial; CONSORT: Consolidated Standards Of Reporting Trials. aItems are reported as frequency (proportion) for categorical variables, unless otherwise specified for continuous variables. bModified impact factor is the impact factor for the journal at the time of publication calculated using the median rather than mean of the citation distribution. Inter-reviewer agreement For all 10 items in the truncated CONSORT checklist, inter-reviewer agreement was high, ranging from κ = 0.96 to 1.00 for individual items (online appendix). Completeness of reporting In all 177 RCTs, the mean total truncated CONSORT score was 5.8 (SD 2.2) out of 10, (ranging from 1 to 10 in individual RCTs). Completeness of reporting of each of the CONSORT items varied considerably (Figure 2). Explicit definitions of the primary outcome measure and the estimated treatment effect size and its precision were most frequently reported and did not decrease over time. Details of trial registration and the availability of the protocol were least frequently reported, but like many other items (other than harms and funding) there was a statistically significant trend of increasing completeness of reporting individual items over time. Figure 2. Reporting of individual truncated CONSORT checklist items.

Completeness of reporting In all 177 RCTs, the mean total truncated CONSORT score was 5.8 (SD 2.2) out of 10, (ranging from 1 to 10 in individual RCTs). Completeness of reporting of each of the CONSORT items varied considerably (Figure 2). Explicit definitions of the primary outcome measure and the estimated treatment effect size and its precision were most frequently reported and did not decrease over time. Details of trial registration and the availability of the protocol were least frequently reported, but like many other items (other than harms and funding) there was a statistically significant trend of increasing completeness of reporting individual items over time. Figure 2. Reporting of individual truncated CONSORT checklist items. From left to right: p=0.042, 0.009, 0.005, <0.001, <0.001, <0.001 and 0.040. *Cochran–Armitage test shows a trend of improvement with time.

Completeness of reporting In all 177 RCTs, the mean total truncated CONSORT score was 5.8 (SD 2.2) out of 10, (ranging from 1 to 10 in individual RCTs). Completeness of reporting of each of the CONSORT items varied considerably (Figure 2). Explicit definitions of the primary outcome measure and the estimated treatment effect size and its precision were most frequently reported and did not decrease over time. Details of trial registration and the availability of the protocol were least frequently reported, but like many other items (other than harms and funding) there was a statistically significant trend of increasing completeness of reporting individual items over time. Figure 2. Reporting of individual truncated CONSORT checklist items. From left to right: p=0.042, 0.009, 0.005, <0.001, <0.001, <0.001 and 0.040. *Cochran–Armitage test shows a trend of improvement with time. Associations with completeness of reporting In univariable analyses, there was a significant improvement in total CONSORT score by 1.9 (95% Confidence Interval (CI) 1.0–2.8) from 1997–2000 until 2010–2016 (p < 0.001) and by 1.1 (95% CI 0.2–2.0) from 2001–2009 until 2010–2016 (p = 0.013) (Table 4 and Figure 3). Journal endorsement of CONSORT at the time of an RCT’s publication, higher modified journal impact factor, having a commercial funding source, multicentre recruitment and larger sample size were all associated with higher total CONSORT reporting scores (Table 4). In multivariable analysis, publication during epochs following a revision of CONSORT reporting guidelines was independently associated with higher completeness of reporting (Table 5), as was journal endorsement of the CONSORT reporting guideline at the time of RCT publication, and journal modified impact factor.

le 4). In multivariable analysis, publication during epochs following a revision of CONSORT reporting guidelines was independently associated with higher completeness of reporting (Table 5), as was journal endorsement of the CONSORT reporting guideline at the time of RCT publication, and journal modified impact factor. Table 4. Univariable analyses of associations with truncated CONSORT score. Categorical covariates Number of RCTs Mean total CONSORT score (standard deviation) p Year of publication <0.001 1997–2000 59 4.9 (2.0) 2001–2009 59 5.8 (2.1) 2010–2016 59 6.8 (2.1) TIA/stroke typea 0.28 Haemorrhagic (ICH or SAH) 66 6.1 (2.1) Ischaemic (all other groups) 111 5.7 (2.3) Trial type 0.11 Acute 46 5.4 (2.3) Other 131 6.0 (2.1) Type of intervention 0.30 Drug 112 6.0 (2.2) Other 65 5.6 (2.3) Number of recruiting sites <0.001 Multicentre 87 6.6 (2.0) Single centre/not specified 90 5.1 (2.1) Outcome estimate 0.43 Positive 92 5.7 (2.3) Negative or neutral 85 6.0 (2.1) Funding source 0.022 Purely commercial 35 6.6 (1. 7) Other/not specified 142 5.7 (2.3) Journal endorsed CONSORT <0.001 Endorsed 108 6.6 (2.0) Not endorsed 69 4.7 (2.0) Continuous covariates Spearman rank-order coefficient p Modified impact factor rs = 0.51 <0.001 Sample size rs = 0.38 <0.001 RCTs: randomised controlled trials; CONSORT: Consolidated Standards Of Reporting Trials. aHaemorrhagic sub-category included intracerebral haemorrhage and subarachnoid haemorrhage, ischaemic encompassed all other sub-categories as this was the most common sub-type within them. Figure 3. Distribution of CONOSRT Scores by Epoch.

p Modified impact factor rs = 0.51 <0.001 Sample size rs = 0.38 <0.001 RCTs: randomised controlled trials; CONSORT: Consolidated Standards Of Reporting Trials. aHaemorrhagic sub-category included intracerebral haemorrhage and subarachnoid haemorrhage, ischaemic encompassed all other sub-categories as this was the most common sub-type within them. Figure 3. Distribution of CONOSRT Scores by Epoch. Table 5. Multivariable linear regression analysis of associations with truncated CONSORT score.

aHaemorrhagic sub-category included intracerebral haemorrhage and subarachnoid haemorrhage, ischaemic encompassed all other sub-categories as this was the most common sub-type within them. Figure 3. Distribution of CONOSRT Scores by Epoch. Table 5. Multivariable linear regression analysis of associations with truncated CONSORT score. Mean CONSORT score (SD) Multiple linear regression RCTs (n) β Coefficient 95% CI p Year of publication 1997–2000 59 4.9 (2.0) Ref 2001–2009 59 5.8 (2.1) 1.071 0.435 to 1.709 0.001 2010–2016 59 6.8 (2.1) 2.248 1.559 to 2.937 <0.001 TIA/stroke typea Haemorrhagic 66 6.1 (2.1) Ref Ischaemic 111 5.7 (2.3) −0.204 −0.794 to 0.387 0.497 Trial type Other 131 Ref Acute 46 0.118 −0.626 to 0.863 0.754 Type of intervention Other 65 5.4 (2.3) Ref Drug 112 6.0 (2.1) 0.098 −0.505 to 0.702 0.748 Number of recruiting sites Other 90 6.6 (2.0) Ref Multicentre 87 5.1 (2.1) 0.672 0.132 to 1.211 0.015 Sample size Per n=100 increase 177 0.009 −0.001 to 0.019 0.066 Outcome estimate Negative or neutral 85 5.7 (2.3) Ref Positive 92 6.0 (2.1) −0.216 −0.722 to 0.290 0.400 Funding source Other 142 6.6 (1. 7) Ref Purely commercial 35 5.7 (2.3) 0.543 −0.127 to 1.212 0.112 Journal endorsed CONSORT Not endorsed 69 6.6 (2.0) Ref Endorsed 108 4.7 (2.0) 1.382 0.726 to 2.038 <0.001 Modified impact factor For each unit increase 177 0.127 0.028 to 0.226 0.012 RCTs: randomised controlled trials; CONSORT: Consolidated Standards Of Reporting Trials.

mercial 35 5.7 (2.3) 0.543 −0.127 to 1.212 0.112 Journal endorsed CONSORT Not endorsed 69 6.6 (2.0) Ref Endorsed 108 4.7 (2.0) 1.382 0.726 to 2.038 <0.001 Modified impact factor For each unit increase 177 0.127 0.028 to 0.226 0.012 RCTs: randomised controlled trials; CONSORT: Consolidated Standards Of Reporting Trials. Multiple linear regression model adjusted for year of publication, TIA/stroke type, trial type, type of intervention, number of recruiting sites, sample size, outcome estimate, funding source, CONSORT endorsement and modified impact factor. ‘Ref' indicates which categories were used as reference categories in the multiple linear regression. Discussion In this systematic review of 177 stroke RCTs published over a 20-year period, we found that stroke RCTs on average have reported ∼6 out of 10 items on a truncated CONSORT checklist (Figure 2). Encouragingly, the overall completeness of reporting has increased with time such that stroke RCTs published after the 2010 revision of CONSORT reported ∼7 out of 10 items (Table 4). This improvement may reflect, at least in part, the awareness or endorsement of CONSORT guidelines by the International Committee of Medical Journal Editors in 2001 and individual journals since then.

sed with time such that stroke RCTs published after the 2010 revision of CONSORT reported ∼7 out of 10 items (Table 4). This improvement may reflect, at least in part, the awareness or endorsement of CONSORT guidelines by the International Committee of Medical Journal Editors in 2001 and individual journals since then. The reporting of primary outcomes and estimates of treatment effect on these outcomes have been consistently good, while the least well-reported items were the method of allocation concealment, trial registration and protocol location (Figure 2), perhaps because repositories for the latter two were not available in the earlier epochs in this study.9,28 This is similar to findings for trials in other diseases,8,11,16,21,28 although the proportion of stroke RCTs adequately reporting registration and the location of protocol is particularly low. Each item that was poorly reported in 1997–2000, excluding harms and funding, has shown an improvement in completeness of reporting with time. The reporting of harms appears to have decreased over time, although this might be confounded by the decrease in the proportion of stroke RCTs investigating drug or surgical interventions over time (Table 3), in which the reporting of harms is more of a requirement than for other interventions (e.g. rehabilitation).

e. The reporting of harms appears to have decreased over time, although this might be confounded by the decrease in the proportion of stroke RCTs investigating drug or surgical interventions over time (Table 3), in which the reporting of harms is more of a requirement than for other interventions (e.g. rehabilitation). We did not find that the type of intervention was associated with completeness of reporting in stroke RCTs in contrast to previous studies in other sub-specialties.15 The development and implementation of extensions to the CONSORT statement, addressing the weaknesses in reporting of non-drug interventions, may be responsible for this difference.33 Similar to the results of others, this study has shown that the time period of publication,7–10,14,15 journal endorsement of CONSORT at the time of publication,11–13,20 and multicentre recruitment21 are all independently associated with a higher completeness of reporting in stroke RCTs. Others have also identified commercial funding as being associated with better completeness of reporting.4,34,35 Our study used a modified journal impact factor, which was associated with a higher total CONSORT checklist score. This finding is in agreement with those of recent studies,7,8,21 but in contrast to the earlier stroke RCT study.9 We speculate that this could be explained by the increased scrutiny and stricter peer-review processes implemented by higher impact journals in recent years, influenced by the drive to improve completeness of reporting nowadays.2 Our study differs from the results of the only previous study of the completeness of reporting of stroke RCTs, which found that completeness of reporting was associated with estimates of treatment effect.9

by higher impact journals in recent years, influenced by the drive to improve completeness of reporting nowadays.2 Our study differs from the results of the only previous study of the completeness of reporting of stroke RCTs, which found that completeness of reporting was associated with estimates of treatment effect.9 As far as we are aware this is the only study to assess completeness of reporting and the associated factors for stroke RCTs in the 21st century. However, it is not without its weaknesses. We scored RCTs using a modified, truncated CONSORT checklist, to give a score out of 10 as a measure of completeness of reporting of key items. Each of the factors was weighted equally, but these factors vary in their importance; however, any attempt to implement a weighted system to this list would be arbitrary and introduce a degree of subjectivity which would limit the generalisability of our results. An advantage of this binary scoring system was the high inter-reviewer agreement in the assessment of reporting. Other scoring systems are subject to lower and more variable kappa values, e.g. ranging from 0.02 to 0.92.8 Lastly, some of our findings may be confounded by factors relating to publication culture: we found that lower impact factor journals exhibited poorer reporting of RCTs, possibly reflecting the fact that higher quality stroke RCTs may be first submitted to higher impact factor journals.

e.g. ranging from 0.02 to 0.92.8 Lastly, some of our findings may be confounded by factors relating to publication culture: we found that lower impact factor journals exhibited poorer reporting of RCTs, possibly reflecting the fact that higher quality stroke RCTs may be first submitted to higher impact factor journals. In summary, the standard of reporting of stroke RCTs has improved with time, but there is room for improvement, particularly in lower impact factor journals. This study provides evidence for areas which can be improved. Authors of stroke RCTs should focus on better reporting of the method of allocation concealment, trial registration and protocol availability. The independent associations that we found between journal-level covariates and completeness of stroke RCT reporting suggest that journals may be best placed to improve reporting completeness by endorsement and enforcement of the CONSORT checklist. One study, but not all, shows that making adherence to CONSORT guidelines mandatory improves completeness of reporting.36 We therefore suggest that journals require the submission of a completed CONSORT checklist with stroke RCT manuscripts, and that this becomes an integrated part of the peer-review assessment. This may help make reporting standards uniform across journals, and therefore rectify the disparity between journals of high- and low-impact factors. Continuous monitoring of reporting completeness2 and other sources of research waste27 will be necessary and can be done by researchers in collaboration with the REWARD Alliance (http://rewardalliance.net).

ds uniform across journals, and therefore rectify the disparity between journals of high- and low-impact factors. Continuous monitoring of reporting completeness2 and other sources of research waste27 will be necessary and can be done by researchers in collaboration with the REWARD Alliance (http://rewardalliance.net). Supplemental Material Supplemental material for Completeness of reporting of randomised controlled trials including people with transient ischaemic attack or stroke: A systematic review Click here for additional data file. Supplemental material for Completeness of reporting of randomised controlled trials including people with transient ischaemic attack or stroke: A systematic review by Blair Wilson, Peter Burnett, David Moher, Douglas G Altman and Rustam Al-Shahi Salman in European Stroke Journal Acknowledgements We dedicate this article to Doug Altman (12 July 1948 – 3 June 2018), for inspiring us to improve the reliability and reporting of randomised trials. We are very grateful to Hazel Fraser, the Managing Editor of the Cochrane Stroke Group, and all the members who facilitated the database search and helped us obtain full manuscripts of selected articles. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: DM and DGA helped to develop the CONSORT Statement; they are members of the CONSORT executive. DM and RA-SS are also members of the REWARD alliance.

of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: DM and DGA helped to develop the CONSORT Statement; they are members of the CONSORT executive. DM and RA-SS are also members of the REWARD alliance. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Informed consent Not applicable. Ethical approval Not applicable. Guarantor RAAS. Contributorship All authors were involved with the planning and design of the study. BW and PB carried out the study. BW wrote the manuscript which was edited and revised by all authors.

ore to after the stroke was used to estimate change in disability caused by the stroke and attribute changes in health and social care utilisation to the stroke itself rather than pre-existing or co-morbid illnesses. Health-Related Quality of Life (health utility on a 0–1 scale) was estimated by mapping from the mRS.12 Probabilities and distributions were largely estimated using SSNAP data, with additional data on long-term outcomes, stroke recurrence and resource use obtained from the South London Stroke Register (SLSR). SSNAP data included in the model included all patients aged 40–100 years admitted for acute stroke from April 2013 to 2015 (n = 111,846). The SLSR is a population-based register with prospective long-term follow-up of all adults with first ever stroke in South London,13 including data on 6000 patients. The SLSR was used to provide data to model survival after stroke, stroke recurrence and to estimate long-term health and social care utilisation after stroke. Resource use was estimated based on the parameters in Table 1, which were combined with unit costs (online Appendix) to estimate the costs generated by individuals in the model.

Session 1: Management of cervical artery dissection (CAD) Chair: T. Tatlisumak (Gothenburg), Secretary: E. Lundström (Stockholm), Speakers: S. Debette (Bordeaux); H. Markus (Cambridge), Contributors: S. T. Engelter (Basel), M. Arnold (Bern) What is the best method to diagnose CAD? Contrast enhanced magnetic resonance imaging (MRI) angiography (MRA) and MRI with T1-fat suppression sequences is the recommended imaging modality to diagnose extra- and intracranial CAD. When not available computed tomography (CT) and CT angiography (CTA) might be alternatives grade C. Acute stroke in the setting of CAD: Is thrombolysis safe? Acute ischaemic stroke (AIS) patients with suspected or confirmed extracranial CAD should not be excluded from intravenous or intra-arterial thrombolysis or mechanical thrombectomy (grade C). Should we use anticoagulants or antiplatelet drugs to prevent CAD? For extracranial CAD: Antithrombotic treatment is strongly recommended (Grade C). There is no evidence of any difference between antiplatelets and anticoagulants (heparin followed by warfarin) (Grade B). For intracranial dissection in the absence of SAH, antiplatelet drugs are recommended (Grade C). Is there a role for angioplasty and stenting? Angioplasty and stenting may be considered in CAD patients with recurrent ischaemic symptoms despite antithrombotic treatment (Grade C). What is the optimal duration of medical treatment?

For intracranial dissection in the absence of SAH, antiplatelet drugs are recommended (Grade C). Is there a role for angioplasty and stenting? Angioplasty and stenting may be considered in CAD patients with recurrent ischaemic symptoms despite antithrombotic treatment (Grade C). What is the optimal duration of medical treatment? Antithrombotic treatment is recommended for at least 6–12 months. In patients in whom full recanalisation of the dissected artery has occurred and there have been no recurrent symptoms stopping antithrombotic treatment may be considered. In case of a residual dissecting aneurysm or stenosis, long-term antiplatelet treatment is recommended (Grade C). Session 2: Update on secondary treatment in AIS Chairs: N. Bornstein, Tel-Aviv, N. Ahmed, Stockholm, Secretary: C. Cooray, Stockholm, Speakers: M. Paciaroni/V. Caso, Perugia, R. Bulbulia (Oxford), H. Mattle (Bern), N. Bornstein (Tel Aviv) Patients with atrial fibrillation and AIS-timing of anticoagulation When is the best time for initiating anticoagulation treatment after AIS based on RAF study?

Session 2: Update on secondary treatment in AIS Chairs: N. Bornstein, Tel-Aviv, N. Ahmed, Stockholm, Secretary: C. Cooray, Stockholm, Speakers: M. Paciaroni/V. Caso, Perugia, R. Bulbulia (Oxford), H. Mattle (Bern), N. Bornstein (Tel Aviv) Patients with atrial fibrillation and AIS-timing of anticoagulation When is the best time for initiating anticoagulation treatment after AIS based on RAF study? In patients with AIS and atrial fibrillation, we recommend that oral anticoagulation treatment may be started at day 4 in mild stroke and small infarct, at day 7 in moderate stroke with medium infarcts, and at day 14 in severe stroke with large infarcts from index stroke. More data from randomised controlled trials (RCTs) and prospective registries are needed to verify these time-points, in particular for direct oral anticoagulants (Grade C). Should low molecular weight heparin (LMWH) not be used alone or prior to start of oral anticoagulation treatment in patients with AF and ischaemic stroke? Based on observational study results, bridging therapy with LMWH, prior to oral anticoagulation therapy may not be used in patients with atrial fibrillation and ischaemic stroke (Grade C). Prevention of stroke in patients with patent foramen ovale (PFO): An update Are there sufficient data from the available RCTs to recommend device closure of a symptomatic (Stroke/TIA) PFO? To whom?

Based on observational study results, bridging therapy with LMWH, prior to oral anticoagulation therapy may not be used in patients with atrial fibrillation and ischaemic stroke (Grade C). Prevention of stroke in patients with patent foramen ovale (PFO): An update Are there sufficient data from the available RCTs to recommend device closure of a symptomatic (Stroke/TIA) PFO? To whom? We recommend that percutaneous PFO closure should be offered to patients with cryptogenic stroke and a PFO provided that the PFO is likely stroke-related according to the RoPE score (Grade A). Considering the best medical treatment-antiplatelets vs. anticoagulation. Long-term follow-up with no crossover and loss of follow-up in the studies is a serious concern. Are further studies feasible? Current evidence did not show any difference in outcome comparing oral anticoagulation and antiplatelet therapy for secondary stroke prevention in patients with PFO. We recommend future randomized trials comparing different antithrombotic/anticoagulant approaches in patients with cryptogenic stroke and PFO, especially trials that include the non-vitamin K antagonist (VKA) oral anticoagulants (Grade B). Is the RoPE score good enough to differentiate between ‘incidental’ and ‘causal’ PFO? Currently, the Risk of Paradoxical Embolism (RoPE) score represents the best tool to estimate the probability whether a discovered PFO is likely stroke-related or incidental. It is desirable that the ROPE score be validated in a prospective large cohort (Grade B). Update on carotid surgery and stenting

dimou, GA Ford, M Lantz, C Sjöstrand, H Christensen, A Steinberg, T Tomson, M Holtkamp, C Cordonnier, KR Lees, E Eriksson, B Norrving, R Veltkamp, M Dichgans, K Kostulas, T Robinson, W Hacke, D Russell, M Thorèn, P Ringleb, M Söderman, I Markaki, M Brainin, D Leys, U Fischer, M Mazya, G Andersen, D Damgaard, A Davalos. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: N Ahmed is vice chair of SITS International, which receives a grant from Boehringer Ingelheim for the SITS-ISTR. T Steiner received research funding from Octapharma, speakers honoraria from Boehringer Ingelheim, BMS Pifzer, Bayer, Daiichy Sanky, consultancy fees from Boehringer Ingelheim, BMS Pifzer, Bayer, Daiichy Sanky, holds shares from NovoNordisk. V Caso received advisory boards & speaker fees from Boehringher-Ingelheim, BMS-PFIZER, EVER PHARMA, Chair of Respect ESUS Safety Board, PI for MindMaze Motion Pro Study. N Wahlgren is chair of SITS International, which receives a grant from Boehringer Ingelheim for the SITS-ISTR. N Wahlgren has also received lecture or consultancy fees from AstraZeneca, Boehringer Ingelheim and Ferrer.

Current evidence did not show any difference in outcome comparing oral anticoagulation and antiplatelet therapy for secondary stroke prevention in patients with PFO. We recommend future randomized trials comparing different antithrombotic/anticoagulant approaches in patients with cryptogenic stroke and PFO, especially trials that include the non-vitamin K antagonist (VKA) oral anticoagulants (Grade B). Is the RoPE score good enough to differentiate between ‘incidental’ and ‘causal’ PFO? Currently, the Risk of Paradoxical Embolism (RoPE) score represents the best tool to estimate the probability whether a discovered PFO is likely stroke-related or incidental. It is desirable that the ROPE score be validated in a prospective large cohort (Grade B). Update on carotid surgery and stenting Given the recent improvements in medical therapy, should we continue to base our treatment decisions on data from ‘old’ symptomatic carotid trials? Patients with symptomatic carotid stenosis and a high risk of recurrent stroke (e.g. >70% carotid stenosis, ischaemic event <2 weeks previously) should be offered timely intervention with carotid intervention (Grade A). Patients with symptomatic carotid stenosis and lower-risk of recurrent stroke (e.g. moderate carotid stenosis, retinal symptoms only, event > 2 weeks previously) may be randomised to trials comparing carotid intervention plus medical therapy vs. medical therapy alone (ECST-2/CREST-2) if clinician and patient substantially uncertain about the benefits of intervention (Grade B). Is it ever appropriate to intervene on a <50% symptomatic stenosis?

only, event > 2 weeks previously) may be randomised to trials comparing carotid intervention plus medical therapy vs. medical therapy alone (ECST-2/CREST-2) if clinician and patient substantially uncertain about the benefits of intervention (Grade B). Is it ever appropriate to intervene on a <50% symptomatic stenosis? Almost all patients with <50% symptomatic carotid stenosis should not be treated with intervention. However, intervention in certain patients may be considered if the stenosis causes recurrent symptoms despite optimal medical therapy (Grade C). Does gender matter – Do women really derive less benefit from carotid intervention than men? Decisions on whether or not to intervene on patients with carotid stenosis should not be based on gender (Grade A). With more experience, better case selection and technological advances, can CAS compete with carotid endarterectomy? Carotid artery stenting (CAS) is an effective alternative intervention in selected cases (e.g. not recently symptomatic, age <70 years, no prior ischaemic brain damage) when done by experienced interventionists. Technological advances in cerebral protection, access and stent design should be considered in patients treated with CAS (Grade A). Session 3: Lipid lowering for primary and secondary stroke prevention – New guideline? Chair: E. Berge (Oslo), Secretary: T. Prazeres Moreira (Stockholm), Speakers: G. Ntaios (Larissa) and A. Charidimou (London) Should aggressive lipid lowering therapy be given for secondary prevention of stroke?

Carotid artery stenting (CAS) is an effective alternative intervention in selected cases (e.g. not recently symptomatic, age <70 years, no prior ischaemic brain damage) when done by experienced interventionists. Technological advances in cerebral protection, access and stent design should be considered in patients treated with CAS (Grade A). Session 3: Lipid lowering for primary and secondary stroke prevention – New guideline? Chair: E. Berge (Oslo), Secretary: T. Prazeres Moreira (Stockholm), Speakers: G. Ntaios (Larissa) and A. Charidimou (London) Should aggressive lipid lowering therapy be given for secondary prevention of stroke? We recommend that statins be used as a part of standard secondary prophylactic treatment after an ischaemic stroke or a transient ischaemic attack (TIA). Benefits were observed both with atorvastatin 80 mg and with simvastatin 40 mg (Grade A). The use of statins in secondary prevention of ischaemic stroke caused by less frequent non-atherosclerotic etiologies such as arterial dissection and PFO requires further investigations. Should lipid lowering therapy be given in the acute phase of stroke?

th with atorvastatin 80 mg and with simvastatin 40 mg (Grade A). The use of statins in secondary prevention of ischaemic stroke caused by less frequent non-atherosclerotic etiologies such as arterial dissection and PFO requires further investigations. Should lipid lowering therapy be given in the acute phase of stroke? There is no evidence from RCTs to support the routine use of statins in the acute phase of stroke (first 2 weeks). However, observational studies do not show an increase in symptomatic ICH in patients previously treated with statins or to whom statin was given within 3 days after stroke. Statin treatment is thus recommended to start before discharge from hospital after an AIS or at least during follow-up (Grade C). Should statins be used after intracerebral haemorrhage (ICH)? Statins should be used with caution in patients with previous spontaneous ICH (Grade C) – changed from previous KSU recommendation. Avoiding high-dose statin regimens in patients with ICH should be considered (Grade A) – new. In a subgroup of patients with cerebral amyloid angiopathy-related lobar ICH, statin use should probably be reserved for compelling indications (Grade C). Is there a place for PCSK9 inhibitors for patients with dyslipidaemia and previous stroke or transient ischaemic attack?

lheim, BMS-PFIZER, EVER PHARMA, Chair of Respect ESUS Safety Board, PI for MindMaze Motion Pro Study. N Wahlgren is chair of SITS International, which receives a grant from Boehringer Ingelheim for the SITS-ISTR. N Wahlgren has also received lecture or consultancy fees from AstraZeneca, Boehringer Ingelheim and Ferrer. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: ESO-Karolinska Stroke Update Conference was sponsored by Bayer, Boehringer Ingelheim, Medtronic, Penumbra, St. Jude Medical, Stryker and SITS International. No funding sources had part in the recommendations and consensus statements, or preparation, review, or approval of the recommendations and consensus statements; or the decision to submit the recommendations and consensus statements for publication.

th ICH should be considered (Grade A) – new. In a subgroup of patients with cerebral amyloid angiopathy-related lobar ICH, statin use should probably be reserved for compelling indications (Grade C). Is there a place for PCSK9 inhibitors for patients with dyslipidaemia and previous stroke or transient ischaemic attack? Proprotein convertase subtilisin–kexin type 9 (PCSK9) inhibitors could be considered for patients with previous ischaemic stroke or TIA who (a) have elevated LDL-cholesterol despite aggressive lipid-lowering treatment (defined as atorvastatin 40/80 mg (or rosuvastatin 20/40 mg) plus ezetimibe 10 mg), or (b) have specific statin-related complications (e.g. myopathy, rhabdomyolysis, other idiosyncratic side-effects) (Grade B). Should lipid lowering therapy be given for primary prevention? Lipid lowering treatment in combination with lifestyle changes is recommended for primary prevention in patients who have high 10-year risk for cardiovascular events (Grade A). The drug-class and the intensity of the lipid-lowering treatment as well as the treatment goals are thus depend on patient characteristics (Grade A). Session 4: Guideline for prophylaxis for venous thromboembolism (VTE) (deep vein thrombosis (DVT)) in immobile patients with AIS Chair: G.A Ford (Oxford), Secretary: M. Lantz (Stockholm), Speakers: V. Caso (Perugia), C. Sjöstrand (Stockholm) To endorse the proposed guideline on prophylaxis for VTE in immobile patients with AIS as follows: We recommend that graduated compression stockings should not be used in patients with ischaemic stroke (Grade A).

: G.A Ford (Oxford), Secretary: M. Lantz (Stockholm), Speakers: V. Caso (Perugia), C. Sjöstrand (Stockholm) To endorse the proposed guideline on prophylaxis for VTE in immobile patients with AIS as follows: We recommend that graduated compression stockings should not be used in patients with ischaemic stroke (Grade A). We recommend that intermittent pneumatic compression (IPC, thigh-length, sequential) should be used for immobile patients with ischaemic stroke. It should not be used in patients with open wounds on the legs and should be used with caution in those with existing DVT, heart failure, severe peripheral vascular disease or confusion (Grade A). To consider prophylactic anticoagulation with unfractionated heparin (UFH), LMWH or heparinoid in immobile patients with ischaemic stroke in whom the benefits of reducing the risk of VTE is high enough to offset the increased risks of intracranial and extracranial bleeding associated with their use (Grade A). Where prophylactic anticoagulation is indicated LMWH or heparinoid should be considered instead of UFH because of its greater reduction in risk of DVT, the greater convenience, reduced staff costs and patient comfort. These advantages should be weighed against the higher risk of extracranial bleeding, higher drug costs and risks in elderly patients with poor renal function (Grade A). To ask the ESO to consider the following remarks in relation to the new guidelines: IPC should be used for 30 days or until the patient is mobilizing independently.

Where prophylactic anticoagulation is indicated LMWH or heparinoid should be considered instead of UFH because of its greater reduction in risk of DVT, the greater convenience, reduced staff costs and patient comfort. These advantages should be weighed against the higher risk of extracranial bleeding, higher drug costs and risks in elderly patients with poor renal function (Grade A). To ask the ESO to consider the following remarks in relation to the new guidelines: IPC should be used for 30 days or until the patient is mobilizing independently. IPC should not be commenced if more than 72 h post stroke, unless pre-existing DVT has been ruled out. Prophylactic anticoagulation should be used if IPC is not tolerated. Treatment should be used for 30 days or until mobilized. Prophylactic anticoagulation may be used in combination with IPC in patients with high risk of VTE (e.g. active cancer, coagulation disorder or previous DVT). If prophylactic anticoagulation with LMWH is used, standard prophylaxis doses should be applied. For Enoxaprin subcutaneous injection of 40 mg once daily (20 mg if creatinine clearance <30 ml/min) and for Dalteparin subcutaneous injection of 5000 IE once daily (2500 IE if creatinine clearance <30 ml/min). The risk for bleeding should be assessed before VTE prophylaxis is administered. Research is needed to validate a risk assessment tool to evaluate bleeding risk in patients with ischaemic stroke.

If prophylactic anticoagulation with LMWH is used, standard prophylaxis doses should be applied. For Enoxaprin subcutaneous injection of 40 mg once daily (20 mg if creatinine clearance <30 ml/min) and for Dalteparin subcutaneous injection of 5000 IE once daily (2500 IE if creatinine clearance <30 ml/min). The risk for bleeding should be assessed before VTE prophylaxis is administered. Research is needed to validate a risk assessment tool to evaluate bleeding risk in patients with ischaemic stroke. In patients with poor renal function (creatinine clearance <30 ml/min), or at higher risk for extracranial bleeding (e.g. recent GI bleeding, known gastric ulceration), UFH can be considered before LMWH. In other clinical settings, non-vitamin K oral antagonists (NOACs) have been shown to be effective for prophylactic treatment of VTE prophylaxis. Further research is warranted to investigate if NOAC may be an option for prophylaxis of VTE in patients with ischaemic stroke.

In patients with poor renal function (creatinine clearance <30 ml/min), or at higher risk for extracranial bleeding (e.g. recent GI bleeding, known gastric ulceration), UFH can be considered before LMWH. In other clinical settings, non-vitamin K oral antagonists (NOACs) have been shown to be effective for prophylactic treatment of VTE prophylaxis. Further research is warranted to investigate if NOAC may be an option for prophylaxis of VTE in patients with ischaemic stroke. Session 5: Stroke, seizures and epilepsy Chair: H. Christensen (Copenhagen), Secretary: A. Steinberg (Stockholm), Speakers: T. Tomson (Stockholm), M. Holtkamp (Berlin) Should primary prophylaxis of acute symptomatic or unprovoked seizures be recommended after stroke? RCTs are few and underpowered, and the quality of evidence is generally low. As the risk of acute symptomatic and unprovoked seizures in stroke is low, we do not suggest general use of anti-epileptic drugs (AEDs) in primary prevention after stroke. If treatment is initiated for primary prevention of acute symptomatic seizures, it should be withdrawn after the acute post-stroke phase. Although the risk of unprovoked seizures is considerably higher in patients with large ICH and cortical involvement as well as sinus VTE, primary prevention is rarely justified (Grade C). RCTs are needed to assess the benefits of short- and long-term prophylaxis with AEDs for prevention of acute symptomatic and unprovoked seizures.

Session 5: Stroke, seizures and epilepsy Chair: H. Christensen (Copenhagen), Secretary: A. Steinberg (Stockholm), Speakers: T. Tomson (Stockholm), M. Holtkamp (Berlin) Should primary prophylaxis of acute symptomatic or unprovoked seizures be recommended after stroke? RCTs are few and underpowered, and the quality of evidence is generally low. As the risk of acute symptomatic and unprovoked seizures in stroke is low, we do not suggest general use of anti-epileptic drugs (AEDs) in primary prevention after stroke. If treatment is initiated for primary prevention of acute symptomatic seizures, it should be withdrawn after the acute post-stroke phase. Although the risk of unprovoked seizures is considerably higher in patients with large ICH and cortical involvement as well as sinus VTE, primary prevention is rarely justified (Grade C). RCTs are needed to assess the benefits of short- and long-term prophylaxis with AEDs for prevention of acute symptomatic and unprovoked seizures. Should secondary prophylaxis of seizures be recommended after one or more acute symptomatic or unprovoked seizure in patients after stroke? RCTs are absent and quality of evidence generally low. Acute symptomatic seizures have a low risk of recurrence and thus short- and long-term prevention is not suggested. If treatment is initiated for secondary prevention of acute symptomatic seizures, it should be withdrawn after the acute post-stroke phase. Unprovoked seizures carry a high risk of recurrence and based on observational data, long-term AED should be considered. There are no conclusive RCT data specific to post-stroke populations to guide the choice of AEDs (Grade C evidence).

te symptomatic seizures, it should be withdrawn after the acute post-stroke phase. Unprovoked seizures carry a high risk of recurrence and based on observational data, long-term AED should be considered. There are no conclusive RCT data specific to post-stroke populations to guide the choice of AEDs (Grade C evidence). RCTs are needed, both to assess potential benefit in reduction in risk of seizure recurrence and its consequences, but also in tolerability and adverse effects in this patient population. Session 6: Management of acute stroke (ischaemic or haemorrhagic) under oral anticoagulant therapy Chairs: C. Cordonnier (Lille) and K.R. Lees (Glasgow), Secretary: E. Eriksson (Stockholm), Speakers: B. Norrving (Lund), T. Steiner (Frankfurt/Heidelberg) R. Veltkamp (London). Issue 1. How should we approach neurological emergencies when patients are on OACs? 1. In AIS, laboratory testing before intravenous thrombolysis (IVT) is necessary if relevant anticoagulant activity cannot be ruled out by medical history (Grade C). 2. In acute ICH, reversal of anticoagulation should be started as soon as possible after diagnosis of ICH unless relevant anticoagulant activity is regarded unlikely by medical history or has been ruled out by laboratory testing (Grade C). 3. Recommendation relating to ‘pharmacodynamically relevant (i.e. active) drug concentrations’ (Grade C). a. For VKA: In acute stroke patients on VKA, INR should be measured. An INR ≤1.7 allows IVT in AIS. For ICH patients, i. an INR >2 should trigger reversal treatment with prothrombin complex concentrate (PCC) 30 U/kg.

3. Recommendation relating to ‘pharmacodynamically relevant (i.e. active) drug concentrations’ (Grade C). a. For VKA: In acute stroke patients on VKA, INR should be measured. An INR ≤1.7 allows IVT in AIS. For ICH patients, i. an INR >2 should trigger reversal treatment with prothrombin complex concentrate (PCC) 30 U/kg. ii. an INR >1.2 should trigger reversal treatment with PCC 10 U/kg. b. For NOACs: Relevant drug concentrations in patients on NOACs should be assumed if: i. Global routine tests are above normal 1. Activated Partial Thromboplastin Time (aPTT) for dabigatran 2. Prothrombin time (PT) for rivaroxaban and edoxaban; however, PT should not guide therapy in cases involving apixaban ii. Non-calibrated tests are above normal 1. Ecarin clotting time (ECT) for dabigatran 2. Factor Xa-activity tests for factor Xa-inhibitors iii. Calibrated tests provide information as below: 1. If diluted thrombin time (dTT) for dabigatran indicates concentration >30 ng/dl 2. If factor Xa-activity tests calibrated for factor Xa-inhibitors indicate concentration >30 ng/dl If calibrated tests are available, their thresholds may guide therapy Issue 2A: Management of AIS and indication for reperfusion therapy during treatment with VKAs 1. In patients with AIS and indication for reperfusion therapy during therapy with VKA and an INR ≤1.7, thrombolysis should be performed (Grade C). 2. In patients with AIS during therapy with VKA and an INR >1.7, thrombolysis should not be performed (Grade C).

If calibrated tests are available, their thresholds may guide therapy Issue 2A: Management of AIS and indication for reperfusion therapy during treatment with VKAs 1. In patients with AIS and indication for reperfusion therapy during therapy with VKA and an INR ≤1.7, thrombolysis should be performed (Grade C). 2. In patients with AIS during therapy with VKA and an INR >1.7, thrombolysis should not be performed (Grade C). 3. Patients with AIS during therapy with VKA who suffer from large vessel occlusion with indication for reperfusion therapy should be offered thrombectomy (Grade C). Issue 2B: Management of acute ICH during treatment with VKAs 1. In adult patients with ICH related to VKA and with an INR ≥2, intravenous 4-factor-PCC in a dose of at least 30 U/kg should be administered to normalise the INR and limit haematoma expansion (Grade B). Reversal of anticoagulation with PCC may also be initiated at INR between 1.2 and 2.0 with lower PCC-dose of 10 U/kg (Grade C). 2. Reversal with fresh frozen plasma is not recommended (Grade C). 3. Administration of vitamin K (10 mg, iv) may be considered if the initial INR ≥1.2 on repeated measurements (Grade C). Issue 3A: Management of AIS and acute ICH occurring during treatment with non-vitamin K oral anticoagulants 1. In adult patients with AIS related to factor Xa-inhibitors and suspicion or evidence of relevant drug concentrations, IVT should not be performed (Grade C). 2. In adult patients with AIS related to dabigatran and the suspicion or evidence of relevant drug concentrations, IVT cannot presently be recommended (Grade C).

Issue 3A: Management of AIS and acute ICH occurring during treatment with non-vitamin K oral anticoagulants 1. In adult patients with AIS related to factor Xa-inhibitors and suspicion or evidence of relevant drug concentrations, IVT should not be performed (Grade C). 2. In adult patients with AIS related to dabigatran and the suspicion or evidence of relevant drug concentrations, IVT cannot presently be recommended (Grade C). 3. In adult patients with AIS related to NOACs, thrombectomy should be performed consistent with recommendations for non-anticoagulated patients (Grade C). Issue 3B: Management of acute ICH occurring during treatment with NOAC 1. In patients with ICH related to dabigatran, idarucizumab 2 × 2.5 g should be injected (Grade B). 2. If idarucizumab is not available, PCC may be infused (30–50 U/kg) (Grade C). 3. In patients with ICH-related to apixaban, edoxaban or rivaroxaban PCC (30–50 U/kg) should be used (Grade C). 4. Reversal of NOAC with fresh frozen plasma is not recommended (Grade C). Session 7: IV thrombolysis in AIS-dosing of alteplase Chair: M. Dichgans (Munich), Session secretary: K. Kostulas (Stockholm), Speakers: T. Robinson (Leicester), W. Hacke (Heidelberg) Do the results of the ENCHANTED study support a recommendation of a dose of 0.6 mg/kg of alteplase for iv thrombolysis for an Asian population? Standard-dose intravenous alteplase (0.9 mg/kg body weight, maximum 90 mg), with 10% of the dose given as a bolus followed by a 60-min infusion, is recommended within 4.5 h of onset of ischaemic stroke (Grade A).

y support a recommendation of a dose of 0.6 mg/kg of alteplase for iv thrombolysis for an Asian population? Standard-dose intravenous alteplase (0.9 mg/kg body weight, maximum 90 mg), with 10% of the dose given as a bolus followed by a 60-min infusion, is recommended within 4.5 h of onset of ischaemic stroke (Grade A). Ethnicity should not be used as a reason for not offering best treatment, i.e. standard-dose alteplase (Grade B). Do the results of the ENCHANTED study support a recommendation of a dose of 0.6 mg/kg of alteplase for iv thrombolysis for a European population? Where there is concern over symptomatic ICH risk, further RCTs are required to define the patient populations in whom low-dose intravenous alteplase (0.6 mg/kg body weight, maximum 60 mg) may be considered (Grade C). Session 8: Management of symptomatic intracranial stenosis Chair: D. Russell (Oslo), Secretary: M. Thorèn (Stockholm), Speakers: P. Ringleb (Heidelberg), M. Söderman (Stockholm) Is intensive medical management the primary recommended therapy for the management of symptomatic intracranial stenosis? Strict risk factor management and optimal medical therapy is the primary recommended treatment for the management of symptomatic intracranial stenosis (Grade B evidence). If so, are there subgroups of patients for which angioplasty and/or stent placement would offer a better or equivalent alterative?

Session 8: Management of symptomatic intracranial stenosis Chair: D. Russell (Oslo), Secretary: M. Thorèn (Stockholm), Speakers: P. Ringleb (Heidelberg), M. Söderman (Stockholm) Is intensive medical management the primary recommended therapy for the management of symptomatic intracranial stenosis? Strict risk factor management and optimal medical therapy is the primary recommended treatment for the management of symptomatic intracranial stenosis (Grade B evidence). If so, are there subgroups of patients for which angioplasty and/or stent placement would offer a better or equivalent alterative? There is not enough evidence to recommend situations where angioplasty and/or stent placement would offer a better or equivalent alterative. Although there is no evidence, the role of angioplasty and stenting, carried out by experienced personnel, may be considered in a few special situations (Grade C evidence). RCTs or prospective registry studies are therefore required. Session 9: How to reach a cognitive endpoint in stroke trials? Chair: V. Caso (Perugia), Secretary: I. Markaki (Stockholm), Speakers: M. Brainin (Krems), D. Leys (Lille) Aims for this session: Strategies that guarantee that cognitive endpoints are included in future major stroke studies/trials Neuropsychological tests for best identifying cognitive endpoints Appropriate tailor strategies for the education of clinicians and researchers on the interplay between stroke and dementia Cognitive endpoints should be included in all stroke trials (Grade C).

Aims for this session: Strategies that guarantee that cognitive endpoints are included in future major stroke studies/trials Neuropsychological tests for best identifying cognitive endpoints Appropriate tailor strategies for the education of clinicians and researchers on the interplay between stroke and dementia Cognitive endpoints should be included in all stroke trials (Grade C). The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) or equivalent should be included in acute stroke trials to be sure that groups are balanced for pre-existing cognitive impairment. Two versions of neuropsychological test batteries may be considered within 3 to 6 months post stroke: a short version that can be conducted by trained nurses or physicians, and a more comprehensive long version that has to be performed mostly by trained neuropsychologists. The short test battery could include the Montreal Cognitive Assessment (MoCA), the Trail Making Test A and B and the digit span forward and backward. An extended test battery should assess multiple domains and be composed of validated neuropsychological tests fulfilling different criteria regarding psychometrics, usability, costs, time, language and culture. Sample sizes and duration of follow-up should be taken into account in prevention trials to evaluate cognitive outcomes. It is advisable to include also a short depression scale, a self-rating scale such as the Beck Depression inventory or the Centre of Epidemiologic Studies Depression scale.

An extended test battery should assess multiple domains and be composed of validated neuropsychological tests fulfilling different criteria regarding psychometrics, usability, costs, time, language and culture. Sample sizes and duration of follow-up should be taken into account in prevention trials to evaluate cognitive outcomes. It is advisable to include also a short depression scale, a self-rating scale such as the Beck Depression inventory or the Centre of Epidemiologic Studies Depression scale. Focus on longstanding effects of interventions should also consider assessment of fatigue and apathy, as well as caregiver status. Session 10: Prehospital triage for mechanical thrombectomy Chair: U. Fischer (Bern), Secretary: M. Mazya (Stockholm), Speakers: D. Damgaard (Aarhus), A. Davalos (Barcelona), M. Mazya (Stockholm) A. Clinical identification of stroke patients with large vessel occlusion: Current evidence and limitations Several published clinical scores to predict large artery occlusion (LAO) appear to have similar predictive performance in the range of 70–80%, resulting in 20–30% of patients with LAO being missed at optimal score cut-off levels. At the same cut-off levels, 12–25% of triage positive patients would not have a LAO (Grade C). Studies validating the predictive performance of currently available LAO prediction scores should be performed in pre-hospital settings in unselected patients with a suspicion of stroke following initial contact with emergency medical services (Grade C). B. Mechanical thrombectomy: ‘Drip and ship’ or ‘load and go’?

Several published clinical scores to predict large artery occlusion (LAO) appear to have similar predictive performance in the range of 70–80%, resulting in 20–30% of patients with LAO being missed at optimal score cut-off levels. At the same cut-off levels, 12–25% of triage positive patients would not have a LAO (Grade C). Studies validating the predictive performance of currently available LAO prediction scores should be performed in pre-hospital settings in unselected patients with a suspicion of stroke following initial contact with emergency medical services (Grade C). B. Mechanical thrombectomy: ‘Drip and ship’ or ‘load and go’? For patients with a suspected LAO based on current clinical tools on field, there is uncertainty about the equipoise between drip and ship (that prioritizes early IVT and other standard of care therapies) and mother-ship (that prioritizes early endovascular thrombectomy) models. Data based on randomized controlled trials are needed to determine the most beneficial model for each particular patient (eligible or not for iv-tPA) in different geographical regions and to establish isochrones where a particular model may be beneficial (Grade C). In the absence of evidence, for patients considered eligible to IVT in the field, if estimated transfer time to the nearest primary stroke centre is considerably shorter than time to a comprehensive stroke centre (approximately more than 30–45 min), the drip and ship model should be considered (Grade C).

For patients with a suspected LAO based on current clinical tools on field, there is uncertainty about the equipoise between drip and ship (that prioritizes early IVT and other standard of care therapies) and mother-ship (that prioritizes early endovascular thrombectomy) models. Data based on randomized controlled trials are needed to determine the most beneficial model for each particular patient (eligible or not for iv-tPA) in different geographical regions and to establish isochrones where a particular model may be beneficial (Grade C). In the absence of evidence, for patients considered eligible to IVT in the field, if estimated transfer time to the nearest primary stroke centre is considerably shorter than time to a comprehensive stroke centre (approximately more than 30–45 min), the drip and ship model should be considered (Grade C). In the absence of evidence, in a scenario where a primary stroke centre and comprehensive stroke centre are equidistant (approximately not more than 30–45 min apart) or when contraindications to IVT are known in the field, patients with suspected LAO in the field, should be considered for transfer directly to a comprehensive stroke centre, bypassing any closer primary stroke centres (Grade C).

comprehensive stroke centre are equidistant (approximately not more than 30–45 min apart) or when contraindications to IVT are known in the field, patients with suspected LAO in the field, should be considered for transfer directly to a comprehensive stroke centre, bypassing any closer primary stroke centres (Grade C). In case of primary admission to a primary stroke centre, evaluation and treatment for patients with a possible LAO must be expeditious, to ensure a rapid secondary transfer to a comprehensive stroke centre, avoiding any sources of delay such as complex neuroimaging studies (i.e. perfusion studies) or waiting for effect of IVT. First picture to puncture time should be less than 90 min (Grade A). Ethics approval Not applicable. Provenance The Recommendations from the ESO-Karolinska Stroke Update Conference were not externally peer reviewed, as they are consensus documents that have been previously peer reviewed by the writing committee for each session. The manuscript was approved for publication by Bo Norrving as Editor-in-Chief and Didier Leys as Vice Editor. Supplementary Material Supplementary material Acknowledgement The authors acknowledge Charlotte Alme, Lena Liljeblad and Marius Matusevicius for their assistance with collection and formatting the document. Congrex Switzerland Ltd assisted in organising the conference.

Provenance The Recommendations from the ESO-Karolinska Stroke Update Conference were not externally peer reviewed, as they are consensus documents that have been previously peer reviewed by the writing committee for each session. The manuscript was approved for publication by Bo Norrving as Editor-in-Chief and Didier Leys as Vice Editor. Supplementary Material Supplementary material Acknowledgement The authors acknowledge Charlotte Alme, Lena Liljeblad and Marius Matusevicius for their assistance with collection and formatting the document. Congrex Switzerland Ltd assisted in organising the conference. * Collaborators: T Tatlisumak, E Lundström, S Debette, H Markus, ST Engelter, M Arnold, N Bornstein, C Cooray, M Paciaroni, R Bulbulia, H Mattle, E Berge, T Prazeres Moreira, G Ntaios, A Charidimou, GA Ford, M Lantz, C Sjöstrand, H Christensen, A Steinberg, T Tomson, M Holtkamp, C Cordonnier, KR Lees, E Eriksson, B Norrving, R Veltkamp, M Dichgans, K Kostulas, T Robinson, W Hacke, D Russell, M Thorèn, P Ringleb, M Söderman, I Markaki, M Brainin, D Leys, U Fischer, M Mazya, G Andersen, D Damgaard, A Davalos.

Introduction Most in-hospital deaths of patients with acute stroke occur after a decision to withhold or withdraw life-sustaining therapies.1,2 The process to make decisions about treatment restrictions in patients with acute stroke differs from that in patients with progressive disease such as cancer because stroke patients often cannot fully participate in this process and because continuation of treatment potentially allows patients to live for months or years at the cost of being left in a state of disability that might be against their wishes.3 Treatment restrictions in the first 2 days after intracerebral haemorrhage have been independently associated with an increased risk of early death,2,4,5 and avoidance of new do-not-resuscitate (DNR) orders during the first 5 days after intracerebral haemorrhage has been associated with a substantially lower 30-day mortality rate than predicted.6 Treatment restrictions are also frequently installed in a later stage,7 but it has not been investigated whether these are also associated with early mortality. Postponing the instalment of treatment restrictions increases the window of opportunity for patients to express their wishes regarding life-sustaining treatments. In this prospective observational study, we assessed the relation between the placement of treatment restrictions and mortality in patients who had survived the first 4 days after severe ischaemic stroke or intracerebral haemorrhage. We also assessed functional outcome and quality of life in survivors.

Treatment restrictions in the first 2 days after intracerebral haemorrhage have been independently associated with an increased risk of early death,2,4,5 and avoidance of new do-not-resuscitate (DNR) orders during the first 5 days after intracerebral haemorrhage has been associated with a substantially lower 30-day mortality rate than predicted.6 Treatment restrictions are also frequently installed in a later stage,7 but it has not been investigated whether these are also associated with early mortality. Postponing the instalment of treatment restrictions increases the window of opportunity for patients to express their wishes regarding life-sustaining treatments. In this prospective observational study, we assessed the relation between the placement of treatment restrictions and mortality in patients who had survived the first 4 days after severe ischaemic stroke or intracerebral haemorrhage. We also assessed functional outcome and quality of life in survivors. Methods This is a prospective two-centre cohort study. Consecutive patients admitted at the stroke unit with an acute severe ischaemic or haemorrhagic stroke with a very small chance of functional independency after 6 months (defined as Barthel Index (BI) ≤ 6 out of 20 at day 4)8 were included. Patients with a subarachnoid haemorrhage and incompetent patients without an available legal representative were excluded from the study. Patients were included between September 2012 and December 2013 in the University Medical Center Utrecht, and between January and December 2013 in the St. Elisabeth hospital in Tilburg, a large regional teaching hospital in the Netherlands.

atients without an available legal representative were excluded from the study. Patients were included between September 2012 and December 2013 in the University Medical Center Utrecht, and between January and December 2013 in the St. Elisabeth hospital in Tilburg, a large regional teaching hospital in the Netherlands. We collected information on patient characteristics, type of stroke (ischaemic or haemorrhagic), stroke severity on admission (by means of National Institutes of Health Stroke Scale (NIHSS) and pre-stroke comorbidity (by means of the Charlson Comorbidity Index (CCI)).9 Treatment restrictions were assessed by a semi-structured questionnaire administered to the treating physician at the day of inclusion. Treatment restrictions were coded for the following categories: (a) DNR order, (b) withhold admission to intensive care unit (ICU), (c) withhold curative treatment of complications and (d) withhold artificial nutrition and hydration. These are incremental steps: each treatment restriction is added up to the before-mentioned treatment restrictions. We assessed all in-hospital treatment restrictions that were installed at study inclusion.

unit (ICU), (c) withhold curative treatment of complications and (d) withhold artificial nutrition and hydration. These are incremental steps: each treatment restriction is added up to the before-mentioned treatment restrictions. We assessed all in-hospital treatment restrictions that were installed at study inclusion. One trained investigator (FASdK) visited each patient and his/her caregiver at 6 months (±6 weeks) after stroke to assess functional outcome and quality of life. Functional outcome was assessed with the modified Rankin Scale (mRS); poor outcome was defined as mRS > 3. Patients’ quality of life was measured with a visual analogue scale (VAS).10 The VAS was a vertical line of 10 cm with a ‘’ at the top demarcating the best possible quality of life and a ‘’ at the lower end for the worst possible quality of life. Scores were calculated as the indicated level in (cm/10) × 100. Quality of life was considered acceptable if VAS ≥ 60. The primary outcome measure was mortality at 6 months. Secondary outcome measures were functional outcome (mRS) and quality of life (VAS) at 6 months. The association between treatment restrictions and these outcomes was calculated with Poisson regression analysis with a robust error after adjustment for age, sex, NIHSS on admission, BI at day 4, CCI and type of stroke (ischaemic or haemorrhagic). We expressed associations as adjusted relative risk (aRR) with 95% CI.

The association between treatment restrictions and these outcomes was calculated with Poisson regression analysis with a robust error after adjustment for age, sex, NIHSS on admission, BI at day 4, CCI and type of stroke (ischaemic or haemorrhagic). We expressed associations as adjusted relative risk (aRR) with 95% CI. We performed post-hoc subgroup analyses in patients with acute ischaemic stroke and intracerebral haemorrhage separately. In this subgroup analyses, we adjusted for age, sex, NIHSS on admission, BI at day 4 and CCI. The study was approved by the institutional review board of each centre, and written informed consent was obtained from each patient or a legal representative. Results Of 874 stroke patients admitted during the course of the study, 127 fulfilled the inclusion criteria and 60 were included. Eight patients were excluded because they had no legal representative available, 48 patients declined participation and 11 were missed (Figure 1). Figure 1. Flow of patients through this study. The median time between stroke onset and inclusion was 6 days (range, 4–10). The mean age of the patients was 72 years (SD 15); 30 (50%) were male; the median NIHSS on admission was 16 (3–28) and the median BI at day 4 was 2 (0–6). Additional patient characteristics are presented in Table 1. Table 1. Baseline characteristics.

ian time between stroke onset and inclusion was 6 days (range, 4–10). The mean age of the patients was 72 years (SD 15); 30 (50%) were male; the median NIHSS on admission was 16 (3–28) and the median BI at day 4 was 2 (0–6). Additional patient characteristics are presented in Table 1. Table 1. Baseline characteristics. All patients n = 60 Full care n = 18 DNR-ordera n = 42 Withhold admission at ICU n = 30 No curative treatment of complications n = 12 Withhold artificial nutrition and hydration n = 10 Age (years) 72 (15) 56 (11) 79 (11) 80 (12) 78 (15) 80 (8) Men 30 (50) 13 (72) 17 (41) 11 (37) 4 (33) 3 (30) Ischaemic stroke 36 (60) 10 (56) 26 (62) 19 (63) 5 (42) 4 (40) NIHSS on admission 16 (6) 16 (6) 16 (6) 16 (7) 19 (7) 19 (7) CCI 1 (0–6) 0 (0–4) 1 (0–6) 1 (0–6) 1 (0–4) 1 (0–4) Barthel Index at day 4 0 (0–6) 2 (0–3) 0 (0–6) 0 (0–6) 0 (0–0) 0 (0–0) Data are n (%), median (range) or mean (standard deviation (SD)) where appropriate. a DNR-order represents all treatment restrictions. DNR: do not resuscitate; ICU: Intensive Care Unit; NIHSS: National Institutes of Health Stroke Scale; CCI: Charlson Comorbidity Index. Forty-two patients (70%) had one or more treatment restrictions. Patients without treatment restrictions were younger than patients with treatment restrictions (56 vs 79 years, p < 0.001), and were more often men (72 vs 41%, p = 0.02) (Table 1). At 6 months, 30 (50%) patients had died, of whom 12 during admission. The median time from stroke onset to in-hospital death was 9 days (range, 3–18). Twenty-eight of the patients who died (93%) had a treatment restriction.

Forty-two patients (70%) had one or more treatment restrictions. Patients without treatment restrictions were younger than patients with treatment restrictions (56 vs 79 years, p < 0.001), and were more often men (72 vs 41%, p = 0.02) (Table 1). At 6 months, 30 (50%) patients had died, of whom 12 during admission. The median time from stroke onset to in-hospital death was 9 days (range, 3–18). Twenty-eight of the patients who died (93%) had a treatment restriction. The presence of any treatment restriction at study inclusion was independently associated with mortality at 6 months (aRR, 1.30; 95% confidence interval, 1.06–1.59; p = 0.01). Each individual type of treatment restriction was also associated with mortality at 6 months (Table 2). Table 2. Results on adjusted Poisson regression analysis on the relation between type of treatment restrictions and mortality. aRRa 95%CI P DNR-orderb 1.30 1.06–1.59 0.01 Withhold admission at ICU 1.41 1.20–1.65 <0.001 No curative treatment of complications 1.26 1.11–1.44 0.001 Withhold artificial nutrition and hydration 1.19 1.05–1.34 0.01 a Adjusted for age, sex, National Institutes of Health Stroke Scale score on admission, Barthel Index at day 4, Charlson Comorbidity Index and type of stroke. b DNR-order represents all treatment restrictions. aRR: adjusted relative risk; CI: confidence interval; DNR: do not resuscitate; ICU: Intensive Care Unit.

aRRa 95%CI P DNR-orderb 1.30 1.06–1.59 0.01 Withhold admission at ICU 1.41 1.20–1.65 <0.001 No curative treatment of complications 1.26 1.11–1.44 0.001 Withhold artificial nutrition and hydration 1.19 1.05–1.34 0.01 a Adjusted for age, sex, National Institutes of Health Stroke Scale score on admission, Barthel Index at day 4, Charlson Comorbidity Index and type of stroke. b DNR-order represents all treatment restrictions. aRR: adjusted relative risk; CI: confidence interval; DNR: do not resuscitate; ICU: Intensive Care Unit. At 6 months, 19 of 30 survivors (63%) had a poor functional outcome (Table 3, Figure 2). Quality of life could be assessed in 26 survivors. Mean score on the VAS was 60 (SD 17). Quality of life was considered satisfactory in 11 of 16 (69%) survivors with a poor functional outcome, and in 6 of 10 (60%) patients with a good functional outcome (Table 3). Table 3. Outcome of survivors at 6 months. All patients n = 60 Full care n = 18 DNR-ordera n = 42 Withhold admission at ICU n = 30 No curative treatment of complications n = 12 Withhold artificial nutrition and hydration n = 10 Alive at 6 months 30 (50) 16 (89) 14 (33) 5 (17) 0 (0) 0 (0) Good functional outcomeb 10 (34) 7 (44) 3 (23) 1 (20) 0 (0) 0 (0) Satisfactory quality of lifec 17 (65) 10 (63) 7 (70) 2 (66) 0 (0) 0 (0) Data are n (%), median (range) or mean (standard deviation (SD)) where appropriate. a DNR represents all treatment restrictions. b n = 29 survivors, 1 patient declined follow-up. c n = 26 survivors, 3 patients declined follow-up and one was aphasic.

All patients n = 60 Full care n = 18 DNR-ordera n = 42 Withhold admission at ICU n = 30 No curative treatment of complications n = 12 Withhold artificial nutrition and hydration n = 10 Alive at 6 months 30 (50) 16 (89) 14 (33) 5 (17) 0 (0) 0 (0) Good functional outcomeb 10 (34) 7 (44) 3 (23) 1 (20) 0 (0) 0 (0) Satisfactory quality of lifec 17 (65) 10 (63) 7 (70) 2 (66) 0 (0) 0 (0) Data are n (%), median (range) or mean (standard deviation (SD)) where appropriate. a DNR represents all treatment restrictions. b n = 29 survivors, 1 patient declined follow-up. c n = 26 survivors, 3 patients declined follow-up and one was aphasic. DNR: do not resuscitate; ICU: Intensive Care Unit; NIHSS: National Institutes of Health Stroke Scale; CCI: Charlson Comorbidity Index. Figure 2. Functional outcome at 6 months. Treatment restrictions were not associated with a poor functional outcome in survivors (Table 4), but patient numbers were small. Table 4. Results on adjusted Poisson regression analysis on the relation between treatment restrictions and poor functional outcome in survivors. aRRa 95%CI p DNR-orderb 0.78 0.49–1.23 0.28 Withhold admission at ICU 0.87 0.64–1.17 0.34 No curative treatment of complications 0.96 0.81–1.15 0.68 Withhold artificial nutrition and hydration 1.02 0.87–1.18 0.84 a Adjusted for age, sex, National Institutes of Health Stroke Scale score on admission, Barthel Index at day 4, Charlson Comorbidity Index and type of stroke. b DNR-order represents all treatment restrictions. aRR: adjusted relative risk; CI: confidence interval; DNR: do not resuscitate; ICU: Intensive Care Unit.

aRRa 95%CI p DNR-orderb 0.78 0.49–1.23 0.28 Withhold admission at ICU 0.87 0.64–1.17 0.34 No curative treatment of complications 0.96 0.81–1.15 0.68 Withhold artificial nutrition and hydration 1.02 0.87–1.18 0.84 a Adjusted for age, sex, National Institutes of Health Stroke Scale score on admission, Barthel Index at day 4, Charlson Comorbidity Index and type of stroke. b DNR-order represents all treatment restrictions. aRR: adjusted relative risk; CI: confidence interval; DNR: do not resuscitate; ICU: Intensive Care Unit. Subgroup analysis In a post-hoc subgroup analysis in the 36 patients with ischaemic stroke, results were essentially the same. The presence of any treatment restriction at study inclusion was independently associated with mortality at 6 months (aRR, 1.33; 95% confidence interval, 1.01–1.76; p = 0.04). Each individual type of treatment restriction was also associated with mortality at 6 months (Supplemental Table 1). In 24 patients with intracerebral haemorrhage, results were comparable but did not reach statistical significance (aRR, 1.15; 95% confidence interval, 0.97–1.36; p = 0.11) (Supplemental Table 2). Discussion This study shows that in patients with severely disabling ischaemic stroke or intracerebral haemorrhage, treatment restrictions installed several days after stroke onset are associated with mortality at 6 months, independent of age, sex, stroke severity or pre-stroke comorbidity.

In 24 patients with intracerebral haemorrhage, results were comparable but did not reach statistical significance (aRR, 1.15; 95% confidence interval, 0.97–1.36; p = 0.11) (Supplemental Table 2). Discussion This study shows that in patients with severely disabling ischaemic stroke or intracerebral haemorrhage, treatment restrictions installed several days after stroke onset are associated with mortality at 6 months, independent of age, sex, stroke severity or pre-stroke comorbidity. This association between treatment restrictions and mortality is probably at least partially causal, because the aim of these restrictions is to withhold potentially life-prolonging treatments when future quality of life expected to be insufficient, prioritising comfort care. The associations persist after adjustment for other factors that might affect survival such as age, pre-stroke comorbidity and stroke severity. Therefore, our findings suggest that treatment restrictions after the first 4 days increase the risk of death.

y of life expected to be insufficient, prioritising comfort care. The associations persist after adjustment for other factors that might affect survival such as age, pre-stroke comorbidity and stroke severity. Therefore, our findings suggest that treatment restrictions after the first 4 days increase the risk of death. Previous studies have also shown that treatment restrictions are associated with mortality in patients with intracerebral haemorrhage2,4,11,12 and in study populations with both ischaemic and haemorrhagic stroke patients.13,14 Avoidance of early DNR orders has been associated with a substantially lower risk of death,6 supporting a causal relationship between treatment restrictions and early mortality. Whether this relation is causal indeed can only be tested in randomised trials of full medical support during a prespecified time period vs. ‘standard’ care, which includes the placement of treatment restrictions, but this design will likely be considered unethical.

p between treatment restrictions and early mortality. Whether this relation is causal indeed can only be tested in randomised trials of full medical support during a prespecified time period vs. ‘standard’ care, which includes the placement of treatment restrictions, but this design will likely be considered unethical. Our findings have important consequences. In clinical practice, physicians should realise that treatment restrictions on their own may increase the risk of death, and that a poor functional outcome does not necessarily implicate an unsatisfactory quality of life. Therefore, physicians should be cautious to withhold their patients a chance on recovery by installing treatment restrictions too early. With respect to intervention trials and prognostic studies, confounding by treatment restrictions should also be avoided, and where this is not possible, the placement of treatment restrictions should be assessed. Confounding by treatment restrictions could be controlled by the adoption of a standard for withdrawal of life-sustaining treatment in the study protocol.

ic studies, confounding by treatment restrictions should also be avoided, and where this is not possible, the placement of treatment restrictions should be assessed. Confounding by treatment restrictions could be controlled by the adoption of a standard for withdrawal of life-sustaining treatment in the study protocol. Treatment restrictions can be appropriate after severe stroke to prevent a patient for staying alive at the cost of being left in a state of disability that might be against his or her wishes. What constitutes a poor outcome is however difficult to adequately define. Although the majority of patients in our study had a poor functional outcome, the majority of the survivors had a satisfactory quality of life. While increasing disability is generally associated with a reduction in quality of life, this is not the first time that quality of life has been reported satisfactory in patients with a disabling stroke.15,16 Assessment of quality of life by these patients is probably influenced by a response shift, which includes a change in the internal standards and values in the self-assessment of quality of life,17 and by the capacity of patients with chronic illness or disability to adapt to their circumstances, a phenomenon often referred to as the disability paradox.18,19 Unfortunately, in the early phase after stroke, it is still unclear how to identify patients who will adapt well to their new situation and recapture a good quality of life.

ity of patients with chronic illness or disability to adapt to their circumstances, a phenomenon often referred to as the disability paradox.18,19 Unfortunately, in the early phase after stroke, it is still unclear how to identify patients who will adapt well to their new situation and recapture a good quality of life. We aimed to include patients with a very small chance on regaining functional independence, because treatment restrictions are probably most often installed in this patient group. The BI is an easy accessible and widely used scale to measure ADL dependency. A cut-off point of six on the BI at day 5 has previously been shown to be an accurate predictor of ADL independency at 6 months.8 We measured the BI on day 4 as part of routine clinical practice in both participating centres. According to the high rate of patients with poor outcome, the cut-off point of 6 on the BI at day 4 was appropriate.

on the BI at day 5 has previously been shown to be an accurate predictor of ADL independency at 6 months.8 We measured the BI on day 4 as part of routine clinical practice in both participating centres. According to the high rate of patients with poor outcome, the cut-off point of 6 on the BI at day 4 was appropriate. This study has limitations. We could not include half of the eligible patients, because the majority of these patients declined participation. In addition, patients with more severe strokes or their relatives might have been more likely to decline consent, which may have led to selection bias. Our primary outcome was mortality at 6 months, and 60% of deaths occurred after discharge. We have no data on mortality at an earlier time point after discharge and not on treatment restrictions after discharge. We consider it likely that most restrictions already installed were not changed. Moreover, our findings were observed in 60 highly selected patients with severe stroke, and our findings do not apply to patients who are not severely disabled at day 4. We included both patients with severe ischaemic stroke or intracerebral haemorrhage, whereas patients who survive intracerebral haemorrhage to the point of rehabilitation have greater improvement in functional abilities than similarly affected patients with ischaemic stroke.12 However, our findings were independent of stroke type. We adjusted for pre-stroke comorbidities but did not collect data on the presence of complications that occurred after stroke, which may have had on impact on prognosis. Finally, quality of life data should be interpreted with caution because patients could have given desired answers during the home visit.

stroke type. We adjusted for pre-stroke comorbidities but did not collect data on the presence of complications that occurred after stroke, which may have had on impact on prognosis. Finally, quality of life data should be interpreted with caution because patients could have given desired answers during the home visit. In conclusion, both clinicians and researchers should realise that placement of treatment restrictions by itself may increase the risk of death after stroke. ‘Our results need further confirmation. Randomised controlled trials on this topic will not be feasible for ethical reasons. Larger multi-centre cohort studies, prospectively assessing the relation between treatment limitations and mortality should further confirm our findings’. Future research should clarify the clinical practices in end-of-life decisions in stroke patients and focus on identifying patients who will recapture a good quality of life a severely disabling stroke. Supplementary Material Supplementary material Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: M. Geurts and H.B. van der Worp are supported by the Dutch Heart Foundation (2010B239 and 2010T075, respectively). Ethical approval The study was approved by the institutional review board of the University Medical Center Utrecht and of the Elisabeth hospital.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: M. Geurts and H.B. van der Worp are supported by the Dutch Heart Foundation (2010B239 and 2010T075, respectively). Ethical approval The study was approved by the institutional review board of the University Medical Center Utrecht and of the Elisabeth hospital. Informed consent Written informed consent was obtained from each patient or a legal representative. Guarantor HBW. Contributorship MG contributed to the study design, data collection, performed the data analysis, interpreted the data and prepared the first draft and subsequent versions. FASK helped to refine the study idea, contributed to the data collection and data analysis, helped interpreting the data and contributed to writing. JHT helped with data collection and contributed to writing. GJMWT, PLMK and LJK helped to refine the study idea, helped interpreting the data and contributed to writing. HBW conceived the study, helped to refine the study idea, helped interpreting the data and contributed to writing.

Introduction Extracranial internal carotid artery (ICA) stenosis of ≥50% or occlusion caused 12% of strokes in patients who were admitted to the hospital because of first-ever or recurrent ischaemic stroke.1 Both carotid artery stenting (CAS) and carotid endarterectomy (CEA) reduce the risk of new ipsilateral stroke in patients with symptomatic ICA stenosis, but carry a periprocedural complication risk. The risk of stroke or death within 30 days after treatment is higher after CAS than after CEA, which is partly explained by the effect of age,2 but the postprocedural rate of ipsilateral stroke does not differ between both treatments.3 For patients with asymptomatic ICA stenosis, preferred treatment (carotid revascularisation or best medical treatment) remains unclear.4 Prediction of clinical outcome after CAS or CEA could aid physicians to estimate periprocedural and postprocedural risks and to choose the appropriate treatment in individual patients. Our aim was to identify existing prediction models or risk scores for short- or long-term outcome after CAS or CEA in patients with symptomatic or asymptomatic ICA stenosis from the literature, and to summarise their most important predictors and predictive performance. We focused on prediction models that included patient characteristics that are known before a procedure is initiated rather than procedural characteristics, because such models may aid in choosing the appropriate treatment (CAS, CEA, or optimal medical treatment alone).

important predictors and predictive performance. We focused on prediction models that included patient characteristics that are known before a procedure is initiated rather than procedural characteristics, because such models may aid in choosing the appropriate treatment (CAS, CEA, or optimal medical treatment alone). Methods Literature search We performed a literature search in PubMed and Embase to identify articles on prediction models for short- or long-term outcome after CAS or CEA published until 22 December 2016. We defined short term as models that predicted in-hospital risk of outcome or risk up to 30 days after the procedure; we classified all other models as long-term models. We searched with synonyms and index terms for ‘carotid artery stenting OR carotid endarterectomy’ AND ‘prediction model’ AND ‘vascular events’. Supplementary Table 1 shows the full syntax.

redicted in-hospital risk of outcome or risk up to 30 days after the procedure; we classified all other models as long-term models. We searched with synonyms and index terms for ‘carotid artery stenting OR carotid endarterectomy’ AND ‘prediction model’ AND ‘vascular events’. Supplementary Table 1 shows the full syntax. Eligibility criteria We included all articles that reported the development or external validation of one or more multivariable prediction models for individual risk estimation of clinical outcome in patients with symptomatic or asymptomatic carotid artery stenosis who underwent CAS or CEA. We classified articles as model development studies if the authors reported the development of a prediction model in their objectives or conclusions, or if it was clear from other information in the article that they developed a prediction model for individual risk estimation (e.g. if they presented a risk score). Articles were included if they reported models for predicting periprocedural or postprocedural risk of vascular events (stroke, myocardial information, or death). In addition, the reported models had to include at least one patient characteristic that is known before a procedure is initiated and that can easily be retrieved from a patient’s medical record (e.g. demographics, vascular risk factors, comorbidities).

isk of vascular events (stroke, myocardial information, or death). In addition, the reported models had to include at least one patient characteristic that is known before a procedure is initiated and that can easily be retrieved from a patient’s medical record (e.g. demographics, vascular risk factors, comorbidities). We excluded studies of patients with carotid artery stenosis who received medical treatment only or who underwent combined procedures for coronary and carotid artery stenosis. In addition, we excluded studies that used laboratory, imaging, or procedural characteristics as predictors in the final prediction model only. However, we did include studies that used both patient characteristics and characteristics of additional examinations or procedural characteristics in the final prediction model. Screening process One author (EJV) screened all potentially eligible studies on title and abstract and subsequently reviewed full-text copies of the selected publications. Uncertainty about eligibility assessment was resolved with discussion between two authors (EJV and JPG). All prognostic factor studies that were identified through our search were screened full text to assess whether the authors developed a prediction model that met our inclusion criteria. References and related citations of the studies selected for critical appraisal were screened for additional relevant model development and external validation studies.

were identified through our search were screened full text to assess whether the authors developed a prediction model that met our inclusion criteria. References and related citations of the studies selected for critical appraisal were screened for additional relevant model development and external validation studies. Critical appraisal and data extraction We designed a standardised form for critical appraisal and data extraction based on items used in the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies.5 Data on the following items were extracted for each study: study design, data source, recruitment period, characteristics of study population, outcome type, number of study participants and outcome events, modelling method, predictors used in the final model, model performance, and method of internal validation. If a study described multiple prediction models, data were extracted separately for each model. For studies describing an external validation of a prediction model (i.e. assessment of a model’s predictive accuracy in a different study population), we extracted the type of external validation (e.g. temporal, geographical), whether the external validation was performed by the same investigators who developed the model, characteristics of external validation study population, geographical location, number of study participants and outcome events, and model performance.

xtracted the type of external validation (e.g. temporal, geographical), whether the external validation was performed by the same investigators who developed the model, characteristics of external validation study population, geographical location, number of study participants and outcome events, and model performance. Data extraction and critical appraisal was performed by one author (EJV); in case of doubt this was resolved with discussion between two authors (EJV, JPG). Statistical analysis Results on modelling methods, model presentation, and predictors used in the final models were summarised with descriptive statistics. We did not perform a meta-analysis of predictive performance of the included prediction models, because predictors and predicted outcomes differed between the models. To get an impression on the average effect size of the individual predictors, we performed a random-effects meta-analysis of all effect estimates (odds/risk/hazard ratios) for predictors that were used in at least 25% of prediction models for short-term outcome after CAS, short-term outcome after CEA or long-term outcome after CAS or CEA. Subsequently, we classified them into strong (≥2.5 or <0.4), moderate (1.5 to <2.5 or 0.4 to <0.7), or weak (1.1 to <1.5 or 0.7 to <0.9) predictors. All results in this study are reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses statement.6

To get an impression on the average effect size of the individual predictors, we performed a random-effects meta-analysis of all effect estimates (odds/risk/hazard ratios) for predictors that were used in at least 25% of prediction models for short-term outcome after CAS, short-term outcome after CEA or long-term outcome after CAS or CEA. Subsequently, we classified them into strong (≥2.5 or <0.4), moderate (1.5 to <2.5 or 0.4 to <0.7), or weak (1.1 to <1.5 or 0.7 to <0.9) predictors. All results in this study are reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses statement.6 Results Literature search and study selection We identified 6366 unique studies in PubMed and Embase with our systematic search (Supplementary Figure 1). After screening titles and abstracts, 103 studies remained. Thirty-six model development studies met our inclusion criteria after full-text screening. Screening references and related citations of these studies revealed three additional articles (one model development study and two external validation studies).

ure 1). After screening titles and abstracts, 103 studies remained. Thirty-six model development studies met our inclusion criteria after full-text screening. Screening references and related citations of these studies revealed three additional articles (one model development study and two external validation studies). The 37 model development studies that we selected for critical appraisal and data extraction described 46 unique prediction models: 11 models in patients who underwent CAS, 34 models in patients who underwent CEA, and one model in both patients who underwent CAS or CEA (Supplementary Table 2). Five of these studies developed more than one prediction model that met our inclusion criteria. Thirty-six models predicted short-term outcome (≤30 days after the procedure), and 10 models predicted long-term outcome (>30 days after the procedure) after CAS or CEA. Table 1 summarises the modelling methods separately for models that predict short- and long-term outcome. Table 1. Reported methods of prediction models for short-and long-term outcome.

erm outcome (≤30 days after the procedure), and 10 models predicted long-term outcome (>30 days after the procedure) after CAS or CEA. Table 1 summarises the modelling methods separately for models that predict short- and long-term outcome. Table 1. Reported methods of prediction models for short-and long-term outcome. Short-term outcome Long-term outcome All models Total n = 36 Total n = 10 Total n = 46 Data source Single centre 7 (19%) 5 (50%) 12 (26%) Multicentre 26 (72%) 5 (50%) 31 (67%) Trial 1 (3%) 0 (0%) 1 (2%) Other 2 (6%) 0 (0%) 2 (4%) Publication year, median (range) 2010 (1993–2016) 2014 (2012–2016) 2012 (1993–2016) Most commonly predicted outcomes Stroke or death 12 (33%) 0 (0%) 12 (26%) Mortality 2 (6%) 7 (70%) 9 (20%) Stroke, myocardial infarction, or death 6 (17%) 1 (10%) 7 (15%) Stroke 3 (8%) 0 (0%) 3 (7%) Other 13 (36%) 2 (20%) 15 (33%) Most common period of follow-up Postprocedural in-hospital 8 (22%) 0 (0%) 8 (17%) Seven days after procedure 2 (6%) 0 (0%) 2 (4%) 30 days after procedure 26 (72%) 0 (0%) 26 (57%) One year after procedure 0 (0%) 2 (20%) 2 (4%) Three years after procedure 0 (0%) 3 (30%) 3 (7%) Five years after procedure 0 (0%) 2 (20%) 2 (4%) Other 0 (0%) 3 (30%) 3 (7%) Presentation of prediction model Regression coefficients 14 (39%) 1 (10%) 15 (33%) Risk score 2 (6%) 1 (10%) 3 (7%) Both 20 (56%) 8 (80%) 28 (61%) Modelling method Logistic regression 35 (97%) 2 (20%) 37 (80%) Cox proportional hazards 0 (0%) 8 (80%) 8 (17%) Other 1 (3%) 0 (0%) 1 (2%) Shrinkage of predictor weights Uniform 9 (25%) 3 (30%) 12 (26%) Other 1 (3%) 0 (0%) 1 (2%) Any 10 (28%) 3 (30%) 13 (28%) None 26 (72%) 7 (70%) 33 (72%) Internally validated modelsa Bootstrapping 10 (28%) 3 (30%) 13 (28%) Cross-validation 7 (19%) 2 (20%) 9 (20%) Split sample 3 (8%) 0 (0%) 3 (7%) Any 16 (44%) 5 (50%) 21 (46%) None 20 (56%) 5 (50%) 25 (54%) Externally validated models Temporal 4 (11%) 0 (0%) 4 (9%) Geographical 2 (6%) 1 (10%) 3 (7%) Fully external 3 (8%)b 2 (20%)c 5 (11%) Any 9 (25%) 3 (30%) 12 (26%) None 27 (75%) 7 (70%) 34 (74%) Results are presented as numbers (%), unless stated otherwise. Models for short-term outcome predict risk ≤30 days after the procedure; models for long-term outcome predict risk >30 days after the procedure.

(10%) 3 (7%) Fully external 3 (8%)b 2 (20%)c 5 (11%) Any 9 (25%) 3 (30%) 12 (26%) None 27 (75%) 7 (70%) 34 (74%) Results are presented as numbers (%), unless stated otherwise. Models for short-term outcome predict risk ≤30 days after the procedure; models for long-term outcome predict risk >30 days after the procedure. a Multiple internal validation methods may be used for one model. b Three prediction models were externally validated in a different paper. c Type of external validation unclear for one model.7 Data source Most models were developed with data from multiple institutions (31 models; 67%). For the majority of these 31 models, data from large registries or health care databases were used (26 models; 57%). One model was developed with randomised trial data. Predicted outcomes In total, the 46 models predicted 16 different single and composite outcomes. For short-term prediction models, the most commonly predicted outcome was a composite of stroke or death (12 models; 33%); for long-term prediction models this was mortality (seven models; 70%) (Table 1). The other 12 predicted outcomes included myocardial infarction (two models; 4%), and composites of neurological and cardiac complications, such as transient ischaemic attack and congestive heart failure (Supplementary Table 2).

els; 33%); for long-term prediction models this was mortality (seven models; 70%) (Table 1). The other 12 predicted outcomes included myocardial infarction (two models; 4%), and composites of neurological and cardiac complications, such as transient ischaemic attack and congestive heart failure (Supplementary Table 2). Predictors In total, the models used 309 predictors (Figure 1, Supplementary Table 3). Median number of predictors per model was 7 (range 2–14). For short-term models, the most common predictors were age (67 versus 56% for CAS versus CEA), diabetes mellitus (33 versus 37%), history of TIA or stroke (22 versus 37%), and symptomatic status (44 versus 19%). Other common predictors for short-term outcome after CAS were anatomical characteristics (e.g. tortuous carotid arterial system or ulceration of carotid plaque on angiogram; 67%) and procedural characteristics (e.g. procedural hemodynamic depression or long duration of procedure; 56%). Other common predictors for short-term outcome after CEA were female sex (22%), ischaemic heart disease (26%), heart failure (41%), coronary artery disease (22%), renal insufficiency or dialysis (22%), and contralateral stenosis ≥ 50% or occlusion (33%). For long-term models, the most common predictors were age (100%), smoking (30%), diabetes mellitus (80%), heart failure (50%), chronic obstructive pulmonary disease or dyspnoea (60%), renal insufficiency or dialysis (50%), symptomatic status (20%), contralateral stenosis ≥ 50% or occlusion (50%), and statin use (30%). Symptomatic status was used as predictor in 11 (34%) of the 32 prediction models that were developed in both symptomatic and asymptomatic patients. Conflicting results regarding the direction of the predictor effect were found for female sex, high blood pressure, prior coronary intervention, ipsilateral carotid stenosis ≥70%, and antiplatelet use, but these predictors were only included in a few prediction models. Figure 1. Predictors used per type of prediction model. Models for short-term outcome predict risk ≤30 days after the procedure; models for long-term outcome predict risk >30 days after procedure. CABG: coronary artery bypass graft; CAS: carotid artery stenting; CEA: carotid endarterectomy; COPD: chronic obstructive pulmonary disease; PCI: percutaneous coronary intervention; TIA: transient ischaemic attack.

ict risk ≤30 days after the procedure; models for long-term outcome predict risk >30 days after procedure. CABG: coronary artery bypass graft; CAS: carotid artery stenting; CEA: carotid endarterectomy; COPD: chronic obstructive pulmonary disease; PCI: percutaneous coronary intervention; TIA: transient ischaemic attack. Supplementary Table 3 shows the pooled effect size for each predictor with classification of its strength. We observed weak to moderate predictors for short-term outcome after CEA and long-term outcomes, but strong predictors for short-term outcome after CAS. Anatomical and procedural characteristics are the strongest predictors for short-term outcome after CAS. Sample size The number of patients used to develop the prediction models ranged from 221 to 49,411 (Table 2). Number of events per candidate variable could be calculated for 13 (28%) models; the total number of candidate predictors was unclear for the remainder. Median number of events per candidate variable was 10.6 (range 1.9–52.3). Table 2. Results of all developed and externally validated prediction models.

,411 (Table 2). Number of events per candidate variable could be calculated for 13 (28%) models; the total number of candidate predictors was unclear for the remainder. Median number of events per candidate variable was 10.6 (range 1.9–52.3). Table 2. Results of all developed and externally validated prediction models. CAS CEA CAS and CEA All models Short-term outcome Long-term outcome Short-term outcome Long-term outcome Long-term outcome Results not reported (%) Developed models Total n = 9 Total n = 2 Total n = 27 Total n = 7 Total n = 1 Total n = 46 Total n = 46 Number of patientsa 606 (221–11,122) 460 (317–602) 6553 (741–49,411) 2001 (291–4114) 506 3095 (221–49,411) 0 (0%) Proportion of events 4.3% (2.4–11.5) 24.5% (13.9–35.0) 3.0% (0.5–6.9) 14.2% (6.7–26.8) 13.8% 4.1% (0.5–35.0) 0 (0%) Age in years 71.5 (66.7–76.3) 70.8 (70.5–71.0) 70.6 (62.5–74.5) 70.0 (67.3–74.2) 69.0 70.6 (62.5–76.3) 7 (15%) Proportion of males 63.4% (61.0–74.0) 72.8% (67.6–77.9) 59.1% (39.7–95.4) 60.6% (56.7–78.7) 62.3% 61.0% (39.7–95.4) 3 (7%) Proportion of symptomatic patients 36.0% (5.1–100) 40.4% 43.9% (0–100) 25.3% (0–84.0) 0 40.9% (0–100) 6 (13%) c-statisticb 0.73 (0.69–0.94) 0.73 (0.66–0.79) 0.68 (0.58–0.74) 0.72 (0.69–0.74) NR 0.71 (0.58–0.94) 11 (24%) External validations Total n = 1 Total n = 2 Total n = 11c Total n = 0 Total n = 1 Total n = 15 Total n = 15 Number of patients 1544 345 (137–552) 1998 (134–71,222) NA 352 1026 (134–71,222) 0 (0%) Proportion of events NR 29.0% (18.9–39.0) 3.2% (0.7–22.1) NA NR 3.3% (0.7–39.0) 4 (27%) c-statistic 0.68 0.68 (0.66–0.69) 0.65 (0.55–0.72) NA NR 0.66 (0.55–0.72) 3 (20%) CAS: carotid artery stenting; CEA: carotid endarterectomy; NA: not applicable; NR: not reported.

52) 1998 (134–71,222) NA 352 1026 (134–71,222) 0 (0%) Proportion of events NR 29.0% (18.9–39.0) 3.2% (0.7–22.1) NA NR 3.3% (0.7–39.0) 4 (27%) c-statistic 0.68 0.68 (0.66–0.69) 0.65 (0.55–0.72) NA NR 0.66 (0.55–0.72) 3 (20%) CAS: carotid artery stenting; CEA: carotid endarterectomy; NA: not applicable; NR: not reported. Results are presented as median (range), unless stated otherwise. Models for short-term outcome predict risk ≤30 days after the procedure; models for long-term outcome predict risk >30 days after the procedure. a For some models only the total number of procedures was reported. b Optimism-adjusted c-statistic was used in case this was reported. c Three prediction models for short-term outcome after CEA were externally validated twice. Modelling methods Most prediction models for short-term outcome were developed with logistic regression analysis (35 models; 97%) and those for long-term outcome with Cox proportional hazards analysis (eight models; 80%). Out of all 46 prediction models, 21 (46%) models were internally and 12 (26%) models were validated at least once externally (Table 1). In 25 models (54%), the authors did not report how they handled missing predictor or outcome data. In the remaining 21 models, complete case analysis was performed in 15, imputation was performed in three, and a combination of these methods was performed in three models. Twenty-eight (61%) of all models reported the full regression formula with intercept or baseline hazard, or the complete risk score with classification of risk according to this score.

case analysis was performed in 15, imputation was performed in three, and a combination of these methods was performed in three models. Twenty-eight (61%) of all models reported the full regression formula with intercept or baseline hazard, or the complete risk score with classification of risk according to this score. Twenty-one (46%) newly developed models selected candidate predictors for the initial multivariable model based on univariable associations with the outcome. Only 12 (26%) models included all candidate predictors or chose variables based on previous literature or clinical experience. Model presentation Supplementary Table 4 shows which performance measures were reported for newly developed models and for external validations of these models. The c-statistic was the most often reported discrimination measure, both for newly developed models and for external validations. The Hosmer–Lemeshow test was the most often reported calibration measure for developed models. Predictive performance Newly developed models Discriminative performance in terms of the c-statistic ranged from 0.58 to 0.94 for all newly developed models (Table 2). Median c-statistic was higher for models that predict short- or long-term outcome after CAS than after CEA. The highest c-statistics were reported for models that predict short-term outcome after CAS and included anatomical and procedural characteristics, and the lowest c-statistics for models that predict short-term outcome after CEA.

c was higher for models that predict short- or long-term outcome after CAS than after CEA. The highest c-statistics were reported for models that predict short-term outcome after CAS and included anatomical and procedural characteristics, and the lowest c-statistics for models that predict short-term outcome after CEA. External validation studies In 13 articles, 15 external validations were performed. The majority of the 46 developed models (34 models; 74%) has never been validated externally. Of the 12 models that were validated externally, nine (75%) were validated once, and three models were validated twice (Supplementary Table 2). Out of the 12 externally validated models, nine (75%) were solely externally validated in the same paper in which their development was described, and three (25%) were validated by independent researchers.8–10 The number of patients in the 15 external validation cohorts ranged from 134 to 71,222 (Table 2). The c-statistic ranged from 0.55 to 0.72 and was consistently lower than the corresponding c-statistic in the model development cohort, except for the model that was validated in the same study population in which the original model was developed.9

the 15 external validation cohorts ranged from 134 to 71,222 (Table 2). The c-statistic ranged from 0.55 to 0.72 and was consistently lower than the corresponding c-statistic in the model development cohort, except for the model that was validated in the same study population in which the original model was developed.9 Discussion This systematic review provides an overview of prediction models for short- and long-term outcome that are developed for patients who underwent carotid revascularisation for symptomatic or asymptomatic ICA stenosis. Age, diabetes mellitus, and contralateral carotid stenosis ≥50% or occlusion were most often used as predictors in all models. For short-term outcome models, symptomatic status, anatomical characteristics, and procedural characteristics were most important predictors after CAS; heart failure and history of stroke or TIA were most important predictors after CEA. For long-term outcome models after CAS or CEA, heart failure, COPD or dyspnoea, and renal insufficiency or dialysis were most important predictors. Predictive performance of the prediction models in terms of discrimination was in general poor to moderate. Only a minority of these prediction models was validated internally or externally. Methodological quality was low for the majority of the prediction models due to incomplete presentation and methodological limitations.

tive performance of the prediction models in terms of discrimination was in general poor to moderate. Only a minority of these prediction models was validated internally or externally. Methodological quality was low for the majority of the prediction models due to incomplete presentation and methodological limitations. From the results of our systematic review, we can draw conclusions on most common predictors and their strength. However, predictors that were less often used in the included prediction models cannot be designated as unimportant, because not all studies included the same candidate predictors when they developed their model. Discriminative performance of the prediction models differed between models developed in CAS and CEA patients, but was in general better for models developed in CAS patients. Specifically for short-term outcome models after CAS, addition of anatomical and procedural characteristics improved predictive performance; c-statistics were higher in short-term outcome models with anatomical and procedural characteristics after CAS compared with models without such characteristics, whereas this was not the case for short-term outcome models after CEA. Highest c-statistic (0.94) was reported for a risk score that predicted risk of a composite of stroke, death, or myocardial infarction within 30 days after CAS.11 However, the risk of overfitting was high for this model because of the low number of patients and the lack of internal or external validation.

ls after CEA. Highest c-statistic (0.94) was reported for a risk score that predicted risk of a composite of stroke, death, or myocardial infarction within 30 days after CAS.11 However, the risk of overfitting was high for this model because of the low number of patients and the lack of internal or external validation. The prediction models included in our systematic review have several methodological and statistical shortcomings that hamper assessment of clinical usefulness in other study populations. First, most models were developed in large study populations from registries or health care databases; such studies may have a higher risk of bias due to missing outcome data and the lack of consistent definitions of predictors and outcomes. Moreover, patients included in these study populations were selected for CAS or CEA by their treating physicians based on vascular risk factors, comorbidities, and clinical experience. In general, multicentre prospective cohort studies with consistent definitions and more accurate event registration are preferred when developing a new prediction model.12 Second, 20 (43%) of all prediction models were published more than five years ago (before 2012), while improved medical therapy with antiplatelets or anticoagulants, treatment of vascular risk factors, and healthy lifestyle changes have probably lowered short- and long-term risk of vascular events in carotid stenosis patients in more recent years. Furthermore, information on regression coefficients and model intercepts or baseline hazards was often lacking, which makes it difficult to externally validate the prediction model. Regarding model presentation, most studies reported the Hosmer-Lemeshow test as measure for calibration, whereas calibration is preferably reported graphically with a calibration plot.13 Also, most studies did not report a measure for precision of the predictive performance, such as a confidence interval. Last, risk of overfitting is high for most prediction models included in our systematic review due to selection of candidate predictors based on statistical testing, lack of shrinkage and internal or external validation, and small sample sizes.

for precision of the predictive performance, such as a confidence interval. Last, risk of overfitting is high for most prediction models included in our systematic review due to selection of candidate predictors based on statistical testing, lack of shrinkage and internal or external validation, and small sample sizes. A major strength of our study is that this is the first systematic review of prediction models for short- and long-term outcome after CAS or CEA. In addition, we performed a comprehensive systematic search and used a standardised form for critical appraisal and data extraction of the selected studies. Nevertheless, our systematic review has some limitations. First, we included prediction models that also used procedural characteristics, whereas we aimed to identify models that predict clinical outcome based on preprocedural patient characteristics. However, the models with procedural characteristics also included preprocedural patient characteristics and thus contributed to our analysis of most frequently used predictors. Second, we used a broad outcome (vascular events) which may limit direct comparison of predictive performance of the included prediction models. Yet, our systematic review provides a comprehensive overview of all published prediction models after CAS or CEA. Third, we separately extracted data from each prediction model in case one study described multiple models (e.g. for different outcome types) or in case multiple studies used data from the same source and time period; this may have influenced our analysis of most important predictors and the meta-analysis of predictor strength. Fourth, some studies may not have reported all results in case of poor predictive performance of their developed prediction model (e.g. c-statistic close to 0.5), which is referred to as selective reporting bias. Selective reporting of predictive performance may have caused overestimation of the median c-statistics in our systematic review. Last, we could not retrieve full-text copies of three studies,14–16 so we may have missed some prediction models.

c-statistic close to 0.5), which is referred to as selective reporting bias. Selective reporting of predictive performance may have caused overestimation of the median c-statistics in our systematic review. Last, we could not retrieve full-text copies of three studies,14–16 so we may have missed some prediction models. We found it difficult to evaluate clinical usefulness of most of the prediction models included in our systematic review because of incomplete reporting and lack of validation. A new prediction model should be applied to other study populations to assess generalisability before the model can be used in daily practice.17 In addition, all but one of the included prediction models predicted outcome after either CAS or CEA, so a common use of one of these models to predict risk after both revascularisation procedures is not possible. Hence, we cannot recommend the use of one specific prediction model for outcome prediction after CAS or CEA from our systematic review. Therefore, we are planning to conduct an external validation study of prediction models included in this systematic review. In 2015, the TRIPOD statement was introduced, which contains a set of recommendations for the reporting of prediction modelling studies.13 Use of this statement for the reporting of model development studies may facilitate assessment of methodological quality and clinical usefulness, and may enable external validation of newly developed prediction models by other researchers.

ns a set of recommendations for the reporting of prediction modelling studies.13 Use of this statement for the reporting of model development studies may facilitate assessment of methodological quality and clinical usefulness, and may enable external validation of newly developed prediction models by other researchers. Conclusion Many prediction models or risk scores for short- and long-term outcome after carotid revascularisation have been developed. Usefulness of most models remains limited because of methodological shortcomings, incomplete presentation, and lack of external validation. External validation of these models is necessary before they can be used to aid physicians in predicting clinical outcome and choosing an appropriate treatment in daily practice. Supplementary Material Supplementary material Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: JPG received funding from the Rudolf Magnus Young Talent Fellowship (University Medical Centre Utrecht). Ethical approval Not applicable. Informed consent Not applicable. Guarantor JPG.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: JPG received funding from the Rudolf Magnus Young Talent Fellowship (University Medical Centre Utrecht). Ethical approval Not applicable. Informed consent Not applicable. Guarantor JPG. Contributorship EJV, AA, and JPG designed the study plan. EJV and JPG performed the systematic review, critical appraisal, and data extraction. All authors were involved in data-analysis and interpretation of the data. EJV wrote the first version of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

Introduction In 2015, over 26,000 patients were admitted to hospitals because of ischemic stroke in the Netherlands.1 Most of these patients need rehabilitation to achieve better recovery in the first months after stroke and reduce long-term disability. In the Netherlands, around 8% of all stroke patients is referred to an inpatient rehabilitation centre.2 Typically, these patients are too disabled to be discharged home, but they are cognitively and physically fit enough to participate in intensive therapy sessions and have sufficient social support to return home within two to four months. Alternatively, patients may be referred to skilled nursing and geriatric rehabilitation facilities. These patients are often elderly, suffer from comorbidities and have a poorer functional prognosis. Still, the majority of stroke patients (60%) is discharged home, mostly with community rehabilitation.2 Discharge planning may depend on multiple factors such as comorbidities and contextual factors (e.g. the presence of a healthy caregiver and premorbid level of functioning). The importance of the contextual factors increases as the functional prognosis of the stroke decreases. Therefore, early prediction of functional outcome may contribute to efficient discharge planning.

as comorbidities and contextual factors (e.g. the presence of a healthy caregiver and premorbid level of functioning). The importance of the contextual factors increases as the functional prognosis of the stroke decreases. Therefore, early prediction of functional outcome may contribute to efficient discharge planning. The most widely used functional outcome measure in acute stroke is the modified Rankin Scale (mRS). The mRS measures the degree of disability in daily activities. It is scored on an ordinal scale ranging from 0 (no symptoms) to 6 (death).3 Another frequently used outcome measure in rehabilitation is the Barthel Index (BI), measuring performance in 10 basic activities of daily living (ADL).4 BI is associated with duration of hospital stay.5 Previous studies identified many prognostic factors for outcome (measured by BI or mRS) after acute stroke.6 Prognostic factors can be combined in a model to identify patients at risk for poor outcome.7 Although several prognostic models exist to predict outcome in stroke, very few are adequately validated for use in daily clinical practice.8 We aimed to develop and validate a prognostic score for disability (BI) at discharge and functional outcome (mRS) at three months after acute ischemic stroke based on clinical information available on admission.

s exist to predict outcome in stroke, very few are adequately validated for use in daily clinical practice.8 We aimed to develop and validate a prognostic score for disability (BI) at discharge and functional outcome (mRS) at three months after acute ischemic stroke based on clinical information available on admission. Methods Derivation cohort Data from the Paracetamol (Acetaminophen) In Stroke (PAIS) study were used for model development.9 PAIS was a multicentre, randomised placebo-controlled phase III trial assessing the effect of high dose paracetamol on the functional outcome in patients with acute stroke. In short, patients were eligible for inclusion if they were diagnosed with acute ischemic stroke or intracerebral hemorrhage, had a prestroke mRS < 2 and study treatment could be started within 12 h after onset of symptoms. We used data of all patients with ischemic stroke included in PAIS.

patients with acute stroke. In short, patients were eligible for inclusion if they were diagnosed with acute ischemic stroke or intracerebral hemorrhage, had a prestroke mRS < 2 and study treatment could be started within 12 h after onset of symptoms. We used data of all patients with ischemic stroke included in PAIS. Outcome measures We used the BI at discharge as the outcome measure for short-term disability. The BI is an ordinal scale used to measure performance in ADL. The scale ranges from 0 to 20, with higher scores indicating a greater likelihood of being able to carry out ADL independently.4 In PAIS, the BI was measured at 14 days after enrolment or at hospital discharge if this occurred earlier (70% of the patients stayed for ≥3 days).9 However, choice of the optimal rehabilitation route mostly depends on more than just discharge outcome.10 Therefore, we additionally evaluated functional outcome at three months with the mRS. The mRS is an ordinal scale used to measure the degree of disability in daily activities and ranges from 0 (no symptoms) to 6, with mRS 5 indicating severe disability and mRS 6 indicating death.3

on more than just discharge outcome.10 Therefore, we additionally evaluated functional outcome at three months with the mRS. The mRS is an ordinal scale used to measure the degree of disability in daily activities and ranges from 0 (no symptoms) to 6, with mRS 5 indicating severe disability and mRS 6 indicating death.3 Model development To identify predictors of disability and functional outcome, we selected variables that were clinically relevant and/or previously reported to predict outcome after stroke in the literature.6 These variables were sex, age, National Institutes of Health Stroke Scale (NIHSS) score, diabetes, previous stroke, atrial fibrillation and hypertension. All predictors were entered into multivariable ordinal regression with backward selection with p < 0.2 for inclusion, separately for BI at discharge and mRS at three months. The final associations were presented as a set of odds ratios (ORs) and 95% confidence intervals (CIs) to indicate the individual predictor effects. ORs from an ordinal logistic regression model can be interpreted as a common OR for shifting over the full outcome range.11

ly for BI at discharge and mRS at three months. The final associations were presented as a set of odds ratios (ORs) and 95% confidence intervals (CIs) to indicate the individual predictor effects. ORs from an ordinal logistic regression model can be interpreted as a common OR for shifting over the full outcome range.11 The resulting models, the Dutch Stroke Score (DSS) for BI at discharge (‘DSS-discharge’) and mRS at three months (‘DSS-3 months’), were internally validated using standard bootstrapping procedures to avoid an optimistic estimate of the model performance, which often occurs when model performance is only evaluated directly in the derivation cohort (apparent validation). In the bootstrap procedure, random samples are drawn from the original sample, each with the same number of patients as the original sample. In each of these samples the modeling steps are repeated and the resulting models are subsequently evaluated on the original sample. The mean model performance in all 500 bootstrap models represents the expected performance of the models in future, similar patients.12

the same number of patients as the original sample. In each of these samples the modeling steps are repeated and the resulting models are subsequently evaluated on the original sample. The mean model performance in all 500 bootstrap models represents the expected performance of the models in future, similar patients.12 Validation cohorts For external validation, we used data from the PRomoting ACute Thrombolysis in Ischemic StrokE (PRACTISE) study and Preventive Antibiotics in Stroke Study (PASS). PRACTISE was a cluster-randomised trial designed to evaluate an implementation strategy to increase the proportion of patients treated with intravenous thrombolysis.13 PRACTISE registered adult patients with acute stroke admitted within 24 h after onset of symptoms and had no age restrictions. We used data from ischemic stroke patients admitted within 4 h as in these patients detailed clinical data were available. PASS was a multicentre, randomised, open-label trial designed to assess whether or not preventive antimicrobial therapy with ceftriaxone improves functional outcome in patients with acute stroke.14 PASS included adult patients with clinical symptoms of a stroke (ischemic or hemorrhagic) admitted within 24 h after symptom onset. We used data of all patients with ischemic stroke included in PASS.

er or not preventive antimicrobial therapy with ceftriaxone improves functional outcome in patients with acute stroke.14 PASS included adult patients with clinical symptoms of a stroke (ischemic or hemorrhagic) admitted within 24 h after symptom onset. We used data of all patients with ischemic stroke included in PASS. Model validation The validity of the DSS-3 months was assessed in terms of discrimination and calibration. The external validation cohorts did not have data on BI at discharge. Discrimination refers to how well the model distinguishes between those who have good outcome (mRS 0–2) vs. those who have poor outcome (mRS 3–6) at three months. Discrimination was assessed by calculating the ordinal area under the curve (AUC) of the receiver operating characteristic (ROC) curve.15 The AUC ranges from 0.5 for non-informative models to 1.0 for perfect models.12 Calibration indicates the agreement between predicted and observed probabilities. Calibration was assessed graphically in a calibration graph, and expressed as the calibration slope and an intercept. The calibration slope is ideally equal to 1 and describes the effect of the predictors in the validation cohort versus in the derivation cohort. The intercept indicates whether predictions are systematically too high or too low, and should ideally be zero.12

aph, and expressed as the calibration slope and an intercept. The calibration slope is ideally equal to 1 and describes the effect of the predictors in the validation cohort versus in the derivation cohort. The intercept indicates whether predictions are systematically too high or too low, and should ideally be zero.12 At external validation, the discriminative power of a model may be influenced by differences in predictor effects, but also by differences in distribution of patient characteristics (case-mix) between the derivation and validation cohort.16 In a more homogeneous population, discrimination between patients with good vs. poor outcome is more difficult than in a heterogeneous population. To take this into account, we calculated the case-mix-corrected AUC. The case-mix-corrected AUC reflects the discriminative power of a model, assuming that the regression coefficients are correct for the validation population. It was calculated by simulating new outcome values for all patients in the validation dataset, based on the predicted risks for each patient.16 After external validation, we fitted the DSS-3 months on the combined data of all three trials to get the best estimates for the regression coefficients.17 The DSS-discharge and DSS-3 months were presented in a score chart, as a score plot simplified to five BI and mRS outcome classes (based on clinically relevant cutoffs), and as formulas to calculate the predicted outcomes.

he combined data of all three trials to get the best estimates for the regression coefficients.17 The DSS-discharge and DSS-3 months were presented in a score chart, as a score plot simplified to five BI and mRS outcome classes (based on clinically relevant cutoffs), and as formulas to calculate the predicted outcomes. All statistical analyses were performed using R software, version 3.3.2 (R foundation for statistical computing, Vienna, Austria). The calibration plots were created with an updated version of the val.prob function (rms library in R). Missing values in the development and validation cohorts were statistically imputed using a multiple imputation method exploiting correlations between predictor variables and between predictor variables and the outcome variables (mice function in R). Complete case analyses were done for comparison with the imputed analyses.

ng values in the development and validation cohorts were statistically imputed using a multiple imputation method exploiting correlations between predictor variables and between predictor variables and the outcome variables (mice function in R). Complete case analyses were done for comparison with the imputed analyses. Results Study population For model development, we included 1227 patients with ischemic stroke from the PAIS trial. Missing data on hypertension (3.1%) were statistically imputed; all other baseline variables and outcomes were complete. For the external validation of the model predicting mRS at three months, we included, 1657 ischemic stroke patients from the PRACTISE study. Sixty-eight patients with missing data on mRS at three months were excluded, resulting in an external validation sample of 1589 patients. Other missing data (0.6%) were statistically imputed. Additionally, we externally validated the model for functional outcome at three months in, 2125 ischemic stroke patients from the PASS study. Eighteen patients with missing data on the mRS at three months were excluded, resulting in an external validation sample of 2107 patients. Other missing data (0.4%) were statistically imputed.

we externally validated the model for functional outcome at three months in, 2125 ischemic stroke patients from the PASS study. Eighteen patients with missing data on the mRS at three months were excluded, resulting in an external validation sample of 2107 patients. Other missing data (0.4%) were statistically imputed. In all three studies, most patients (55–58%) were male and the mean age was around 70 years (Table 1). The three populations are comparable concerning baseline characteristics, except for time from stroke onset to inclusion (PAIS and PRACTISE had a smaller time window compared to PASS), previous stroke (33% in PASS vs. 20% in the other trials) and diabetes (20% in PASS vs. 15–17% in PAIS and PRACTISE). The number of patients with poor outcome (mRS 3–6) was lower in PASS compared to PAIS and PRACTISE (online supplemental Figure 1(a)). In PAIS, this is reflected in the substantial proportion of patients with favorable outcome on the BI at discharge (online supplemental Figure 1(b)). Figure 1. Calibration plots of the DSS-3 months in (a) PRACTISE and (b) PASS. DSS: Dutch Stroke Score; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study. Table 1. Baseline characteristics of the included patients from the PAIS, PRACTISE and PASS studies.

In all three studies, most patients (55–58%) were male and the mean age was around 70 years (Table 1). The three populations are comparable concerning baseline characteristics, except for time from stroke onset to inclusion (PAIS and PRACTISE had a smaller time window compared to PASS), previous stroke (33% in PASS vs. 20% in the other trials) and diabetes (20% in PASS vs. 15–17% in PAIS and PRACTISE). The number of patients with poor outcome (mRS 3–6) was lower in PASS compared to PAIS and PRACTISE (online supplemental Figure 1(a)). In PAIS, this is reflected in the substantial proportion of patients with favorable outcome on the BI at discharge (online supplemental Figure 1(b)). Figure 1. Calibration plots of the DSS-3 months in (a) PRACTISE and (b) PASS. DSS: Dutch Stroke Score; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study. Table 1. Baseline characteristics of the included patients from the PAIS, PRACTISE and PASS studies. PAIS (n = 1227) PRACTISE (n = 1589) PASS (n = 2107) Male sex 675 (55%) 872 (55%) 1212 (58%) Age in years (mean, sd) 70.1 (13.4) 70.6 (13.4) 71.9 (12.5) Time from onset to CT in hours (median, IQR) 3.0 (1.8–5.9) 2.0 (1.4–3.0) NA NIHSS (median, IQR) 6.0 (3.0–11.0) 5.0 (3.0–12.0) 5.0 (3.0–9.0) Diabetes mellitus 181 (15%) 266 (17%) 423 (20%) Previous ischemic stroke 245 (20%) 318 (20%) 698 (33%) Atrial fibrillation 190 (16%) 290 (18%) 326 (16%) Hypertension 601 (49%)a 811 (51%) 1154 (55%) Current smoking 380 (31%) 374 (24%) 524 (25%) NIHSS: National Institutes of Health Stroke Scale; IQR: interquartile range; NA: not available; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study; PAIS: Paracetamol (Acetaminophen) In Stroke.

(49%)a 811 (51%) 1154 (55%) Current smoking 380 (31%) 374 (24%) 524 (25%) NIHSS: National Institutes of Health Stroke Scale; IQR: interquartile range; NA: not available; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study; PAIS: Paracetamol (Acetaminophen) In Stroke. a38 Missings. Model development in PAIS The relation between age as a continuous variable and the log odds of disability (BI) in the development data was non-linear and intensified when age was above 60 years (online supplemental Figure 2). Because of this non-linearity, we considered different age effects for patients older vs. younger than 60 years. Figure 2. DSS score charts simplified to five outcome classes of the (a) BI at discharge and (b) mRS at three months. Legend of (a): Dark red=0, Red=1–9, Orange=10–14, Yellow=15–18, Green=19–20 and legend of (b): Dark red=6, Red=4–5, Orange=3, Yellow=2, Green=0–1. DSS: Dutch Stroke Score; BI: Barthel Index; mRS: modified Rankin Scale.

Figure 2. DSS score charts simplified to five outcome classes of the (a) BI at discharge and (b) mRS at three months. Legend of (a): Dark red=0, Red=1–9, Orange=10–14, Yellow=15–18, Green=19–20 and legend of (b): Dark red=6, Red=4–5, Orange=3, Yellow=2, Green=0–1. DSS: Dutch Stroke Score; BI: Barthel Index; mRS: modified Rankin Scale. Of the variables considered, age per decade above 60, NIHSS per point and diabetes were the strongest predictors of BI at discharge, both in univariable (data not shown) and multivariable analysis (Table 2) and were included in the model for disability at discharge. The internally validated ordinal AUC was 0.76 (95%CI 0.75–0.79). Age per decade above 60, NIHSS per point, diabetes, previous stroke and atrial fibrillation were the strongest predictors of mRS at three months, both in univariable (data not shown) and multivariable analysis (Table 2) and were included in the final model for mRS at three months. The internally validated ordinal AUC was 0.75 (95%CI 0.74–0.77). Table 2. Associations of predictors in multivariable ordinal regression with lower BI at discharge in PAIS and higher mRS at three months in in PAIS, PRACTISE and PASS.

Of the variables considered, age per decade above 60, NIHSS per point and diabetes were the strongest predictors of BI at discharge, both in univariable (data not shown) and multivariable analysis (Table 2) and were included in the model for disability at discharge. The internally validated ordinal AUC was 0.76 (95%CI 0.75–0.79). Age per decade above 60, NIHSS per point, diabetes, previous stroke and atrial fibrillation were the strongest predictors of mRS at three months, both in univariable (data not shown) and multivariable analysis (Table 2) and were included in the final model for mRS at three months. The internally validated ordinal AUC was 0.75 (95%CI 0.74–0.77). Table 2. Associations of predictors in multivariable ordinal regression with lower BI at discharge in PAIS and higher mRS at three months in in PAIS, PRACTISE and PASS. PAIS (n = 1227) BI at discharge PAIS (n = 1227) mRS at three months PRACTISE (n = 1589)mRS at three months PASS (n = 2107)mRS at three months Variable OR (95%CI) P value OR (95%CI) P value OR (95%CI) P value OR (95%CI) P value Male sex 1.01 (0.79–1.23) 0.923 0.87 (0.71–1.07) 0.189 0.81 (0.67–0.97) 0.022 0.77 (0.61–0.93) 0.002 Age per decadeif over 60a,b 1.55 (1.41–1.68) <0.001 1.86 (1.64–2.12) <0.001 1.80 (1.61–2.01) <0.001 1.55 (1.41–1.70) <0.001 Age per decadeif under 60 1.07 (0.83–1.30) 0.589 0.93 (0.76–1.15) 0.514 0.89 (0.74–1.07) 0.211 0.70 (0.56–0.86) <0.001 NIHSS per pointa,b 1.24 (1.22–1.26) <0.001 1.19 (1.17–1.22) <0.001 1.19 (1.17–1.21) <0.001 1.21 (1.19–1.23) <0.001 Diabetesa 1.62 (1.32–1.91) 0.002 1.87 (1.40–2.51) <0.001 1.70 (1.34–2.17) <0.001 1.31 (1.11–1.51) 0.007 Previous strokeb 1.18 (0.91–1.45) 0.225 1.67 (1.29–2.16) <0.001 1.59 (1.27–1.99) <0.001 1.14 (0.98–1.31) 0.111 Atrial fibrillationb 1.09 (0.78–1.39) 0.592 1.41 (1.05–1.89) 0.022 1.24 (0.98–1.57) 0.076 1.14 (0.91–1.36) 0.264 Hypertension 1.06 (0.84–1.28) 0.594 1.02 (0.83–1.26) 0.844 1.08 (0.90–1.30) 0.384 0.91 (0.75–1.07) 0.246 BI: Barthel Index; mRS: modified Rankin Scale; OR: odds ratio; CI: confidence interval; NIHSS: National Institutes of Health Stroke Scale; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study; PAIS: Paracetamol (Acetaminophen) In Stroke.

0) 0.384 0.91 (0.75–1.07) 0.246 BI: Barthel Index; mRS: modified Rankin Scale; OR: odds ratio; CI: confidence interval; NIHSS: National Institutes of Health Stroke Scale; PRACTISE: PRomoting ACute Thrombolysis in Ischemic StrokE; PASS: Preventive Antibiotics in Stroke Study; PAIS: Paracetamol (Acetaminophen) In Stroke. aParameter included in final model on BI at discharge. bParameter included in final model on mRS at three months. External validation in PRACTISE and PASS In PRACTISE, the DSS-3 months had an ordinal AUC of 0.74 and an AUC for the cutoff mRS ≥ 3 of 0.81 (95%CI 0.81–0.84) (online Supplemental Table 1). The model predicted 49.4% poor outcome (mRS ≥ 3); whereas the observed probability of poor functional outcome was 45.2%. The calibration slope was 1.022 and the intercept was −0.238, indicating that the model’s predictions of poor outcome were systematically higher than the observed probability of poor outcome (Figure 1(a)). In PASS, the DSS-3 months had an ordinal AUC of 0.69 and an AUC for the cutoff mRS ≥ 3 of 0.81 (95%CI 0.81–0.83) (online Supplemental Table 1). The predicted probability of poor outcome was 48.6%, compared to an observed probability of poor functional outcome of 38.5%. The calibration slope was 1.058 and the intercept was −0.555, indicating that the model’s predictions of poor outcome were systematically too high (Figure 1(b)). This overestimation was higher than in PRACTISE. The internal and external validation in the complete cases (PAIS n = 1227, PRACTISE n = 1581, PASS n = 2098) yielded similar results (not shown).

In PASS, the DSS-3 months had an ordinal AUC of 0.69 and an AUC for the cutoff mRS ≥ 3 of 0.81 (95%CI 0.81–0.83) (online Supplemental Table 1). The predicted probability of poor outcome was 48.6%, compared to an observed probability of poor functional outcome of 38.5%. The calibration slope was 1.058 and the intercept was −0.555, indicating that the model’s predictions of poor outcome were systematically too high (Figure 1(b)). This overestimation was higher than in PRACTISE. The internal and external validation in the complete cases (PAIS n = 1227, PRACTISE n = 1581, PASS n = 2098) yielded similar results (not shown). The lower discriminative ability of the DSS-3 months in the external validation cohorts was largely explained by a less heterogeneous case-mix compared to the development cohort. This is illustrated by small differences between the development AUC and case-mix-corrected AUCs (online Supplemental Table 1). The lower discriminative ability in PASS compared to PAIS and PRACTISE was due to both case-mix and differences in predictor effects (relatively large difference between AUC in external validation and case-mix-corrected AUC in PASS). The final DSS-3 months was developed on the combined data of all three cohorts (n = 4923). The model had an ordinal AUC of 0.73 and an AUC for the cutoff mRS ≥3 of 0.81 (95%CI 0.81–0.83) (online Supplemental Table 1).

The lower discriminative ability of the DSS-3 months in the external validation cohorts was largely explained by a less heterogeneous case-mix compared to the development cohort. This is illustrated by small differences between the development AUC and case-mix-corrected AUCs (online Supplemental Table 1). The lower discriminative ability in PASS compared to PAIS and PRACTISE was due to both case-mix and differences in predictor effects (relatively large difference between AUC in external validation and case-mix-corrected AUC in PASS). The final DSS-3 months was developed on the combined data of all three cohorts (n = 4923). The model had an ordinal AUC of 0.73 and an AUC for the cutoff mRS ≥3 of 0.81 (95%CI 0.81–0.83) (online Supplemental Table 1). The final models are presented as the DSS score chart (Table 3, and simplified to five outcome classes in Figure 2), with higher scores indicating worse outcome. For example, a patient of 70 years with an NIHSS of 13 and a history of previous stroke and diabetes has a DSS-discharge score of 8 and a predicted probability of 17% for BI 19–20 at discharge and a DSS-3 months score of 13 and a predicted probability of 76% for mRS ≥ 3 at three months (online Appendix 1). Table 3. DSS score chart based on ordinal analysis of the BI and mRS. A higher score indicates a worse outcome (lower predicted BI and higher mRS).

The final models are presented as the DSS score chart (Table 3, and simplified to five outcome classes in Figure 2), with higher scores indicating worse outcome. For example, a patient of 70 years with an NIHSS of 13 and a history of previous stroke and diabetes has a DSS-discharge score of 8 and a predicted probability of 17% for BI 19–20 at discharge and a DSS-3 months score of 13 and a predicted probability of 76% for mRS ≥ 3 at three months (online Appendix 1). Table 3. DSS score chart based on ordinal analysis of the BI and mRS. A higher score indicates a worse outcome (lower predicted BI and higher mRS). Variable Points for predicting BI at discharge Points for predicting mRS score at 3 months Age <60 0 0 60–70 1 2 70–80 2 4 80–90 3 6 90+ 4 8 NIHSS 0 0 0 1–4 1 1 5–15 5 5 16–20 10 10 21–42 15 15 Diabetes 1 2 Previous stroke – 2 Atrial fibrillation – 1 Total 0–20 0–28 BI: Barthel Index; mRS: modified Rankin Scale; NIHSS: National Institutes of Health Stroke Scale; DSS: Dutch Stroke Score. Discussion We propose the DSS, consisting of two simple prediction models for disability (BI) at discharge and functional outcome (mRS) at three months after acute ischemic stroke based on clinical information available on admission. The DSS-discharge consists of three variables: age per decade above 60 years, NIHSS per point and diabetes. The DSS-3 months additionally includes previous stroke and atrial fibrillation. Both models showed reasonable performance in internal and external validation.

based on clinical information available on admission. The DSS-discharge consists of three variables: age per decade above 60 years, NIHSS per point and diabetes. The DSS-3 months additionally includes previous stroke and atrial fibrillation. Both models showed reasonable performance in internal and external validation. Relation with previous literature Previously, several models to estimate the probability of unfavourable outcome after stroke have been developed, with a high variability in endpoints, time between symptom onset and assessment of the variables, and patient populations. Literature reviews have shown that many of these prediction models have methodological shortcomings that limit their use for early discharge planning. For instance, assessment of predictors multiple days after stroke onset18,19 and the use of a dichotomous outcome such as mortality.20–26 In addition, previously developed models were not validated, and hence their use in clinical practice is limited.8,27 One tool has been developed specifically to predict unfavorable discharge destination from the hospital stroke unit. Functional disability, poor sitting balance, depression, cognitive disability and old age were identified as predictors of poor discharge outcome.10 However, this model was only applicable for decision-making at 7–10 days post stroke. Moreover, this study had some methodological shortcomings, including dichotomisation of predictors, a small sample size and dichotomisation of the outcome.

ability and old age were identified as predictors of poor discharge outcome.10 However, this model was only applicable for decision-making at 7–10 days post stroke. Moreover, this study had some methodological shortcomings, including dichotomisation of predictors, a small sample size and dichotomisation of the outcome. Implications of study findings Prediction models in acute stroke are useful to inform patients and relatives on prognosis and identify patients at risk for poor outcome before treatment decisions are made.7 On population level, prediction models can be used for adjustment when comparing quality of care for stroke patients across institutions. Additionally, prediction models could be relevant in design and analysis of randomised controlled trials, e.g. for covariate adjustment.28,29 Further, prediction of functional outcome may contribute to discharge planning. If functional outcome is expected to be poor, contextual factors, such as housing circumstances, financial problems and whether or not a patient is living alone, become more important.

controlled trials, e.g. for covariate adjustment.28,29 Further, prediction of functional outcome may contribute to discharge planning. If functional outcome is expected to be poor, contextual factors, such as housing circumstances, financial problems and whether or not a patient is living alone, become more important. We developed the DSS to be used by stroke unit nurses during the first day after admission. In clinical practice, the NIHSS is mostly scored shortly after the administration of alteplase. Therefore, we did not add treatment with alteplase as a covariable to our analysis. Recently, intra-arterial treatment administered within six hours after stroke onset has been shown beneficial in patients with a proximal intracranial arterial occlusion.30 However, the majority (90%) of acutely admitted ischemic stroke patients still receives intravenous alteplase as only treatment. Therefore, the DSS is potentially suitable for use in present neurovascular practice. To facilitate discharge planning in endovascular-treated patients, a next step could be to update the models by including treatment (thrombolysis, thrombectomy or both) as a predictor. Moreover, no imaging or laboratory tests are required for clinicians to be able to use the DSS, which allows bedside use of the models early after admission. The DSS score chart can be easily incorporated in clinical practice since it consists of a few readily obtainable clinical variables at admission. Stroke unit nurses will be able to score all variables, including the NIHSS, provided that they are well trained and certified.

side use of the models early after admission. The DSS score chart can be easily incorporated in clinical practice since it consists of a few readily obtainable clinical variables at admission. Stroke unit nurses will be able to score all variables, including the NIHSS, provided that they are well trained and certified. The DSS-discharge still needs to be externally validated to give reliable estimates on model performance and study generalisability. At external validation, the discriminative ability of the DSS-3 months was generally lower than in the development sample. Discrimination was better in PRACTISE compared to PASS, both for the ordinal analysis of the mRS and for three different cutoffs of the mRS (online Supplemental Table 1). These higher AUCs were partly explained by differences in case-mix, as reflected in the case-mix-corrected AUCs. In addition, the predictor effects were slightly stronger in PRACTISE than in PASS. These differences in regression coefficients were most evident for diabetes and previous stroke, and could be explained by discrepancies in predictor definitions. For instance, in PASS, previous stroke comprised both Transient Ischemic Attack (TIA) and ischemic stroke, while in PRACTISE only ischemic stroke was considered. This implicates that the DSS-3 months is valid, but the definitions of the predictors should be identical to those in the development cohort.

tor definitions. For instance, in PASS, previous stroke comprised both Transient Ischemic Attack (TIA) and ischemic stroke, while in PRACTISE only ischemic stroke was considered. This implicates that the DSS-3 months is valid, but the definitions of the predictors should be identical to those in the development cohort. The reasonable discriminative ability of the DSS-3 months was associated with an overall overestimation of the probability of poor outcome. This overestimation was higher in PASS compared to PRACTISE, which might be due to the difference in outcome distribution between these cohorts (lower proportion of patients with poor outcome in PASS). This difference is most likely caused by the exclusion of patients with imminent death and neurological deterioration in PASS. The overestimation of the probability of poor outcome implies that the DSS-3 months needs updating (e.g. adjustment of the intercept (recalibration)) before it is suitable for individualised predictions in clinical practice.

caused by the exclusion of patients with imminent death and neurological deterioration in PASS. The overestimation of the probability of poor outcome implies that the DSS-3 months needs updating (e.g. adjustment of the intercept (recalibration)) before it is suitable for individualised predictions in clinical practice. Strengths and limitations Strengths of this study are the internal and (partial) external validation of the DSS, and the large size of the development and two independent validation cohorts. Even though many models have been developed for prediction of outcome after stroke, the large sample size and the aim of contributing to efficient discharge planning makes that our study has added value compared to already existing evidence. Also, we predicted outcomes over the whole range from no symptoms to death. Furthermore, we used two well-known and widely implemented outcome measures for functional outcome in our models. The BI is a reliable and valid scale to measure ADL.31 Since discharge destination (partially) depends on the patient’s ability to carry out ADL, the BI is a suitable outcome for our model. Additionally, we selected potential predictors based on the literature and clinical knowledge. This is preferred over selection based on the data as the latter may result in overfitting (model perfect for the development data but performing poor in new patients).12 The robustness of our approach is represented in the reasonable performance of the models in internal and external validation.

nd clinical knowledge. This is preferred over selection based on the data as the latter may result in overfitting (model perfect for the development data but performing poor in new patients).12 The robustness of our approach is represented in the reasonable performance of the models in internal and external validation. Several limitations of our study need to be considered. We included only hospitalised patients with an ischemic stroke in our analysis. Consequently, our chart does not apply for patients with intracerebral hemorrhage. Further, the development and validation cohorts originated from randomised controlled trials conducted in the Netherlands, potentially limiting the generalisability of the chart. To evaluate the performance of the models beyond the Dutch setting, external validation in observational data from settings with a different healthcare system configuration is necessary. However, the Dutch stroke population is representative for stroke populations in developed countries. Moreover, our external validation cohorts consist of unselected, prospectively included patients, originating from hospitals representative in size, geographic distribution and frequency of stroke treatment procedures. We were able to externally validate the DSS-3 months, but not the DSS-discharge as no data on BI at discharge were available. Also, discharge policy is variable between and within different healthcare systems, which makes it a difficult outcome for prediction purposes. However, these differences in discharge timing resemble the variation in clinical practice. Additionally, in the field of rehabilitation, predicting functional outcome in terms of the mRS has limitations. Important aspects that can contribute to the level of disability and the need for rehabilitation (e.g. pain, communication, cognition) are not entirely covered by the mRS.32 However, the mRS is a widely used outcome measure in stroke management.

tation, predicting functional outcome in terms of the mRS has limitations. Important aspects that can contribute to the level of disability and the need for rehabilitation (e.g. pain, communication, cognition) are not entirely covered by the mRS.32 However, the mRS is a widely used outcome measure in stroke management. The prognostic performance of the DSS after validation could be classified as satisfactory. This does not disqualify the usefulness of the models for clinical practice, because in general, multivariable prediction models are able to incorporate and accurately weigh more factors than a human mind.33 Nevertheless, the results should always be regarded as a mere recommendation and should be placed in the context of the personal circumstances, needs and wishes of the patient. Other factors that are worth considering when planning patients’ discharge are the presence of social support, cognitive disability, the therapeutic needs of the patient and the expected future residence destination (e.g. home or nursing facility). Conclusion The DSS has satisfactory performance in predicting BI at discharge and mRS at three months in ischemic stroke patients. If further validated, the DSS may contribute to efficient stroke unit discharge planning alongside patients’ contextual factors (e.g. social support, housing circumstances and cognitive disability) and therapeutic needs.

tory performance in predicting BI at discharge and mRS at three months in ischemic stroke patients. If further validated, the DSS may contribute to efficient stroke unit discharge planning alongside patients’ contextual factors (e.g. social support, housing circumstances and cognitive disability) and therapeutic needs. Supplementary Material Supplementary material Acknowledgements The Dutch Stroke Score was developed by researchers from the Erasmus MC Rotterdam and validated in data provided by researchers from the Erasmus MC Rotterdam and AMC Amsterdam. The authors wish to thank the investigators and patients participating in the PAIS, PRACTISE and PASS trials. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: SD receives revenue from the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI).

Supplementary Material Supplementary material Acknowledgements The Dutch Stroke Score was developed by researchers from the Erasmus MC Rotterdam and validated in data provided by researchers from the Erasmus MC Rotterdam and AMC Amsterdam. The authors wish to thank the investigators and patients participating in the PAIS, PRACTISE and PASS trials. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: SD receives revenue from the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI). Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is funded by the Foundation for Neurovascular Research Rotterdam (SNOR, Stichting Neurovasculair Onderzoek Rotterdam). The PAIS study was funded by the Dutch Heart Foundation (grant number 2002B148). The PRACTISE study was funded by the Netherlands Organization for Health Research and Development (ZON-MW, grant number 945–14-217). The PASS study was funded by the Netherlands Organization for Health Research and Development (ZON-MW, grant numbers 171002302 and 016116358), the Netherlands Heart Foundation (grant number 2009B095), and the European Research Council (ERC Starting Grant). Informed consent Not applicable. Ethical approval All trial protocols were approved by the local ethics committees of the corresponding centres. Guarantor DD.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is funded by the Foundation for Neurovascular Research Rotterdam (SNOR, Stichting Neurovasculair Onderzoek Rotterdam). The PAIS study was funded by the Dutch Heart Foundation (grant number 2002B148). The PRACTISE study was funded by the Netherlands Organization for Health Research and Development (ZON-MW, grant number 945–14-217). The PASS study was funded by the Netherlands Organization for Health Research and Development (ZON-MW, grant numbers 171002302 and 016116358), the Netherlands Heart Foundation (grant number 2009B095), and the European Research Council (ERC Starting Grant). Informed consent Not applicable. Ethical approval All trial protocols were approved by the local ethics committees of the corresponding centres. Guarantor DD. Contributorship DD, ES and HF conceived and supervised the study. IR, SD and HF were involved in data analysis. IR and SD wrote the manuscript. MS, HH, MD, WW, PN, DB and GR provided critical revision of the manuscript for intellectual content. All authors approved the final version of the manuscript. IR and SD contributed equally to this article. Trial registration PAIS study – Netherlands Trial Register: NTR2365. PRACTISE study – ISRCTN registry: ISRCTN20405426. PASS study – ISRCTN registry: ISRCTN66140176.

Introduction Stroke is one of the major global causes of mortality and disability. Each year, 10 million people are estimated to have a first ever stroke, and 6.5 million people die as a result of stroke.1 In the United Kingdom, stroke is the third leading cause of years of life lost and of disability-adjusted life years.2 This human burden is mirrored by the very large cost of providing healthcare to people with stroke, with stroke care accounting for approximately 3–5% of all healthcare expenditure.3,4 Many health systems are faced with the challenge of tackling the rising costs of healthcare services, and it is now increasingly recognised that cost is a fundamental dimension of healthcare quality. Cost reduction is one of the components of the Institute of Healthcare Improvement’s ‘Triple Aim’5 and some have argued that the concept of quality in healthcare should be reframed in terms of ‘value’: health outcomes achieved per unit of expenditure.6 Although registries have been established in many settings7–9 to measure stroke care and outcomes after stroke, they typically do not include information about healthcare costs and are not specifically designed to measure the value that healthcare services provide.

e’: health outcomes achieved per unit of expenditure.6 Although registries have been established in many settings7–9 to measure stroke care and outcomes after stroke, they typically do not include information about healthcare costs and are not specifically designed to measure the value that healthcare services provide. The Sentinel Stroke National Audit Programme (SSNAP) is the national stroke register of England, Wales and Northern Ireland, collecting data on an estimated 95% of all patients admitted to hospital with acute stroke.10 It provides individualised data and analytics about the quality of stroke care to hospitals, rehabilitation services, payers (‘Commissioners’ in the terminology of the National Health Service), stroke survivors and policy makers. The aims of this study were to estimate patient-level health and social care costs of stroke for all patients admitted to hospital with stroke in England, Wales and Northern Ireland and integrate this alongside near real-time reporting of SSNAP data on quality of care and patient outcomes. Methods The study was carried out in two phases (Figure 1). Firstly, a health economic simulation was developed to produce patient-level estimates of the health and social care cost of acute stroke, taking account of patient’s age, stroke type, stroke severity and sex. Secondly, these estimates were integrated into SSNAP, and interactive tools were developed to model the health economic consequences of increasing access to intravenous thrombolysis and early supported discharge. Figure 1. Design of the study.

roke, taking account of patient’s age, stroke type, stroke severity and sex. Secondly, these estimates were integrated into SSNAP, and interactive tools were developed to model the health economic consequences of increasing access to intravenous thrombolysis and early supported discharge. Figure 1. Design of the study. Health economic simulation A detailed description of the health economic methods is provided in the online Appendix and is summarised below. The model was a time-to-event individual patient simulation model.11 In this model, patients go through different treatment units for healthcare (Table 1 and Figure A1, online Appendix). Their health status, stroke-related resource use and costs and outcomes are modelled through the simulation, up to five years after stroke. Table 1. Summary of methods used in the model.

lation model.11 In this model, patients go through different treatment units for healthcare (Table 1 and Figure A1, online Appendix). Their health status, stroke-related resource use and costs and outcomes are modelled through the simulation, up to five years after stroke. Table 1. Summary of methods used in the model. Treatment unit Health state change Length of stay Mortality Next destinations ASU and thrombolysis Thrombolysed patients: mRS when leaving ASU was dependent on NIHSS 24 h after thrombolysis and age group Not thrombolysed patients – mRS when leaving ASU was dependent on NIHSS when admitted to hospital and age group Generalised linear model with family of gamma and log link • Age • Sex • Thrombolysed patients – NIHSS after thrombolysis • Not thrombolysed patients – NIHSS when admitted to hospital • Discharge destination Bounded at 90th percentile of the SSNAP data: 36.0 days Mortality probability in ASU was dependent on age group and mRS and was different in thrombolysed and not thrombolysed patients. All causes mortality was used for both groups. ESD SU CRT Discharge with no need for rehabilitation GMWs No health state change involved in GMW due to the lack of data and relatively short period of time in GMW Fixed average length of stay No mortality in the model. ASU SU – inpatient rehabilitation Health state change was dependent on age group and mRS when entering SU Length of stay was sampled from a lognormal distribution with the parameters dependent on the next destination (including death), age group and mRS when entering SU Bounded at 90th percentile of the SSNAP data: 70.8 days Dependent on age and mRS on arrival at SU ESD CRT Discharge with no need for rehabilitation ESD Health state change is dependent on age group and mRS when entering ESD Length of stay was sampled from an exponential distributions with the parameters dependent on age and mRS when entering ESD Bounded at 90th percentile of the SSNAP data: 63.1 days Dependent on age and mRS when arrived to ESD CRT Discharged to own home or nursing home CRT Health state change is dependent on age group and mRS when admitted by a CRT Length of stay with CRT is determined by a discrete event algorithm Survival curve was fitted dependent on age, sex, mRS at discharge and stroke type (Cox regression) Discharged to own home or care home Discharge to own home or care home No health state change after discharge (unless they have a recurrence).

d by a CRT Length of stay with CRT is determined by a discrete event algorithm Survival curve was fitted dependent on age, sex, mRS at discharge and stroke type (Cox regression) Discharged to own home or care home Discharge to own home or care home No health state change after discharge (unless they have a recurrence). Admission to care home only included if newly institutionalised after stroke Length of stay at own home/care home after discharge is determined by a discrete event algorithm Survival curve is dependent on age, sex, mRS at discharge and stroke type (Cox regression) Patients might have stroke recurrence after discharge Stroke recurrence Recurrence severity is measured in NIHSS, dependent on recurrence type, independent from age and previous severity Same pathway as first stroke (dependent on age and severity) Same pathway as first stroke Same pathway as first stroke ASU: Acute stroke unit; mRS: modified Rankin scale; ESD: early supported discharge; SU: stroke unit; CRT: community rehabilitation team; NIHSS: National Institutes of Health Stroke Scale, or NIH Stroke Scale; GMW: general medical ward.

ndent on age and severity) Same pathway as first stroke Same pathway as first stroke ASU: Acute stroke unit; mRS: modified Rankin scale; ESD: early supported discharge; SU: stroke unit; CRT: community rehabilitation team; NIHSS: National Institutes of Health Stroke Scale, or NIH Stroke Scale; GMW: general medical ward. The principal model parameters were: age, sex, stroke type (ischaemic stroke and primary intracerebral haemorrhage) and stroke severity. Health status as well as other parameters including mortality, next event and length of stay in each treatment unit were probabilistically updated at the end of each treatment unit, depending on their characteristics when entering the unit. Health states were measured by the National Institute of Health Stroke Score (NIHSS; a measure of stroke severity) at stroke onset and then subsequently by the modified Rankin Score (mRS; a measure of functional status). The change in mRS from before to after the stroke was used to estimate change in disability caused by the stroke and attribute changes in health and social care utilisation to the stroke itself rather than pre-existing or co-morbid illnesses. Health-Related Quality of Life (health utility on a 0–1 scale) was estimated by mapping from the mRS.12

d to provide data to model survival after stroke, stroke recurrence and to estimate long-term health and social care utilisation after stroke. Resource use was estimated based on the parameters in Table 1, which were combined with unit costs (online Appendix) to estimate the costs generated by individuals in the model. Cost estimates took a health and social care perspective (healthcare is free for users of the National Health Service in the United Kingdom but social care may incur charges). Unit costs for events and treatment units were derived from existing cost sources (online Appendix) and were not discounted due to the relatively short time-horizon used. Quality-adjusted life-years (QALYs), the primary health outcomes, were estimated by weighting survival with health utility. Health cost estimates included the costs of pre-hospital care, acute care, diagnostics, prescribing, inpatient rehabilitation, community rehabilitation, early supported discharge, primary care, secondary prevention and stroke recurrence. Social care included nursing home care, formal care at home, supported meals and day services. The population mean of these costs and QALY were then estimated according to 80 different combinations of baseline characteristics (age, sex, stroke type and stroke severity).

care, secondary prevention and stroke recurrence. Social care included nursing home care, formal care at home, supported meals and day services. The population mean of these costs and QALY were then estimated according to 80 different combinations of baseline characteristics (age, sex, stroke type and stroke severity). Baseline results were estimated to reflect the current situation of stroke care. Inputs estimated from SSNAP were used directly for this purpose. Scenario analyses were conducted to estimate the cost-effectiveness results of increasing the proportion of patients treated with intravenous thrombolysis or early supported discharge (online Appendix). The treatment effects of thrombolysis and early supported discharge were then calibrated to the results of Cochrane reviews of randomised controlled trials of these interventions,14,15 by comparing models using treatment effects from the Cochrane reviews to those obtained using the real-world treatment effects observed in SSNAP. Treatment effects reported by the Cochrane reviews were used to optimise the inputs of the model. Mean healthcare cost, mean social care cost and mean QALYs over all stroke patients at one and five years after primary stroke were plotted and the line of best fit calculated to estimate the cost savings and QALYs gained for each extra person thrombolysed or discharged to ESD (online Appendix). Probabilistic sensitivity analysis (PSA) was conducted using Monte Carlo simulation to assess uncertainty of our results and estimate 95% confidence intervals for the base case results.

it calculated to estimate the cost savings and QALYs gained for each extra person thrombolysed or discharged to ESD (online Appendix). Probabilistic sensitivity analysis (PSA) was conducted using Monte Carlo simulation to assess uncertainty of our results and estimate 95% confidence intervals for the base case results. Statistical analyses were carried out with R and Stata 13, and the health economic model was built in MS Excel VBA (2010). Integration into SSNAP All patients in SSNAP were assigned one of the model subgroups based on sex, age group, stroke severity and stroke type. The data were then linked with a table of cost estimates from the health economics model to generate individualised cost estimates for all patients in SSNAP. Total and average costs were integrated into reports alongside SSNAP outputs describing quality of care metrics and outcomes (survival, mRS at discharge, new institutionalisation). SSNAP outputs incorporating health economic data include executive summaries designed for hospital executive boards and dashboards designed for commissioners. The results of the model were illustrated for one year of SSNAP data (n = 84,184 patients admitted between April 2015 and March 2016), plotting five-year health and social care cost against age, stroke severity and stroke type.

summaries designed for hospital executive boards and dashboards designed for commissioners. The results of the model were illustrated for one year of SSNAP data (n = 84,184 patients admitted between April 2015 and March 2016), plotting five-year health and social care cost against age, stroke severity and stroke type. Ethics Data analyses were carried out using fully anonymised datasets from SSNAP and the SLSR. SSNAP has approval from the Clinical Advisory Group of the NHS Health Research Authority to collect patient-level data under section 251 of the NHS Act 2006. All patients or their relatives gave written informed consent to participate in the SLSR. The design of the SLSR was approved by the ethics committees of Guy's and St Thomas' National Health Service Foundation Trust, King's College Hospital Foundation Trust, St George's University Hospital, National Hospital for Nervous Diseases, and Westminster Hospital. Results The mean per-patient cost of healthcare attributable to stroke was £13,452 in the first year after stroke and £17,963 after five years (Table 2). Mean social care costs attributable to stroke were lower than healthcare costs in the first year (£8977) but by five years after stroke social care costs accounted for the larger burden of cost (£28,076). Total mean costs were £22,429 by one year and £46,039 at five years. Table 2. Baseline mean costs per patient at one and five years.

costs attributable to stroke were lower than healthcare costs in the first year (£8977) but by five years after stroke social care costs accounted for the larger burden of cost (£28,076). Total mean costs were £22,429 by one year and £46,039 at five years. Table 2. Baseline mean costs per patient at one and five years. 1 Year 5 Year Mean healthcare costs per patienta £13,452 £17,963 Mean social care costs per patientb £8977 £28,076 Mean total health and social care costs per patient £22,429 £46,039 Combined total cost for all patients included in SSNAP, April 2015–March 2016 (n = 84,184) £1,736,338,300 £3,604,672,200 Mean health and social costs per patient with ischaemic stroke £20,121 £41,432 Mean health and social costs per patient with ICH stroke £24,297 £52,726 ICH: intracerebral hemorrhage; SSNAP: Sentinel Stroke National Audit Programme. a Healthcare costs include: ambulance, MRI or CT scan, thrombolysis, acute stroke unit care, rehabilitation stroke unit care, general medical ward care, community rehabilitation, GP visits, secondary prevention, and ESD therapists. b Social care costs include: care home, home help, meals on wheels, and social service day centre visits. In England, Wales and Northern Ireland, there were 84,184 patients admitted with stroke between April 2015 and March 2016 and entered onto SSNAP. Of these, 73,318 (87.1%) had ischaemic stroke, 10,267 (12.2%) had ICH, 599 (0.7%) had undetermined stroke type. Total health and social care costs attributable to stroke for this cohort of patients were £1.74 billion at one year and £3.60 billion at five years.

between April 2015 and March 2016 and entered onto SSNAP. Of these, 73,318 (87.1%) had ischaemic stroke, 10,267 (12.2%) had ICH, 599 (0.7%) had undetermined stroke type. Total health and social care costs attributable to stroke for this cohort of patients were £1.74 billion at one year and £3.60 billion at five years. The estimated stroke cost varied widely according to patient characteristics, ranging from £7322 to £44,854 at one year and £19,101 to £107,336 at five years. Costs increased with age and with stroke severity (Figure 2). Mean costs were £4000 higher for ICH stroke than for ischaemic stroke at one year and £11,000 higher by five years (Figure 3) Figure 2. Five-year health and social costs by age and stroke severity, for all patients (n = 84,184) admitted between April 2015 and March 2016. Each dot on the scatter plot is one patient with stroke; the best fit line is a restricted cubic spline with four knots. Figure 3. Five-year health and social cost age and stroke type, for all patients (n = 84,184) admitted between April 2015 and March 2016. Each dot on the scatter plot is one patient with stroke; the best fit line is a restricted cubic spline with four knots.

The estimated stroke cost varied widely according to patient characteristics, ranging from £7322 to £44,854 at one year and £19,101 to £107,336 at five years. Costs increased with age and with stroke severity (Figure 2). Mean costs were £4000 higher for ICH stroke than for ischaemic stroke at one year and £11,000 higher by five years (Figure 3) Figure 2. Five-year health and social costs by age and stroke severity, for all patients (n = 84,184) admitted between April 2015 and March 2016. Each dot on the scatter plot is one patient with stroke; the best fit line is a restricted cubic spline with four knots. Figure 3. Five-year health and social cost age and stroke type, for all patients (n = 84,184) admitted between April 2015 and March 2016. Each dot on the scatter plot is one patient with stroke; the best fit line is a restricted cubic spline with four knots. Scenario analysis We found that increasing the proportion of patients receiving thrombolysis or early supported discharge was cost saving, particularly for long-term social care costs. For thrombolysis, we estimated healthcare savings of £3200 and £4000, social care savings of £2900 and £5300 and 0.08 and 0.26 QALYs gained in total for each extra patient thrombolysed after one and five years, respectively (Figure A2, online Appendix). For ESD, the cost savings for health and social care were estimated to be £1600 and £2400 after one year, £1600 and £8100 after five years, with QALY gains of 0.04 and 0.14 per at one and five years, respectively, for each extra patient discharged with ESD (Figure A3, online Appendix). In the sensitivity analyses, 95% confidence intervals were estimated for three scenarios: base case, 95% of the patients who met the SSNAP minimum criteria were thrombolysed and 35% of patients who were not discharged to ESD redirected to ESD treatment (online Appendix).

ischarged with ESD (Figure A3, online Appendix). In the sensitivity analyses, 95% confidence intervals were estimated for three scenarios: base case, 95% of the patients who met the SSNAP minimum criteria were thrombolysed and 35% of patients who were not discharged to ESD redirected to ESD treatment (online Appendix). We used these results to support quality improvement and service development in stroke care services by building interactive Excel tools for SSNAP participants (illustration of the thrombolysis tool in Figure 4). These enable SSNAP participants to compare their own data to the results of the health economic model and estimate the potential cost effectiveness of increasing access to thrombolysis and ESD in their patient cohorts. The tool allows participants to model different scenarios (such as increasing the number of patients arriving at hospital within the thrombolysis time window), allowing them to identify opportunities for improvement. Figure 4. Screenshot from the Thrombolysis tool.

cess to thrombolysis and ESD in their patient cohorts. The tool allows participants to model different scenarios (such as increasing the number of patients arriving at hospital within the thrombolysis time window), allowing them to identify opportunities for improvement. Figure 4. Screenshot from the Thrombolysis tool. Discussion The care of patients with stroke is expensive, both at the individual level and at the level of health systems and society as a whole. These costs vary considerably between patients, in the first five years after stroke ranging fivefold between patient subgroups with the lowest and highest care costs. Five-year costs were highest in patients with haemorrhagic strokes, older patients and those with more severe stroke. Costs were lowest in patients with ischaemic stroke and mild strokes. Costs over the first year after stroke largely arise from the costs of acute treatment, but over time the costs of providing social care account for a greater proportion of the total care costs of stroke. We found that both thrombolysis and early supported discharge services were both cost effective and cost saving, resulting in financial savings to both the healthcare and social care costs of stroke. Increasing the proportion of patients receiving these interventions therefore has the potential not only to improve patient outcomes after stroke but also help to control the high financial burden of stroke.

ffective and cost saving, resulting in financial savings to both the healthcare and social care costs of stroke. Increasing the proportion of patients receiving these interventions therefore has the potential not only to improve patient outcomes after stroke but also help to control the high financial burden of stroke. As far as we are aware, this is the first time that a detailed patient-level health economic model has been integrated into a large-scale stroke register. The development of a stratified model enabled patient-level cost estimates to take account of how costs vary according to patients’ age, stroke type, stroke severity and sex, rather than simply using an average across all patients. By integrating this model into an existing stroke register, we were able to extend SSNAP to report patient-level data on estimated costs alongside data about stroke care quality and outcomes. Although we have not yet evaluated the effectiveness and usefulness of feeding back cost data in this way, the methodology could be adopted by other stroke registries and has the potential to expand the role of registries in supporting quality improvement and research.

gside data about stroke care quality and outcomes. Although we have not yet evaluated the effectiveness and usefulness of feeding back cost data in this way, the methodology could be adopted by other stroke registries and has the potential to expand the role of registries in supporting quality improvement and research. These cost estimates are consistent with previous studies of the costs of stroke. Recent estimates of the costs in high-income countries with health systems similar to the United Kingdom have reported costs in the first year after stroke of €29,484 in the Netherlands16 and €21,200 in Sweden.17 A previous national analysis of stroke data in England in 2009 estimated a mean cost of £28,568 (in 2017 prices).18 Since that time there have been major changes in the organisation and delivery of stroke care in the United Kingdom, resulting in reduced length of stay, increase in the provision and use of stroke units and greater use of interventions which reduce long-term disability (e.g. thrombolysis)10 and recurrence (e.g. antiplatelet therapy). These improvements in care may have contributed to the reduction in the estimated costs of stroke care since 2009.

educed length of stay, increase in the provision and use of stroke units and greater use of interventions which reduce long-term disability (e.g. thrombolysis)10 and recurrence (e.g. antiplatelet therapy). These improvements in care may have contributed to the reduction in the estimated costs of stroke care since 2009. Collecting data on large cohorts of patients, typically at the regional or national level, registries are one of the major sources of ‘real world data’ about human health and healthcare services.19,20 It is too early to evaluate the impact of including data on health costs within a large-scale clinical registry, but we anticipate that these data will have several uses. For example, bespoke and clearly sourced estimates of costs could be used by clinicians and managers to make business cases for quality improvement or service development. Similarly, these data make it simpler to estimate the financial return on investment in spending money on stroke prevention, through for example detection and management of atrial fibrillation20 and in predicting the future financial burden of stroke care. These data could also be useful in the planning, design and conduct of observational and interventional research studies. For example, embedding randomised controlled trials (RCTs) within registries that include data on health economic outcomes could help researchers carry out cheaper and more useful RCTs of new interventions in stroke care21 and aid in the health economic evaluation of new interventions.

interventional research studies. For example, embedding randomised controlled trials (RCTs) within registries that include data on health economic outcomes could help researchers carry out cheaper and more useful RCTs of new interventions in stroke care21 and aid in the health economic evaluation of new interventions. Strengths and limitations The strength of this study was that it estimated a fuller range of costs than are usually available in healthcare payment or reimbursement systems – in the UK, for example, previous sources of stroke cost data included only inpatient care but not the longer term costs of rehabilitation, nursing and medical care. By making use of the detailed real-world stroke data about stroke collected through registries, we were able to model stroke care with a level of detail that has typically not been possible in previous studies of the cost of stroke.22

tient care but not the longer term costs of rehabilitation, nursing and medical care. By making use of the detailed real-world stroke data about stroke collected through registries, we were able to model stroke care with a level of detail that has typically not been possible in previous studies of the cost of stroke.22 There are, however, a number of limitations. Firstly, the model only takes a health and social care perspective and does not include the wider societal costs of stroke (such as costs to patients or family members through loss of income and informal care). Cost estimates were also only based on patients who had a full NIHSS score in SSNAP (76% out of the all stroke patients). In comparison with the entire patient population, these patients were slightly younger (median = 76 vs. 80) and slightly more independent (median mRS = 0 vs. 1), likely resulting in a bias towards lower cost estimates. Cost estimates were also based only on patients aged above 40 years due to insufficient data being available in the SLSR of patients aged less than 40 years to fit the models. In implementing these estimates in SSNAP, we therefore assumed that costs in younger patients are the same as for those aged 40–64 years. The model describes the typical care pathways experienced by the majority of patients admitted with acute stroke in the UK, but this is by necessity a simplification of the real world. For example, the model assumes that all patients are admitted to a stroke unit, and although this is true for 95% of stroke patients in England, Wales and Northern Ireland,10 the model does not take account of the 5% of patient which do not. It also does not take account of the impact of thrombectomy in patients with acute ischemic stroke. Although in the UK this is still not in widespread use and so is unlikely to have made a significant difference to current cost estimates, future estimates of the cost of stroke should take account of this emerging therapy, as it is likely to result in lower long-term costs.23 To gain more power from the data, we combined all SSNAP data collected from England, Wales, and Northern Ireland and applied the same unit costs to all patients. In the real world, costs might vary between these three countries, although the differences are likely to be small as their health systems are similar.

sts.23 To gain more power from the data, we combined all SSNAP data collected from England, Wales, and Northern Ireland and applied the same unit costs to all patients. In the real world, costs might vary between these three countries, although the differences are likely to be small as their health systems are similar. Conclusion We developed and implemented a detailed health economic model within a national stroke register. Not only does this help to provide more up to date estimates of the large financial costs arising from stroke in the United Kingdom, but provides new data about how costs vary between patients and a new source of information to help plan and improve stroke care services. In particular, these results imply that improving access to thrombolysis and early supported discharge services can contribute to reducing the financial burden of stroke on health and social care services. Health economic models are inevitably a simplification of the real world, and it is important to understand the assumptions that support the estimates from models such as this. Nonetheless, integrating the measurement and reporting of health economic outcomes data into clinical registries could help them become increasingly useful resources for quality improvement and research. Data sharing Data from SSNAP are available at www.strokeaudit.org. This includes extensive and unrestricted access to publicly available datasets and contact details to request access to anonymised patient-level data.

Conclusion We developed and implemented a detailed health economic model within a national stroke register. Not only does this help to provide more up to date estimates of the large financial costs arising from stroke in the United Kingdom, but provides new data about how costs vary between patients and a new source of information to help plan and improve stroke care services. In particular, these results imply that improving access to thrombolysis and early supported discharge services can contribute to reducing the financial burden of stroke on health and social care services. Health economic models are inevitably a simplification of the real world, and it is important to understand the assumptions that support the estimates from models such as this. Nonetheless, integrating the measurement and reporting of health economic outcomes data into clinical registries could help them become increasingly useful resources for quality improvement and research. Data sharing Data from SSNAP are available at www.strokeaudit.org. This includes extensive and unrestricted access to publicly available datasets and contact details to request access to anonymised patient-level data. Supplementary Material Supplementary material Acknowledgements This work uses data provided by patients and collected by the National Health Service as part of their care and support. We would like to thank the many hundreds of people and organisations participating in SSNAP, and on their behalf specifically thank the members of the SSNAP collaboration (www.strokeaudit.org/Research/SSNAP-Collaboration.aspx).

by patients and collected by the National Health Service as part of their care and support. We would like to thank the many hundreds of people and organisations participating in SSNAP, and on their behalf specifically thank the members of the SSNAP collaboration (www.strokeaudit.org/Research/SSNAP-Collaboration.aspx). The SLSR and CDAW are funded/supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London, and the NIHR Collaboration for Leadership in Applied Health Research (award number NIHR CLAHRC-2013-10022) and Care South London at King's College Hospital NHS Foundation Trust. Declaration of conflicting interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: AGR is the National Clinical Director for Stroke, NHS England. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was funded by a grant from NHS England. SSNAP is funded by the Healthcare Quality Improvement Partnership on behalf of NHS England. The SLSR and CDAW are funded/supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London, and the NIHR Collaboration for Leadership in Applied Health Research (award number NIHR CLAHRC-2013-10022) and Care South London at King's College Hospital NHS Foundation Trust.

Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London, and the NIHR Collaboration for Leadership in Applied Health Research (award number NIHR CLAHRC-2013-10022) and Care South London at King's College Hospital NHS Foundation Trust. Informed consent SSNAP has permission from the NHS Health Research Authority under section 251 of the Health and Social Care Act 2006 to collect patient data without prospective consent. Patients can opt out of data collection. Ethical approval SSNAP: SSNAP has been granted permission by the NHS Health Research Authority to collect patient data under Section 251 of the Health and Social Care Act 2006. Patients may opt out of data collection and record linkage. South London Stroke Register (SLSR): All patients and/or their relatives gave written informed consent to participate in the study, and over the study period very few patients have declined to be registered. The design of the study was approved by the ethics committees of Guy's and St Thomas' NHS Foundation Trust, King's College Hospital Foundation Trust, St George's University Hospital, National Hospital for Nervous Diseases, and Westminster Hospital. Guarantor BDB.

Ethical approval SSNAP: SSNAP has been granted permission by the NHS Health Research Authority to collect patient data under Section 251 of the Health and Social Care Act 2006. Patients may opt out of data collection and record linkage. South London Stroke Register (SLSR): All patients and/or their relatives gave written informed consent to participate in the study, and over the study period very few patients have declined to be registered. The design of the study was approved by the ethics committees of Guy's and St Thomas' NHS Foundation Trust, King's College Hospital Foundation Trust, St George's University Hospital, National Hospital for Nervous Diseases, and Westminster Hospital. Guarantor BDB. Contributions XMX, EV, LP, BDB carried out the analyses. BDB, XMX, DW designed the study. DW supervised the health economic methods. AD prepared data extracts and analyses from the SLSR. AH, CDAW, AGR supervised the study and provided clinical input. BDB and XMX wrote the first draft of the paper. All authors contributed to the final version of the manuscript.

Introduction Aphasia occurs in about one third of stroke patients and has severe consequences for verbal communication and quality of life.1,2 Several randomised controlled trials (RCT) have reported a benefit of speech and language therapy (SLT) over no treatment for patients with aphasia due to stroke.3 Hence, most patients receive SLT as part of their rehabilitation program. The relationship between timing of SLT, i.e. the interval between stroke onset and start of treatment, and its efficacy is unclear.4 In a meta-analysis comparing studies with different starting points of SLT, the average effect size in studies evaluating treatment initiated in the first 3 months after stroke was larger than that in studies performed in a later stage.5 However, this analysis was mainly based on uncontrolled and non-randomised studies. The efficacy of early initiated SLT has been studied in four trials with contradictory findings; two large studies were neutral, two smaller suggested an effect of early treatment.6–9 The need for more research on the effect of timing of SLT was explicitly accentuated in a Cochrane Review on efficacy of SLT for post-stroke aphasia.10

early initiated SLT has been studied in four trials with contradictory findings; two large studies were neutral, two smaller suggested an effect of early treatment.6–9 The need for more research on the effect of timing of SLT was explicitly accentuated in a Cochrane Review on efficacy of SLT for post-stroke aphasia.10 In the early phase after stroke, impairment-based cognitive-linguistic treatment (CLT) is often preferred over other types of SLT, as it targets specific linguistic functions supposedly stimulating functional neural networks.11–13 As most recovery occurs within the first 3 months after stroke,5,14–16 standard practice early after stroke often comprises CLT.17 When linguistic performance reaches a plateau, SLT may be continued with compensatory treatment instead of CLT. There is some evidence suggesting that high-intensity treatment may be more effective than less frequent therapy.3,18,19 However, the feasibility of high-intensity treatment is questionable, as in several trials compliance with treatment was significantly lower in intervention groups with intensive language treatment.3 Experts in language rehabilitation suggest a best-practice regime of early initiated intensive CLT.13,17 Scientific evidence underpinning this recommendation is frail. The objective of the Rotterdam Aphasia Therapy Study (RATS)-3 was to study whether early intensive CLT for 4 weeks is more effective than no language treatment in the first 4 to 6 weeks after stroke, and whether this approach generates a long-lasting benefit.

ientific evidence underpinning this recommendation is frail. The objective of the Rotterdam Aphasia Therapy Study (RATS)-3 was to study whether early intensive CLT for 4 weeks is more effective than no language treatment in the first 4 to 6 weeks after stroke, and whether this approach generates a long-lasting benefit. Methods Essential elements of the study design are described below. Detailed methods were published elsewhere and are provided as online Appendix 2.20 RATS-3 is a prospective multicentre controlled clinical trial with randomised treatment allocation, open label treatment and blinded evaluation of the primary outcome measure (PROBE-design). Fourteen regional networks for integrated stroke care across the Netherlands participated (online Appendix 1). Within 2 weeks of stroke onset, patients were randomised to 4 weeks of either intensive CLT or no language treatment. After 4 weeks, both groups received regular SLT. The study protocol was approved by the Medical Ethical Committee of the Erasmus MC (MEC-2005-347) and the study was registered in the Netherlands Trial Register (NTR3271). Participants Speech and language therapists (SL-therapists) from participating centres checked eligibility criteria (Table 1) and requested informed consent from patients and/or their proxy. Information about RATS-3 was provided to patients and their relatives orally and on paper, including simplified information leaflets adapted to people with aphasia. Patients who were not eligible or who did not consent to participation were not registered. Table 1. Eligibility criteria for RATS-3.

s and/or their proxy. Information about RATS-3 was provided to patients and their relatives orally and on paper, including simplified information leaflets adapted to people with aphasia. Patients who were not eligible or who did not consent to participation were not registered. Table 1. Eligibility criteria for RATS-3. Inclusion criteria 1 Aphasia after stroke, diagnosed by a neurologist or rehabilitation physician and speech and language therapist 2 Aphasia ascertained with shortened Token Test (score < 29) or Goodglass Aphasia Severity Rating Scale (score < 5) 3 Testable with ScreeLing 4 Treatment can be started within 2 weeks after stroke onset 5 Age 18–85 years 6 Language near-native Dutch 7 Life expectancy >6 months Exclusion criteria 1 Pre-existing aphasia 2 Subarachnoid/subdural haemorrhage/haematoma 3 Language therapy is not feasible because of: Severe dysarthria Premorbid dementia Illiteracy Severe developmental dyslexia Severe visual perceptual disorders Recent psychiatric history Randomisation The trial-coordinator verified inclusion criteria and, after written informed consent was obtained, included and randomised participants within 2 weeks of stroke onset. Independent trial-assistants concealed computer-generated allocation sequences in consecutively numbered, opaque, sealed envelopes. Randomisation was stratified according to baseline aphasia severity (Aphasia Severity Rating Scale: ASRS-score 0–2 = severe; ASRS-score 3–4 = moderate/mild) and including centre.

ke onset. Independent trial-assistants concealed computer-generated allocation sequences in consecutively numbered, opaque, sealed envelopes. Randomisation was stratified according to baseline aphasia severity (Aphasia Severity Rating Scale: ASRS-score 0–2 = severe; ASRS-score 3–4 = moderate/mild) and including centre. Baseline tests At baseline, a short test battery was conducted including the ScreeLing, the 36-item Token Test and a semi-standardised interview for eliciting spontaneous speech, which was rated with the ASRS.20 An experienced SL-therapist blinded to treatment allocation classified the spontaneous speech samples as fluent or non-fluent. Baseline characteristics and the Barthel Index were recorded, as well as treatment with intravenous alteplase, as this is associated with rapid recovery from stroke.21

as rated with the ASRS.20 An experienced SL-therapist blinded to treatment allocation classified the spontaneous speech samples as fluent or non-fluent. Baseline characteristics and the Barthel Index were recorded, as well as treatment with intravenous alteplase, as this is associated with rapid recovery from stroke.21 Intervention Patients in the intervention-group were to receive at least 1 h of CLT every day of the week for a period of 4 weeks. The hour of treatment could be delivered in more than one session per day, if preferable. We chose an intervention period of 4 weeks for three reasons. First, intervention in the control-group had to reflect usual care in the Netherlands, where SLT for aphasia starts on average 3 to 6 weeks after onset. Second, we specifically aimed to study the effect of early initiated treatment. With a maximal inclusion period of 2 weeks and a 4-week intervention period, this early phase was not exceeded. Last, we expected that a longer intervention period with high intensity would be too burdensome for many patients. Treatment was directed at semantics using the therapy program BOX22 and/or phonology using the therapy program FIKS23 to improve word finding deficits. Participating SL-therapists had ample experience in using both Dutch therapy-programs and carefully selected exercises for face-to-face treatment and homework, registered as part of the total amount of treatment provided. The control-group received no language treatment during the first 4 weeks after randomisation. Minimal counselling was allowed, aimed at preventing communication problems and included elaborate information about aphasia and providing communication advice. Concise diagnostics for therapy goal setting was allowed also. The trial-coordinator had at least two-weekly contact with the SL-therapists to ensure no treatment was provided in the control-group and to monitor compliance in the intervention-group. After 4 weeks, further SLT was left to the discretion of the local SL-therapist in both groups.

herapy goal setting was allowed also. The trial-coordinator had at least two-weekly contact with the SL-therapists to ensure no treatment was provided in the control-group and to monitor compliance in the intervention-group. After 4 weeks, further SLT was left to the discretion of the local SL-therapist in both groups. Assessments An extensive linguistic test-protocol was conducted at three time-points; 4 weeks, 3 months and 6 months after randomisation, with the following tests for language and communication: Amsterdam-Nijmegen Everyday Language Test (ANELT) for everyday functional verbal communication,24 a semi-standardised interview from the Aachen Aphasia Test (AAT) rated with the reliable and valid ordered categorical six-point ASRS, the ScreeLing, the Token Test and the Boston Naming Test. The battery also included tests for semantic processing: Semantic Association Test (SAT), verbal version; Comprehensive Aphasia Test (CAT), word comprehension; and Category Fluency; and for phonological processing: Nonword repetition and Auditory Lexical Decision from the Psycholinguistic Assessment of Language Processing in Aphasia (PALPA) and Letter Fluency. In addition, we assessed general functional outcome with the EQ-5D-3L for quality of life, and the modified Rankin Scale (mRS) and Barthel Index for level of independency.

Nonword repetition and Auditory Lexical Decision from the Psycholinguistic Assessment of Language Processing in Aphasia (PALPA) and Letter Fluency. In addition, we assessed general functional outcome with the EQ-5D-3L for quality of life, and the modified Rankin Scale (mRS) and Barthel Index for level of independency. Outcomes Primary outcome was the ANELT A-score ‘understandability’ (range: 10–50, higher scores equal better performance), measuring the adequacy of verbal communication, 4 weeks after randomisation. This valid and reliable test was chosen to verify whether the impairment-based CLT generalises to everyday communication.24 All ANELTs were audio-recorded and rated by two independent assessors, blinded to intervention and time-point. The mean of these two scores was used for analyses. Secondary outcomes were scores on the linguistic tests, EQ-5D-3L and mRS at 4 weeks, and scores on the ANELT-A, the linguistic tests, EQ-5D-3L, and mRS at 3 and 6 months after randomisation. Sample size We considered a four-point difference between both groups on the ANELT-A a clinically worthwhile treatment effect. This is 50% of the critical difference for individual improvement and half a standard deviation of average ANELT-A scores in previous RATS trials.24–26 We estimated that a sample of 150 participants would provide 84% power to find a statistically significant treatment effect at a 5% two-sided significance level.

ffect. This is 50% of the critical difference for individual improvement and half a standard deviation of average ANELT-A scores in previous RATS trials.24–26 We estimated that a sample of 150 participants would provide 84% power to find a statistically significant treatment effect at a 5% two-sided significance level. Statistical analyses Primary analyses were performed on intention-to-treat basis. In addition, on-treatment analyses were performed, with on-treatment being defined for the intervention-group as having accomplished at least the intended intensity of 28 h in 4 weeks and for the control-group as having received no language treatment during 4 weeks after randomisation. We used linear regression to analyse the treatment effect as a mean difference in ANELT-A scores between the intervention and control-group 4 weeks after randomisation, adjusted for age (years), sex, education (high or low), baseline aphasia severity (ASRS-score), type of stroke (ischaemic or haemorrhagic), location of stroke (right or left hemisphere) and baseline Barthel Index score. Linear regression was also used to analyse the effect of treatment on the specific linguistic measures and measures of general functional outcome at 4 weeks, 3 months and 6 months after randomisation, with the same adjustments as in the primary analysis. For the ordered categorical variable mRS, we used multivariable ordinal logistic regression. Post-hoc subgroup analyses were performed with the variables used for baseline-adjustment (online Appendix 3). The results were combined with available evidence from previous trials on early aphasia treatment in a meta-analysis (online Appendix 4).

orical variable mRS, we used multivariable ordinal logistic regression. Post-hoc subgroup analyses were performed with the variables used for baseline-adjustment (online Appendix 3). The results were combined with available evidence from previous trials on early aphasia treatment in a meta-analysis (online Appendix 4). Results From 1 January 2012 until 2 December 2014, we included 153 participants with first-ever aphasia due to stroke, of whom 80 were allocated to the intervention-group. One participant in the control-group was excluded after randomisation, because more detailed assessment revealed that a brain tumour had been misdiagnosed as haemorrhagic infarct. The baseline distribution of clinical characteristics was similar for both groups (Table 2). Table 2. Baseline characteristics of participants in RATS-3.

nt in the control-group was excluded after randomisation, because more detailed assessment revealed that a brain tumour had been misdiagnosed as haemorrhagic infarct. The baseline distribution of clinical characteristics was similar for both groups (Table 2). Table 2. Baseline characteristics of participants in RATS-3. Intervention group (n = 80) Control group (n = 72) Age (years; mean, SD) 66 (12) 66 (12) Sex (male; n, %) 48 (60%) 37 (51%) Handedness (n, %) Right 63 (79%) 61 (85%) Left 6 (8%) 7 (10%) Ambidextrous 5 (6%) 1 (1%) Unknown 6 (8%) 3 (4%) Level of education (n, %) No/unfinished elementary school 3 (4%) 0 Elementary school 13 (16%) 11 (15%) Unfinished junior secondary vocational education 4 (5%) 8 (11%) Junior secondary vocational education 27 (34%) 13 (18%) Total low education 47 (59%) 32 (44%) Senior vocational education 17 (21%) 16 (22%) Higher education 13 (16%) 18 (25%) University 2 (3%) 3 (4%) Total high education 32 (40%) 37 (51%) Unknown 1 (1%) 3 (4%) Type of stroke (n, %) Ischaemic 60 (75%) 61 (85%) Haemorrhagic 20 (25%) 11 (15%) Location of lesion (n, %) Left hemisphere 77 (96%) 69 (96%) Right hemisphere 3 (4%) 3 (4%) Treatment with intravenous alteplase (n, %) Yes 28 (35%) 16 (22%) No 50 (63%) 55 (76%) Unknown 2 (3%) 1 (1%) Time between stroke and randomisation (days; mean, range) 8 (1–18) 8 (2–15) Time between stroke and start treatment (days; mean, range) 12 (5–22) n.a. Barthel Index score (median, IQR) 15 (14) 17 (12.5) Aphasia severity (n, %) Severe (ASRS-score = 0 to 2) 44 (55%) 30 (42%) Mild-moderate (ASRS-score = 3 to 4) 36 (45%) 42 (58%) Fluency (n, %) Fluent aphasia 26 (33%) 30 (42%) Non-fluent aphasia 52 (65%) 42 (58%) Missing 2 (3%) 0 SD: standard deviation; n: number; IQ: Interquartile Range; ASRS: Aphasia Severity Rating Scale; n.a.: not applicable.

evere (ASRS-score = 0 to 2) 44 (55%) 30 (42%) Mild-moderate (ASRS-score = 3 to 4) 36 (45%) 42 (58%) Fluency (n, %) Fluent aphasia 26 (33%) 30 (42%) Non-fluent aphasia 52 (65%) 42 (58%) Missing 2 (3%) 0 SD: standard deviation; n: number; IQ: Interquartile Range; ASRS: Aphasia Severity Rating Scale; n.a.: not applicable. In the intervention-group, two patients died in the intervention period, and in the control-group, one patient died in the intervention period and one just afterwards, before testing could be performed (Figure 1). During follow-up, in each group two patients died. Five participants from the intervention-group did not receive the allocated treatment; one was very ill and four refused intensive treatment. In the control-group, 10 participants refused deferred treatment and received regular SLT. The trial-coordinator did not interfere with treatment, and details on the content of SLT provided to these patients were not recorded. Figure 1. Flow-chart Rotterdam Aphasia Therapy Study-3. Compliance A treatment intensity of 28 h in 4 weeks in the intervention-group was achieved by 23 of 80 patients (29%). The median treatment intensity was 24.5 h in 4 weeks (IQR: 9.5) (Figure 2). Figure 2. Distribution of treatment intensity in the intervention group (total number of hours in 4 weeks).

In the intervention-group, two patients died in the intervention period, and in the control-group, one patient died in the intervention period and one just afterwards, before testing could be performed (Figure 1). During follow-up, in each group two patients died. Five participants from the intervention-group did not receive the allocated treatment; one was very ill and four refused intensive treatment. In the control-group, 10 participants refused deferred treatment and received regular SLT. The trial-coordinator did not interfere with treatment, and details on the content of SLT provided to these patients were not recorded. Figure 1. Flow-chart Rotterdam Aphasia Therapy Study-3. Compliance A treatment intensity of 28 h in 4 weeks in the intervention-group was achieved by 23 of 80 patients (29%). The median treatment intensity was 24.5 h in 4 weeks (IQR: 9.5) (Figure 2). Figure 2. Distribution of treatment intensity in the intervention group (total number of hours in 4 weeks). Intention-to-treat analysis The mean score on the primary outcome, the ANELT-A at 4 weeks, was 33.2 in the intervention-group and 36.2 in the control-group, with a difference of −3.01; 95% confidence interval (CI): [−7.15 to 1.14]. Baseline aphasia severity and baseline Barthel Index were strong prognostic factors in the regression model (online Table 3). The adjusted mean difference in scores on the ANELT-A was 0.39; 95% CI: [−2.70 to 3.47], p = 0.805 (Figure 3). There were also no statistically significant differences on the ANELT-A between groups at 3 months (adjusted difference = 0.54, 95% CI: [−3.04 to 4.12], p = 0.767) and 6 months (adjusted difference = −0.41, 95% CI: [−3.70 to 2.89], p = 0.807) after randomisation (Figure 3). Figure 3. Differences in outcome and treatment effect between intervention and control on the ANELT-A. 95%CI: 95% Confidence Interval; ANELT: Amsterdam-Nijmegen Everyday Language Test; unadj. diff: unadjusted differences *) Statistically significant at a 95% confidence level ◊) Primary outcome

randomisation (Figure 3). Figure 3. Differences in outcome and treatment effect between intervention and control on the ANELT-A. 95%CI: 95% Confidence Interval; ANELT: Amsterdam-Nijmegen Everyday Language Test; unadj. diff: unadjusted differences *) Statistically significant at a 95% confidence level ◊) Primary outcome No statistically significant treatment effects were observed on the linguistic tests and on the measures for general functional outcome, at any time-point (online Table 4). Post-hoc subgroup analyses are provided as online Appendix 3. The meta-analysis showed no beneficial effect of early SLT (online Appendix 4). On-treatment analysis In the on-treatment analysis, we included all patients of the intervention-group who received at least the prespecified intensity of 28 h in 4 weeks (n = 23, 29%) and all subjects in the control-group who did not receive any treatment (n = 62, 86%). Baseline characteristics of the intervention and control-group included in the on-treatment analyses were similar (online Table 5). When on-treatment criteria were applied, the intervention-group reached significantly higher scores than the control-group after 4 weeks on the primary outcome ANELT-A (adjusted difference = 5.41, 95% CI: [1.52 to 9.31], p = 0.007); SAT verbal (adjusted difference = 3.57, 95% CI: [0.36 to 6.78], p = 0.030) and CAT word comprehension (adjusted difference = 3.64, 95% CI: [0.58 to 6.69], p = 0.020) (Figure 3, online Table 6). On all other outcome measures and time points, results did not differ from those of the intention-to-treat analyses.

= 0.007); SAT verbal (adjusted difference = 3.57, 95% CI: [0.36 to 6.78], p = 0.030) and CAT word comprehension (adjusted difference = 3.64, 95% CI: [0.58 to 6.69], p = 0.020) (Figure 3, online Table 6). On all other outcome measures and time points, results did not differ from those of the intention-to-treat analyses. Discussion Principal findings In this multicentre RCT in 152 patients with aphasia due to stroke, we found that 4 weeks of early intensive CLT did not result in better everyday verbal communication than no early language treatment. The 95% CIs for the adjusted differences between groups did not include the pre-specified clinically relevant difference of four points on the ANELT-A, which allows us to conclude that early intensive CLT is not effective.

sive CLT did not result in better everyday verbal communication than no early language treatment. The 95% CIs for the adjusted differences between groups did not include the pre-specified clinically relevant difference of four points on the ANELT-A, which allows us to conclude that early intensive CLT is not effective. This contradicts the findings from two smaller RCTs in which a benefit of early intensive treatment was reported. In 59 patients, 30–80 min of impairment-based SLT per workday for 4 weeks initiated 3 days after stroke improved communication more than usual care (<80 min per week).7 Although nearly 20% of the patients in the intervention-group did not achieve the minimum treatment intensity of 150 min per week, the authors conclude that daily treatment is feasible early after stroke and, if tolerated, is effective for recovery of aphasia. In another study, 12 patients were randomly allocated to 2 weeks of either 1-h sessions of impairment-based SLT on workdays starting on average 2.2 days after stroke or no SLT.8 In addition to statistically significant better scores on the AAT subparts Naming and Written language processing in the early treatment group, the authors report significant differences between groups in post treatment recruitment of brain areas on functional MRI scans. However, this is a very small trial with only six participants per treatment-arm.

ificant better scores on the AAT subparts Naming and Written language processing in the early treatment group, the authors report significant differences between groups in post treatment recruitment of brain areas on functional MRI scans. However, this is a very small trial with only six participants per treatment-arm. Our findings are in line with those from two larger RCTs on early initiated SLT. In a trial among 123 patients, Laska et al. found no effect of 3 weeks of early intensive impairment-based SLT on ANELT-A scores 3 weeks and 6 months after stroke onset.6 Bowen et al. randomly allocated 170 stroke patients with communication deficits to either agreed best-practice SLT or social support provided by trained volunteers for 16 weeks starting on average 2 weeks after stroke onset.9 They found no differences regarding functional communication at follow-up and conclude that SLT is not more effective than social support. This trial differs from ours, as stroke patients with either aphasia, dysarthria or both were included, which makes the results difficult to interpret. Furthermore, treatment intensity was tailored to the individuals’ needs and possibilities. Consequently, treatment intensity was on average only 1.5 h per week, which may not have been sufficient to reach a sizeable treatment effect.3,18,19

sarthria or both were included, which makes the results difficult to interpret. Furthermore, treatment intensity was tailored to the individuals’ needs and possibilities. Consequently, treatment intensity was on average only 1.5 h per week, which may not have been sufficient to reach a sizeable treatment effect.3,18,19 While the concept of early language rehabilitation after stroke is attractive, the summary of evidence in our meta-analysis shows that SLT, whether or not intensive, when started within 4 weeks after stroke onset, is not more effective in improving verbal communication or language functioning, than regular, less intensive or deferred treatment.

language rehabilitation after stroke is attractive, the summary of evidence in our meta-analysis shows that SLT, whether or not intensive, when started within 4 weeks after stroke onset, is not more effective in improving verbal communication or language functioning, than regular, less intensive or deferred treatment. Strengths and weaknesses Strengths of RATS-3 are its large size, multicentre design, a clearly defined clinically relevant intervention contrast, and representative cohort of patients with post-stroke aphasia. The treatment programs used in the intervention-group are frequently applied in daily practice in the Netherlands and have good potential to generate an effect on language recovery, as exercises are directed at facilitating word finding, an essential problem in aphasia. Consequently, results of our trial are highly generalizable to daily practice. We could have opted for a more distinct intervention contrast by actively limiting all language-related activities in the control-group, e.g. reading, writing and computer use, but that would not reflect daily reality. In fact, our aim was to study whether intensive CLT, added to language-related activities people with aphasia engage in naturally, is effective for the recovery of aphasia.

ly limiting all language-related activities in the control-group, e.g. reading, writing and computer use, but that would not reflect daily reality. In fact, our aim was to study whether intensive CLT, added to language-related activities people with aphasia engage in naturally, is effective for the recovery of aphasia. Many efficacy studies on impairment-based treatment have used impairment-based language tests as outcome measures, e.g. naming or word-comprehension, as these are closely related to the intervention being studied.27 However, scores on linguistic tests are rather artificial and do not necessarily reflect adequate functional communication in daily life, which should be the ultimate goal of aphasia treatment.3 Therefore, a relevant and reliable measure of communication, most closely reflecting the patients’ sense of recovery and return to normal functioning, is preferable.10 Hence, in line with our previous trials, both in which we found that improvement on the ANELT-A was correlated with improvement at the impairment level, we used the ANELT-A as primary outcome measure.16,25,26

ion, most closely reflecting the patients’ sense of recovery and return to normal functioning, is preferable.10 Hence, in line with our previous trials, both in which we found that improvement on the ANELT-A was correlated with improvement at the impairment level, we used the ANELT-A as primary outcome measure.16,25,26 Our study has limitations. Although we accomplished a high median treatment intensity of 24.5 h in 4 weeks, achieving the intended intensity of 28 h appeared a major challenge. Even with a strictly protocolled treatment regime and highly motivated SL-therapists who were frequently contacted by the trial-coordinator, less than 30% of the intervention-group achieved the requested intensity. Patients were often too tired or ill to practise 1 h per day, even if treatment was spread over the day. Although poor adherence to the protocol was mainly caused by patient-related issues, organisational problems such as limited availability of therapists, or priority given to motor rehabilitation also played a role, albeit minor. Although this trial was no feasibility study, the results demonstrate that even if intensive treatment had been found more effective for selected patients, feasibility is improbable for all stroke patients with aphasia early after onset. This is in line with findings from the most recent Cochrane review.3

albeit minor. Although this trial was no feasibility study, the results demonstrate that even if intensive treatment had been found more effective for selected patients, feasibility is improbable for all stroke patients with aphasia early after onset. This is in line with findings from the most recent Cochrane review.3 Patient selection seems essential to generate a potential beneficial effect of early intensive CLT on recovery of aphasia, as the on-treatment analyses did show a limited effect. However, this finding should be interpreted with great caution, as on-treatment analyses could only be performed in patients in the intervention-group who could tolerate intensive treatment, whereas the control-group comprised both patients who may and may not tolerate this intensive regime. Completeness of follow-up for the primary outcome was 93%, which is in line with other studies in this field.3 At 6 months after stroke, 19% of participants had refused follow-up testing. This may have reduced the validity of our findings, but the measurements at 3 and 6 months follow-up are secondary outcomes and are in line with the primary outcome.

primary outcome was 93%, which is in line with other studies in this field.3 At 6 months after stroke, 19% of participants had refused follow-up testing. This may have reduced the validity of our findings, but the measurements at 3 and 6 months follow-up are secondary outcomes and are in line with the primary outcome. Implications Despite the lack of unequivocal proof for a beneficial effect of early SLT, deferring treatment in aphasia due to stroke has long been considered unethical.28 However, early after stroke, patients may suffer from concomitant illnesses or fatigue and may not tolerate intensive impairment-based treatment. Our findings demonstrate that it is not detrimental to delay CLT in the first weeks after stroke onset in these vulnerable patients, which also occasionally happens unintentionally due to waiting lists or lengthy diagnostic pathways.

mitant illnesses or fatigue and may not tolerate intensive impairment-based treatment. Our findings demonstrate that it is not detrimental to delay CLT in the first weeks after stroke onset in these vulnerable patients, which also occasionally happens unintentionally due to waiting lists or lengthy diagnostic pathways. However, our findings do not justify the conclusion that the work of SL-therapists is redundant in the first weeks after stroke, as patients with aphasia and their proxies definitely need guidance and help in coping with their deficits early after stroke. In times of radical changes in health care policy and budget cutbacks, SL-therapists are urged to utilise their limited resources effectively for patients with acute stroke. Instead of focusing on impairment-based treatment, they might better put more emphasis on counselling and providing communication support, which are essential for coping with communication problems and prevention of social isolation. CLT may be more effective later in the course of this disabling condition.

ute stroke. Instead of focusing on impairment-based treatment, they might better put more emphasis on counselling and providing communication support, which are essential for coping with communication problems and prevention of social isolation. CLT may be more effective later in the course of this disabling condition. Future research Future studies should aim to find the optimal timing of commonly used treatment types, either impairment-based or functional approaches. New studies may be focussed on patient selection also, as results from our on-treatment analyses indicate that some patients might benefit from early intensive treatment. International cooperation is one way to conduct large aphasia trials that allow for more reliable pre-specified subgroup analyses, which is of great importance to identify factors contributing to treatment success and may enable individualisation of SLT. Conclusion Our study shows that 4 weeks of intensive CLT aimed at semantic and phonological processing started within 2 weeks after stroke onset does not improve the recovery of aphasia, either in the short or long term. Supplementary Material Supplementary material Acknowledgements We thank Marjolein de Jong-Hagelstein and Carolina Mendez Orellana for contributing to the conception and outset of RATS-3. We acknowledge all participating centres in the Netherlands for recruiting patients and treating them according to the trial protocol. We are thankful to Irma Adbegovic, Liset Bergevoets, Yvonne Hendrick, Nienke Wolthuis and Marjolein Zomerdijk for their work as blinded ANELT assessors.

conception and outset of RATS-3. We acknowledge all participating centres in the Netherlands for recruiting patients and treating them according to the trial protocol. We are thankful to Irma Adbegovic, Liset Bergevoets, Yvonne Hendrick, Nienke Wolthuis and Marjolein Zomerdijk for their work as blinded ANELT assessors. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: All authors have completed the ICMJE Form for Disclosure of Potential Conflicts of Interest (available on request from the corresponding author) and declare no competing interests. LL has received a fellowship of the Brain Foundation Netherlands. DD, PK, EV and FN are employees of the Department of Neurology of the Erasmus MC. This department has received grants from Zawabas Fund, Coolsingel Fund, Dura Fund, Blokland Fund and the Erasmus MC Care Research Fund for conducting this trial.

interests. LL has received a fellowship of the Brain Foundation Netherlands. DD, PK, EV and FN are employees of the Department of Neurology of the Erasmus MC. This department has received grants from Zawabas Fund, Coolsingel Fund, Dura Fund, Blokland Fund and the Erasmus MC Care Research Fund for conducting this trial. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: RATS-3 was funded by a fellowship granted to LL by the Brain Foundation Netherlands (project number: 2011(1)-20) and grants from Zawabas Fund, Coolsingel Fund, Dura Fund, Blokland Fund and the Erasmus MC Care Research Fund. The funding sources of RATS-3 had no role in study design, data collection, data analysis, data interpretation or writing of the report. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. Ethical approval Ethical approval for this study protocol was obtained from the Medical Ethical Committee of the Erasmus MC (MEC-2005-347). Informed consent Written informed consent was obtained from all participants or their next of kin before inclusion in the study. Trial registration The study protocol for this randomised controlled trial was registered in the Netherlands Trial Register (NTR3271). Guarantor FN is guarantor and takes responsibility for the accuracy of the manuscript.

Informed consent Written informed consent was obtained from all participants or their next of kin before inclusion in the study. Trial registration The study protocol for this randomised controlled trial was registered in the Netherlands Trial Register (NTR3271). Guarantor FN is guarantor and takes responsibility for the accuracy of the manuscript. Contributorship RATS-3 was conceived by EV, DD, LL, PK and MS. LL was principal investigator of this trial and FN was trial-coordinator. FN primarily acquired and analysed the data. SG and DB were local principal investigators of centres that recruited more than 10% of the study population and critically reviewed the manuscript. HL and FN conducted statistical analyses. EV, MS, DD, PK, LL, HL and FN interpreted the data. All data and statistical analysis reports were available to all authors. FN drafted the first version of the manuscript. All authors critically read and revised the manuscript. The final version of the manuscript was approved by all authors. Data sharing Anonymous data can be requested from the corresponding author with a detailed description containing the aims and methods of the study for which the data are intended to be used. Data will be contributed to the project titled: ‘REhabilitation and recovery of peopLE with Aphasia after StrokE’ (RELEASE) funded by the National Institute of Health Research in the UK (NIHR). Informed consent was not obtained from participants for data sharing, but the data will be anonymised and risk of identification is low.

Introduction In the first days after stroke, about half of all patients develop complications, including infections and fever. The risk of developing these events is greater in patients of higher age or with more severe stroke.1–3 Aspiration, infections and fever can impede functional recovery, prolong hospital admissions, and are independently associated with an increased risk of death or long-term dependency.1,2,4–9 In addition, systematic review of animal studies modelling ischaemic stroke has shown that hyperthermia during or shortly after the onset of ischaemia substantially increases infarct size, suggesting that the relation between fever and poor outcome observed in patients is at least in part causal.10

m dependency.1,2,4–9 In addition, systematic review of animal studies modelling ischaemic stroke has shown that hyperthermia during or shortly after the onset of ischaemia substantially increases infarct size, suggesting that the relation between fever and poor outcome observed in patients is at least in part causal.10 The risk of developing these complications can be reduced by very simple, safe and inexpensive measures, such as metoclopramide for the management of dysphagia, antibiotics for the prevention of infections and paracetamol for the prevention of fever, but it is uncertain whether these measures also improve functional outcome.11–14 In some, generally small, randomised trials, preventive treatment with these drugs not only convincingly reduced the risks of aspiration, infections, or fever by one third to one half, but was also associated with clear trends towards a lower risk of death or poor outcome.11–14 The cluster-randomised Quality in Acute Stroke Care (QASC) study demonstrated that implementation of nursing protocols for the management of fever, hyperglycaemia and swallowing dysfunction on a stroke unit resulted in better outcomes.15 However, in two recent large trials, preventive treatment with antibiotics did not improve functional outcomes.16,17

Acute Stroke Care (QASC) study demonstrated that implementation of nursing protocols for the management of fever, hyperglycaemia and swallowing dysfunction on a stroke unit resulted in better outcomes.15 However, in two recent large trials, preventive treatment with antibiotics did not improve functional outcomes.16,17 American guidelines for the treatment of patients with acute ischaemic stroke advocate screening for dysphagia; the use of antibiotics in patients with infections; and antipyretic drugs such as paracetamol in patients with subfebrile temperatures or fever.18 Guidelines of the European Stroke Organisation concluded that there is insufficient evidence from randomised trials to make strong recommendations on whether, when and to whom preventive antibiotic or antipyretic treatment should be given after ischaemic stroke or intracerebral haemorrhage.19,20 The authors called for randomised trials to allow for better-informed recommendations in the future.20 The PREvention of Complications to Improve OUtcome in elderly patients with acute Stroke (PRECIOUS) trial will assess whether a pharmacological strategy to prevent aspiration, infections, or fever with metoclopramide, ceftriaxone, paracetamol, or any combination of these in elderly patients with a moderately severe to severe acute stroke is more effective at reducing the risk of death and improving functional outcome than current clinical practice of waiting until these complications are manifest before initiating treatment.

ramide, ceftriaxone, paracetamol, or any combination of these in elderly patients with a moderately severe to severe acute stroke is more effective at reducing the risk of death and improving functional outcome than current clinical practice of waiting until these complications are manifest before initiating treatment. Design Overview and timelines PRECIOUS is an international, multi-centre, 3 × 2-factorial, randomised, controlled, open-label clinical trial with blinded outcome assessment (PROBE) of the preventive use of metoclopramide, ceftriaxone, paracetamol, or any combination of these, for four days in elderly patients with acute ischaemic stroke or intracerebral haemorrhage. The primary outcome measure is the score on the modified Rankin Scale (mRS) at 90 days (±14 days).21 3800 patients will be recruited over a period of about four years in about 80 hospitals (both academic and regional) in 9 European countries (Figure 1). The first patient was included in May 2016 and the main results are anticipated to be available in 2020. The complete and most recent version of the study protocol is available at www.precious-trial.eu. Figure 1: Participating countries in PRECIOUS.

Design Overview and timelines PRECIOUS is an international, multi-centre, 3 × 2-factorial, randomised, controlled, open-label clinical trial with blinded outcome assessment (PROBE) of the preventive use of metoclopramide, ceftriaxone, paracetamol, or any combination of these, for four days in elderly patients with acute ischaemic stroke or intracerebral haemorrhage. The primary outcome measure is the score on the modified Rankin Scale (mRS) at 90 days (±14 days).21 3800 patients will be recruited over a period of about four years in about 80 hospitals (both academic and regional) in 9 European countries (Figure 1). The first patient was included in May 2016 and the main results are anticipated to be available in 2020. The complete and most recent version of the study protocol is available at www.precious-trial.eu. Figure 1: Participating countries in PRECIOUS. Study population The study population consists of patients aged 66 years or older who are hospitalised with moderately severe to severe acute ischaemic stroke or intracerebral haemorrhage and can be treated within 24 h of stroke onset. In order to be eligible to participate, a patient must meet all inclusion criteria listed in Table 1 and none of the exclusion criteria listed in Table 2. Patients with an active infection are excluded. Table 1. Inclusion criteria. A clinical diagnosis of acute ischaemic stroke or intracerebral haemorrhage, confirmed with CT or MRI scana A score on the NIHSS ≥ 6, indicating moderately severe to severe strokeb Age 66 years or older The possibility to start treatment within 24 h of symptom onsetc

Study population The study population consists of patients aged 66 years or older who are hospitalised with moderately severe to severe acute ischaemic stroke or intracerebral haemorrhage and can be treated within 24 h of stroke onset. In order to be eligible to participate, a patient must meet all inclusion criteria listed in Table 1 and none of the exclusion criteria listed in Table 2. Patients with an active infection are excluded. Table 1. Inclusion criteria. A clinical diagnosis of acute ischaemic stroke or intracerebral haemorrhage, confirmed with CT or MRI scana A score on the NIHSS ≥ 6, indicating moderately severe to severe strokeb Age 66 years or older The possibility to start treatment within 24 h of symptom onsetc Written informed consentd CT: computed tomography; MRI: magnetic resonance imaging; NIHSS: National Institutes of Health Stroke Scale. aA normal CT scan is considered compatible with ischaemic stroke. bNIHSS is assessed at the time of inclusion. cIn case of a stuttering stroke, treatment should start within 24 h of the moment the first symptoms occurred. dInformed consent is given by the patient, legal representative or independent physician (depending on local and national regulations). Table 2. Exclusion criteria. Active infection requiring antibiotic treatmenta Pre-stroke score on the mRS ≥4b Death appearing imminent at the time of assessment Criteria for censoring a treatment stratum:For the metoclopramide stratum: Hypersensitivity to metoclopramide or to any of the excipients;

dInformed consent is given by the patient, legal representative or independent physician (depending on local and national regulations). Table 2. Exclusion criteria. Active infection requiring antibiotic treatmenta Pre-stroke score on the mRS ≥4b Death appearing imminent at the time of assessment Criteria for censoring a treatment stratum:For the metoclopramide stratum: Hypersensitivity to metoclopramide or to any of the excipients; Gastrointestinal haemorrhage, mechanical obstruction or gastro-intestinal perforation for which the stimulation of gastrointestinal motility constitutes a risk; Confirmed or suspected pheochromocytoma; History of neuroleptic or metoclopramide-induced tardive dyskinesia; Epilepsy; Parkinson's disease; Use of levodopa or dopaminergic agonists; Known history of methaemoglobinaemia with metoclopramide or of NADH cytochrome-b5 deficiency. Clinical indication for the use of metoclopramide. Incidental use of metoclopramide before screening is not an exclusion criterion. For the ceftriaxone stratum: Known hypersensitivity to beta-lactam antibiotics; Clinical indication for antibiotic treatment. The use of an antibiotic before screening is not an exclusion criterion. For the paracetamol stratum: Known hypersensitivity to paracetamol or any of the excipients; Known severe hepatic insufficiency; Chronic alcoholism; Clinical indication for the use of paracetamol. Incidental use of paracetamol before screening is not an exclusion criterion. mRS: modified Rankin Scale. aAs judged by the treating clinical physician.

For the paracetamol stratum: Known hypersensitivity to paracetamol or any of the excipients; Known severe hepatic insufficiency; Chronic alcoholism; Clinical indication for the use of paracetamol. Incidental use of paracetamol before screening is not an exclusion criterion. mRS: modified Rankin Scale. aAs judged by the treating clinical physician. bScore 4 mRS: Moderately severe disability. Unable to attend to own body needs without assistance and unable to walk unassisted. Patient enrolment After written informed consent, patients are randomly allocated in a 3 × 2 factorial design to any combination of open-label oral, rectal, or intravenous metoclopramide (10 mg thrice daily); intravenous ceftriaxone (2000 mg once daily); oral, rectal, or intravenous paracetamol (1000 mg four times daily); or usual care, started within 24 h after symptom onset and continued for four days or until complete recovery or discharge from hospital, if earlier (Figure 2). The daily dose of metoclopramide is reduced to 3 times 5 mg in patients with moderate to severe renal impairment or with severe hepatic impairment, and to 3 times 2.5 mg in patients with end-stage renal disease. Figure 2: Treatment allocation will be based on proportional minimisation. Investigators will have the opportunity to censor a single randomisation stratum in a specific patient before randomisation. Each of the 8 subgroups is expected to consist of approximately 475 patients.

Patient enrolment After written informed consent, patients are randomly allocated in a 3 × 2 factorial design to any combination of open-label oral, rectal, or intravenous metoclopramide (10 mg thrice daily); intravenous ceftriaxone (2000 mg once daily); oral, rectal, or intravenous paracetamol (1000 mg four times daily); or usual care, started within 24 h after symptom onset and continued for four days or until complete recovery or discharge from hospital, if earlier (Figure 2). The daily dose of metoclopramide is reduced to 3 times 5 mg in patients with moderate to severe renal impairment or with severe hepatic impairment, and to 3 times 2.5 mg in patients with end-stage renal disease. Figure 2: Treatment allocation will be based on proportional minimisation. Investigators will have the opportunity to censor a single randomisation stratum in a specific patient before randomisation. Each of the 8 subgroups is expected to consist of approximately 475 patients. Allocation is based on proportional minimisation through a web-based allocation service. Treatment allocation is stratified by country and includes the following minimisation factors for balance in baseline characteristics: age (66–75 years vs. > 75 years); sex (male vs. female); stroke type (ischaemic stroke vs. intracerebral haemorrhage); stroke severity (NIHSS 6–12 vs. > 12); and diabetes mellitus (yes vs. no). Investigators have the opportunity to censor a single randomisation stratum in a specific patient before randomisation, for example in case of an allergy to one of the study medications (Table 2). Alongside the study treatment, all patients receive standard care as recommended by national or international guidelines or local protocols. This may include thrombolysis and endovascular treatment for acute ischaemic stroke, and treatment of hypertension for intracerebral haemorrhage.

the study medications (Table 2). Alongside the study treatment, all patients receive standard care as recommended by national or international guidelines or local protocols. This may include thrombolysis and endovascular treatment for acute ischaemic stroke, and treatment of hypertension for intracerebral haemorrhage. Data collection and follow-up Baseline characteristics assessed are listed in Table 3. The presence of any treatment restriction, the method of food intake and the vital signs (including body temperature) are recorded at baseline and during the first seven days of hospitalisation. The recording and reporting period for all severe or serious adverse events begins after randomisation and ends on day 7, except for serious adverse reactions and suspected unexpected serious adverse reactions (SUSARs), which are recorded and reported up to 90 days. Death occurring before day 90 (± 14) is a study secondary outcome and is always documented and recorded. Table 3. Baseline characteristics. Demographics: age; sex; ethnicity Comorbidities/medical history: atrial fibrillation; diabetes mellitus; hypertension; pre-stroke mRS Concurrent drugs: use of any antipyretic, antibiotic, or antiemetic drug in the three days before randomisation.a Way of food intake on the day before the strokeb Treatment restrictionsc Dates and times: stroke onset, hospital admission Vital signs: blood pressure; pulse; body temperatured Neurological examination: NIHSS; location of the lesion Laboratory examinationse Results of chest X-ray and urine analysis if performed as part of routine clinical practice

Way of food intake on the day before the strokeb Treatment restrictionsc Dates and times: stroke onset, hospital admission Vital signs: blood pressure; pulse; body temperatured Neurological examination: NIHSS; location of the lesion Laboratory examinationse Results of chest X-ray and urine analysis if performed as part of routine clinical practice Imaging results: stroke type: ischaemic stroke or intracerebral haemorrhage Previous treatment: intravenous thrombolysis with alteplase; intra-arterial treatment. mRS: modified Rankin Scale; NIHSS: National Institutes of Health Stroke Scale. aAspirin in any formulation and in a daily dose of up to 300 mg is not considered an antipyretic drug. bThe method of feeding on the day before the stroke and at noon of the relevant day will be recorded and classified as 1. normal food; 2. oral, soft or fluids only; 3. nasogastric tube; 4. percutaneous endoscopic gastrostomy (PEG); 5. intravenous only; 6. none. cThe presence of any treatment restriction will be recorded at baseline and during the patients stay in the hospital, and will be classified as 1. Do not resuscitate; 2. Do not intubate and ventilate; 3. Withholding other treatments that may prolong life; 4. Withholding food; 5. Withholding fluids; and 6. Palliation with morphine or a benzodiazepine. Any combination of these strategies is possible. dBlood pressure, pulse and body temperature will be assessed at baseline and at 12-h (± 3 h) intervals (where assessed as part of routine clinical practice). Both rectal and tympanic thermometry are allowed.

cThe presence of any treatment restriction will be recorded at baseline and during the patients stay in the hospital, and will be classified as 1. Do not resuscitate; 2. Do not intubate and ventilate; 3. Withholding other treatments that may prolong life; 4. Withholding food; 5. Withholding fluids; and 6. Palliation with morphine or a benzodiazepine. Any combination of these strategies is possible. dBlood pressure, pulse and body temperature will be assessed at baseline and at 12-h (± 3 h) intervals (where assessed as part of routine clinical practice). Both rectal and tympanic thermometry are allowed. eIf assessed at baseline as part of routine clinical practice, the following laboratory tests will be collected: serum glucose; glomerular filtration rate; C-reactive protein (CRP); alkaline phosphatase (ALP); gamma-glutamyl transferase (GGT); alanine aminotransferase (ALT); and aspartate aminotransferase (AST); leucocyte count and differential.

part of routine clinical practice, the following laboratory tests will be collected: serum glucose; glomerular filtration rate; C-reactive protein (CRP); alkaline phosphatase (ALP); gamma-glutamyl transferase (GGT); alanine aminotransferase (ALT); and aspartate aminotransferase (AST); leucocyte count and differential. At day 7 after admission to the hospital, or at discharge if earlier, the score on the mRS is assessed. During a follow-up visit at day 90 (± 14), the mRS is assessed by a trained, certified investigator in a standard fashion according to each centre’s normal practice, and the interview is recorded with a digital video camera. During this recording, no reference to the treatment allocation is made. The videos are uploaded and distributed for independent and blinded scoring by three certified expert raters from the same country as the patient. Additionally, the Barthel index (BI),22 Montreal Cognitive Assessment (MoCA)23 and EuroQol 5D-5L (EQ-5D-5L) are assessed at 90 days, as well as the patient’s location and number of nights spent at home over the first 90 days after stroke. Outcome events The primary outcome measure is the score on the mRS at 90 days (±14).24 The mRS is an ordinal hierarchical scale incorporating seven categories from 0 up to and including 6 and describes the range of disability encountered post stroke. ‘Death’ is assigned a score of 6. Secondary outcomes are outlined in Table 4. Table 4. Study outcomes. Primary outcome Score on the mRSa

Outcome events The primary outcome measure is the score on the mRS at 90 days (±14).24 The mRS is an ordinal hierarchical scale incorporating seven categories from 0 up to and including 6 and describes the range of disability encountered post stroke. ‘Death’ is assigned a score of 6. Secondary outcomes are outlined in Table 4. Table 4. Study outcomes. Primary outcome Score on the mRSa Secondary outcomesAt 7 days (± 1 day) or at discharge, if earlier: Infections in the first seven days (± 1 day; frequency, type and C. difficile infections)b 3rd generation cephalosporin resistance in the first seven days (± 1 day)c Antimicrobial use during the complete hospital admission for stroked SAEs in the first seven days In a subgroup of patients: presence of ESBL-producing bacteria.e At 90 days (± 14 days): Death Unfavourable functional outcomef Disabilityg Cognitionh Quality of lifei Home timej Patient locationk mRS: modified Rankin Scale; SAE: serious adverse event; ESBL: extended-spectrum beta-lactamase. aAs assessed by three independent and blinded adjudicators based on a video recording of an mRS interview at the follow-up visit after 90 days. bInfections will be categorised as diagnosed by the clinician, and as judged by an independent adjudication committee (masked to treatment allocation). cDetected as part of routine clinical practice. dConverted to units of defined daily doses according to the classification of the WHO Anatomical Therapeutic Chemical Classification System with Defined Daily Doses Index. eAs detected by PCR in a rectal swab. fDefined as mRS 3 to 6. gAssessed with the Barthel index (BI).28

bInfections will be categorised as diagnosed by the clinician, and as judged by an independent adjudication committee (masked to treatment allocation). cDetected as part of routine clinical practice. dConverted to units of defined daily doses according to the classification of the WHO Anatomical Therapeutic Chemical Classification System with Defined Daily Doses Index. eAs detected by PCR in a rectal swab. fDefined as mRS 3 to 6. gAssessed with the Barthel index (BI).28 hAssessed with the Montreal Cognitive Assessment (MoCA).29 iAssessed with the EuroQol 5D-5L (EQ-5D-5L). jThe number of nights among the first 90 since stroke onset that are spent in the patient’s own home or a relative’s home. Where final follow-up occurs earlier, the last known placement will be extrapolated to 90 days. kHospital; rehabilitation service; chronic nursing facility; home. Infections will be categorised as diagnosed by the clinician, and as judged by an independent adjudication committee (masked to treatment allocation) according to modified Centres for Disease Control and Prevention criteria.25 The scoring algorithms for infections that will be used by this committee have been described previously and are in line with recommendations of the Pneumonia in Stroke Consensus Group.26 Clostridium difficile infection will be defined as diarrhoea in combination with a positive Clostridium difficile toxin test.

ria.25 The scoring algorithms for infections that will be used by this committee have been described previously and are in line with recommendations of the Pneumonia in Stroke Consensus Group.26 Clostridium difficile infection will be defined as diarrhoea in combination with a positive Clostridium difficile toxin test. Substudy To detect selection of bacteria with third generation cephalosporin resistance caused by increased antibiotic pressure, a nested case-control substudy will be performed in 1000 patients in 30 centres in different participating countries. The presence of extended spectrum beta-lactamase (ESBL) producing bacteria will be assessed with polymerase chain reaction (PCR). For this purpose, two rectal swabs will be collected in each patient, after specific informed consent, on admission and at day 7 (± 1 day, or at discharge, if earlier).

ticipating countries. The presence of extended spectrum beta-lactamase (ESBL) producing bacteria will be assessed with polymerase chain reaction (PCR). For this purpose, two rectal swabs will be collected in each patient, after specific informed consent, on admission and at day 7 (± 1 day, or at discharge, if earlier). Sample size calculation and statistical analysis plan The primary effect estimate will be the difference between groups in the mRS scores obtained through centralised adjudications and assessed using multiple regression, and will be expressed as a mean difference with 95% confidence interval. PRECIOUS is powered to detect a statistically significant shift in the difference in the proportion of patients with mRS 0 to 2 at 90 days, assuming an effect that leads to a 5% absolute increase (from 36 to 41%)26 in the cumulative proportion of patients with mRS 0 to 2 in any intervention group, compared with controls. The effect size of 5% is based on previous smaller studies and/or meta-analyses thereof, performed in more general stroke populations.12–14,16 The statistical analyses will be performed according to the intention-to-treat principle and adjusted for the minimisation factors mentioned, other relevant baseline characteristics, and treatment allocation for the other two strata of the trial. Three separate primary analyses will be performed, looking at the main effects of each of the three interventions compared with their respective controls. Although the study is not powered to detect interactions between the three interventions, such interactions will be investigated in secondary analyses. Two sensitivity analyses will be performed in which all patients who are lost to follow-up will be classified as having the worst possible outcome (death) or the best possible outcome (mRS = 0), respectively.

ect interactions between the three interventions, such interactions will be investigated in secondary analyses. Two sensitivity analyses will be performed in which all patients who are lost to follow-up will be classified as having the worst possible outcome (death) or the best possible outcome (mRS = 0), respectively. Secondary efficacy outcomes will be analysed using similar methods as for the primary efficacy analysis, with binary logistic regression used for binary outcomes, including death, unfavourable outcome and SAEs. Ordinal logistic regression will be used for ordered categorical data and multiple regression for continuous outcomes. Wilcoxon rank sum test will be used for continuous outcome measures which are not normally distributed. Several subgroup analyses will be performed based on age, sex, stroke type and severity, diabetes mellitus, presence of atrial fibrillation, pre-stroke mRS score, treatment with alteplase or other recanalisation method, treatment allocation for the other two trial strata and time to treatment. A full statistical analysis plan will be completed before the final follow-up of the last patient.

e and severity, diabetes mellitus, presence of atrial fibrillation, pre-stroke mRS score, treatment with alteplase or other recanalisation method, treatment allocation for the other two trial strata and time to treatment. A full statistical analysis plan will be completed before the final follow-up of the last patient. Discussion Because several complications in the first days after stroke have consistently been associated with a higher risk of death or poor functional outcome, prevention of these complications appears a logical, simple and safe approach to improve outcome after stroke. In the past two decades, several trials aimed at prevention of complications have been performed, but – besides organised care in a designated stroke unit – no treatment to prevent complications has convincingly shown to improve the functional outcome in patient with stroke.13,14,16 However, most of these trials were underpowered, started treatment too late after stroke onset, or aimed at only one specific complication after stroke. Strengths of PRECIOUS are the assessment in an elderly population with moderately severe to severe stroke (with an increased risk of complications and poor outcome), the start of treatment within 24 h after stroke onset, and its multifactorial design. The trial will provide high-quality evidence that could be broadly generalisable. Because of its pragmatic design and the use of safe, inexpensive, and generally available drugs, the results of PRECIOUS could be implemented rapidly throughout Europe and the rest of the world.

troke onset, and its multifactorial design. The trial will provide high-quality evidence that could be broadly generalisable. Because of its pragmatic design and the use of safe, inexpensive, and generally available drugs, the results of PRECIOUS could be implemented rapidly throughout Europe and the rest of the world. It may be questioned whether the effects of prophylactic antibiotics in patients with stroke should still be assessed after the neutral results in two recent phase III clinical trials. Ceftriaxone is an off-patent, broad-spectrum antibiotic with proven efficacy against bacteria most frequently causing infection in patients with acute stroke.12,27 In the PASS trial, 2550 patients with stroke were randomly assigned to standard care or intravenous ceftriaxone, started within 24 h of stroke onset, continued for four days. Preventive ceftriaxone reduced the incidence of infections in general (from 7% to 3%; p < 0.0001), but did not have an effect on the occurrence of pneumonia or the risk of a poor outcome at 90 days.16 However, the median score on the NIHSS of patients in PASS-was 5, which could explain the relatively low incidence of infections. In the cluster-randomised STROKE-INF trial, which included 1217 stroke patients with dysphagia, prophylactic antibiotics did not change the incidence of post-stroke pneumonia or poor functional outcome.17 However, antibiotic treatment may have started too late (up to 48 h after stroke onset) to prevent early infections. In addition, a considerable proportion of patients in the treatment group received only a limited number of antibiotic doses, while 34% of the patients in the control group received an antibiotic at least once during the first seven days. Finally, individual centres included only a small number of patients over an extended period of time; in a cluster-randomised study, this may induce selection bias decreasing the discriminative power. Because of the limitations of these two trials, and the strong association between infections and a poor functional outcome,4,28 additional evidence on the effect of preventive antibiotics is still strongly needed.

time; in a cluster-randomised study, this may induce selection bias decreasing the discriminative power. Because of the limitations of these two trials, and the strong association between infections and a poor functional outcome,4,28 additional evidence on the effect of preventive antibiotics is still strongly needed. The PRECIOUS trial will also be able to assess whether antibiotics work in isolation, or whether the effect is dependent on the combination of drugs that are used in the trial. The results of PASS and STROKE-INF support the concept that post-stroke pneumonia might be a respiratory syndrome resulting from complex bacterial, chemical and immunological causes that might not be prevented by antibiotics alone. The combination of treatments in PRECIOUS, especially the combination of metoclopramide and ceftriaxone, targets different pathways in the development of post-stroke pneumonia, potentially resulting in a larger reduction in the risk of complications than with the individual treatments alone. The prevention of complications with the treatments proposed in PRECIOUS was safe in previous trials and not associated with an increased risk of SAEs.13,14,16 In addition, the risk of developing Clostridium difficile overgrowth was smaller than 1% in previous studies with ceftriaxone, and there was no association with an increase in antimicrobial resistance.16,17

ments proposed in PRECIOUS was safe in previous trials and not associated with an increased risk of SAEs.13,14,16 In addition, the risk of developing Clostridium difficile overgrowth was smaller than 1% in previous studies with ceftriaxone, and there was no association with an increase in antimicrobial resistance.16,17 PRECIOUS uses paracetamol for the prevention of increases in body temperature because this was safe in doses up to 6 g per day in randomised clinical trials in patients with acute stroke, reduced the risk of subfebrile temperatures or fever at 24 h by 50% and was associated with a trend towards an improvement in functional outcome in the PAIS trial. This trial was underpowered to detect a benefit on functional outcome because this was terminated prematurely due to lack of funding after inclusion of 1400 patients, against a target of 2500 patients.10 For PRECIOUS, we have selected a maximum daily dose of 4 g to comply with the drug’s summary of product characteristics. Given the potential benefit of the prevention of complications to the patients included in PRECIOUS, future stroke patients, their caregivers, and society, the risk-benefit balance is strongly in favour of conducting this clinical trial.

PRECIOUS uses paracetamol for the prevention of increases in body temperature because this was safe in doses up to 6 g per day in randomised clinical trials in patients with acute stroke, reduced the risk of subfebrile temperatures or fever at 24 h by 50% and was associated with a trend towards an improvement in functional outcome in the PAIS trial. This trial was underpowered to detect a benefit on functional outcome because this was terminated prematurely due to lack of funding after inclusion of 1400 patients, against a target of 2500 patients.10 For PRECIOUS, we have selected a maximum daily dose of 4 g to comply with the drug’s summary of product characteristics. Given the potential benefit of the prevention of complications to the patients included in PRECIOUS, future stroke patients, their caregivers, and society, the risk-benefit balance is strongly in favour of conducting this clinical trial. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Philip Bath is Stroke Association Professor of Stroke Medicine and is an NIHR Senior Investigator; Janika Kõrv has received fees as a consultant or lecture fees from Bayer, Pfizer, Boehringer Ingelheim, ReNeuron; Götz Thomalla has received fees as a consultant or lecture fees from Acandis, Bayer Vital, Bristol-Myers Squibb/Pfizer, Boehringer Ingelheim, Daichii Sankyo, GlaxoSmithKline, and Stryker and received a research grant from Bayer. Bart van der Worp has received speaker’s fees from Boehringer Ingelheim and Bayer; the other authors report no conflict of interest.

ultant or lecture fees from Acandis, Bayer Vital, Bristol-Myers Squibb/Pfizer, Boehringer Ingelheim, Daichii Sankyo, GlaxoSmithKline, and Stryker and received a research grant from Bayer. Bart van der Worp has received speaker’s fees from Boehringer Ingelheim and Bayer; the other authors report no conflict of interest. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: PRECIOUS has received funding from the European Union’s Horizon, 2020 research and innovation programme under grant agreement No 634809. Ethical approval The trial has been approved by national or local research ethics boards for all active clinical sites. Informed consent Not applicable Guarantor HBW. Contributorship All co-authors have contributed to the trial design, protocol development, and the writing of this manuscript.

Introduction Stroke is the second most common single cause of death in Europe. It is responsible for more than one million deaths per year and is the leading cause of long-term disability.1,2 Twenty to 35% of stroke patients die within the first month after a stroke, and up to one-third of survivors lose their independence.3,4,5 Consequently, the socioeconomic impact of stroke is considerable: the annual cost of stroke in Europe is estimated to be €45 billion: €20 billion for direct care, €9 billion related to loss of productivity and €16 billion for informal care.1 The burden of stroke differs widely across Europe. Specifically, the incidence and case fatality of stroke in central and eastern European countries remain higher than in northern, western and southern European countries.1,6–9

r direct care, €9 billion related to loss of productivity and €16 billion for informal care.1 The burden of stroke differs widely across Europe. Specifically, the incidence and case fatality of stroke in central and eastern European countries remain higher than in northern, western and southern European countries.1,6–9 The main pillars of acute ischaemic stroke treatment are stroke unit (SU) care and treatments promoting revascularisation. SUs, defined as dedicated areas or wards in hospitals open round the clock, where stroke patients are admitted and cared for by a multidisciplinary team including medical, nursing and therapy staff, are most effective in reducing mortality and morbidity.10–12 Since 2002, intravenous thrombolysis (IVT) has been approved in Europe for use in acute stroke treatment.13 In 2015, following randomised controlled trials demonstrating that endovascular treatment (EVT) dramatically improves the outcome of stroke patients with large vessel occlusion in the anterior circulation, European guidelines recommended the use of EVT in comprehensive stroke centres (CSCs).14–16 However, pan-European data on the access to and delivery of SU care, IVT and EVT are lacking. Therefore, the European Stroke Organisation (ESO), the European Society of Minimally Invasive Neurological Therapy (ESMINT), the European Academy of Neurology (EAN) and the Stroke Alliance for Europe (SAFE) surveyed the access to and delivery rates of acute SU care, IVT and EVT throughout Europe.

The main pillars of acute ischaemic stroke treatment are stroke unit (SU) care and treatments promoting revascularisation. SUs, defined as dedicated areas or wards in hospitals open round the clock, where stroke patients are admitted and cared for by a multidisciplinary team including medical, nursing and therapy staff, are most effective in reducing mortality and morbidity.10–12 Since 2002, intravenous thrombolysis (IVT) has been approved in Europe for use in acute stroke treatment.13 In 2015, following randomised controlled trials demonstrating that endovascular treatment (EVT) dramatically improves the outcome of stroke patients with large vessel occlusion in the anterior circulation, European guidelines recommended the use of EVT in comprehensive stroke centres (CSCs).14–16 However, pan-European data on the access to and delivery of SU care, IVT and EVT are lacking. Therefore, the European Stroke Organisation (ESO), the European Society of Minimally Invasive Neurological Therapy (ESMINT), the European Academy of Neurology (EAN) and the Stroke Alliance for Europe (SAFE) surveyed the access to and delivery rates of acute SU care, IVT and EVT throughout Europe. Objectives We aimed to collect national data on access to and delivery rates for acute SU care, IVT and EVT throughout Europe. We compared both pooled data of all countries and individual country rates with highest country rates. We also aimed to estimate the number of SUs necessary to cover the needs of the European population and to calculate how many patients could be treated with IVT and EVT if best practice were to be followed in all European countries.

h pooled data of all countries and individual country rates with highest country rates. We also aimed to estimate the number of SUs necessary to cover the needs of the European population and to calculate how many patients could be treated with IVT and EVT if best practice were to be followed in all European countries. Methods Study design and participants This pan-European study surveyed European chairs of national stroke societies and stroke experts on the access to and delivery rates of acute SU care, IVT and EVT in their countries. The study’s steering committee consisted of 10 representatives from ESO, ESMINT, EAN and SAFE, including at least 1 board member from each society. The survey was pre-announced six months before it started in the ESO, ESMINT, and EAN newsletters. The World Health Organization (WHO)’s definition of the European region as including 51 countries was adopted. Countries with less than 100,000 inhabitants (Monaco, Liechtenstein, Andorra, and San Marino) were excluded, and North Cyprus, which was not listed under the WHO definition, was added. National stroke society chairs were invited to be country coordinators by the steering committee. In the absence of a national stroke or neurological society, national coordinators were selected by the steering committee, following a thorough review of proposals proffered by board members and/or country representatives of ESO, EAN or ESMINT. Each country coordinator was responsible for selecting two additional national stroke experts, one of whom was a neurointerventionalist, whenever possible. Coordinators and experts were responsible for identifying the most reliable and most recent national data sources (i.e. stroke registries, governmental data sources, etc.) to answer the survey questions. In the absence of national or local stroke registries, the coordinator and experts were asked to perform best estimates by consensus, and took full responsibility for the validity of the responses provided. The affiliations of the 44 coordinators and 58 experts are shown in Supplementary Appendix 1.

e survey questions. In the absence of national or local stroke registries, the coordinator and experts were asked to perform best estimates by consensus, and took full responsibility for the validity of the responses provided. The affiliations of the 44 coordinators and 58 experts are shown in Supplementary Appendix 1. Data collection The survey was drafted by the steering committee after a series of meetings, audio conferences and e-mail correspondence. Thereafter, it was externally reviewed by four stroke experts with extensive research skills and proven experience in drafting European stroke surveys (see Acknowledgments and Supplementary Appendix 1). The survey consisted of 64 items (Supplementary Appendix 2). The first section collected information on the professional background of the coordinators and experts. Subsequent sections dealt with stroke care and prehospital stroke care pathways, acute stroke treatment strategies, information on stroke registries and quality control, as well as stroke awareness campaigns. The definition of an acute SU was based on local and/or national definitions. A pilot survey was performed in Austria, Hungary, Poland, Spain and Switzerland to assess feasibility. The survey was performed between 30 October 2016 and 24 February 2017. Collected data were independently reviewed by two authors (RVM, DAS). Whenever there was ambiguity and/or missing or conflicting responses, the steering committee requested clarifications, and final approvals were granted by the steering committee.

The survey was performed between 30 October 2016 and 24 February 2017. Collected data were independently reviewed by two authors (RVM, DAS). Whenever there was ambiguity and/or missing or conflicting responses, the steering committee requested clarifications, and final approvals were granted by the steering committee. Data analyses Our analyses focused on access to and delivery rates of acute SU care, IVT and EVT as well as the number of centres delivering IVT and EVT. All data were analysed using appropriate descriptive methods. Whenever a response was ‘unknown’, it was eliminated. We calculated crude rates of acute SUs per one million inhabitants using United Nations population estimates (2015 Revision of World Population Prospects),17 with the exception of Serbia, for which estimates from the Statistical Office of the Republic of Serbia were used (not including Kosovo and Metohija). Similar calculations were done for annual numbers of IVT and EVT performed, and number of centres providing IVT and EVT within each country. The annual incidence of ischaemic stroke per country was based on the estimates from the Global Burden of Disease Report (2016).8 These data were used to test the associations between SUs, IVT and EVT rates per million population using Pearson correlation. We calculated the number of acute SUs necessary to reach the defined target of three acute SUs per one million inhabitants or one SU per 1000 annual incident strokes.18,19 We also calculated how many additional patients could be treated if an IVT rate of 18% could be achieved in all countries, and similar calculations were done for patients treated with EVT if an EVT rate of 5% could be achieved in all countries. Both cut-offs were data driven and were based on the next lower whole-numbered value of the three countries with highest rates. The data obtained from the survey were collated and analysed in Microsoft Excel, version 2010 (Microsoft Corporation, Redmond, WA, USA). Selected variables were imported into Stata 14 (StataCorp, College Station, TX, USA) for further analysis.

e next lower whole-numbered value of the three countries with highest rates. The data obtained from the survey were collated and analysed in Microsoft Excel, version 2010 (Microsoft Corporation, Redmond, WA, USA). Selected variables were imported into Stata 14 (StataCorp, College Station, TX, USA) for further analysis. Best practice We compared both pooled and individual national data per one million inhabitants and per annual incident ischaemic stroke rates regarding access to and delivery of SU care and IVT and EVT therapies with highest country rates. Patient involvement A representative from a patient organisation (SAFE) was involved in all stages of the research project (i.e. generating hypothesis, design of the survey, interpretation of the results and critical revision of the manuscript). This patient representation is intended to insure sufficient integration of the patient perspective, and better dissemination of results to patients and next of kin. Results Overall, 44/51 invited countries participated (Supplementary Figure 1). Seven countries did not respond to multiple attempts at contact. The total number of inhabitants in the 44 participating countries was estimated to be 835 million.17 Overall, 22 (50%) surveys were completed by a coordinator and two experts, 9 (20%) by a coordinator and one expert, and 13 (30%) by a coordinator alone.

. Seven countries did not respond to multiple attempts at contact. The total number of inhabitants in the 44 participating countries was estimated to be 835 million.17 Overall, 22 (50%) surveys were completed by a coordinator and two experts, 9 (20%) by a coordinator and one expert, and 13 (30%) by a coordinator alone. Acute SUs, IVT hospitals and EVT centres Information on acute SU care was provided for 42/44 countries. National requirements for SUs were defined in 29 countries (68%) and planned in seven countries (16%) (Supplementary Figure 2). Overall, there were 2139 acute SUs in 42 countries, corresponding to a pooled mean of 2.9 SUs per million inhabitants (95% CI; 2.3–3.6) (Table 1, Supplementary Figure 3) and 1.5 (95% CI; 1.1–1.9) per 1000 annual incident ischaemic strokes (Table 2, Supplementary Figure 6). There was a considerable heterogeneity among the 42 countries (Figures 1 and 2). Countries with highest rates had 9.2 acute SUs per one million population and 5.8 per 1000 annual incident ischaemic strokes. Ten countries had less than one acute SU per one million inhabitants. Table 1. Absolute and relative numbers per million inhabitants of stroke units, annual number of intravenous thrombolysis treatments (IVT), number of IVT centres, annual number of endovascular treatments (EVT) and EVT centres per country. Country No. of stroke unitsa Stroke units per million Annual no. of IVTb Annual no. of IVT per millionb No. of IVT centres IVT centres per million Annual no. of EVTc Annual no. of EVT per millionc No. of EVT centres No. of EVT centres 24/7 EVT centres per million

Table 1. Absolute and relative numbers per million inhabitants of stroke units, annual number of intravenous thrombolysis treatments (IVT), number of IVT centres, annual number of endovascular treatments (EVT) and EVT centres per country. Country No. of stroke unitsa Stroke units per million Annual no. of IVTb Annual no. of IVT per millionb No. of IVT centres IVT centres per million Annual no. of EVTc Annual no. of EVT per millionc No. of EVT centres No. of EVT centres 24/7 EVT centres per million Albania 1 0.3 0 0.0 0 0.0 6 2.1 1 0 0.3 Austria 38 4.4 3000 351.1 38 4.4 650 76.1 11 9 1.3 Belarus – – – – 12 1.3 – – 1 1 0.1 Belgium – – 2250 199.1 – – 706 62.5 17 12 1.5 Bosnia andHerzegovina 4 1.0 150 39.4 4 1.0 50 13.1 1 1 0.3 Bulgaria 30 4.2 296 41.4 28 3.9 34 4.8 4 – 0.6 Croatia 18 4.2 300 70.7 20 4.7 30 7.1 2 1 0.5 Czech Republic 45 4.3 3800 360.4 45 4.3 1063 100.8 15 15 1.4 Denmark 20 3.5 1632 287.9 10 1.8 280 49.4 3 3 0.5 Estonia 4 3.0 541 412.2 6 4.6 109 83.0 3 2 2.3 Finland 21 3.8 1600 290.7 21 3.8 407 74.0 5 5 0.9 France 140 2.2 8000 124.2 140 2.2 4589 71.3 37 – 0.6 Georgia 2 0.5 4 1.0 1 0.3 0 0.0 0 0 0.0 Germany 295 3.7 30,000 371.8 350 4.3 9000 111.5 135 110 1.7 Greece 5 0.4 225 20.5 18 1.6 40 3.7 6 1 0.5 Hungary 39 4.0 2000 202.9 39 4.0 265 26.9 6 3 0.6 Iceland 1 3.0 30 91.1 5 15.2 0 0.0 0 0 0.0 Ireland 22 4.7 550 117.3 24 5.1 210 44.8 2 1 0.4 Israel 10 1.2 1200 148.8 20 2.5 300 37.2 9 9 1.1 Italy 178 3.0 8000 133.8 182 3.0 1882 31.5 50 16 0.8 Kyrgyzstan 5 0.8 0 0.0 0 0.0 0 0.0 0 0 0.0 Latvia 7 3.6 540 274.0 7 3.6 95 48.2 2 1 1.0 Lithuania 8 2.8 699 242.8 10 3.5 276 95.9 6 3 2.1 Luxembourg 3 5.3 30 52.9 3 5.3 3 5.3 1 – 1.8

Ireland 22 4.7 550 117.3 24 5.1 210 44.8 2 1 0.4 Israel 10 1.2 1200 148.8 20 2.5 300 37.2 9 9 1.1 Italy 178 3.0 8000 133.8 182 3.0 1882 31.5 50 16 0.8 Kyrgyzstan 5 0.8 0 0.0 0 0.0 0 0.0 0 0 0.0 Latvia 7 3.6 540 274.0 7 3.6 95 48.2 2 1 1.0 Lithuania 8 2.8 699 242.8 10 3.5 276 95.9 6 3 2.1 Luxembourg 3 5.3 30 52.9 3 5.3 3 5.3 1 – 1.8 FYROM 1 0.5 60 28.9 1 0.5 2 1.0 1 0 0.5 Malta 1 2.4 40 95.5 2 4.8 30 71.7 1 1 2.4 Montenegro 0 0.0 20 32.0 3 4.8 0 0.0 0 0 0.0 Netherlands 85 5.0 4821 284.8 85 5.0 1088 64.3 20 20 1.2 Northern Cyprus 2 6.4 30 95.7 2 6.4 20 63.8 2 2 6.4 Norway 48 9.2 1100 211.1 49 9.4 150 28.8 5 – 1.0 Poland 171 4.4 6493 168.2 171 4.4 175 4.5 21 4 0.5 Portugal 25 2.4 1516 146.5 25 2.4 845 81.6 9 4 0.9 Republic of Moldova 2 0.5 45 11.1 2 0.5 20 4.9 0 0 0.0 Romania 10 0.5 200 10.3 10 0.5 15 0.8 2 0 0.1 Russia 451 3.1 11,651 81.2 451 3.1 260 1.8 134 37 0.9 Serbia 10 1.4 400 56.8 20 2.8 10 1.4 4 3 0.6 Slovakia 36 6.6 972 179.1 44 8.1 385 71.0 7 3 1.3 Slovenia 3 1.5 400 193.5 12 5.8 142 69.2 1 1 0.5 Spain 60 1.3 5002 108.5 59 1.3 2408 52.2 37 35 0.8 Sweden 72 7.4 2600 265.9 72 7.4 390 39.9 6 3 0.6 Switzerland 23 2.8 1000 120.5 23 2.8 626 75.4 9 9 1.1 Turkey 33 0.4 1480 18.8 41 0.5 456 5.8 21 10 0.3 Ukraine 10 0.2 300 6.7 27 0.6 10 0.2 4 – 0.6 United Kingdom 200 3.1 10,290 159.0 200 3.1 478 7.4 28 2 0.1 Total /Mean [95% CI] 2139 2.9[2.3–3.6] 113,267 142.0[107.4–176.7] 2282 3.6[2.7–4.4] 27,505 37.1[26.7–47.5] 629 327 0.9[0.6–1.2] FYROM: former Yugoslav Republic of Macedonia. aAccording to the local definition of stroke unit. bAnnual number of intravenous thrombolysis in 2015 or 2016.

United Kingdom 200 3.1 10,290 159.0 200 3.1 478 7.4 28 2 0.1 Total /Mean [95% CI] 2139 2.9[2.3–3.6] 113,267 142.0[107.4–176.7] 2282 3.6[2.7–4.4] 27,505 37.1[26.7–47.5] 629 327 0.9[0.6–1.2] FYROM: former Yugoslav Republic of Macedonia. aAccording to the local definition of stroke unit. bAnnual number of intravenous thrombolysis in 2015 or 2016. cAnnual number of endovascular treatments (EVT) for acute ischaemic stroke performed in 2016, except Denmark, Sweden and Malta (2015). Table 2. Relative numbers of stroke units, annual number of intravenous thrombolysis (IVT) treatments, IVT centres, annual endovascular treatment (EVT) and EVT centres per annual incident ischaemic strokes.a Country Annual no. of incident ischaemic strokesa Stroke units per 1000 ischaemic strokes Annual no. of IVTs per 1000 ischaemic strokes Proportion of stroke patients with IVT (%) IVT hospitals per 1000 ischaemic strokes Annual no. of EVTs per 1000 ischaemic strokes Proportion of ischaemic stroke patients with EVT (%) EVT centres per 1000 ischaemic strokes Albania 4815 0.2 0.0 0.0 0.0 1.2 0.1 0.2 Austria 16,314 2.3 183.9 18.4 2.3 39.8 4.0 0.7 Belarus 30,056 – – – 0.4 – - 0.0 Belgium 19,689 – 114.3 11.4 – 35.9 3.6 0.9 Bosnia and Herzegovina 13,580 0.3 11.0 1.1 0.3 3.7 0.4 0.1 Bulgaria 28,869 1.0 10.3 1.0 1.0 1.2 0.1 0.1 Croatia 20,278 0.9 14.8 1.5 1.0 1.5 0.1 0.1 Czech Republic 30,877 1.5 123.1 12.3 1.5 34.4 3.4 0.5 Denmark 8339 2.4 195.7 19.6 1.2 33.6 3.4 0.4 Estonia 3757 1.1 144.0 14.4 1.6 29.0 2.9 0.8 Finland 12,685 1.7 126.1 12.6 1.7 32.1 3.2 0.4 France 87,372 1.6 91.6 9.2 1.6 52.5 5.3 0.4 Georgia 8973 0.2 0.4 0.0 0.1 0.0 0.0 0.0

Bulgaria 28,869 1.0 10.3 1.0 1.0 1.2 0.1 0.1 Croatia 20,278 0.9 14.8 1.5 1.0 1.5 0.1 0.1 Czech Republic 30,877 1.5 123.1 12.3 1.5 34.4 3.4 0.5 Denmark 8339 2.4 195.7 19.6 1.2 33.6 3.4 0.4 Estonia 3757 1.1 144.0 14.4 1.6 29.0 2.9 0.8 Finland 12,685 1.7 126.1 12.6 1.7 32.1 3.2 0.4 France 87,372 1.6 91.6 9.2 1.6 52.5 5.3 0.4 Georgia 8973 0.2 0.4 0.0 0.1 0.0 0.0 0.0 Germany 171,801 1.7 174.6 17.5 2.0 52.4 5.2 0.8 Greece 22,182 0.2 10.1 1.0 0.8 1.8 0.2 0.3 Hungary 32,335 1.2 61.9 6.2 1.2 8.2 0.8 0.2 Iceland 419 2.4 71.6 7.2 11.9 0.0 0.0 0.0 Ireland 4809 4.6 114.4 11.4 5.0 43.7 4.4 0.4 Israel 7411 1.3 161.9 16.2 2.7 40.5 4.0 1.2 Italy 108,811 1.6 73.5 7.4 1.7 17.3 1.7 0.5 Kyrgyzstan 5238 1.0 0.0 0.0 0.0 0.0 0.0 0.0 Latvia 10,282 0.7 52.5 5.3 0.7 9.2 0.9 0.2 Lithuania 12,539 0.6 55.7 5.6 0.8 22.0 2.2 0.5 Luxembourg 700 4.3 42.9 4.3 4.3 4.3 0.4 1.4 FYROM 5939 0.2 10.1 1.0 0.2 0.3 0.0 0.2 Malta 537 1.9 74.5 7.4 3.7 55.9 5.6 1.9 Montenegro 1315 0.0 15.2 1.5 2.3 0.0 0.0 0.0 Netherlands 23,458 3.6 205.5 20.6 3.6 46.4 4.6 0.9 Northern Cyprus – – – – – – – – Norway 8230 5.8 133.7 13.4 6.0 18.2 1.8 0.6 Poland 99,397 1.7 65.3 6.5 1.7 1.8 0.2 0.2 Portugal 18,208 1.4 83.3 8.3 1.4 46.4 4.6 0.5 Republic of Moldova 9592 0.2 4.7 0.5 0.2 2.1 0.2 0.0 Romania 79,247 0.1 2.5 0.3 0.1 0.2 0.0 0.0 Russia 538,984b 0.8 21.6 2.2 0.8 0.5 0.0 0.2 Serbia 31,632 0.3 12.6 1.3 0.6 0.3 0.0 0.1 Slovakia 16,273 2.2 59.7 6.0 2.7 23.7 2.4 0.4 Slovenia 4629 0.6 86.4 8.6 2.6 30.7 3.1 0.2 Spain 66,594 0.9 75.1 7.5 0.9 36.2 3.6 0.6 Sweden 17,361 4.1 149.8 15.0 4.1 22.5 2.2 0.3 Switzerland 12,896 1.8 77.5 7.8 1.8 48.5 4.9 0.7 Turkey 69,064 0.5 21.4 2.1 0.6 6.6 0.7 0.3 Ukraine 160,904 0.1 1.9 0.2 0.2 0.1 0.0 0.0

Serbia 31,632 0.3 12.6 1.3 0.6 0.3 0.0 0.1 Slovakia 16,273 2.2 59.7 6.0 2.7 23.7 2.4 0.4 Slovenia 4629 0.6 86.4 8.6 2.6 30.7 3.1 0.2 Spain 66,594 0.9 75.1 7.5 0.9 36.2 3.6 0.6 Sweden 17,361 4.1 149.8 15.0 4.1 22.5 2.2 0.3 Switzerland 12,896 1.8 77.5 7.8 1.8 48.5 4.9 0.7 Turkey 69,064 0.5 21.4 2.1 0.6 6.6 0.7 0.3 Ukraine 160,904 0.1 1.9 0.2 0.2 0.1 0.0 0.0 United Kingdom 87,594 2.3 117.5 11.7 2.3 5.5 0.5 0.3 Total /Mean [95% CI] 1,913,985 1.5[1.1–1.9] 72.7[54.2–91.2] 7.3%[5.4–9.1] 1.9[1.3–2.5] 19.3[13.5–25.1] 1.9%[1.3–2.5] 0.4[0.3–0.5] FYROM: former Yugoslav Republic of Macedonia. aAccording to Global Burden of Disease estimates (2016).8 bAccording to the local experts, the incidence of ischemic stroke is overestimated. Figure 1. Choropleth map showing number of stroke units per million population in 42 European countries (mean 2.9, 95% CI 2.3–3.6). Overall, 43/44 countries reported the number of hospitals delivering IVT (Table 1). IVT was performed at 2282 hospitals, corresponding to a mean number of 3.6 (95% CI; 2.7–4.4) IVT hospitals per one million inhabitants and 1.9 (95% CI; 1.3–2.5) per 1000 annual incident ischaemic strokes. Countries with highest rates had 15.2 IVT hospitals per one million population and 11.9 per 1000 annual incident ischaemic strokes.

d at 2282 hospitals, corresponding to a mean number of 3.6 (95% CI; 2.7–4.4) IVT hospitals per one million inhabitants and 1.9 (95% CI; 1.3–2.5) per 1000 annual incident ischaemic strokes. Countries with highest rates had 15.2 IVT hospitals per one million population and 11.9 per 1000 annual incident ischaemic strokes. All countries reported the number of centres delivering EVT (Table 1). EVT was performed at 629 stroke centres, corresponding to a mean number of 0.9 (95% CI; 0.6–1.2) EVT centres per one million inhabitants and 0.4 (95% CI; 0.3–0.5) per 1000 annual incident ischaemic strokes (Table 2). Countries with highest rates had 6.4 EVT centres per one million population and 1.9 per 1000 annual incident ischaemic strokes. Twenty-nine countries had less than one stroke centre capable of performing EVT per one million inhabitants. Centres offering EVT round the clock were provided by 33 countries (Table 1). According to coordinators and experts, the available centres met the needs for EVT in 15/43 countries. In eight of the countries currently lacking full EVT coverage, it was already being planned. In the remaining 20 countries, there were no plans to provide full national access to EVT in the near future. National protocols for EVT were available in 20 countries (Supplementary Table 1 and 2).

for EVT in 15/43 countries. In eight of the countries currently lacking full EVT coverage, it was already being planned. In the remaining 20 countries, there were no plans to provide full national access to EVT in the near future. National protocols for EVT were available in 20 countries (Supplementary Table 1 and 2). IVT Overall, 43/44 countries provided figures on annual IVTs performed. IVT was not available in 2/44 countries. The number of IVTs performed came from national IVT registries in 26 countries. The remaining countries provided data from national offices of statistics, service reports or estimates made by coordinators and experts. For all countries, data were from the years 2015 or 2016 (Tables 1 and 2). Overall, the total annual number of patients receiving IVT in 43 countries was 113,267. The estimated mean number of IVTs per one million inhabitants for these 43 countries was 142.0 (95% CI; 107.4–176.7) and 72.7 (95% CI; 54.2–91.2) per 1000 annual incident ischaemic strokes, while the highest country rates were 412.2 and 205.5, respectively (Supplementary Figure 4). In 10 countries, the estimated annual numbers of IVT treatments delivered per one million inhabitants were fewer than 50, whereas four countries had rates above 350 (Figure 3). Overall, 7.3% (95% CI; 5.4–9.1) of all patients with an ischaemic stroke in Europe received IVT, whereas 13 countries reported IVT rates of 10% or more (Table 2, Figure 4, Supplementary Figure 7).

treatments delivered per one million inhabitants were fewer than 50, whereas four countries had rates above 350 (Figure 3). Overall, 7.3% (95% CI; 5.4–9.1) of all patients with an ischaemic stroke in Europe received IVT, whereas 13 countries reported IVT rates of 10% or more (Table 2, Figure 4, Supplementary Figure 7). Figure 2. Choropleth map showing number of stroke units per 1000 annual incident ischaemic strokes in 42 European countries (mean 1.5; 95% CI 1.1–1.9). Figure 3. Choropleth map showing contemporary annual rates of intravenous thrombolysis (IVT) per million population in 42 European countries (mean 142.0, 95% CI 107.4–176.7). The two most frequent reasons for not performing IVT were late patient admission (27 countries) and a lack of personnel with stroke expertise on site (15 countries). Other reasons included a lack of immediate access to brain imaging (5 countries), and a lack of a round the clock SU and/or medical laboratory availability in five countries. In Kyrgyzstan, alteplase was scheduled to be approved for the treatment of acute ischaemic stroke treatment in the near future. The cost was singled out as a barrier to providing IVT in four countries. The number of acute SUs per one million population was significantly associated with the number of IVTs delivered per million population (Pearson correlation coefficient 0.54, p = 0.0002) (Supplementary Figure 9).

ke treatment in the near future. The cost was singled out as a barrier to providing IVT in four countries. The number of acute SUs per one million population was significantly associated with the number of IVTs delivered per million population (Pearson correlation coefficient 0.54, p = 0.0002) (Supplementary Figure 9). EVT EVT was available in 40/44 countries and 39 countries provided figures on annual numbers of EVTs performed. The number of EVTs performed came from national registries in 14 countries. The remaining countries provided data from national offices of statistics, service reports, extrapolation from locally obtained figures or estimates by coordinators and experts. For most countries, data were from 2016 (Tables 1 and 2). Overall, 27,505 procedures were performed, corresponding to a mean number of 37.1 procedures per one million inhabitants (95% CI; 26.7–47.5) and 19.3 (95% CI; 13.5–25.1) per 1000 annual incident ischaemic strokes, while highest country rates were 111.5 and 55.9, respectively (Supplementary Figure 5 and 8). The annual number of treatments delivered per million inhabitants was fewer than 10 per one million inhabitants in 13 countries, whereas three countries reported EVT rates above 100 per one million (Table 1, Figure 5). Overall, 1.9% (95% CI; 1.3–2.5) of all patients with an ischaemic stroke in Europe received EVT, 15 countries reported EVT rates of 3% or more (Table 2, Figure 6).

ximum of 24 h. The classification of nonfocal symptoms was adapted from the Rotterdam Study.1,2 Symptoms that were compatible with a different diagnosis, such as migraine or epilepsy, were excluded from the analysis. The researcher who completed the interview was aware of the presence or absence of CAO in each patient. Table 1. Occurrence of different types of nonfocal TNA symptoms in the past six months. Nonfocal symptoms, n (%) Carotid occlusion (n = 67) No carotid occlusion (n = 62) Blurred vision 6 (9%) 2 (3%) Bilateral leg weakness 4 (6%) 1 (2%) Unsteadiness 6 (9%) 0 (0%) Nonrotatory dizziness 18 (27%) 13 (21%) Paresthesias 0 (0%) 4 (7%) Unconsciousness 4 (6%) 0 (0%) Confusion 2 (3%) 0 (0%) Amnesia 1 (2%) 0 (0%) TNA: transient neurological attack. In addition, we recorded demographic and clinical characteristics, including the presence of vascular risk factors, use of antihypertensive, lipid lowering, antiplatelet or anticoagulant medication, blood pressure, side of symptomatic CAO and presence of other extracranial arterial stenosis.

nts was fewer than 10 per one million inhabitants in 13 countries, whereas three countries reported EVT rates above 100 per one million (Table 1, Figure 5). Overall, 1.9% (95% CI; 1.3–2.5) of all patients with an ischaemic stroke in Europe received EVT, 15 countries reported EVT rates of 3% or more (Table 2, Figure 6). Figure 4. Choropleth map showing contemporary annual estimates of the proportion of patients with incident ischaemic stroke treated with intravenous thrombolysis (IVT) in 42 European countries (mean 7.3%; 95% CI 5.4–9.1). The most common reasons reported for not providing EVT to all eligible patients were lack of specifically trained personnel (34 countries), lack of facilities (22 countries) and costs (16 countries). Overall, we observed that the number of EVTs delivered per one million population tended to be higher in countries with a greater number of SUs (Pearson correlation coefficient 0.36, p = 0.02). Discussion Our study shows that (1) for most European countries, access to and delivery of SUs, IVT and EVT are far below highest country rates and there are considerable inequalities among and within the different countries, (2) only 7.3% of all acute ischaemic stroke patients receive IVT and only 1.9% receive EVT, (3) and there is a significant correlation between the number of SUs per million inhabitants and delivery rates for both IVT and EVT.

ountry rates and there are considerable inequalities among and within the different countries, (2) only 7.3% of all acute ischaemic stroke patients receive IVT and only 1.9% receive EVT, (3) and there is a significant correlation between the number of SUs per million inhabitants and delivery rates for both IVT and EVT. Although there has been a dense network of SUs in northern European countries for the past two decades, no such network is present in most eastern and southern European countries. The European Brain Council has recently estimated that only one in three stroke patients in Europe has access to acute SU care.19 We have calculated that, in order to provide three SUs per one million inhabitants across Europe, at least 628 additional SUs would be needed in 20 countries (Table 3), and in order to reach one SU per 1000 annual incident ischaemic strokes, 447 SUs would be necessary in 18 countries. However, these calculations are rough estimates, based on the assumption that existing SUs are equally distributed within countries and that there is a uniform population distribution throughout Europe. In remote rural areas, more SUs might be necessary to meet the needs of the population, whereas in urban areas, high volume SUs can care for more stroke patients. Furthermore, we did not take into consideration the number of beds in the SUs and some SUs are likely to be too small to meet the needs of patients in their area. In line with the results regarding the availability of SUs, the number of EVT centres per country also varies significantly throughout Europe; 28 countries did not reach the benchmark of one CSC per one million inhabitants.20 To achieve this goal, at least 286 more CSCs delivering EVT would be needed in these 28 countries (Table 3). It is also noteworthy that some countries, as Albania, Bulgaria, Greece, Luxembourg, FYROM, Romania, Russia, Serbia and Ukraine, had a very low mean annual number of EVT per CSC. Although we recognise that in some cases this might be accounted by a low but growing case volume in recently created centres in 2016, which was a year of change in many countries regarding the organisation of EVT centres, this also indicates the need to implement basic requirements for CSC.

nual number of EVT per CSC. Although we recognise that in some cases this might be accounted by a low but growing case volume in recently created centres in 2016, which was a year of change in many countries regarding the organisation of EVT centres, this also indicates the need to implement basic requirements for CSC. Table 3. Estimated number of additional stroke units and comprehensive stroke centres required to achieve three stroke units per one million inhabitants and one comprehensive stroke centre per one million inhabitants, and estimated number of additional treatments with intravenous thrombolysis and endovascular interventions if rates of 18% and 5%, respectively, are to be achieved. Country No. of additional stroke units required No. of additional comprehensive stroke centres required No. of additional intravenous thrombolysis treatments per year(target rate 18%) No. of additional endovascular treatments per year(target rate 5%) Albania 8 2 867 235 Austria 0 0 0 166 Belarus – 8 – – Belgium – 0 1294 278 Bosnia and Herzegovina 7 3 2294 629 Bulgaria 0 3 4900 1409 Croatia 0 2 3350 984 Czech Republic 0 0 1758 481 Denmark 0 3 0 137 Estonia 0 0 135 79 Finland 0 1 683 227 France 53 27 7727 0 Georgia 10 4 1611 449 Germany 0 0 924 0 Greece 28 5 3768 1069 Hungary 0 4 3820 1352 Iceland 0 0 45 21 Ireland 0 3 316 30 Israel 14 0 134 71 Italy 1 10 11,586 3559 Kyrgyzstan 13 6 943 262 Latvia 0 0 1311 419 Lithuania 1 0 1558 351 Luxembourg 0 0 96 32 FYROM 5 1 1009 295 Malta 0 0 57 0 Montenegro 2 1 217 66 Netherlands 0 0 0 85 Northern Cyprus 0 0 – – Norway 0 0 381 262 Poland 0 18 11,398 4795 Portugal 6 1 1761 65 Republic of Moldova 10 4 1682 460 Romania 49 18 14,064 3947

Ireland 0 3 316 30 Israel 14 0 134 71 Italy 1 10 11,586 3559 Kyrgyzstan 13 6 943 262 Latvia 0 0 1311 419 Lithuania 1 0 1558 351 Luxembourg 0 0 96 32 FYROM 5 1 1009 295 Malta 0 0 57 0 Montenegro 2 1 217 66 Netherlands 0 0 0 85 Northern Cyprus 0 0 – – Norway 0 0 381 262 Poland 0 18 11,398 4795 Portugal 6 1 1761 65 Republic of Moldova 10 4 1682 460 Romania 49 18 14,064 3947 Russia 0 9 85,366a 26,689a Serbia 11 3 5294 1572 Slovakia 0 0 1957 429 Slovenia 3 1 433 89 Spain 78 9 6985 922 Sweden 0 4 525 478 Switzerland 2 0 1321 19 Turkey 203 58 10,952 2997 Ukraine 124 41 28,663 8035 United Kingdom 0 37 5477 3902 Total 628 286 226,662 67,347 FYROM: former Yugoslav Republic of Macedonia. aAccording to the local experts, the incidence of ischemic stroke is overestimated. Figure 5. Choropleth map showing contemporary annual rates of endovascular treatments (EVT) for ischaemic stroke per million population in 43 European countries (mean 37.1, 95% CI 26.7–47.5). Figure 6. Choropleth map showing contemporary annual estimates of the proportion of patients with incident ischaemic stroke receiving endovascular treatment (EVT) in 42 European countries (mean 1.9%; 95% CI 1.3–2.5).

Figure 5. Choropleth map showing contemporary annual rates of endovascular treatments (EVT) for ischaemic stroke per million population in 43 European countries (mean 37.1, 95% CI 26.7–47.5). Figure 6. Choropleth map showing contemporary annual estimates of the proportion of patients with incident ischaemic stroke receiving endovascular treatment (EVT) in 42 European countries (mean 1.9%; 95% CI 1.3–2.5). Governments, health care professionals and stroke specialists should now carefully plan and implement networks of SUs providing IVT and CSCs delivering EVT, taking into account geographical conditions and regional and local characteristics while still being affordable. CSCs will require round the clock stroke physicians with expertise in endovascular stroke management, and fully trained neurointerventionalists with qualifications based on current models of certification.21,22 Too many competing EVT centres will not be cost-effective, will compromise quality and will dilute the expertise of specialists. Universal adoption of a uniform definition of SUs and CSCs together with an appropriate certification process would help to guarantee the achievement of set quality targets throughout Europe.

Too many competing EVT centres will not be cost-effective, will compromise quality and will dilute the expertise of specialists. Universal adoption of a uniform definition of SUs and CSCs together with an appropriate certification process would help to guarantee the achievement of set quality targets throughout Europe. IVT and EVT Overall, 7.3% of all patients received IVT and 1.9% EVT. So far, population-based studies on IVT and EVT rates are scarce and there are no established and well-accepted benchmarks for determining what proportion of patients with acute ischaemic stroke should receive IVT and EVT. The burden of stroke report provided thrombolysis rates in national and large regional audits, but the denominator of populations varied between ‘all stroke patients’, ‘all ischaemic stroke patients’, ‘hospitalised stroke patients’, ‘stroke unit patients’ and others, making reliable comparisons impossible.9 Furthermore, using hospitalised stroke patients as the denominator can be misleading, since hospitalisation rates for stroke vary across Europe.9 In the same report, average annual thrombolysis rates per 100,000 population were only provided for 11 countries, within different time frames. The denominators of our survey were the population at risk and the number of incident ischaemic strokes, based on the 2015 Global Burden of Disease Report.8 Highest practice rate for IVT was 20.6%, whereas 13 countries had IVT rates of 10% or more. Highest EVT rates were 5%, with 15 countries achieving 3% or more. The estimated number of additional patients who could be treated with IVT taking 18% as the threshold is 226,662, and the estimated number of additional patients who could be treated with EVT at a threshold of 5% is 67,347 (Table 3). However, these are still rather conservative estimates, since even the IVT and EVT rates in countries with the highest rates can be improved. In Germany, for instance, only 60% of eligible patients were reported to have been treated by IVT in 2012.9 Based on the available evidence from intervention trials and prospective registries for EVT, a recent study in the UK has estimated that approximately 10% of stroke patients admitted to hospital were eligible for EVT.23

any, for instance, only 60% of eligible patients were reported to have been treated by IVT in 2012.9 Based on the available evidence from intervention trials and prospective registries for EVT, a recent study in the UK has estimated that approximately 10% of stroke patients admitted to hospital were eligible for EVT.23 Implications Our findings have implications for the future organisation of acute stroke care in Europe as we have shown in which countries rates of SUs, IVT and EVT are currently below best practice. In 2006, the second Helsingborg Declaration, co-sponsored by the WHO Regional Office for Europe, set new targets for stroke management and care.24 One specific goal for 2015 was that all patients with acute stroke who were potentially eligible for acute, specific treatment, were transferred to hospitals with adequate capacity and expertise to administer treatments – defined as a SU or stroke dedicated area. Our survey shows that many patients still have no access to appropriate acute stroke treatment and a clear action plan to address these inequalities is urgently needed. Therefore, a third Action Plan for Stroke in Europe is currently being drafted to set achievement goals for 2030 and the present survey will provide important information for future health care planning. By increasing the awareness of what needs to change, this knowledge may also be useful for local and national stroke patient and professional organisations, in order to develop tailored approaches that consider the specific barriers faced by national healthcare systems and take the first vital steps to improve overall stroke care. To develop a successful strategy for change, you need to understand the types of barriers faced in healthcare. Using this data, you can consider which barriers and levers may operate in your organisation and which may be relevant to a particular problem. Following careful analysis, it is possible to develop a tailored approach to overcome the barriers, encourage changes in behaviour and ultimately implement guidance.

ealthcare. Using this data, you can consider which barriers and levers may operate in your organisation and which may be relevant to a particular problem. Following careful analysis, it is possible to develop a tailored approach to overcome the barriers, encourage changes in behaviour and ultimately implement guidance. Strengths and limitations A major strength of our survey – apart from the large number of participating countries – is that data were mostly based on national stroke registries. Furthermore, the survey questions were answered by experienced stroke and neurointerventional experts. Our survey does, however, have several important limitations: (1) The population density is not reflected in the survey owing to the lack of well-established indicators for each individual European country. (2) Several countries lack prospective stroke registries and coordinators and experts had to use multiple sources of regional and local information to extrapolate national figures. (3) Given the lack of a uniform definition of SUs, some differences in SU rates are likely to be related to differences in definitions. Furthermore, we did not assess the number of patients treated in SUs. (4) Estimated stroke incidence rates of the 2015 Global Burden of Disease Report – the most accepted global data on stroke incidence – might eventually be too high for some countries, while for other countries, rates might be too low. We have therefore presented data both ways, per one million inhabitants and per 1000 annual incident ischaemic strokes. Finally, the methods used to collect information in countries lacking national stroke registries were not standardised and therefore in these cases, data should be seen as exploratory. The development of stroke registries, internal and external audits in those countries should be a goal, in order to improve the quality of data gathered. However, since in most cases this is not expected to happen in the near future, we do believe that communication of those preliminary results provides the opportunity to highlight the need for change and can lead to effective results.

n those countries should be a goal, in order to improve the quality of data gathered. However, since in most cases this is not expected to happen in the near future, we do believe that communication of those preliminary results provides the opportunity to highlight the need for change and can lead to effective results. Conclusions This is the first comprehensive survey on access to and delivery rates of acute SU care, IVT and EVT in Europe. We found major inequalities in treatment of acute stroke patients between and within 44 European countries, and in many countries rates are far below highest country rates, leaving many patients untreated. Individual country level data indicate where access to and delivery of acute stroke care is insufficient or lacking and allows comparison with highest country rates. These data will support governments, health care providers and European politicians to draw up action plans to implement acute stroke treatment, tailored for each European country, in order to reduce stroke-related mortality and morbidity in Europe. Supplemental Material Supplemental appendix - Supplemental material for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries Click here for additional data file.

Introduction Stroke is the second most common cause of death and one of the main causes of disability in the world.1 Therefore, a lot of research into treatment of strokes is carried out and as a result the field of acute stroke care is rapidly evolving. Intravenous thrombolysis (IVT), and the more recently introduced intra-arterial thrombectomy (IAT), both currently play an important part of the treatment of acute ischaemic stroke (AIS). In IVT as well as IAT, the treatment should be given in the first 4.5 or 6 h, respectively, and within these time frames as soon as possible (‘time is brain’).2 The faster these treatments are provided, the higher the odds for a favourable outcome.3,4 The treatment with IVT and IAT is logistically and technically complex and performed by a multidisciplinary and experienced team, consisting of neurologists, (intervention) radiologists, anaesthesiologists and nurses in often different hospitals. Therefore, a well-organised acute stroke care is a necessity.

Conclusions This is the first comprehensive survey on access to and delivery rates of acute SU care, IVT and EVT in Europe. We found major inequalities in treatment of acute stroke patients between and within 44 European countries, and in many countries rates are far below highest country rates, leaving many patients untreated. Individual country level data indicate where access to and delivery of acute stroke care is insufficient or lacking and allows comparison with highest country rates. These data will support governments, health care providers and European politicians to draw up action plans to implement acute stroke treatment, tailored for each European country, in order to reduce stroke-related mortality and morbidity in Europe. Supplemental Material Supplemental appendix - Supplemental material for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries Click here for additional data file. Supplemental material, Supplemental appendix for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries by Diana Aguiar de Sousa, Rascha von Martial, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora, Valery Feigin, Valeria Caso and Urs Fischer in European Stroke Journal

troke treatments: A survey of national scientific societies and stroke experts in 44 European countries by Diana Aguiar de Sousa, Rascha von Martial, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora, Valery Feigin, Valeria Caso and Urs Fischer in European Stroke Journal Supplemental Material Supplemental Table - Supplemental material for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries Click here for additional data file. Supplemental material, Supplemental Table for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries by Diana Aguiar de Sousa, Rascha von Martial, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora, Valery Feigin, Valeria Caso and Urs Fischer in European Stroke Journal Acknowledgements We thank Bo Norrving, Didier Leys, Kennedy Lees, M and Turgut Tatlisumak for their useful comments regarding the questionnaire, data collection methodology and policy implications of this paper. We thank Susan Edith Kaplan for editorial comments.

Supplemental material, Supplemental Table for Access to and delivery of acute ischaemic stroke treatments: A survey of national scientific societies and stroke experts in 44 European countries by Diana Aguiar de Sousa, Rascha von Martial, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora, Valery Feigin, Valeria Caso and Urs Fischer in European Stroke Journal Acknowledgements We thank Bo Norrving, Didier Leys, Kennedy Lees, M and Turgut Tatlisumak for their useful comments regarding the questionnaire, data collection methodology and policy implications of this paper. We thank Susan Edith Kaplan for editorial comments. “ESO ESMINT EAN SAFE Survey on Stroke Care in Europe” collaborators M Petrela, O Taka, E Enesi (Albania); S Kiechl, M Brainin, F Fazekas (Austria); P Konovalov (Belarus); A Peeters, G Vanhooren, L Defreyne (Belgium); H Haris, M Moranjkic (Bosnia and Herzegovina); S Andonova, I Staikov (Bulgaria); H Budincevic, D Ozretic, B Malojčić (Croatia); A Tomek (Czech Republic); H Iversen, SP Jonhsen (Denmark); J Korv, R Vibo, V Malikov (Estonia); T Sairanen, D Strbian, K Lappalainen (Finland); S Timsit, E Touze, D Leys (France); A Tsiskaridze, I Burduladze (Georgia); M Dichgans, J Röther, J Fiehler (Germany); G Tsivgoulis, E Brountzos (Greece); D Bereczki. C Óváry, I Szikora (Hungary); B Thorarinsson, V Vilmarsson (Iceland); J Harbison (Ireland); N Bornstein, D Tanne, A Horev (Israel); D Toni, S Mangiafico, V Caso (Italy); A Murzaliev, I Lutsenko, A Artykbaev, (Kyrgyzstan); E Miglãne, K Kupcs, I Kikule (Latvia); D Jatuzis, D Rastenyte, M Kurminas (Lithuania); D Ulbricht (Luxembourg); A Arsovska, M Smiceska, E Lickova (The former Yugoslav Republic of Macedonia); M Mallia, R Grech, M Schembri (Malta); V Lisnic, E Manole, E Zota (Moldova); M Cukic (Montenegro); D Dippel, C Majoie, H B van der Worp (Netherlands); S Akpinar (Northern Cyprus); E Berge, Hanne Ellekjær (Norway); A Czlonkowska, W Poncyljusz, A Kobayashi (Poland); E Azevedo, V Tedim-Cruz, M Ribeiro (Portugal); C Tiu (Romania); T Kharitonova (Russia); L Bumbasirevic (Serbia); Z Gdovinova, I Vulev, P Turcani (Slovakia); B Zvan, V Svigelj, Z Milosevic (Slovenia); J Gállego, M Rodríguez Yáñez, E Palacio-Portilla, A Gonzallez Mandly, M Gallofré, S Abilleira, C Gimenez, S Calleja-Puerta, J Marta-Moreno, M Óscar Ayo, A Lago (Spain); M Arnold (Switzerland); M Mazya, T Moreira (Sweden); N Uzuner, L Gungor, O Ozdemir (Turkey); H Rodgers, J Dawson, P White (United Kingdom); Y Flomin, D Shcheglov (Ukraine).

z Yáñez, E Palacio-Portilla, A Gonzallez Mandly, M Gallofré, S Abilleira, C Gimenez, S Calleja-Puerta, J Marta-Moreno, M Óscar Ayo, A Lago (Spain); M Arnold (Switzerland); M Mazya, T Moreira (Sweden); N Uzuner, L Gungor, O Ozdemir (Turkey); H Rodgers, J Dawson, P White (United Kingdom); Y Flomin, D Shcheglov (Ukraine). Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Diana Aguiar de Sousa, Rascha von Martial, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Valery Feigin and Valeria Caso: nothing to disclose. Sònia Abilleira: Co-Principal Investigator of the RACECAT trial (NCT02795962). Istvan Szikora: First Past President of the European Society of Minimally Invasive Neurological Therapy (ESMINT), investigator in the TREVO Registry, consultant to Medtronic Neurovascular, Stryker Neurovascular, Codman Neurovascular and Sequent Medical, outside the submitted work. Urs Fischer: Secretary General of the European Stroke Organisation. Received research grants from the Swiss National Science Foundation and the Swiss Heart Foundation. He is principal investigator of the SWITCH trial, the ELAN trial and the SWIFT DIRECT trial. Received consultancy for Medtronic/Covidien. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research received no specific grant from any funding agency in the public, commercial, or not-forprofit sectors.

Urs Fischer: Secretary General of the European Stroke Organisation. Received research grants from the Swiss National Science Foundation and the Swiss Heart Foundation. He is principal investigator of the SWITCH trial, the ELAN trial and the SWIFT DIRECT trial. Received consultancy for Medtronic/Covidien. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research received no specific grant from any funding agency in the public, commercial, or not-forprofit sectors. Ethical approval Not required. Informed consent The lead author (UF) affirms that this manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained. All authors, external and internal, had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. Guarantor Urs Fischer is the study guarantor. Contributorship Idea and concept: Urs Fischer, Valeria Caso. Study design: Urs Fischer, Valeria Caso, Diana Aguiar de Sousa, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora. Design of the questionnaire: Urs Fischer, Valeria Caso, Diana Aguiar de Sousa, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora. Literature search: Diana Aguiar de Sousa, Urs Fischer.

Study design: Urs Fischer, Valeria Caso, Diana Aguiar de Sousa, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora. Design of the questionnaire: Urs Fischer, Valeria Caso, Diana Aguiar de Sousa, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora. Literature search: Diana Aguiar de Sousa, Urs Fischer. Figures: Diana Aguiar de Sousa. Data collection: Rascha von Martial, Diana Aguiar de Sousa, Valery Feigin, Urs Fischer. Data analysis: Rascha von Martial, Diana Aguiar de Sousa. Data interpretation: Rascha von Martial, Diana Aguiar de Sousa, Urs Fischer, Sònia Abilleira, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora. Writing: Urs Fischer, Diana Aguiar de Sousa, Valeria Caso. Editorial comments: Susan Edith Kaplan. Relevant intellectual inputs to the manuscript: Sònia Abilleira, Rascha von Martial, Thomas Gattringer, Adam Kobayashi, Miquel Gallofré, Franz Fazekas, Istvan Szikora, Valery Feigin.

Introduction Nonfocal transient neurological attacks (TNAs) are defined as attacks with atypical, nonlocalizing cerebral symptoms such as unsteadiness, confusion or bilateral weakness.1,2 Patients with nonfocal TNAs are at higher risk of cardiac events, stroke and dementia than patients without nonfocal TNAs.2,3 Even though they are by no means benign, the underlying pathophysiology of nonfocal TNAs is still poorly understood. Both cerebral ischemia and cerebral hypoperfusion have been suggested to play a role.1,3–5 Factors that might contribute to cerebral hypoperfusion are severe obstruction of cerebropetal arteries and impaired cerebral autoregulation.6 We therefore hypothesized that in patients with carotid artery occlusion (CAO), nonfocal TNAs are more prevalent than in patients without CAO. Methods Study population and design We selected patients from two prospective observational cohort studies.7,8

Introduction Nonfocal transient neurological attacks (TNAs) are defined as attacks with atypical, nonlocalizing cerebral symptoms such as unsteadiness, confusion or bilateral weakness.1,2 Patients with nonfocal TNAs are at higher risk of cardiac events, stroke and dementia than patients without nonfocal TNAs.2,3 Even though they are by no means benign, the underlying pathophysiology of nonfocal TNAs is still poorly understood. Both cerebral ischemia and cerebral hypoperfusion have been suggested to play a role.1,3–5 Factors that might contribute to cerebral hypoperfusion are severe obstruction of cerebropetal arteries and impaired cerebral autoregulation.6 We therefore hypothesized that in patients with carotid artery occlusion (CAO), nonfocal TNAs are more prevalent than in patients without CAO. Methods Study population and design We selected patients from two prospective observational cohort studies.7,8 Patients with CAO were selected from the Heart-Brain Connection (HBC) study, which is a multicenter cohort study in the Netherlands that focusses on the relation between cardiovascular and hemodynamic measures and cognitive impairment.7 From all consecutive patients with CAO who participated in the HBC study from November 2014 to March 2017 (data release 1 October 2017), we included all patients with CAO who had experienced an ipsilateral transient ischemic attack (TIA) or nondisabling ischemic stroke more than six months before inclusion. CAO was defined as a complete occlusion of the internal carotid artery and was measured either with ultrasound, magnetic resonance angiography or computed tomography angiography.

CAO who had experienced an ipsilateral transient ischemic attack (TIA) or nondisabling ischemic stroke more than six months before inclusion. CAO was defined as a complete occlusion of the internal carotid artery and was measured either with ultrasound, magnetic resonance angiography or computed tomography angiography. Patients without CAO were selected from the Second Manifestations of ARTerial disease (SMART) study from November 2015 until September 2017. The SMART study is a cohort study executed by a single academic hospital in the Netherlands, including patients with vascular risk factors or clinical manifest vascular disease.8 Eligible patients had a history of TIA or nondisabling ischemic stroke of the anterior circulation that had occurred more than six months before inclusion and did not have a stenosis of the ipsilateral or contralateral carotid artery of ≥50% on ultrasound. Imaging of the internal carotid arteries was performed <1 year before inclusion into our study. In both groups, patients were independent in activities of daily living (modified Rankin Scale of 0–3). The HBC study and SMART study were approved by the Ethics Committee of the Leiden UMC and UMC Utrecht, respectively.7,8 All patients provided written informed consent.

Patients without CAO were selected from the Second Manifestations of ARTerial disease (SMART) study from November 2015 until September 2017. The SMART study is a cohort study executed by a single academic hospital in the Netherlands, including patients with vascular risk factors or clinical manifest vascular disease.8 Eligible patients had a history of TIA or nondisabling ischemic stroke of the anterior circulation that had occurred more than six months before inclusion and did not have a stenosis of the ipsilateral or contralateral carotid artery of ≥50% on ultrasound. Imaging of the internal carotid arteries was performed <1 year before inclusion into our study. In both groups, patients were independent in activities of daily living (modified Rankin Scale of 0–3). The HBC study and SMART study were approved by the Ethics Committee of the Leiden UMC and UMC Utrecht, respectively.7,8 All patients provided written informed consent. Study parameters All patients were interviewed by a trained physician or research nurse with a standardized questionnaire on the occurrence of eight nonfocal symptoms in the preceding six months (Table 1). Nonfocal TNAs were defined as attacks of one or more nonfocal signs or symptoms with an acute onset, a minimum duration of 30 s and a maximum of 24 h. The classification of nonfocal symptoms was adapted from the Rotterdam Study.1,2 Symptoms that were compatible with a different diagnosis, such as migraine or epilepsy, were excluded from the analysis. The researcher who completed the interview was aware of the presence or absence of CAO in each patient.

Nonfocal symptoms, n (%) Carotid occlusion (n = 67) No carotid occlusion (n = 62) Blurred vision 6 (9%) 2 (3%) Bilateral leg weakness 4 (6%) 1 (2%) Unsteadiness 6 (9%) 0 (0%) Nonrotatory dizziness 18 (27%) 13 (21%) Paresthesias 0 (0%) 4 (7%) Unconsciousness 4 (6%) 0 (0%) Confusion 2 (3%) 0 (0%) Amnesia 1 (2%) 0 (0%) TNA: transient neurological attack. In addition, we recorded demographic and clinical characteristics, including the presence of vascular risk factors, use of antihypertensive, lipid lowering, antiplatelet or anticoagulant medication, blood pressure, side of symptomatic CAO and presence of other extracranial arterial stenosis. Statistical analysis We calculated the prevalence of nonfocal TNAs in both patient groups. We assessed the association between CAO and the occurrence of nonfocal TNAs with Poisson regression analysis with robust standard errors.9 We calculated crude and adjusted risk ratios (RR) with corresponding 95% confidence intervals (CIs) for the occurrence of ≥1 and ≥2 different nonfocal TNAs. Adjustments were made for age, sex, systolic blood pressure and time interval between most recent TIA or ischemic stroke and questionnaire, as these factors were considered a priori to be potential confounders of the association between CAO and the occurrence of nonfocal TNAs. Subgroup analyses were performed in patients without steno-occlusion of the vertebral arteries, and in patients with CAO who also had had a contralateral carotid or vertebral artery steno-occlusion versus patients without any carotid or vertebral artery steno-occlusion.

between CAO and the occurrence of nonfocal TNAs. Subgroup analyses were performed in patients without steno-occlusion of the vertebral arteries, and in patients with CAO who also had had a contralateral carotid or vertebral artery steno-occlusion versus patients without any carotid or vertebral artery steno-occlusion. Results We included 67 patients with CAO and 62 patients without CAO. Mean age was 66 (±9) years. Patients with CAO were more often male, had more vascular risk factors and had higher systolic blood pressure than those without CAO (Table 2). Seven (11%) of patients with CAO had a contralateral asymptomatic carotid artery stenosis of ≥50% and six (9%) had a contralateral CAO. Fifteen (23%) patients with CAO also had >50% stenosis of a vertebral artery. Table 2. Patient characteristics.

Results We included 67 patients with CAO and 62 patients without CAO. Mean age was 66 (±9) years. Patients with CAO were more often male, had more vascular risk factors and had higher systolic blood pressure than those without CAO (Table 2). Seven (11%) of patients with CAO had a contralateral asymptomatic carotid artery stenosis of ≥50% and six (9%) had a contralateral CAO. Fifteen (23%) patients with CAO also had >50% stenosis of a vertebral artery. Table 2. Patient characteristics. Carotid occlusion (n = 67) No carotid occlusion (n = 62) Age in years, mean (SD) 65.3 (±7.9) 65.8 (±9.8) Male sex, n (%) 51 (76%) 37 (60%) Hypertension, n (%) 52 (78%) 39 (63%) Hyperlipidemia, n (%) 61 (91%) 24 (39%) Current smoking, n (%) 17 (25%) 11 (18%)a Diabetes mellitus, n (%) 21 (31%) 11 (18%) History of, n (%) Myocardial infarction 9 (13%) 5 (8%) Peripheral arterial disease 21 (31%) 2 (3%) TIA (including amaurosis fugax) 50 (75%) 22 (36%) Ischemic stroke 38 (57%) 42 (68%) Medication use, n (%) Antiplatelet 56 (84%) 48 (77%) Anticoagulant 6 (9%) 10 (16%) Antihypertensive 47 (70%) 32 (52%) Lipid lowering 59 (88%) 50 (81%) Systolic blood pressure in mmHg, mean (SD) 152.1 (±21.1) 136.3 (±15.0)b Diastolic blood pressure in mmHg, mean (SD) 82.7 (±12.1) 80.3 (±10.8)b Stenosis of extracranial cerebral arteries, n (%) Contralateral ICA stenosis ≥50% or occlusion 13 (20%)c NA VA stenosis ≥50% or occlusion 15 (23%)c 3 (5%) Side of symptomatic carotid occlusion, n (%) Right 31 (46%) NA Left 30 (45%) NA Both 6 (9%) NA Interval between most recent TIA or ischemic stroke and questionnaire in years, median (IQR) 3.7 (1.9–8.9) 1.8 (1.0–10.5) SD: standard deviation; TIA: transient ischemic attack; ICA: internal carotid artery; VA: vertebral artery; IQR: interquartile range; NA: not applicable.

31 (46%) NA Left 30 (45%) NA Both 6 (9%) NA Interval between most recent TIA or ischemic stroke and questionnaire in years, median (IQR) 3.7 (1.9–8.9) 1.8 (1.0–10.5) SD: standard deviation; TIA: transient ischemic attack; ICA: internal carotid artery; VA: vertebral artery; IQR: interquartile range; NA: not applicable. aMissing in n = 1 (2%). bBlood pressure was measured before administration of questionnaire (<1 year earlier). cMissing in n = 3 (5%). Table 1 summarizes the different nonfocal symptoms. Forty-three of all patients (33%) had had one or more nonfocal TNAs in the preceding six months (Supplementary Figure 1). The prevalence of at least one nonfocal TNA was not significantly different between patients with and without CAO (39% vs. 27%; adjusted RR 1.47, 95% CI 0.83–2.61; Table 3). However, the prevalence of two or more different nonfocal TNAs was significantly higher in patients with CAO (16% vs. 3%; adjusted RR 4.77, 95% CI 1.20–18.98; Table 3). Table 3. Occurrence of nonfocal TNAs in the past six months in patients with and without CAO, and in the subgroup of patients without vertebral artery stenosis or occlusion.

Table 1 summarizes the different nonfocal symptoms. Forty-three of all patients (33%) had had one or more nonfocal TNAs in the preceding six months (Supplementary Figure 1). The prevalence of at least one nonfocal TNA was not significantly different between patients with and without CAO (39% vs. 27%; adjusted RR 1.47, 95% CI 0.83–2.61; Table 3). However, the prevalence of two or more different nonfocal TNAs was significantly higher in patients with CAO (16% vs. 3%; adjusted RR 4.77, 95% CI 1.20–18.98; Table 3). Table 3. Occurrence of nonfocal TNAs in the past six months in patients with and without CAO, and in the subgroup of patients without vertebral artery stenosis or occlusion. All patients Carotid occlusion (n = 67) No carotid occlusion (n = 62) RR crude (95% CI) RR adjusted (95% CI) Occurrence of ≥1 different TNA symptoms, n (%) 26 (39%) 17 (27%) 1.42 (0.86–2.34) 1.47 (0.83–2.61)a Occurrence of ≥2 different TNA symptoms, n (%) 11 (16%) 2 (3%) 5.09 (1.17–22.06) 4.77 (1.20–18.98)b No VA steno-occlusion Carotid occlusion (n=49) No carotid occlusion (n=59) RR crude(95% CI) RR adjusted(95% CI) Occurrence of ≥1 different TNA symptoms, n (%) 16 (33%) 16 (27%) 1.20 (0.67–2.15) 1.08 (0.56–2.10)a Occurrence of ≥2 different TNA symptoms, n (%) 8 (16%) 2 (3%) 4.82 (1.07–21.64) 4.02 (0.99–16.35)b TNA: transient neurological attack; CAO: carotid artery occlusion; RR: risk ratio; CI: confidence interval; VA: vertebral artery. aAdjusted for age, sex, systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire.

All patients Carotid occlusion (n = 67) No carotid occlusion (n = 62) RR crude (95% CI) RR adjusted (95% CI) Occurrence of ≥1 different TNA symptoms, n (%) 26 (39%) 17 (27%) 1.42 (0.86–2.34) 1.47 (0.83–2.61)a Occurrence of ≥2 different TNA symptoms, n (%) 11 (16%) 2 (3%) 5.09 (1.17–22.06) 4.77 (1.20–18.98)b No VA steno-occlusion Carotid occlusion (n=49) No carotid occlusion (n=59) RR crude(95% CI) RR adjusted(95% CI) Occurrence of ≥1 different TNA symptoms, n (%) 16 (33%) 16 (27%) 1.20 (0.67–2.15) 1.08 (0.56–2.10)a Occurrence of ≥2 different TNA symptoms, n (%) 8 (16%) 2 (3%) 4.82 (1.07–21.64) 4.02 (0.99–16.35)b TNA: transient neurological attack; CAO: carotid artery occlusion; RR: risk ratio; CI: confidence interval; VA: vertebral artery. aAdjusted for age, sex, systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire. bAdjusted for systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire. Exclusion of patients with vertebral artery steno-occlusion (n = 21, 16%) yielded comparable results (Table 3). In a post-hoc analysis, nonfocal TNAs occurred more often in patients with both CAO and a contralateral carotid or vertebral artery steno-occlusion (n = 28) than in patients without any carotid or vertebral artery steno-occlusion (n = 59) (46% vs. 27%; adjusted RR 2.22, 95% CI 1.08–4.60 for ≥1 and 21% vs. 3%; adjusted RR 8.27, 95% CI 1.83–37.32 for ≥2 nonfocal TNAs) (Table 4).

n patients with both CAO and a contralateral carotid or vertebral artery steno-occlusion (n = 28) than in patients without any carotid or vertebral artery steno-occlusion (n = 59) (46% vs. 27%; adjusted RR 2.22, 95% CI 1.08–4.60 for ≥1 and 21% vs. 3%; adjusted RR 8.27, 95% CI 1.83–37.32 for ≥2 nonfocal TNAs) (Table 4). Table 4. Occurrence of nonfocal TNAs in the past six months in patients with CAO and VA or contralateral carotid steno-occlusion and in those without CAO and without VA steno-occlusion. Carotid occlusiona (n = 28) No steno- occlusionb (n = 59) RR crude (95% CI) RR adjusted (95% CI) Occurrence of ≥1 different TNA symptoms, n (%) 13 (46%) 16 (27%) 1.71 (0.96–3.05) 2.22 (1.08–4.60)c Occurrence of ≥2 different TNA symptoms, n (%) 6 (21%) 2 (3%) 6.32 (1.36–29.36) 8.27 (1.83–37.32)d TNA: transient neurological attack; CAO: carotid artery occlusion; VA: vertebral artery; RR: risk ratio; CI: confidence interval. aPatients with both CAO and a contralateral carotid artery steno-occlusion or vertebral artery steno-occlusion. bPatients without steno-occlusion of the carotid and vertebral arteries. cAdjusted for age, sex, systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire. dAdjusted for systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire.

aPatients with both CAO and a contralateral carotid artery steno-occlusion or vertebral artery steno-occlusion. bPatients without steno-occlusion of the carotid and vertebral arteries. cAdjusted for age, sex, systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire. dAdjusted for systolic blood pressure, time interval between most recent TIA or ischemic stroke and questionnaire. Discussion We found that one-third of patients with a distant history of TIA or nondisabling ischemic stroke of the anterior circulation experienced nonfocal TNAs in the preceding six months. The prevalence of multiple nonfocal TNAs was higher in patients with CAO than in those patients without CAO. Also, the prevalence of one or more nonfocal TNAs was higher in patients with CAO who also had a contralateral carotid or vertebral artery steno-occlusion than in patients without any cerebropetal steno-occlusion. Studies on the prevalence of nonfocal TNAs are limited. In a previous population-based study, 2% of patients (>55 years) experienced a nonfocal TNA in the preceding three years.1 In another hospital-based study, nonfocal symptoms occurred more frequently in patients with a recent TIA or nondisabling ischemic stroke; 20% of patients without carotid or vertebral artery stenosis, 36% of patients with carotid artery stenosis and 54% of patients with vertebral artery stenosis experienced nonfocal TNAs in the six months preceding the cerebral ischemic event.10 Our findings are compatible with these results.

nondisabling ischemic stroke; 20% of patients without carotid or vertebral artery stenosis, 36% of patients with carotid artery stenosis and 54% of patients with vertebral artery stenosis experienced nonfocal TNAs in the six months preceding the cerebral ischemic event.10 Our findings are compatible with these results. Although we can only hypothesize, the higher prevalence of multiple nonfocal TNAs in patients with versus without CAO might be caused by cerebral hypoperfusion due to CAO. Our finding that these differences in prevalence were even more profound in patients with both CAO and a contralateral carotid or vertebral artery steno-occlusion supports a hemodynamic etiology.

evalence of multiple nonfocal TNAs in patients with versus without CAO might be caused by cerebral hypoperfusion due to CAO. Our finding that these differences in prevalence were even more profound in patients with both CAO and a contralateral carotid or vertebral artery steno-occlusion supports a hemodynamic etiology. Our study has some limitations. To start, although we tried to make the group without CAO comparable to the CAO group with regards to vascular risk factors and risk of ischemic stroke by including only patients with a distant history of TIA or nondisabling ischemic stroke as our control population, vascular risk factors and previous vascular diseases were more often present in the CAO group. Although we cannot rule out residual confounding, we have adjusted our analysis for those variables that we considered potential confounders of the association between presence of CAO and occurrence of nonfocal TNAs. Second, the time interval between the most recent cerebral ischemic event and administration of the questionnaire was longer in patients with CAO than in patients without CAO. Patients with a more recent event may have been more aware of what symptoms they experienced in the previous six months which may have underestimated the prevalence of nonfocal TNAs in patients with CAO. However, this would only have enhanced the difference between patients with and without CAO and adjustment for this time interval did not have a large influence on the effect estimates. Third, the diagnosis of a nonfocal TNA can be difficult and the questionnaire was administered by several different unblinded interviewers which carries the risk of misclassification. However, we used a standardized questionnaire and systematically asked for nonfocal TNAs. Last, we had no detailed information regarding diagnostic examinations in patients with nonfocal TNAs and as such could not be completely sure that other diagnoses were ruled out.

ewers which carries the risk of misclassification. However, we used a standardized questionnaire and systematically asked for nonfocal TNAs. Last, we had no detailed information regarding diagnostic examinations in patients with nonfocal TNAs and as such could not be completely sure that other diagnoses were ruled out. In conclusion, our study indicates that nonfocal TNAs are highly common in patients with a distant TIA or nondisabling stroke. Patients with CAO experience multiple nonfocal TNAs more often than patients without CAO. Since nonfocal TNAs are highly prevalent and it is unknown whether they should be a reason to adapt clinical management, future research is needed to understand underlying mechanisms and prognosis. Supplemental Material Supplemental material for Nonfocal transient neurological attacks in patients with carotid artery occlusion Click here for additional data file. Supplemental Material for Nonfocal transient neurological attacks in patients with carotid artery occlusion by Eline A Oudeman, Eline J Volkers, Jacoba P Greving, Catharina JM Klijn, Ale Algra, L.Jaap Kappelle and on Behalf of the HBC and SMART Study Group in European Stroke Journal Acknowledgements We gratefully acknowledge the contribution of researchers and participants of the SMART Study Group and HBC Study Group. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental Material for Nonfocal transient neurological attacks in patients with carotid artery occlusion by Eline A Oudeman, Eline J Volkers, Jacoba P Greving, Catharina JM Klijn, Ale Algra, L.Jaap Kappelle and on Behalf of the HBC and SMART Study Group in European Stroke Journal Acknowledgements We gratefully acknowledge the contribution of researchers and participants of the SMART Study Group and HBC Study Group. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Heart Brain Connection (HBC) consortium is supported by the Netherlands CardioVascular Research Initiative: the Dutch Heart Foundation (CVON 2012-06 HBC), the Netherlands Organisation for Health Research and Development and the Royal Netherland Academy of Sciences. CJM Klijn is supported by grants from the Dutch Heart Foundation (2012T077) and ZonMW (015008048).The SMART study is supported by a grant from the University Medical Center Utrecht. Ethical approval The HBC study and SMART study were approved by the Ethics Committee of the Leiden UMC and UMC Utrecht, respectively. Informed consent Written informed consent was obtained from all participants. Guarantor EAO, EJV and LJK.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Heart Brain Connection (HBC) consortium is supported by the Netherlands CardioVascular Research Initiative: the Dutch Heart Foundation (CVON 2012-06 HBC), the Netherlands Organisation for Health Research and Development and the Royal Netherland Academy of Sciences. CJM Klijn is supported by grants from the Dutch Heart Foundation (2012T077) and ZonMW (015008048).The SMART study is supported by a grant from the University Medical Center Utrecht. Ethical approval The HBC study and SMART study were approved by the Ethics Committee of the Leiden UMC and UMC Utrecht, respectively. Informed consent Written informed consent was obtained from all participants. Guarantor EAO, EJV and LJK. Authors’ contribution EAO and EJV performed the interviews and data extraction; EAO, EJV, JPG, AA and LJK designed the study plan and were involved in data analysis and interpretation of the results. EAO and EJV wrote the first version of the manuscript. All authors contributed to revision of the manuscript and gave final approval to submit the manuscript for publication. Appendix Collaborators of the Heart Brain Connection study are R.J. van Oostenbrugge, H.P. Brunner-La Rocca, A.C. van Rossum, M.A. van Buchem, A. de Roos, W.M. van der Flier, W.J. Niessen, R.J. an der Geest, M.A. Ikram, P.J. Koudstaal, M. Daemen, M.J.P. van Osch.

Authors’ contribution EAO and EJV performed the interviews and data extraction; EAO, EJV, JPG, AA and LJK designed the study plan and were involved in data analysis and interpretation of the results. EAO and EJV wrote the first version of the manuscript. All authors contributed to revision of the manuscript and gave final approval to submit the manuscript for publication. Appendix Collaborators of the Heart Brain Connection study are R.J. van Oostenbrugge, H.P. Brunner-La Rocca, A.C. van Rossum, M.A. van Buchem, A. de Roos, W.M. van der Flier, W.J. Niessen, R.J. an der Geest, M.A. Ikram, P.J. Koudstaal, M. Daemen, M.J.P. van Osch. Collaborators of the SMART study group are A. Algra, Y. van der Graaf, D.E. Grobbee, G.E.H.M. Rutten, G.J. de Borst, L.J. Kappelle, T. Leiner, H.M. Nathoe, F.L.J. Visseren.

outcome.3,4 The treatment with IVT and IAT is logistically and technically complex and performed by a multidisciplinary and experienced team, consisting of neurologists, (intervention) radiologists, anaesthesiologists and nurses in often different hospitals. Therefore, a well-organised acute stroke care is a necessity. By measuring everyday practice, more information can be obtained regarding quality of stroke care compared to the information provided by clinical trials alone. Clinical auditing is an instrument that seeks to assure the quality of patient care and aids to improve outcomes. It uses quality indicators to measure the performance of an individual hospital over time and compares the quality of care provided between hospitals. Therefore, clinical auditing can identify areas for improvement. The main clinical auditing tool for stroke in the Netherlands since 2014 is the Dutch Acute Stroke Audit (DASA). The DASA is a clinical audit concerning stroke care for patients with AIS and intracranial haemorrhage (ICH). The DASA can serve as an instrument to determine variation of care between hospitals to investigate the reason for this variation and to find ways to improve, for example by an in-depth study of best practices in acute stroke care. Aims and/or hypothesis The aim of this study was to present the structure of the DASA and evaluate the results of the registration of acute stroke care parameters in the DASA from 2015 up to and including 2016.

By measuring everyday practice, more information can be obtained regarding quality of stroke care compared to the information provided by clinical trials alone. Clinical auditing is an instrument that seeks to assure the quality of patient care and aids to improve outcomes. It uses quality indicators to measure the performance of an individual hospital over time and compares the quality of care provided between hospitals. Therefore, clinical auditing can identify areas for improvement. The main clinical auditing tool for stroke in the Netherlands since 2014 is the Dutch Acute Stroke Audit (DASA). The DASA is a clinical audit concerning stroke care for patients with AIS and intracranial haemorrhage (ICH). The DASA can serve as an instrument to determine variation of care between hospitals to investigate the reason for this variation and to find ways to improve, for example by an in-depth study of best practices in acute stroke care. Aims and/or hypothesis The aim of this study was to present the structure of the DASA and evaluate the results of the registration of acute stroke care parameters in the DASA from 2015 up to and including 2016. Materials and methods Initiation of the DASA In the Netherlands, the foundation ‘Kennisnetwerk CVA NL’ (KNCN) was set up in 2006 to secure and improve care by registering the multidisciplinary care of stroke patients on a regional level. In 2014, KNCN joined the Dutch Institute for Clinical Auditing (DICA) to initiate the Cerebrovascular Accident Benchmark (CVAB) to register stroke patients to evaluate stroke care in each hospital in the Netherlands. In 2016, the Netherlands Society of Neurology (NVN) took over the governance of the audit. Due to the high registration burden of the audit and the developments in acute stroke care treatment, the primary focus of the registry was shifted towards the acute treatment of AIS. In 2017, the CVAB registry was renamed to DASA.

2016, the Netherlands Society of Neurology (NVN) took over the governance of the audit. Due to the high registration burden of the audit and the developments in acute stroke care treatment, the primary focus of the registry was shifted towards the acute treatment of AIS. In 2017, the CVAB registry was renamed to DASA. DICA as a facilitator of clinical audits in the Netherlands DICA is an independent organisation, founded by medical specialists, that facilitates national audits for various medical professions, including the DASA. The National Health Care Institute utilises DICA to fulfil the role given by the government of maintaining the quality and affordability of health care in the Netherlands as well as provide transparency in quality of care to the public. Funding for the audit is ensured by ‘Zorgverzekeraars Nederland’ (i.e. the umbrella organisation of nine health insurers in the Netherlands).

the role given by the government of maintaining the quality and affordability of health care in the Netherlands as well as provide transparency in quality of care to the public. Funding for the audit is ensured by ‘Zorgverzekeraars Nederland’ (i.e. the umbrella organisation of nine health insurers in the Netherlands). Dataset defined by experts in the field The NVN formed a clinical audit board, consisting of mandated clinical experts in acute stroke care. Alongside neurologists, this board consists of a vascular surgeon (specialised in carotid interventions), an epidemiologist and additionally includes representation of the board of the NVN. The clinical audit board defines the indicators, containing performance indicators as well as outcome measures. These are based on Dutch evidence-based guidelines5 and emerge from conferences with neurologists active in the NVN and are updated with the latest developments in stroke care. They are used to assess the provided quality of care and identify variation between hospitals. Each variable in the dataset is evaluated by the clinical audit board for clinical relevancy on an annual basis.

rom conferences with neurologists active in the NVN and are updated with the latest developments in stroke care. They are used to assess the provided quality of care and identify variation between hospitals. Each variable in the dataset is evaluated by the clinical audit board for clinical relevancy on an annual basis. Data collection The data dictionary of the DASA is shown in Supplement 1. The data are collected with a waiver of patient consent as is common in clinical audits. Hospitals are free to decide who carries out the data registration (for instance (research) nurses, data managers or neurologists), but the final responsibility rests with the neurologist. Besides DICA and the NVN, a third party, Medical Research Data Management (MRDM), is involved to anonymise the data to comply with privacy legislature. Hospitals have three ways to provide the collected data to this data processor. First, an online survey through a secured web environment is available for hospitals to record the data. Second, hospitals can distribute the data in batches, i.e. data files in which large amounts of data can be transferred directly to the data processor. Third, to minimise registration burden, some hospitals took initiative to implement data linkage, i.e. extracting the data from their individual electronic patient health record to be automatically forwarded to MRDM. Thereafter, the data are reported in de-identified format to the clinical researcher of the DASA and the clinical audit board, who perform the nationwide data analysis.

ative to implement data linkage, i.e. extracting the data from their individual electronic patient health record to be automatically forwarded to MRDM. Thereafter, the data are reported in de-identified format to the clinical researcher of the DASA and the clinical audit board, who perform the nationwide data analysis. Feedback of data Results of the data are provided to participating hospitals through a secured website, MyDASA. Results are weekly updated. Annually, indicators shown to be valid (i.e. indicator is a valid tool to measure the defined care) and complete are selected by the clinical audit board for public transparency. These indicators are discussed with other stakeholders, such as medical specialists, hospitals, organisations representing patients and insurers and are then included in next year’s transparent indicators. During the next year of inclusion, the results of the established set of public indicators can be monitored in MyDASA and at the end of the year hospital-specific indicator results are published online and therefore visible to all relevant parties through DICA’s web-based Transparency Portal. In this way, the public, as well as the health care insurers and the government, are informed with meaningful information about the quality of the care provided by individual hospitals.

fic indicator results are published online and therefore visible to all relevant parties through DICA’s web-based Transparency Portal. In this way, the public, as well as the health care insurers and the government, are informed with meaningful information about the quality of the care provided by individual hospitals. Patient domain The DASA is a voluntary prospective clinical audit. Consecutive patients over 18 years of age that are presented in the acute phase with AIS and ICH at the hospital are included. Patients are considered to be in the acute phase of the stroke when they are presented to the emergency room or when they suffer a stroke while admitted to hospital. Excluded are patients with a transient ischaemic attack (TIA), cerebral venous thrombosis, subarachnoid haemorrhage, subdural and epidural haematoma. With respect to the DASA, TIA was defined as a neurological event caused by temporary lack of blood flow with neurological deficits that were not apparent at presentation anymore. Imaging-proven infarct with time of deficit of less than 24 h was classified as AIS. The minimal data requirements to consider a patient eligible for analysis are date of birth, type of stroke and date of presentation with symptoms of stroke at the hospital. The date of presentation at the hospital is used to determine the year of inclusion.

with time of deficit of less than 24 h was classified as AIS. The minimal data requirements to consider a patient eligible for analysis are date of birth, type of stroke and date of presentation with symptoms of stroke at the hospital. The date of presentation at the hospital is used to determine the year of inclusion. Indicators The collected data of each patient can be divided into three categories: patient and/or disease characteristics, process indicators and outcome indicators. Age, gender and type of stroke are registered as relevant patient characteristics. Severity of stroke, i.e. National Institutes of Health Stroke Scale (NIHSS) score, was not registered in every day practice (i.e. outside of trial setting) for each stroke patient until 2017. Process indicators measure the activities concerning aspects of health care that were delivered by the health care providers. In the DASA, onset-to-door time was registered for both patients with AIS as well as patients with ICH. Onset time was defined as the time when symptoms of stroke started reported by patient or observer. In case the patient had a stroke during sleep, or the patient could not recall the time when symptoms began, the onset time was defined as unknown. Door time is defined as the time of arrival at emergency room. In case the patient has a stroke while admitted to the hospital, then the time when the neurologist sees the patient is the door time. Onset-to-door time is defined as the difference between onset time and door time. Other process indicators have been implemented, such as application of IVT and/or IAT including the derived door-to-needle time (DTNT) and door-to-groin time (DTGT), for only patients with AIS. DTNT is defined as the door time to the start of IVT. DTGT is the door time at the IAT centre to the groin puncture at the start of the IAT. To assess the functional outcome after three months, the modified Rankin Scale (mRS) score is used. Data for this indicator are collected by a research nurse over telephone. Additionally, during this time point, possible mortality will be recorded.

r time at the IAT centre to the groin puncture at the start of the IAT. To assess the functional outcome after three months, the modified Rankin Scale (mRS) score is used. Data for this indicator are collected by a research nurse over telephone. Additionally, during this time point, possible mortality will be recorded. Methods R Studio version 3.4.3 was used for statistical analysis. Patients with AIS and ICH registered in registration years 2014 up to and including 2016 were included. First, the volume of the DASA was described of the whole cohort. For further analysis, the year 2014 was excluded as this was a start-up year and was believed to be incomplete. Differences between patients and treatment characteristics are described using descriptive statistics. Categorical variables were compared using the chi-square trend test. Kruskal–Wallis test was used to assess age, onset-to-door time, DTNT and DTGT over the years of registration. Ordinal regression was used to determine the difference in mRS scores for each year. Logistic regression was used to determine the effect of IAT centre on dichotomised mRS (0–2 compared to 3–6). For this study, no informed consent or ethical approval was required under Dutch law. Results During the first three years of the audit, the DASA has included over 86,388 patients: 10,680 (12%) with ICH and 75,708 (88%) with AIS. In 2014, the first year of the audit, 75 hospitals participated. By 2016, 81 hospitals participated in the audit.

Methods R Studio version 3.4.3 was used for statistical analysis. Patients with AIS and ICH registered in registration years 2014 up to and including 2016 were included. First, the volume of the DASA was described of the whole cohort. For further analysis, the year 2014 was excluded as this was a start-up year and was believed to be incomplete. Differences between patients and treatment characteristics are described using descriptive statistics. Categorical variables were compared using the chi-square trend test. Kruskal–Wallis test was used to assess age, onset-to-door time, DTNT and DTGT over the years of registration. Ordinal regression was used to determine the difference in mRS scores for each year. Logistic regression was used to determine the effect of IAT centre on dichotomised mRS (0–2 compared to 3–6). For this study, no informed consent or ethical approval was required under Dutch law. Results During the first three years of the audit, the DASA has included over 86,388 patients: 10,680 (12%) with ICH and 75,708 (88%) with AIS. In 2014, the first year of the audit, 75 hospitals participated. By 2016, 81 hospitals participated in the audit. Patient characteristics From January 2015 to December 2016, 55,854 patients with AIS and 7727 patients with ICH were registered in the DASA. The median age of patients with AIS was 74.0 years (IQR 64–82) and 52% was male. Of the patients with ICH, the median age was 76.0 years (IQR 66–83) and 52% was male. The median age and sex distribution did not significantly change over the years for both AIS and ICH, as shown in Table 1.

H were registered in the DASA. The median age of patients with AIS was 74.0 years (IQR 64–82) and 52% was male. Of the patients with ICH, the median age was 76.0 years (IQR 66–83) and 52% was male. The median age and sex distribution did not significantly change over the years for both AIS and ICH, as shown in Table 1. Table 1. Patient and disease characteristics, process indicators and outcome indicators registered in the DASA. Acute ischaemic stroke (n = 55,854) Intracranial haemorrhage (n = 7727) 2015 2016 p-value 2015 2016 p-value Number of patients 28,820 27,034 4145 3582 Number of hospitals 78 80 79 75 Patient and disease characteristics Age in years (median, IQR) 74 (64–82) 74 (64–82) 0.92 76 (65–83) 76 (66–84) 0.17 Male sex (n,%) 14,555 (52.4) 13,839 (52.6) 0.62 1997 (51.8) 1802 (52.5) 0.61 Onset-to-door time in minutes (median, IQR) 171 (74–544) 175 (74–570) 0.42 138 (61–430) 130 (61–408) 0.45 Process indicators Intravenous thrombolysis (n, %) 5338 (20.4) 5299 (21.7) <0.001 – – – Door-to-needle time in minutes (median, IQR) 27 (20–37) 25 (19–35) <0.001 – – – Intra-arterial thrombectomy (n, %) 755 (3.1) 985 (4.1) <0.001 – – – Door-to-groin time in minutes (median, IQR) 66 (41–99) 64 (35–95) 0.02 – – – Outcome indicators In-hospital mortality (n, %) 1310 (5.0) 1161 (4.8) 0.19 995 (25.4) 863 (26.4) 0.35 Modified Rankin Scale (mRS) scorea No symptoms, mRS 0 (n, %) 2144 (17.8) 2294 (19.8) <0.001 89 (7.1) 87 (7.5) <0.001

Intravenous thrombolysis (n, %) 5338 (20.4) 5299 (21.7) <0.001 – – – Door-to-needle time in minutes (median, IQR) 27 (20–37) 25 (19–35) <0.001 – – – Intra-arterial thrombectomy (n, %) 755 (3.1) 985 (4.1) <0.001 – – – Door-to-groin time in minutes (median, IQR) 66 (41–99) 64 (35–95) 0.02 – – – Outcome indicators In-hospital mortality (n, %) 1310 (5.0) 1161 (4.8) 0.19 995 (25.4) 863 (26.4) 0.35 Modified Rankin Scale (mRS) scorea No symptoms, mRS 0 (n, %) 2144 (17.8) 2294 (19.8) <0.001 89 (7.1) 87 (7.5) <0.001 Mild symptoms, mRS 1–2 (n, %) 6178 (51.2) 5893 (51.0) 478 (37.9) 459 (39.6) Moderate to severe symptoms, mRS 3-5 (n, %) 2783 (23.1) 2608 (22.5) 478 (37.9) 414 (35.8) Death, mRS 6 (n, %) 953 (7.9) 769 (6.6) 215 (17.1) 198 (17.1) IQR: interquartile range. P-values < 0.05 are printed bold. amRS was not obtained of patients that died during admission. Process indicators Onset-to-door time. Median onset-to-door time was 173 min (IQR 74–544) in patients with AIS and 134 min (IQR 61–421) in patients with ICH. Onset-to-door time did not change over time in both AIS and ICH. IVT. Seventy-five hospitals administrating IVT are registered in the DASA. In 2015, 20.4% (n = 5338) of the patients with AIS received IVT. This percentage increased to 21.7% (n = 5299) in 2016 (p < 0.001). During the registry, the median DTNT reduced from 27 min (IQR 20–37) to 25 min (IQR 19–35) (p < 0.001). The range of distribution of IQR decreased over the years, implying that variation in DTNT reduced, as shown in the density plot in Figure 1. Of all patients receiving IVT during the study period, 92% was given within 60 min.

e registry, the median DTNT reduced from 27 min (IQR 20–37) to 25 min (IQR 19–35) (p < 0.001). The range of distribution of IQR decreased over the years, implying that variation in DTNT reduced, as shown in the density plot in Figure 1. Of all patients receiving IVT during the study period, 92% was given within 60 min. Figure 1. Distribution of DTNT in patients with AIS treated with IVT for each year. The vertical lines represent the annual median. On a hospital level, the median DTNT ranged from 15.5 to 48 min over the course of 2015 and 2016, as shown in Figure 2(a). Figure 2(b) shows the delta for each hospital, i.e. the difference of median DTNT between 2015 and 2016. Forty-three hospitals lowered their median DTNT when comparing 2015 and 2016, six hospitals have similar median DTNTs and 23 hospitals show an increase of median DTNT. Figure 2. (a) Boxplots of DTNT with range of distribution from fifth to 95th percentile for 2015 and 2016 combined for each hospital registering in the DASA and (b) difference in median DTNT in minutes (i.e. delta) between 2015 and 2016 for each hospital registering in the DASA. The dotted line reflects the nationwide trend of reduction of median DTNT. DTNT: door-to-needle time.

m fifth to 95th percentile for 2015 and 2016 combined for each hospital registering in the DASA and (b) difference in median DTNT in minutes (i.e. delta) between 2015 and 2016 for each hospital registering in the DASA. The dotted line reflects the nationwide trend of reduction of median DTNT. DTNT: door-to-needle time. IAT. The Multicenter Randomized Clinical trial of Endovascular treatment for Acute ischemic stroke in the Netherlands (MR CLEAN) proved irrevocably as first international randomised controlled trial that IAT is an effective and safe therapy, when administered within 6 h after AIS caused by proximal intracranial occlusion of the anterior circulation.6 All 17 IAT hospitals, which also participated in the MR CLEAN registry, registered their IAT patients in the DASA. In 2015, 3.1% of patients (n = 755) with IAT treatment were registered; by 2016, this increased to 4.1% (n = 986; p < 0.001). During the study period, the median DTGT reduced from 66 min (IQR 41–99) to 64 min (IQR 35–95). Similar to the DTNT in treatment with IVT, the variation in DTGT reduced, as shown in Figure 3. Figure 3. Annual distribution of DTGT in patients with AIS treated with IAT. The vertical lines represent the annual median.

IAT. The Multicenter Randomized Clinical trial of Endovascular treatment for Acute ischemic stroke in the Netherlands (MR CLEAN) proved irrevocably as first international randomised controlled trial that IAT is an effective and safe therapy, when administered within 6 h after AIS caused by proximal intracranial occlusion of the anterior circulation.6 All 17 IAT hospitals, which also participated in the MR CLEAN registry, registered their IAT patients in the DASA. In 2015, 3.1% of patients (n = 755) with IAT treatment were registered; by 2016, this increased to 4.1% (n = 986; p < 0.001). During the study period, the median DTGT reduced from 66 min (IQR 41–99) to 64 min (IQR 35–95). Similar to the DTNT in treatment with IVT, the variation in DTGT reduced, as shown in Figure 3. Figure 3. Annual distribution of DTGT in patients with AIS treated with IAT. The vertical lines represent the annual median. Outcome indicators Of patients with AIS registered from 2014 to and including 2016, the mortality during admission was 4.9% with no significant difference between the years of inclusion (p = 0.19). Of the patients that were discharged, 7.3% had died at three-month follow-up. After three months, 70% of patients were functionally independent (mRS score of 0–2) of which 19% had no symptoms at all (mRS of 0). When comparing IAT centres to non-IAT centres, the IAT centres have significantly more patients with severe outcome (mRS score of 3–6) than non-IAT centres (OR 1.09; 95% CI 1.04–1.15; p = 0.001).

ree months, 70% of patients were functionally independent (mRS score of 0–2) of which 19% had no symptoms at all (mRS of 0). When comparing IAT centres to non-IAT centres, the IAT centres have significantly more patients with severe outcome (mRS score of 3–6) than non-IAT centres (OR 1.09; 95% CI 1.04–1.15; p = 0.001). Of patients with ICH from 2014 to and including 2016, the mortality during admission was 26%. At the three-month follow-up, of the patients that were discharged from hospital another 17% had died. After three months, 47% of patients was functionally independent (mRS score of 0–2) of which 7% had no symptoms at all (mRS score of 0). Moderate or severe disability (mRS of 3–5) was assessed in 37% of the follow-up patients. When comparing both years, no significant differences in independent functioning at three months were found (p = 0.35). After discharge, the follow-up was registered in 49% of the patients with AIS and 45% of the patients with ICH. For both AIS and ICH, the percentage of completeness of registration for each hospital ranges from 0 to 99.7% and 0 to 100%, respectively. Discussion This is the first report of the DASA. Since the start of the DASA in 2014, over 86,000 stroke patients have been registered. During the DASA, the median time to treatment for both IVT and IAT decreased and showed a narrower range of distribution of treatment time, indicating less variability.

After discharge, the follow-up was registered in 49% of the patients with AIS and 45% of the patients with ICH. For both AIS and ICH, the percentage of completeness of registration for each hospital ranges from 0 to 99.7% and 0 to 100%, respectively. Discussion This is the first report of the DASA. Since the start of the DASA in 2014, over 86,000 stroke patients have been registered. During the DASA, the median time to treatment for both IVT and IAT decreased and showed a narrower range of distribution of treatment time, indicating less variability. Clinical auditing, in this case by using DASA, can be used to define a standard by using process indicators and outcome indicators to assess the quality of the treatment provided by hospitals.7–9 Due to the ongoing continuous character of the auditing cycle, an up-to-date benchmark can be provided. Several studies have shown that a reduction of the DTNT in stroke care can be reached, using quality registration as a tool for improvement.10–12 As a result, the requirement of a median DTNT under 45 min for each hospital has now been formulated in the Dutch national quality standard of acute stroke to stimulate local interventions aiming at a reduction of time to treatment. Local quality improvement processes in the Netherlands for DTNT have shown to be effective. For instance, Zinkstok et al.13 and Van Schaik et al.14 used information obtained from an auditing cycle to significantly reduce the median DTNT in IVT. With the addition of IAT as a second acute treatment option for patients with AIS, the DASA can be utilised to evaluate the DTGT. IAT is only performed in selected centres, therefore referral from a centre where the patient primarily presented is often necessary. On short term and based on the data in the DASA, it will be possible to calculate door-to-door-to-groin times to evaluate the regional organisation of stroke care.

to evaluate the DTGT. IAT is only performed in selected centres, therefore referral from a centre where the patient primarily presented is often necessary. On short term and based on the data in the DASA, it will be possible to calculate door-to-door-to-groin times to evaluate the regional organisation of stroke care. There are several limitations to the DASA. First, severity of stroke was not registered in the DASA during the first three years of the registry. However, such influencing prognostic patient-related factors are necessary to register to understand the measured outcome. From January 2014 until September 2016, five hospitals in the DASA registered a more extensive set of variables to identify possible influential factors. Three were determined: age, type of stroke (either AIS or ICH) and the severity of the stroke (defined by the NIHSS score). Earlier research verified these as important prognostic factors for which 30-day outcome measures must be adjusted.15 Since 2017, the NIHSS score was added to the dataset and these case-mix factors are used in order to provide reliable benchmark information.

and the severity of the stroke (defined by the NIHSS score). Earlier research verified these as important prognostic factors for which 30-day outcome measures must be adjusted.15 Since 2017, the NIHSS score was added to the dataset and these case-mix factors are used in order to provide reliable benchmark information. A second possible limitation is that to date the mRS score after three months has been registered in 49% of the patients with AIS and in 45% of patients with ICH. This percentage of completion needs to improve to better assess quality improvement. When correcting for patient-related factors, the outcome measure could be able to rank hospitals, as mentioned earlier. It is challenging for hospitals in the Netherlands to complete these data, as they do not get financial support to aid this process. Other national audits seem to experience the same difficulty with completeness range from 5 to 100%.16 The NVN is actively appealing to the neurologists to implement NIHSS score registration in the acute phase of AIS as well as to increase the completeness concerning three-month follow-up. A final limitation is that data completeness is difficult to determine as only patients presented to the emergency room in the acute phase of the stroke are registered in the DASA. Patients seen in the outpatient clinic are not included, but are assigned the same diagnosis–treatment–combination hospital reimbursement code. To assess data accuracy, we will perform data verification on a random sample in the near future.

he emergency room in the acute phase of the stroke are registered in the DASA. Patients seen in the outpatient clinic are not included, but are assigned the same diagnosis–treatment–combination hospital reimbursement code. To assess data accuracy, we will perform data verification on a random sample in the near future. Even with the significant reduction of DTNT and DTGT, the DASA can continue to play an important role in monitoring and improving stroke care in the future. The DAWN trial17 and the DEFUSE-3 trial18 have recently proven that IAT is effective in a selection of patients with large vessel occlusion with an onset of stroke symptoms for more than 6 h and less than 24 h. This indicates that these patients with long-standing symptoms must be assessed in the acute setting as well, resulting in extra strain on the acute stroke services. This warrants an adjustment in regional logistic protocol. The DASA can be used to monitor the implementation of this adjusted protocol and indicate domains for improvement.

these patients with long-standing symptoms must be assessed in the acute setting as well, resulting in extra strain on the acute stroke services. This warrants an adjustment in regional logistic protocol. The DASA can be used to monitor the implementation of this adjusted protocol and indicate domains for improvement. Conclusion From 2014 to and including 2016 the DASA, a national audit in the Netherlands, has registered over 86,000 patients with AIS and ICH. Over the course of time, the number of patients who received IVT and/or IAT and have been enrolled into the registry has significantly risen and the median time from the door of the hospital to treatment has been reduced. Increasing completeness of registration of the outcome, in combination with adjustment for patient-related factors, is necessary to define and further improve the quality of the acute stroke care provided by each hospital. Due to the rapidly evolving field of acute stroke care and the broadening of the selection of patients that are eligible for treatment, there is a continuing need for monitoring and assessing stroke care as done by the DASA. Supplemental Material Supplemental material for The Dutch Acute Stroke Audit: Benchmarking acute stroke care in the Netherlands Click here for additional data file. Supplemental material for The Dutch Acute Stroke Audit: Benchmarking acute stroke care in the Netherlands by Laurien S Kuhrij, Michel WJM Wouters, Renske M van den Berg-Vos, Frank-Erik de Leeuw and Paul J Nederkoorn in European Stroke Journal

Supplemental Material Supplemental material for The Dutch Acute Stroke Audit: Benchmarking acute stroke care in the Netherlands Click here for additional data file. Supplemental material for The Dutch Acute Stroke Audit: Benchmarking acute stroke care in the Netherlands by Laurien S Kuhrij, Michel WJM Wouters, Renske M van den Berg-Vos, Frank-Erik de Leeuw and Paul J Nederkoorn in European Stroke Journal Acknowledgements We thank the Clinical Audit board of the DASA for its contribution: Lingsma HF, de Borst GJ, van Norden AGW and Eysink Smeets MM.

Supplemental material for The Dutch Acute Stroke Audit: Benchmarking acute stroke care in the Netherlands by Laurien S Kuhrij, Michel WJM Wouters, Renske M van den Berg-Vos, Frank-Erik de Leeuw and Paul J Nederkoorn in European Stroke Journal Acknowledgements We thank the Clinical Audit board of the DASA for its contribution: Lingsma HF, de Borst GJ, van Norden AGW and Eysink Smeets MM. We also want to thank the DASA consortium: Aerden LAM (Reinier de Graaf Ziekenhuis), Alblas CL (Franciscus Vlietland Ziekenhuis), de Beer F (Spaarne Gasthuis), Bienfait PH (Gelre Ziekenhuis), Boon AE (St Anna Ziekenhuis), Bor S (Rode Kruis Ziekenhuis), Boreas AMHP (Antonius Ziekenhuis Sneek), Bronner I (Flevoziekenhuis), Brouns R (St ZorgSaam Terneuzen), Brouwers PJAM (Medisch Spectrum Twente), Brugman F (Ziekenhuis Rivierenland), Dane ML (Rivas Beatrix Ziekenhuis), Fransen PSS (Isala Klinieken), van Gemert HMA (Meander Medisch Centrum), van Golde AEL (Ziekenhuisgroep Twente), de Graaf MT (Slotervaart ziekenhuis), Hani L (Noordwest Ziekenhuis Den Helder), Hilkens PH (St Antonius Ziekenhuis), ten Holter JBM (Deventer ziekenhuis), de Jong SW (St Jansdal Harderwijk), Kapelle LJ (University Medical Center Utrecht), Keizer K (Catharina ziekenhuis), Keunen R (HagaZiekenhuis), Kloppenborg RP (St Franciscus Gasthuis), Kok AJM (Elkerkliek Ziekenhuis), Koops L (Isala Diaconessenhuis Meppel), Kruyt ND (Leiden University Medical Center), de Leeuw FE (Radboud University Medical Center), Lövenich H (St Jans Gasthuis), Luijckx GJ (University Medical Center Groningen), Maasland E (Bethesda Ziekenhuis), van der Meijden AMHG (VieCuri Medisch Centrum), Miedema I (Ziekenhuis Gelderse Vallei), Nederkoorn PJ (Academic Medical Center), van Norden AGW (Amphia Ziekenhuis), Persoon S (Wilhelmina Ziekenhuis), Peters EW (Admiraal de Ruyter Ziekenhuis), van der Ree TC (Westfries Gasthuis), Rozeman AD (Albert Schweitzer Ziekenhuis), Saxena R (Maasstad Ziekenhuis), van Schaik S (Ons Lieve Vrouwe Gasthuis), de Schryver ELLM (Alrijne Ziekenhuis), Schuiling WJ (Medisch Centrum Leeuwarden), Schut ES (Martiniziekenhuis), Staals JEA (Maastricht University Medical Center), Stalpers X (Maxima medisch centrum), Tjeerdsma H (Bravis Ziekenhuis), van Tuijl JH (St Elisabeth Ziekenhuis), Vermeer SE (Rijnstate ziekenhuis), Visser MC (Vrije Universiteit Medical Center), van den Wijngaard I (Haaglanden Medical Center), van Zagten MSG (Jeroen Bosch Ziekenhuis) and Zylicz SA (Langeland Zoetermeer).

Stalpers X (Maxima medisch centrum), Tjeerdsma H (Bravis Ziekenhuis), van Tuijl JH (St Elisabeth Ziekenhuis), Vermeer SE (Rijnstate ziekenhuis), Visser MC (Vrije Universiteit Medical Center), van den Wijngaard I (Haaglanden Medical Center), van Zagten MSG (Jeroen Bosch Ziekenhuis) and Zylicz SA (Langeland Zoetermeer). Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Informed consent Not applicable. Ethical approval For this study, no informed consent or ethical approval was required under Dutch law. Guarantor PJN. Contributorship LSK and PJN wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

Introduction Cerebral small vessel disease (SVD) is the most important vascular contributor to cognitive decline and dementia and causes up to 25% of all ischaemic strokes worldwide.1,2 SVD affects the structure and function of the smallest cerebral blood vessels, including the perforating arterioles, capillaries and venules.3 Although these small vessels themselves cannot yet be visualised on conventional magnetic resonance imaging (MRI), MRI reliably detects a spectrum of tissue alterations thought to arise from SVD. These include white matter hyperintensities (WMH), lacunes, microbleeds, enlarged perivascular spaces, brain atrophy, and more recently, acute (micro)infarcts and loss of white matter microstructural integrity.2

ic resonance imaging (MRI), MRI reliably detects a spectrum of tissue alterations thought to arise from SVD. These include white matter hyperintensities (WMH), lacunes, microbleeds, enlarged perivascular spaces, brain atrophy, and more recently, acute (micro)infarcts and loss of white matter microstructural integrity.2 Within this spectrum of imaging findings, WMH are the most ubiquitous and extensively studied. The established paradigm for WMH development is that they arise slowly over the years and are caused by chronic hypoperfusion.3 However, this notion is mainly based on animal studies, which do not reliably capture the complex pathophysiology of a disease that develops over decades.4 In a recent meta-analysis including human in vivo studies measuring cerebral blood flow (CBF) using various techniques, evidence of reduced CBF in individuals with more severe WMH was observed, both globally and in the majority of grey and white matter regions.6 However, this association was not confirmed in all longitudinal studies.6 Studies comparing CBF within WMH and potentially at risk normal-appearing white matter show variable results.7–9 These data suggest that additional processes may play a role in the conversion of normal-appearing white matter towards WMH.

egions.6 However, this association was not confirmed in all longitudinal studies.6 Studies comparing CBF within WMH and potentially at risk normal-appearing white matter show variable results.7–9 These data suggest that additional processes may play a role in the conversion of normal-appearing white matter towards WMH. The notion of gradual progression of SVD caused by chronic hypoperfusion may also have arisen because progression has usually been studied with inter-scan intervals of several years. For instance, the majority of de novo WMH over a four-year course appeared to be due to growth of existing lesions, supporting the notion of a slow, continuous process rather than a series of acute events.10 However, sudden, rather than chronic progression or even regression11,12 may go unnoticed with MRI scan intervals of years. Similar to WMH, a previous study demonstrated incidental lacunes to occur predominantly at the edge of WMH.13 These findings suggest that WMH and incidental lacunes share an underlying pathological mechanism, potentially being acute ischaemia.

The notion of gradual progression of SVD caused by chronic hypoperfusion may also have arisen because progression has usually been studied with inter-scan intervals of several years. For instance, the majority of de novo WMH over a four-year course appeared to be due to growth of existing lesions, supporting the notion of a slow, continuous process rather than a series of acute events.10 However, sudden, rather than chronic progression or even regression11,12 may go unnoticed with MRI scan intervals of years. Similar to WMH, a previous study demonstrated incidental lacunes to occur predominantly at the edge of WMH.13 These findings suggest that WMH and incidental lacunes share an underlying pathological mechanism, potentially being acute ischaemia. Indeed, a recent study among five subjects with moderate to severe WMH with 16 weekly MRI assessments showed evidence of acute progression of WMH.14 In three participants, a total of nine incidental diffusion-weighted imaging lesions (DWI+ lesions) were observed in the white matter, considered to be suggestive of acute infarcts, which in the weeks thereafter approached the imaging characteristics of WMH.14 Of note, patients did not experience any clinical symptom, although detailed serial neuropsychological examinations had not been performed.

ions (DWI+ lesions) were observed in the white matter, considered to be suggestive of acute infarcts, which in the weeks thereafter approached the imaging characteristics of WMH.14 Of note, patients did not experience any clinical symptom, although detailed serial neuropsychological examinations had not been performed. A growing body of evidence now suggests that DWI+ lesions are rather common in SVD, but often go unnoticed because most of them remain clinically silent and the imaging evidence for a DWI+ lesion is strongest within the first four weeks.15 In cross-sectional studies, the prevalence of DWI+ lesions ranged from 0% in a population-based study including relatively young (58% of the cohort being younger than 60 years) and cognitively healthy individuals with low SVD burden16 to 8% in patients with severe SVD and a history of a lacunar stroke.17 In addition to the evolution of DWI+ lesions into WMH, other studies have shown that these lesions have different fates and may as well develop into a lacune, transform into a (micro)haemorrhage or even disappear.2,18–21 However, the exact temporal dynamics of DWI+ lesions are largely unknown. Apart from a focal effect of SVD on brain structure and function, converging evidence suggests that SVD also affects remote areas of the brain, well beyond the original lesion.22–25 However, the time course of these events is largely unknown as this cannot be accurately monitored by longitudinal studies with large follow-up intervals.

al effect of SVD on brain structure and function, converging evidence suggests that SVD also affects remote areas of the brain, well beyond the original lesion.22–25 However, the time course of these events is largely unknown as this cannot be accurately monitored by longitudinal studies with large follow-up intervals. Therefore, to investigate the origin, evolution and consequences of SVD, we set up the RUN DMC – InTENse study (Radboud University Nijmegen Diffusion tensor and Magnetic resonance imaging Cohort – Investigating The origin and EvolutioN of cerebral small vessel disease), a single-centre longitudinal study performing MRI assessments every month during 10 consecutive months (totalling up to 10 scans per subject), complemented with clinical, motor function and cognitive examinations among non-demented survivors of the RUN DMC study. For these subjects, nine years of follow-up imaging and clinical data were already available. Specifically, we aim to investigate the monthly incidence of DWI+ lesions and to assess to which extent they explain progression of SVD imaging markers. Furthermore, we aim to investigate the effects of DWI+ lesions on cortical thickness, on structural and functional connectivity and on cognitive and motor performance. Finally, in this study we will also explore the potential role of the innate immune system in the pathophysiology of SVD. Here, we present the design and protocol of the RUN DMC – InTENse study.

ects of DWI+ lesions on cortical thickness, on structural and functional connectivity and on cognitive and motor performance. Finally, in this study we will also explore the potential role of the innate immune system in the pathophysiology of SVD. Here, we present the design and protocol of the RUN DMC – InTENse study. Methods Participants Individuals were recruited from the 503 subjects of the RUN DMC study. This prospective study, on the causes and consequences of SVD, comprised baseline MRI and clinical data collection in 2006 and follow-up examinations in 2011 and 2015.26 Individuals for the RUN DMC – InTENse study were recruited between February and September 2016.

recruited from the 503 subjects of the RUN DMC study. This prospective study, on the causes and consequences of SVD, comprised baseline MRI and clinical data collection in 2006 and follow-up examinations in 2011 and 2015.26 Individuals for the RUN DMC – InTENse study were recruited between February and September 2016. Inclusion and exclusion criteria for the RUN DMC – InTENse study are summarised in Table 1. In short, we aimed to include 50 individuals with a high likelihood of progression of SVD markers during the study period, which could be attributed as much as possible to the underlying SVD pathology. Therefore, we first carefully scrutinised the medical history and previous data collected in the RUN DMC study of all participants who underwent MRI in 2006 and 2015 and excluded those with any evidence of other presumed causes of ischaemia,27 including large-artery disease, cardioembolic source and other determined cause of stroke, i.e. vasculitis, or with evidence of intracranial haemorrhage other than a microbleed on MRI. As cognitive and motor decline were among our secondary outcomes, patients with dementia and Parkinson’s disease (according to international diagnostic criteria28,29) were also excluded. Subsequently, considering previous progression of WMH as the most important determinant of future WMH progression, all remaining eligible individuals were ranked by their WMH change between 2006 and 2015. Individuals were then invited by volume of WMH progression, those with the highest progression first, until we attained a sample of 50 individuals. Figure 1 summarises the subject recruitment in a flowchart.

uture WMH progression, all remaining eligible individuals were ranked by their WMH change between 2006 and 2015. Individuals were then invited by volume of WMH progression, those with the highest progression first, until we attained a sample of 50 individuals. Figure 1 summarises the subject recruitment in a flowchart. Table 1. Inclusion and exclusion criteria of the RUN DMC – InTENse study. Inclusion criteria Participated at least in RUN DMC waves 2006 and 2015 Progression of WMH between 2006 and 2015 Able to visit clinic monthly Exclusion criteria Large artery disease defined as a carotid artery stenosis >50% based on medical history or on ultrasound during the pre-visit data collection of the RUN DMC – InTENse study Cardioembolism defined as atrial fibrillation (based on medical history or detected on ECG at baseline or pre-visit data collection of the RUN DMC or the RUN DMC – InTENse study, respectively), use of oral anticoagulants (either oral anticoagulants or direct oral anticoagulants) prescribed for arterial thromboembolism or any other cardioembolic source (e.g. mitral insufficiency) Radiological or clinical evidence of a cortical ischaemic stroke or transient ischaemic attack (e.g. aphasia or hemianopia) Evidence of vasculitis Any intracranial haemorrhage other than a microbleed on MRI Dementia Parkinson’s disease 3T MRI contraindication Pre-existing structural brain lesion preventing MRI analysis Any disease with a life expectancy less than one year Figure 1. Flowchart of subject inclusion.

Radiological or clinical evidence of a cortical ischaemic stroke or transient ischaemic attack (e.g. aphasia or hemianopia) Evidence of vasculitis Any intracranial haemorrhage other than a microbleed on MRI Dementia Parkinson’s disease 3T MRI contraindication Pre-existing structural brain lesion preventing MRI analysis Any disease with a life expectancy less than one year Figure 1. Flowchart of subject inclusion. Finally, the RUN DMC – InTENse study included 54 individuals with mild to severe SVD as documented on preceding MRIs (2006, 2011, 2015). The median[interquartile range] WMH volume at the RUN DMC follow-up in 2015 and the annual WMH progression between 2006 and 2015 were 5.6[2.5;9.8] ml and 0.35[0.20;0.58] ml/year, respectively. All individuals gave written informed consent. The study was approved by the medical ethics committee region Arnhem–Nijmegen. Study design The RUN DMC – InTENse study is a longitudinal observational study encompassing 12 visits, that is, a pre-visit, 10 monthly visits including a MRI and a follow-up visit one year after the start of the study. Table 2 depicts the type of data that was collected for each study visit. Data collection took place between March 2016 and November 2017. Table 2. Schedule of all assessments in the RUN DMC – InTENse study.

Study design The RUN DMC – InTENse study is a longitudinal observational study encompassing 12 visits, that is, a pre-visit, 10 monthly visits including a MRI and a follow-up visit one year after the start of the study. Table 2 depicts the type of data that was collected for each study visit. Data collection took place between March 2016 and November 2017. Table 2. Schedule of all assessments in the RUN DMC – InTENse study. Study visit Month Assessment Pre-visit 1 2 3 4 5 6 7 8 9 10 1-y FU Screening Ultrasonography carotid arteries x ECG x MRI x x x x x x x x x x Cognitive assessment Full cognitive assessment x x Test of Attentional Performance x x x x x x x x x x x x Motor assessment Timed Up & Go test x x x x x x x x x x Six-meter walk test x x x x x x x x x x Physical assessment Blood pressure, pulse rate x x x x x x x x x x x Weight, length, BMI x x Abdominal circumference x x Additional laboratory investigations Glucose level x Lipid profile x Structured questionnaires Educational level x Barthel index x x IADL x x CES-D x x Substance use x x x x x x x X x x x x Trigger factors and events x x x x x x x X x x x x Medication use x x x x x x x X x x x x Blood sampling x x x Note: Physical activity was assessed once in the month March to take out seasonal effects. 1-y FU: 1-year follow-up; IADL: Instrumental Activities of Daily Living; CES-D: Center of Epidemiologic Studies Depression Scale.

tors and events x x x x x x x X x x x x Medication use x x x x x x x X x x x x Blood sampling x x x Note: Physical activity was assessed once in the month March to take out seasonal effects. 1-y FU: 1-year follow-up; IADL: Instrumental Activities of Daily Living; CES-D: Center of Epidemiologic Studies Depression Scale. Screening for exclusion criteria Ultrasonography of the carotid arteries. During the pre-visit, ultrasonography of the carotid arteries was performed to detect an internal carotid artery stenosis >50%, as indicated by a peak systolic velocity ratio between the internal and common carotid artery >2. The intima media thickness (mm) was determined and averaged over a length of 1 cm in the far wall of the left and right distal common carotid artery near the bifurcation. ECG. During the pre-visit, an ECG was made to detect atrial fibrillation. All ECGs were assessed by a cardiologist. MRI acquisition Participants were scanned on a 3T MRI system (MAGNETOM Prisma, Siemens Medical Solutions, Erlangen, Germany) with a 32-channel head coil. To allow for detection of a spectrum of SVD consequences, the following sequences were applied: 3D fluid-attenuated inversion recovery (FLAIR) with repetition time/echo time/inversion time (TR/TE/TI) 5000/394/1800 ms, 0.85 × 0.85 × 0.85 mm, 192 slices, acquisition time (TA) 7.02 min; Magnetisation Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE),30 to obtain a quantitative T1 map and a uniform bias-corrected T1-weighted image, TR/TI1/TI2 5500/700/2500 ms, flip angle α1/α2 7/4°, 0.85 × 0.85 × 0.85 mm, 256 slices, TA 11.51 min;

MRI acquisition Participants were scanned on a 3T MRI system (MAGNETOM Prisma, Siemens Medical Solutions, Erlangen, Germany) with a 32-channel head coil. To allow for detection of a spectrum of SVD consequences, the following sequences were applied: 3D fluid-attenuated inversion recovery (FLAIR) with repetition time/echo time/inversion time (TR/TE/TI) 5000/394/1800 ms, 0.85 × 0.85 × 0.85 mm, 192 slices, acquisition time (TA) 7.02 min; Magnetisation Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE),30 to obtain a quantitative T1 map and a uniform bias-corrected T1-weighted image, TR/TI1/TI2 5500/700/2500 ms, flip angle α1/α2 7/4°, 0.85 × 0.85 × 0.85 mm, 256 slices, TA 11.51 min; Presaturated turbo flash sequence to obtain a quantitative B1 map (transmit radiofrequency [RF] map) to correct for residual RF inhomogeneities in the T1 map with TR/TE 11310/2.23 ms, 3.3 × 3.3 × 2.5 mm, 42 slices, 100% slice gap, TA 23 s; Multi-shell DWI using multi-band accelerated echo planar imaging (EPI, developed at the Center for Magnetic Resonance Research, CMRR) including 99 diffusion-weighted directions (3 × b = 200, 6 × b = 500, 30 × b = 1000, 60 × b = 3000 s/mm2) with uniform coverage within and across shells31 and 10 b = 0 images, one acquired before each series of 10 diffusion-weighted images, multi-band acceleration factor 3, TR/TE 3220/74 ms, 1.7 × 1.7 × 1.7 mm, 87 slices, TA 6.36 min; One b = 0 image to correct for susceptibility-induced distortions in DWI32 with acquisition parameters equal to the previous b=0 images, but acquired in opposite phase-encoding direction, TA 48 s;

Multi-shell DWI using multi-band accelerated echo planar imaging (EPI, developed at the Center for Magnetic Resonance Research, CMRR) including 99 diffusion-weighted directions (3 × b = 200, 6 × b = 500, 30 × b = 1000, 60 × b = 3000 s/mm2) with uniform coverage within and across shells31 and 10 b = 0 images, one acquired before each series of 10 diffusion-weighted images, multi-band acceleration factor 3, TR/TE 3220/74 ms, 1.7 × 1.7 × 1.7 mm, 87 slices, TA 6.36 min; One b = 0 image to correct for susceptibility-induced distortions in DWI32 with acquisition parameters equal to the previous b=0 images, but acquired in opposite phase-encoding direction, TA 48 s; Multiple spin echo sequence to obtain a quantitative T2-map (via model-based nonlinear inverse reconstruction33,34) with TR/ΔTE 4000/10 ms, 16 echoes, 0.7 × 0.7 × 3.0 mm acquisition voxel size reconstructed at 0.36 × 0.36 × 3.0 mm, 48 slices, 10% slice gap, TA 3.22 min; 3D multi-echo fast low angle shot (FLASH) providing magnitude and phase images for quantitative susceptibility imaging and R2* mapping, TR/ΔTE 35/4.92 ms, 6 echoes, 0.8 × 0.8 × 2.0 mm, 72 slices, no slice gap, TA 5.57 min; Resting-state functional MRI (rs-fMRI) using multi-band accelerated EPI (CMRR) with multi-band acceleration factor 8, 700 measurements, TR/TE 700/39 ms, 2.4 × 2.4 × 2.4 mm, 64 slices, TA 8.19 min;

3D multi-echo fast low angle shot (FLASH) providing magnitude and phase images for quantitative susceptibility imaging and R2* mapping, TR/ΔTE 35/4.92 ms, 6 echoes, 0.8 × 0.8 × 2.0 mm, 72 slices, no slice gap, TA 5.57 min; Resting-state functional MRI (rs-fMRI) using multi-band accelerated EPI (CMRR) with multi-band acceleration factor 8, 700 measurements, TR/TE 700/39 ms, 2.4 × 2.4 × 2.4 mm, 64 slices, TA 8.19 min; Two spin-echo EPI acquisitions acquired with opposite phase-encoding direction (anterior-posterior) to compute displacement maps to correct for susceptibility-induced distortions in rs-fMRI images32 with TR/TE 7100/66 ms, 2.4 × 2.4 × 2.4 mm, 64 slices, TA 7.1 s each. During the last two MRI sessions, the following sequences were applied instead of rs-fMRI: Triggered single-slice quantitative flow of the carotid and vertebral arteries, coupled to the peak of each cardiac cycle, TR/TE 23.6/7.44 ms, 0.6 × 0.6 × 5.0 mm, TA 2.26 min; pseudo-continuous arterial spin labelling (PCASL), labelling duration 3000 ms and post-labelling delay 2000 ms, TR/TE 5500/29.6 ms, 3.8 × 3.8 × 3.8 mm acquisition voxel size reconstructed at 1.9 × 1.9 × 3.8 mm, 24 slices, TA 5.30 min, including two M0 acquisitions with opposite phase-encoding direction and TA 11 s each.

pseudo-continuous arterial spin labelling (PCASL), labelling duration 3000 ms and post-labelling delay 2000 ms, TR/TE 5500/29.6 ms, 3.8 × 3.8 × 3.8 mm acquisition voxel size reconstructed at 1.9 × 1.9 × 3.8 mm, 24 slices, TA 5.30 min, including two M0 acquisitions with opposite phase-encoding direction and TA 11 s each. All imaging sequences were automatically aligned using an auto-align localiser sequence. If necessary, manual adjustments were made. For PCASL, the labelling plane was manually positioned perpendicular to the orientation of the internal carotid arteries distal to the bifurcation using acquired single-slice coronal and sagittal phase contrast vessel images covering the head and neck. To reduce within-subject variability in MRI scans, we followed recommendations as previously described,35 that is, careful positioning of the participant in the scanner, use of same scanner and head coil throughout the study, automated checks of sequence parameters on every acquired dataset and standardised visual image quality control.

ubject variability in MRI scans, we followed recommendations as previously described,35 that is, careful positioning of the participant in the scanner, use of same scanner and head coil throughout the study, automated checks of sequence parameters on every acquired dataset and standardised visual image quality control. Cognitive assessment Full cognitive assessment. During the pre-visit and one-year follow-up visit, participants underwent an extensive cognitive assessment, directed to measure especially (change in) information processing speed and attention and executive functioning, being the cognitive domains particularly affected by SVD.36,37 Information processing speed was assessed using cards I and II of the Stroop Color-Word test,38 the Symbol Digit Modalities task39 and the Trail Making Test A (TMT-A).40 Attention and executive functioning was assessed using Stroop card III, TMT-B, the Brixton Spatial Anticipation Test41 and a verbal fluency task in which participants had to name as many animals as possible in one minute. Furthermore, the Mini-Mental State Examination (MMSE)42 was administered to evaluate global cognitive functioning. Working memory was investigated with the Digit Span Forward and Backward of the Wechsler Adult intelligence Scale-III43 and verbal memory with the three-trial version of the Rey Auditory Verbal Learning Test (RAVLT)44 including delayed free recall and recognition after approximately 30 min, during which no other memory or language tests were carried out. At one-year follow-up, a parallel version of the RAVLT was used to prevent material-specific learning effects. Additionally, during the pre-visit premorbid intelligence level was determined with the Dutch version of the National Adult Reading Test (NART).45

ring which no other memory or language tests were carried out. At one-year follow-up, a parallel version of the RAVLT was used to prevent material-specific learning effects. Additionally, during the pre-visit premorbid intelligence level was determined with the Dutch version of the National Adult Reading Test (NART).45 Test of Attentional Performance. To investigate possible acute effects of SVD, participants performed two subtasks of the Test of Attentional Performance (TAP)46 on a laptop, parallel to each monthly MRI session. The computer session started with the Alertness subtask, a sensitive test for attention and processing speed, which has also found to correlate with executive dysfunction, working memory deficits and apathy in patients with CADASIL.47 The Alertness subtask consists of four sessions including 20 trials, in which participants are instructed to press a response button as quickly as possible once an X (target stimulus) appears on the screen. During sessions two and three, the trials are preceded by an auditory warning cue. Furthermore, to examine mental flexibility as part of executive functioning, participants performed successively the letter, digit and alternating sessions of the Flexibility subtask, in which the target stimulus is a letter, digit or alternating a letter or digit. During each trial, a letter and digit are presented on each side of the screen. Participants were instructed to respond as soon as possible by pressing the button on the side of the target stimulus. The letter and digit sessions contain 50 trials and the alternate session 100 trials. For both subtasks, reaction times are given as output. To reduce non-specific learning effects on the TAP, which are generally observed between the first two test administrations, participants performed the TAP for practice purposes during the pre-visit, but these results are not taken into account.

session 100 trials. For both subtasks, reaction times are given as output. To reduce non-specific learning effects on the TAP, which are generally observed between the first two test administrations, participants performed the TAP for practice purposes during the pre-visit, but these results are not taken into account. Motor assessment Parallel to each MRI session, participants performed two motor function tasks. Functional mobility was examined using the Timed Up & Go test.48 The number of steps and the time required (s) were reported. Gait speed (m/s) was determined over a distance of 6 m. For both tests, participants were instructed to walk at their preferred walking speed. Physical assessment Blood pressure and pulse rate. Blood pressure and pulse rate were measured in sitting position after 5 min of rest. During the pre-visit, blood pressure was assessed once on both sides. Next, blood pressure and pulse rate were measured three times with 1-min rest between each measurement, both during the pre-visit and every subsequent monthly visit at the arm with the highest recording. Furthermore, blood pressure and pulse rate were measured once during the pre-visit after 1 min in standing position. All measurements were performed with the same blood pressure monitor and time of day was reported.

ent, both during the pre-visit and every subsequent monthly visit at the arm with the highest recording. Furthermore, blood pressure and pulse rate were measured once during the pre-visit after 1 min in standing position. All measurements were performed with the same blood pressure monitor and time of day was reported. Weight, length and waist circumference. Body weight and length were measured without shoes in light clothing. The body mass index (BMI) was calculated as weight divided by height squared (in meters). Additionally, waist circumference was measured between the lowest rib and the iliac crest after a normal expiration. Additional laboratory investigations Glucose level. During the pre-visit, random plasma glucose level was tested through a finger prick test. In case of a glucose level 7.8–11.0 mmol/l, the overnight fasting plasma glucose level was measured parallel to the third MRI visit. In case of a random plasma glucose level >11.0 mmol/l, we considered this as indication for diabetes mellitus. Lipid profile: Lipid profile after overnight fasting was determined, including total cholesterol, high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), non-HDL cholesterol and triglycerides in EDTA plasma. Structured questionnaires Educational level During the pre-visit, educational level was determined using a seven-point Dutch rating scale,49 ranging between one (less than primary school) and seven (academic degree).

Lipid profile: Lipid profile after overnight fasting was determined, including total cholesterol, high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), non-HDL cholesterol and triglycerides in EDTA plasma. Structured questionnaires Educational level During the pre-visit, educational level was determined using a seven-point Dutch rating scale,49 ranging between one (less than primary school) and seven (academic degree). Physical activity The Physical Activity for the Elderly (PASE)50 was used to assess physical activity. This standardised questionnaire evaluates leisure, household and occupational or voluntary activities of the past seven days. To take out seasonal effects, the PASE was administered once in March in all participants. Activities of daily living During the pre-visit and the one-year follow-up visit, disability and level of independence were assessed using the Barthel Index51 and the Instrumental Activities of Daily Living (IADL).52 Depressive symptoms During the pre-visit and the one-year follow-up visit using the Center of Epidemiologic Studies Depression Scale (CES-D),53 the presence of depressive symptoms was assessed. Substance use Participants were asked about smoking habits, alcohol consumption and drug use. For each we recorded during the pre-visit whether they ever used the substance, age started, current consumption and if quitted, age quitted and previous consumption. Thereafter, during each visit, changes in substance use were recorded.

se Participants were asked about smoking habits, alcohol consumption and drug use. For each we recorded during the pre-visit whether they ever used the substance, age started, current consumption and if quitted, age quitted and previous consumption. Thereafter, during each visit, changes in substance use were recorded. Trigger factors and events Any clinical event for which participants sought medical attention or any stressful life event, investigated through the List of Threatening Experiences (LTE),54 during the past year and subsequently since each previous study visit was reported. Furthermore, prior to MRI scanning, participants were asked about fever, influenza and alcohol consumption during the previous 24 h and cigarette smoking and liquid consumption during the previous 1 h. In case of hospitalisation during the study period, the treating physician was contacted to obtain the relevant information on the event, which was adjudicated by the appropriate specialist to confirm the diagnosis. In case of death, the general practitioner was contacted to obtain information about the cause of death. Medication use Baseline medication use and any change therein during the study period were reported and classified according to the Anatomical Therapeutic Chemical (ATC) classification system (World Health Organization, Collaborating Centre for Drug Statistics and Methodology, www.whocc.no/atc_ddd_index/).

Trigger factors and events Any clinical event for which participants sought medical attention or any stressful life event, investigated through the List of Threatening Experiences (LTE),54 during the past year and subsequently since each previous study visit was reported. Furthermore, prior to MRI scanning, participants were asked about fever, influenza and alcohol consumption during the previous 24 h and cigarette smoking and liquid consumption during the previous 1 h. In case of hospitalisation during the study period, the treating physician was contacted to obtain the relevant information on the event, which was adjudicated by the appropriate specialist to confirm the diagnosis. In case of death, the general practitioner was contacted to obtain information about the cause of death. Medication use Baseline medication use and any change therein during the study period were reported and classified according to the Anatomical Therapeutic Chemical (ATC) classification system (World Health Organization, Collaborating Centre for Drug Statistics and Methodology, www.whocc.no/atc_ddd_index/). Blood sampling 30 ml blood was collected at three different time points. During the pre-visit and the last study visit, non-fasting blood (serum and plasma) was collected for future biochemical analyses. Part of the samples was stored for future DNA and biomarker analyses. Parallel to the third MRI visit, fasting EDTA blood was drawn for immunological analyses. Briefly, flow cytometry analysis was used to determine monocyte subsets and ex vivo stimulation of peripheral blood mononuclear cells was performed to explore cytokine production capacity, as described previously.55

rker analyses. Parallel to the third MRI visit, fasting EDTA blood was drawn for immunological analyses. Briefly, flow cytometry analysis was used to determine monocyte subsets and ex vivo stimulation of peripheral blood mononuclear cells was performed to explore cytokine production capacity, as described previously.55 Primary and secondary outcomes The primary outcome is the monthly incidence of DWI+ lesions. Secondary outcomes are the evolution of DWI+ lesions on MRI (into WMH, lacune, microbleed or disappear) and the effects of these lesions on cortical thickness, structural and functional connectivity and cognitive and motor performance. Sample size consideration This study is powered to detect an increase in WMH, proposed to be caused by DWI+ lesions. The progression of WMH is low (<0.5 ml/year) in individuals with mild SVD, but higher (>2.0 ml/year) in individuals with severe SVD.56 The current study is powered to detect a mean increase of 1.2 ml WMH over a 40-week period in individuals with mild to severe SVD. To detect this increase (with a power of 80% and an α = .05), a sample size of 39 participants is required. Taking into account a loss to follow-up of 20%, we aimed to include 50 participants. Discussion To the best of our knowledge, the RUN DMC – InTENse study is the first study performing both neuroimaging and extensive clinical assessments with such a high frequency among a relatively large number of individuals with SVD.

Sample size consideration This study is powered to detect an increase in WMH, proposed to be caused by DWI+ lesions. The progression of WMH is low (<0.5 ml/year) in individuals with mild SVD, but higher (>2.0 ml/year) in individuals with severe SVD.56 The current study is powered to detect a mean increase of 1.2 ml WMH over a 40-week period in individuals with mild to severe SVD. To detect this increase (with a power of 80% and an α = .05), a sample size of 39 participants is required. Taking into account a loss to follow-up of 20%, we aimed to include 50 participants. Discussion To the best of our knowledge, the RUN DMC – InTENse study is the first study performing both neuroimaging and extensive clinical assessments with such a high frequency among a relatively large number of individuals with SVD. The main innovative aspect of the study includes its high-frequency serial imaging design. Due to our monthly visits, we are less likely to miss acute events and we are able to closely monitor the evolution of SVD lesions. Another strong element of our study includes the in-depth phenotyping of subjects with SVD. Since all individuals were retrieved from the RUN DMC study, nine years of prior imaging and clinical data was available. Therefore, we could carefully select our participants, making sure to include individuals with a high likelihood of SVD progression during the study period and to exclude individuals with other possible or additional causes of ischaemia, such as carotid stenosis or atrial fibrillation. Moreover, our participants have extensive experience with participation in research, reducing the chance of drop-out. Another powerful aspect of the study includes the state of the art multimodal imaging protocol, enabling us to apply advanced imaging analyses such as advanced diffusion modelling and iron mapping.

ion. Moreover, our participants have extensive experience with participation in research, reducing the chance of drop-out. Another powerful aspect of the study includes the state of the art multimodal imaging protocol, enabling us to apply advanced imaging analyses such as advanced diffusion modelling and iron mapping. However, a few limitations should also be noted. First, although recruitment from the RUN DMC is a major strength, the sample of relatively healthy individuals selected in the RUN DMC – InTENse study may not be representative of the entire population, limiting the external validity of our results. Second, although the current study contains the largest sample size of individuals with short-term serial imaging data in its field, we acknowledge that the sample size is still relatively small. The RUN DMC – InTENse study will shed new light on the role of DWI+ lesions in the pathophysiology of SVD, which will be of importance for clinical practice. Determining the role of acute, potentially ischaemic, processes in SVD progression might be informative for the development of new treatment strategies. Furthermore, the occurrence of DWI+ lesions may be used as surrogate marker in future clinical trials aimed at slowing SVD progression. Finally, our study will provide a time lapse of (remote) sequelae in the brain following the development of a DWI+ lesion and will provide insight in the effects of these lesions on cognitive and motor performance.

DWI+ lesions may be used as surrogate marker in future clinical trials aimed at slowing SVD progression. Finally, our study will provide a time lapse of (remote) sequelae in the brain following the development of a DWI+ lesion and will provide insight in the effects of these lesions on cognitive and motor performance. To conclude, the RUN DMC – InTENse study will provide a better understanding of the role of DWI+ lesions in the pathophysiology of SVD and will further unravel the structural and functional consequences and the clinical importance of these lesions, with an unprecedented temporal resolution. Understanding the role of acute, potentially ischaemic, processes in SVD may provide new strategies for therapies. Acknowledgement We would like to thank Sjacky Cooijmans, Surekha Gadgil, Daniek van Gils, Karlijn Keizer, Jabke de Klerk, Iridi Stollman and Joost Wissink for their help in collection of the data. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Acknowledgement We would like to thank Sjacky Cooijmans, Surekha Gadgil, Daniek van Gils, Karlijn Keizer, Jabke de Klerk, Iridi Stollman and Joost Wissink for their help in collection of the data. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: AMT was supported by the Dutch Heart Foundation (grant number, 2016 T044). NPR received funding from the European Union’s Horizon, 2020 research and innovation program under grant agreement no. 667837. CJMK was supported by a clinical established investigator grant of the Dutch Heart Foundation (grant no. 2012 T077) and an Aspasia grant from The Netherlands Organisation for Health Research and Development (ZonMw grant no. 015.008.048). FEdL was supported by a clinical established investigator grant of the Dutch Heart Foundation (grant no. 2014 T060) and by a VIDI innovational grant from The Netherlands Organisation for Health Research and Development (ZonMw grant no. 016.126.351). This work was supported by the Vascular Dementia Research Foundation. Informed consent Written informed consent was obtained from all subjects before the study. Ethical approval Ethical approval for this study was obtained from the medical ethics committee region Arnhem-Nijmegen on 6 October, 2015 (ID: NL53939.091.15). Guarantor FEdL.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: AMT was supported by the Dutch Heart Foundation (grant number, 2016 T044). NPR received funding from the European Union’s Horizon, 2020 research and innovation program under grant agreement no. 667837. CJMK was supported by a clinical established investigator grant of the Dutch Heart Foundation (grant no. 2012 T077) and an Aspasia grant from The Netherlands Organisation for Health Research and Development (ZonMw grant no. 015.008.048). FEdL was supported by a clinical established investigator grant of the Dutch Heart Foundation (grant no. 2014 T060) and by a VIDI innovational grant from The Netherlands Organisation for Health Research and Development (ZonMw grant no. 016.126.351). This work was supported by the Vascular Dementia Research Foundation. Informed consent Written informed consent was obtained from all subjects before the study. Ethical approval Ethical approval for this study was obtained from the medical ethics committee region Arnhem-Nijmegen on 6 October, 2015 (ID: NL53939.091.15). Guarantor FEdL. Contributorship AtT and FEdL wrote the first draft of the manuscript and designed the tables and figure. Study concept and idea: FEdL, MD, AMT and AtT. All authors were involved in designing aspects of the study related to his/her field or assessment of the data. AtT, KW and MPN collected the data. All authors revised the article for intellectual content and approved the final version of the manuscript.

Introduction Stroke remains one of the leading causes of death and disability in Europe, and projections show that with a ‘business as usual’ approach, the burden of stroke will not decrease in the next decade or beyond. An important contributing factor to this is that the number of older persons in Europe is rising, with a projected increase of 35% between 2017 and 2050.1 Fortunately, there is compelling evidence that stroke is highly preventable, treatable and manageable, and the potential exists to drastically reduce the burden of stroke and its long-term consequences. However, this requires the joint actions of ministries of health, other governmental bodies, scientific and stroke support organisations, healthcare professionals, clinical and preclinical researchers and the pharmaceutical and device industries.

o drastically reduce the burden of stroke and its long-term consequences. However, this requires the joint actions of ministries of health, other governmental bodies, scientific and stroke support organisations, healthcare professionals, clinical and preclinical researchers and the pharmaceutical and device industries. To this end, two previous pan-European consensus meetings, the 1995 and 2006 Helsingborg meetings,2,3 were convened to review the scientific evidence and the state of current services and to set targets for the development of stroke care for the decade to follow. The European Stroke Organisation (ESO) has prepared a European Stroke Action Plan (ESAP) for the years 2018 to 2030, in cooperation with the Stroke Alliance for Europe (SAFE). The ESAP adheres to the format of the Helsingborg Declarations, presenting a review of the ‘state of the art’, the state of current services, research and development priorities and targets for a series of domains in stroke care (organisation of stroke services, management of acute stroke, prevention, rehabilitation, evaluation of stroke outcome and quality assessment). The ESAP includes two additional domains, on primary prevention and life after stroke, along with research and development priorities for translational stroke research. ESAP 2018‒2030 complements the WHO Global Action Plan on non-communicable diseases (NCDs) 2013–2020, the WHO-Europe NCD Action Plan and the UN Sustainable Development Goals for 2015 to 2030.

s, on primary prevention and life after stroke, along with research and development priorities for translational stroke research. ESAP 2018‒2030 complements the WHO Global Action Plan on non-communicable diseases (NCDs) 2013–2020, the WHO-Europe NCD Action Plan and the UN Sustainable Development Goals for 2015 to 2030. Overarching targets for 2030 For each domain of the 2018 to 2030 ESAP, specific targets are being set, as detailed in the following sections. Beyond these targets, four overarching targets for 2030 have been identified: to reduce the absolute number of strokes in Europe by 10% to treat 90% or more of all patients with stroke in Europe in a dedicated stroke unit as the first level of care to have national plans for stroke encompassing the entire chain of care from primary prevention to life after stroke. to fully implement national strategies for multisector public health interventions to promote and facilitate a healthy lifestyle, and reduce environmental (including air pollution), socioeconomic and educational factors that increase the risk of stroke. Methods The work was led by a steering committee. The documents were prepared by working groups for each of the seven domains, with patient organisation representatives in each domain. Research priorities were identified by two persons for each domain. One working group prepared research priorities for translational stroke research. The draft documents were open to public comments during a one-month period. The final document was prepared after a workshop in Munich on 21–23 March 2018 that was streamed live.

priorities were identified by two persons for each domain. One working group prepared research priorities for translational stroke research. The draft documents were open to public comments during a one-month period. The final document was prepared after a workshop in Munich on 21–23 March 2018 that was streamed live. Primary prevention State of the art Primary prevention of stroke is part of both primary healthcare and public health. Stroke prevention, which involves both pharmacological and non-pharmacological interventions, targets many of the same risk factors as those involved in other cardiovascular diseases (CVDs) and other NCDs. Thus, primary prevention of stroke has two integral components, which can be implemented at different levels. First, measures to promote a healthy lifestyle, such as smoking cessation and reducing alcohol intake, can be implemented at the individual patient level. Second, reductions in air pollution and improvements in socioeconomic and educational status require action at governmental and societal level.

at different levels. First, measures to promote a healthy lifestyle, such as smoking cessation and reducing alcohol intake, can be implemented at the individual patient level. Second, reductions in air pollution and improvements in socioeconomic and educational status require action at governmental and societal level. Risk factor modification Ten modifiable risk factors may account for the vast majority of the population attributable risk of stroke, across age groups, genders and ethnicity. These risk factors are: hypertension, smoking, dyslipidaemia, unhealthy diet, physical inactivity, obesity, diabetes mellitus, cardiac disease, excessive alcohol intake and psychosocial factors.4 There is overwhelming evidence that the treatment of cardiovascular risk factors reduces the risk of stroke, but the target levels in primary prevention are less stringent than those in secondary prevention and depend on comorbidities such as diabetes mellitus and active smoking; these targets may differ between men and women.5 There is good evidence to support risk factor modification as primary prevention in individuals aged 40‒75 years, but the optimal target levels for the management of hypertension and dyslipidaemia in persons outside this age range are less well established.5,6

e smoking; these targets may differ between men and women.5 There is good evidence to support risk factor modification as primary prevention in individuals aged 40‒75 years, but the optimal target levels for the management of hypertension and dyslipidaemia in persons outside this age range are less well established.5,6 Hypertension, diabetes mellitus, dyslipidaemia, smoking, unhealthy diet, obesity, physical inactivity and atrial fibrillation (AF) are the strongest and most common modifiable risk factors for stroke,7,8 and most guidelines include these risk factors. Based on an extensive literature, most primary prevention guidelines recommend the use of risk tables, which take age and comorbidities into account. In a European setting, the Systematic COronary Risk Evaluation (SCORE) tables, accessible through the European Society of Cardiology website, are appropriate, since they are derived from a mixed European population. An adapted version for older people, SCORE-OP, is available.9 Self-management, with or without the use of technology such as eHealth and mobile health (mHealth) initiatives, may offer the possibility of delivering primary prevention in a range of healthcare settings, including geographically remote areas in Europe. However, the expected impact on risk factor improvement is modest at best, and sustainability of effects is a major challenge.10 A consistent and constructive collaboration between healthcare professionals in primary, secondary and tertiary care is likely to be essential for sustainable primary prevention.

s in Europe. However, the expected impact on risk factor improvement is modest at best, and sustainability of effects is a major challenge.10 A consistent and constructive collaboration between healthcare professionals in primary, secondary and tertiary care is likely to be essential for sustainable primary prevention. Primary prevention in public health Primary prevention can be delivered at the individual, community and population levels. Public health interventions targeting highly prevalent risk factors that do not require pharmacological intervention, including discouraging smoking and encouraging a healthy lifestyle, should be implemented at several levels; such interventions may include legislative changes, media campaigns, labelling of food and educational and preventive measures in schools, workplaces and the community. Although there is only limited high-quality evidence for a direct effect on the incidence of stroke and CVD, initiatives such as building cycle lanes, guiding people to stairs, serving healthy food in public places, smoking bans, decreasing the amount of salt and sugar in processed food and soft drinks, health education and public health campaigns to increase awareness of modifiable stroke risk factors are sensible public health interventions and should be pursued.

, guiding people to stairs, serving healthy food in public places, smoking bans, decreasing the amount of salt and sugar in processed food and soft drinks, health education and public health campaigns to increase awareness of modifiable stroke risk factors are sensible public health interventions and should be pursued. From a population perspective, it is not sufficient to target only those individuals at high risk of CVD. The majority of strokes will develop in persons with low or intermediate risk,11 and hence broad prevention strategies targeting the wider population are indicated, particularly considering the high prevalence of some of the strongest risk factors for stroke. A combination of opportunistic screening of those at increased risk (based, for example, on demographic characteristics) and systematic screening for risk factors is probably most appropriate.5 In particular, raising awareness of hypertension as a risk factor for stroke could increase early detection rates and lead to earlier initiation of treatment. Many risk factors are more prevalent among people with low socioeconomic status, and extra effort to reach these people with preventive strategies is warranted.12 In all situations, the potential benefits and risks ‒ such as over-medicalisation ‒ should be carefully evaluated.

d lead to earlier initiation of treatment. Many risk factors are more prevalent among people with low socioeconomic status, and extra effort to reach these people with preventive strategies is warranted.12 In all situations, the potential benefits and risks ‒ such as over-medicalisation ‒ should be carefully evaluated. State of current services There are major differences among European countries, indirectly reflecting the state of local services, in terms of risk factor prevalence and control.7 Moreover, awareness of the modifiable risk factors for stroke can vary markedly between populations.8 The organisation of primary prevention services differs greatly between European countries. Some countries, such as the Netherlands and the United Kingdom, have a strong primary care system, and extensive guidelines that include the primary prevention of CVD are in place, whereas in other countries, it is less clear who is principally responsible for primary prevention. There are currently no generally accepted European guidelines as to which populations should be screened, for which risk factors and at which age. A national NCD action plan is a prerequisite for successful primary prevention strategies that are feasible and realistic for each country, but not all European countries have such a plan.12 Blood pressure management The prevalence of hypertension in people above the age of 18 years ranges from 20 to 40% in Europe.8 Of those on medical treatment, only a minority sustainably reach their target blood pressure.13

State of current services There are major differences among European countries, indirectly reflecting the state of local services, in terms of risk factor prevalence and control.7 Moreover, awareness of the modifiable risk factors for stroke can vary markedly between populations.8 The organisation of primary prevention services differs greatly between European countries. Some countries, such as the Netherlands and the United Kingdom, have a strong primary care system, and extensive guidelines that include the primary prevention of CVD are in place, whereas in other countries, it is less clear who is principally responsible for primary prevention. There are currently no generally accepted European guidelines as to which populations should be screened, for which risk factors and at which age. A national NCD action plan is a prerequisite for successful primary prevention strategies that are feasible and realistic for each country, but not all European countries have such a plan.12 Blood pressure management The prevalence of hypertension in people above the age of 18 years ranges from 20 to 40% in Europe.8 Of those on medical treatment, only a minority sustainably reach their target blood pressure.13 Smoking A steady decline in smoking since 1990 has occurred among men and women in most European countries. However, 10–28% of the adult populations in European countries still smoke.14,15

Blood pressure management The prevalence of hypertension in people above the age of 18 years ranges from 20 to 40% in Europe.8 Of those on medical treatment, only a minority sustainably reach their target blood pressure.13 Smoking A steady decline in smoking since 1990 has occurred among men and women in most European countries. However, 10–28% of the adult populations in European countries still smoke.14,15 Obesity, diet and physical activity The prevalence of overweight and obesity has increased dramatically over the last 25 years. National programmes to reduce salt consumption have been successful in many European countries, with high evidence of a beneficial effect on cardiovascular mortality in some countries. Furthermore, 6 out of 10 European citizens do not reach the WHO-recommended 150 min of moderate intensity physical activity per week.7 Dyslipidaemia The number of people with untreated dyslipidaemia has decreased over the last 25 years, but adherence to treatment with cholesterol-lowering drugs is still poor, with less than half of patients being sustainably adherent to their medication.16 Diabetes mellitus Diabetes care can be organised in outpatient clinics or in primary care, coordinated by specialised nurses. The percentage of patients with adequate diabetes control can be estimated based on the glycated haemoglobin (HbA1C) level and varies between European countries, reflecting different levels of service.

s mellitus Diabetes care can be organised in outpatient clinics or in primary care, coordinated by specialised nurses. The percentage of patients with adequate diabetes control can be estimated based on the glycated haemoglobin (HbA1C) level and varies between European countries, reflecting different levels of service. Atrial fibrillation AF-related stroke is increasing in incidence and prevalence: this may be related to better detection but is at least partly the result of the increasing incidence of AF with age and the ageing population in most countries. In people with AF with an appropriate CHA2DS2VASc score, the beneficial effect of anticoagulation is evident, but the value of wider screening for AF and the clinical significance of short episodes of paroxysmal AF are currently under debate, particularly if it concerns primary prevention. The balance between benefits and harms of long-term anticoagulation as primary prevention is questionable in some AF patients, particularly those with very short episodes of AF or a low CHA2DS2VASc score.17 Because stroke and CVDs share many common risk factors, research initiatives in primary prevention of stroke should not be undertaken in isolation. Close collaboration with primary prevention initiatives from cardiology, primary healthcare and public health is strongly recommended and should be covered by a regional or national NCD action plan. Such initiatives should be aligned with WHO initiatives such as the Action Plan for NCD, and updated as necessary when these are revised. Research and development priorities

Because stroke and CVDs share many common risk factors, research initiatives in primary prevention of stroke should not be undertaken in isolation. Close collaboration with primary prevention initiatives from cardiology, primary healthcare and public health is strongly recommended and should be covered by a regional or national NCD action plan. Such initiatives should be aligned with WHO initiatives such as the Action Plan for NCD, and updated as necessary when these are revised. Research and development priorities Which factors are responsible for major health disparities in Europe, including risk factor prevalence and control, and access and adherence to primary prevention? What are the effects of low socioeconomic status and other social factors? Can the current risk prediction models be improved by extending the current 10-year risk prediction to 20-year or life-time risk for those at younger age (<40 years), and five-year or life-time risk for those over 75 years? Can further individualisation of primary prevention strategies, for example, by considering multiple comorbidities, poly-pharmacy, geographic and ethnic differences and polygenetic risk profiles, enhance effectiveness? Can public awareness of the potential for primary prevention of stroke be improved by personalised and public health education about modifiable risk factors? What are the benefits and harms of screening for stroke risk factors in different populations, using different approaches such as systematic and opportunistic screening?

Can public awareness of the potential for primary prevention of stroke be improved by personalised and public health education about modifiable risk factors? What are the benefits and harms of screening for stroke risk factors in different populations, using different approaches such as systematic and opportunistic screening? Can adherence to primary prevention interventions be improved by using eHealth or mHealth approaches to encourage self-management, using drug combinations (for example, in ‘polypill’ formulations), and combining individualised approaches with public health interventions? Targets for 2030 Achieving universal access in Europe to primary preventive treatments based on improved and more personalised risk prediction. Full implementation of national strategies for multisectorial public health interventions promoting and facilitating a healthy lifestyle, and reducing environmental, socioeconomic and educational factors that increase the risk of stroke. Making available evidence-based screening and treatment programmes for stroke risk factors in all European countries. Having blood pressure detected and controlled in 80% of persons with hypertension. As with the research and development priorities, close collaboration and alignment with cardiovascular disease prevention targets are essential to meet these proposed targets.

Making available evidence-based screening and treatment programmes for stroke risk factors in all European countries. Having blood pressure detected and controlled in 80% of persons with hypertension. As with the research and development priorities, close collaboration and alignment with cardiovascular disease prevention targets are essential to meet these proposed targets. Organisation of stroke services State of the art The organisation of stroke services is fundamental to quality care across the spectrum of care that progresses from prevention via acute therapy to long-term care. Definitions of the different facilities for stroke care included into this document are provided in Table 1. The focus of this section is on the organisation of acute stroke treatment services, since other elements will be presented in the sections dealing with specific domains. Although organised services exist in most European countries, there is large variability in the practical application of treatment guidelines and adherence to quality indicators.19 Awareness programmes have been shown to have an impact on the number of acute stroke calls, but this effect is not sustained in the long term.20 National or regional quality registries have been implemented in several countries to monitor key performance indicators, and programmes to certify stroke units and stroke centres have been established in others.21 Research activities, covering both experimental and clinical studies, and funding have significantly increased during the last decades.22

istries have been implemented in several countries to monitor key performance indicators, and programmes to certify stroke units and stroke centres have been established in others.21 Research activities, covering both experimental and clinical studies, and funding have significantly increased during the last decades.22 Table 1. Definitions of selected terms used in the European Stroke Action Plan.

istries have been implemented in several countries to monitor key performance indicators, and programmes to certify stroke units and stroke centres have been established in others.21 Research activities, covering both experimental and clinical studies, and funding have significantly increased during the last decades.22 Table 1. Definitions of selected terms used in the European Stroke Action Plan. Stroke unit: A dedicated geographically clearly defined area or ward in a hospital, where stroke patients are admitted and cared for by a multi-professional team (medical, nursing and therapy staff) who have specialist knowledge of cerebral function, training and skills in stroke care with well-defined individual tasks, regular interaction with other disciplines and stroke leadership. This team coordinates care through regular, multidisciplinary meetings.18 Stroke centre: A hospital infrastructure and related processes of care that provide the full pathway of stroke unit care. A stroke centre is the coordinating body of the entire chain of care. This covers pre-hospital care, ongoing rehabilitation and secondary prevention and access to neurosurgical and vascular intervention. A stroke unit is the most important component of a stroke centre. The Stroke Centre provides stroke unit services for the population of its own catchment area and serves as a referral centre for peripheral hospitals with stroke units in case their patients need services which are not available locally.18 Comprehensive stroke unit: A dedicated area where acute stroke management is combined with early mobilisation and rehabilitation and secondary prevention, according to the needs of the patient. Early supported discharge: Early supported discharge is designed to enable the accelerated discharge of stroke patients to their home, providing specialist rehabilitation and social support in a home setting rather than an acute hospital ward. The early supported discharge team comprises a variety of specialist therapists, social and support workers. The team visits patients in their home setting, enabling patients to undergo rehab in a familiar home setting, and thus increasing patient flow and bed availability within the acute hospital. Registry: A system for collecting process and outcome data at regional or national level that achieves near universal coverage.

The team visits patients in their home setting, enabling patients to undergo rehab in a familiar home setting, and thus increasing patient flow and bed availability within the acute hospital. Registry: A system for collecting process and outcome data at regional or national level that achieves near universal coverage. Appropriate training of emergency medical services (EMS) personnel and dispatchers increases the number of patients arriving early at hospital, and several countries have established clear transportation routes to the closest suitable hospital.23 Pre-hospital identification of patients with acute stroke by use of validated tools and scales has been recognised as being important for prompt treatment: the Face Arm Speech Test (FAST) is the easiest and the most commonly used, but all pre-hospital scales demonstrate suboptimal specificity, even with adequate training.24,25 The development of stroke teams (including all disciplines required for acute stroke management including radiology) has been shown to lead to shorter delays and more rapid management, particularly when the team is pre-notified of the imminent arrival of the patient.26,27 The concept of delivering personnel and equipment to the patient via mobile stroke units seems to be effective, but this option is not widely available.28 The same is true for helicopter transportation, which may be useful in certain settings.29

when the team is pre-notified of the imminent arrival of the patient.26,27 The concept of delivering personnel and equipment to the patient via mobile stroke units seems to be effective, but this option is not widely available.28 The same is true for helicopter transportation, which may be useful in certain settings.29 Admission to centralised, rather than non-centralised, facilities for acute hospital care increases the likelihood of receiving thrombolytic treatment, which has a direct impact on stroke outcome.30 Hospitals with greater use of thrombolysis achieve statistically and clinically significantly shorter delays in administering tissue plasminogen activator (tPA) after arrival in hospital.31 This may be also true for mechanical thrombectomy, but data on this are limited at present. Treatment in dedicated stroke units significantly reduces the risk of disability, institutional care and death by 20%, irrespective of stroke severity and age.32 All hospitals should ensure that patients with suspected stroke have rapid access to brain imaging, including vascular imaging at all times, since immediate brain imaging is the most cost-effective approach in stroke.33 Early supported discharge (ESD) in a multidisciplinary setting is beneficial for patients with mild to moderate stroke and is a likely cause of the significant reduction in length of stay in stroke units in these patients.34,35

at all times, since immediate brain imaging is the most cost-effective approach in stroke.33 Early supported discharge (ESD) in a multidisciplinary setting is beneficial for patients with mild to moderate stroke and is a likely cause of the significant reduction in length of stay in stroke units in these patients.34,35 In remote regions without specialist stroke expertise, telemedicine in community hospitals may help to provide evidence-based acute stroke treatment, including thrombolysis.36 Certification and sentinel stroke audit programmes support the delivery of high-quality organised care: it has been shown that high performance in audited process indicators reduces mortality by up to 25%.37 This emphasises the need for continuing quality improvement processes, particularly in view of the known differences in care quality related to time and day of admission.38 Auditing measures lead to a better quality of stroke care, and more sustained stroke care performance over time.39

educes mortality by up to 25%.37 This emphasises the need for continuing quality improvement processes, particularly in view of the known differences in care quality related to time and day of admission.38 Auditing measures lead to a better quality of stroke care, and more sustained stroke care performance over time.39 State of current services Political awareness of the burden of stroke has increased in recent years, but still varies significantly across Europe. Although significant progress has been made, there is still large heterogeneity in the organisation of stroke care in Europe, as shown by a recent ESO survey in 44 countries.40 Hence, reliable and precise information about the structure and organisation of stroke care, and the implementation of stroke management pathways, is still lacking in many countries. In most, but not all, European countries, health services and political structures are linked to a national stroke society, and these linkages help to coordinate stroke services and foster quality improvements of stroke care. Most countries have an EMS system with regional organisation and written protocols for acute stroke. However, many do not have obligatory transport routes to the closest suitable hospital or procedures for pre-hospital notification of hyper-acute stroke. Although stroke symptoms and the importance of immediate action have repeatedly been communicated to the public, awareness is still unsatisfactory among the general population.

any do not have obligatory transport routes to the closest suitable hospital or procedures for pre-hospital notification of hyper-acute stroke. Although stroke symptoms and the importance of immediate action have repeatedly been communicated to the public, awareness is still unsatisfactory among the general population. The crucial impact of time in acute stroke has constantly been stressed, but fewer than 10% of stroke patients reach the hospital within 60 min of symptom onset. In many countries, the symptom-to-hospital delay has not decreased in recent years. Several countries have built a nationwide network of hospitals with stroke units or stroke centres following written protocols. However, only limited information on the definitions of stroke units and comprehensive stroke centres in the various countries is available. Moreover, only a minority of countries have established a certification system with well-defined quality criteria, or a monitored system for benchmarking delivery of care, that is evaluated at regular intervals. Among the options for acute stroke care, mechanical thrombectomy is the latest and the most challenging component, and there is currently huge variation in the delivery of this treatment across Europe.40

d quality criteria, or a monitored system for benchmarking delivery of care, that is evaluated at regular intervals. Among the options for acute stroke care, mechanical thrombectomy is the latest and the most challenging component, and there is currently huge variation in the delivery of this treatment across Europe.40 Only a few countries have established a continuous quality improvement system with a predefined set of criteria that are regularly measured and compared with benchmarks. The reimbursement structures for stroke care are still highly variable between European countries, leading to gaps in care quality in some countries. The challenges for acute stroke care differ markedly between countries, and between different regions (e.g. cities, rural areas), but only a few countries monitor the delivery of stroke care at a patient level to ensure a defined standard of service. Research and development priorities Can we identify the most relevant barriers to the implementation of evidence-based stroke care? What is the role of economical aspects? What is the health-economic impact of strokes and the return of investment in stroke care? Which are the most cost-effective concepts to improve the organisation of stroke care in countries with limited resources? Which are the minimum educational criteria for stroke experts (physicians, nurses or therapists)? What are the optimum numbers and ratios of stroke centres and stroke units for municipal and rural areas? What is the role of telemedicine systems for acute stroke, rehabilitation and long-term care?

Which are the most cost-effective concepts to improve the organisation of stroke care in countries with limited resources? Which are the minimum educational criteria for stroke experts (physicians, nurses or therapists)? What are the optimum numbers and ratios of stroke centres and stroke units for municipal and rural areas? What is the role of telemedicine systems for acute stroke, rehabilitation and long-term care? What is the appropriate structure to manage childhood stroke? How can regional networks of EMS, stroke units and rehabilitation centres be developed most effectively? How can evidence-based media campaigns be organised in order to promote public awareness and knowledge of stroke signs and the importance of immediate action? What elements are needed to enable more effective participation in decision making among patients and relatives? Targets for 2030 Establishing a medical society and stroke support organisation in each country, which collaborates closely with the responsible body in developing, implementing and auditing the national stroke plan. Guiding national stroke care by evidence-based pathways that cover the entire chain of care. These pathways are understood by the public and may be adapted to meet regional circumstances to ensure equal access to stroke care irrespective of patient characteristics, region and time of hospitalisation. Managing and delivering stroke care by competent personnel and teams, and creating plans for effective recruitment and training as part of a national stroke plan.

Guiding national stroke care by evidence-based pathways that cover the entire chain of care. These pathways are understood by the public and may be adapted to meet regional circumstances to ensure equal access to stroke care irrespective of patient characteristics, region and time of hospitalisation. Managing and delivering stroke care by competent personnel and teams, and creating plans for effective recruitment and training as part of a national stroke plan. All stroke units and other stroke services undergo regular certifications or equivalent auditing processes for quality improvement. Management of acute stroke State of the art Acute stroke is a medical emergency. The benefit of recanalisation therapies in patients with acute ischaemic stroke is strongly time-dependent, with earlier intervention achieving better outcomes.41 Stroke care systems should therefore minimise the time to assessment and initiation of treatment, before brain injury becomes irreversible.42,43 All patients with ischaemic stroke or intracranial haemorrhage (ICH) benefit from specialist attention and organised care within a designated stroke unit (number needed to treat (NNT) to prevent poor outcome, 18) including rapid diagnostic imaging, probably due in part to the prevention or early treatment of complications such as infections and fever.32 Randomised trials testing strategies to prevent complications are ongoing.32,44

d care within a designated stroke unit (number needed to treat (NNT) to prevent poor outcome, 18) including rapid diagnostic imaging, probably due in part to the prevention or early treatment of complications such as infections and fever.32 Randomised trials testing strategies to prevent complications are ongoing.32,44 For ischaemic stroke, aspirin ‒ started within two days of the onset of symptoms ‒ has a small benefit (NNT 79) but is suitable for a large number of patients. Although unproven, the benefit of early treatment with clopidogrel is probably comparable. In selected patients with acute ischaemic stroke or TIA, dual antiplatelet therapy in the first weeks confers additional benefit.45 Selected interventions, including intravenous thrombolysis, endovascular thrombectomy (EVT) and hemicraniectomy, offer greater benefit but are applicable to fewer patients than antiplatelet therapy. Timely restoration of blood flow through intravenous thrombolysis (IVT, NNT 5‒9) or EVT (NNT 3) is the first priority. The earlier the treatment with IVT is initiated, the greater the benefit, irrespective of age or stroke severity.42 In patients with acute ischaemic stroke caused by occlusion of a proximal intracranial artery of the anterior circulation, EVT strongly increases the chance of a good outcome and, like alteplase, improves global functional outcomes. The initial evidence of benefit with EVT was largely limited to patients in whom treatment could be started within 6 h of symptom onset, but recent evidence suggests that, in highly selected patients with acute ischaemic stroke and considerable residual mismatch between perfusion deficit and infarct volume, EVT can offer benefit when used up to 24 h of symptom onset. In recent trials, EVT was associated with a 19% absolute reduction in the risk of a poor outcome, but nevertheless 29‒67% of the patients randomised to the intervention arm were dead or dependent at three months.46–50

een perfusion deficit and infarct volume, EVT can offer benefit when used up to 24 h of symptom onset. In recent trials, EVT was associated with a 19% absolute reduction in the risk of a poor outcome, but nevertheless 29‒67% of the patients randomised to the intervention arm were dead or dependent at three months.46–50 The wake up study reported that in selected patients with acute stroke with an unknown time of onset, intravenous alteplase guided by a mismatch between diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) in the region of ischemia is beneficial.51 For patients aged 60 years or younger with large, space-occupying infarctions, surgical decompression involving hemicraniectomy and duraplasty strongly reduces the risk of death or dependency (NNT 2‒4).52 The benefit in older patients is questionable because surgical decompression in the elderly reduces the risk of death at the cost of a large increase in the risk of long-term dependency in activities of daily living (ADL).53 For patients with ICH, care on a stroke unit is the only treatment that has been demonstrated to be beneficial. Early intensive blood pressure lowering may improve outcome, although the NNT is relatively large.54 To date, the use of haemostatic agents to prevent early haematoma growth has not been proven effective. Surgical treatment may be life-saving in some patients, but a surgical intervention policy has not been shown to improve outcome in patients with ICH in general.55,56

outcome, although the NNT is relatively large.54 To date, the use of haemostatic agents to prevent early haematoma growth has not been proven effective. Surgical treatment may be life-saving in some patients, but a surgical intervention policy has not been shown to improve outcome in patients with ICH in general.55,56 In patients with subarachnoid haemorrhage (SAH) caused by a rupture of an intracranial aneurysm, the primary goal is the prevention and treatment of complications such as rebleeding, delayed cerebral ischaemia and hydrocephalus. The risk of rebleeding can be reduced by occlusion of the aneurysm with endovascular or microneurosurgical techniques; coiling is preferred in cases where both treatment options appear equally feasible. Oral nimodipine reduces the risk of delayed cerebral ischaemia and increases the chance of a favourable outcome.57,58 Early, short-term, treatment with tranexamic acid to reduce the risk of recurrent SAH before the aneurysm is secured, is currently under study. Many in-hospital deaths of stroke patients occur after a decision to withhold or withdraw life-sustaining treatments. Observational studies have demonstrated a strong relation between treatment restrictions, such as do-not-resuscitate orders and an increased risk of in-hospital mortality. It remains uncertain whether this relation is causal, and whether refraining from do-not-resuscitate orders would result in better outcomes.59

ents. Observational studies have demonstrated a strong relation between treatment restrictions, such as do-not-resuscitate orders and an increased risk of in-hospital mortality. It remains uncertain whether this relation is causal, and whether refraining from do-not-resuscitate orders would result in better outcomes.59 State of current services A recent survey of access to, and delivery rates of, acute stroke unit care, IVT and EVT in 44 European countries found major inequalities in acute stroke treatment between and within countries.40 Overall, there were 1.5 stroke units per 1000 annual incident ischaemic strokes, but individual country-level data indicate that both access to and delivery of acute stroke care are insufficient or lacking in many countries. Furthermore, regional differences were reported within 28 countries. Fewer than 20% of patients with acute ischaemic stroke had access to treatment with IVT, and the overall rate of IVT in incident ischaemic stroke was 7.3%, with considerable inter- and intra-country variability. EVT was performed in only 0.4 centres per 1000 annual incident ischaemic strokes, and only 1.9% of all acute ischaemic stroke patients were estimated to have received this treatment. Moreover, 28 countries lacked full EVT coverage.40

ent ischaemic stroke was 7.3%, with considerable inter- and intra-country variability. EVT was performed in only 0.4 centres per 1000 annual incident ischaemic strokes, and only 1.9% of all acute ischaemic stroke patients were estimated to have received this treatment. Moreover, 28 countries lacked full EVT coverage.40 The mainly Western European SITS-MOST registry has reported a median onset-to-treatment time of 140 min, compared with 150 min reported by the Eastern European SITS-EAST registry. A median door-to-needle (DNT) time of 70 min has been reported in international European registries; continuous improvement of processes can limit median DNT times to just 20 min in experienced and high-volume centres. Guidelines for the management of ICH generally recommend early blood pressure control in patients with acute ICH. Unfortunately, European data on the use of antihypertensive treatment in this setting, and the degree of blood pressure control achieved, are scarce. Research and development priorities How can disparities in access to stroke unit care across European countries be diminished? Which reperfusion options should be used, based on patient-, service- and cost-specific factors? How can the speed, safety and effectiveness of reperfusion approaches (drugs or devices) be optimised in Europe? Which pharmacological or other strategies will reduce the extent of irreversible brain damage in ischaemic stroke patients before recanalisation therapies are started?