The Clinical Research Center for Stroke (CRCS; director, Dr. Byung-Woo Yoon, Seoul National University Hospital), supported by the Korea Healthcare Technology R&D Project, Ministry of Health, Republic of Korea (HI10C2020), was established in 2006 to facilitate multi-center collaborative clinical research and to develop a series of clinical practice guidelines based on data from the Korean population. Six divisions comprise the CRCS and the fifth division, CRCS-5 (principal investigator, Dr. Hee-Joon Bae, Seoul National University Bundang Hospital), is designated to develop clinical guidelines for secondary prevention of stroke through clinical epidemiological research.

To this end, CRCS-5 tried to establish a prospective, multi-center clinical registry for stroke cases from the start of CRCS. Before the commencement of CRCS, the Korean Stroke Registry (KSR; director, Dr. Jae-Kyu Roh, Professor Emeritus at Seoul National University Hospital) was operated through a web-based database to which more than 30 hospitals submitted acute stroke cases.5 The CRCS decided to continue with the baseline structure of database tables from the KSR. CRCS-5 also maintained the overall KSR database structure and added a few tables and fields for its own purposes. To ensure standardization of registration process, CRCS-5 workbook for data collection and variable definitions was established and disseminated throughout the participating center. The CRCS-5 steering committee also regularly held a database workshop to help investigators and registrars being accustomed to the database structure and data entry system at least once a year. Participating researchers in the CRCS-5 registry tried to integrate the registration, auditing, and reporting of the clinical registry into daily clinical practice, and thus established "quality of stroke care monitoring and improvement" as a primary objective of the registry.

The CRCS-5 registry was initiated in April 2008 with nine hospitals. In April 2008, participating centers were mostly located in the Seoul metropolitan area, with one from the central area and one from the southeastern area of South Korea. Such geographical preponderance was slowly alleviated with the participation of Chonnam National University (located in the southwestern part of South Korea) and Dongguk University Ilsan Hospital in January 2011. Jeju National University Hospital joined the CRCS-5 registry in October 2011, and in February 2013, two hospitals joined the CRCS-5 registry-Chungbuk National University Hospital and Ulsan University Hospital. Keimyung University Dongsan Medical Cetner, located in Daegu, began enrolling acute stroke cases in the CRCS-5 registry in May 2014. As of September 2014, 15 hospitals, covering most areas of South Korea, are actively registering stroke cases in the CRCS-5 registry (Figure 1).

The database tables of the CRCS-5 registry have expanded since its initiation. The initial KSR database included demographic information locations and parenchymal lesions, arterial occlusion, stroke mechanism, National Institute of Health Stroke Scale (NIHSS) score at arrival, laboratory values, and treatment information. The CRCS-5 registry maintained the structure of the KSR database as "initial and extended tables" with some minor modifications. In October 2008, a clinical practice-monitoring database table was introduced for quality improvement of stroke care. In November 2009, the CRCS-5 registry set up a new database table dedicated to hyperacute treatment and acute management information. The CRCS-5 registry steering committee has long been in need of prospective stroke outcomes, and thus it implemented a systematic capture strategy for prospective outcomes, including functional status and vascular events in November 2009, selecting five centers with sufficient resources (Eulji General Hospital, Seoul National University Bundang Hospital, Seoul Medical Center, Soonchunhyang University Hospital Seoul, and Inje University Ilsan Paik Hospital). With a piloting operation duration of 14 months, the prospective outcome capture strategy was expanded to all participating CRCS-5 registry centers in January 2011. Prospective outcome information included functional and event outcomes at 3 months and 1 year after stroke onset as well as early neurological deterioration (END; details will be discussed later). Further, temporary database tables were set up for any specific research purposes (Figure 2).

The structure of the CRCS-5 registry and the epidemiological profile of 14,792 stroke cases enrolled up to January 2012 have been discussed previously.6 In the current article, the case profile is up-to-date as of August 2014, and details on hyperacute treatment and outcome information after the index stroke will be discussed.

Acute stroke cases were registered to the CRCS-5 web-based database (http://www.stroke-crc.or.kr/ecrf), and the registration process was required to be started within 48 hours of arrival. The central data manager monitors the number of registered cases from each hospital biweekly, and when the number deviates from the usual level, an inquiry is made to the check for any bias or missing data. The data manager also checks the integrity and completeness of the database bimonthly. Pre-specified queries are used, and peripheral registrars revise erroneous entries upon inquiries. The CRCS-5 registry steering committee has monthly gatherings to review and supervise the process. Details have previously been presented as supplemental data.6 The overall process of case registration, monitoring, inquiry and correction of erroneous data, and outcome capture usually takes 6-7 months to be completed, and then, the up-to-date CRCS-5 database is opened to participating researchers.

As of August 2014, the most recent available data are for cases from April 2008 to November 2013. The figures in the current article are derived from this database. Two-tailed significance values were set as P<0.05. Values are presented as frequencies (percentages), means±standard deviations, or medians [interquartile ranges], as appropriate. Statistical analyses were performed using STATA/MP 13.1 for Mac (STATACorp, College Station, TX, USA).

The most recent database of the CRCS-5 registry as of August 2014 contained stroke cases admitted between April 2008 and November 2013. During these 5.5 years, a total of 28,348 cases were enrolled in the registry, and 27,851 ischemic stroke or transient ischemic attack (TIA) cases were available to be analyzed. In Korea, hemorrhagic stroke cases are usually managed by neurosurgeons, and thus the 497 hemorrhagic stroke cases included in the CRCS-5 registry are not representative of Korean cases. Accordingly, they were removed for the present analyses.

The profile of the analyzable cases is presented in Table 1. The demographic information and profile of vascular risk factors were relatively comparable to previous reports, in spite of a longer inclusion period and more participating centers.6 The overall profile was similar to that from the report of 34,000 cases from the Japan Standard Stroke Registry Study (JSSRS).7 Regarding stroke mechanisms, the CRCS-5 registry showed a higher proportion of cardioembolism than that observed in Taiwan,8 and also a higher proportion of undetermined etiologies in comparison to the JSSRS.7 The discordant rates of stroke mechanisms may originate from differences in work-up intensity and requirements for etiologic evaluation. Onset (defined as last seen normal) to arrival delay was median 12.7 hours in the CRCS-5 registry as of August 2014, a decrease from 14.2 hours observed in the previous report.6 Median NIHSS score decreased by 1 point to 3 points in the current analysis.

Among the 27,851 cases, 13% received hyperacute recanalization treatment, higher than that observed in the US, Taiwan,8,9 and the South Korean general population.10 However, our data should be interpreted with caution, given that the participating centers in the CRCS-5 registry mainly consist of tertiary academic hospitals, meaning that the figure is likely not representative of the primary care environment. The median delay between stroke onset and the initiation of recanalization treatment was 2.3 hours.

On discharge, the proportion of cases with a modified Rankin Scale (mRS) score of 0-1 was 41%, similar to the number from JSSRS.7 The median duration of hospitalization was 7.4 days, and 68% of cases were discharged to their home. Discharge disposition is dependent on cultural background. Specifically, the proportion observed here was lower than data from Taiwan (89%), but much higher than the US (46%) and France (47%).11,12 In-hospital mortality was 2.8%.

Over the 6 years, a few interesting trends were noted in the CRCS-5 registry. The proportion of acute recanalization treatment of any sort increased from 8.8% in 2008 to 14.8% in 2013 (P-for-trend over the 6 years <0.01; Figure 3A). Increasing utilization of acute treatment was also detected in the US,13 but the dramatic increase by 6% over 6 years observed here was exceptional. During the 6 years, the proportion of endovascular recanalization treatment with or without preceding IV alteplase remained unchanged at around 30%-37% (P-for-trend, 0.13). Even after the introduction of the Solitaire neurothrombectomy stent device (ev3 Inc., Plymouth, MN, USA) to South Korea in 2011, the proportion did not change. Such improvement in the utilization of alteplase would result from earlier arrival of acute stroke cases. The proportion of stroke cases who arrived within 3 hours of onset was 27% in 2008 (median onset-to-arrival delay of the whole population, 14.8 hours) but elevated to 32% in 2013 (median 11.9 hours; P-for-trend of arrival within 3 hours from onset <0.01; Figure 3B). However, the vascular risk factors showed relatively consistent profiles over the 6 years (Figure 3C), contrasting the increased prevalence and attributable risk from obesity and hypertension observed in the US over the last 10 years.13,14 One interesting finding was that the length of hospital stay has been consistently decreasing in CRCS-5 registry cases (Figure 3D). The duration of hospitalization is usually considered a major source of healthcare costs for stroke patients, and it can be inferred from the trends that economical pressure for cost reduction is growing in the stroke care system in South Korea.15,16,17

The demographic and vascular risk factor profiles of admitted stroke cases to 14 participating centers were relatively comparable, but the acute stroke treatment showed noticeable variability across hospitals (Table 2). Among the differences, arrival-to-IV alteplase injection time and the utilization of endovascular recanalization treatment were most discordant across centers. Median in-hospital delay ranged from 29 minutes (hospital #13) to 55 minutes (hospital #3), and the proportion of endovascular treatment was between 6% (hospital #10) and 64% (hospital #2). Still, real-world evaluation based on the Korean population is not available to determine whether such center disparity affects outcomes and quality of stroke care. Thus, this issue warrants further investigation.

In November 2009, a separate database table was established in the CRCS-5 registry dedicated to acute stroke and recanalization treatment information, collecting NIHSS scores 24 hours after recanalization treatment, occlusion and recanalization status of cerebral artery, hemorrhagic transformation, utilization of stroke unit, and detailed information on hyperacute recanalization treatment modalities. In this section of the article, a total of 2,724 stroke cases receiving hyperacute recanalization treatment across 13 hospitals will be analyzed (Table 3).

The demographic and vascular risk factor profiles of the 2724 cases were relatively similar to those of the overall CRCS-5 registered stroke cases. Despite similar age, the prevalence of atrial fibrillation was higher among the treated cases and thus NIHSS score at arrival was elevated and more cases were designated as having cardioembolic stroke. The median arrival-to-IV alteplase injection delay was 45 minutes, which was lower than that from the Get With The Guidelines registry data.13 Median procedure time of endovascular treatment was 75 minutes, and 17% of cases were treated beyond the 4.5-hours time window for IV alteplase, 37% (176 cases) of which only had IV alteplase injection without endovascular treatment. Considering recent advances in IV alteplase treatment that note the possibility of extending the IV time window beyond 4.5 hours,18,19 such off-label use of alteplase requires further investigation.

Symptomatic hemorrhagic transformation, associated with NIHSS score increment ≥4 points, occurred in 7% of the treated cases, and any hemorrhagic transformation (including asymptomatic ones) was detected in 27% of the cases. In-hospital mortality was 7%, and at 3 months after stroke, mRS scores of 0-1 were observed in 33%, and mortality was 14%. The prevalence of hemorrhagic transformation was higher in the CRCS-5 registry than it was in the data from Taiwan,20 which may be due to the treatment beyond the conventional time window and the high utilization of endovascular recanalization treatment.

Endovascular recanalization treatment became widely available in clinical practice due to its intuitive outcomes and its function as a rescue treatment.21 Two separate phase II clinical trials in 2,012 reported superior efficacy of newer thrombectomy stent devices, and thus supported wider acceptance of endovascular modalities.22,23,24 Among the 2,724 recanalization-treated cases in the CRCS-5 registry, endovascular treatment with or without preceding IV alteplase was utilized in 38% of cases (Table 4). Combined IV-endovascular recanalization treatment was usually performed for cases arriving within the time window of IV thrombolysis and for those who had higher prevalence of atrial fibrillation, higher NIHSS scores, and lower pre-stroke disability. The arrival-to-initiation of IV alteplase injection was comparable between IV alteplase-only and combined IV and endovascular treatment groups. Between endovascular-only and combined treatment groups, preceding IV alteplase did not cause a significant delay in the intrahospital logistics of stroke cases. Interestingly, a lower dose of IV alteplase (0.6 mg/kg) was utilized in 29% (653) of the 2,724 cases, but the proportion was much higher in the combined IV and endovascular groups at 55% (P-for difference <0.01). In the current descriptive analyses, the combined IV and endovascular treatment group showed a larger degree of NIHSS score decrement over the first 24 hours following treatment, but the prevalence of hemorrhagic transformation was elevated. Mortality rate at discharge and at 3 months after stroke was comparable between the IV-only and endovascular-treated groups, but the proportion of mRS scores of 0-1 was lower in the endovascular group. Although the efficacy of endovascular recanalization was suggested in the prior phase II trials, the scientific evidence supporting the treatment modality is still lacking, as demonstrated in phase III trials with older-generation modalities.25,26,27 Such confusion is prevalent among CRCS-5 registry centers (Figure 4). The real-world effectiveness of hyperacute endovascular intervention warrants further investigation, and the clinical registry with disparity between centers on this issue would be a valuable source for clinical analyses.

The CRCS-5 registry initiated prospective outcome collection in November 2009 with a unified definition of stroke outcomes and capture strategy. The detailed definitions of major outcome variables are presented in Table 5. Long-term recurrent events and functional outcomes have been collected since November 2009 for selected centers, and this was expanded to all participating centers in January 2011 with the introduction of END. The CRCS-5 registry steering committee checks and audits the occurrence of END and captures the rate of long-term outcomes at monthly intervals.

A total of 14,730 cases admitted after January 2011 were available to analyze for ENDs. Among them, END occurred in 2,615 cases (15%), including two ENDs in 342 cases and three separate END events in 47 cases. The total number of END events was 3,004. The following causes led to ENDs: 300 recurrent strokes (10%), 2059 stroke progressions (69%), 171 symptomatic hemorrhagic transformations (6%), 147 other (5%), 300 unknown (10%), and 22 transient ischemic attacks (1%; missing information in 5 cases). Ischemic strokes characterized 92% (275 cases) of recurrent strokes, and 15 cases were found to have hemorrhagic stroke. The median time from stroke onset to END occurrence was 35 hours [interquartile range, 15-80].

END cases tended to be older, have higher prevalence of vascular risk factors, and were dependent before the index stroke (Table 6). They had higher NIHSS scores at arrival and arrived earlier, with higher proportions of large artery atherosclerosis and cardioembolism. Stroke cases with END often had poor outcomes, with a longer length of stay and higher in-hospital mortality, as well as poor 3-month outcomes (Figure 5).

Prospective collection of long-term functional and vascular event outcomes was performed for 19,441 cases. Discharge outcome was available in 19,388 of these cases (99.7%), 3-month outcome in 18,564 cases (96%), and 1-year outcome in 13,703 cases (71%). The availability of 1-year outcome will increase, as the long-term outcome capture is an ongoing process. Cross tables comparing mRS scores at different time points after stroke are presented in Table 7. Overall, the proportion of acute stroke cases with mRS scores 0-1 was 42% at discharge, 50% at 3 months, and 55% at 1 year after stroke (Figure 6). Mortality rates also increased as time passed. In the China National Stroke Registry data, the proportion of mRS scores of 3-5 (dependent but not dead) at 3 months was 25.6% and 19.4% at 1 year. In comparison, the CRCS-5 registry had higher values (37% and 29%, respectively).28

Among the 19,441 cases whose long-term outcomes were available, recurrent stroke occurred in 4.5%, vascular death in 3.0%, and composite events of recurrent stroke, myocardial infarction and vascular death in 6.8% within 365 days of the index stroke (Table 8). The median time between stroke onset and the first event was 34 [7-116] days for recurrent stroke, 78 [21-237] days for myocardial infarction, 8 [3-44] days for vascular death, and 18 [5-90] days for composite events. The median observation period was 365 [116-382] days. Overall, the rate of vascular events after stroke was higher in the early period after stroke but decreased thereafter, for both recurrent stroke and composite events (Figure. 7).

The stroke recurrence rate was higher among older individuals and cases with aggravated initial stroke severity (Table 9). It also differed according to ischemic stroke subtypes; the lowest recurrent rate was detected in small vessel occlusions, and similar rates were observed in large artery atherosclerosis and cardioembolism. Interestingly, the recurrence rate of stroke was lower in cryptogenic stroke cases. This discrepancy across ischemic stroke subtypes was documented previously,29,30,31 but the absolute recurrent event rates from the CRCS-5 registry were lower than previously reported. This might be explained by advances in the management of stroke cases and secondary prevention of ischemic stroke as well as ethnic or cultural differences, on which further investigations with regional collaborations among East Asian countries are warranted for clarification.