One-Year Systolic Blood Pressure Trajectory After Acute Ischemic Stroke


 Although the effect of blood pressure on post-stroke outcome is well-recognized, the long-term trajectory of blood pressure after acute ischemic stroke and its influence on outcomes have not yet been fully elucidated. From a multicenter prospective registry of acute ischemic stroke patients, 5,514 patients with measurements of systolic blood pressure (SBP) at more than 2 of 7 prespecified time-points, up to 1-year after stroke onset, were analyzed. Outcome measures, a composite of stroke recurrence, myocardial infarction and mortality, and each stroke recurrence and mortality, were prospectively collected up to 1-year after stroke onset. The study subjects were categorized into 4 groups according to their SBP trajectories: Low (27.0%), Moderate (59.5%), Persistently high (1.2%), and Slowly dropping (12.4%). After adjustments for pre-determined covariates, the Slowly dropping SBP Group was at higher risk of the composite outcome (hazard ratio, 1.32; 95% confidence interval, 1.05‒1.65), and mortality (1.35; 1.03‒1.78) compared to the Moderate SBP Group. Four main 1-year longitudinal SBP trajectories were identified after acute ischemic stroke. One trajectory, slowly dropping SBP, was particularly prone to adverse outcomes after stroke. These findings provide possible leads for future investigations of SBP control targets after stroke.


Introduction
Clinicians are accustomed to managing blood pressure (BP) according to single measurements with limited attention to longitudinal changes in BP over time. Analysis of BP data based on trajectories is another option, although this approach is often not explored by practitioners in routine o ce practice 1 .
Studies adopting group-based approaches according to BP trajectory patterns are primarily populationbased studies of the effect of longitudinal BP changes over a lifetime. These studies have shown that individuals with sustained, poorly controlled high BP have a higher risk of cardiovascular events or mortality 2-6 .
Whereas an elevation or early surge of BP after acute stroke is well known 7 , some recent studies on BP trajectory after acute ischemic stroke show that BP changes during the acute period have distinct patterns that are associated with prognosis 8- 10 . Speci cally, individuals with distinguishable BP trajectory patterns, whose BP does not drop or remains elevated during the rst few hours after stroke onset, may be at higher risk of a poor prognosis 8 . However, the dynamics of BP beyond the acute stroke time period are unknown, as there is a paucity of studies on the prognostic impact of the long-term time course of BP changes after ischemic stroke. Such knowledge may be important as it may facilitate more appropriate BP management after stroke.
In this study, we describe the patterns of BP changes up to 1-year after ischemic stroke using group-based trajectory models to categorize BP and to explore the associations between BP trajectory groups and stroke outcomes.

Study subjects
Patients with acute ischemic stroke, who were admitted to the 10 participating centers of the Clinical Research Collaboration for Stroke in Korea (CRCS-K) registry 27,28 between January 2010 and December 2011 and who met the study's eligibility criteria, i.e., 1) hospitalization within 7 days of symptom onset (N = 6,547) and 2) documentation of ischemic lesions relevant to stroke symptoms on diffusion-weighted images (DWI) (N = 5,791), were identi ed in the CRCS-K registry database. Those who died during hospitalization due to the index stroke were excluded from this study (N = 158).
Informed consent was waived for collection of clinical information in the CRCS-K registry with approval by the Institutional Review Board (IRB) of all participating centers (Seoul National University Bundang Hospital, Nowon Eulji Medical Center, Inje University Ilsan Paik Hospital, Soonchunhyang University Hospital, Eulji University Hospital, Seoul Medical Center, Yeungnam University Medical Center, Dong-A University Hospital, Chonnam National University Hospital, Hallym University Sacred Heart Hospital) as the data was collected under the purpose of quality of stroke care monitoring and improvement and the data was provided to the researchers after de-identi cation. In addition, analysis of the registry database with additional collection of data for this study was approved by all local ethics committees mentioned above. All methods were carried out in accordance with relevant guidelines and regulations.
Blood pressure and clinical data collection For the 5,633 eligible patients, BP data were collected at 7 time-points after onset (day 0, day 3, day 7, day 30, day 90, day 180, and day 365) by medical record review. BP data obtained after outcome events were excluded from the analysis. BP was measured during hospitalization or at outpatient clinic during routine practice, following the institutional protocols of each hospital. It was recommended to use a standard mercury sphygmomanometer or a non-invasive BP monitoring device on the non-paralytic arm. A total of 452,654 systolic BP (SBP) measurements were collected, along with the date and time of measurement. The measured SBP was allocated to one of the aforementioned 7 time-points that was closest to the measured date or time. The number of patients with allocated SBP data at each time-point and median duration from the measured time-points to the allocated time-points are described in Supplemental Table 1.
Basic demographics and clinical information on vascular risk factors (hypertension, diabetes, hyperlipidemia, atrial brillation, coronary heart disease, current smoking, and previous history of stroke or transient ischemia attack); stroke characteristics, including initial stroke severity according to the National Institute of Health stroke scale (NIHSS) and stroke subtypes according to the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classi cation, with some modi cation 29 ; premorbid functional status, as recorded in the modi ed Rankin scale; symptomatic steno-occlusion (more than 50% of stenosis or occlusion) of relevant major cerebral arteries; acute treatment modalities; and medications at discharge were obtained directly from the CRCS-K registry database.
Information on antihypertensive agents used during hospitalization and in outpatient clinics was collected from the reimbursement claims database of each hospital. The number of antihypertensive agents, date of their prescription, and number of days of treatment were extracted and used for analysis in this study.

Outcome measures
The primary outcome measure was a composite of stroke recurrence, myocardial infarction, and all-cause mortality. The secondary outcome measures were stroke recurrence and mortality. All outcome events were captured prospectively up to 1 year after the index stroke, based on structured telephone interview or during routine follow-up visits in the outpatient clinics. Detailed de nitions of outcomes and the protocols of the CRCS-K registry are published elsewhere 27,28 .

Statistical analysis
We applied a group-based trajectory model approach using the TRAJ procedure of SAS software to determine the SBP trajectories during 1 year after the index stroke and categorizes patients according to the trajectory groups 8,30 . Brie y, this approach is an application of a nite mixture model, in which the longitudinal SBP data were tted and grouped by a maximum likelihood method as a mixture of multiple latent trajectories in a censored normal model with a polynomial function of time 1 . Patients with 2 or more SBP data entries at the aforementioned 7 time-points were eligible for this analysis. The optimal number of groups were determined using the Bayesian information criterion (BIC) comparing 2 × ΔBIC between each number of groups and polynomial orders for time function (Supplemental Tables 2 and 3). Each group was named according to the visual description of the SBP trajectory.
In each group of patients, parameters were described as mean ± SD for interval variables, frequency (%) for categorical variables, and median with interquartile range (IQR) for ordinal variables. Comparisons among groups were made by chi-square tests, one-way analysis of variance, or the Kruskal-Wallis test according to the type of data. The cumulative incidence of the primary and secondary outcomes in each SBP trajectory group was estimated using the Kaplan-Meier (product-limit) method, and crude cumulative incidence was compared among groups using the log-rank test.
For multivariable analysis, a shared frailty model with the participating centers as a random effect was used along with predetermined covariates. Hazard ratios for each outcome among groups were provided by: 1) unadjusted models, 2) model 1 adjusting for age, sex, time interval from onset to hospital arrival stroke subtype, and initial NIHSS score, and 3) model 2 with further adjustments for pre-stroke mRS score; history of hypertension, diabetes, hyperlipidemia, TIA, or stroke; atrial brillation; coronary heart disease; current smoking; intravenous thrombolysis; endovascular thrombectomy; and prescription of antiplatelet, anticoagulant, statin, and/or antihypertensive agents (angiotensin-converting enzyme inhibitor or angiotensin receptor blocker, beta-blocker, calcium channel blocker, diuretics) at discharge. Information on the symptomatic steno-occlusion of a relevant artery was provided.
As a sensitivity analysis, the group-based trajectory modeling approach was restricted to patients who had 3 or more SBP data entries at the 7 time-points, to prove the robustness of the study results.
All statistical analyses were conducted using SAS software version 9.4. (SAS Institute Inc. Cary, NC, USA), and R software version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria). A p-value < 0.05 was considered as statistically signi cant.

Results
A total of 5,514 patients who had at least 2 SBP data entries at the 7 time-points were included in the nal analysis. Using the group-based trajectory model, the patients were grouped into 4 SBP trajectory categories ( Fig. 1 and Supplemental Table 2). Based on the visual depiction of the SBP curves over time, the SBP trajectory groups were named the "Low SBP" (N = 1,487), "Moderate SBP" (N = 3,280), "Persistently high SBP" (N = 66), and "Slowly dropping SBP" (N = 681) groups. In the rst 3 groups, SBP decreased in the rst 3-7 days and remained steady thereafter. The mean SBP in these groups was in the range of approximately 114-116 mmHg in the Low SBP Group, 130-135 mmHg in the Moderate SBP Group, and 147-171 mmHg in the Persistently high SBP Group. In the Slowly dropping SBP Group, the SBP trajectory decreased more slowly over the rst month, from 182 mmHg to 135 mmHg, and then paralleled the SBP trajectory of the Moderate SBP Group.
The patient characteristics differed among the SBP trajectory groups ( Table 1). The Persistently high SBP Group were younger than the other groups and were more likely to have vascular risk factors, such as hypertension or diabetes, whereas the Low SBP Group were more likely to have atrial brillation. Nearly 90% of individuals in the Slowly dropping SBP Group and Persistently high SBP Group had hypertension. More than 70% of individuals in these two groups were on antihypertensive medications at discharge. In terms of antihypertensive drug class, renin-angiotensin-aldosterone system inhibitors were most frequently prescribed at discharge, followed by calcium channel blockers. Beta-blockers or diuretics were prescribed infrequently. The proportion of individuals diagnosed with hypertension before the index stroke was markedly higher in the Slowly dropping SBP Group and in the Persistently high SBP Group than in the other two groups. The antihypertensive prescription rate among those diagnosed with hypertension was 78% at stroke onset in the Slowly dropping SBP Group and exceeded 85% in the other 3 groups. In terms of stroke subtypes, large artery atherosclerosis was most common in the Persistently high SBP Group (59%), while cardioembolic stroke was most common in the Low SBP Group (30%). The median follow-up duration was 373 (IQR, 363-399) days. Overall, the 1-year cumulative incidence was 11.9% for the primary outcome, 5.0% for stroke recurrence, and 8.2% for all-cause mortality. The Slowly dropping SBP Group showed the highest cumulative incidence for the primary outcome during most of the 1-year follow-up period, but the nal 1-year cumulative incidence was comparable in both the Slowly dropping SBP Group and the Persistently high SBP Group (15.7% vs. 15.8%). The 1-year cumulative incidence of mortality was highest in the Slowly dropping SBP Group (11.1%). Interestingly, the 1-year cumulative incidence for all outcomes in the Low SBP Group was comparable with that of the Moderate SBP Group, although numerically, the value was higher in the Moderate SBP Group than in the Low SBP Group (Fig. 2 and Supplemental Table 4).
In both adjusted models 1 and 2, only the Slowly dropping SBP Group had a signi cantly higher hazard ratio, of 1.3, for the primary outcome than the Moderate SBP Group, which had the lowest 1-year cumulative incidence for the primary outcome. The hazard ratio for all-cause mortality was also increased signi cantly in the Slowly dropping SBP Group, but the hazard ratio for stroke recurrence was not. The hazard ratio for all outcomes was not increased signi cantly in the Low SBP Group or the Persistently high SBP Group as compared to the Moderate SBP Group (Table 2). Information on the antihypertensive agent prescription during the 1-year follow-up period was available for 3,627 patients. The overall rate of prescription of antihypertensive agents increased to approximately 70% during the rst 2 months post-stroke and then decreased slightly (Fig. 3). The proportion of those receiving multiple antihypertensive agents reached about 40% at 2 months after stroke onset. The antihypertensive prescription rate was markedly higher in the Persistently High SBP Group and in the Slowly dropping SBP Group than in the other 2 groups, and more than 50% of participants in these groups were on multiple agents after day 30 (Fig. 4).
The sensitivity analysis, which was restricted to subjects who had more than 3 SBP data entries among the 7 possible time-points (N = 4,603), showed similar results for the 4 SBP trajectory groups, with a higher risk for the primary outcome in the Persistently high SBP Group and the Slowly dropping SBP Group (Supplemental Tables 5 and 6, Supplemental Figure).

Discussion
We identi ed 4 distinctive categories of SBP trajectory: the Low SBP, Moderate SBP, Persistently high SBP, and Slowly dropping SBP groups. Longitudinal changes in the mean SBP in the study subjects overall were similar to that of previous ndings, in that more than 80% of patients with acute ischemic stroke had elevated SBP above 140 mmHg early after ischemic stroke; 15 this largely stabilized by 24 hours after stroke onset 7,11 . Using the BP trajectory model, we were able to nd distinguishable patterns that could not be detected by observing the overall group mean of SBP. Furthermore, we observed that distinct patterns were associated with differences in clinical outcomes.
The most noteworthy category among the 4 SBP trajectory groups was the Slowly dropping SBP Group. Compared to the Moderate SBP Group, the Slowly dropping SBP Group had markedly higher SBP (approximately 180 mmHg at stroke onset), which decreased slowly during the rst month, and nally reached a level of 120-130 mmHg; the SBP level at 30 days after the index stroke was similar to that in the Moderate SBP Group. However, the risk of the primary outcome was higher in the Slowly dropping SBP Group than in the Moderate SBP Group, despite the similarity in SBP levels after the rst month. This result is concordant with the ndings from previous studies focusing on BP trajectories in a more acute stroke time period. Our prior study showed that SBP during the rst 24 hours after stroke onset had distinct trajectory groups, and patients who were classi ed by categories as having SBP above 150 mmHg at 24 hours had a higher risk of adverse events, including mortality 8 . Additionally, a secondary analysis of the CATIS trial about the SBP trajectories during the rst week after stroke onset showed a similar result, in that patients with SBP above 160 mmHg had the highest risk of adverse events. 9 Interestingly, in the latter study, patients who initially had a high SBP (approximately 180 mmHg), but which rapidly dropped to 140 mmHg (within 3 days), had a lower risk of mortality than those whose SBP remained high. In our study, the Slowly dropping SBP Group and Persistently high SBP Group may have been comparable to those with high early SBPs in these prior studies.
One possible explanation for the poor outcome in ischemic stroke patients with higher SBP at baseline is that high SBP is a marker of elevated sympathetic activity, which may lead to subsequent cardiovascular complications, and consequently makes an individual more prone to comorbidities, such as infection 12,13 . However, in the Slowly dropping SBP Group, more than one-third of the patients who were diagnosed with hypertension before the index stroke were not on antihypertensive medications at stroke onset. This implies undertreatment of hypertension in this patient group. Uncontrolled and sustained high BP throughout the lifetime is also known to result in poor outcomes. Previous studies analyzing the longterm BP trajectory in the general population consistently showed that those with higher BP have a higher risk of adverse events than those who are normotensive. BP trajectory groups with a high BP level show a higher risk of mortality 2,5,6,14 ; cardiovascular events, such as stroke 4,6 , myocardial infarction 4 , heart failure 14 , and atrial brillation 15 ; and also subclinical markers, such as increased carotid intima-media thickness or left ventricular mass index 3 .
BP drop during the early stage of ischemic stroke is known to result in subsequent neurological deterioration by decreasing cerebral perfusion 16 , and current practice guidelines mention that initiating BP-lowering therapy within the rst 48 or 72 hours of onset may have no bene t 17 . However, eventually lowering SBP to guideline-based levels (e.g., < 140 mmHg or < 130 mmHg) in patients with stroke or transient ischemic attack is recommended to prevent subsequent cardiovascular events 18 . It is not clear when to begin lowering of BP or how quickly target BP levels should be reached in patients with acute ischemic stroke. There have been several clinical trials, such as the CATIS, ENOS, and SCAST trial, all of which failed to show a bene t in the primary outcome endpoint with more intensive BP-lowering therapy in patients with acute stroke [19][20][21] . Our study results imply that the BP trajectory immediately after the acute stage of ischemic stroke is a potential target for BP-lowering interventions. More than half of the Slowly dropping SBP Group received no antihypertensive medication or only one medication at 30 days after stroke onset (Fig. 4). Intense treatment, particularly during the rst 30 days after stroke onset, might improve outcomes in these patients. This area of research needs to be explored further, as physicians may feel that clinical equipoise exists in relation to when and how much to lower BP after stroke; this may explain, at least in part, the nding of no or only 1 antihypertensive medication being prescribed at 30 days in our Slowly dropping SBP Group.
It is interesting to note that the Low SBP Group did not have an evidently lower risk of mortality than the Moderate SBP Group (Table 2 and Supplemental Table 4). This result supports the ndings from previous observational studies that there may be a "J-shaped" association between BP and outcomes 22,23 . About 30% of the Low SBP Group were classi ed as having cardioembolic stroke and about 25% had atrial brillation (Table 1). These ndings might explain both the low SBP and poor outcome in these patients, as atrial brillation results in decreased cardiac contractility and is associated with cardiovascular comorbidities, such as myocardial infarction and heart failure 24,25 . Our data showed that the Low SBP Group were more likely to have atrial brillation and coronary heart disease and to present with more severe neurologic de cits at arrival, all of which could increase mortality (Supplemental Table 7).
Our study had several limitations. First, as we included patients who had SBP measurements taken at no fewer than 2 of 7 time-points, there might be a potential selection bias. On the other hand, only 2 measurements are not adequate for estimating BP trajectories. We intended to maximize the inclusion of such patients to minimize possible selection bias and performed a sensitivity analysis with subjects with 3 or more SBP measurements, which demonstrated the robustness of our study results. Second, although we were able to nd associations between 1-year SBP trajectories and outcomes, we cannot conclude that there is a causal relationship between our main outcome ndings and SBP trajectory results. However, we analyzed SBP data that were obtained before outcome events, in order to maintain the temporal relationships between SBP measurements and outcome events. Third, all the centers participating in the CRCS-K registry are academic hospitals, and therefore the generalizability of the study results to the entire stroke population might be limited. However, the age and sex distributions of the CRCS-K registry subjects are similar to those of the ischemic stroke population in South Korea 26 . Forth, BP measurement protocol and device were not standardized between centers. Although we tried to consider the center effect using the Shared Frailty Model, this heterogeneity should be noted.
In conclusion, SBP trajectory in acute ischemic stroke patients were categorized into four distinct groups and patients in the Slowly dropping SBP group had poor outcome after index stroke event. While only small proportion of patients used antihypertensive agents despite high SBP during the early period in this group, meticulous BP management in these patients might be a potential target for improving outcome.

Declarations
Data availability: The data on which the ndings of this study are based are available from the corresponding author upon reasonable request.   Number of prescribed antihypertensive agents according to day after stroke onset