Five-Day Glycemic Patterns and 3-Month Functional Outcomes After Endovascular Thrombectomy

Article information

J Stroke. 2026;28(3):462-474
Publication date (electronic) : 2026 September 30
doi : https://doi.org/10.5853/jos.2026.03867
1Department of Neurology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea
2Bonghwa-Gun Public Health Center, Bonghwa, Korea
3Clinical Research Center, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea
4Department of Neurology, Research Institute of Clinical Medicine of Jeonbuk National University-Biomedical Research Institute of Jeonbuk National University Hospital, Jeonju, Korea
5Department of Neurology, Chosun University Hospital, Chosun University College of Medicine, Gwangju, Korea
6Department of Neurology, Ewha Womans University Seoul Hospital, Ewha Womans University College of Medicine, Seoul, Korea
7Department of Neurology, Chungbuk National University Hospital, Chungbuk National University College of Medicine, Cheongju, Korea
8Department of Neurology, Seoul National University Bundang Hospital, Seoul National University College of Medicine, Seongnam, Korea
9Department of Neurology, Gyeongsang National University Hospital, Gyeongsang National University School of Medicine, Jinju, Korea
10Department of Neurology, Kyung Hee University Hospital, Kyung Hee University College of Medicine, Seoul, Korea
11Department of Neurology, Ajou University School of Medicine, Ajou University Medical Center, Suwon, Korea
12Department of Neurology, Dong-A University Hospital, College of Medicine, Dong-A University, Busan, Korea
13Department of Neurology, Gil Medical Center, Gachon University College of Medicine, Incheon, Korea
14Department of Neurology, Busan Paik Hospital, College of Medicine, Inje University, Busan, Korea
15Department of Neurology, Chonnam National University Medical School, Chonnam National University Hospital, Gwangju, Korea
16Department of Neurology, Inha University Hospital, Inha University College of Medicine, Incheon, Korea
17Department of Neurology, Yongin Severance Hospital, Yonsei University College of Medicine, Yongin, Korea
18Department of Neurology, Korea University Anam Hospital, Korea University Medicine, Seoul, Korea
19Department of Neurology, Dongtan Sacred Heart Hospital, Hallym University College of Medicine, Hwaseong, Korea
20Department of Neurology, Keimyung University Dongsan Hospital, Keimyung University School of Medicine, Daegu, Korea
21Department of Neurology, Ulsan University Hospital, University of Ulsan College of Medicine, Ulsan, Korea
22Department of Neurology, Daejeon Eulji Medical Center, Eulji University School of Medicine, Daejeon, Korea
23Department of Neurology, Wonkwang University Hospital, School of Medicine, Wonkwang University, Iksan, Korea
24Department of Neurology, Jeju National University Hospital, Jeju National University College of Medicine, Jeju, Korea
25Department of Neurology, Kyungpook National University Hospital, School of Medicine, Kyungpook National University, Daegu, Korea
26Department of Neurology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea
27Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea
Correspondence: Sun U. Kwon Department of Neurology, Asan Medical Center, University of Ulsan College of Medicine, 88, Olympic-ro 43-gil, Songpa-gu, Seoul 05505, Korea Tel: +82-2-3010-3960 E-mail: sunkwon7@gmail.com
Received 2026 June 15; Revised 2026 September 9; Accepted 2026 September 9.

Abstract

Background and Purpose

Admission hyperglycemia is associated with poor outcomes after endovascular thrombectomy (EVT) for acute ischemic stroke, but whether post-EVT glycemic pattern determines functional outcome remains unknown.

Methods

A Retrospective Study for the Prediction of Clinical Outcomes in Acute Ischemic Stroke Patients with Thrombectomy study was a hybrid prospective-retrospective multicenter cohort study nested within the Rescue on Reperfusion Damage in Cerebral Infarction by Nelonemdaz trial (NCT05041010) across 24 tertiary stroke centers in South Korea. Patients with ≥80% measurements within 80–140 mg/dL and no value ≥200 mg/dL were classified as normoglycemic (Group 1); the remainder were classified by group-based trajectory modeling into persistently elevated (Group 2) and variable hyperglycemia (Group 3). The primary outcome was 3-month modified Rankin Scale (mRS) distribution.

Results

Among 457 patients, 319 were included (mean age 73.6±12.2 years; 53.6% male): 97 (30.4%) in Group 1, 159 (49.8%) in Group 2, and 63 (19.7%) in Group 3. Groups 2 and 3 had more diabetes and higher hemoglobin A1c; stroke severity, Alberta Stroke Program Early Computed Tomography Scores, and reperfusion status were comparable. Both hyperglycemic groups had lower odds of a favorable mRS shift than Group 1 (adjusted common odds ratio [OR] 0.35, 95% confidence interval [CI] 0.20–0.61 and 0.49, 0.25–0.94). Functional independence (mRS 0–2) was achieved in 63.9% of Group 1 versus 36.5% and 38.1% in Groups 2 and 3 (adjusted OR 0.26, 95% CI 0.13–0.53 and 0.31, 0.13–0.73). No significant difference was observed between Groups 2 and 3.

Conclusions

Five-day post-EVT hyperglycemic status was independently associated with worse 3-month functional outcomes beyond chronic glycemic burden, supporting re-evaluation of targeted glycemic interventions during this vulnerable window.

Introduction

Hyperglycemia is a well-established predictor of poor outcome, increased mortality, and hemorrhagic transformation (HT) in acute ischemic stroke (AIS) [1-5]. Despite this consistent association, randomized trials on intensive glucose control have largely failed to demonstrate clinical benefit [6,7], leaving the optimal glycemic management strategy in the acute stroke setting unresolved [1].

Patients undergoing endovascular thrombectomy (EVT) may represent a particularly vulnerable population in whom post-procedural glycemic management is critical. Mechanical reperfusion restores cerebral blood flow but simultaneously initiates a phase of metabolic instability, during which the salvaged penumbra remains highly susceptible to secondary injury from glucose-mediated oxidative stress and blood-brain barrier disruption [8]. However, most clinical evidence to date has relied on static, single-point measurement such as admission glucose [9-11]. Other studies have examined glycemic change, by comparing measurements at admission and at 24 hours [12,13]. Such a two-point comparison distinguishes persistent from transient hyperglycemia but cannot characterize the shape of the glycemic course over the days, nor identify patients whose hyperglycemia emerges beyond the first day. Furthermore, prior intervention trials enrolled heterogeneous stroke populations without stratification by reperfusion status, potentially diluting any treatment effect among those most likely to benefit. Recent evidence suggests that active management of post-stroke hyperglycemia may improve outcomes [14], and that novel glucose-lowering agents such as glucagon-like peptide-1 receptor agonists show promise in patients with acute cerebrovascular events [15].

Therefore, we aimed to characterize distinct longitudinal glycemic trajectory patterns during the first five days following EVT and to evaluate their association with 3-month functional outcomes with the goal of informing the design of future targeted intervention trials in this high-risk population.

Methods

Study design and population

PRESTO (A Retrospective Study for the Prediction of Clinical Outcomes in Acute Ischemic Stroke Patients with Thrombectomy) study is a multicenter cohort study with a hybrid prospective-retrospective design, nested within the RODIN (Rescue on Reperfusion Damage in Cerebral Infarction by Nelonemdaz, NCT05041010) trial, supplemented by retrospective collection of longitudinal blood glucose data. The RODIN trial was a multicenter double-blind placebo-controlled randomized phase 3 clinical trial evaluating the efficacy and safety of nelonemdaz in patients with AIS undergoing EVT conducted across 24 tertiary stroke centers in South Korea from December 2021 to June 2023 [16]. Detailed eligibility criteria and study procedures have been described previously [16]. Briefly, patients were enrolled when they presented with AIS caused by large-vessel occlusion and underwent EVT within 12 hours from the last known well time. Additional inclusion criteria included age 19 years or older, initial National Institutes of Health Stroke Scale (NIHSS) score of 8 or greater, premorbid modified Rankin Scale (mRS) score of 0 or 1, and baseline Alberta Stroke Program Early Computed Tomography Scores (ASPECTS) of 4 or greater.

For the PRESTO study, the prospectively collected RODIN trial dataset-comprising baseline characteristics, neuroimaging data, periprocedural parameters, and 3-month functional outcomes-served as the primary data source. Additional longitudinal data, including serial blood glucose measurement, hemoglobin A1c (HbA1c) and pharmacological treatment records, were retrospectively retrieved from medical records at each participating center. From the 457 patients in the RODIN full analysis set, those with at least five blood glucose measurements during the first five days following EVT and with an available 3-month mRS score were included in the present analysis.

The PRESTO study was approved by independent Institutional Review Board (IRB) at each participating center (e.g., 2025-0694 of IRB of Asan Medical Center, which is the coordinating center of the RODIN and PRESTO study). Given the retrospective nature of the additional data collection, the requirement for informed consent was waived by each institutional review board. This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [17].

Data collection

Baseline characteristics, neuroimaging data, and periprocedural parameters were obtained from the prospectively collected RODIN trial database. ASPECTS and the final reperfusion status (extended thrombolysis in cerebral infarction [eTICI] grade) were analyzed by the imaging core laboratory of the RODIN trial [16]. Primary and secondary outcome variables, including 3-month mRS scores, were adopted from the RODIN trial database which was assessed by independent investigators blinded to all clinical and treatment information.

Longitudinal blood glucose data were retrospectively retrieved from medical records. Point-of-care testing (POCT) values served as the primary data source. When POCT values were unavailable, laboratory chemistry-based glucose values were substituted. To minimize chronologic discrepancies attributable to specimen processing, laboratory-based glucose values were assigned a timestamp 30 minutes earlier than the recorded time. Blood glucose measurements obtained from the time of EVT procedure initiation through the subsequent five days were included in the analysis.

Variable definition

Baseline variables were defined according to the RODIN trial protocol and included age, sex, medical history, stroke severity, and imaging markers.

Normoglycemia was defined as having 80% or more of blood glucose measurements within the 80–140 mg/dL range during the first 5 days after EVT according to previous studies [7,9,10], with the upper threshold of 140 mg/dL corresponding to the American Diabetes Association’s definition of hyperglycemia for inpatients [18]. Regardless of this criterion, patients with any single blood glucose measurement of 200 mg/dL or more were not classified as normoglycemic.

Symptomatic HT was defined as brain imaging–confirmed intracranial hemorrhage accompanied by neurological deterioration, defined as an increase of 4 or more points in the NIHSS score.

Group-based trajectory model analysis

Patients meeting the normoglycemia definition were classified as the normoglycemic group (Group 1) before modeling and served as the reference group in all outcome comparisons. Among patients not meeting normoglycemic criteria, the group-based trajectory modeling (GBTM) was applied to identify distinct longitudinal glycemic subgroups [19]. This two-step classification approach–first identifying normoglycemic patients based on established clinical thresholds, then applying data-driven modeling to the remaining patients–was adopted to preserve the clinical interpretability of the normoglycemic reference group while allowing objective characterization of distinct hyperglycemic patterns. We implemented the GBTM framework using the hlme function from the lcmm package in R. Longitudinal blood glucose data over a five-day post-EVT period were modeled as a mixture of multiple latent trajectories under a normal distribution assumption, incorporating natural cubic splines of time to capture non-linear trends without individual random effects. Single-group models were evaluated initially to determine the optimal functional form of time based on the lowest Bayesian Information Criterion (BIC) value. The number of trajectory groups was then selected on the basis of the lowest BIC; the two-group solution had the lowest BIC and was selected as the final model.

Statistical analysis

Baseline characteristics across glycemic trajectory groups were compared using chi-square or Fisher’s exact test for categorical variables and one-way analysis of variance or Kruskal-Wallis test for continuous variables, as appropriate.

For the primary outcome, the distribution of 3-month mRS scores across groups was compared using the Cochran-Mantel-Haenszel shift test, with treatment effect estimated as a common odds ratio (cOR) with 95% confidence interval (CI) using an ordinal logistic regression model. For the secondary outcomes, multivariable logistic regression was used to evaluate the association between glycemic trajectory group and favorable functional outcome (mRS 0–2) at 3 months, excellent functional outcome (mRS 0–1) at 3 months, poor functional outcome (mRS 4–6) at 3 months, and symptomatic HT. For cerebral infarct volume at week 12, between-group differences were assessed using ranked analysis of covariance (ANCOVA). Firth’s penalized logistic regression was applied for symptomatic HT due to the small number of events. As an exploratory outcome, any HT on day 5 post-procedure was classified according to the European Cooperative Acute Stroke Study (ECASS) classification [20].

Three adjustment models were applied to all outcome analyses. Model 1 was adjusted for age, sex, initial NIHSS score, and ASPECTS. Model 2 additionally included the non-glycemic baseline variables that differed across groups at P<0.10 (hypertension, diabetes mellitus, and intravenous thrombolysis). Model 3 further added HbA1c to the covariates in Model 2 to account for chronic glycemic burden. For the primary outcome, an additional model (Model 4) further added admission glucose to the covariates in Model 3. Admission glucose was also modeled without the glycemic trajectory classification, in a parallel series of unadjusted and adjusted models. For the ranked ANCOVA of week 12 infarct volume, cerebral infarct volume at day 1 was additionally included as a covariate in all models.

The normoglycemic group (Group 1) served as the reference group in all primary comparisons. An additional analysis using Group 2 as the reference was performed to assess outcome differences between two hyperglycemic trajectory groups. To assess whether the requirement of at least five glucose measurements introduced selection bias, a sensitivity analysis was performed in an expanded cohort of patients with at least three measurements during the first five days after EVT, applying the same two-step classification procedure and the same adjustment models.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P-value of less than 0.05 was considered statistically significant.

Results

Study population and trajectory classification

Of 457 patients enrolled in the full analysis set of the RODIN trial, 138 were excluded due to fewer than five blood glucose measurements within the first 5 days, and 319 met the inclusion criteria. Of these, 97 patients (30.4%) were classified as the normoglycemic group (Group 1). The remaining 222 patients were categorized by GBTM into two distinct trajectory subgroups: Group 2 (persistently elevated stable hyperglycemic trajectory, n=159), and Group 3 (variable hyperglycemic trajectory, n=63) as shown in Figure 1.

Figure 1.

Glycemic trajectory patterns within 5 days after endovascular thrombectomy. Lines represent model-estimated mean blood glucose for each group, and shaded areas indicate 95% CIs. CI, confidence interval.

Baseline characteristics

Baseline characteristics across the three trajectory groups are presented in Table 1. The overall mean age was 73.6±12.2 years and 53.6% were male. Age differed significantly across groups (Group 1: 71.0±12.8, Group 2: 75.1±11.6, Group 3: 73.9±12.3 years; P=0.034), while sex distribution was comparable. Diabetes mellitus was markedly more prevalent in Group 2 and 3 (Group 1: 8.5%, Group 2: 50.6%, Group 3: 55.6%; P<0.001), as was hypertension (59.6%, 77.2%, and 77.8%, respectively; P=0.006). Similarly, baseline HbA1c levels were also significantly higher in non-normoglycemic groups (Group 1: 5.6%±0.4%, Group 2: 6.4%±1.1%, Group 3: 6.6%±1.9%; P<0.001). However, baseline stroke severity as measured by NIHSS score (median 13, interquartile range [IQR] 10–17; P=0.419) and ASPECTS (median 8, IQR 6–9; P=0.471), as well as other clinical parameters such as occlusion site, Trial of Org 10172 in Acute Stroke Treatment classification, intravenous thrombolysis rate, eTICI grade distribution, and procedural time metrics, showed no significant differences across the groups. Admission glucose differed across trajectory groups (Group 1: 123.1±25.2 mg/dL, Group 2: 148.7±43.8, Group 3: 152.2±58.7; P<0.001), as did the use of glucose-lowering treatment during the observation period (4 patients [4.1%], 48 [30.2%], and 28 [44.4%], respectively; P<0.001), including both oral agents (4.1%, 14.5%, and 19.0%; P=0.009) and injectable agents (0.0%, 20.8%, and 39.7%; P<0.001). Allocation to nelonemdaz was balanced across groups (54.6%, 52.8%, and 49.2%; P=0.796).

Characteristics of patients at baseline

Clinical outcomes

Outcome data are presented in Table 2 and Figure 2. On the primary outcome, Group 1 showed a significantly more favorable 3-month mRS distribution compared with Group 2 and 3 (median mRS 2 [IQR 1–3] vs. 4 [1–5] and 4 [1–4], respectively; P<0.001 by shift analysis). After adjustment for age, sex, baseline NIHSS score, ASPECTS, hypertension, diabetes mellitus, intravenous thrombolysis, and HbA1c (Model 3), both Group 2 and 3 showed significantly lower odds of a favorable mRS shift compared with Group 1 (adjusted cOR 0.35, 95% CI 0.20–0.61, P<0.001 for Group 2; adjusted cOR 0.49, 95% CI 0.25–0.94, P=0.032 for Group 3) (Table 3).

Primary and secondary outcomes by glycemic trajectory group

Figure 2.

Distribution of mRS at 3 months. mRS, modified Rankin Scale.

Multivariable analysis of clinical outcomes comparing glycemic trajectory groups with the normoglycemic group

When admission glucose was added to Model 3 (Model 4), Group 2 remained associated with a less favorable mRS shift (adjusted cOR 0.36, 95% CI 0.20–0.63; P<0.001), whereas the estimate for Group 3 was essentially unchanged in magnitude but no longer reached statistical significance (adjusted cOR 0.52, 95% CI 0.26–1.03; P=0.061) (Table 3). Admission glucose was not independently associated with outcome (cOR per 10 mg/dL 0.94, 95% CI 0.87–1.01; P=0.073) (Supplementary Table 1). Modeled without the trajectory classification, admission glucose was associated with a less favorable mRS shift under Model 1 (cOR per 10 mg/dL 0.90, 95% CI 0.86–0.95; P<0.001) but not under Model 3 (cOR 0.94, 95% CI 0.88–1.01; P=0.079) (Supplementary Table 2).

For secondary outcomes, Group 1 achieved the highest rate of functional independence (mRS 0–2, 63.9%), compared with Group 2 (36.5%) and Group 3 (38.1%; P<0.001). After adjustment (Model 3), the odds of functional independence were significantly lower in Group 2 (adjusted OR [aOR] 0.26, 95% CI 0.13–0.53) and Group 3 (aOR 0.31, 95% CI 0.13–0.73) compared with Group 1. Poor functional outcome (mRS 4–6) remained more than four times higher in Group 2 (aOR 4.15, 95% CI 1.94–8.88) and Group 3 (aOR 4.25, 95% CI 1.72–10.53) compared with Group 1. When Group 2 was used as the reference, no significant differences in outcomes were observed between the two hyperglycemic trajectory groups (Table 4). Symptomatic HT was observed only in the hyperglycemic groups (3.3% of both Group 2 and 3 vs. 0.0% in Group 1); however, this did not reach statistical significance. Due to the small number of events (n=7), Firth’s penalized logistic regression was applied and no significant association was identified. Regarding any HT regardless of symptoms, the distribution of HT subtypes based on ECASS classification on day 5 was significantly different across groups (P=0.009), driven primarily by a higher rate of parenchymal hematoma type 1 (PH1) in Group 2 (13.7%) compared with Group 1 (1.0%) and Group 3 (3.3%) (Supplementary Table 3). Cerebral infarct volume was significantly larger in Group 2 than in Group 1 at day 1 (median 55 [IQR 18–171] vs. 23 [8–69]; P=0.002) and day 5 (69 [25–195] vs. 26 [12–82]; P<0.001), while Group 3 showed intermediate values (day 1: 23 [10–95]; day 5: 37 [17–145]). At week 12, infarct volume was available for 181 patients and did not differ significantly across groups, either unadjusted or in any adjusted model (Table 2).

Multivariable analysis comparing the normoglycemic and variable hyperglycemic groups with the persistently elevated stable group

In the sensitivity analysis including 361 patients with at least three glucose measurements during the first 5 days, 115 patients (31.9%) met the normoglycemic criteria, and GBTM again identified two hyperglycemic trajectories with shapes closely resembling those of the primary analysis (Group 2, n=178; Group 3, n=68) (Supplementary Figure 1). Baseline characteristics differed across groups in the same pattern, although the between-group difference in age was attenuated (P=0.081) (Supplementary Table 4). The 3-month mRS distribution favored Group 1 (median 2 [IQR 1–3] vs. 4 [1–5] in Group 2 and 3 [1–4] in Group 3; P<0.001), and functional independence was achieved in 64.3%, 38.2%, and 41.2%, respectively (P<0.001); other outcomes followed the same pattern as in the primary analysis (Supplementary Table 5). Under Model 3, both hyperglycemic groups had lower odds of functional independence (aOR 0.27, 95% CI 0.14–0.52 for Group 2; aOR 0.36, 95% CI 0.16–0.79 for Group 3) and higher odds of poor functional outcome (aOR 4.18, 95% CI 2.08–8.42 for Group 2; aOR 3.57, 95% CI 1.53–8.31 for Group 3) compared with Group 1 (Table 5). For the 3-month mRS shift, the estimate for Group 2 was essentially unchanged (adjusted cOR 0.38, 95% CI 0.23–0.63; P<0.001), whereas the estimate for Group 3 was of similar magnitude to the primary analysis but no longer reached statistical significance (adjusted cOR 0.55, 95% CI 0.30–1.02; P=0.058). The two hyperglycemic groups did not differ significantly from each other when Group 2 served as the reference (Supplementary Table 6).

Sensitivity analysis of clinical outcomes comparing glycemic trajectory groups in the expanded cohort of patients with at least three glucose measurements

Discussion

In this nested cohort study within the RODIN trial, both persistently elevated stable hyperglycemia (Group 2) and variable hyperglycemia (Group 3) were independently associated with significantly worse 3-month functional outcomes compared with sustained normoglycemia (Group 1), even after adjustment for HbA1c. For the ordinal mRS shift, further adjustment for admission glucose left both estimates essentially unchanged, although that for Group 3 no longer reached statistical significance.

The detrimental impact of admission hyperglycemia on post-EVT outcomes is well established. Large registry studies, pooled meta-analyses, and post hoc analyses of major EVT trials have consistently demonstrated that admission hyperglycemia is independently associated with worse functional outcomes, higher mortality, and increased symptomatic intracranial hemorrhage [9,11,21]. However, these investigations relied on glucose measured at admission alone, which does not capture the extent to which hyperglycemia persists over the days following reperfusion. In the present cohort, admission glucose was associated with the 3-month mRS shift when modeled without the glycemic groups, but this association was attenuated after adjustment for HbA1c, and admission glucose itself carried no independent association when added to the fully adjusted model (Model 4). The 5-day post-EVT glycemic pattern therefore appears to convey prognostic information that neither a single admission value nor chronic glycemic burden accounts for.

The independence from HbA1c is of particular clinical importance, since HbA1c reflects chronic glycemic burden accumulated over the preceding two to three months. This implies that the acute post-procedural glycemic environment carries independent prognostic significance beyond what is attributable to pre-existing metabolic disease. This finding aligns with the concept of stress hyperglycemia, in which acute neurological insult and procedural stress trigger catecholamine and glucocorticoid release, driving glucose elevation through increased hepatic glucose production and peripheral insulin resistance [22]. Prior work has shown that hyperglycemic non-diabetic stroke patients carry a particularly poor prognosis, with higher rates of HT and cerebral edema than both normoglycemic non-diabetic and diabetic patients, suggesting that the acute stress response is the principal driver of harm [23].

The exploratory analysis of HT on day 5 imaging provides additional mechanistic insight. The higher rate of PH1 in Group 2 compared with Group 1 suggests that sustained post-procedural hyperglycemia may contribute to worse functional outcomes in part through a greater burden of hemorrhagic parenchymal injury. Hyperglycemia is known to impair blood-brain barrier integrity and exacerbate reperfusion injury in the ischemic penumbra, mechanisms that may facilitate HT following mechanical reperfusion [5]. Although symptomatic HT did not differ significantly across groups, likely reflecting the limited statistical power attributable to the small number of events (n=7), the imaging-based finding supports a biologically plausible mechanism linking sustained hyperglycemia to parenchymal injury in the post-EVT setting.

Groups 2 and 3 showed broadly comparable functional outcomes despite exhibiting distinct glycemic trajectory shapes, though this finding should be interpreted with caution given the small sample size of Group 3 (n=63). This observation directs attention to the overall magnitude of glucose elevation early in the post-EVT course. However, serial measurement remained necessary to capture the glycemic burden of patients whose glucose was not elevated throughout, as in Group 3, whose glucose passed through the near-normal range during the first post-procedural day before peaking on day 3.

Patients in Group 3 exhibited lower glucose levels during the first post-procedural day followed by a delayed rise peaking on day 3, whereas Group 2 patients sustained persistently elevated glucose throughout the monitoring period. Prior studies reported that glycemic variability itself carries independent prognostic significance in AIS [24,25]. Although group-level trajectories do not quantify within-patient variability, the marked fluctuation of the Group 3 trajectory suggests that temporal instability of glucose, in addition to its magnitude, may be relevant to outcome. Other studies that analyzed glucose measurements at admission and at 24 hours have reported that persistent hyperglycemia sustained through the first 24 hours was independently associated with poor functional outcomes [12,13]. The comparability of outcomes between Group 2 and Group 3 may reflect offsetting contributions of early glycemic level and a delayed peak, but the present data cannot distinguish these components and this interpretation remains speculative.

Despite the evidence linking hyperglycemia to adverse stroke outcomes, clinical trials of intensive glucose control have failed to demonstrate functional benefit in the general AIS population [6,7]. The present trajectory-based findings offer a potential framework for interpreting these results. Patients who maintain normoglycemia throughout the post-EVT period (Group 1) demonstrated markedly better outcomes. Conversely, patients who exhibit persistent or variable hyperglycemia (Groups 2 and 3) constitute a distinct high-risk subgroup that may be a more informative population in which to test targeted glycemic management. The uniform application of glucose targets across all patients without stratification by glycemic trajectory or reperfusion status may have diluted a treatment effect among the subgroup most likely to benefit. Moreover, with the recent advances in glucose management strategies offering improved glycemic control with reduced hypoglycemic risk [14,15,26], the therapeutic landscape for acute glycemic management has substantially evolved since the completion of earlier intensive glucose control trials. Future studies should investigate whether the prognostic impact of glycemic trajectories differs according to the use of glucose-lowering interventions to help identify the subgroup most amenable to targeted glycemic intervention.

Several limitations of this study warrant acknowledgment. First, as a nested cohort study within a randomized trial not designed for glycemic monitoring, the frequency and timing of blood glucose measurements varied across patients and centers, which may have introduced measurement heterogeneity. Second, patients with fewer blood glucose measurements were excluded from analysis, potentially introducing selection bias. These patients had comparable baseline stroke severity, ASPECTS, and HbA1c, but less diabetes, more complete reperfusion, smaller infarct volumes, and better 3-month outcomes, consistent with an uncomplicated course requiring less intensive glucose monitoring (Supplementary Table 7). Our findings may therefore be less generalizable to patients with fewer metabolic comorbidities and complete reperfusion. Because excluded patients would predominantly be normoglycemic, however, this selection would attenuate rather than exaggerate the observed contrast between trajectory groups. A sensitivity analysis in an expanded cohort requiring at least three measurements reproduced the same group structure and yielded consistent estimates, although the association of the variable hyperglycemic group with the ordinal mRS shift did not reach statistical significance in that cohort. Third, we did not perform a formal analysis of the effect of glucose-lowering treatment, which was administered more frequently in the hyperglycemic groups. Because such treatment is initiated in response to the glucose values that define the trajectories, it is both a consequence of the exposure and a determinant of its later course, and conventional adjustment for it would yield biased estimates. The timing and dose data required for a valid analysis were not captured in sufficient detail. If treatment is beneficial or has no effect on functional outcome, as randomized data indicate for intensive glycemic control after AIS [7], residual confounding by treatment would bias toward the null. Whether glucose-lowering treatment itself influences outcome in this setting requires further study. In addition, because few patients in Group 1 had diabetes, residual confounding by unmeasured diabetes duration, severity, and pre-stroke treatment cannot be excluded despite adjustment for diabetes and HbA1c. Fourth, a larger infarct elicits a greater systemic stress response and may thereby raise post-procedural glucose. Group 2 had a day 1 infarct volume more than twice that of the normoglycemic group, and we cannot exclude the possibility that lesion burden contributed to the glycemic pattern observed in this group. However, persistent post-stroke hyperglycemia has itself been shown to be an independent determinant of infarct expansion [27], so that the direction of this relationship cannot be resolved from cross-sectional volumes. Group 3 is informative in this regard: its day 1 volume was comparable to that of the normoglycemic group, yet its odds of poor outcome were similarly elevated, suggesting that the glycemic pattern may carry prognostic information beyond lesion burden. Our findings should nonetheless be interpreted as identifying a pattern associated with poor outcome rather than as establishing a causal contribution of hyperglycemia, and residual confounding by unmeasured severity cannot be excluded. Fifth, the stress hyperglycemia ratio (SHR), which accounts for the degree of acute glucose elevation relative to chronic glycemic status [28], was not incorporated into the present analysis; future studies examining whether trajectory-based stratification provides discriminative value beyond SHR alone would be informative. Sixth, cerebral infarct volume at week 12 was available for 181 of 319 patients, and imaging was less likely in those who died or remained severely disabled; this comparison should therefore be regarded as exploratory. Finally, the study population was derived from a Korean multicenter trial, and generalizability to other ethnic populations should be evaluated in future studies.

Conclusions

This study demonstrates that the 5-day post-procedural hyperglycemic pattern following EVT is independently associated with 3-month functional outcomes beyond chronic glycemic burden. Both persistently elevated and variable hyperglycemia were associated with worse prognosis compared with sustained normoglycemia, without a significant difference between the two groups, and estimates were largely unchanged after further adjustment for admission glucose. Serial measurement was nonetheless required to identify patients whose hyperglycemia peaked after the first post-procedural day. These findings indicate that the acute glycemic environment following mechanical reperfusion carries prognostic information. As an observational analysis, this study cannot determine whether modifying post-procedural glycemia would alter outcome; our results may, however, inform the design of future trials incorporating systematic glycemic monitoring and targeted glycemic intervention during this vulnerable post-procedural window.

Supplementary materials

Supplementary materials related to this article can be found online at https://doi.org/10.5853/jos.2026.03867.

Supplementary Table 1.

Full multivariable model of the 3-month mRS score including glycemic trajectory group and admission glucose

jos-2026-03867-Supplementary-Table-1.pdf
Supplementary Table 2.

Association of admission glucose with 3-month mRS score

jos-2026-03867-Supplementary-Table-2.pdf
Supplementary Table 3.

Distribution of hemorrhagic transformation subtypes on day 5 by glycemic trajectory group

jos-2026-03867-Supplementary-Table-3.pdf
Supplementary Table 4.

Baseline characteristics of patients in the expanded cohort with at least three glucose measurements

jos-2026-03867-Supplementary-Table-4.pdf
Supplementary Table 5.

Primary and secondary outcomes by glycemic trajectory group in the expanded cohort with at least three glucose measurements

jos-2026-03867-Supplementary-Table-5.pdf
Supplementary Table 6.

Sensitivity analysis of clinical outcomes comparing the normoglycemic and variable hyperglycemic groups with the persistently elevated stable group in the expanded cohort

jos-2026-03867-Supplementary-Table-6.pdf
Supplementary Table 7.

Comparison of baseline characteristics and outcomes between included and excluded patients

jos-2026-03867-Supplementary-Table-7.pdf
Supplementary Figure 1.

Glycemic trajectory patterns within 5 days after endovascular thrombectomy in the expanded cohort with at least three glucose measurements.

jos-2026-03867-Supplementary-Figure-1.pdf

Notes

Funding statement

This research was supported by a grant of Patient-Centered Clinical Research Coordinating Center (PACEN) funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2021-KH120281).

Conflicts of interest

The authors have no financial conflicts of interest.

Author contribution

Conceptualization: Kwanju Song, Sun U. Kwon. Study design: Kwanju Song, Sun U. Kwon. Methodology: Kwanju Song, Ji Sung Lee, Sun U. Kwon. Data collection: Hyun Goo Kang, Seong Hwan Ahn, Tae-Jin Song, Dong-Ick Shin, Hee-Joon Bae, Chang Hun Kim, Sung Hyuk Heo, Jin Soo Lee, Jae-Kwan Cha, Yeong Bae Lee, Eung Gyu Kim, Man Seok Park, Hee-Kwon Park, Jinkwon Kim, Sungwook Yu, Heejung Mo, Sung Il Sohn, Jee Hyun Kwon, Jae Guk Kim, Young Seo Kim, Jay Chol Choi, Yang-Ha Hwang, Keun Hwa Jung, Soo-Kyoung Kim, Woo Keun Seo, Jung Hwa Seo, Joonsang Yoo, Jun Young Chang, Mooseok Park, Wookjin Yang, Bum Joon Kim, Sun U. Kwon. Investigation: Kwanju Song, Ji Sung Lee, Sun U. Kwon. Statistical analysis: Ji Sung Lee. Writing—original draft: Kwanju Song, Sun U. Kwon. Writing—review & editing: all authors. Funding acquisition: Sun U. Kwon. Approval of final manuscript: all authors.

References

1. Prabhakaran S, Gonzalez NR, Zachrison KS, Adeoye O, Alexandrov AW, Ansari SA, et al. 2026 guideline for the early management of patients with acute ischemic stroke: a guideline from the American Heart Association/American Stroke Association. Stroke 2026;57:e316–e436.
2. Capes SE, Hunt D, Malmberg K, Pathak P, Gerstein HC. Stress hyperglycemia and prognosis of stroke in nondiabetic and diabetic patients: a systematic overview. Stroke 2001;32:2426–2432.
3. Ntaios G, Egli M, Faouzi M, Michel P. J-shaped association between serum glucose and functional outcome in acute ischemic stroke. Stroke 2010;41:2366–2370.
4. Southerland AM, Mayer SA, Chiota-McCollum NA, Bolte AC, Pauls Q, Pettigrew LC, et al. Glucose control and risk of symptomatic intracerebral hemorrhage following thrombolysis for acute ischemic stroke: a SHINE trial analysis. Neurology 2024;102e209323.
5. Aziz YN, Vagal V, Mehta TV, Siegler JE, Mistry AM, Yaghi S, et al. Higher blood glucose is associated with the severity of hemorrhagic transformation after endovascular therapy for stroke. J Stroke Cerebrovasc Dis 2024;33:107823.
6. Gray CS, Hildreth AJ, Sandercock PA, O'Connell JE, Johnston DE, Cartlidge NE, et al. Glucose-potassium-insulin infusions in the management of post-stroke hyperglycaemia: the UK Glucose Insulin in Stroke Trial (GIST-UK). Lancet Neurol 2007;6:397–406.
7. Johnston KC, Bruno A, Pauls Q, Hall CE, Barrett KM, Barsan W, et al. Intensive vs standard treatment of hyperglycemia and functional outcome in patients with acute ischemic stroke: the SHINE randomized clinical trial. JAMA 2019;322:326–335.
8. Smith M, Reddy U, Robba C, Sharma D, Citerio G. Acute ischaemic stroke: challenges for the intensivist. Intensive Care Med 2019;45:1177–1189.
9. Kim JT, Jahan R, Saver JL, ; SWIFT Investigators. Impact of glucose on outcomes in patients treated with mechanical thrombectomy: a post hoc analysis of the solitaire flow restoration with the intention for thrombectomy study. Stroke 2016;47:120–127.
10. Rinkel LA, Nguyen TTM, Guglielmi V, Groot AE, Posthuma L, Roos YBWEM, et al. High admission glucose is associated with poor outcome after endovascular treatment for ischemic stroke. Stroke 2020;51:3215–3223.
11. Yazawa Y, Ohira T, Itabashi R, Uchida K, Sakai N, Yamagami H, et al. Association of admission hyperglycemia with clinical outcomes in Japanese patients with acute large vessel occlusion stroke: a post hoc analysis of the Recovery by Endovascular Salvage for Cerebral Ultra-Acute Embolism Japan Registry 2. Cerebrovasc Dis 2021;50:12–19.
12. Yong M, Kaste M. Dynamic of hyperglycemia as a predictor of stroke outcome in the ECASS-II trial. Stroke 2008;39:2749–2755.
13. Merlino G, Smeralda C, Sponza M, Gigli GL, Lorenzut S, Marini A, et al. Dynamic hyperglycemic patterns predict adverse outcomes in patients with acute ischemic stroke undergoing mechanical thrombectomy. J Clin Med 2020;9:1932.
14. Bladin CF, Wah Cheung N, Dewey HM, Churilov L, Middleton S, Thijs V, et al. Management of poststroke hyperglycemia: results of the TEXAIS randomized clinical trial. Stroke 2023;54:2962–2971.
15. Zhu H, Yang B, Lu L, Li Y, Sui R, Liu K, et al. Liraglutide in acute minor ischemic stroke or high-risk transient ischemic attack with type 2 diabetes: the LAMP randomized clinical trial. JAMA Intern Med 2026;186:46–54.
16. Lee JS, Kang HG, Ahn SH, Song TJ, Shin DI, Bae HJ, et al. Nelonemdaz and patients with acute ischemic stroke and mechanical reperfusion: the RODIN randomized clinical trial. JAMA Netw Open 2025;8e2456535.
17. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 2007;335:806–808.
18. American Diabetes Association Professional Practice Committee for Diabetes. 16. Diabetes care in the hospital: standards of care in diabetes-2026. Diabetes Care 2026;49(Supplement_1):S339–S355.
19. Nagin DS. Group-based trajectory modeling: an overview. Ann Nutr Metab 2014;65:205–210.
20. Fiorelli M, Bastianello S, von Kummer R, del Zoppo GJ, Larrue V, Lesaffre E, et al. Hemorrhagic transformation within 36 hours of a cerebral infarct: relationships with early clinical deterioration and 3-month outcome in the European Cooperative Acute Stroke Study I (ECASS I) cohort. Stroke 1999;30:2280–2284.
21. Chamorro Á, Brown S, Amaro S, Hill MD, Muir KW, Dippel DWJ, et al. Glucose modifies the effect of endovascular thrombectomy in patients with acute stroke: a pooled-data meta-analysis. Stroke 2019;50:690–696.
22. Song G, Liu X, Lu Z, Guan J, Chen X, Li Y, et al. Relationship between stress hyperglycaemic ratio (SHR) and critical illness: a systematic review. Cardiovasc Diabetol 2025;24:188.
23. Muscari A, Falcone R, Recinella G, Faccioli L, Forti P, Pastore Trossello M, et al. Prognostic significance of diabetes and stress hyperglycemia in acute stroke patients. Diabetol Metab Syndr 2022;14:126.
24. Kim YS, Kim C, Jung KH, Kwon HM, Heo SH, Kim BJ, et al. Range of glucose as a glycemic variability and 3-month outcome in diabetic patients with acute ischemic stroke. PLoS One 2017;12e0183894.
25. Kim TJ, Lee JS, Park SH, Ko SB. Short-term glycemic variability and hemorrhagic transformation after successful endovascular thrombectomy. Transl Stroke Res 2021;12:968–975.
26. Spanakis EK, Urrutia A, Galindo RJ, Vellanki P, Migdal AL, Davis G, et al. Continuous glucose monitoring-guided insulin administration in hospitalized patients with diabetes: a randomized clinical trial. Diabetes Care 2022;45:2369–2375.
27. Baird TA, Parsons MW, Phan T, Butcher KS, Desmond PM, Tress BM, et al. Persistent poststroke hyperglycemia is independently associated with infarct expansion and worse clinical outcome. Stroke 2003;34:2208–2214.
28. Roberts GW, Quinn SJ, Valentine N, Alhawassi T, O’Dea H, Stranks SN, et al. Relative hyperglycemia, a marker of critical illness: introducing the stress hyperglycemia ratio. J Clin Endocrinol Metab 2015;100:4490–4497.

Article information Continued

Figure 1.

Glycemic trajectory patterns within 5 days after endovascular thrombectomy. Lines represent model-estimated mean blood glucose for each group, and shaded areas indicate 95% CIs. CI, confidence interval.

Figure 2.

Distribution of mRS at 3 months. mRS, modified Rankin Scale.

Table 1.

Characteristics of patients at baseline

Characteristics Total (n=319) Normoglycemic (Group 1, n=97) Persistently elevated stable (Group 2, n=159) Variable hyperglycemic (Group 3, n=63) P
Age (yr) 73.6±12.2 71.0±12.8 75.1±11.6 73.9±12.3 0.034
Male sex 171 (53.6) 52 (53.6) 88 (55.3) 31 (49.2) 0.710
Medical history*
 Hypertension 227 (72.1) 56 (59.6) 122 (77.2) 49 (77.8) 0.006
 Diabetes 123 (39.0) 8 (8.5) 80 (50.6) 35 (55.6) <0.001
 Dyslipidemia 128 (40.6) 34 (36.2) 63 (39.9) 31 (49.2) 0.255
 Coronary artery disease 53 (16.8) 16 (17.0) 25 (15.8) 12 (19.0) 0.844
 Atrial fibrillation 168 (53.3) 50 (53.2) 87 (55.1) 31 (49.2) 0.733
 Current smoking 38 (12.1) 12 (12.8) 17 (10.8) 9 (14.3) 0.744
 Current alcohol consumption 69 (21.9) 23 (24.5) 37 (23.4) 9 (14.3) 0.258
 Family history of ischemic stroke 48 (15.2) 17 (18.1) 20 (12.7) 11 (17.5) 0.439
mRS at one day prior to the ischemic stroke onset 0.869
 0 244 (76.5) 76 (78.4) 120 (75.5) 48 (76.2)
 1 75 (23.5) 21 (21.6) 39 (24.5) 15 (23.8)
Initial NIHSS score 13 (10–17) 13 (10–17) 13 (10–18) 13 (9–17) 0.419
ASPECTS 8 (6–9) 8 (6–9) 8 (6–9) 8 (5–9) 0.471
Occlusion site
 Intracranial ICA 82 (25.7) 21 (21.6) 39 (24.5) 22 (34.9) 0.153
 MCA M1 190 (59.6) 60 (61.9) 89 (56.0) 41 (65.1) 0.395
 M1-equivalent M2 78 (24.5) 25 (25.8) 42 (26.4) 11 (17.5) 0.352
TOAST classification 0.570
 CE 194 (60.8) 62 (63.9) 97 (61.0) 35 (55.6)
 Non-CE 125 (39.2) 35 (36.1) 62 (39.0) 28 (44.4)
Intravenous thrombolysis 149 (46.7) 46 (47.4) 81 (50.9) 22 (34.9) 0.096
eTICI grade† 0.451
 0 14 (4.5) 4 (4.1) 7 (4.5) 3 (4.9)
 1 7 (2.2) 0 (0.0) 5 (3.2) 2 (3.3)
 2a 25 (8.0) 4 (4.1) 16 (10.3) 5 (8.2)
 2b 129 (41.1) 45 (46.4) 60 (38.5) 24 (39.3)
 2c 70 (22.3) 18 (18.6) 37 (23.7) 15 (24.6)
 3 69 (22.0) 26 (26.8) 31 (19.9) 12 (19.7)
Time from onset to ED arrival (min) 96 (49–208) 83 (42–207) 94 (54–172) 152 (52–263) 0.153
Time from arrival to thrombectomy (min) 108 (81–141) 108 (83–153) 110 (86–138) 98 (76–127) 0.237
Time from arrival to intravenous alteplase (min) 42 (29–57) 39 (30–56) 45 (31–59) 38 (19–57) 0.276
HbA1c (%) 6.2±1.2 5.6±0.4 6.4±1.1 6.6±1.9 <0.001
Admission glucose (mg/dL) 141.4±44.2 123.1±25.2 148.7±43.8 152.2±58.7 <0.001
Glucose-lowering treatment‡ 80 (25.1) 4 (4.1) 48 (30.2) 28 (44.4) <0.001
Oral agent 39 (12.2) 4 (4.1) 23 (14.5) 12 (19.0) 0.009
Injectable agent 58 (18.2) 0 (0.0) 33 (20.8) 25 (39.7) <0.001
Nelonemdaz 168 (52.7) 53 (54.6) 84 (52.8) 31 (49.2) 0.796

Values are presented as mean±standard deviation, n (%), or median (interquartile range).

mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early Computed Tomography Score; ICA, internal carotid artery; MCA, middle cerebral artery; TOAST, Trial of Org 10172 in Acute Stroke Treatment; CE, cardioembolism; eTICI, extended thrombolysis in cerebral infarction; ED, emergency department; HbA1c, hemoglobin A1c.

*

Medical history was unavailable in 4 patients; percentages for all medical history variables are based on patients with available data;

†

eTICI grade was unavailable in 5 patients; percentages were calculated among the patients with available data;

‡

Categories are not mutually exclusive; 17 patients received both an oral agent and injectable agent.

Table 2.

Primary and secondary outcomes by glycemic trajectory group

Outcomes Normoglycemic (Group 1, n=97) Persistently elevated stable (Group 2, n=159) Variable hyperglycemic (Group 3, n=63) P*
3-month mRS (shift analysis) 2 (1–3) 4 (1–5) 4 (1–4) <0.001
Functional independence (mRS 0–2) 62 (63.9) 58 (36.5) 24 (38.1) <0.001
Excellent outcome (mRS 0–1) 44 (45.4) 42 (26.4) 18 (28.6) 0.005
Poor outcome (mRS 4–6) 23 (23.7) 82 (51.6) 32 (50.8) <0.001
Symptomatic hemorrhagic transformation on day 5† 0 (0.0) 5 (3.3) 2 (3.3) 0.162
Cerebral infarct volume (mL)
 Day 1 23 (8–69) 55 (18–171) 23 (10–95) 0.002
 Day 5 26 (12–82) 69 (25–195) 37 (17–145) <0.001
 Week 12 9 (3–36) 20 (3–55) 10 (4–63) 0.173
 Model 1‡ 0.534
 Model 2§ 0.743
 Model 3ǁ 0.781

Values are presented as median (interquartile range) or n (%). Cerebral infarct volume at week 12 was available for 181 patients; the Model 3 analysis included 116 owing to missing covariate data.

mRS, modified Rankin Scale; ANCOVA, analysis of covariance; NIHSS, National Institutes of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early Computed Tomography Score; HTN, hypertension; DM, diabetes mellitus; IVT, intravenous thrombolysis; HbA1c, hemoglobin A1c.

*

P-value by Cochran-Mantel-Haenszel shift test, chi-square test, Fisher’s exact test, and Kruskal-Wallis test;

†

Data were missing in 9 patients; percentages were calculated among the patients with available data;

‡

Model 1: ranked ANCOVA adjusted for age, sex, initial NIHSS score, ASPECTS, and cerebral infarct volume at day 1;

§

Model 2: adjusted for age, sex, initial NIHSS score, ASPECTS, cerebral infarct volume at day 1, HTN, DM, and IVT;

ǁ

Model 3: adjusted for age, sex, initial NIHSS score, ASPECTS, cerebral infarct volume at day 1, HTN, DM, IVT, and HbA1c.

Table 3.

Multivariable analysis of clinical outcomes comparing glycemic trajectory groups with the normoglycemic group

Outcomes Normoglycemic (Group 1, n=97)
Persistently elevated stable (Group 2, n=159)
Variable hyperglycemic (Group 3, n=63)
Reference Common OR or OR (95% CI) P* Common OR or OR (95% CI) P*
Unadjusted
 3-month mRS (shift analysis)
  Unadjusted common OR (95% CI)† 1 (Ref) 0.33 (0.21–0.53) <0.001 0.42 (0.24–0.74) 0.002
 Functional independence (mRS 0–2)
  Unadjusted OR (95% CI) 1 (Ref) 0.32 (0.19–0.55) <0.001 0.35 (0.18–0.67) 0.002
 Excellent outcome (mRS 0–1)
  Unadjusted OR (95% CI) 1 (Ref) 0.43 (0.25–0.74) 0.002 0.48 (0.24–0.95) 0.035
 Poor outcome (mRS 4–6)
  Unadjusted OR (95% CI) 1 (Ref) 3.43 (1.95–6.01) <0.001 3.32 (1.68–6.56) <0.001
 Symptomatic hemorrhagic transformation at day 5‡
  Unadjusted OR (95% CI)‡ 1 (Ref) 7.15 (0.39–132.64) 0.187 8.11 (0.38–174.81) 0.182
Model 1§
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.32 (0.20–0.52) <0.001 0.39 (0.21–0.72) 0.003
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.29 (0.16–0.53) <0.001 0.27 (0.12–0.58) <0.001
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.43 (0.23–0.79) 0.007 0.46 (0.21–1.01) 0.054
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.02 (2.08–7.81) <0.001 4.98 (2.20–11.32) <0.001
Model 2ǁ
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.32 (0.19–0.55) <0.001 0.45 (0.24–0.87) 0.017
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.27 (0.14–0.53) <0.001 0.31 (0.13–0.71) 0.006
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.37 (0.19–0.74) 0.005 0.47 (0.20–1.09) 0.080
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.41 (2.11–9.20) <0.001 4.56 (1.88–11.06) <0.001
Model 3
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.35 (0.20–0.61) <0.001 0.49 (0.25–0.94) 0.032
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.26 (0.13–0.53) <0.001 0.31 (0.13–0.73) 0.008
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.40 (0.20–0.81) 0.011 0.48 (0.20–1.14) 0.096
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.15 (1.94–8.88) <0.001 4.25 (1.72–10.53) 0.002
Model 4**
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.36 (0.20–0.63) <0.001 0.52 (0.26–1.03) 0.061

OR, odds ratio; CI, confidence interval; mRS, modified Rankin Scale; Ref, reference; NIHSS, National Institutes of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early Computed Tomography Score; HTN, hypertension; DM, diabetes mellitus; IVT, intravenous thrombolysis; HbA1c, hemoglobin A1c.

*

P-value by ordinal logistic regression and binary logistic regression;

†

Common OR: ordinal shift across the range of mRS scores toward a better outcome;

‡

Due to the very small number of symptomatic hemorrhagic transformation events (n=7), Firth’s penalized logistic regression model was applied;

§

Model 1: adjusted for age, sex, initial NIHSS, and ASPECTS;

ǁ

Model 2: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, and IVT;

Model 3: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, IVT, and HbA1c;

**

Model 4: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, IVT, HbA1c, and admission glucose.

Table 4.

Multivariable analysis comparing the normoglycemic and variable hyperglycemic groups with the persistently elevated stable group

Outcomes Normoglycemic (Group 1, n=97)
Persistently elevated stable (Group 2, n=159)
Variable hyperglycemic (Group 3, n=63)
Common OR or OR (95% CI) P* Reference Common OR or OR (95% CI) P*
Unadjusted
 3-month mRS (shift analysis)
  Unadjusted common OR (95% CI)† 3.00 (1.90–4.74) <0.001 1 (Ref) 1.26 (0.75–2.10) 0.382
 Functional independence (mRS 0–2)
  Unadjusted OR (95% CI) 3.08 (1.82–5.22) <0.001 1 (Ref) 1.07 (0.59–1.96) 0.822
 Excellent outcome (mRS 0–1)
  Unadjusted OR (95% CI) 2.31 (1.36–3.94) 0.002 1 (Ref) 1.11 (0.58–2.14) 0.744
 Poor outcome (mRS 4–6)
  Unadjusted OR (95% CI) 0.29 (0.17–0.51) <0.001 1 (Ref) 0.97 (0.54–1.74) 0.917
 Symptomatic hemorrhagic transformation at day 5‡
  Unadjusted OR (95% CI)‡ 0.14 (0.01–2.60) 0.187 1 (Ref) 1.13 (0.24–5.27) 0.872
Model 1§
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 3.14 (1.92–5.12) <0.001 1 (Ref) 1.23 (0.71–2.14) 0.459
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 3.45 (1.88–6.34) <0.001 1 (Ref) 0.93 (0.46–1.86) 0.837
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 2.35 (1.27–4.34) 0.007 1 (Ref) 1.09 (0.52–2.29) 0.825
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 0.25 (0.13–0.48) <0.001 1 (Ref) 1.24 (0.62–2.46) 0.542
Model 2ǁ
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 3.09 (1.81–5.27) <0.001 1 (Ref) 1.40 (0.80–2.45) 0.236
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 3.73 (1.88–7.39) <0.001 1 (Ref) 1.14 (0.55–2.35) 0.722
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 2.69 (1.35–5.36) 0.005 1 (Ref) 1.26 (0.59–2.69) 0.556
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 0.23 (0.11–0.47) <0.001 1 (Ref) 1.03 (0.51–2.11) 0.925
Model 3
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 2.84 (1.65–4.90) <0.001 1 (Ref) 1.38 (0.78–2.43) 0.265
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 3.82 (1.89–7.73) <0.001 1 (Ref) 1.19 (0.57–2.50) 0.647
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 2.50 (1.23–5.07) 0.011 1 (Ref) 1.20 (0.55–2.61) 0.640
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 0.24 (0.11–0.52) <0.001 1 (Ref) 1.02 (0.49–2.13) 0.949

OR, odds ratio; CI, confidence interval; mRS, modified Rankin Scale; Ref, reference; NIHSS, National Institutes of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early Computed Tomography Score; HTN, hypertension; DM, diabetes mellitus; IVT, intravenous thrombolysis; HbA1c, hemoglobin A1c.

*

P-value by ordinal logistic regression and binary logistic regression;

†

Common OR: ordinal shift across the range of mRS scores toward a better outcome;

‡

Due to the very small number of symptomatic hemorrhagic transformation events (n=7), Firth’s penalized logistic regression model was applied;

§

Model 1: adjusted for age, sex, initial NIHSS, and ASPECTS;

ǁ

Model 2: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, and IVT;

Model 3: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, IVT, and HbA1c.

Table 5.

Sensitivity analysis of clinical outcomes comparing glycemic trajectory groups in the expanded cohort of patients with at least three glucose measurements

Outcomes Normoglycemic (Group 1, n=115)
Persistently elevated stable (Group 2, n=178)
Variable hyperglycemic (Group 3, n=68)
Reference Common OR or OR (95% CI) P* Common OR or OR (95% CI) P*
Unadjusted
 3-month mRS (shift analysis)
  Unadjusted common OR (95% CI)† 1 (Ref) 0.37 (0.24–0.57) <0.001 0.49 (0.29–0.83) 0.008
 Functional independence (mRS 0–2)
  Unadjusted OR (95% CI) 1 (Ref) 0.34 (0.21–0.56) <0.001 0.39 (0.21–0.72) 0.003
 Excellent outcome (mRS 0–1)
  Unadjusted OR (95% CI) 1 (Ref) 0.47 (0.29–0.77) 0.003 0.54 (0.29–1.02) 0.057
 Poor outcome (mRS 4–6)
  Unadjusted OR (95% CI) 1 (Ref) 3.33 (1.98–5.62) <0.001 2.90 (1.52–5.51) 0.001
 Symptomatic hemorrhagic transformation at day 5‡
  Unadjusted OR (95% CI)‡ 1 (Ref) 7.61 (0.41–140.62) 0.173 8.87 (0.41–190.45) 0.163
Model 1§
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.34 (0.22–0.53) <0.001 0.43 (0.24–0.77) 0.004
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.29 (0.17–0.51) <0.001 0.29 (0.14–0.60) <0.001
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.46 (0.26–0.81) 0.007 0.51 (0.25–1.06) 0.071
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.25 (2.30–7.86) <0.001 4.50 (2.08–9.77) <0.001
Model 2ǁ
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.35 (0.21–0.57) <0.001 0.51 (0.28–0.94) 0.030
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.27 (0.14–0.50) <0.001 0.34 (0.16–0.75) 0.007
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.43 (0.23–0.80) 0.008 0.55 (0.25–1.21) 0.138
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.55 (2.30–9.00) <0.001 3.93 (1.72–9.01) 0.001
Model 3
 3-month mRS (shift analysis)
  Adjusted common OR (95% CI) 1 (Ref) 0.38 (0.23–0.63) <0.001 0.55 (0.30–1.02) 0.058
 Functional independence (mRS 0–2)
  Adjusted OR (95% CI) 1 (Ref) 0.27 (0.14–0.52) <0.001 0.36 (0.16–0.79) 0.011
 Excellent outcome (mRS 0–1)
  Adjusted OR (95% CI) 1 (Ref) 0.47 (0.25–0.90) 0.022 0.58 (0.26–1.28) 0.177
 Poor outcome (mRS 4–6)
  Adjusted OR (95% CI) 1 (Ref) 4.18 (2.08–8.42) <0.001 3.57 (1.53–8.31) 0.003

OR, odds ratio; CI, confidence interval; mRS, modified Rankin Scale; Ref, reference; NIHSS, National Institutes of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early Computed Tomography Score; HTN, hypertension; DM, diabetes mellitus; IVT, intravenous thrombolysis; HbA1c, hemoglobin A1c.

*

P-value by ordinal logistic regression and binary logistic regression;

†

Common OR: ordinal shift across the range of mRS scores toward a better outcome;

‡

Due to the very small number of symptomatic hemorrhagic transformation events (n=7), Firth’s penalized logistic regression model was applied;

§

Model 1: adjusted for age, sex, initial NIHSS, and ASPECTS;

ǁ

Model 2: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, and IVT;

Model 3: adjusted for age, sex, initial NIHSS, ASPECTS, HTN, DM, IVT, and HbA1c.