Abstract
Background
Elevated blood glucose levels following acute ischemic stroke have been associated with adverse clinical outcomes in thrombolytic and nonthrombolytic treated patients. The current study examined multiple blood glucose parameters and their association with modified Rankin Scale (mRS) score at 3 months following mechanical thrombectomy and hospital discharge.
Methods
Acute ischemic stroke patients undergoing mechanical thrombectomy with a retrievable stent at two stroke centers were studied. Admission blood glucose level, maximum blood glucose during the hospital stay, and serial blood glucose measurements within the first 24 h of hospital admission were recorded. Variability in blood glucose level was represented by the standard deviation of the serial measurements within the first 24 h. The following demographic and clinical data was also collected: age, sex, baseline NIHSS score, onset-to-reperfusion times, hemoglobin A1c, and stroke mechanism.
Results
79 patients were identified; at 3 months, 35 patients had an mRS score of 0–2 and 44 had had an mRS of 3–6. Among the blood glucose variables, standard deviation of blood glucose in the first 24 h following admission and maximum blood glucose during hospital stay were significantly higher in the mRS 3–6 group. In multivariate logistic regression analysis, only the standard deviation of blood glucose remained significant (OR = 1.07, 95% CI = 1.02–1.11, p = 0.003) in a model that adjusted for admission NIHSS score (p = 0.016) and number of stent retriever passes (p = 0.042).
Conclusions
Greater blood glucose variability following acute ischemic stroke is associated with worse clinical outcome in patients undergoing mechanical thrombectomy.
Keywords: Acute stroke, Endovascular procedures, Functional recovery in stroke, Glucose, Stroke thrombectomy
Introduction
Elevated blood glucose (BG) levels at admission following acute ischemic stroke (AIS) have been associated with adverse clinical outcomes in both thrombolytic and nonthrombolytic treated patients [1, 2]. Hyperglycemia has also been shown to impair recanalization and decrease the reperfusion rates in patients treated with intravenous thrombolysis (IVT) [3, 4]. It has been hypothesized that the observed negative association between poststroke hyperglycemia and clinical outcome may be due to the overproduction of superoxides, while other theories suggest increased blood-brain barrier (BBB) disruption and reperfusion injury [5].
The BG parameter that is most closely related to outcome in previous studies of AIS patients receiving thrombolysis is inconsistent. Data from a recent study suggest that serial measurements of BG, not measurements obtained at admission alone, are a better predictor of functional outcome in patients with AIS [6]. This study focused on AIS patients treated with IVT using tissue plasminogen activator; however, standard deviation of BG has never been examined in patients undergoing mechanical thrombectomy (MT).
A meta-analysis of the five multicenter randomized controlled trials (MR CLEAN, ESCAPE, SWIFT PRIME, EXTEND-IA, and REVASCAT) demonstrated that an early intervention with a retrievable stent plus IVT resulted in greater functional recovery than IVT alone [7]. With MT as a definitive therapeutic option in ischemic stroke patients, more data is needed on the predictors of good outcome in this population [8]. The primary aim of this study was to look at multiple BG parameters in the first 24 h following admission for AIS and their association with functional outcome following MT with a retrievable stent. The secondary aim of our study was to investigate the association of a history of diabetes and poor glycemic control with functional outcome following endovascular therapy.
Methods
We retrospectively studied consecutive AIS patients undergoing MT with a retrievable stent between August 2012 and March 2016 at two stroke centers within the SSM Hospital System (SSM Saint Louis University Hospital and SSM DePaul Hospital). The institutional review board at Saint Louis University approved the study. The IRB waived the need for written informed consent due to the retrospective design of this study. Patients were enrolled in the study if they had AIS and were treated with MT. We excluded patients with loss to follow-up evaluations and without at least 3 results of BG within the first 24 h after admission.
Patient's demographic characteristics, stroke risk factors, National Institutes of Health Stroke Scale (NIHSS) on admission, hemoglobin A1c, symptom onset-to-reperfusion times, stroke mechanism, and clinical outcomes were collected retrospectively. The number of stent retriever passes during the MT, the stent retriever type used, and the time-to-recanalization were also recorded. Per hospital protocol, BG measurements were recorded upon arrival to the emergency room and subsequently 4–6 times during the first 24 h after admission. Per our institution's standard of practice, sliding scale insulin was used for each patient to hold BG in the range of 120–160 mg/dL. Various BG parameters were included: for each patient, the mean of the patient's multiple BG values obtained during the first 24 h after admission; for each patient, the standard deviation of the patient's multiple BG values obtained during the first 24 h after admission; and for each patient, the maximum BG level obtained during the patient's entire hospital stay. Standard deviation was defined as the square root of the average squared difference between each individual number and the mean of the multiple BG values obtained during a patient's first 24 h after admission.
The primary study outcome was to compare measurements of BG and identify the best correlate(s) of modified Rankin Scale (mRS) score at 3 months following MT in AIS. Time from stroke onset to reperfusion was relevant only for cases with successful recanalization (n = 73, 92.4%); associations with outcome were estimated from applicable cases. Univariate comparisons of study variables between the two outcome groups - mRS 0–2 vs. 3–6 - were made using the independent t test (pooled variance estimate, unless otherwise indicated) or the χ2 test of association (Pearson χ2, unless cell sizes required Fisher exact test). In multivariate analysis, forward conditional-selection logistic regression was used to identify variables with multivariate associations with mRS 0–2 vs. 3–6. To evaluate consistency of the multivariate model, backward conditional-elimination logistic regression was also used. Odds ratios (OR) and 95% confidence intervals were calculated; sensitivity and specificity values for outcome classification were generated. Statistical significance for all analyses was evaluated at p < 0.05.
Results
79 patients were included in this study. Baseline characteristics and descriptive results are shown in Table 1. The study group had a mean age of 69 years with a range from 31 to 98. 49% of enrolled patients were male. 43 patients had diabetes and 15 had hyperglycemia at admission with no prior diagnosis of diabetes. 19 patients had had a previous stroke or transient ischemic attack. 35 of the 79 included patients had a 90-day mRS score less than 2. Comparisons of variables as a function of outcome (mRS score at 3 months: 0–2 vs. 3–6) are also represented. Standard deviation of BG measurements, intracranial hemorrhage (ICH), maximum BG obtained, admission NIHSS, unsuccessful recanalization, and number of stent retriever passes were higher or more prevalent (all p < 0.05) in the mRS score 3–6 group, relative to the mRS score 0–2 group. No other significant differences were obtained.
Table 1.
Descriptive results for total study sample and as a function of outcome
| Variables | Total (n = 79) | mRS 0–2 (n = 35, 44.3%) | mRS 3–6 (n = 44, 55.7%) | P |
|---|---|---|---|---|
| Age, years | 69.2±13.8 | 68.3±14.7 | 69.8±13.3 | 0.634 |
| Female gender | 40 (50.6) | 18 (51.4) | 22 (50.0) | 0.900 |
| Risk factors | ||||
| Hypertension | 63 (79.7) | 25 (71.4) | 38 (86.4) | 0.101 |
| Atrial fibrillation | 37 (46.8) | 15 (42.9) | 22 (50.0) | 0.527 |
| Diabetes | 43 (54.4) | 19 (54.3) | 24 (54.5) | 0.982 |
| Dyslipidemia | 35 (44.3) | 14 (40.0) | 21 (47.7) | 0.492 |
| Coronary artery disease | 33 (41.8) | 13 (37.1) | 20 (45.5) | 0.457 |
| Prior stroke/TIA | 19 (24.1) | 6 (17.1) | 13 (29.5) | 0.200 |
| Smoking | 34 (43.0) | 13 (37.1) | 21 (47.7) | 0.345 |
| Stroke etiology | ||||
| Cardioembolism | 36 (45.6) | 20 (57.1) | 16 (36.4) | 0.065 |
| Large artery atherosclerosis | 35 (44.3) | 13 (37.1) | 22 (50.0) | 0.253 |
| Occlusion site | ||||
| MCA, MCA1, and/or MCA2 | 54 (68.4) | 27 (77.1) | 27 (61.4) | 0.134 |
| ICA, ICA-T, or ICA-C | 18 (22.8) | 7 (20.0) | 11 (25.0) | 0.599 |
| Basilar | 7 (8.9) | 2 (5.7) | 5 (11.4) | 0.454* |
| Admission NIHSS | 16.4±6.3 | 14.6±7.1 | 17.9±5.3 | 0.018 |
| Admission SBP | 156.9±28.0 | 157.4±31.4 | 156.5±25.3 | 0.894 |
| Admission BG | 153.7±66.3 | 143.7±53.5 | 161.7±74.7 | 0.234 |
| Mean BG in first 24 h | 151.9±57.7 | 138.0±52.4 | 163.0±59.8 | 0.056 |
| SD BG in first 24 h | 30.8±23.8 | 20.2±13.0 | 39.2±27.0 | <0.001† |
| Maximum BG | 199.8±78.2 | 172.4±62.2 | 221.6±83.3 | 0.005 |
| Hemoglobin A1c | 6.9±1.9 | 6.8±1.8 | 7.0±1.9 | 0.719 |
| Hours to reperfusion‡ | 5.4±2.9 | 5.3±3.0 | 5.5±2.9 | 0.753 |
| Recanalization success | 73 (92.4) | 35 (100.0) | 38 (86.4) | 0.031* |
| MT devices¶ | 0.988 | |||
| Trevo | 30 (39.0) | 13 (37.1) | 17 (40.5) | |
| Solitaire | 11 (14.3) | 5 (14.3) | 6 (14.3) | |
| Penumbra | 8 (10.4) | 4 (11.4) | 4 (9.5) | |
| Stent retriever passes | 2.1±1.1 | 1.8±1.0 | 2.3±1.0 | 0.036 |
| ICH | 24 (30.4) | 1 (2.9) | 23 (52.3) | <0.001* |
Data are given as mean ± SD or n (%). mRS, modified Rankin Scale; SD, standard deviation; TIA, transient ischemic attack; MCA, middle cerebral artery; ICA, internal carotid artery; NIHSS, National Institutes of Health Stroke Scale; SBP, systolic blood pressure; BG, blood glucose; MT, mechanical thrombectomy; ICH, intracranial hemorrhage.
Fisher exact test; this variable was not considered in the logistic regression analysis.
Significance for pooled and separate variance estimates.
n = 73.
p value is for 4 × 2 χ2 analysis; this variable was not considered in the logistic regression analysis.
Table 2 displays the results of the multivariable, forward conditional-selection and backward conditional-elimination logistic regression analyses. In these analyses, all variables were considered for inclusion in the logistic regression equation, with the exception of basilar occlusion site, recanalization success, and ICH (variables for which sample sizes were too low or high as a function of mRS groups for inclusion in a logistic model; see Table 2). In separate analyses, variables were added (conditional-selection) or retained (conditional-elimination) based on statistically significant associations with mRS (0–2 vs. 3–6). A significant logistic model was obtained, χ2(3) = 26.3, p < 0.001, Nagelkerke R2 = 0.38, with three variables emerging as significant correlates of mRS category: standard deviation of BG measurements (OR = 1.07, p = 0.003), admission NIHSS (OR = 1.12, p = 0.016), and number of stent retriever passes (OR = 1.91, p = 0.042). Together, these variables generated a sensitivity of 77% and a specificity of 74% for predicting outcome. Identical solutions were obtained for both logistic regression analyses (forward conditional-selection and backward conditional-elimination).
Table 2.
Multivariate associations with outcome: logistic regression analysis
| Variables | Odds ratio | 95% CI | p |
|---|---|---|---|
| SD BG | 1.07 | 1.02–1.11 | 0.003 |
| Admission NIHSS | 1.12 | 1.02–1.23 | 0.016 |
| Stent retriever passes | 1.91 | 1.03–3.57 | 0.042 |
Outcome variable: modified Rankin Scale score at 3 months. Model Nagelkerke R2 = 0.38; sensitivity = 77%, specificity = 74%. CI, confidence interval; SD, standard deviation; BG, blood glucose; NIHSS, National Institutes of Health Stroke Scale.
Due to low or high prevalence of events within one or both mRS groups, basilar occlusion, successful recanalization, and ICH variables could not be considered for inclusion in the logistic regression analysis. Inclusion of such variables leads to overfitting of logistic regression models. It was important, however, to consider whether the standard deviation of BG measurements would still be significantly associated with mRS with adjustment, where applicable, for these variables (since basilar occlusion was not significantly associated with mRS, it was not given further consideration). Multiple linear regression was used to investigate this issue, with mRS expressed as a continuous criterion variable (mRS = 0–6). Multiple linear regression was able to accommodate the dichotomous variables for successful recanalization (0 = no, 1 = yes) and ICH (0 = no, 1 = yes). With standard deviation of BG measurements, admission NIHSS, number of stent retriever passes, ICH, and recanalization success included in the analysis, the multiple regression equation was significant, F(5, 73) = 13.9, p < 0.001, adjusted R2 = 0.45. As shown in Table 3, standard deviation of BG measurements, admission NIHSS, ICH, and recanalization success - but not number of stent retriever passes - were associated with higher mRS scores. Thus, standard deviation of BG measurements was associated with mRS even with adjustment for recanalization success and ICH.
Table 3.
Multivariate associations with outcome: linear regression analysis
| Variables | B | 95% CI | β | P |
|---|---|---|---|---|
| ICH | 1.88 | 1.10 to 2.66 | 0.42 | <0.001 |
| SD BG | 0.03 | 0.02 to 0.05 | 0.36 | <0.001 |
| Recanalization success | −2.02 | −3.34 to −0.69 | −0.26 | 0.003 |
| Admission NIHSS | 0.07 | 0.01 to 0.12 | 0.20 | 0.024 |
| Stent retriever passes | 0.18 | −0.17 to 0.53 | 0.09 | 0.302 |
Outcome variable: modified Rankin Scale score at 3 months. Model adjusted R2 = 0.45. B, unstandardized regression coefficient; CI, confidence interval; β, standardized regression coefficient; ICH, intracranial hemorrhage; SD, standard deviation; BG, blood glucose; NIHSS, National Institutes of Health Stroke Scale.
Discussion
This retrospective study assessed various BG parameters as predictors of clinical outcome in patients being treated with MT using stent retrievers. In contrast to our analysis, previous studies have focused on admission BG as a predictor of outcome [8, 9, 10]. The time point of glucose measurement is not clear in these studies, and measurements taken upon admission are not indicative of when the stroke actually occurred. Our study actually showed that the mean of BG measurements obtained in the first 24 h had a nonsignificant yet positive trend with mRS score > 2 at 3 months, and, more importantly, the standard deviation of those measurements in the first 24 h along with the maximum BG value obtained over the course of their hospital stay was significantly associated with mRS score > 2 at 3 months. These results would suggest that values obtained following intervention with thrombectomy could be better for predicting prognosis than admission glucose alone.
There have been several theories regarding the contribution of higher BG levels to poor outcomes following IVT or MT. It has been observed that hyperglycemia is associated with larger infarct volumes and reduced salvage of perfusion-diffusion mismatch tissue [11]. This means that hyperglycemia may be associated with a larger increase in the infarct volume leading to a worse clinical outcome despite recanalization [12]. A common theory for the occurrence of ICH in patients with hyperglycemia after MT is that it is due to the BBB disruption and microvasculature impairments [13, 14]. If worsened outcome after ischemic stroke is truly related to superoxide production and/or dysfunction of the BBB, then it is probable that this is an ongoing process that is highly dependent on stroke severity and BG measured over time.
We did not find a significant difference in hemoglobin A1c among those with good clinical outcomes compared to those without. This may indicate that a patient's glucose control prior to AIS is not a major contributor to their clinical outcome. Our study points to poststroke hyperglycemia - likely secondary to stress-related cortisol release - as the more important factor. Overproduction of counterregulatory hormones and cytokines leads to failure of insulin to suppress hepatic gluconeogenesis in a hyperglycemic environment [15]. Many studies have been done to demonstrate the significant portions of hospitalized patients exhibiting hyperglycemia, many of whom meet the criteria for diabetes, with no history of a previous diagnosis of the disease [16, 17].
In our study, BG variability was more associated with clinical outcome in patients with AIS treated with MT than any single BG reading taken prior to the procedure. Yoo et al. [6] found that BG after IVT, mean BG, and maximum BG were all associated with mRS score > 2 at 3 months. We obtained similar results in our study with MT, showing mean BG and maximum BG as predictors of outcome. However, Yoo et al. [6] found that standard deviation of BG showed a nonsignificant yet positive association with death within 3 months. The results of our study showed standard deviation of BG to have the highest predictive value for good versus poor outcome. There was also a recent study that showed BG variability as a strong predictor of unfavorable clinical outcome in patients with subarachnoid hemorrhage. This association has been well established in critically ill patients without stroke, but only in regard to mortality and not functional or neurological outcome [18].
There are a few limitations to this study. First, it is limited by its retrospective design with a modest sample size taken from two medical centers. In addition, we were not able to collect data on the insulin treatments given to the patients in the study who had hyperglycemia. Variability in insulin regimens is likely a confounding variable in this study, specifically in regards to the postadmission BG levels used to obtain mean and standard deviation. Also, patients with elevated blood sugar would be more likely to have multiple BG measurements during the course of their hospital stay. In addition, there is currently no universal standard of practice for diet or hydration following MT at our institution. The cases are discussed amongst the interventionalist and the critical care team to tailor to the needs of each individual patient.
In conclusion, the results of our study suggest that using standard deviation of serial measurements of BG might be better than single BG measurements, such as admission BG, at predicting clinical outcome in patients with AIS treated with MT. It remains unclear, however, whether correction of this variability in BG would result in improved functional outcome. Current guidelines from the AHA recommend treating hyperglycemia to achieve a BG level in a range of 140–180 mg/dL following AIS. Tight glycemic control using insulin can result in hypoglycemia, and a strict protocol to maintain BG within this small target range would be difficult to implement given any necessary adjustments in oral intake along with the importance of early mobilization in poststroke care. Further research into the association of BG variability and clinical outcome after MT for AIS is needed.
Disclosure Statement
All authors report no conflict of interest, financial or other, relevant to this paper.
References
- 1.Ahmed N, Dávalos A, Eriksson N, Ford GA, Hennerici M, et al. Association of admission blood glucose and outcome in patients treated with intravenous thrombolysis. Arch Neurol. 2010;67:1123–1130. doi: 10.1001/archneurol.2010.210. [DOI] [PubMed] [Google Scholar]
- 2.Alvarez-Sabín J, Molina CA, Montaner J, Arenillas JF, Huertas R, Ribo M, et al. Effects of admission hyperglycemia on stroke outcome in reperfused tissue plasminogen activator-treated patients. Stroke. 2003;34:1235–1240. doi: 10.1161/01.STR.0000068406.30514.31. [DOI] [PubMed] [Google Scholar]
- 3.Ribo M, Molina C, Montaner J, Rubiera M, Delgado-Mederos R, Arenillas JF, et al. Acute hyperglycemia state is associated with lower tPA-induced recanalization rates in stroke patients. Stroke. 2005;36:1705–1709. doi: 10.1161/01.STR.0000173161.05453.90.9f. [DOI] [PubMed] [Google Scholar]
- 4.Ribo M, Molina CA, Delgado P, Rubiera M, Delgado-Mederos R, Rovira A, et al. Hyperglycemia during ischemia rapidly accelerates brain damage in stroke patients treated with tPA. J Cereb Blood Flow Metab. 2007;27:1616–1622. doi: 10.1038/sj.jcbfm.9600460. [DOI] [PubMed] [Google Scholar]
- 5.Suh SW, Shin BS, Ma H, Van Hoecke M, Brennan AM, Yenari MA, et al. Glucose and NADPH oxidase drive neuronal superoxide formation in stroke. Ann Neurol. 2008;64:654–663. doi: 10.1002/ana.21511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Yoo DS, Chang J, Kim JT, Choi MJ, Choi J, Choi KH, et al. Various blood glucose parameters that indicate hyperglycemia after intravenous thrombolysis in acute ischemic stroke could predict worse outcome. PLoS One. 2014;9:e94364. doi: 10.1371/journal.pone.0094364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Goyal M, Menon BK, van Zwam WH, Dippel DW, Mitchell PJ, Demchuk AM, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet. 2016;387:1723–1731. doi: 10.1016/S0140-6736(16)00163-X. [DOI] [PubMed] [Google Scholar]
- 8.Ozdemir O, Giray S, Arlier Z, Baş DF, Inanc Y, Colak E. Predictors of a good outcome after endovascular stroke treatment with stent retrievers. Sci World J. 2015;2015:403726. doi: 10.1155/2015/403726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Singer OC, Haring HP, Trenkler J, Nolte CH, Bohner G, Riech A, et al. Age dependency of successful recanalization in anterior circulation stroke: the ENDOSTROKE study. Cerebrovasc Dis. 2013;36:437–445. doi: 10.1159/000356213. [DOI] [PubMed] [Google Scholar]
- 10.Desilles JP, Meseguer E, Labreuche J, Lapergue B, Sirmimarco G, Gonzalez-Valcarcel J, et al. Diabetes mellitus, admission glucose, and outcomes after stroke thrombolysis: a registry and systematic review. Stroke. 2013;44:1915–1923. doi: 10.1161/STROKEAHA.111.000813. [DOI] [PubMed] [Google Scholar]
- 11.Natarajan SK, Dandona P, Karmon Y, Yoo AJ, Kalia JS, Hao Q, et al. Prediction of adverse outcomes by blood glucose level after endovascular therapy for acute ischemic stroke. J Neurosurg. 2011;114:1785–1799. doi: 10.3171/2011.1.JNS10884. [DOI] [PubMed] [Google Scholar]
- 12.Mazighi M, Labreuche J, Amarenco P. Glucose level and brain infarction: a prospective case & control study and prospective study. Int J Stroke. 2009;4:346–351. doi: 10.1111/j.1747-4949.2009.00329.x. [DOI] [PubMed] [Google Scholar]
- 13.Hawkins BT, Lundeen TF, Norwood KM, Brooks HL, Egleton RD. Increased blood-brain barrier permeability and altered tight junctions in experimental diabetes in the rat: contribution of hyperglycaemia and matrix metalloproteinases. Diabetologia. 2007;50:202–211. doi: 10.1007/s00125-006-0485-z. [DOI] [PubMed] [Google Scholar]
- 14.Kamada H, Yu F, Nito C, Chan PH. Influence of hyperglycemia on oxidative stress and matrix metalloproteinase-9 activation after focal cerebral ischemia/reperfusion in rats: Relation to blood-brain barrier dysfunction. Stroke. 2007;38:1044–1049. doi: 10.1161/01.STR.0000258041.75739.cb. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.McCowen KC, Malhotra A, Bistrian BR. Stress-induced hyperglycemia. Crit Care Clin. 2016;17:107–124. doi: 10.1016/s0749-0704(05)70154-8. [DOI] [PubMed] [Google Scholar]
- 16.Karunakar MA, Staples KS. Does stress-induced hyperglycemia increase the risk of perioperative infectious complications in orthopaedic trauma patients? J Orthop Trauma. 2010;24:752–756. doi: 10.1097/BOT.0b013e3181d7aba5. [DOI] [PubMed] [Google Scholar]
- 17.Kopelman TR, O'Neill PJ, Kanneganti SR, Davis KM, Drachman DA. The relationship of plasma glucose and glycosylated hemoglobin A1C levels among nondiabetic trauma patients. J Trauma. 2008;64:30–33. doi: 10.1097/TA.0b013e318161b0ab. [DOI] [PubMed] [Google Scholar]
- 18.Okazaki T, Hifumi T, Kawakita K, Shishido H, Ogawa D, Okauchi M, et al. Blood glucose variability: a strong independent predictor of neurological outcomes in aneursymal subarachnoid hemorrhage. J Intensive Care Med. 2016:1–7. doi: 10.1177/0885066616669328. [DOI] [PubMed] [Google Scholar]