Hyperglycaemia, Insulin Therapy and Critical Penumbral Regions for Prognosis in Acute Stroke: Further Insights from the INSULINFARCT Trial

Background Recently, the concept of ‘clinically relevant penumbra’ was defined as an area saved by arterial recanalization and correlated with stroke outcome. This clinically relevant penumbra was located in the subcortical structures, especially the periventricular white matter. Our aims were to confirm this hypothesis, to investigate the impact of admission hyperglycemia and of insulin treatment on the severity of ischemic damages in this area and to study the respective contributions of infarct volume and ischemic damage severity of the clinically relevant penumbra on 3-month outcome. Methods We included 99 patients from the INSULINFARCT trial. Voxel-Based Analysis was carried on the Apparent Diffusion Coefficient (ADC) maps obtained at day one to localize the regions, which were more damaged in patients i) with poor clinical outcomes at three months and ii) without arterial recanalization. We determined the intersection of the detected areas, which represents the clinically relevant penumbra and investigated whether hyperglycemic status and insulin regimen affected the severity of ischemic damages in this area. We performed logistic regression to examine the contribution of infarct volume or early ADC decrease in this strategic area on 3-month outcome. Findings Lower ADC values were found in the corona radiata in patients with poor prognosis (p< 0.0001) and in those without arterial recanalization (p< 0.0001). The tracking analysis showed that lesions in this area interrupted many important pathways. ADC values in this area were lower in hyperglycemic than in normoglycemic patients (average decrease of 41.6 ± 20.8 x10−6mm2/s) and unaffected by the insulin regimen (p: 0.10). ADC values in the clinically relevant penumbra, but not infarct volumes, were significant predictors of 3-month outcome. Conclusion These results confirm that the deep hemispheric white matter is part of the clinically relevant penumbra and show that hyperglycaemia exacerbates the apparition of irreversible ischemic damage within 24 hours in this area. However, early intensive insulin therapy fails to protect this area from infarction. Trial Registration ClinicalTrials.gov NCT00472381

P oststroke hyperglycemia is an independent predictor of poor functional outcome and death in the acute phase of stroke [1][2][3][4] but this statistical relationship does not prove causality. Indeed, there is still controversy about whether stroke-related hyperglycemia is a cause or effect of the more severe damage found in patients with stroke with elevated blood sugars. However, the hyperglycemia "toxicity" has been suggested by animal studies that reported accelerated penumbra-into-infarction conversion and no-reflow phenomenon. 5 Recent MRI and transcranial Doppler studies have suggested that a similar phenomenon may occur in humans. 6 -10 In these studies, evidence has accumulated to define that the glucose toxicity threshold was low (approximately 7 mmol/L, between 6 and 8 mmol/L). 11 These data explain why hyperglycemia is increasingly considered to be a potential therapeutic target in acute stroke and is of growing interest for intensive insulin treatment (IIT). 11,12 Currently published randomized trials do not conclude on the clinical efficacy of IIT in patients with stroke. [13][14][15][16][17][18][19][20] The UK Glucose Insulin in Stroke Trial (GIST), 14 which has enrolled 933 patients with stroke in the first 24 hours of stroke onset, is one such example. It has been criticized because of the heterogeneous population included in the study, its slow recruitment rate, late treatment initiation, and inefficient glucose control. Seven other small studies did not have the statistical power to detect clinical efficiency. 13,[15][16][17][18][19][20] They all showed that IIT carries a high risk of hypoglycemia (4%-76%). One of these studies 18 found that IIT was associated with greater infarct growth (IG) on MRI in patients with persistent arterial occlusion. However, the median delay of treatment initiation was 20 hours after stroke onset, so the study was unlikely to detect a putative protective effect on the ischemic penumbra. Additionally, better glucose control was not obtained in the intensive group, except in a small time window. To date, a larger study examining an intensive glycemic protocol versus usual care on the effects of glucose control and infarct expansion would be of great interest in patients with stroke. In such a study, it would be critical to start treatment as early as possible. 21 The INSULINFARCT study was designed to address these issues. INSULINFARCT is a randomized proof-of-concept trial comparing IIT versus control subcutaneous treatment initiated within 6 hours of stroke onset in carotid infarction. The study was powered to detect an increased efficiency of IIT on serum glucose control during the next 24 hours compared with usual care with subcutaneous insulin. The working hypothesis was that IIT, by better glucose control, would reduce subsequent IG on MRI because hyperglycemia is independently associated with increased IG. 7,8,10,22

Study Design
The INSULINFARCT study was an academic monocenter (Pitié-Salpêtrière Hospital, Paris, France), prospective, randomized, unblinded trial in patients with hyperacute stroke (Ͻ6 hours) to investigate insulin therapy management in the first 24 hours after admission. The first objective was to assess the efficiency of glycemic control achieved with intravenous insulin perfusion versus control subcutaneous insulin therapy over 24 hours. The second objective was to compare the MRI infarct growth between the 2 groups. The third objective was to compare the clinical and safety outcomes 90 days after stroke. The first and second objectives were planned to valid the working hypothesis, that is, that IIT, by better glucose control, would reduce subsequent IG on MRI. For recruitment reasons, the study initially designed for 2 years was extended to 4 years. There were no changes in the selection criteria or design during the recruitment period. The trial ended when the number of patients reached the required sample size for the primary objective.

Population
Eligible patients were prospectively enrolled and were consecutive patients from the Pitié-Salpêtrière Hospital admitted between June 2007 and March 2011 who met the following criteria: (1) an ischemic stroke in the carotid territory proven by initial diffusionweighted imaging (DWI); (2) an initial MRI with DWI Ͻ5 hours of stroke onset; (3) initiation of insulin treatment within 1 hour of MRI (Ͻ6 hours after stroke onset); and (4) an admission score on the National Institutes of Health Stroke Scale (NIHSS) between 5 and 25. Patients were eligible whatever their baseline serum glucose level and their non-or diabetic status. Symptom onset was defined as the last time the patient was seen in normal health. The exclusion criteria were life-threatening conditions that limited follow-up visits, preadmission modified Rankin Scale Ͼ2, or patients under legal protection. The neurological examination was assessed using the NIHSS at admission (before the MRI and any treatment), on Days 1 and 7. Blood pressure, heart rate, and temperature were recorded at admission and on Day 1. Body mass index and umbilical perimeter were measured.
The study was conducted according to good clinical practice guidelines and was approved by the local ethical committee. Written informed consent was obtained by the on-duty physician from each participant or from a legal proxy/family member before randomization. Consent was not immediately required if the patient had severe language disturbances, neglect, or loss of consciousness so that the insulin treatment could be initiated as soon as possible.

Randomization and Masking
Each participant was randomized 1:1 by a secure web site (Cleanweb, Telemedicine Technologies) between 2 groups: the IIT (intravenous insulin continuously) or the SIT (subcutaneous insulin every 4 hours) group. Randomization was previously entered into the system using a prespecified randomization list (random-number generator). The randomization list was made of 4 blocks of 45 patients. None of the investigators were aware of the randomization list or the number or size of the blocks. Treatment allocation was performed during the 24 hours after randomization. Physicians were unblinded to the patients' treatment.

Treatment Protocol
In the IIT group, soluble human Actrapid insulin was administered in an intravenous continuous infusion with hourly dose adaptation to the capillary glucose test control (CGT) according to the nomogram shown in Figure 1A. In the SIT group, insulin was administered subcutaneously every 4 hours based on CGT values ( Figure 1B). Nomograms were designed by the endocrinologic department. Insulin was stopped when the CGT value reached the lowest limit of the nomogram in each group. The stop point was Ͻ5.5 mmol/L in the IIT group and 8 mmol/L in the SIT group. Both groups have different target glucose thresholds because the aim of the study was to compare our usual management of glucose control (SIT group) with aggressive insulin therapy (IIT). Nevertheless, both nomograms were designed to achieve a targeted glucose control under the "toxic" threshold of 7 mmol/L. In both groups, the patients received saline infusion with potassium. Oral feeding was allowed if the neurological deficit was minor.
In case of hypoglycemia (defined as a CGT Ͻ3 mmol/L), the patient received 10 mL of 30% glucose infusion and CGT was checked every 15 minutes. If CGT reached the objective of the arm treatment, the doses of the allocated treatment were divided by 2.
After the 24 hours of the study treatment, the treatment was decided by the on-duty physician. The recommendation was to use either subcutaneous insulin every 4 or 8 hours, antidiabetic drugs, or no treatment at all. Intravenous insulin infusion was recommended to be stopped.

MRI Parameters
The initial MRI was performed at admission (Ͻ5 hours) before randomization and the control MRI was performed between Days 1 and 3.
MRI was performed on a 3.0-T whole-body MR unit (Signa 3.0-T HDx; General Electric Medical System, Milwaukee, WI) with an 8-channel head coil. The MRI protocol included 4 sequences: DWI, fluid-attenuated inversion recovery, intracranial time-of-flight MR angiography, and T2*-weighted sequence. Parameters could be found in Appendix I in the online-only Data Supplement.

MRI Assessment
The admission infarct volume was defined as the hyperintense area on initial DWI (b valueϭ1000 s/mm 2 ). It was delineated by interactive manual outlining using the Neurinfarct software (Intelligence in Medical Technology, Paris, France). The follow-up infarct volume was determined using identical methodology applied to the follow-up DWI. This method has been shown to be highly reproducible. 23,24 The IG was defined as the difference between follow-up and admission DWI volumes. All volume determinations were performed by one author (C.R.) who was blinded to arm treatment.
Intracranial occlusion and arterial recanalization were assessed on initial and follow-up MR angiography, respectively. Recanalization was considered in a 3-item scale: (1) patent or complete recanalization; (2) partial or minimal flow-related signal in the region of the arterial clot or displacement of the initial clot in a distal portion of the occluded artery; and (3) persistent occlusion.

Blood Sample Collection
Patients had venous blood glucose measurement taken at admission before MRI and at the end of the treatment. Blood samples were obtained on Day 1 to detect diabetes mellitus and low-density lipoprotein cholesterol level. Blood glucose, diabetes mellitus, and low-density lipoprotein cholesterol measurements were centralized at the Pitié-Salpêtrière Hospital biochemical laboratory. CGTs were performed using an Optium H apparatus according to manufacturer procedures (ABBOTT-MediSense; www.abbottdiabetescare.co.uk/healthcare-professionals/clinical-papers).

Outcome Scales
The modified Rankin Scale (mRS) was used to assess the outcome at discharge and at 90 days. A good outcome was defined as independency (mRS 0 -2) and a poor outcome as moderate to severe disability (mRS [3][4][5]. This scale was performed by trained neurologists who were not blinded to the arm treatment.

Statistical Analysis
The analysis was performed as intention-to-treat to reflect the clinical practice in a stroke unit where patients with acute stroke are treated as soon as possible. Analysis was performed by the statistician of the trial (J.-C.C), blinded for treatment allocation, with Statistica software (StatSoft, Inc, Maison-Alfort, France).
Descriptive statistics are median and interquartile range (IQR). Comparisons of proportions were made by a 2 test; the quantitative variables were compared by t tests or repeated-measure analysis of variance. MRI volumes were compared by Mann-Whitney U tests because their distribution is not normal. 23,24 ORs and their 95% CI were computed for all binarized data as a measure of the effect size.

Glucose Control
The primary objective of the study was to test the hypothesis that the percentage of patients with a mean CGT Ͻ7 mmol/L during the 24-hour protocol would be higher in the IIT than in the SIT group.
This was a necessary step to prove that intensive management of stroke-related hyperglycemia would modify infarct growth.
The mean CGT was defined as the averaged of CGT values determined at H4 (4 hours after randomization), H8, H12, H16, H20, and H24. We calculated that it was necessary to include 82 patients per group to detect a difference of 20% in the proportion of patients with a mean CGT Ͻ7 mmol/L with an ␣ risk of 5% and a 90% power. Considering possible exclusions due to missing data, the sample size of both groups was fixed at 90. It was prespecified in the protocol that patients with Ͼ3 missing CGT values would be excluded. We also compared the CGT values in both groups at specific time points (H4, H8, H12, H16, H20, and H24), the proportion of patients with CGT Ͻ7 mmol/L at each of these time points.

Secondary Objectives
The subsequent prespecified secondary efficacy outcome variable was the comparison of IG on MRI. Based on previous data, 10,24 the study was sufficiently powered to detect a difference of 15 cm 3 in the infarct growth volume assuming a SD of 30 cm 3 : 70 patients per group were necessary with an ␣ risk of 5% and a 80% power. In addition, although this was not prespecified by the protocol, we had searched for treatment group and vessel patency effects on IG using a 2-way analysis of variance as was performed in a previous study. 18 We also performed a broad analysis of determinants of infarct growth by running a stepwise multiple regression analysis to adjust with the baseline variables. The IG was the dependent variable (as a continuous variable) and the independent variables were the baseline clinically relevant variables such as: age, arm of insulin treatment (SITϭ1 and IITϭ2), baseline NIHSS, diabetes mellitus binary status (Ն6.2%), thrombolytic treatment, time between stroke and MRI, hypoglycemic events, and DWI volume at admission.
The other efficacy outcome, more exploratory, was the comparison of proportions of patients with good functional outcome at 3 months. We also analyzed functional outcome across the entire distribution of the mRS scores at Day 90 ( 2 test, dlϭ6). We finally ran a logistic regression analysis with good functional outcome as the dependent variable and with the same independent variables as the previous regression analysis on IG. In addition, we compared NIHSS at Day 1 and Day 7 after stroke onset.
The prespecified safety outcomes were hypoglycemias (defined by CGT Ͻ3 mmol/L), death within 3 months, and serious adverse events including symptomatic intracerebral hemorrhages (as defined as any increase in NIHSS Ͼ3 points during the first week attributed to CT-or MRI-documented parenchymal hematomas), any neurological worsening (NIHSS Ͼ3 points), and any event that lengthened the hospital stay or was life-threatening.

Population
Between June 2007 and March 2011, 230 patients were eligible and 180 were included in the study (Figure 2). The functional outcome and safety analyses were performed on the 180 patients in the intention-to-treat analysis. The analysis of glucose control was performed on 176 patients (89 in the SIT and 87 in the IIT groups) because 4 patients with Ͼ3 missing CGT values were excluded as prespecified in the protocol. In the per-protocol analysis, 6 patients were excluded because of violation of inclusion/exclusion criteria: vertebrobasilar stroke (nϭ2), mimicked stroke related to a dural fistula (nϭ1), preadmission mRS Ͼ2 (nϭ1), or because they were under legal protection (nϭ2). The MRI analysis was performed in 160 patients (80 in the control and 80 in the intensive group) because the control MRI was missing in 16 patients (6 early deaths, 3 comas, 7 follow-up MRIs performed after 3 days). Appendix II in the online-only Data Supplement provides details and sensitivity analysis on patients with missing MRI data in each group.
Patient characteristics are presented in Table 1. The body mass index was higher in the IIT group (Pϭ0.02). Time to initial MRI tends to be earlier in the IIT group (Pϭ0.06). Although there was no statistical difference between groups for the other variables, patients tend to be older and less frequently treated by recombinant tissue-type plasminogen activator in the SIT than in the IIT group. Fasting patients were 71% (63 of 89) in the SIT and 70% (61 of 87) in the IIT group. The number of patients under oral hypoglycemic agents before admission was 12.5% (13 of 89) in the SIT and 7% (6 of 87) in the IIT group. The initial CGTs were similar in both groups with a median at 6.6 mmol/L in the SIT group versus 6.7 mmol/L in the IIT group (Pϭ0.48). Proportion of patients with initial CGT Ն7 mmol/L were 44% (39 of 89) in the SIT group versus 39% (34 of 87) in the IIT group (Pϭ0.6). The median delay between initial MRI and randomization was similar (39 minutes; IQR, 32-49 versus 37 minutes; IQR, 29 -48; Pϭ0.89).

Glucose Control
The primary end point of a mean CGT Ͻ7 mmol/L was reached in 95.4% (83 of 87) of patients in the IIT and in 67.4% (60 of 89) of patients in SIT group (PϽ0.0001). The OR for the IIT group was 9.5 (95% CI, 3.2-28.6). The mean 24-hour CGT was lower in the IIT group (5.7 mmol/L; IQR, 5.2-5.9 versus 6.6 mmol/L; IQR, 5.6 -7.3; PϽ0.0001). Figure  3 shows the CGT time course in both groups. The time course of CGT showed a significant group*time interaction ( Figure  3; repeated-measure analysis of variance, Pϭ0.005). The proportion of patients with a CGT Ͻ7 mmol/L at H4 was 86% in the IIT group and 58% in the SIT group (Pϭ0.001; Figure 3). Nearly all patients (98%) received some insulin in the intensive group (24-hour dose, 18.5 UI; IQR, 9 -24), whereas only 55% of patients received insulin in the control group (dose, 2 UI; IQR, 0 -4; PϽ0.0001 between groups).

MRI Outcome
As shown in Table 1, the initial infarct volume was similar in both groups as was the percentage of intracranial arterial occlusions. The control MRI was performed at a similar delay in the IIT and SIT groups (29.5

Safety Outcome
Five patients (5.7%) had 8 asymptomatic hypoglycemic episodes in the IIT group versus none in the SIT group (Pϭ0.07 for the number of patients and Pϭ0.02 for the number of events). None of the hypoglycemic episodes was considered symptomatic (worsening in NIHSS or consciousness disturbance). If we consider a less conservative threshold to define hypoglycemia (Յ3.6 mmol/L), these rates were 1.1% (one of 89) versus 34.5% (30 of 87) of hypoglycemic events for 1.1% (one of 89) and 19.5% (17 of 87) patients in the SIT versus the IIT group. Numbers are median and interquartile range. SIT indicates subcutaneous insulin therapy; IIT, intensive insulin therapy; NIHSS, National Institutes of Health Stroke Scale; tPA, tissue-type plasminogen activator; HBA1C, diabetes mellitus; LDL, low-density lipoprotein; BMI, body mass index; BP, blood pressure; DWI, diffusion-weighted imaging; MCA, middle cerebral artery; ICA, internal carotid artery. *PϽ0.05 for between-group comparison.  Analysis of safety data found 32 versus 25 patients with serious adverse events in the SIT and IIT groups, respectively ( Table 2; Pϭ0.33). There was no significant difference in death at 3 months in the 2 groups (Pϭ0.36) with an OR for the IIT group of 0.6 (95% CI, 0.25-1.5). Serious adverse events including or not including death occurred in the same proportions in the 2 treatment arms. The most frequent serious adverse event was neurological worsening due to infarct extension or malignant edema. Symptomatic intracranial hemorrhage occurred in 7 patients in the IIT group and 6 in the SIT group. Asymptomatic intracranial hemorrhages, with no neurological worsening, were similar in the SIT and IIT groups (

Discussion
This trial has confirmed a better control of CGT by an intensive intravenous treatment with insulin versus a usual subcutaneous insulin protocol. The median delay of treatment initiation was Ͻ3 hours and efficient glucose control was quickly achieved in the IIT group. The IIT treatment was more effective than the SIT treatment in controlling capillary glucose level early after stroke with 95.4% of patients having a mean CGT Ͻ7 mmol/L during the 24 hours of treatment. Furthermore, efficient glucose control was observed in 86% of the patients at the first time point, which was 4 hours after the initiation of treatment. In addition, the intravenous salineinsulin regimen appeared to be safe despite more hypoglycemic episodes but not associated with more serious adverse events. Hypoglycemia occurred only in 5.7% (nϭ5) of the IIT group, a rate in the lower range of previously reported stroke IIT studies, 13,14,16,18 showing that this saline infusion insulin nomogram was at least as safe as glucose-potassiuminsulin regimens previously reported. 14 In addition, the study was performed in an intensive care unit with highly trained nurses, previously formed to other protocols using scaled therapies. However, hypoglycemic events were not retained in the logistic regression analysis predicting good functional outcome. There were no hypoglycemic episodes in the SIT group.
Despite the excellent efficiency-safety profile of the IIT regimen, the infarct growth was not smaller in the IIT group. This might be disappointing because animal 5,25 and human 6 -9 imaging data have shown that acute hyperglycemia increases infarct growth by exaggerating the transformation of penumbra into infarction and perhaps reperfusion injury. However, our results are consistent with those of McCormick and colleagues 18 who reported that IIT was not associated with reduced infarct growth but with an increased IG in patients with persistent occlusion in a small series of patients. In the present study, the largest infarct growth was also observed in the IIT persistent occlusion group. Furthermore, several studies have shown that preischemic insulin reduces IG in hyperglycemic animals, but there are very few reports on the effects of postischemic insulin treatment. 5,26,27 In one study 27 involving 12 rats, early death occurred in 5 animals and infarct size in the remaining rats was similar to those of the control groups. Similarly, 4 of 10 hyperglycemic cats receiving postinfarct insulin died from malignant edema and the remaining had the largest infarct size. 26 We have no explanation for this discrepancy between preclinical and clinical data but they clearly indicate the need for a reappraisal of the pathophysiological models of glucose energy metabolism alterations in the early phase of focal cerebral ischemia.
Despite the increased infarct growth in the IIT group, the functional outcome was similar in both groups perhaps because the study was not powered to detect clinical changes.
This study has some limitations due to its monocentric and open-label design. However, the patient sample was representative of the usual care of the population treated in our stroke center because 78% of eligible patients were included in the study. Although DWI was found highly correlated with relevant clinical outcome in previous studies, DWI volumetry only allows an approximation of initial and final infarct size. 10,28 Nevertheless, follow-up DWI volume performed between Day 1 and 3 could be also an advantage because it decreases lost at follow-up and dead at follow-up patients and it has been shown to be highly correlated with 30-days fluid-attenuated inversion recovery-determined volume. 29 The study was probably underpowered to detect a modest clinical change on the functional outcome although a favorable functional outcome in favor of IIT seems unlikely when considering the negative result found on DWI. Another and independent limit concerns that only few patients included in this study were diabetic (18% and 24%, respectively, in the IIT and SIT groups) and even few have very high glucose levels. This may limit the conclusions for this specific population for whom the poststroke hyperglycemia profile might be managed differently. 30

Summary
In conclusion, in the INSULINFARCT study, the working hypothesis was not confirmed; despite rapid and efficient control of poststroke hyperglycemia in acute stroke, intravenous insulin-controlled glucose did not reduce infarct growth but rather increased it. IIT cannot be recommended at the present time.