This is a randomized, open-label, pilot clinical trial. Patients with acute stroke admitted to the stroke ICU of National Taiwan University Hospital within 72 h of onset (onset defined as the last known time at which the patient was well) were screened. The entry criteria for admittance to the stroke ICU included ischemic stroke with thrombolytic therapy or endovascular treatment, intracerebral hemorrhage with aggressive blood pressure control treatment, severe neurological deficits (e.g., a National Institute Health Stroke Scale score higher than 8), stroke-in-evolution, or other medical conditions requiring intensive care.[20, 21] Patients who met the inclusion criteria (age ≥ 20 years, capillary blood glucose > 200 mg/dL after admission) were enrolled. Patients were excluded if they had any of the following conditions: autoimmune disease, human immunodeficiency virus infection, sepsis, pregnancy, treatment with corticosteroids or vasopressors, end-stage renal disease requiring dialysis, type I diabetes mellitus, or hypersensitivity to any insulin products.
The diagnosis of stroke was confirmed through diffusion-weighted magnetic resonance imaging of the head or repeated computed tomography scanning (if the first scan did not clearly reveal the infarct region). Patients with ischemic stroke were further classified into 5 major subtypes according to the Trial of ORG 10172 in Acute Stroke Treatment criteria: large-artery atherosclerosis, small-vessel occlusion, cardioembolism, specific etiology, and undetermined etiology.[22]
A detailed history of each patient’s clinical presentation, vascular risk factors, and comorbidity was obtained. Body mass index was calculated as weight divided by the square of height in meters. Stroke severity on admission was assessed using the National Institute Health Stroke Scale. Mortality and functional outcome 3 months poststroke were assessed using the modified Rankin scale. Acute stroke care for both groups followed the latest guidelines from the American Heart Association and the American Stroke Association.
Randomization
Patients were assigned to the IG group or the NPH group through computer-generated randomization. Allocation was concealed by the use of opaque, consecutively numbered envelopes. Treatment allocation was 1:1.
Insulin Regimens
All participants received only insulin for glycemic control during the intervention period, which was the first 72 h after study enrollment. Administration of oral hypoglycemic agents and noninsulin injections, regardless of whether they were used before stroke, was stopped. The insulin regimens in both groups are described as follows. A basal–bolus regimen of IG or NPH was used for glycemic control. All patients received regular insulin before meals as a correctional component. During the period when food was prohibited (nothing-by-mouth status), IG was administered because it does not demonstrate a pronounced peak. NPH was withheld to prevent hypoglycemia.
The total daily insulin dose (TDD) was based on weight. For most patients, the TDD was 0.6 U/kg/day. For patients with anticipated insulin sensitivity, such as those aged over 80 years or those whose creatinine clearance rate was below 30 mL/min, the TDD was empirically reduced to 0.5 U/kg/day. The TDD of patients receiving insulin therapy before admission was based on their outpatient dose and modified by an endocrinologist. For the IG group, 50% of the TDD was basal insulin and the other 50% was evenly administered before each meal. For the NPH group, 25% of the TDD was basal insulin, and similarly, the remaining 75% was administered evenly before meals. After the regimens were administered, an endocrinologist carefully assessed glycemic levels and made daily adjustments to achieve glycemic levels between 80 and 180 mg/dL. Scheduled prandial or basal insulin administration was temporarily stopped during treatment of hypoglycemia (defined as blood glucose < 60 mg/dL), which was performed in accordance with a previously published guideline.[23]
Blood Glucose Measurements
Participants’ capillary blood glucose was measured before each meal and once during fasting, at the frequency of every 4 to 6 h, depending on their eating schedule. In addition, a Medtronic Enlite Glucose Sensor (Medtronic, Northridge, CA, USA) was used to monitor glycemia during the intervention period.[24] Interstitial glucose levels were recorded every 5 min with a detection range of 40 to 400 mg/dL. Glucose levels from capillary blood and continuous glucose monitoring (CGM) were compared and calibrated at randomization and 1, 3, and 4 h after randomization. The glycemic data from CGM were used in the analyses.
Study Outcomes
The primary efficacy outcomes were the percentage of glucose levels within the range of 80 to 180 mg/dL and the percentage of glucose reduction compared with prerandomization levels during the 72-h intervention period. The primary safety outcome was the percentage of time with hypoglycemia (glucose < 60 mg/dL). The secondary clinical outcomes were determined according to whether a favorable outcome was achieved (modified Rankin scale score ≤ 2), whether a poor outcome was achieved (modified Rankin scale score ≥ 4), mortality, and Barthel index scores at 3 months poststroke. The secondary laboratory outcomes were glycemic mean and variability, determined from CGM data. Glycemic variability was calculated through linear and nonlinear analyses using previously described methods. We computed two-time domain measures as standard deviations and the root mean square of successive beat-to-beat differences. The nonlinear analysis of sample entropy was also applied to the glucose data using methods described previously.[20, 25, 26]
Statistical Analysis
Sample size calculation was not directly applicable in this trial because the expected effect size of our primary outcome, namely the percentage of time before return to normoglycemia, was yet unknown. Instead, we followed a previously proposed method of enrolling at least 50 participants in a pilot study.[27] The analysis was performed as intention-to-treat to reflect clinical practice in stroke ICUs. Descriptive statistics of baseline characteristics of the study population are expressed as numbers (percentages) and means ± standard deviations.
Because the primary efficacy and safety outcomes were nonnormally distributed quantitative variables, the Mann–Whitney U test was used to determine whether the between-group differences in mean percentage of time were significant. Between-group mean differences and 95% confidence intervals of the primary outcomes were also analyzed. The chi-square test or the Fisher’s exact test were applied for categorical secondary outcomes, whereas the Mann–Whitney U test was applied for the continuous variables. In addition, factors associated with a favorable outcome (modified Rankin scale score ≤ 2) were examined through univariate analysis. A logistic regression method was used to adjust for age and sex and univariate analysis of was conducted to identify factors that were significant predictors (P < .05) of a favorable outcome. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, NY, USA). Significance was defined as P < .05.
Ethics
This study was approved by the Institutional Review Board and the Medical Ethics Committee of National Taiwan University Hospital (Clinicaltrials.gov number: NCT02607943). All participants provided informed consent or were recruited with consent provided by their first-degree relatives. The study was conducted in accordance with the revised version of the 2008 Declaration of Helsinki.