This study was a post hoc exploratory study of blood samples that were obtained and analyzed from participants in a prospective randomized-controlled trial of intranasal insulin in surgical patients (ClinicalTrials.gov NCT02729064; registered on 5 April, 2016). The primary aim of the original trial was to study the effect of two intranasal doses of insulin (40 IU and 80 IU) on glycemic control during cardiac surgery requiring CPB.7 This sub-study was conducted at the Royal Victoria Hospital (McGill University Health Center MUHC, Montreal, Quebec, Canada) between 21 September 2016 and 14 March 2018.
All patients gave written informed consent, and the study was approved by our institutional ethics review board on 1 September, 2016. We enrolled patients (> 18 yr old) undergoing elective cardiac surgery requiring CPB. The exclusion criteria included procedures with anticipated deep hypothermic circulatory arrest, planned use of drugs that effect glycemia during the first two hours of surgery (e.g., insulin, steroids, epinephrine), allergy to insulin, acromegaly, Cushing’s syndrome, hyperthyroidism, pheochromocytoma, pregnancy, and a baseline blood glucose < 3.9 mMol·L−1.7
Before surgery, we recorded the patients’ age, sex, height, weight, body mass index, Euroscore-II,8 hematocrit, plasma creatinine, left ventricular ejection function, and comorbidities.
Oral hypoglycemic drugs were discontinued 12 hr before surgery, and a subcutaneous sliding scale insulin regimen was used in patients at risk of hyperglycemia (> 10 mMol·L−1) or hypoglycemia (< 4 mMol·L−1). Anesthetic care included the use of standard anesthesia monitors9 that were supplemented by central venous or pulmonary artery catheters and transesophageal echocardiography. Midazolam, propofol, sufentanil, sevoflurane, and a depolarizing or non-depolarizing muscle relaxant were given during induction and maintenance of anesthesia. The inhaled oxygen concentration was adjusted to 50%–100% based on the surgical procedure and the patients’ blood oxygen saturation. Normal saline solution and/or Ringer’s lactate were used as intravenous fluids. Systolic blood pressure (BP) was maintained at 100 mmHg before and after CPB with the mean arterial pressure during CPB maintained at 50–70 mmHg using norepinephrine (1–10 μg·kg−1·hr−1) as needed. Body temperature (BT) was maintained at 32–36° during CPB. Red blood cells were administrated when the hematocrit was < 25%. Before CPB, heparin 400 IU·kg−1 was administered intravenously followed by additional doses, if necessary, to maintain an activated clotting time (ACT) of > 480 sec. Protamine was administered on a 1:1 ratio after separation from CPB. Cardioplegia solution was free of glucose and mannitol and consisted of high-dose (100 mEq·L−1) potassium to induce and maintain cardiac arrest.
Blood glucose was measured every ten to 30 min during surgery to clarify whether intranasal insulin affected glucose values.7 When the blood glucose measured by BGA was < 4.0 mMol·L−1, 10ml of 20% dextrose with phosphate 30 mMol·L−1 was administered and a continuous dextrose infusion was started at 20 mL·hr−1. When the blood glucose concentration was > 10.0 mMol·L−1, an insulin infusion of 2 units·hr−1 was initiated to maintain the blood glucose between 4.0 and 10.0 mMol·L−1.
We also recorded the duration of anesthesia, surgery, CPB, and aortic cross clamping.
Glucose measurement protocol
Arterial whole blood samples were drawn from a catheter placed in either the radial, femoral, or brachial artery. Glucose concentrations were collected in a 3-mL syringe (without any additives) and analyzed by the StatStrip glucometer with additional blood collected in a 3-mL lithium heparin blood gas syringe (or a regular syringe during heparinization) for analysis in a GEM® Premier™ 3000® (Instrumentation Laboratory Company, Bedford, MA, USA) blood gas analyzer. Blood glucose analyses were performed within two minutes of blood collection.
Samples were collected at five time-points: baseline (before surgery), pre-CPB (before CPB after heparinization), during early CPB (30 min after establishing full CPB; CPB1), during late CPB (one to two hours after establishing full CPB; CPB2), and post-CPB (30 min after separation of CPB and the administration of protamine).
The following data were recorded at the five above time-points: arterial pH, pO2, pCO2, lactate, hematocrit, ACT, (BT, nasal, and bladder), BP (systolic, diastolic, and mean), heart rate (HR), and catecholamine infusion rate.
The primary outcome was the accuracy of blood glucose values obtained by the StatStrip glucometer, which was defined by the criteria specified by the POCT12-A3 guideline4 of the CLSI and/or criteria established by the ISO 15197:2013.6
The two CLSI POCT12-A3 criteria, both of which needed to be met, were as follows4: criterion 1: 95% of samples should be within ± 0.66 mMol·L−1 of reference glucose values < 5.5 mMol·L−1 and ± 12.5% for reference glucose values > 5.5 mMol·L−1; and criterion 2: 98% of samples should be within ± 0.83 mMol·L−1 of reference glucose values < 4.1 mMol·L−1 or 20% of the reference glucose for values > 4.1 mMol·L−1 (Table 4) (Fig. 2). We estimated that a minimum of 100 samples would be needed to determine the accuracy. When 98 of 100 samples meet the criterion, the 95% confidential interval (CI) is 93 to 100%, and when 95 of 100 samples meet the criterion, the 95% CI is 89 to 98% consistent with the ISO 15197:2013.6 Clinical accuracy was acceptable when 99% of samples were within zones A and B on the Parkes error grid.5,10 The Parkes error grid is divided into five risk zones.5 The five risk zones were illustrated in Figure 3 and described in Table 1. In agreement with previous protocols, arterial blood glucose values measured by BGA were considered the reference.11,12
When samples did not meet the above criteria, an additional post hoc exploratory analysis was performed to determine whether clinical variables, laboratory or hemodynamic parameters, and heparinization were associated with the discrepancy of glucose values obtained by the two methods and the accuracy of the StatStrip glucose measurements.
Preoperative, surgical, and laboratory data are summarized using descriptive statistics. Categorical variables are described as counts and percentages. All data were tested for normality using the Kolmogorov–Smirnov test. Continuous variables with normal distribution are presented as mean (standard deviation) and variables with skewed distribution as median [interquartile range (IQR)].
The differences between laboratory and hemodynamic data at five time-points were compared using the Friedman test. Patients with missing data were excluded.
Accuracy was analyzed using the criteria defined by the CLSI POCT12-A3 guideline4 for hospital POCT glucometers and by ISO 15197:2013 Parkes error grid for self-monitoring blood glucose glucometers used for patients with type 1 diabetes mellitus.
The Spearman’s rank correlation was used to determine whether BT, laboratory (arterial pH, pO2, pCO2, lactate, Hematocrit, ACT) and hemodynamic parameters (mean BP, catecholamine infusion rate) were related to the difference of glucose values obtained by the two methods. The Mann–Whitney U test was used to determine whether blood transfusion and heparinization had an influence on the absolute difference of glucose values obtained by the two methods.
A logistic regression model was used for all samples to assess the relationship between accuracy defined by the CLSI POCT12-A3 criteria—i.e., whether each point-of-care glucose value is within ± 0.66 mMol·L−1 of reference glucose values < 5.5 mMol·L−1 and ± 12.5% for reference glucose values > 5.5 mMol·L−1 (criterion 1 in CLSI POCT12-A3) and possible factors. Considering the Spearman’s rank correlation and the Mann–Whitney U test, possible factors were put into the logistic regression model. Because multi-colinearity among covariates can give spurious results, backward stepwise procedures were performed to identify independently associated variables, and odds ratios were calculated.
When significant continuous variables were detected, the areas under the receiver operating characteristic (ROC) curves were calculated. The area under the curve (AUC) is a measure of the parameter’s accuracy (AUC = 0.5, no better than chance and no prediction possible; AUC = 1.0, best possible prediction). The optimal hematocrit cut-off point for StatStrip glucometer usage was defined by the Youden index [maximum (sensitivity + specificity − 1)].
According to the CLSI POCT12-A3 guidelines, a minimum of 100 samples are needed to determine accuracy in a given patient population. Assuming that approximately 20% of patients would have missing data, we intended to analyze a minimum of 120 patients. To evaluate whether the data set was suitable for accurately characterizing the relationship between two measurement procedures, we calculated the Spearman correlation after data collection.4
All tests used for statistical analysis were two-sided and P values < 0.05 were considered statistically significant. Data were analyzed in SPSS version 23 (SPSS Inc, Chicago, IL, USA).