Introduction
Critical illness is often accompanied by acute hyperglycaemia, which appears to be a negative prognostic factor for critically ill patients [
1‐
4]. Worse outcome in hyperglycaemic patients has been explained mainly by complications such as infections and decreased wound healing [
5]. Moreover, acute hyperglycaemia has a direct effect on the culprit organ(s). A factor that may also contribute to bad prognosis is still the apparent lack of treatment of acute hyperglycaemia.
Recent research suggests that glycaemic control in specific groups of patients reduces morbidity and even mortality. Van den Berghe
et al. [
6] showed that blood glucose regulation toward normoglycaemia decreased mortality and morbidity in a cardiosurgical intensive care unit (ICU) population. Krinsley [
7] found a beneficial effect of glucose regulation on mortality in a study using a historical control group. However, in a retrospective study of 1,085 consecutive mixed ICU patients (mortality 20%), using a multivariate analysis model, hyperglycaemia was not an independent factor contributing to mortality [
8]. In a recent prospective study in a medical ICU population, van den Berghe
et al. [
9] found that reduced blood glucose levels did not significantly reduce in-hospital mortality for the whole group, only for the subgroup of patients with an ICU stay of 3 days or more. So, the available incidence is inconsistent in regard to the beneficial effects of normoglycaemia. Further progress in establishing the role of glycaemic control in critically ill patients certainly requires an excellent glucose measuring and reporting system. It should give fast and reliable results that can be used in a nurse-driven insulin infusion algorithm that reduces hyperglycaemia without inducing hypoglycaemia [
10,
11].
A central hospital laboratory delivers glucose values after approximately 30 to 60 minutes, which is too slow to use in a glucose regulation protocol. Literature on point-of-care testing reveals varying accuracy of different handheld meters [
12]. The ABL blood gas/glucose analyser used in the study of van den Berghe
et al. [
6], to our knowledge, has not been validated in a critically ill ICU population. Before implementation of a strict glucose protocol in our ICU, we chose to evaluate accuracy and feasibility of the ICU-based blood gas/glucose analyser ABL715; afterward, we tested the reliability of a contemporary handheld point-of-care meter using test strips (Precision PCx; Abbott Diabetes Care, Amersfoort, The Netherlands) and a subcutaneous continuous glucose monitoring system (CGMS) (CGMS System Gold; Medtronic MiniMed, Inc., Northridge, CA, USA).
Materials and methods
Patients with an expected ICU stay of more than 48 hours were included. Data were collected prospectively. No formal glucose regulation protocol was used; as a rule, continuous insulin infusion therapy was initiated at blood glucose levels exceeding 9 to 10 mmol/l (162 to 180 mg/dl) with a blood glucose target between 6 and 8 mmol/l (at the discretion of the attending physician or nurse). We first validated the ICU-based blood gas/blood glucose analyser ABL715 against the central hospital laboratory (part 1 of the study). Heparinised arterial blood samples were drawn from an arterial line every 6 hours by the same nurse practitioner. One part was sent in a vacuum-sealed plasma separation tube to the central hospital laboratory and analysed by a laboratory technologist using a conventional blood glucose analyser (YSI 2300; Yellow Springs Instruments, Yellow Springs, OH, USA). The YSI analyser uses glucose oxidase to measure the glucose concentration in whole blood (by generation of hydrogen peroxide and gluconic acid from glucose and oxygen via the enzyme glucose oxidase). The other part was immediately analysed on the blood gas/blood glucose analyser ABL715.
Next, part 2 was performed. After inserting (and calibrating) the subcutaneous CGMS System Gold, heparinised arterial blood samples were drawn from an arterial line every 6 hours and analysed by both the Precision PCx point-of-care meter using test strips and the blood gas/glucose analyser ABL715. CGMS glucose data were downloaded after 24 to 72 hours. The continuous glucose monitoring system sensor was inserted in the abdominal subcutis and calibrated every 6 hours according to the manufacturer's indications (using the ABL715 blood glucose values). ICU nurses inspected the insertion site twice daily for signs of infection or bleeding. Data of the CGMS devices were downloaded using the MiniMed Com-Station® and CGMS Solutions™ software (Medtronic MiniMed, Inc.). The institutional medical ethical committee was informed and agreed with the study protocol.
ABL715
The ABL715 Series (Radiometer Medical ApS, Brønshøj, Denmark) can be set up to measure pH and blood gas and any combination of oximetry, electrolyte, and metabolite parameters. Glucose concentration is measured by the glucose-oxidase method from 95 μl of whole blood to provide results within 2 minutes. Maintenance, calibration, and quality control are performed on a regular basis by the central hospital laboratory. The ABL715 has the capability of cross-linking between the hospital and laboratory information systems, making manual data entry into medical records superfluous. The network connectivity reduces the workload by automating data collection.
Precision PCx
The Precision PCx point-of-care system (Abbott Diabetes Care) is a portable, glucose oxidase-based, whole-blood testing system using test strips, requiring 3.5 μl of whole blood to provide results within 20 seconds. Results of 4,000 tests can be stored in the monitor itself. The monitor also features a data port and a docking station to automatically download these results in the laboratory or hospital information systems. According to the manufacturer, glucose levels can be measured between 1.1 and 75 mmol/l (20 to 600 mg/dl).
Continuous Glucose Monitoring System
The CGMS System Gold features a sensor that can be used for up to 72 hours. The sensor consists of a flexible, platinum-plated electrode residing inside a permeable membrane and is inserted in the subcutaneous tissue just under the skin of the abdomen. Subcutaneous glucose is measured by the glucose-oxidase method, and every 10 seconds an interstitial glucose measurement is sent to a monitor that records an average glucose value every 5 minutes. Sensor blood glucose values are calculated using MiniMed Solutions® version 3.0 software. The manufacturer recommends at least four calibrations a day for the CGMS System Gold. Monitor data can be downloaded onto a spreadsheet using the MiniMed Com-Station® and software. The CGMS calculates blood glucose values in the range of 2.2 to 22.2 mmol/l (40 to 400 mg/dl).
Statistical analysis
Results of paired measurements (laboratory versus ABL715, CGMS versus ABL715, and Precision PCx versus ABL715) were analysed in three ways. Paired values were plotted on a Bland-Altman plot, which is a plot of differences expressed as a percentage of averages between the two methods [
13].
The Pearson correlation coefficient (
r) between the different methods was determined by linear regression. Finally, error-grid analysis was performed to obtain information of clinical relevance from the measurements [
14]. Results are plotted in zones of different significance: points in zone A have no clinical implications (clinically accurate measurement), and points in zone B lead to an appropriate clinical decision. Only points lying in zones C, D, and E would lead to inappropriate intervention or lack of intervention. Zone C means misinterpretation of euglycaemia for hyper- or hypoglycaemia (unnecessary overcorrection is possible). Points in zones D and E mean overestimation of hypoglycaemia or underestimation of hyperglycaemia, which may lead to dangerous treatment.
Discussion
In this study on seriously ill ICU patients, the blood gas/glucose analyser ABL715 appears to be a good alternative for laboratory testing and can be used for fast and reliable glucose measurements in the ICU. The glucose value is available to the ICU nurse within 2 minutes and is immediately visible in the hospital information system. The ABL715 requires a very small amount of blood; the chance of spilling blood from the syringe or the device is low.
The nursing staff gave the device a high score in terms of user-friendliness and feasibility. As a minor point of criticism, it was mentioned that one ABL715 at a 12-bed ICU occasionally causes a short waiting time. The blood gas/glucose analyser ABL715 systematically gives values approximately 18% higher than the laboratory, although the correlation is very good. This is important when deciding which blood glucose target one aims for in a tight glucose regulation protocol (laboratory value of 6 mmol/l = ABL715 value of approximately 7 mmol/l). When subsequently using this analyser in the ICU, this difference is of minor importance.
The CGMS System Gold also showed a fair correlation. Subcutaneous placing of the sensor's needle was judged to be easy by the research nurse who performed the insertions during this study. The device has to be calibrated four times a day, which requires the availability of another fast and validated blood glucose analyser. This type of CGMS is not yet useful for a glucose regulation protocol, because it provides no real-time glucose data. (They have to be downloaded every 24 to 72 hours.) Soon, newer types will be available which will deliver directly visible 'online' glucose data every 5 minutes.
The Precision PCx point-of-care glucometer showed a fair correlation coefficient, was judged to be user-friendly (in spite of the use of test strips and the higher chance of spilling blood), and delivers fast results.
The turnaround time for glucose determinations by the central hospital laboratory was 30 to 60 minutes, which is too long for fast adjustment of the insulin infusion dose or other therapeutic actions as required in a glucose regulation protocol and might cause excess hypoglycaemia.
Current literature on feasibility of bedside glucometry in critically ill patients is sparse. In a recent hyperinsulinaemic-euglycaemic clamp study in 16 patients with diabetes, Clarke
et al. [
15] found good accuracy for the subcutaneous CGMS in the euglycaemic range: 93.7% of the readings were in error grid zones A and B. In the hypoglycaemic zone (250 blood glucose measurements <3.7 mmol/l [<70 mg/dl]), however, they found the CGMS to be less accurate, with only 62.8% in zones A and B. Goldberg
et al. [
16] also investigated the accuracy of the subcutaneous CGMS System Gold in 21 critically ill patients admitted to a medical ICU and found a Pearson correlation coefficient of 0.88. Clarke error grid analysis categorised 98.7% within the clinically acceptable zones A and B. Only four of more than 500 paired readings were in the hypoglycaemic range in this study [
16].
Tang
et al. [
17] tested six handheld glucose meters and a portable glucose analyser. At a glucose concentration of 4.8 mmol/l (86.4 mg/dl) and at a pH between 6.94 and 7.84, glucometry remained accurate. However, at a higher glucose level of 11.2 mmol/l (200 mg/dl), such extreme pH conditions significantly lowered its precision [
17]. In another study, bedside reflectance glucometry in 105 arterial blood samples of 10 critically ill adults showed a correlation coefficient of 0.86 compared with the central laboratory [
18]. An early report described that fingerstick glucometry was inaccurate among patients in shock [
19]. Louie
et al. [
12] compared two point-of-care meters using arterial blood samples obtained from 247 critically ill patients and found that 98% to 100% of the SureStepPro and 91% to 95% of the Precision G measurements fell within an error tolerance of ± 15% versus the hospital chemistry analyser. In a study by Maser
et al. [
20] at a cardiovascular ICU, capillary whole-blood glucometry had a correlation coefficient of 0.88 with arterial plasma samples analysed in the hospital laboratory. In the last study, capillary whole-blood glucose levels were a mean of 0.5 mmol/l (9 mg/dl) lower than arterial plasma specimens.
We found, despite the good correlation coefficient, that glucose values measured by the blood gas/glucose analyser ABL715 were constantly higher than laboratory values and that glucose concentrations measured by the Precision PCx and the CGMS System Gold tended to be lower than the ABL715 values. It is therefore important to realise that defining targets for glucose regulation (for instance, from 4.4 to 6.1 mmol/l [79.2 to 109.8 mg/dl] [
6]) also depends on the specific analytic method used at a particular ICU.
It could be questioned which error tolerances are acceptable in a critical care setting. Error tolerance criteria proposed by Kost
et al. [
21] are (a) within ± 0.8 mmol/l (15 mg/dl) of the reference measurement for glucose levels ≤5.6 mmol/l (100 mg/dl) and (b) within approximately 15% of the reference measurement for glucose levels greater than 5.6 mmol/l (100 mg/dl). The ISO (International Organization for Standardization) criteria are as follows: if the reference value is greater than 4.1 mmol/l, the sensor reading should be within 20%; if the reference value is less than 4.1 mmol/l, the sensor reading should be within ± 0.8 mmol/l.
Another way to define the error tolerance and to obtain information of clinical relevance from the measurements is to calculate the Clarke error grid for the glucometer that is evaluated [
14]. Defined in this way, all three tested glucometers would be acceptable for use in critically ill patients. We are not completely sure whether this is also true in the hypoglycaemic range, because we did not have many measurements in this range.
Introducing a glucose regulation protocol requires a fast and accurate way to measure blood glucose levels. Because implementing such a protocol would increase nursing workload, it has to be feasible as well. This means that, in all likelihood, not the most accurate, but the most feasible and nevertheless still fairly accurate, glucose analyser would be chosen. On the other hand, the risk of hypoglycaemic episodes could increase. In the sedated, critically ill patient, hypoglycaemic warning symptoms are absent. This means that, apart from safety rules in the protocol, the glucose analyser has to be very reliable in the low range too.
We deliberately choose seriously ill ICU patients, as shown in the patient characteristics, because we aimed to test the reliability of the various analysers under conditions such as shock, vasopressor use, edema, sepsis, and renal replacement therapy. Our study has too few patients and therefore too little data points under extreme conditions of pH, temperature, electrolyte disturbances, and hypoglycaemia to make statements about the reliability of specific analysers under these circumstances.
Competing interests
ICCH serves as a scientific consultant for the continuous glucose measurement research program of Medtronic Europe.
Authors' contributions
ICCH and JJML initiated and planned the study. AMC, RS, and JJML conducted the study, collected data, and drafted the manuscript. JSWL, JGZ, JET, ICCH, and JHJMM assisted in writing the manuscript. All authors read and approved the final manuscript.