Performance evaluation of three i‐SENS glucometers using arterial blood samples compared with the YSI 2300 Glucose Analyzer

Abstract Background Capillary blood is the most commonly used sample for point‐of‐care (POC) glucometers. However, in critically ill patients, the glucose levels measured from capillary blood may not be reliable. Thus, we aimed to evaluate and compare the accuracy of glucose levels measured with POC glucometers and the YSI 2300 glucose analyzer using leftover arterial blood samples. Methods In total, 100 leftover heparinized arterial blood samples were used to evaluate the performance of three i‐SENS glucometers (BAROzen H Expert plus, CareSens PRO, and CareSens H Beat) and the ACCU‐CHEK® Inform II glucometer. The reference value was obtained using the YSI 2300 glucose analyzer. The results were analyzed based on International Organization for Standardization 15197:2013 guidelines. Results More than 95% of results obtained using POC glucometers were within ±15 mg/dL of the reference value for glucose concentrations <100 mg/dL and within ±15% of the reference value for glucose concentrations ≥100 mg/dL. In the consensus error grid analysis, more than 99% of results were found to be within zones A and B. An excellent correlation was found between the values obtained using POC glucometers and the YSI 2300 glucose analyzer (R 2 > .99). Conclusion The i‐SENS glucometers showed stable and accurate results when leftover arterial blood samples were used. Therefore, POC glucometers could be useful in critical care settings, such as intensive care units, where arterial samples are routinely used.


| Samples
We retrospectively evaluated 100 heparinized arterial blood samples obtained between January 2019 and March 2019. The samples were originally collected for arterial blood gas analysis (ABGA) from patients who visited the Korea University Guro Hospital. Specimens remaining after ABGA were used to evaluate the performance of three i-SENS glucometers, as well as that of the ACCU-CHEK ® Inform II glucometer and the YSI 2300 glucose analyzer. We excluded samples showing hemolysis or contamination and those with volume <350 µL or hematocrit levels out of 15-65%. All samples were handled anonymously and were assigned a new identification number.

This study was approved by the Institutional Review Board of Korea
University Guro Hospital (IRB No: 2019GR0016).

| i-SENS glucometers
Three commercially available i-SENS glucometers (BAROzen H Expert plus, CareSens PRO, and CareSens H Beat) from Korea were evaluated in this study. BAROzen H Expert plus is used in medical institutions, while the other two glucometers are used for SMBG in general diabetes healthcare. Apart from this, there are no significant differences in the methodology of these glucometers. All i-SENS glucometers use a GDH system, while the ACCU-CHEK ® Inform II glucometer uses mutant quinone GDH in order to prevent maltose interference.

| Study protocols
The samples examined in this study were allowed to rest at room temperature before measurement; analyses were performed within 30 minutes for each sample. Three different lots were prepared for each i-SENS glucometer, and one lot was prepared for ACCU-CHEK ® Inform II. Hematocrit levels were determined using a Sorvall Legend Micro 17 centrifuge (Thermo Fisher Scientific). Whole blood glucose levels were measured using the YSI 2300 glucose analyzer and converted to plasma glucose values using a preprogrammed algorithm that incorporated the given hematocrit level. 6 Using an arterial blood gas analyzer (ABL FLEX 700; Radiometer), the oxygen level in the samples was also measured, in order to exclude unsuitable specimens. Glucose levels were simultaneously measured using the three i-SENS glucometers (BAROzen H Expert plus, CareSens PRO, and CareSens H Beat) and ACCU-CHEK ® Inform II. Three different strips (lot A, B, and C) for i-SENS glucometers and one strip for ACCU-CHEK ® Inform II were inserted into the glucometer. Each strip loaded with 0.4 µL (0.5 uL for CareSens H Beat) of arterial blood and all measurements were performed twice. The difference between the reference value (from the YSI 2300 glucose analyzer) and the value obtained from each glucometer was calculated; Passing-Bablok regression analysis was used to determine the correlation between measured values. All statistical analyses were performed using Microsoft Excel 2016 (Microsoft, NY, USA) and Analyse-it (Analyse-it Software Ltd., The Tannery).

| RE SULTS
The distribution of blood glucose concentrations measured using the YSI 2300 glucose analyzer, as defined by ISO 15197:2013, is presented in Table 1 The correlation between the values obtained using the POC glucometers and the YSI 2300 glucose analyzer, including the slope, y-intercept, 95% confidence intervals (CI), and correlation coefficient (R 2 ) calculated using Passing-Bablok regression, are summarized in Table 3. All glucometers showed excellent correlation (R 2 > .95) with the reference value. Additional comparison study between the three i-SENS glucometers and ACCU-CHEK ® Inform II also showed an ex-

| D ISCUSS I ON
Our study was designed to confirm whether arterial blood was suitable for glucose measurements using commonly used blood glucometers. The laboratory measurements of blood glucose using serum or plasma are widely considered the "gold standard" but often fail to meet expectations, mainly due to time-related delays. In contrast, POC glucometers that require capillary blood-which can be obtained by pricking the patient's finger-can be used to quickly measure the patient's glucose levels. However, in some patients with severe anemia, metabolic acidosis, and hypoxia, the results obtained from capillary blood specimens may be inaccurate owing to various sources of interference. [8][9][10] Recently developed glucometers apply a unique algorithm developed by manufacturers to minimize hematocrit interference and offer stable results within a hematocrit range of 10-70%. 11 Nevertheless, it is recommended that arterial samples-rather than capillary blood-be used to test glucose levels in critically ill patients, and especially in patients with shock who are on vasopressor support. 12 A systematic review showed that in critically ill patients, the results obtained with arterial blood were significantly more accurate than those obtained with capillary blood when arterial blood gas analyzers and/or glucometers were used. 13 Arterial blood gas analysis is an essential and routine test performed to monitor oxygenation status and acid-base balance in ICU patients. 14 It may be beneficial to measure blood glucose levels using arterial blood samples that remain after routine ABGA. To ensure that POC glucometers provide accurate and stable results with arterial blood samples, a verification procedure is necessary. We used several types of glucometers to confirm the stability of leftover arterial blood samples in this study.

TA B L E 1 Distribution of blood glucose concentrations
As mentioned previously, hematocrit values are known to strongly affect measured plasma glucose levels. 8 Glucometers measure glucose in whole blood and correct the measured value using a specified formula under the assumption that the hematocrit level is normal. 8,15 Hence, in patients with anemia, hypoglycemia can go undetected due to false high glucose values and can lead to the administration of incorrect insulin dosages. 16 Therefore, ISO 15197:2013 recommends the use of test procedures and acceptance criteria to evaluate the influence of hematocrit, which can interfere with glucose measurement. i-SENS glucometers extend the hematocrit range, with no effect on glucose values, by applying specific algorithms that convert internal signals into measured values, similar to other recently developed glucometers that show no significant interference of hematocrit levels. 17 The most commonly used enzymes for measuring blood glucose levels are GDH, which is used in i-SENS glucometers, and GOD. GOD-based glucometers are prone to oxygen interference because oxygen is a physiological electron acceptor and is naturally affected by both low and high oxygen levels. In contrast, GDH is not affected by oxygen levels because oxygen is not involved in its electrochemical reaction. 18 In our study, we also measured oxygen levels in samples to identify any unsuitable specimens, but oxygen levels were found to have no significant effect on the results because all i-SENS glucometers used in the study were GDH-based.
There are some limitations to our study. First, the number of samples with low and high blood glucose concentrations was limited owing to challenges with sample collection. We could not gen-