Abstract
Background:
Point-of-care (POC) testing devices for monitoring glucose and ketones can play a key role in the management of dysglycemia in hospitalized diabetes patients. The accuracy of glucose devices can be influenced by biochemical changes that commonly occur in critically ill hospital patients and by the medication prescribed. Little is known about the influence of these factors on ketone POC measurements. The aim of this study was to assess the analytical performance of POC hospital whole-blood glucose and ketone meters and the extent of glucose interference factors on the design and accuracy of ketone results.
Methods:
StatStrip glucose/ketone, Optium FreeStyle glucose/ketone, and Accu-Chek Performa glucose were also assessed and results compared to a central laboratory reference method. The analytical evaluation was performed according to Clinical and Laboratory Standards Institute (CLSI) protocols for precision, linearity, method comparison, and interference.
Results:
The interferences assessed included acetoacetate, acetaminophen, ascorbic acid, galactose, maltose, uric acid, and sodium. The accuracies of both Optium ketone and glucose measurements were significantly influenced by varying levels of hematocrit and ascorbic acid. StatStrip ketone and glucose measurements were unaffected by the interferences tested with exception of ascorbic acid, which reduced the higher level ketone value. The accuracy of Accu-Chek glucose measurements was affected by hematocrit, by ascorbic acid, and significantly by galactose. The method correlation assessment indicated differences between the meters in compliance to ISO 15197 and CLSI 12-A3 performance criteria.
Conclusions:
Combined POC glucose/ketone methods are now available. The use of these devices in a hospital setting requires careful consideration with regard to the selection of instruments not sensitive to hematocrit variation and presence of interfering substances.
Point-of-care testing (POCT) devices for monitoring glucose and ketones can play a key role in the management of dysglycemia in hospitalized diabetes patients. It is well recognized that POC monitoring of blood glucose levels in hospitalized acute and critical care patients is important for managing and maintaining normal glycemia to reduce patient morbidity and mortality and improve patient recovery.1-3 The use of handheld glucose meters for intermittent monitoring of patient glucose levels in hospitalized patients requiring intravenous insulin allows for rapid clinical decision making and immediate treatment responses. 4 Similarly the measurement of whole-blood ketone (beta hydroxybutyrate [BHB]) plays an important role in the management of adult and pediatric diabetes ketoacidosis (DKA). 5 Recent guidelines have advocated the use of whole-blood measurement of BHB for the management of DKA and for monitoring target directed therapy.6,7
Combined testing for glucose and BHB is often used in accident and emergency departments8,9 and can be an aid on ambulances and in a prehospital setting to identify and triage severely hyperglycemic patients. 10 A high level of accuracy is required for POC glucose and BHB devices to ensure that patients are managed optimally. However, it has been well reported that the accuracy of many hospital glucose meters can be affected by the biochemical and physiological changes that commonly occur in acute, chronic, and critically ill hospital patients and also by the medication prescribed.11,12 A wide range of endogenous and exogenous substances present in the blood of critically ill patients can donate electrons to nonspecifically initiate the electrochemical reaction and cause interference with glucose results, giving elevated values. 13 An increasingly large number of interfering factors have been associated with inaccurate glucose readings including abnormal hematocrit, oxygen tension and pH, uric acid, sodium, lipids, and administration of drugs (ascorbic acid, acetaminophen, dopamine). It is also well known that some of the enzymes used in glucose meters are not specific for glucose and can give falsely elevated results when other nonglucose sugars are present in patient samples. 14 POC meters for whole-blood ketone testing are based on the same design principle as commonly established glucose meter. Little has been reported on the influence of known glucose meter interferences on the accuracy of ketone meters.
The aim of this study was to undertake an analytical assessment of POC hospital whole-blood glucose and ketone meters to determine performance reliability and to determine the extent of glucose interference factors on the design and accuracy of ketone results.
Materials and Methods
Instrumentation
Two POC devices offering both glucose and ketone testing were assessed: StatStrip glucose/ketone (Nova Biomedical Waltham, MA, U.S.A) and Optium FreeStyle glucose/ketone (Abbott Diabetes Care Roma, Italy). In addition, the Accu-Chek Performa glucose only meter (Roche Diagnostic Milan, Italy) was also assessed. All 3 meter strip technologies represented the latest version provided for hospital use at the time of the study available. The StatStrip glucose strip technology is a modified glucose oxidase-based amperometric test system, and the ketone strip technology uses a BHB dehydrogenase enzyme based amperometric strip. Both strips incorporate hematocrit and other interference correction; Optium Freestyle uses an electrochemical glucose dehydrogenase/coenzyme nicotinamide adenine dinucleotide–based amperometric strip, and the ketone strip technology uses a BHB dehydrogenase enzyme based amperometric strip; Accu-Chek uses a glucose dehydrogenase/coenzyme pyrroloquinolone quinone–based amperometric strip recently modified to reduce the influence of maltose interference. The calibration of the StatStrip strips is predetermined by the manufacturer and requires no additional calibration code, whereas the Accu-Chek and Optium meters require the use of a calibration code for calibration of each strip lot. The ADVIA 2400 (Siemens Healthcare Diagnostics Milan, Italy) laboratory analyzer plasma hexokinase method was used as the reference method for glucose measurements. For BHB testing the Ranbut enzymatic reagents (Randox Laboratories Ltd London, UK) run on Uvikon 940 spectrophotometer (Tegimenta Basel, Switzerland) was used as the reference method.
Precision Testing
This was performed as according to Clinical and Laboratory Standards Institute (CLSI) guideline EP05-A2. 15 Within-day precision was assessed using quality control material specific for each meter supplier and with 3 whole-blood samples containing low, medium, and high levels of glucose or BHB. For glucose testing lithium heparin whole blood was collected from a volunteer and mixed continuously 18 to 24 hours before commencing the study. In this specimen, red blood cell glycolytic activity reduced the glucose concentration to less than 35 mg/dL. For BHB testing lithium heparin whole blood was collected from a volunteer just prior to performing the precision testing. The 3 different levels of samples were prepared by adding varying amounts of a concentrated glucose or BHB stock solution to aliquots of the donated sample. For within-day precision 20 replicates of each of the 3 blood sample levels and quality control samples were tested by each of the meters. Day-to-day precision was assessed according to the CLSI EP5-A2 protocol using quality control (QC) material specific to each meter. Each sample was measured in duplicate 2 times a day for 20 days on each meter and with the reference method.
Method Comparison
The comparison was performed according to CLSI guideline EP09-A3. 16 Forty lithium heparin samples were collected from hospitalized patients. Each sample was measured in duplicate on each POC meter and for each analyte. The order of testing was randomized, and testing was completed in 5 minutes to reduce the influence of glycolsis on the pattern of results. This approach was taken throughout the analytical studies undertaken. The remainder of each aliquot was immediately centrifuged to prepare a plasma sample for glucose and BHB measurement with the respective reference methods. To ensure glucose and BHB sample concentrations reflected the entire analytical range, higher level samples were prepared by spiking varying amounts of a concentrated glucose (20% wt/vol) solution or BHB (0.5 mol/L) solution to samples. Low-level glucose samples were obtained by leaving samples mixing for 12 to 24 hours.
Hematocrit Interference
A lithium heparin whole-blood specimen was collected, as previously described. Three glucose and BHB levels were prepared by adding the appropriate volume of a concentrated glucose (20% wt/vol) solution or BHB (0.5 mol/L) solution to achieve target glucose values of 20-60, 200-275, and 325-400 mg/dL and target ketone values of approximately 0.3-0.7, 2.2-2.8, and 3.7-4.3 mmol/L. For each glucose and BHB sample, 5 further aliquots were prepared and the hematocrit levels were adjusted after centrifugation and dilution to provide hematocrit levels of 28%, 38%, 48%, 59%, and 68% across each of the 3 concentrations. The actual hematocrit value for each of the aliquot preparations was confirmed using a StatSpin MP microhematocrit centrifuge (Iris Sample Processing, Westwood, MA). Each sample was tested 6 times by each strip meter system. The remainder of the sample was centrifuged immediately, and the plasma glucose and BHB level were tested by the respective reference methods.
Specificity Interference Studies
This was performed according to CLSI guideline EP07-A2. 17 Three different interfering substances (acetaminophen, ascorbic acid, and acetoacetate) were assessed for influence on the accuracy of BHB measurements, and 6 different interfering substances (acetaminophen, ascorbic acid, acetoacetate, galactose, maltose monohydrate, uric acid, and sodium) were assessed for influence on the accuracy of glucose measurements. Three samples at increasing concentration of BHB or glucose were prepared as described previously. For each concentration level a further 3 samples were prepared with increasing concentrations of each interfering substance. Each whole-blood sample was analyzed 6 times with each of the meters. The remainder of the aliquot was centrifuged, and the plasma glucose and BHB levels were assayed by the laboratory reference methods.
Linearity
This was performed according to CLSI guideline EP06-A. 18 For glucose linearity testing 11 samples were prepared by scalar dilutions of a high-concentrated sample with a low-level one. For BHB linearity testing 7 samples were prepared by adding varying amounts of a concentrated BHB (0.5 mol/L) stock solution to aliquots of a donated sample. Each sample was measured 10 times on each POCT meter.
Data Analysis
All calculations were performed on MS Excel. For an assessment of the accuracy of each glucose device, the percentage bias of each glucose meter result compared with the respective reference method result was calculated and assessed by comparison to performance criteria specified in ISO 15197 2013 19 and CLSI guideline POCT12-A3. 20
Results
Precision Testing
All ketone and glucose testing showed good within-day precision for whole-blood samples as well as for the manufacturers QC materials (Tables 1 and 2). The ketone and glucose meters also showed good day-to-day reproducibility (Tables 3 and 4). For glucose, the 3.3% desirable coefficient of variation (CV) according to biological variability was reached in most of the cases.
Within-Day Precision for Ketone POC Devices.
Within-Day Precision for Glucose POC Devices.
Day-to-Day Precision for POC Ketone Devices.
CV = overall imprecision; CVbd = between-day imprecision, CVbr = between-run imprecision, CVr = repeatability of duplicate; CVw = within-lab imprecision.
Day-to-Day Precision for POC Glucose Devices.
CV = overall imprecision; CVbd = between-day imprecision, CVbr = between-run imprecision, CVr = repeatability of duplicate; CVw = within-lab imprecision.
Method Comparison
For the comparison of BHB methods, 48 samples were tested and the mean plasma BHB value was 2.17 mmol/L (range 0.11-5.65 mmol/L). Both methods showed a good correlation to the laboratory reference method (Table 5 and Figure 1), but both also demonstrated a positive bias. In the case of StatStrip there was a significant constant bias of about 0.2 mmol/L (Figure 1A), while Optium showed a relevant proportional bias (Figure 1B).
Method Comparison for POC Ketone Methods and Concordance of Classification According to International Society for Pediatric and Adolescent Diabetes (ISPAD) Decision-Making Criteria.
Method comparison between reference method and beta hydroxybutyrate concentrations measured by StatStrip (A) and Optium (B) (device 1 closed symbols, device 2 open symbols).
For Optium the positive bias increased at higher levels of BHB, and this resulted in a higher level of discordance when assessing concordance of results to International Society for Pediatric and Adolescent Diabetes (ISPAD) decision-making criteria (Table 5). 7
For the comparison of glucose methods, 40 samples were tested and the mean plasma reference glucose value was 109 mg/dL (range, 28-398 mg/dL). All 3 POC glucose methods showed a good correlation to the reference method (Table 6). Although the method correlation wasn’t set up to assess accuracy in compliance to current standards and guidance, there was some differences seen between the meters in comparison to ISO 15197 19 and CLSI 12-A3 20 performance criteria (Table 6 and Figure 2).
Method Comparison and Accuracy Indication for POC Glucose Methods.
Bias distribution for glucose method correlation data: Difference of glycemia values between each POCT system (A StatStrip, B Optium, and C Accu-Chek) and the reference method plotted against the reference method, with solid lines representing ISO 15197 2013 and broken lines representing CLSI POCT 12-A3 limits.
Hematocrit Interference
The final adjusted hematocrit levels were confirmed to be 24%, 38%, 43%, 56%, and 66% for the BHB samples and 24%, 36%, 45%, 56%, and 62% for the glucose samples. For StatStrip over the hematocrit range of 38-56% there was a minimal effect on BHB measurements across the 3 levels tested, while at a hematocrit level of 24% the StatStrip BHB reading was 20% lower and at a hematocrit level of 66% the BHB reading was 12% higher compared to 43% hematocrit level (Figure 3A). For glucose there was a minimal effect of varying hematocrit levels over the range tested (Figure 3B)
Influence of varying levels of hematocrit on beta hydroxybutyrate (BHB) and glucose values. (A) BHB values measured with StatStrip and Optium for 3 different levels of BHB samples. (B) Glucose values measured with StatStrip, Optium, and Accu-Chek for 3 different levels of glucose samples.
The influence of hematocrit was pronounced for Optium device for both glucose and ketone measurements across all concentration levels tested (Figure 3A). For the glucose samples there is up to a 60% difference between the lowest and highest hematocrit levels, for the BHB samples up to 300%.
For Accu-Chek there was a minimal effect of varying hematocrit for the low-level glucose sample, but at higher glucose levels there was an influence on the accuracy of results, with up to a 15% difference between the lowest and highest hematocrit levels (Figure 3B).
Effect of Interfering Substances
Acetoacetate at concentration of 5 and 10 mmol/L had no influence on the 3 levels of BHB measurements for both StatStrip and Optium (Figure 4). Acetaminophen at concentrations of 0.33 and 0.66 mmol/L had only a slight negative influence on the 3 levels of BHB and no influence on glucose measurements for both StatStrip and Optium (Figures 4 and 5).
Influence of acetaminophen, acetoacetate, and ascorbic acid on the beta hydroxybutyrate (BHB) values measured with StatStrip and Optium for 3 different levels of BHB samples
Influence of acetaminophen, ascorbic acid, galactose, maltose, uric acid, and sodium on the glucose values measured with StatStrip, Optium, and Accu-Chek for 3 different levels of glucose.
The influence of ascorbic acid at a concentration of 0.29 and 0.59 mmol/L was pronounced on BHB measurements for both StatStrip and Optium, with falsely elevated BHB measurements for Optium and falsely reduced measurements for StatStrip (Figure 4). For the glucose samples only the lower level glucose was affected with Optium and Accu-Chek (up to 39% and 31% increase, respectively, as the ascorbic acid concentration increased) (Figure 5). Ascorbic acid had no influence on StatStrip glucose measurements.
For StatStrip and Optium none of the other interferences assessed (galactose 5.6 and 11.1 mmol/L, maltose 2.8 and 5.6 mmol/L, sodium 25 and 50 mmol/L, and uric acid 0.47 and 1.18 mmol/L) affected the accuracy of glucose measurements (Figure 5). For Accu-Chek maltose, sodium and uric acid at the concentrations tested had little effect on the accuracy of glucose measurements. However galactose had a significant effect across all 3 glucose levels with up to an 84% difference at low glucose levels and up to an 18% difference at higher levels (Figure 5).
Discussion
The benefit of using POC testing as an alternative to centralized laboratory testing is often debated, but it is recognized that the availability of faster results leads to more efficient diagnosis and treatment, particularly when managing critically ill patients or patients with dysglycemia.21,22
Whole-blood POCT for BHB is widely accepted as part of the management of diabetes type 1 patients with hyperglycemia and DKA. A number of guidelines now recommend the use of whole-blood BHB testing, 6 and recent UK Joint British Diabetes guidelines also advocate goal-directed treatment of adult DKA to lower BHB levels and follow-up BHB testing to monitor BHB levels. 7 Many handheld POC devices for markers such as BHB are often based on the design of POC glucose methods, as it is well recognized that the accuracy of whole-blood POC glucose methods can be influenced by a number of interfering factors that may be present in patients with critical or chronic diseases; 13 the same limitations may affect the accuracy of measurement results for the other markers. This analytical study was performed on 2 commercially available combined glucose/BHB POC devices (StatStrip and Optium Freestyle), 1 of which (StatStrip) corrected for interferences including hematocrit. The interference study data showed that the hematocrit interference correction of StatStrip did ensure that the accuracy of glucose and BHB measurements was unaffected across a wide hematocrit range. The influence of hematocrit on the accuracy of glucose measurements of Optium Freestyle has previously been reported.23,24 In this study the BHB device also was affected, resulting in the characteristic interference pattern seen with glucose measurements, that is, falsely elevated BHB values when hematocrit is low and vice versa (Figures 3A, 3B). Although the Accu-Chek strip technology has recently been modified, there is still evidence of an influence of hematocrit as previously reported. 24
The association of hematocrit interferences on glucose measurements directly leading to adverse risk in hospitalized patients managed as part of a tight glycemic control policy has recently been reported.25,26 In DKA patients it is not clear if the combined use of a POC Glucose/BHB method that exhibits hematocrit interference will lead to adverse decision making. A recent study indicated that venous blood POC error was not considered clinically significant in patients with DKA/HHS. 27 The authors concluded that there is a low risk of causing hypoglycemia from excess insulin administration based on POC glucose error but stated that patient harm could occur by delaying correction of metabolic disturbances in patients with DKA/HHS. Therefore further investigation is required to determine if POC BHB methods affected by hematocrit will influence the management of ketonemia particularly in patients following a goal directed targeted treatment for DKA. 7 A recent Canadian study 28 indicated that the range of hematocrit levels seen in patients managed in the primary care varied from 20% to 60% and from 10% to 70% in a hospital setting. Both primary care and hospital populations included patients with diabetes, and this reiterates the importance of considering hematocrit as a risk factor for both glucose and BHB monitoring.
As disease progresses in diabetes patients, they may become more exposed to hospitalization and increased uptake of medication. We looked at the influence of acetaminophen and ascorbic acid as medications that have been reported to adversely influence the accuracy of handheld POC glucose measurements. Acetaminophen at the concentrations assessed had no or very small influence, whereas ascorbic acid did. Intravenous ascorbic acid is used in the management of burns patients 29 and has been advocated for use in managing oncology patients 30 and also in diabetes patients on hemodialysis. 31 A recent report highlighted a spurious elevation of glucose measurements with a POC device during administration of high-dose ascorbic acid in a patient with type 2 diabetes on hemodialysis, which masked a hypoglycemic event in the patient. 31 In our study the Optium device exhibited falsely elevated readings for both BHB and glucose measurements with ascorbic acid (Figures 4 and 5), while only the BHB measurement was affected on StatStrip causing an underestimation of values (Figures 4 and 5).
The enzymes used in some handheld POC glucose devices have been reported not to be specific for glucose and shown to react in the presence of maltose and galactose. 14 Similarly it was recently reported that Optium BHB measurements were adversely influenced by acetoacetate, suggesting poor specificity of the BHB dehydrogenase enzyme used. 32 This was contradicted in this study as acetoacetate at the concentrations tested did not influence the accuracy of BHB measurements (Figure 4).
The Accu-Chek glucose strip technology has been recently modified to eliminate maltose interference. This study confirmed that the presence of maltose at the concentrations tested had no influence on the accuracy of results. As previously reported however, 33 the modification has not altered the specificity for other nonglucose sugars as glucose measurements were falsely elevated in the presence of galactose (Figure 5). The consequences of inaccurate readings in terms of managing neonates with galactosemia have previously been reported.34,35
This study focused primarily on assessing the current versions of combined glucose/BHB meter combinations with interferences reported to be associated with glucose meter adverse events. We did not assess other pathophysiological factors such as pH, pO2, and electrolytes in the in-laboratory specificity study as these are more appropriate to assess in patient clinical samples. StatStrip utilizes a modified glucose oxidase enzyme that potentially could be influenced by oxygen tension. This has previously been assessed in adult and neonatal critical care patients exhibiting a wide range of pO2 levels, and the accuracy of StatStrip glucose was unaffected.36,37 In an adult ICU study 36 it was reported that StatStrip achieved a much higher level of accuracy than the GDH-based meters across 4 partial pressure oxygen ranges. In the NICU study 37 abnormal levels of pH and electrolytes were also assessed and shown not to influence the accuracy of StatStrip. In a recent study it was reported that Accu-Chek POC glucose accuracy was influenced by acidemia, which can be present in DKA/HHS patients, 27 and as such further investigation into the role of fluctuating pH ranges on POC meter performance is required.
The study was not set up to assess the accuracy of the devices because the glucose measurements do not reflect the distribution of samples as outlined in new standards published since completion of the study. However an assessment of the correlation data in comparison to the new CLSI POCT 12 A3 20 performance criteria indicates that it will be challenging for some of the current POC glucose methods (Figure 2). At present there are no performance criteria established for BHB testing. Although StatStrip and Optium BHB methods correlated closely to the laboratory method, both methods showed a positive bias, even if of a different nature (constant for StatStrip, proportional for Optium) (Figure 1). If required, this can be compensated for with StatStrip as it allows a calibration adjustment to be implemented. The bias of Optium has previously been reported in a correlation assessment to a liquid chromatography–mass spectrometry BHB method. 38
The specificity deficiencies for some of the POC glucose methods evaluated in this analytical study parallel those of similar studies, and there is evidence emerging of the direct consequences of these deficiencies on clinical decision making.23,24 However, although this analytical study also demonstrated that hematocrit and ascorbic acid can influence the accuracy of BHB results, the clinical significance and relevance of this to patient decision making require further investigation. In addition, we assessed only a small number of known glucose interference factors, and further investigation of other known factors may help determine the extent of the inaccuracy problem.
Conclusions
In summary, combined POC glucose/BHB methods are now available to aid the management of dysglycemia in hospitalized diabetes patients. Their use in a hospital setting requires careful selection of instruments not sensitive to hematocrit variation and presence of interfering substances.
Footnotes
Acknowledgements
We thank Dr Andrei Malic for his valuable input in data evaluation and discussion.
Abbreviations
BHB, beta hydroxybutyrate; CLSI, Clinical and Laboratory Standards Institute; DKA, diabetes ketoacidosis; ISO, International Organization for Standardization; ISPAD, International Society for Pediatric and Adolescent Diabetes; POC, point of care; QC, quality control.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: FM, FL, and FV are full-time employees of A. Menarini Diagnostics.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by A. Menarini Diagnostics, Florence, Italy.