Correlation of capillary and venous blood glucometry with laboratory determination

ORIGINAL CONTRIBUTIONS CORRELATION

OF

CAPILLARY AND VENOUS BLOOD GLUCOMETRY LABORATORY DETERMINATION

WITH

CPT Gautam Kumar, MBBS, CPT Ban Leong Sng MBBS, COL Surya Kumar, MBBS, MSS, MMed Cap and Lab results. Conclusion. Capillary whole-blood glucose values best approximated venous plasma glucose values from the laboratory. Measuring the venous wholeblood glucose using the glucometer resulted in an overestimation of the venous plasma glucose compared with the laboratory result by about 0.97 mmol/L (17.46 mg/dL). This may result in the withholding of intravenous glucose for patients who are actually hypoglycemic. Key words: blood glucose analysis; emergency treatment methods; hypoglycemia; diagnosis; Singapore.

ABSTRACT Background. During resuscitation in the Singapore Armed Forces, blood glucose samples are often obtained for analysis using the capillary glucometer. A drop of venous blood from the intravenous cannula is sometimes used to ascertain the patient’s blood glucose level. Venous samples may be sent to a commercial laboratory, but this does not allow immediate results. Objective. To establish the correlation between the glucose levels of the capillary fingerprick sample analyzed by the glucometer (Cap), the venous sample analyzed by the glucometer (Ven), and the venous sample tested by the commercial laboratory (Lab). Methods. This multicenter, prospective study enrolled subjects from Selarang, Clementi, and Maju Camp Medical Centers in the Singapore Armed Forces from November 2002 to January 2003. All subjects provided at least two samples; because provision of the capillary blood glucose sample was voluntary, only some gave consent for capillary fingerprick and provided three samples. There were no exclusion criteria. Bland-Altman plots were then constructed to compare the capillary and venous-on-capillary values with the laboratory results. Results. A total of 270 subjects were recruited. One hundred seventy subjects (63.0%) gave consent for capillary glucose measurement and, thus, had three readings for comparison. There was a mean difference of 0.97 mmol/L (17.46 mg/dL) between the Ven and Lab results. There was an insignificant mean difference of 0.01 mmol/L (0.18 mg/dL) between the

PREHOSPITAL EMERGENCY CARE 2004;8:378–383

During evacuation and resuscitation in the Singapore Armed Forces (SAF), a blood glucose sample is almost always taken for analysis. The device commonly used, the glucometer, is designed for capillary blood samples. The portability and convenience of this machine make it an ideal choice for ambulatory use in the clinic, the field, the ambulance, and other out-of-hospital (OOH) settings. In addition, our medical centers do not have on-site laboratory facilities to provide instant determination of blood glucose values. In the interest of expediency and minimizing pain to the patient, venous blood from the intravenous cannula or from elective venipuncture is often used for the determination of the blood glucose level on glucometers designed for capillary blood samples. We have witnessed this practice in the hospital environment, in SAF medical centers, and in OOH settings. The clinical significance of this practice and its impact on patient management have not been adequately studied. A small study involving 97 healthy volunteers reported a relatively poor correlation between venous and capillary blood glucose measurements.1 Our study was designed to further elucidate this topic. The objective of our study was to establish the correlations and differences between the glucose levels measured from three arms: the capillary fingerprick whole-blood glucose level analyzed by the glucometer (Cap), the

Received July 24, 2003, from the Singapore Armed Forces Medical Corps, Singapore (GK, BLS, SK). Dr. G. Kumar is currently in the Department of Emergency Medicine, Singapore General Hospital, Singapore, and Dr. Sng is currently in the Department of Anesthesia, Tan Tock Seng Hospital, Singapore. Revisions received January 19, 2004, and June 3, 2004; accepted for publication June 3, 2004. Presented at the 4th Singapore Armed Forces Military Medicine Conference, Singapore, February 8, 2003. Address correspondence and reprint requests to: CPT Gautam Kumar, MBBS, C/O Headquarters, Army Medical Services, AFPN 0041 #03-02, 701 Transit Road, Singapore 778910. e-mail: . doi:10.1016/j.prehos.2004.06.010

378

Kumar et al. CAPILLARY/VENOUS BLOOD GLUCOMETRY

VS.

379

LABORATORY VALUES

venous whole-blood glucose level analyzed by the glucometer (Ven), and the venous plasma glucose level analyzed by the commercial laboratory to which we routinely send our samples (Lab).

METHODS Study Design This study involved three medical centers in the SAF—Clementi, Maju, and Selarang Camps. It was a prospective, nonrandomized comparison study with two or three arms. The protocol was approved by the Joint Medical Conference (Research) in November 2002. The enrollment period was from November 2002 to January 2003.

Subjects In the SAF, operationally ready national servicemen are subjected to health screening for coronary risk factors and diabetes at the age of 35 years. Hence, they report to the respective medical centers for venous fasting blood glucose and lipid samples, in addition to other tests. Thus, to acquire a capillary blood sample, an extra fingerprick is needed. Informed consent for this extra fingerprick was obtained.

Equipment The Bayer Elite glucometer (Bayer Corporation, Elkhart, IN) was used to determine venous wholeblood and capillary whole-blood glucose levels in all three medical centers. This device uses a glucose oxidase sensor coupled to amperometric measurement. Previous studies have established the reliability, accuracy, and precision of this glucometer.2–4 All three glucometers used had been calibrated monthly according to our internal quality audit (IQA) requirements.

Samples Capillary whole-blood samples were obtained by fingerprick using a disposable spring-loaded lancet and were tested using the glucometer. Venous whole blood was obtained by elective venipuncture at the cubital fossa using the conventional syringe-and-needle technique. A drop of this blood was put on the glucose strip to test the venous whole-blood glucose level. The rest of the blood was injected into fluoride tubes to inhibit glycolysis. After collection, venous blood samples were stored in a refrigerator at 28C to 88C while awaiting transport.

Laboratory Services Since medical centers in the SAF do not have on-site laboratory services, all blood samples were sent to Quest Laboratories Pte Ltd. There was a maximal lag time of about ten hours from the time of collection to the time at which the samples were processed in the laboratory. The laboratory uses the enzymatic reference method with hexokinase to assay the plasma glucose level.5,6

Data Management A Microsoft Access (Microsoft Corp., Redmond, WA) database was created for this study to store and process raw data. Subsequently, the database was converted to SPSS for Windows, release 10 (SPSS Inc., Chicago, IL), for data analysis. In the analysis of these results, we used BlandAltman plots to compare the two sets of data.7,8 The difference between two measurements of the same sample was plotted against the mean of the same two measurements within the three data sets, i.e., Cap and Lab, Ven and Lab, and Ven and Cap.

RESULTS A total of 270 healthy subjects were enrolled for this study. Of these, 157 (58%) were from Clementi Camp Medical Center, 83 (31%) were from Selarang Camp Medical Center, and 30 (11%) were from Maju Camp Medical Center. In terms of race, 219 (81%) of the study subjects were Chinese, 38 (14%) were Malay, 11 (4.1%) were Indian, and two (1%) were from other groups. All 270 subjects had venous plasma glucose levels from the laboratory (Lab) and venous whole-blood glucose levels from the glucometer (Ven) determined. Of these subjects, 170 (63%) consented to have the additional fingerprick to determine capillary wholeblood glucose levels using the glucometer. The mean Cap glucose reading was 5.10 mmol/L (91.8 mg/dL), with a standard deviation of 0.86 mmol/ L (15.48 mg/dL); the mean Ven glucose reading was 6.07 mmol/L (109.26 mg/dL), with a standard deviation of 0.66 mmol/L; and the mean Lab glucose reading was 5.10 mmol/L (91.8 mg/dL), with a standard deviation of 0.57 mmol/L (10.26 mg/dL). The mean difference between the Cap and Lab results was 0.01 mmol/L (0.18 mg/dL). The 95% confidence interval for this mean was ÿ0.11 to 0.12 mmol/L (ÿ1.98 to 2.16 mg/dL). In 82% of all subjects, the difference was 0.83 mmol/L (15 mg/dL) or less (Figure 1). The Pearson product–moment correlation coefficient between the Cap and Lab results was 0.5. The mean difference between Ven and Lab results was 0.97 mmol/L (17.46 mg/dL). The 95% confidence interval for this mean was 0.91 to 1.03 mmol/L (16.38

380

PREHOSPITAL EMERGENCY CARE OCTOBER / DECEMBER 2004 VOLUME 8 / NUMBER 4

FIGURE 1. Bland-Altman plot of the difference between each capillary whole-blood glucose sample from the glucometer (Cap) and each corresponding venous plasma glucose sample (Lab) plotted against the mean of the two samples. 2s = 2 SD.

to 18.54 mg/dL) (Figure 2). The Pearson product– moment correlation coefficient between the Ven and Lab results was 0.7. The mean difference between the Ven and Cap results was 0.92 mmol/L (16.56 mg/dL). The 95% confidence interval for this mean was 0.80 to 1.04 mmol/L (14.4 to 18.72 mg/dL) (Figure 3). The Pearson product– moment correlation coefficient between the Ven and Cap results was 0.5.

DISCUSSION Hypoglycemia is a medical emergency that is usually characterized by a venous plasma glucose level below 2.5 to 2.8 mmol/L (45 to 50.4 mg/dL). However, it is well known that the threshold for hypoglycemiainduced symptoms and the physiologic response to hypoglycemia vary significantly in different clinical settings with different patients. In view of the fact that there is no exact cutoff value for hypoglycemia and the possibility that it can occur in the paucity of symptoms and signs, e.g., when a patient is taking beta-blockers, it is important for medical per-

sonnel to be able to assess the patient’s glucose levels rapidly and accurately. This has led to the emergence of several laboratory procedures that are used by hospital personnel and several point-of-care handheld glucose measurement devices for OOH medical personnel. The Bland-Altman plot was used for clinical comparison of a ‘‘new’’ measurement technique (i.e., venous-on-capillary, capillary samples) with a more established form of glucose measurement (i.e., venous samples). The Bland-Altman plot is easy to compute and interpret. By plotting the difference of two measurements of the same sample against the mean of the same two measurements, this method allows both a graphical representation of the data and a trend of the difference. The mean difference between the two sets of values and the 2s of this mean difference between the two values are also plotted on the graphs to provide an overview of the spread of the differences between the two readings. It is understandable that this plot is not as commonly used as Pearson’s correlation coefficient. However, the Bland-Altman plot provides greater insight into comparison of two measurement techniques.7,8 A recent concise description of the

Kumar et al. CAPILLARY/VENOUS BLOOD GLUCOMETRY

VS.

LABORATORY VALUES

381

FIGURE 2. Bland-Altman plot of the difference between each venous whole-blood glucose sample from the glucometer (Ven) and each corresponding venous plasma glucose sample (Lab) plotted against the mean of the two samples. 2s = 2 SD.

various methods in correlational analysis provides a more detailed discussion on this topic.9 There was a very small difference between the Cap and Lab glucose values at this range of glucose concentration. This difference was not clinically significant. We arbitrarily selected a difference between Cap and Lab values of 0.83 mmol/L (15 mg/dL) and computed that in 82% of all subjects, the Cap values would be within these ‘‘tolerance’’ limits from the Bland-Altman plots. At the range of glucose concentrations that we studied, this difference is not likely to be clinically significant but could be an issue at other settings. The use of the Bland-Altman plot enables us to set ‘‘tolerance’’ limits, which may vary according to circumstances and from which we can determine the percentage of values that would be acceptable in a certain scenario. We found a Pearson correlation coefficient of 0.5 between the Ven and Cap values; this compares favorably with the value of 0.24 from a smaller study1 that also used the Bayer Glucometer Elite. Figure 3 shows that at lower blood glucose values of about 4 to 5 mmol/L (72 to 90 mg/dL), the Ven reading appears to

overestimate the Cap reading. This may be clinically significant when OOH care providers use blood from the intravenous cannula to measure the venous wholeblood glucose and then decide not to give intravenous glucose based on this value, which may be higher than the actual capillary whole-blood glucose value. In Figure 1, the Cap glucose value may appear to underestimate its Lab counterpart at lower blood glucose values of 3 to 4 mmol/L (54 to 72 mg/dL). Based on the impression that these readings provide, some patients in OOH settings may receive intravenous glucose when their actual Lab readings indicate euglycemia. This phenomenon also may be due to an instrument inaccuracy at lower glucose values. Indeed, the accuracy of glucometers at the low glucose range has been questioned.10,11 Unfortunately, we were unable to analyze this trend in detail since most of our subjects tended not to have blood glucose readings within this area of interest. In a small study involving 25 severely hypotensive and 39 normotensive patients, simultaneous fingerstick reagent strip glucose readings, venous reagent strip glucose readings, and laboratory venous plasma

382

PREHOSPITAL EMERGENCY CARE OCTOBER / DECEMBER 2004 VOLUME 8 / NUMBER 4

FIGURE 3. Bland-Altman plot of the difference between each venous whole-blood glucose sample from the glucometer (Ven) and each corresponding capillary whole-blood glucose sample from the glucometer (Cap) plotted against the mean of the two samples. 2s = 2 SD.

glucose readings were compared. Fingerstick glucose values in the hypotensive group were a mean of 67.5% of the laboratory glucose values in the same group; only 36% of hypotensive patients had fingerstick glucose values within the acceptable range of 20% of the laboratory glucose value. Thirty-two percent of the hypotensive patients were incorrectly labeled as hypoglycemic [glucose <3.9 mmol/L (<70.2 mg/dL)] and, interestingly, two of these patients were actually hyperglycemic [glucose >11.1 mmol/L (>199.8 mg/ dL)].12 An earlier paper discussed case reports of misdiagnosed hypoglycemia using capillary whole-blood samples in patients with poor peripheral circulation and suggested that ‘‘a drop of venous blood should be used for bedside glucose measurement if the diagnosis is in doubt.’’13 A Danish study reported that venous plasma glucose measurements, and not capillary whole-blood glucose measurements, should be used for the diagnosis of diabetes mellitus.14 From these trials, we can thus understand the limitations of capillary whole-blood glucose measurements in various clinical settings.

CONCLUSION There are several modalities for determining blood glucose levels, and each has its advantages and disadvantages. From our study, capillary whole-blood glucose levels from the glucometer best approximate venous plasma glucose levels from the laboratory, but it must be noted that at lower glucose ranges and in patients with poor peripheral circulation, e.g., shock, capillary whole-blood glucose levels will not be accurate. A drop of venous whole blood may be used to measure blood glucose levels from the glucometer, but this practice appears to overestimate the venous plasma glucose level from the laboratory by about 1mmol/L (18 mg/ dL). This may result in the withholding of intravenous glucose in a patient who is actually hypoglycemic. However, both methods (capillary whole blood using the glucometer and venous whole blood using the glucometer) provide rapid estimates of the patient’s blood glucose level and act as surrogate values until laboratory ‘‘gold standards’’ can be obtained. Hence, they do facilitate clinical management decisions. As care providers in the OOH setting and in emergency

Kumar et al. CAPILLARY/VENOUS BLOOD GLUCOMETRY

VS.

LABORATORY VALUES

departments, we must appreciate these differences in order to make sound clinical decisions. The authors thank the following people for their help: BG (Dr.) Wong Yue Sie, Chief Medical Corps, Singapore Armed Forces, for his approval and support of this study; Dr. Chan Yiong Huak, Clinical Trials and Epidemiology Research Unit, National Medical Research Council, Singapore, for his statistical advice and assistance with data analysis; A/P Sunil Sethi, Head, Department of Laboratory Medicine, National University Hospital, for his advice on aspects of clinical biochemistry; CPT (Dr.) Nicholas Foo, Medical Officer, Maju Medical Center, Singapore Armed Forces, for his assistance in data collection from patients at his center; and medical personnel from Selarang Camp Medical Center, Clementi Camp Medical Center and Maju Camp Medical Center, for their assistance in data collection.

References 1. Funk DL, Chan L, Lutz N, Verdile VP. Comparison of capillary and venous glucose measurements in healthy volunteers. Prehosp Emerg Care. 2001;5:275–7. 2. Harrison B, Markes R, Bradley P, Ismail IA. A comparison of statistical techniques to evaluate the performance of the Glucometer Elite blood glucose meter. Clin Biochem. 1996; 29:521–7. 3. Grunert C, Wood WG. Comparison between the Bayer Elite glucometer and the hexokinase method for glucose determination on the Eppendorf EPOS 5060 using capillary whole blood samples. Eur J Clin Chem Clin Biochem. 1995;33:153–6.

383

4. Rheney CC, Kirk JK. Performance of three blood glucose meters. Ann Pharmacother. 2000;34:317–21. 5. Wright WR, Rainwater JC, Tolle LD. Glucose assay systems: evaluation of a colorimetric hexokinase procedure. Clin Chem. 1971;17:1010–5. 6. Yee HY. Automated hexokinase procedure for assaying glucose in urine, serum or plasma. Clin Chem. 1972;18:1416–9. 7. Altman DG, Bland JM. Measurement in medicine: the analysis of method comparison studies. Statistician. 1983;32:307–17. 8. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–10. 9. Chan YH. Biostatistics 104: correlational analysis. Singapore Med J. 2003;44:614–9. 10. Moberg E, Lundblad S, Lins PE, Adamson U. How accurate are home blood-glucose meters with special respect to the low glycemic range? Diabetes Res Clin Pract. 1993;19:239–43. 11. Choubtum L, Mahachoklertwattana P, Udomsubpayakul U, Preeyasombat C. Accuracy of glucose meters in measuring low blood glucose levels. J Med Assoc Thai. 2002;85(suppl 4): S1104–10. 12. Atkin SH, Dasmahapatra A, Jaker MA, Chorost MI, Reddy S. Fingerstick glucose determination in shock. Ann Intern Med. 1991;114:1020–4. 13. Lee KO, Wong JE, Cheah JS. Hypoglycemia misdiagnosed by capillary glucose measurement. Ann Acad Med Singapore. 1989;18:453–5. 14. Stahl M, Brandslund I, Jorgensen LG, Hyltoft Petersen P, BorchJohnsen K, de Fine Olivarius N. Can capillary whole blood glucose and venous plasma glucose measurements be used interchangeably in the diagnosis of diabetes mellitus? Scand J Clin Lab Invest. 2002;62:159–66.