| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
1 Clinical Chemistry Branch, Division of Laboratory Sciences, National Center for Environmental Health, 2 Behavioral and Clinical Surveillance Branch, Division of HIV and AIDS Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, Atlanta, GA; and 3 International Medical Press, Atlanta, GA
aaddress correspondence to this author at: Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, Mailstop F25, 4770 Buford Hwy NE, Atlanta, GA 30341-3724; fax 770-488-4192, e-mail mkimberly{at}cdc.gov
Many diabetic patients use home-use glucose monitors for glycemic control (1)(2). Ideally, in studies comparing and assessing the variability among monitors, testing should be done with the same type of sample used by patients, i.e., capillary whole blood. Simultaneous evaluation of multiple monitors requires more sample than can effectively be collected by lancing a single finger one or more times. Samples may therefore need to be collected from multiple fingers, and if so, the among-finger variability will need to be quantified before a study of among-monitor variability can be conducted. To our knowledge, no studies have been published that report among-finger variability in blood glucose measurements within an individual patient. Because evaluation of multiple monitors using multiple samples from an individual also assumes that the within-person blood glucose will be stable for the duration of the testing period, we tested, in a pilot study using a HemoCue 201 point-of-care glucose analyzer, the assumptions that the within-person, among-finger variability (hereafter referred to as "among-finger" variability) is negligible and that the within-person blood glucose is stable over a period of 35 min.
The Institutional Review Boards at the CDC and at Research Triangle Institute (where the study was conducted) approved the protocol. Free-living, community-dwelling people recruited from the Raleigh-Durham-Chapel Hill metropolitan area of North Carolina gave informed consent. Diabetic and nondiabetic adults over 18 years of age with a fasting glucose of 700-2000 mg/L and hematocrit of 30–55% were included in the sample population. Participants were required to have the middle three fingers on both hands. We excluded people who were taking medications that alter blood viscosity (such as aspirin); people with peripheral vascular disease, kidney disease, or hemophilia; and people with obvious lipemia (based on milky appearance by visual inspection). The admission requirements for diabetic patients included diagnosis of diabetes within the last 15 years, absence of severe hypoglycemic events within the previous 3 months, and ability to recognize the symptoms of hypoglycemia. To achieve stable metabolic status, all participants were required to fast for at least 8 h before participating in the study.
Statistical power calculations were used to determine the sample size needed for the pilot study based on an assumed CV of 3.5% for HemoCue 201 analyzer measurements. Using this CV, we determined that a minimum of 11 participants would be required to detect an among-finger CV 
2.5% (80% power, using a two-sided 0.05 level test). We actually sampled 20 individuals, and the calculated CV (by ANOVA) of the HemoCue 201 measurements was 3.7%. This CV includes the analytical imprecision of the instrument as well as any within-person, within-finger imprecision. When we recalculated the statistical power using a sample size of 20 and a CV of 3.7%, we determined that we should be able to detect an among-finger CV 2.1% with 80% power using a 0.05 level test.
Eight diabetic and 12 nondiabetic individuals participated in the pilot study. One trained licensed practical nurse collected and measured all samples. Fingers were lanced by use of a Tenderlett lancing device (International Technidyne Corporation). The sampling scheme was the same for all participants and involved taking capillary samples from the left and right sides of the middle three fingers of both hands. The left side of all six fingers was lanced before returning to the first finger to lance the right side. Thus, each finger was sampled twice at two different sites. The order of finger sampling was the same for all participants to minimize errors from changes in sampling protocol from person to person. Measurements were performed with a HemoCue 201 using one lot of cuvettes and control solutions. The HemoCue test system is based on the glucose dehydrogenase method. All measurements from a participant were completed within 30–35 min.
The means of the 12 blood glucose measurements ranged from 710 to 1350 mg/L for all 20 participants, from 900 to 1350 mg/L for the 8 diabetic participants, and from 710 to 1040 mg/L for the 12 nondiabetic participants. The mean (SD) difference between the 1st and 12th measurement for each person was 40 (60) mg/L (range, –60 to 180 mg/L). The mean (SD) difference between the maximum and minimum values for each person was 120 (40) mg/L (range, 60–210 mg/L). There was no apparent concentration dependence of the differences between the 1st and 12th measurements or the maximum and minimum measurements. The diabetic and nondiabetic participants did not demonstrate statistically significant differences between the 1st and 12th or between the maximum and minimum blood glucose concentrations. Statistically significant (P <0.05) increases in glucose concentrations were observed in 3 (2 nondiabetic and 1 diabetic) of the 20 participants, and a statistically significant decrease was observed in 1 diabetic individual. For the three individuals with increases, the estimated slope from a regression of glucose concentration on time of measurement predicted an increase in concentration after 35 min of 110, 188, and 130 mg/L, and for the individual with the decrease, the predicted decrease was –56 mg/L after 35 min. For all other participants, there was no significant increase or decrease in glucose concentration over the 35-min period from 1st to 12th measurement. From these data, we did not note any systematic increases or decreases in glucose over time across the population, and we concluded that the changes seen over the course of 35 min could be attributed to among-finger variation and short-term random physiologic variation.
We used ANOVA to determine the among-finger variance. Confidence intervals (CIs) were determined and are based on the 
2 distribution and the estimated degrees of freedom associated with the variance estimate (3)(4). The among-finger variance was 0.98 (95% CI, 0.81–1.21). Therefore, at a blood glucose concentration of 1000 mg/L, the estimated among-finger CV was 0.99% (95% CI, 0.90–1.1%). Correspondingly, the estimated among-finger CVs were 1.4% (95% CI, 1.3–1.6%) and 0.73% (95% CI, 0.67–0.81%) at blood glucose concentrations of 700 and 1350 mg/L, respectively.
These results indicate that the among-finger variability needs to be considered when assessing among-monitor variability using capillary blood from different fingers of the same individual. Although this variability is important for the assessment of variability among monitors, it is too small to be of clinical relevance and does not affect patient care. Therefore, diabetic patients or their physicians do not need to be concerned about which finger to use for consecutive capillary sampling.
Acknowledgments
This work was supported through an Interagency Agreement with the National Institute for Diabetes and Digestive and Kidney Diseases. The study was conducted at Research Triangle Institute; the assistance of John Heinrich, Angela Burroughs, and Michael J. Phillips is acknowledged.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
