Professional Documents
Culture Documents
JR - Fingerstick Precision and Total Error
JR - Fingerstick Precision and Total Error
research-article2019
DSTXXX10.1177/1932296819831273Journal of Diabetes Science and TechnologyArnold et al
Original Article
Abstract
Background: Point-of-care (POC) HbA1c tests hold the promise of reducing the rates of undiagnosed diabetes, provided
they exhibit acceptable analytical performance. The precision and total error of the POC (Afinion™ HbA1c Dx) test were
investigated using whole blood samples obtained by fingerstick and venipuncture.
Methods: Fingerstick samples spanning the assay range were collected from 61 subjects at three representative POC sites.
At each site, six fingerstick samples were obtained from each subject and tested on the POC test across two (Afinion AS100)
instruments. Repeatability, between-operator, and between-instrument components of variance were calculated using
analysis of variance (ANOVA). Four venous samples (low, threshold, medium, and high HbA1c) were measured in duplicate
across three instruments using three reagent lots, twice per day over 20-days. Repeatability, between-run, between-day,
between-lot, and between-instrument components of variance were calculated. These fingerstick and venous blood results,
combined with estimates of imprecision and bias from a prior investigation, allowed for the calculation of the total coefficient
of variation (CV) and total error of the POC test using fingerstick and venous whole blood samples.
Results: The total imprecision ranged from 1.30% to 2.03% CV using fingerstick samples and from 1.31% to 1.64% CV using
venous samples. The total error ranged from 2.87% to 4.75% using fingerstick samples and from 2.93% to 3.80% using venous
samples.
Conclusions: The POC test evaluated here is precise across its measuring range using both fingerstick and venous whole
blood. The calculated total error of the test is well under the accepted quality requirement of ≤6%.
Keywords
diagnosis of diabetes, HbA1c, hemoglobin A1c, point of care, precision, total error
The measurement of HbA1c is an accepted, convenient and The results of a previous investigation of a POC method,
useful means for the diagnosis of diabetes. Despite some the Afinion™ HbA1c Dx test (Study A) first suggested that
limitations, such as interference from common hemoglobin this method may offer acceptable accuracy and precision
variants and conditions associated with increased red blood when used by trained laboratory operators.5
cell turnover, HbA1c measurement presents numerous
advantages including greater preanalytical stability, less day- 1
Abbott Rapid Diagnostics, a division of Abbott Laboratories, San Diego,
to-day variability during stress or illness, and does not CA, USA
require individuals to fast prior to testing.1 The barrier to 2
Abbott Rapid Diagnostics, a division of Abbott Laboratories, Oslo,
evaluating glycemic status could be further reduced by uti- Norway
3
lizing POC HbA1c methods, which can provide access to L-MARC Research Center, Louisville, KY, USA
4
Rochester Clinical Research, Inc, Rochester, NY, USA
care in disadvantaged communities and offer the potential of 5
Rainier Clinical Research Center, Inc, Renton, WA, USA
identifying many more underserved individuals with predia-
betes and diabetes.2-4 The advantages and disadvantages of Corresponding Author:
Richard C. San George, PhD, Abbott Rapid Diagnostics, a division of
HbA1c versus fasting plasma glucose or oral glucose toler- Abbott Laboratories, 9975 Summers Ridge Rd, San Diego, CA 92121,
ance testing for the diagnosis of diabetes have been discussed USA.
extensively elsewhere.1 Email: rick.sangeorge@alere.com
Arnold et al 891
After the completion of Study A, there was interest in eval- used to estimate the bias of the POC test at HbA1c values of
uating additional components of variance not assessed in that 5.0% (31.1 mmol/mol), 6.5% (47.5 mmol/mol), 8.0%
study design. In the study reported here (Study B), a direct (63.9 mmol/mol), and 12% (107.7 mmol/mol). Three POC
evaluation of repeatability, between-operator, and between- test cartridge lots were used across sites, and the between-lot
instrument precision is made using fingerstick (capillary) component of variance, as well as repeatability estimates for
whole blood samples, the principal sample type used at the low, threshold, medium, and high HbA1c levels, were calcu-
POC. This is a first of a kind study to evaluate components of lated as follows. Samples were grouped into the following
variance in fingerstick samples. In addition, precision using four HbA1c interval levels: low (4.00%-5.99% HbA1c, or
venous whole blood samples was evaluated over 20 days, 20.2-42.0 mmol/mol HbA1c), threshold (6.00%-6.99%
consistent with Clinical and Laboratory Standards Institute HbA1c, or 42.1-52.9 mmol/mol HbA1c), medium (7.00%-
(CLSI) Guideline EP05-A3 (Study C).6 Taken together, the 9.99% HbA1c, or 53.0-85.7 mmol/mol HbA1c), and high
data from these studies (A, B, and C) enable the evaluation of (≥10% HbA1c, or 85.8 mmol/mol HbA1c). For each HbA1c
the total imprecision and total error of this POC test for both level the imprecision was separated into repeatability (within-
fingerstick and venous whole blood sample types. run) and between-lot components of variance using a two-
factor ANOVA model without nesting based on Type II sums
Methods of squares (MATLAB ANOVAN and SAS PROC MIXED).
In Study B, at three POC study sites having laboratories
The POC test (manufactured by Alere Technologies AS, operated by trained medical professionals, fingerstick sam-
Oslo, Norway; a subsidiary of Abbott Laboratories) is a fully ples from prospectively enrolled subjects were collected and
automated boronate affinity assay for the determination of tested on the POC test. Each subject signed an informed con-
the percentage of hemoglobin A1c in human whole blood. sent form approved by a central, independent institutional
The test cartridge contains all of the reagents necessary for review board (IRB) prior to any study procedures being con-
the determination of %HbA1c. A patient specimen is col- ducted. Enrollment included those that met the following
lected with the integrated sampling device then reinserted eligibility criteria:
back into the test cartridge, which is then placed in the
Afinion AS100 Analyzer (instrument). The sample is auto- Inclusion criterion:
matically diluted and mixed with a solution that releases
hemoglobin from the erythrocytes. After the hemoglobin is •• At least 18 years of age
precipitated, the sample mixture is transferred to a blue
boronic acid conjugate which binds to the cis-diols of gly- Exclusion criteria:
cated hemoglobin. This reaction mixture is soaked through a
filter membrane and all precipitated hemoglobin, glycated •• Known to have the hemoglobin variant HbF > 7%
and nonglycated hemoglobin, remains on the membrane. •• Diagnosed with hemophilia
Excess conjugate is removed with a washing reagent. The •• Taking glucocorticoid or nicotinic acid medications
analyzer measures the reflectance of the precipitate on the •• Diagnosed with iron deficiency, hemolytic anemia,
membrane as blue (glycated hemoglobin) and red (total thalassemias, hereditary spherocytosis, chronic hepatic,
hemoglobin) color intensities. The percentage of total gly- or renal disease
cated hemoglobin and %HbA1c has been shown to be well •• Pregnant
correlated;7 calibration adjusts the measured total glycated •• Received a blood transfusion or cancer chemotherapy
hemoglobin to report the equivalent level of HbA1c. The within the prior three weeks
POC test is traceable to the International Federation of •• Unwilling or unable to sign a written informed con-
Clinical Chemistry and Laboratory Medicine (IFCC) sent and comply with study procedures, or is deemed
Reference Method for Measurement of HbA1c,8 and is certi- inappropriate for participation by the site investigator
fied by the NGSP as having documented traceability to the
Diabetes Control and Complications Trial (DCCT) Reference Each site targeted enrolling five subjects having HbA1c val-
Method. The NGSP awards certification to manufacturers for ues within one of the low, threshold, medium, and high levels
successfully meeting performance criteria for bias testing.9 as defined above. Fingerstick sample testing was conducted
In Study A, conducted previously,5 a first-replicate POC by using the same lot of POC test cartridges at each of the
test value (from fingerstick and venous whole blood samples) three study sites. At each site, for each subject, there were
was compared against the average of duplicate results from a three test operators who performed testing across two instru-
venous sample tested on the Tosoh glycohemoglobin test on ments. Each operator collected two fingerstick samples and
the G8 HPLC analyzer at the NGSP Secondary Reference tested them, one sample per instrument. The results from the
Laboratory (SRL) at the Diabetes Diagnosis Laboratory, fingerstick testing were evaluated by using ANOVA method-
University of Missouri School of Medicine, Columbia MO. ology based on Type II sums of squares (MATLAB ANOVAN
Passing-Bablok regression equations were calculated and and SAS PROC MIXED), consistent with the methodology
892 Journal of Diabetes Science and Technology 14(5)
Table 1. Summary of Components of Variance Used in the The total CVs, together with the bias estimates from study
Calculation of Total Imprecision, and Study From Which the A, were used to estimate the total error of the POC test at
Estimate Was Obtained. HbA1c values of 5.0% (31.1 mmol/mol), 6.5% (47.5 mmol/
Component of Study where mol), 8.0% (63.9 mmol/mol), and 12% (107.7 mmol/mol)
Sample type variance estimate obtained using fingerstick and venous whole blood samples. The coef-
ficient of variance, bias, and total error are expressed as a
Fingerstick Repeatabilitya A
percentage of the mean, where the mean is expressed in
Between-lot A
NGSP units (%HbA1c).
Between-instrument B
Between-operator B
Between-run C Results
Between-day C
Venous Repeatability C A total of 62 subjects were enrolled across three sites for the
Between-lot C purpose of fingerstick sample testing in Study B. One of the
Between-instrument C 62 subjects had out-of-range HbA1c values and was excluded
Between-run C from analysis. The remaining 61 subjects each provided six
Between-day C fingerstick samples. The population had an average age of
a 60 years, was 56% male, and had HbA1c values evenly dis-
The repeatability estimates from Study A were used in the calculation
of total imprecision using fingerstick samples for all HbA1c levels, except tributed over the measurement range of the POC test.
for the high level, which was estimated from Study B due to larger sample Estimates of between-instrument, between-operator, and
size (15 subjects vs 6 subjects). repeatability components of variance from the testing of
these fingerstick samples are shown in Table 2. Across the
described in CLSI EP05-A3.6 The repeatability (within-run), assay range, the between-instrument imprecision did not
between-operator, and between-instrument components of exceed 0.47% CV, and the between-operator imprecision did
variance were calculated for each HbA1c level by combining not exceed 0.21% CV. While repeatability was the largest
the data from all sites. For each component of variance, the component of variance, it did not exceed 1.39% CV.
CV was calculated as the SD divided by the grand mean of These estimates of fingerstick repeatability were slightly
all the POC test results within the level. smaller than but consistent with those calculated from previ-
In Study C, at the manufacturer’s site, four venous whole ous studies. Table 3 describes the repeatability and between-
blood samples were selected for precision testing, one at lot components of variance that were calculated using prior
each HbA1c level (low, threshold, medium, and high), as duplicate fingerstick measurements in Study A.5
defined for the fingerstick samples. These K2-EDTA- In Study C, four venous whole blood samples, represent-
anticoagulated samples were identified from among deiden- ing low, threshold, medium, and high HbA1c levels, were
tified remnants of specimens from hospital laboratories or tested on the POC test over 20 days. A grand total of 720
fresh samples collected from an IRB approved in-house measurements for each HbA1c level resulted from testing
donation. Each of the four samples was tested on the ana- these samples on each of three cartridge lots and each of
lyzer using each of three reagent lots and each of three instru- three instruments with two runs per day and two measure-
ments. Two replicate results were obtained during each of ments per run. Estimates of the components of variance and
two runs per day over a 20-day period, for a total of 720 total %CV for these venous sample results are shown in
measurements at each HbA1c level. From the results of Table 4. As was the case with fingerstick samples in Study A
Study C, the repeatability (within-run), between-run, and Study B, the repeatability was the single largest compo-
between-day, between-lot, and between-instrument SD and nent of variance for the POC test when using venous whole
CV were calculated (SAS VARCOMP). blood samples.
The total CV of the POC test using fingerstick samples Using the sums of squares of the relevant components of
was calculated based on the sum of squares of six compo- variance, estimates were made of the total imprecision of the
nents of variance: repeatability, between-lot, between-run, POC test when testing fingerstick samples (Table 5). The
between-day, between-instrument, and between-operator total error of the POC test was then estimated according to
(see Table 1). The between-run and between-day compo- the equation:
nents were obtained from the POC test results using venous
samples in Study C, since these components cannot be %TE = % Bias + 1.96 × %CV × (1 + %Bias/100 )
directly estimated using fingerstick samples. The between-
lot and repeatability components were obtained from Study Across the assay range, the total imprecision using finger-
A, because the number of subjects was larger in this study stick samples was 2.03% CV or less, and the total error did
than in Study B. Study B was necessary to estimate the not exceed 4.75% (Table 6). When using venous blood sam-
between-operator and between-instrument components of ples the total CV was 1.64% or less, and the total error was
variance in fingerstick samples. no greater than 3.80%.
Arnold et al 893
Table 2. POC Test Components of Variance Using Fingerstick Whole Blood Samples Estimated in Study B.
Table 3. POC Test Components of Variance Using Fingerstick Whole Blood Samples Estimated in Study A.5
Table 4. POC Test Components of Variance Using Venous Whole Blood Samples Estimated in Study C.
Mean HbA1c Repeatability Between run Between day Between lot Between instrument
HbA1c Level %HbA1c mmol/mol %CV %CV %CV %CV %CV Total %CV
Results using samples from four subjects and 720 test results per subject. The CVs are expressed as a percentage of the mean in NGSP units (%HbA1c).
%HbA1c (mmol/mol)
Study Variance component 4.00-5.99 (20.2-42.0) 6.00-6.99 (42.1-52.9) 7.00-9.99 (53.0-85.7) ≥10 (85.8)
a
A N (M) 45 (90) 68 (136) 51 (102) 6 (12)
Repeatability %CVb 1.52 1.36 1.35 —
Between-lot %CV 1.20 0.42 0.00 0.07
B N (M)a 15 (90) 15 (90) 16 (96) 15 (90)
Repeatability %CVb — — — 1.14
Between-instrument 0.14 0.30 0.00 0.47
Between-operator 0.00 0.00 0.21 0.00
C Ma 720 720 720 720
Between-run 0.00 0.00 0.21 0.12
Between-day 0.60 0.62 0.56 0.39
TOTAL %CV 2.03 1.58 1.49 1.30
The CVs are expressed as a percentage of the mean in NGSP units (%HbA1c).
a
N is the number of subjects, M is the number of test results. bThe repeatability estimates from Study A were used in the calculation of total imprecision
using fingerstick samples for all HbA1c levels, except for the High level, which was estimated from Study B due to larger sample size.
894 Journal of Diabetes Science and Technology 14(5)
Table 6. Total Imprecision, Bias, and Total Error Estimates for POC Test.
The CV, bias, and TE is expressed as a percentage of the mean in NGSP units (%HbA1c).
a
Bias relative to the Tosoh Glycohemoglobin test on the G8 HPLC analyzer measured at an NGSP SRL, estimated in Study A.5
7. Nuttall FQ. Comparison of percent total GHb with percent 12. Rohlfing CL, Parvin CA, Sacks DB, Little RR. Comparing
HbA1c in people with and without known diabetes. Diabetes analytic performance criteria: evaluation of HbA1c certifica-
Care. 1998;21:1475-1480. tion criteria as an example. Clin Chim Acta. 2014;433:259-
8. Jeppsson JO, Kobold U, Barr J, et al. Approved IFCC reference 263.
method for the measurement of HbA1c in human blood. Clin 13. NGSP. College of American Pathologists (CAP) GH5 survey
Chem Lab Med. 2002;40:78-89. data. [Internet]. Columbia, MO: NGSP; [updated 2017 Dec;
9. NGSP. List of NGSP certified methods [Internet]. Columbia cited 2019 Feb 08]. Available at: http://www.ngsp.org/CAP
(MO): NGSP; [updated 2019 Feb; cited 2019 Feb 08]. /CAP17c.pdf. Accessed April 2018.
Available from: http://www.ngsp.org/docs/methods.pdf. 14. Sobolesky PM, Smith BE, Saenger AK, et al. Multicenter
10. Sacks DB, Arnold M, Bakris GL, et al. Guidelines and recom- assessment of a hemoglobin A1c point-of-care device for diag-
mendations for laboratory analysis in the diagnosis and man- nosis of diabetes mellitus. Clin Biochem. 2018;61:18-22.
agement of diabetes mellitus. Clin Chem. 2011;57:e1-e47. 15. Lenters-Westra E, English E. Evaluation of four HbA1c
11. US Department of Health and Human Services. Hemoglobin point-of-care devices using international quality targets: are
A1c test system. [Internet]. Silver Spring, MD: US food and they fit for the purpose? J Diabetes Sci Technol. 2018;12:762-
Drug Adminstration; 2014 Aug 25 [updated 2018 Sep 04; 770.
revised 2018 Apr 01; cited 2019 Feb 08]. Available at: https:// 16. Lenters-Westra E, Slingerland RJ. Three of 7 hemoglobin A1c
www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch point-of-care instruments do not meet generally accepted ana-
.cfm?fr=862.1373. Accessed April 2018. lytical performance criteria. Clin Chem. 2014;60:1062-1072.