CLINICAL CHEMISTRY
Original Article
Laboratory and Bedside Evaluation of
Portable Glucose Meters
JAMES H. NICHOLS, P H D , 1 CAROLINE HOWARD, MT, 1 KIMBERLY LOMAN, RN, 2
CYNTHIA MILLER, RN, 2 DOROTHY NYBERG, MS, RN, 3 AND DANIEL W. CHAN, P H D 1
hematocrit bias and limited linear range when compared to the Nova.
Only meters meeting both minimal analytical performance and computerization requirements, the One Touch II and AccuData Easy, were
selected for further evaluation. At the bedside, the One Touch II demonstrated performance consistent with the lab evaluation, whereas the
AccuData Easy showed greater imprecision in the low glucose range
and a correlation that varied with sample type: capillary, venous, or
arterial blood. This evaluation indicates that the clinician must interpret near-patient glucose results with respect to meter limitations. FDA
approval and marketing statistics, alone, are insufficient to judge the
performance of the meters in routine institutional use. Independent
method validation, under actual operating conditions, is a better means
of predicting future performance of the meters. (Key words: Glucose
meters; Near-patient testing; Decentralized testing; Quality assurance;
Portable monitors) Am J Clin Pathol 1995;103:244-251.
A two-phase, laboratory and bedside, evaluation of blood glucose meters was conducted in this study. Four meters, the AccuData Easy (Boehringer-Mannheim, Indianapolis, IN), HemoCue Glucose (HemoCue,
Mission Viejo, CA), LifeScan One Touch II (LifeScan, Milpitas, CA),
and Miles Encore QA (Miles, Elkhart, IN) systems, were compared
to the Nova Stat Profile 5 (Nova, Waltham, MA) as the laboratory
reference. Precision, linearity, correlation to the laboratory method, interference from hematocrit, data management, and operator preference
were examined. None of the meters were found to satisfy all of the
study's evaluation criteria. Therefore, institutions must weigh which
criteria are most important to their individual settings. Although the
HemoCue Glucose was found to be technically superior, this meter had
no data management capabilities. The Encore QA had greater variance
and low bias, whereas the AccuData Easy had bias affected by hematocrit and glucose concentration, and the One Touch II had a negative
bedside, evaluation was conducted. Four meters were initially
examined during the laboratory phase: the AccuData Easy
(Boehringer-Mannheim, Indianapolis, IN), HemoCue Glucose
(HemoCue, Mission Viejo, CA), LifeScan One Touch II (LifeScan, Milpitas, CA), and Miles Encore QA (Miles, Elkhart, IN)
systems. The selected manufacturers have representative meters currently in our hospital. The HemoCue meter was included because of reports of enhanced analytical performance. 9
The meter models are the latest in reflectance technology by
the selected manufacturers. In the laboratory phase of the evaluation, each meter's performance was tested by one medical
technologist and the precision, linearity and correlation to the
Nova Stat Profile 5 (Nova, Waltham, MA) were compared. Hematocrit and volume effects were also investigated. Based on
the initial results, the AccuData Easy and One Touch II Hospital systems, two meters that met the minimal institutional requirements for analytical performance and computerized data
management, were selected for further bedside testing.
Multiple nurse operators performed quality control, patient
correlations, and completed an ease-of-use/operator preference
survey. Data from both phases of this investigation were used
by an interdepartmental task force on bedside glucose testing
From the 'Department ofPathology, Johns Hopkins Medical Institu- to make recommendations on the purchase of a hospital-wide
tions, Baltimore, Maryland; ^Department ofMedicine, Johns Hopkins instrument and revisions to our overall quality assurance proHospital, Baltimore; *Department ofNursing, Johns Hopkins Hospital, gram.
Baltimore.
MATERIALS A N D METHODS
Manuscript received March 15, 1994; revision accepted August 15,
Reagents
1994.
The AccuData Easy, HemoCue Glucose, One Touch II HosAddress reprint requests to Dr. Nichols: Department of Pathology,
pital, and Encore QA glucose meters were loaned to the hospiJohns Hopkins Hospital, 600 N. Wolfe St., Baltimore MD 21287-7065.
Point-of-care testing has become important in management of
the diabetic patient, especially in the post-surgical and medical
Intensive Care Units, Critical Care Units, Obstetric/gynecology
and Emergency room services1,2 By providing faster turnaround times, the clinician can rapidly respond to changes in
the patient's glucose levels, resulting in optimum insulin regimens and shorter, less expensive hospitalizations.3,4 Ultimately, the ability of a meter to provide a rapid result must be
weighed against its reliability. Early instrument performance
was dependent on operator technique. 5 With the development
of newer technologies, the reliability of results obtained by nonlaboratory operators has improved, provided that the operators
are trained and quality control guidelines are developed and
maintained.6"8 Improvements in computerized data management have made the recording and review of quality control
results easier.
With recent regulatory pressures and present advances in
meter technology, our hospital was prompted to review the status of its bedside testing program. A two-part, laboratory and
244
245
NICHOLS ET AL.
le Glucose Meters
Evaluation ofPor
tal by the manufacturers for this evaluation. The One Touch II
Hospital meter and the Encore QA were set to display plasma
corrected results, and the other meters report whole blood results. All test strips and controls were supplied by the manufacturers as specified for use with each instrument. Reagents were
stored as recommended. HemoCue supplied Boehringer Mannheim Glucose BG controls for use with its meter because
HemoCue does not currently manufacture specific controls.
For the laboratory phase, arterial blood specimens were obtained from samples submitted for blood gas analysis in our
Critical Care Laboratory. All arterial specimens were drawn in
lithium heparin syringes and analyzed within 30 minutes of
phlebotomy. Venous specimens were collected in lithium
heparin Vacutainer brand tubes (Becton Dickinson, Rutherford, NJ).
Methods
For the laboratory phase, two meters from each manufacturer were evaluated. Intraassay (within day) precision was determined from 20 tests in succession using either control material or an arterial patient specimen. Interassay (between day)
precision was determined from 20 control analyses over a 10day period (2 per day).
Linearity was determined using multiple venous specimens
from a single patient. One aliquot was spiked with known
amounts of NBS glucose (National Bureau of Standards, SRM
917a). After gently mixing for 10 minutes at room temperature,
the spiked specimen was diluted with an unspiked aliquot in
amounts appropriate for the construction of a linear curve
(spiked:unspiked = 3:1, 1:1, 1:3). Specimens were mixed for 10
minutes and analyzed in duplicate within 30 minutes of dilution.
Central laboratory agreement was judged by correlation of
arterial whole blood glucose to the Nova Stat Profile 5 analyzer.
Arterial specimens were analyzed simultaneously on meters
and lab instrument.
Hematocrit effects were determined using venous specimens
from a single patient. After centrifugation for 10 minutes at
room temperature, at 1,000-1,300 relative cetrifugal force
(RCF), plasma was removed and remixed with red blood cells
to create a linear hematocrit curve (plasma:RBC = 3:1, 1:1,
1:3). Specimens were mixed for 10 minutes and analyzed in
duplicate. Remaining aliquots were sent to the Hematology
Laboratory for verification of exact hematocrit by Coulter
STKS (Coulter, Haileah, FL) and microhematocrit centrifuge.
When the effects of glucose concentration and hematocrit were
investigated, NBS glucose was employed as a spike before construction of the hematocrit curve.
For the bedside phase, operators were trained by the manufacturer on the correct operation, quality control, and maintenance of each meter. Thirteen inpatient nursing units: Anesthesia (ICU), Emergency Room, Medicine, Neurology,
Pediatrics (ICU), Obstetrics/Gynecology, Ophthalmology, Oncology, Surgery and Geriatrics, and two Medical outpatient
units at three hospitals: Johns Hopkins Hospital, Bayview Hospital and Wyman Park Medical Center, volunteered to participate. Forty-five operators volunteered who were from various
educational backgrounds: licensed practical nurse (5), registered nurse (28), physician's assistant (1), clinical nurse specialist (4), emergency technician (2), and nursing support technician (5). Operators were divided into two groups. Each group
Vol. 1
evaluated only one meter at a time. After completing the first
evaluation, the groups crossed-over and were trained for an
evaluation of the alternate meter. All operators performed
quality control, a patient test and cleaned both the nursing
unit's current meter and the evaluation meter each day. The
AccuChek II (Boehringer Mannheim, Indianapolis, IN) is currently the predominant meter in our institution. The Ames
Glucometer, One Touch II, and AccuChek Easy are in use only
on isolated nursing units. Therefore, the amount of correlation
data from these meters was limited. Results were collected for
5 working days and data were analyzed to compare operator
and meter performance. At the end of the bedside evaluation,
the operators completed a 4-page User Survey, ranking ease-ofuse, training, infection control, maintenance and preference on
a l-to-5 Likert scale. Twenty-five operators (cross-over group)
completed evaluations of both meters. Data from the crossover group were treated separately to determine any differences
from operators who only evaluated a single meter. Nurse managers participated in downloading data from the evaluation
meters and printing summary reports. Each nurse manager
completed a survey pertaining to data management capabilities.
RESULTS
Laboratory Evaluation
Results from the precision evaluation are summarized in Table 1. Overall means and coefficients of variation (CV) were
averaged from the data of the two instruments. Variance from
intraassay precision (Table 1) was slightly higher than the variance obtained by the manufacturer (within 3% of manufacturer's claims). Interassay precision (Table 1) also was higher than
reported (within 2% of manufacturer's claims). Trends noted in
the within-day precision experiment were also noted between
days. The AccuData Easy meter demonstrated higher variance
in the low range, whereas the One Touch II and HemoCue Glucose meters had higher variance in the high range. Only the
One Touch II Hospital and the HemoCue Glucose meters
passed the Association of Clinical Biochemists10 and the Consensus Development Conference on Self-Monitoring of Blood
Glucose6 recommendations for a CV of less than 5%. Comparison of precision data from control solutions and patient specimens indicated differences, most probably due to matrix effects
(Table 1). The manufactured control solutions evidently do not
mimic patient specimens and, thus, do not provide an adequate
indication of meter performance with respect to a whole blood
matrix.
Linearity was validated by comparison of spiked venous
specimens to the Nova Stat Profile 5 and by recovery of known
amounts of National Bureau of Standards glucose (Fig. 1). The
AccuData Easy and the HemoCue Glucose meters demonstrated linearity within their reportable ranges (1.2-25.4 actual
vs 1.1-27.8 mmol/L reportable for AccuData Easy and 0.716.8 actual vs 2.0-22.2 mmol/L reportable for HemoCue Glucose). Recoveries averaged 104.8% and 111.4% in the range of
7.6-24.6 mmol/L for AccuData Easy and HemoCue Glucose
meters, respectively (104.8% recovery for the Nova Stat Profile
5 on the same specimens). The Encore QA meter was linear
(0.8-30.3 actual vs 0.6 - 33.3 mmol/L reportable), but demonstrated greater imprecision above 25 mmol/L (Fig. 1) with av-
• No. 2
246
CLINICAL CHEMISTRY
Original Article
TABLE 1. INTRAASSAY AND INTERASSAY PRECISION
Intraassay precision (glucose, mmol/L)
Mean (control)
CV
N
Accu-Data Easy
Hemocue
One Touch II
Encore QA
Nova Stat Profile
2.28
3.9%
20
3.69
2.1%
20
2.74
2.7%
20
1.91
7.3%
20
—
5.35
5.1%
20
5.30
6.3%
20
—
Mean (control)
CV
N
5.05
2.0%
20
Mean (control)
CV
N
14.34
1.2%
20
20.20
3.8%
20
17.31
4.1%
20
24.06
5.7%
20
—
Mean (patient)
CV
N
5.30
8.0%
10
5.18
2.1%
10
10.31
1.6%
10
10.55
3.5%
10
—
2.09
6.9%
20
3.69
2.6%
20
2.75
2.7%
20
1.88
6.3%
20
3.79
3.8%
469
5.50
3.4%
20
5.16
4.3%
10
7.23
2.9%
492
15.84
3.8%
20
23.40
6.2%
20
13.56
3.6%
503
Interassay precision (glucose, mmol/L)
Mean (control)
CV
N
Mean (control)
CV
N
4.98
4.0%
20
Mean (control)
CV
N
13.81
2.3%
20
19.09
4.6%
20
erage recoveries of 103.3%. However, the One Touch II meter
was not linear to the full extent of its reportable range (0.2-19.4
actual vs 0-33.3 mmol/L reportable) demonstrating a low bias
above 19 mmol/L. Recoveries averaged 113.1% for the One
Touch II between 7.6 mmol/L and the upper, verified limit of
linearity, 19 mmol/L. Good correlation was noted to the Nova
Stat Profile 5 (plasma) within the acceptable linear limits for
the AccuData Easy (whole blood), HemoCue Glucose (whole
blood) and One Touch II (plasma/serum corrected)(Fig. 1).
The Miles Encore QA (plasma) demonstrated a low, constant
bias compared to the Nova Stat Profile 5 (plasma). Additional
test strip lots and instruments were examined in separate experiments to verify the lack of high end linearity with the One
Touch II and high-end precision/low bias with the Encore QA.
Central laboratory agreement was evaluated by comparison
of arterial blood specimens with the Nova Stat Profile 5 (Fig. 2).
The One Touch II meter (in plasma mode) and the HemoCue
Glucose (whole blood) demonstrated excellent correlation to
the Nova plasma glucose. However, the AccuData Easy whole
blood results were biased with respect to the Nova. Analysis of
duplicate correlation samples, on a second meter, verified these
regressions within 1% for slopes and intercepts ± 0.16 mmol/L
of the reported equations for One Touch II, HemoCue Glucose
and AccuData Easy meters. The Encore QA meters exhibited
wider regressions between duplicate meters and demonstrated
a plasma to plasma bias with the Nova (Encore A Meter =
-0.26 + 0.90 (Nova), r = 0.971, Sy.x = 0.63; Encore B Meter =
-0.55 + 0.93 (Nova), r = 0.972, S y.x = 0.65, n = 96). Mean
values of glucose for each instrument indicate the magnitude
of the manufacturer's bias; Nova = 9.03, One Touch II = 9.22,
HemoCue Glucose = 8.17, AccuData Easy = 8.58 and Encore
= 7.85 mmol/L, n = 96. After making a correction for whole
blood to plasma differences (add 10% to whole blood result for
the HemoCue Glucose and AccuData meters), absolute values
were compared to the Nova Stat Profile 5. Based on accuracy
recommendations from the Association of Clinical Biochemists10 and the Consensus Development Conference on Self
Monitoring of Blood Glucose6 (agreement of meter within 15%
of the central laboratory method), the One Touch II (3.7% outside ± 15%) and HemoCue Glucose (7.3% outside ± 15%) meters correlated better than the AccuData Easy and Encore QA
meters (31.8% and 41.7% outside ± 15% from the Nova). When
compared against medically useful criteria limits, 610 " the AccuData Easy and Encore QA would generate unacceptable deviations from our Nova results more than 30% of the time,
leading to changes in the clinical management of patients between the central laboratory and portable meter levels.
Specimen volume was examined to determine the minimum
required sample and volume dependent bias. The HemoCue
was not included in this experiment because the HemoCue
Glucose cuvette accepts a fixed 5 fiL volume. All meters demonstrated linear response between 15 /tL and 50 nL of a specimen. Minor variations in sample volume, thus, do not contribute to meter inaccuracy.
The effect of hematocrit was determined by comparison to
the Nova Stat Profile 5 which is not affected by hematocrits of
less than 70% (viscosity errors above 70%) (Fig. 3). HemoCue
Glucose (•) results matched the Nova response between 0 and
A.J.C.P.-February 1995
247
NICHOLS ET AL.
Evaluation of Portable Glucose Meters
40
35
30
"""'"A
•*ir^
"''
A
° 25
E
J
• / . . • • • '
„, 20
i"
•
>">i
(0
•
O
O 16
.
sirs
10
'*•••
""'
~
X?'
6
10
15
NOVA
ACCU'DATA
20
25
GLUCOSE
30
36
40
(mmol/L)
EASY
.
^C—K
* ^"
•
.
^s
"
v $ & "*
^
^
;
*
NOVA
OLUCOIE
QLUCOMBTtR
tomol/U
INCORI
NOVA
OA
OLUCOSS
ONE T O U C H I I
tenol/U
HOaMTAL
1jjlr
NOVA
QLUCOfe
bnnol/U
NOVA
~
f
~~
QLUC098
FIG. 1. (Top) Verification of high end linearity, range of 8.7 to 38.8 mmol/L glucose, using the Nova Stat Profile 5 as the laboratory reference. The
),
Nova linearity was validated to at least 33.3 mmol/L, by recovery of NBS glucose standards in a separate experiment. Normal line to Nova (
•
). All meters were linear to 30 mmol/
AccuData Easy (— • — ) , Encore QA (--- + ---), HemoCue Glucose ( - - • - - ) , One Touch II (
L except the One Touch II (A). Regression statistics were based on duplicates at each concentration level and represent the optimal least squares fit.
AccuData Easy = -1.24 + 1.05(Nova), range to 25.4 mmol/L, n = 10, r = 0.994, Sy.x = 0.52; Encore QA = -1.61 + 1.00 (Nova), range to 30.3
mmol/L, n = 12, r = 0.993, Sy.„ = 0.93; HemoCue Glucose = -2.15+1.11 (Nova), range to 16.9 mmol/L, n = 6, r = 0.997, Sy.x = 0.25; One Touch
II = -1.18 + 1.05 (Nova), range to 19.4 mmol/L, n = 8, r = 0.995, Sy.x = 0.49.
FIG. 2. (Bottom) Patient correlations (N = 96) for evaluation meters; AccuData Easy (Whole Blood), HemoCue Glucose (Whole Blood), Glucometer Encore QA (Plasma) and One Touch II Hospital (Plasma Mode), against the laboratory reference, Nova Stat Profile 5 (Plasma). Data from
only 1 instrument is shown. Normal line to Nova (
), least squares regression line (---). AccuData Easy = -1.48 + 1.11 (Nova), r = 0.943, Sy.„
= 1.13; HemoCue Glucose = -0.02 + 0.91 (Nova), r = 0.954, S yx = 0.82; Encore QA = -0.26 + 0.90 (Nova), r = 0.971, Sj..x = 0.63; One Touch II
= 0.18 + 1.00 (Nova), r = 0.975, Sy.x = 0.65.
Vol. 103-No. 2
248
CLINICAL CHEMISTRY
Original Article
30
,
T..
,
•
25
3
i 20
o
E
k.
'
v:
^'
m 16
J
•
*"*..
CO
o
u
•
•
*
o
1
20
40
60
HEMATOCRIT
80
(%)
60
I
,
;
!
40
O
a.
20
CO
<
-20
•40
-60
-10
10
20
30
PERCENT
40
60
60
70
80
HEMATOCRIT
FIG. 3. (Top) Effect of hematocrit on glucose result from portable bedside meters: AccuData Easy (— •
Encore QA (--- + ---) and One Touch II Hospital (
A
) compared to the Nova Stat Profile 5 (
X
), HemoCue Glucose (
).
•
),
FIG. 4. (Bottom) Hematocrit and glucose concentration bias of Boehringer Mannheim AccuData Easy: glucose 4.44 - 5.27 mmol/L (— O —), 11.11
- 22.22 mmol/L (—*—) and 5.56 - 16.67 mmol/L (—•—). The mean ± 2 standard deviations of duplicate determinations are noted by
bracketed points (Only one determination for — • — curve). The bias curve changes as the hematocrit passes 447o and glucose 11.11 mmol/L
(middle point in — • — curve).
A.J.C.P.-February 1995
NICHOLS ET AL.
249
Evaluation of Portable Glucose Meters
62% hematocrit. The Encore QA (+) results had an increasingly
negative bias with respect to the Nova above 20% hematocrit
and, like the Nova, was unable to analyze the 82% specimen,
due to high viscosity. The One Touch II ( A ) meter had good
agreement with the Nova between 20% and 41% hematocrit,
but was biased low above 41%. The AccuData Easy (•) had a
positive bias below 41% and a negative bias above 41% hematocrit. Further experimentation (Fig. 4) demonstrated that the
AccuData Easy bias is related to the amount of glucose present
in the specimen. If the glucose is below 11 mmol/L, the AccuData Easy shows a positive bias above 44% and a negative bias
below 44% hematocrit (O). For glucose values above 11 mmol/
L, the bias flips, positive below 44% and negative above 44%
hematocrit (*). The dependence of the AccuData Easy on both
glucose concentration and hematocrit is further demonstrated
in Figure 4 (•). Two specimens from the same patient were
produced ([1] 5.6 mmol/L at 23% hematocrit; and [2] 16.7
mmol/L at 68% hematocrit) and mixed at 1:3, 1:1, and 3:1 dilutions. AccuData bias switched as the glucose level crossed 11
mmol/L and hematocrit 44%. The other meters showed constant hematocrit bias at all levels of glucose. Based on ± 15%
laboratory correlation recommendations, 610 only the HemoCue Glucose analyzer demonstrated Nova comparable results
at all levels of hematocrit.
Bedside
Evaluation
Comparison of precision between laboratory and bedside
testing sites indicate a higher variance when multiple operators,
control and test strip lots are considered (Tables 2 and 3). Statistics analyzed by operator, meter serial number, and lots of
reagent (Tables 2 and 3) showed higher variance across operators, especially at lower concentrations (CV range 0.3%—19.5%
across all operators and glucose levels), whereas the One Touch
II precision was more consistent across different levels of glucose and operators (CV range 0.9%-9.9% across all operators
and glucose levels). This variability is related to operator effects
because differences of more than 5% CV were noted for AccuData Easy operators from a single nursing unit, using identical
meter, test strip, and control vials (Table 3). Analysis of the
AccuChek II meter (Table 2) showed comparable variations
among operators, especially in the low range. Individual statistics (Tables 2 and 3) provide an important means of evaluating
single operator, meter and reagent lot performance (Tables 2
and 3) and can be used for validation of operator skills as required by the Joint Commission on the Accreditation of
Healthcare Organizations.12
Data from the crossover group, operators testing both meters, were not significantly different from total operators, except
for the Ames Glucometer (Table 2). Removal of one operator
from the Ames Glucometer data set resulted in lower variance
(by more than half). Thus, at least three meters, the Ames Glucometer, the Accucheck II, and the AccuData Easy systems
have some dependence on operator technique. Published data
have also noted performance variations based on operator technique with other meters. 13 "' 5
Evaluation meters were compared to the nursing unit's current meter, predominantly the AccuChek II. The evaluation
meters were not compared to the Ames Glucometer because a
statistically relevant number of specimens were not gathered
within the evaluation time period. Correlation samples were a
mixture of capillary fingersticks, arterial and central venous
TABLE 2. BEDSIDE EVALUATION AND INDIVIDUAL
STATISTICS OF CURRENT METER QUALITY CONTROL
Glucose (mmol/L) (mean ± SD)
High
Low
Bedside evaluation: results of current meter quality control
Accu-Chek II (n = 219/223)
17.37 ± 1.37(7.9) 3.69 ±0.48 (12.9)
Glucometer (n = 26/25)
15.37 ±0.67 (4.3) 2.55 ±0.45 (17.6)
Accu-Chek Easy (n = 20)
13.06 ±0.54(4.1) 2.19 ±0.29(13.4)
Crossover group (n = 25)
Accu-Check II (n = 157/161)
Glucometer (n = 16)
Accu-Chek Easy (n = 20)
17.23 ± 1.28(7.4)
3.6 ±0.46 (12.4)
.28(7.4) 3.67
15.48 ± 0.34
(2.2) 2.4
""""""
7.43 ±0.12 (4.8)
13.06 ±0.54(4.1) 2.1'9 ±0.29 (13.4)
Control lot number
Accu-Chek II 55123 (single lot of test strips evaluated)
Unit 55123(n = 16/19)
16.93 ± 1.32(7.8) 3.53 ±0.62(17.5)
Total 55123 (n = 80/83)
17.58 ±1.12 (6.3) 3.70 ±0.41 (11.0)
Unit 253011 (n = 5/5)
14.99 ±0.72(4.8) 2.98 ±0.32(10.7)
Total 253011 (n = 9/10)
15.49 ±0.81 (5.3) 2.98 ±0.32(10.7)
Meter serial number
Accu-Chek II 4267/220
Unit(n = 21/24)
Total (n = 28/32)
16.47 ± 1.46(8.9) 3.42 ±0.61 (17.8)
16.57 ±1.42(8.6) 3.33 ±0.60(18.0)
Operator ID
Accu-Chek II
32 (n = 6/7)
33 (n = 6/8)
35 (n= 10)
17.18 ± 1.57(9.1) 3.52 ±0.69 (19.6)
16.44 ±1.18(7.2) 3.27 ±0.72 (22.0)
16.03 ±1.43 (8.9) 3.39 ±0.49 (14.7)
CV values are given in parentheses (%).
line specimens. The AccuData Easy system correlated with the
AccuChek II (AccuData Easy = 0.302 + 0.995 AccuChek II; n
= 97; r = 0.920; Sy.x = 1.817 mmol/L; AccuChek II mean =
8.71 mmol/L; AccuData Mean = 8.97 mmol/L). However,
39% of the values fell outside 15% from the AccuChek II target
value with about a third of these more than 30% (15% of specimens were 30% outside the Accuchek II value), which is double
the recommended bias.6,10 When only capillary fingersticks
were compared, the AccuData Easy correlation statistics improved (AccuData Easy Capillary = —0.150 + 1.021 AccuChek
II; n = 78; r = 0.954; Sy.x = 1.453 mmol/L; AccuChek II capillary mean = 9.10 mmol/L; AccuData capillary mean = 9.14
mmol/L). Capillary specimens (36%) also had a comparable
number of values falling outside the AccuCheck II target.
However, One Touch II, demonstrated some bias to the AccuChek II meter (One Touch II = 1.281 + 0.888 AccuChek II;
n = 102; r = 0.946; Sy.x = 1.532 mmol/L; AccuChek II mean
= 8.59 mmol/L; One Touch II mean = 8.91 mmol/L). The
One Touch II also had a number of results more than 15% from
the AccuChek II (32%), but had fewer 30% outliers (7%) when
compared to the AccuData Easy/AccuChek II comparison.
These correlation statistics were not affected by sample type
(One Touch II capillary = 1.283 + 0.888 AccuChek II; n = 92;
r = 0.946; Sy.x = 1.572 mmol/L; AccuChek II capillary mean
= 8.67 mmol/L; One Touch II capillary mean = 8.98 mmol/L)
(30% of capillary results were greater than 15% difference from
Vol. 103-No. 2
CLINICAL CHEMISTRY
250
Original Article
TABLE 3. BEDSIDE EVALUATION AND INDIVIDUAL STATISTICS OF EVALUATION METER QUALITY CONTROL
Glucose (mmol/L) (mean ± SD)
Mid
Low
Bedside evaluation: results of evaluation meter quality control
Easy(n= 138/124/136)
13.73 ±0.73 (5.3)
O Touch (n = 129/129/132)
17.51+ 0.91 (5.2)
5.05 ±0.29 (5.9)
5.66 ±0.24 (4.2)
2.16 + 0.29(13.3)
2.84 ±0.15 (5.2)
Crossover group (n = 25)
Easy(n= 102/97/102)
O Touch (n = 93/96/96)
13.75 + 0.80(5.8)
17.52 ±0.89(5.1)
5.01 ±0.25(5.0)
5.67 ±0.24 (4.3)
2.15 ±0.27(12.4)
2.84 ±0.16 (5.7)
Control lot number
Accu-Data Easy 200536 (single lot of test strips evaluated)
Unit(n= 10)
13.64 ±0.46 (3.3)
Total (n = 23/16/23)
13.71 ±0.46(3.3)
4.92 ±0.34 (6.9)
4.93 ±0.30 (6.1)
1.97 ±0.19 (9.5)
2.00 ±0.29 (14.6)
Test strip lot number
One Touch II 301779 (single lot of controls evaluated)
Unit(n = 15)
17.42 ±0.56 (3.2)
Total (n =•• 103/103/106)
17.51 ±0.92(5.2)
5.75 ±0.16(2.7)
5.65 ±0.25 (4.5)
2.84 ±0.07 (2.4)
2.84 ±0.16 (5.6)
13.67 ±0.49 (3.6)
4.94 ±0.31(6.2)
2.03 ±0.24(12.0)
17.42 ±0.56 (3.2)
17.11 ± 1.09(6.4)
5.75 + 0.16(2.7)
5.57 ±0.26 (4.7)
2.84 ±0.07 (2.4)
2.80 ±0.24 (8.5)
13.44±0 (0)
13.57 ±0.37 (2.7)
13.76 ±0.59 (4.3)
5.06 ±0.16(3.1)
4.98 ±0.21(4.2)
4.82 ±0.46 (9.5)
1.97 ±0.19(10.0)
2.02 ±0.14 (6.9)
1.93 ±0.24(12.3)
17.59 + 0.51(2.9)
17.44 ±0.59 (3.4)
17.22 ±0.65 (3.8)
5.70±0.18(3.1)
5.73 ±0.18 (3.2)
5.79 ±0.12 (2.0)
2.83 ±0.07 (2.4)
2.84 ±0.08 (2.9)
2.86 ±0.07 (2.5)
High
Meter serial number
Accu-Data Easy GT0519
Unit = total (n = 15)
One Touch II FCL40E9B
Unit(n= 15)
Total (n = 20/23/23)
Operator ID
Accu-Data Easy
32(n = 2)
33(n = 3)
35(n = 5)
One Touch II
32(n = 5)
33(n = 5)
35 (n = 5/4/4)
CV values are given in parentheses.
the AccuChek II capillary value whereas only 7% were more
than 30% from the capillary AccuChek II). Thus, the AccuData
Easy system, which is FDA approved for only capillary blood,
has significant interference from non-capillary specimens. Although both the AccuChek II and the One Touch II meters
are approved for all types of specimens, the two meters yield
different answers for many patients. Because the One Touch
II meter demonstrated excellent correlation in the laboratory
evaluation to the Nova Stat Profile 5, these data indirectly suggest that the AccuChek II systems, which are the predominant
meters currently in use in our hospital, are yielding acceptable
answers (within 15% of the central laboratory) only about 60%
to 70% of the time. These data demonstrate the importance of
initial method validation with ongoing laboratory correlations
as part of a decentralized quality assurance program.
Responses from the operator surveys provided limited but
important information regarding current bedside testing practices in the hospital. Operators were asked to rate their preference for ease of use, training, maintenance, infection control,
and data management features using a 1 (least desirable) to 5
(most desirable) Likert scale. No difference was found between
the AccuData Easy and One Touch II responses with either the
total or cross-over groups. The responses clearly indicated a
preference for either evaluation meter over the current meters.
In addition, operators preferred a performance skills test over a
written examination. Quality control issues were also reinforced. Although most operators check the expiration dates of
test strips before analysis, many do not consistently update this
date when opening a new vial. This is a concern because most
test strips expire 1 to 4 months after opening a sealed vial.
Time analysis indicates that either evaluation meters (AccuData Easy 5:16 (minutes:seconds), n = 12; One Touch II 6:55,
n = 22 for 3 levels of control) would decrease the amount of
time required to perform QC, maintenance, and a patient test
over the current meters (Accu-Chek II 9:11 (minutes:seconds),
n = 25; Accu-Chek Easy 3:35, n = 4; One Touch II personal 3:
45, n = 1; Ames Glucometer 5:25, n = 1 for 2 levels of control).
Conversion to the One Touch II would save 2 minutes on these
tasks, whereas the AccuData Easy system would save approximately 4 minutes compared to our current Accu-Chek II meters, despite the evaluation meters analyzing three levels of control compared to only two levels on the AccuChek II meter.
A.J.C.P.-February 1995
NICHOLS ET AL.
251
Evaluation of Portable Glucose Meters
DISCUSSION
This investigation demonstrates that none of the evaluated
glucose meters have optimal technical performance, user and
data management features for hospital application, despite
their principal market share in near-patient testing. Meters
manufactured for in-home, fingerstick use by a single patient
cannot be expected to perform identically when exposed to hospital conditions. Multiple operator techniques, patients, disease populations, hematocrits, hydration states, and the use of
noncapillary specimens all contribute to increased result variability under hospital operation of meters. In-home performance characteristics are not indicative of how the meter will
actually perform in any given institution. Clinicians must consider their sampling practices, patient characteristics, and meter limitations when interpreting near-patient glucose results.
In addition, technical performance alone is insufficient to allow
selection of a device for hospital use. Ease of operation, infection control, cost-effectiveness, reagent stability, and data management/computer automation are issues of equal importance
to the clinical staff acceptance of bedside glucose meters.
Our task force decision to select a single meter for use in our
institution was based not just on laboratory technical performance, but was also influenced by ease-of-use, patient education, and ability to analyze non-capillary specimens. These are
important issues to our nursing staff. The results of this evaluation demonstrate that FDA approval and manufacturer data
are only indicative of what an instrument is capable of achieving under optimum operating conditions. We strongly recommend that each institution set minimum standards for meter
performance and independently verify that each meter is capable of achieving those standards under routine operating conditions. Despite the "waived" status of these instruments under
CLIA '88, portable glucose meters should be evaluated both in
the laboratory and at the bedside. Enlistment of nurse operators during the initial evaluation allows operators to voice their
opinions and provides better acceptance of the QA program
revisions. Our evaluations uncovered many institutional quality assurance deficiencies that were addressed by our interdepartmental task force. The conclusions of this investigation
have resulted in the total revision of our QA program, limitations in the number of operators, standardization of operator
training, validation of operator skills proficiency, selection of a
single manufacturer for hospital-wide use, and ongoing program monitoring by a central office. This evaluation has not
only enhanced collaboration between clinical, laboratory, and
administrative staff, but has demonstrated the limitations of
meter results to our clinical staff, improving the overall quality
of patient care.
REFERENCES
1. Chernow B. The bedside laboratory: A critical step forward in ICU
care. Chest 1990;97:183S— 184S.
2. Misiano DR, MeyerhoffME, Collison ME. Current and future directions in the technology relating to bedside testing of critically
ill patients. Chest 1990;97:204S-214S.
3. Zaloga GP. Evaluation of bedside testing options for the critical
care unit. Chest 1990;97:185S-190S.
4. Trundle DS, Weizenecker RA. Capillary glucose testing:A costsaving bedside system. Lab Management 1986; 24:59-62.
5. Latman NS. Evaluation of electronic, digital blood glucose monitors. BiornedInstrum Technol 1991;25:43-49.
6. Consensus Development Conference. Consensus statement on
self-monitoring of blood glucose. Diabetes Care 1987; 10:95-99.
7. Colagiuri R, Colagiuri S, Jones S, Moses RG. The quality of selfmonitoring of blood glucose. DiabetMed 1990;7:800-804.
8. Godine JE, Hurxthal K., Nathan DM. Bedside capillary glucose
measurement by staff nurses in a general hospital. Am J Med
1986;80:803-806.
9. Ashworth L, Gibb I, Alberti K.GMM. Hemocue: Evaluation of a
portable photometeric system for determining glucose in whole
blood. Clin Chem 1992;36:1479-1482.
10. Price CP, Burrin JM, Nattrass M. Extra-laboratory glucose measurement: A policy statement. Diabet Med 1988;5:705-709.
11. Skendzel LP, Barnett RN, Piatt R. Medically useful criteria for
analytic performance of laboratory tests. Am J Clin Pathol
1985;83:200-205.
12. The Joint Commission on the Accreditation of Healthcare Organizations. Accreditation Manual for Pathology and Clinical
Laboratory Services. Oakbrook Terrace, IL: JCAHO, 1993.
13. Portable Blood Glucose Monitors: Evaluation. Health Devices
1992;21:43-78.
14. Phillipou G, Farrant RK, Phillips PJ. Computer based quality assessment of hospital capillary blood glucose monitoring. Diabet
Med 1990;7:234-237.
15. Hilton BA. Nurse's performance and interpretation of urine testing and capillary blood glucose monitoring measures. J Adv
Nurs 1982;7:509-521.
Vol. 103-No. 2