26 Feb 1985 ... (1,2); consequently, the instruments are usu- ... observed, however, that either
instrument .... To determine if use of serum control material as a.
CLIN.CHEM. 31/8,1349-1352(1985)
Preparationand Use of Serum-BasedMaterialas Controland Calibratorin EvaluatingIon-SelectiveElectrodesfor Calcium John Toffalettl”2 and Kin Man Lee1 We analyzed 93 sera and 81 whole-blood samples, plus several aqueous materials and one serum-based material as controls,with two analyzers for ionized calcium (Radiometer
ICA 1 and AVL 980). On 16 days, comparisonsof resultsfor patients’samples were good, with nearlyall between-instrument differences (Ca2) of samples being within 0.04 mmol/L. On 12 days, comparisons weremediocre,withthe Ca2 of samples usually0.04 to 0.10 mmol/L The control materialthatmost consistentlyindicatedthe direction and magnitude of the Ca2 ofpatients’ sampleswastheserumbased control. Adjustment of results for samples from pa-
tients, based on the iCa2 of the serum controlon the same day, substantially improved comparisons between instruments. Our findingssuggest that the use of a serum-based calibratorprovidesthe mostvalid comparisonbetweendifferent calcium-ion analyzers and reliably indicates when electrode replacementis needed. Published comparisons of ionized calcium results by analyzers from different manufacturers typically show marked discrepancies (1,2); consequently, the instruments are usually evaluated individually or compared with older instruments from the same manufacturer (3-5). Efforts to prepare primary standards have been directed toward obtaining better uniformity and intercomparability of results among ionized calcium analyzers from different manufacturers (6). The two ion-selective electrode systems evaluated here are of different designs. The Radiometer ICA 1 ionized calcium analyzer operates at 37#{176}C, an ion-exchange material [calcium di(4.octylphenyl)phosphate] is the electrode sensor, and a cellophane membrane covers the sensor, which improves durability but slows the response time. The AVL 980 utilizes a neutral carrier molecule (alkyl derivative of 3,6dioxaoctane diamide) as calcium ion sensor and measures ionized calcium at room temperature. This instrument has very low sample requirement and excellent throughput, but the durability of the electrodes may be relatively low. In a published comparison of these analyzers (7), the AVL reportedly gave lower values than the Radiometer; we have observed, however, that either instrument may give the lower value. Here we report our comparison of the results with serum and whole-blood samples and our studies of the use of aqueous and serum-based materials as calibrators to achieve equitable comparisons between analyzers and as controls for determining acceptable performance of an instrument. 1Clinical Chemistry Laboratory and2 Department of Pathology, Duke University Medical Center, Durham, NC 27710. Received February 26, 1985, accepted May 9, 1985.
Materials and Methods Instruments We used two automated flow-through calcium-ion selective electrode systems. The Radiometer ICA 1 (Radiometer America, Inc., Westlake, OH 44145) analyzer controls temperature at 37#{176}C, requires 110 pL of sample, has a 3-mm cycle time, and also measures pH and provides the ionized calcium corrected to pH 7.4 (Ca7.4) (3). The AVL 980 (AVL Scientific Corp., Pine Brook, NJ 07058) operates at ambient temperature, requires 60 pL of sample, has a 60-s cycle time, and also measures sodium and potassium concentrations. Both instruments were calibrated with standard solutions supplied by the respective manufacturer. During every analytical run we analyzed the Radiometer aqueous controls and the serum-based control (see below) with both instruments; we also analyzed the Radiometer calibrators with the AVL instrument.
Reagents The compositions of the Radiometer calibrating solution 1 (Ca2 = 1.25 mmolJL, pH = 7.38 at 37 #{176}C) and calibrating solution 2 (Ca2 = 2.50 mmol/L, pH 6.84 at 37 #{176}C) are described in another report (3). Standard A (AVL). The concentrations in mmol/L, as stated by the manufacturer, are: calcium chloride 1, sodium chloride 150, potassium chloride 5, and triethanolamine 1.5. Standard B (AVL). The concentrations in mmol/L, as stated by the manufacturer, are: calcium chloride 4, sodium chloride 50, potassium chloride 1.8, and triethanolamine 1.5. Calcium controls (Radiometer). Composition, mmol/kg water, for low- (and high-) concentration controls: calcium chloride 0.9 (2.0), sodium chloride 108 (105), 2-{[2-hydroxy1,1-bis(hydroxymethyl)ethyl]amino}ethanesulfonic acid (TES) 71.2 (102), at pH 7.55 (7.25). Serum-based control. Pool two- to three-day-old sera from patients’ specimens, and acidil& by adding 10 L of 1 mol/L HC1 per each milliliter of serum. Leave open in a refrigerator for seven to 10 days, or magnetically stir for one day in a refrigerator. Adjust the pH to about 7.4 at 37 #{176}C with 1 mol/L NaOH or HC1, as needed. Then mix, ifiter through ordinary ifiter paper, and store (-20 #{176}C) in sealed 1-mL plastic syringes. Thaw individual syringes when needed for an analysis, mixing carefully to overcome the separations of protein that occur with freezing.
Results Between-day precision of control materials. The summary of precision data in Table 1 shows the Radiometer ICA 1 was consistently more precise than the AVL 980. However, as CLINICALCHEMISTRY,Vol. 31, No. 8, 1985
1349
of the change per month in ionized calcium. The rest of the decrease may be from micro precipitation of ionized calcium from the serum control material. Comparison of results on serum and whole-blood samples. Either serum or whole-blood samples, with ionized calcium results ranging from 0.63 to 2.06 mmol/L, were analyzed on 28 days during the study. On 16 of those days, the Ca2 of samples from patients was low, with nearly all results on a sample differing by 0 to 0.04 mmoIJL. On the other 12 days, the Ca2 was higher, with nearly all patients’ results differing by 0.04 to 0.10 mmol/L. As Figure 2 shows, the only mmol/L
Table 1. Between-Day Precision of Control Materials Ionized calcium,
mmol/L Control material
instrument
AVL Radiometer AVL Radiometer AVL
Serum
Mean 1.20
Serum AqueouslowCa AqueouslowCa Aqueoushigh Ca
1.23 0.73 0.79 1.66
SD
CV, %
0.048
4.0 2.0 2.9 2.0
0.025 0.021 0.016
Radiometer n
=
Aqueous high Ca 39 analyses on dIfferent days.
consistent indicator of whether the Ca2 of results on patients’ samples (both serum and heparinized whole blood) was low or high was the Ca2 of the serum-based control material. To determine if use of serum control material as a calibrator would have lowered the Ca2 of patient samples, we added or subtracted the Ca2 of the serum control on that day from each of the results for patients’ samples. This adjustment substantially decreased both the S, and the means (± SD) of the Ca2 of results for patients’ samples (Table 2).
indicated by Figure 1, we might have been able to attain better precision by installing a new electrode on the AVL 980 when a change in the electrode response was noted. We especially wanted to determine if a control material could accurately indicate when to install a new electrode. Inspection of the daily between-instrument difference (Ca2) in means of patients’ results (Figure 2) indicated that a large shift occurred after day 12 and day 26. Of two aqueous controls and the serum-based control, the serum material best indicated that shifts had occurred (Figure 2) and that these shifts were due to changes in the AVL 980 instrument (Figure 1). As analyzed by the Radiometer ICA 1, the serum control material showed a gradual decline in ionized calcium of about 0.01 mmol/L per month, the study shown in Figure 1 having been done over a five-month period. This decline did not change when a new electrode was installed just before day 22. During the study, the pH of this material increased by about 0.006/month, which would account for about 0.003
.
Discussion Our study indicates that use of a serum-based calibrator could substantially improve agreement of ionized calcium results for either serum or whole blood analyzed by different instruments that agree only relatively poorly when aqueous calibrators are used. The Radiometer and AVL analyzers have distinctly different electrode design and sensor materials, flow-system geometries, and temperature-control mechanisms, yet apparently would have given comparable results if calibrated with the serum-based material. With none of the aqueous calibrators or controls would the Radiometer
1.30 S S
1.20
S
#{149}
S
. #{149}
S
S
#{149} #{149} S
SERUM:AVL
#{149} S #{149} .
S
S
#{149} #{149} .
.
S
#{149}
1.10
1.30 S
S
#{149} #{149} S #{149} #{149}
#{149} S
#{149} #{149} #{149} #{149} #{149}
1.20
S
SERUM #{149} RAD #{149} .
#{149} #{149}
#{149}
.
E
.
.
.
S
#{149} #{149} #{149}
S
#{149} #{149} #{149} #{149} #{149} #{149} S
0.85
LO CALCM #{149}S S
#{149}S
#{149} #{149}#{149}
.
S
.
S
#{149} S
.
S
+
+
#{149} #{149}+S
.
.
S
#{149} #{149} #{149}
.
#{149} S
#{149}
S
S
#{149} S
#{149} #{149}
#{149}#{149}
S
#{149} S
#{149} #{149} #{149}
#{149},
S
0.75
#{149}
S
.
LOCALCM:AVL
0.70
.
.
0.80
E
S
.
#{149} -
#{149} #{149} #{149} #{149} #{149} U
#{149} #{149} #{149} #{149} #{149} #{149} S
#{149} S
: RAD #{149} #{149} #{149} #{149} #{149} S
S
#{149} #{149} S
S
#{149} #{149}
#{149}+S
S
1.75
C-)
S
IS
1.65 1.55 1.80
I
#{149}
S
#{149} #{149} #{149} S
#{149}
HICALCM:AVL
. .
+
S
#{149} S #{149} #{149} #{149}
U
S S
.
S
S S
S
#{149}
#{149}
#{149}
1
#{149}
S
S
2 A
.
B C
3
S
4
6
S
S
#{149} S
#{149}+5
HICALCM:RAD
#{149} #{149} #{149} #{149}
S
1.70
S
#{149}
S
8
#{149} #{149} S
10
12
#{149}
14
#{149} #{149} #{149} #{149} . 16
18 D
S
.
#{149} #{149} #{149} S
20
-
E F G
S
S
S
H
I
J
#{149}#{149} $ #{149}
K 22
24
26
28
Day FIg. 1. Plots of ionized calcium results forthecontrol materials M,mbered daysindtoatedayswhenbothcontrolsand patientssampleswereanalyzed,andcorrespond to thenumbering ofdaysin Fig.2. Lettered days indate days whenonlycontrolswere analyzed.Mows indtoatewhen a newelectrodewas installed.LOCALCM:Radiometertow calcium control; HI CALCM: Radiometerhigh calciumcontrol 1350
CLINICALCHEMISTRY, Vol.31,No.8,1985
+0.1 S
mean of patient results
.
S
#{149}S
S
0
-
#{149} S S
+0.1 0
serumQC
S
.
-
S
S
S
S
a
S
S S
S
S
S #{149} S
S S
-0.1
-J
S
S
S S
S
0
S
E E
low
QC
.
S #{149}
-0.1
S
S
S
S
S
S
#{149} S
S
#{149}
S
S
S
#{149}#{149}
S
V CD
0
#{149}#{149}#{149}S
S
#{149}
S
S
#{149} S
high QC
-0.1
S
#{149}
S
S S
#{149}
#{149}
-
S
#{149}
S
S S
+
0
01 CD
S
S
-
5
S
1.25 std
#{149} #{149} #{149}
/
S
S
S
S
S
S
0
-0.1
S
S
#{149}
S
S
0
#{149} #{149} #{149}
S S
2.50 std -0.1
‘p....--
S
S
S
S
S
S
S
S
. S
#{149}
S
S S
.
-0.2 #{149}
1 2
S S S
4
6
8
10
14
12
16
18
20
22
24
26
28
Day FIg.2. Plotsof the between-instrument differences (ACa2) in ionizedcalcium results for the dailymean of patients’results and of thecontrol matenals Pointsinsquaresindicateresultsfor whole-bloodsamples.Low OC: lowcalciumcontrol;highOC:high calciumcontrol;1.25sat and 2.50sat: Radiometer1.25and2.50 mmot/Lstandards
Table 2. Comparison of ionized Calcium Results with the AVL and Radiometer Analyzers t Ca2 Specimen
All
Serum Wholeblood
determining
Mean
fl
173
0.045
92
(0.027) 0.044
81
(0.024) 0.047 (0.030)
0.038 (0.023) 0.037 (0.025) 0.038 (0.022)
serum-based material also appears best for determining when replacement of an electrode is needed, as concluded elsewhere (8). Potential problems with the serum-based material are (a)
concentrations of ionized calcium in that will give results for patients’ samples that are consistent over time, and (b) ensuring the stability of the pools used in preparing the controls. Development of a reference or consensus method with a primary calibrator used to determine set points on the serum calibratore should give uniformity between preparations. Concerning stability, our serum-based material lost ionized calcium at about0.01 mmol/L per month, as judged by results from theRadiometer ICA 1. The increase in pH of about 0.006 per month, due to loss of C02, partly accounts for this, and a slow precipitation of calcium carbonate and calcium phosphate may also occur. Perhaps this could be prevented by removing bicarbonate and increasing the concentration of a soluble chelator such as citrate. A previous report (7) stated the need for standardization of calibration solutions for instruments based on ion-selective electrode technology. Our results show that a serumbased calibrator can give uniform measurements from two different
(0017) o
(0.018) 0.035 (0.016)
#{149}For the correlation of theAVLresults(y)with theRadiometerresult (4 bPsolute value of the between-instrument difference of results for a sample. CData in parenthesesreflectadjustmentof corresponding values by subtractionof that day’sCa2 (AVL - Radiometer)forthe serum-basedcontrol fromeach result.
and AVL analyzers have consistently and uniformly agreed regarding results for patients’ samples. This has important implications for developing a uniform standard for ionized calcium measurements (6), in that an aqueous standard may not be satisfactory as a secondary calibrator. The
set-point
preparations
CLINICALCHEMISTRY, Vol. 31, No. 8, 1985 1351
dissimilar instruments for ionized calcium in either serum or whole blood. Clinically, a uniformity of ionized calcium results among different instruments and laboratories should lessen the variability in reference ranges and make interpretation of these results easier. Thus, the discrepancies of ionized calcium results between instruments may be due more to the suitability of the calibrating solution used than to electrode design, type of sensor, temperature of measurement, and possibly even the age of the electrode.
References 1. Larsson L, Finnstrom 0, Nilsson B, Ohman S. Evaluation of Radiometer ICA 1 as routine instrument for serum ionized calcium and its application for whole blood capillary samples from newborn infants. Scand J Clin Lab Invest 43 (Suppi 165), 21-26 (1983). 2. Drop LI, Tochka LN, Misiano DR. Comparative evaluation of
two calcium ion-selective monitoring (1982).
steady-state
electrode systems, and their utility for in Ca2. Clin Chem 28, 129-133
changes
3. Fogh-Andersen N. Ionized calcium analyzer with a built-in pH correction. Clin Chem 27, 1264-1267 (1981). 4. Smith SCH, Buckley BM, Wedge G, Bold AM. An evaluation of the ICA lionized calcium analyzer in a clinical chemistry laboratory. Scand J Clin Lab Invest 43 (Suppl 165), 33-37 (1983). 5. Husdan H, Leung M, Oreopoulos D, Rapaport A. Measurement of serum and plasma ionic calcium with “Space-Stat 20 Ionized Calcium Analyzer”. Clin Chem 23, 1775-1777 (1977). 6. Crowell JA, Bowers GN Jr. Apparent binding of ionized calcium by various buffers. Clin Chem 31, 267-270 (1985). 7. Boink ABTJ, Gimpel JA, Mass AHJ. A comparison of four Ca2 analyzers. Saznd J Clin Lab Invest 43 (Suppl 165), 17-19 (1983). 8. Drop U, Misiano DR. Tochka LN. Commercial protein-containing solutions for quality assurance of [Ca2J measurements. Clin Chem 28, 2448 (1982). Letter.
CLIN. CHEM. 31/8,1352-1354 (1985)
Adequacyof InterlaboratoryPrecisionCriteriain MeasuringIntralaboratory Performance Sharon S. Ehrmeyer1and Ronald H. Laessig2 We compared the predictivevalue ofthe variouscriteriaused for grading pH and blood-gas measurements in interlabora-
proficiencytestingprogramswith performanceas determined from actual intralaboratoryquality-assurancedata. The evaluationcriteriawere the two-standard-deviationinterval (2 SDI) proposedbythe Collegeof AmericanPathologists (CAP); CAP’s proposed fixed criteria; and the fixed criteria of tory
the American Thoracic Society (ATS). These were compared with 95% confidencelimitsderivedfromthe individuallaboratories’ actual intralaboratory data. We found that the CAP’s most-stringent criterion (2 SDI) overestimated the number of
outliers(unacceptableresults)for
and 2’
whereasthe
proposed fixed limits underestimated them. For pH and O2 the ATS’s limits, which are more stringent, more closely match the individual laboratory’s actual performance as measured by conventional (mean ± 2 SD) intralaboratory quality-assurance practices. Additional Keyphraees: pH
blood gases
Interlaboratory performance comparisons (including proficiency-testing programs) are an integral part of most laboratories’ overall quality assurance in pH and blood-gas analyses. They serve as a basis for accuracy control, selfevaluation, licensure, and (or) accreditation (1,2). However, it has been long suspected that proficiency-testing results and actual intralaboratory performance are not well correlated (3). The problem of defining “acceptable” performance in pH/blood gas analyses is compounded by the unavailabil1 School of Allied Health Professions, Rm. 6169, 1300 University Ave., University of Wisconsin-Madison, Madison, WI 53706. 2Depssents of Pathology and Laboratory Medicine, and Preventive Medicine, University of Wisconsin-Madison, and State Laboratory of Hygiene, University of Wisconsin-Madison, Madison, WI 53706. Received January 2, 1985; accepted May 10, 1985.
1352
CLINICAL CHEMISTRY, Vol. 31, No. 8, 1985
ity of reference materials and methods (4). As a result, in most survey programs, including that of the CAP, the group mean value and a plus or minus two-standarddeviation interval of participants (2 SD!) are used to evaluate a laboratory’s performance (5, 6). Recently, the CAP Survey Committee proposed an evaluation criterion based on the group mean and fixed limits (7). The rationale is that, when 2 SDI limits calculated from specific instrument group values are applied, laboratories that are using newer instruments with inherently better precision are often judged to have “unacceptable” performance whereas laboratories having older, less precise instrumentation-thus producing poorer results-are “acceptable.” The new CAP fixed criteria, chosen to reflect the levels of accuracy required for satisfactory patient care, also eliminate this grading bias against certain instrument groups. In the ATS interlaboratory survey a variable limits approach similar to the 2 SD! was also used, but now ATS is evaluating more-stringent fixed criteria, based on established clinical needs, as a means of ensuring that laboratories will produce medically useful test results (8).
Materials and Methods Using actual interlaboratory survey data, we compared the response to four evaluation criteria used in interlaboratory surveys (two variable and two fixed) with the participating laboratories’ performances as measured by intralaboratory quality assurance. Each criterion was assessed on the basis of its ability to identify data from individual laboratories that reflected “out of control” performance. The CAP 2 SDI, or variable, intervals were selected from results of CAP’s 1982 interlaboratory survey, in which fluorocarbon-based emulsion-type specimens were used (Table 1) (6). 3Nonstandard abbreviations: CAP, College of American Pathologists; ATS, American Thoracic Society; 2 SD!, two-standard-deviation interval.
