6H2& solution, stock. Dissolve 77 g of the salt in 0.18 molIL sulfuric acid and dilute to 1 L with the same solvent. CoSO4(NH4)2S04 solutions, working. Using.
CUN. CHEM.
27/10,
1651-1654
(1981)
A Candidate Reference Method for Determination of Total Protein in Serum II. Test for Transferability Basil T. Doumas (Chairman),’ David D. Bayse,2 Klaus Borner,3 Richard J. Carter,4 Frank Eievitch,5 Carl C. Garber,6 Robed A. Graby,7 Lawrence L. Hause (statistical analysis),1 Alan Mather,2 Theodore Peters, Jr.,8 Royden N. Rand,7 Dennis J. Reeder,9 Steven M. Russell,5 Robert Schaffer,9 and James 0. Westgard#{176} The transferability of the candidate Reference Method for total serum protein was tested in eight laboratories in the United States and Europe. National Bureau of Standards
SRM 927 (bovIne serum albumin) was used in each analytical run as the calibration standard. The mean absorptivity value obtained for this material was 0.2983 L g cm. Four serum pools prepared at the Centers for Disease Control were analyzed on each of 15 days. Within-run variation of the protein values (expressed as CV) in the eight laboratories ranged from 0.1 to 2.5% and day-to-day (total) variation in six of the laboratories ranged from 0.4 to 1 %. The purpose of the transferability test was to assess the performance of the candidate Reference Method (1) in a number of laboratories. Specifically, we evaluated the accuracy, precision, linearity, and analytical recovery of the method. Eight laboratories-seven in the U.S. and one in Europe-participated in the study, which was conducted between January and June 1978 according to a detailed protocol developed by the Study Group. Included in this protocol were procedures for checking the performance of spectrophotometers and pipetting devices.
Materials and Methods Spectrophotometry The following All instruments
tests were performed by each participant. used had a bandpass of less than 8 nm. Wavelength calibration. This was verified by use of a holmium oxide glass filter (Arthur H. Thomas Co., Philadelphia, PA 19105; cat no. 8470-FlO) or the spectral emission lines (486 and 656 nm) of hydrogen or deuterium lamps. Photometric accuracy. Although we did not intend to establish any absorptivity constants in this study, we nevertheless preferred to make measurements in instruments with good photometric accuracy. Furthermore, we believed it would be desirable to evaluate the range of absorbance (A) values of SRM 927 in the biuret reaction. The recommended proceMedical College of Wisconsin, Milwaukee, WI 53226. 2 Centers for Disease Control, Atlanta, GA 30333. Free University of Berlin, West Berlin, F.R.G. DuPont de Nemours Co., Wilmington, DE 19898. Mount Zion Hospital, San Francisco, CA 94115. 6 University of Wisconsin, Madison, WI 53792. 1
‘
Eastman
Kodak
Co., Rochester,
NY 14650.
The Mary Imogene Bassett Hospital, Cooperstown, NY 13326. National Bureau of Standards, Washington, DC 20234. Prepared for the Committee on Standards of the American Association for Clinical Chemistry by the Study Group on Proteins. Part of this work was presented at the 1979 National Meeting of the AACC, New Orleans, LA (Clin. C/tern. 25: 1072, 1979, abstract). Received June 22, 1981; accepted June 25, 1981.
dures called for use of (a) glass filters (SRM 930) from NBS and (b) potassium dichromate solutions (2). Photometric linearity. Use of one of the following procedures was recommended to each participant: A. CoSO4(NH4)2S04 ‘ 6H2& solution, stock. Dissolve 77 g of the salt in 0.18 molIL sulfuric acid and dilute to 1 L with the same solvent. CoSO4(NH4)2S04 solutions, working. Using one 5-mL volumetric pipet, add 5, 10, 15 and 20 mL, respectively, of the stock Co(II) solution to four 25-mL volumetric flasks. Dilute the contents of each flask to volume with 0.18 mol/L sulfuric acid and mix well. Measure the absorbances of the working and stock solutions at 510 nm vs 0.18 mol/L H2S04.
B. Cyanmethemoglobin
solutions.
Centrifuge
about 5 mL
of blood, collected with EDTA as anticoagulant, aspirate, and discard the plasma. Add 1 mL of erythrocytes to 100 mL of Drabkin’s reagent (3), mix, and let stand for 5 to 10 mm. Centrifuge the solution to remove the “stroma” and without disturbing the sediment, transfer the supernatant fluid into a flask. Measure the absorbance of the supernate vs Drabkin’s reagent at 540 nm. Adjust the absorbance of the solution to between 0.92 and 0.96 by diluting with Drabkin’s reagent, and use this as the stock solution. Prepare cyanmethemoglobin working solutions from the stock solution as described for the Co(II) working solutions. Use Drabkin’s reagent as diluent and measure the absorbance of the working and stock solutions at 540 nm vs Drabkin’s reagent. Criteria for acceptable linearity. Evaluate the linearity of the spectrophotometer by linearregression analysis of the data (y = absorbance and x = relative concentrationof test solution, i.e.,1.0, 0.8, 0.6, etc). The linearityisconsidered acceptable when (a) Pearson’s coefficient(r2) is greaterthan 0.999 and (b) the concentrationscalculatedfrom the regression equation are within ±1% of the nominal values. Photometric drift: This should not exceed 0.002 A per hour, and should be checked as described previously (1). Pipets and cuvets. Specifications for cuvets and pipetting devices, and a procedure for checking the accuracy and precision of the latter have been provided elsewhere (1).
Materials Materials for the study were provided by the Centers for Disease Control (CDC) Atlanta, GA, the National Bureau of Standards (NBS), Washington, DC, and Reheis Chemical Co., Division of Armour Pharmaceutical Co., Phoenix, AZ 85077. Specifications for reagents and water were described previously (1). Protein standard. Ten ampoules of SRM 927 (bovine serum albumin, BSA) were mailed to each participant. According to the NBS directions, the BSA solution should be used promptly after an ampoule is opened. We found that the solution could be used more than once (up to five days) if transferred into a small screw-capped glass container and stored at 4 #{176}C. CLINICAL CHEMISTRY, Vol.27,No. 10,1981
1651
Reheis bovine albumin solution. Vials containing a crystalline BSA solution (10 g/L protein nitrogen, stock no. 3200-01) were provided by R. J. Westfall (Reheis Chemical Co.). Control sera. Four control sera were provided in 1-mL glass vials. Sodium aside (NaN3), 0.5 g/L, was added to each control as a preservative. Twenty-five vials of each control were shipped to each laboratory. Bovine serum albumin solution. About 40 mL of a 160 g/L BSA solution containing 9 g/L NaC1 and 0.5 g/L sodium aside was shipped to each participant.
Table 1. Photometric Accuracy of the Instruments Used: Absorbance of K2Cr207 Solutions at 350 nm Absorbanc Lab, no.
1
2
3 4 5
Note: All protein
solutions (CDC Controls, SRM, BSA and Reheis) were stored at 4 #{176}C. Before use, all protein solutions were allowed to warm to room temperature and then were mixed thoroughly by inversion.
Cobaltous ammonium sulfate solution. A 46 g/L solution of CoSO4 (NH4)2 SO4 . 6H20 in 0.18 mol/L H2S04 was shipped in tightly stoppered plastic bottles to each participant. The absorbance of this solution was measured with every analytical run at 510 nm, against water.
Characterization
of the Biuret Reaction
Method.
All protein analyses were performed in duplicate, and according to the candidate Reference Method developed by the Study Group (1). Linearity. The linearity of the method was evaluated with the SRM 927, the Reheis BSA solution, the CDC control 2, and a fresh serum pool. The protein concentration was varied by using aliquots of 50, 100, 150, and 200 L of each protein solution. Reagent and sample blanks were prepared in the same manner. Corrections to the standard volume (5.1 mL) were made as described previously (1). The linearity was evaluated by linear regression analysis. Criteria for acceptable linearity were the same as those described under Photometric
linearity. Precision.
To become familiar with the method, each participant performed 10 replicate analyses on each of the four control sera and the Reheis BSA solution. An estimate of the within-run precision of the method was obtained from this experiment. The day-to-day precision was evaluated from data obtained from 15 separate analytical runs. Analytical recovery. The BSA stock solution (160 g/L) and solutions of 80 and 120 g/L, prepared by diluting the stock with a solution containing 9 g of NaCl and 0.5 g of NaN3 per liter, were used for the recovery experiment. Equal volumes (3.0 mL) of each of the three BSA solutions were mixed with 3.0 mL of a fresh serum pool (clear, non-hemolyzed, and non-jaundiced). An aliquot of the serum pool was also diluted with an equal volume of the NaCl/NaN3 solution; this diluted pool served as the baseline for calculating the recovery of the added protein (BSA). The amount of protein added was calculated from the protein concentration of the 80 g/L BSA solution found by analysis.
Results Performance
of Spectrophotometers
Data on the photometric accuracy of the spectrophotometers are shown in Table 1. With one exception (Gilford 250), absorbance values for the 50 and 100 mg/L K2Cr2O7 solutions were within 1% of the 0.535 and 1.072 accepted values (2). Instruments checked with NBS glass filters provided absorbance values within the tolerance specified by NBS, i.e., ±0.0022 of the nominal absorbance values. All spectrophotometers met the criteria for acceptable linearity (data not shown). In all cases r2 exceeded 0.999 and absorbance values calculated from the regression equation were within ±1% of the measured values. Furthermore, the 1652
CLINICAL CHEMISTRY, Vol. 27, No. 10, 1981
Sp.ctrophotomot.r
50 mg/L
Cary 14 Cary 16 Cary 14 Cary 14 Beckman ACTA
0.536
100 mg/I
1.076
NBS filters, SRM930 NBS filters, SAM 930 0.536 1.07 1 0.536
1.070
0.534 0.538 0.537
1.069 1.083 1.038
CIII Zeiss PM 4 Beckman DU Gilford 250
6 7 8
standard error of the slope was very small (0.05 to 1.0% of the slopevalue).Although both cobaltous ammonium sulfate and cyanmethemoglobin were found suitable for use in evaluating photometric linearity, we recommend the use of cyanmethemoglobin in future studies, because it provides a regression line having a zero intercept. In retrospect, we have some doubts about the validity of the criteria for linearity. Two instruments met both criteria despite intercepts of 0.015 A and 0.026 A, which suggest either serious blanking problems or lack of linearity. Consequently, in future studies it may be necessary to consider limits for the intercept in defining criteria for acceptable linearity.
Absorbance Values of the Co(ll) Solution and SRM 927 Table 2 lists precision data on the absorbance values of the Co(II) solution and of the SRM 927 with the biuret reagent. A minimum of 15 absorbance readings of the Co(II) solution and 60 readings of the SRM were obtained in each of the laboratories. The SRM values are corrected Test absorbances, i.e., A4 - A3 - (A2 - A1) (1). Both the intra- and interlaboratory precision are excellent. The highly reproducible absorbance value of the SRM permits the establishment of its biuret absorptivity value (a) at 540 nm as 0.2983 L g’cm, with an SD of 0.0018.
Linearity of the Biuret Method The
first criterion
of linearity
(r2 >0.999)
was met by all
Table 2. Absorbance (A) Values for Cobaltous Ammonium Sulfate Solution (46 gIL) and SRM 927 ACO(II) .510 nm Labno.
ASRM 927 , 540 nm
SD
I
SD
1
0.402a
0.0015
0.414
0.0014
2 3
0.573 0.576
0.0008 0.0011
0.414
0.0027
4
0.578
0.0039
0.411 0.411
0.0012 0.0041
5 6 7
0.567 0.573 0.579
0.0007 0.0007
0.408 0.409
0.0021 0.0015 0.0021
8 Grand mean
0.3688
0.0036
0.414
0.0035
0574b
0.0044
O.412’
0.0025
0.0007
a Absorbance of the Co(ll) solution was measured Mean and SD were calculated from 90 absorbance were calculated from 420 absorbance values. ASRM: volumes of sample and reagent delivered by pipets. performed by Lab. 7.
#{149}
at 540 nm in these labs. values. c Mean and SD Corrected for the actual Pipet calibration was not
Table 3. Within-run Precision of 10 Replicate Analyses in Each of Eight Laboratories 1
Lab. no.
2
0.093
Absorbance of biuret reagent Reheis-Armour
Control 1
x
62.2
CV, % x
0.2 100.7
0.096
3
4 Protein
0.101
5 concn,
0.098
6
7
3
gIL
0.100
0.102
0.102
0.095
61.6 0.1
61.7 0.2
61.0 0.7
62.4 0.2
61.7 0.7
62.2 0.3
61.2 0.6
100.6
100.8
100.8
100.9
101.4
98.1
96.7
CV, %
0.2
0.1
k
71.1
70.2
CV, % x
0.3 50.9
0.1
0.3
0.4
0.2
0.3
0.5
0.6
Control 3
51.0
51.5
51.6
51.2
51.5
51.9
0.1
Control 4
CV. % x
59.8
61.1
0.4 63.0
0.2 63.6
0.3 62.0
0.5 65.7
0.5 58.9
50.0 1.2
0.1
0.6
0.5
0.3
0.7
0.6
Control2
CV, %
0.4 0.5
0.1
0.5
71.2
0.1
0.5
70.9
71.4
71.0
0.6 68.6
0.8 67.4
63.0 1.5
Table 4. Intralaboratory Precision Lab. no.
1
2
3
4
Protein
concn,
5
6
7
8
g/L
Control 1 Mean Within-run
101.0
101.7
100.7
101.6
101.3
100.7
97.2
0.2
0.1
0.1
0.3
0.2
0.3
0.4
98.2 0.8
0.4
0.8
0.4
1.0
0.5
0.6
0.8
2.5
CV, %
Total CV, %
Control 2
70.5
71.4
71.2
71.3
71.0
70.4
67.9
68.0
0.2 0.8
0.1 0.8
0.2 0.5
0.5 0.9
0.6 0.9
0.4 1.0
0.5 1.7
1.2 4.3
51.1
52.1
51.5
51.6
51.7
51.2
50.8
50.2
Within-run CV, %
0.4
0.5
Total CV, %
0.6
0.8
0.2 0.4
0.5 0.7
0.3 0.7
0.4 0.7
0.6 1.1
0.8 2.4
59.9 0.4 1.2
63.3 0.3 2.0
62.8 0.3 0.8
65.3 0.5 2.6
61.7 0.3 1.2
662 0.8 1.2
58.3 0.8
63.7 1.8
2.0
3.9
Mean Within-run CV, % Total CV, %
Control 3
Mean
Control 4
Mean Within-run CV, % Total CV, %
laboratories except for Laboratory 8, which experienced some problems with its spectrophotometer; the large intercept (0.012 A) with both the SRM and the fresh serum pool indicate a persistent problem. Several laboratories failed to meet the second criterion of linearity (see Criteria for acceptable linearity). At low protein concentration (35 gIL), a negative deviation of >1% was ob-
served with both the SRM and the fresh pool. Owing to the calibration mode (use of a single standard), the slight suppression of the absorbance at protein concentrations near 35 g/L may introduce an error of
