Clinical Chemistry 49: 463-469, 2003;
10.1373/49.3.463
(Clinical Chemistry. 2003;49:463-469.)
© 2003 American Association for Clinical Chemistry, Inc.
Certification of Standard Reference Material 970, Ascorbic Acid in Serum, and Analysis of Associated Interlaboratory Bias in the Measurement Process
Sam A. Margolis1a,
Mark Vangel2 and
David L. Duewer1
1 Analytical Chemistry Division, Chemical Science and Technology Laboratory, National Institute of Standards and Technology, Gaithersburg, MD 20899-8392.
2 MRI Imaging Department, Massachusetts General Hospital, Charlestown, MA 02129.
aAuthor for correspondence. Fax 301-977-0685; sam.margolis{at}worldnet.att.net.
 |
Abstract
|
|---|
Background: The accurate and reproducible measurement of ascorbic acid is essential in delineating the role of ascorbic acid as a diagnostic tool for human disease and for the comparison of data acquired by different laboratories. A stabilized pair of standards of ascorbic acid in human serum, which is compatible with most analytical methods, have been prepared.
Methods: The certification was based on the gravimetric addition of ascorbic acid to metaphosphoric acid-stabilized, ascorbic acid-depleted serum and NIST liquid chromatography–electrochemical measurements. The NIST results were analyzed statistically for homogeneity, and the expanded uncertainty of each SRM was calculated using all of the NIST data. An interlaboratory comparison exercise was also performed.
Results: These materials, Standard Reference Material (SRM) 970 Ascorbic Acid in Serum, Level I and Level II, are homogeneous and are certified to contain (10.07 ± 0.21) and (30.57 ± 0.28) mmol ascorbic acid/L of solution (expanded uncertainty), respectively. In the interlaboratory comparison (n = 17), the relative SDs for the two materials were 22% and 19%.
Conclusions: Two lots of serum, each containing different amounts of ascorbic acid stabilized in metaphosphoric acid, have been prepared and characterized. Many laboratories provide inaccurate results.
|
Introduction
|
|---|
Ascorbic acid (AA)1
has been proposed as having cancer chemopreventive activity (1). Interlaboratory epidemiologic studies are used by the National Cancer Institute and others as one of the mechanisms for investigating the potential of AA as a cancer chemopreventative agent (1)(2)(3)(4)(5). For these studies to yield the most meaningful results, plasma and (or) serum AA concentrations determined at different times or in different laboratories must be comparable. Numerous studies have addressed the intralaboratory repeatability of AA measurements, generally observing a relative standard deviation (RSD) of 1–2% (6). Several studies have indicated that interlaboratory reproducibility of AA measurements in biological samples is much larger (CV, 15–25%) (7)(8). This large difference between intra- and interlaboratory measurement performance indicates systematic biases among the materials used to assess performance and (or) among the methodologies used to determine the concentration of AA. Assessment of the interlaboratory reproducibility of any chemical measurement requires well-characterized homogeneous samples of demonstrated stability that are compatible with the majority of methods in current use. NIST has demonstrated that total AA (TAA), which is the sum of AA + dehydro-AA in lyophilized plasma samples containing dithiothreitol, is stable for at least a 6-year period when stored at -70 °C. Similarly, TAA in frozen serum containing 50 g/L metaphosphoric acid (MPA) degrades at no more than 1% of the total mass per year over a 2-year period when stored at -70 °C (9). In a series of three collaborative studies with laboratories that used similar analytical methods, the interlaboratory reproducibility for the determination of TAA was 15% (9).
To assist laboratories in improving the accuracy of their measurements and evaluating their quality-control practices, NIST has prepared and certified Standard Reference Material (SRM) 970, AA in human serum. This SRM consists of two types of samples, Level I, containing a low AA concentration, and Level II, containing a high AA concentration, that are at approximately the 25th and 75th percentiles, respectively, of the distribution of AA concentrations in humans. Several different modes have been described for use in value assignment of chemical composition in natural matrix SRMs at NIST (10). In the certification of SRM 970, we used a mode that consisted of two independent methods of TAA analysis at NIST and obtained additional confirmatory data from an interlaboratory collaborative study. In the NIST study, the two methods used to certify this SRM were the gravimetric and liquid chromatography–electrochemical detection (LC-EC) methods. This report describes the preparation and certification of this SRM and the results of the collaborative interlaboratory study.
|
Materials and Methods
|
|---|
AA, dithiothreitol, and MPA were purchased from Sigma-Aldrich Chemical Co. Acetonitrile was HPLC grade. The AA contained <0.1% impurities (determined by proton magnetic resonance spectroscopy), including isomers and degradation products. No D-ascorbic acid was detected by LC at the 1% level (11), and the absorptivity (E1%) in acid solution (50 g/L MPA) is 557, which is essentially identical to the value reported in the literature, 560 (12). Human serum was purchased from Interstate Blood Bank in 
260-mL lots. The serum was free of anticoagulants and tested negative immunologically for hepatitis B and HIV-1 agents.
preparation of srm 970 level i and level ii samples
All components of SRM 970 Level I and Level II were measured gravimetrically into separate 5-L borosilicate bottles. The serum units, which were stored at -70 °C, were defrosted and blended into two separate and distinct lots, one for each level. Each 5-L bottle was placed in an ice bath on a magnetic stirrer. While the serum was stirred mechanically with a Teflon stirring bar, approximately an equal volume of an aqueous solution of MPA (100 g/L) was added gravimetrically. Ten 1-mL aliquots were removed from each serum-MPA solution and frozen for determination of the residual AA in the serum. Each solution was then supplemented by the gravimetric addition of AA stock solution (0.98058 g/g of 100 g/L MPA). The masses of each component, the densities of the serum and the MPA, and the calculated concentration of the supplemented AA are summarized in Table 1
. The supplemented serum solutions were kept in an ice bath with constant stirring, and 2.2-mL aliquots were distributed mechanically into cryules (Wheaton Science Products) that were purged with nitrogen and then sealed. The samples were placed in boxes of 144 ampoules in the order of filling and stored at -70 °C. The total preparation time for each level was 4 h.
sample selection and analysis
A total lot for each AA level consisted of seven boxes, each containing 144 ampoules (i.e., 1008 ampoules). Eight ampoules were randomly selected from each box to provide samples representative of the entire lot for NIST measurements and the interlaboratory study. From each stratified set (Level I and Level II), 12 ampoules spanning the entire fill sequence were randomly selected for evaluation by NIST of the homogeneity of the entire lot and for measurement of the AA and the TAA (AA + dehydro-AA) content. On each of 6 days, four randomly selected samples were alternately selected either from Level I or Level II. These samples were randomly analyzed in duplicate. A series of three calibration solutions that bracketed both levels were analyzed before and after each set of samples on each day. Two sets of calibration solutions were prepared; one set was used for the first 3 days and the second set was used for the following 3 days. The remaining ampoules were randomly assigned to the laboratories participating in the interlaboratory study. The AA and the TAA (AA + dehydro-AA) content were analyzed at NIST using the method of Margolis and Shapira (11).
interlaboratory study
Fifteen laboratories participated in the interlaboratory study. Each laboratory received two ampoules each of SRM 970 Level I and Level II, which were randomly assigned from the selected pool of 56 samples at each level, and a vial of crystalline AA. Each ampoule was analyzed in duplicate by the laboratories. Each laboratory gravimetrically prepared a standard AA solution in aqueous MPA (50 g/L) by weighing 200 mg of the provided crystalline AA into a 100-mL volumetric flask and then weighing the amount of aqueous MPA (50 g/L) that was added to the 100-mL mark. This solution was then diluted 1:200 gravimetrically in a 100-mL volumetric flask with aqueous MPA (50 g/L). The 
max (wavelength of maximum ultraviolet absorbance) and the absorbance at that wavelength were measured, and the E1% was calculated at this wavelength. The amount of AA in this dilute standard was calculated using the gravimetric data and measured along with the AA in the SRM 970 samples. The density of the aqueous MPA (50 g/L) was calculated to be 1.038 (SD = 0.003; n = 10). Each laboratory analyzed the dilute AA solution and the samples with the method(s) that they routinely used for these measurements. Two laboratories (laboratories 3 and 4) analyzed the samples by two different methods, one by photometry and one by LC, and these results were treated as independent sets of data.
safety considerations
All chromatography reagents contained acetonitrile. Some reagents also contained MPA. These compounds are toxic to various degrees with respect to inhalation, ingestion, and/or eye irritation. Acetonitrile is flammable. The serum tested negative immunologically for hepatitis B and HIV-1 agents. However, other unidentified infectious agents may be present, and appropriate precautions were taken. All laboratories were notified of the appropriate precautions to minimize exposure. The reagents and serum used at NIST were handled in a chemical hood and disposed of according to Occupational Safety and Health Administration regulations by the NIST Health and Safety Office.
|
Results and Discussion
|
|---|
The procedures defined in NIST Special Publication 260-136, "Definition of Terms and Modes Used at NIST for Value Assignment Reference Materials for Chemical Measurements" (10) and by Levenson et al. (13) were used for certification of the AA content of SRM 970. These procedures are consistent with the ISO Guides (14)(15). The certification process must meet the following NIST criteria. (a) The material must be chemically stable over the temperature range at which it is to be used and stored. (b) The accuracy and systematic errors of the methods used for value assignment must be defined. (c) The homogeneity of the units of the candidate material must be demonstrated. (d) The analyte should be measured by two independent methods. When possible, the analyte is also evaluated in a collaborative study involving a group of selected laboratories using different types of analytical instrumentation. It is recommended that the samples in this type of study should include the candidate material and a control material that contains a different concentration of the analyte in the same matrix to evaluate more effectively the methodologic bias.
homogeneity and nist certification
The AA concentrations in SRM 970 were value assigned using the NIST LC-EC method (11) and the gravimetric data for the addition of AA to the serum-MPA mixture. The homogeneity was demonstrated as a part of the LC-EC measurements (see Materials and Methods). The results of the gravimetric measurements are summarized in Table 1
. The value for TAA represents the concentration of the added AA in each of the samples. In addition to the total added AA, the serum also contained a small amount of endogenous AA: 0.25 and 0.30 mmol/L in Level I and Level II, respectively, as determined by LC.
The homogeneity of the entire group of samples was evaluated using a randomly selected subset of samples. The samples were analyzed in groups of four and in duplicate in random order over a 3-day period to estimate the within-sample, within-day, and between-day variation. Samples from Levels I and II were measured on alternate days, and a new set of three calibrants that bracketed the AA content of both groups of samples (Levels I and II) was prepared for each pair of measurements. No significant difference was observed between the three groups of calibration solutions, each run on 2 consecutive days. On each day a set of calibration solutions was assayed before and after the serum samples (total of 36 measurements). The combined analysis of all of sets of three calibration solutions indicated that the peak area-vs-concentration curve was linear and passed very close to the origin: y = 1727.6x - 14.7 (R2 = 0.997; n = 36), where y is the peak area and x is the AA concentration. The ANOVA of the measurements on Levels I and II (Table 2
) indicated that there was no significant variation (P >0.05) in the homogeneity of the samples as a function of the order of filling (box), the time of analysis (day), or the combination of these factors (day:box). The expanded uncertainties (11) of the values for the AA in Level I and II samples obtained for this combined homogeneity and LC study were <1% of the mean value for each level (Table 3
, % relative expanded uncertainty). Both of these measurements indicate that the measurement error for both Level I and Level II was very small, that each sample lot was relatively homogeneous, and that at the 95% confidence limit, the combined methodologic error (relative expanded uncertainty) was also <1%. The close agreement between the gravimetric measurements and the LC measurements indicates that the LC method is accurate and that the sum of the sources of systematic bias in the determination of total AA by the NIST method is <0.28 µmol/L of serum. The long-term stability (6 years) with respect to TAA of the samples prepared at NIST by this method has been clearly established (9).
interlaboratory study
The interlaboratory study consisted of two parts. The first assessed the capability of each laboratory to accurately prepare and assay a control material. The second part assessed their ability to measure the AA in the SRM 970 samples. We used robust statistics (16)(17) to analyze these data. This analytical approach eliminates the need to assume that the data fit a gaussian distribution. It uses the trimmed mean, which is insensitive to small numbers of gross deviations and works well for heavy-tailed distributions close to the normal.
The results of the first part of this study are summarized in Table 4
. The values for the absorptivity of AA indicate that in all but two cases the laboratories were able to accurately prepare the control solution. However, 10 of 17 laboratories reported a >10% error in the measurement of the AA in their control solutions, indicating that their assay was not completely in control. The errors were nearly equally distributed negatively and positively. The results of the measurement of the AA in SRM 970 by these laboratories (Tables 5
and 6
) also reflect this variation in accuracy as indicated by the large %RSD of both the means and the medians (Table 7
). The robust %RSD (Table 7
) of the median is particularly informative because this value is less affected by outlying values. The relatively low SDbetween, SDwithin, and SDheterogeneity of the mean compared with the median of the results of the collaborative study indicate that the laboratories were precise in their measurements but that their measurements lacked significant accuracy (Table 7
). This lack of accuracy is further illustrated in Figs. 1
and 2
. The distribution of the results of the interlaboratory study on Level I and Level II samples in the Youden (18) plots on a line 45 degrees from the axis is consistent with intralaboratory precision and a lack of laboratory accuracy. Because of this large interlaboratory variation, the results of this study were not incorporated into the certified value; however, they do confirm the NIST-certified value that lies within the central circle of the Youden plot.
conclusion
Two lots of serum containing different amounts of AA stabilized in MPA have been prepared and characterized at NIST as SRM 970, Level I and Level II.
Each of the five criteria for certification was met. (a) The purity of the AA used to prepare SRM 970 was demonstrated by the absence of D-ascorbic acid by LC analysis and by the correspondence of its absorptivity to that reported in the literature (12). (b) The accuracy was demonstrated by the agreement of the HPLC measurements with the gravimetrically added AA. The data in Table 2
indicate that the samples are homogeneous, i.e., there is no significant variation associated with the filling of the vials (box) or the day of analysis (day). (c) The data in Table 3
indicate that the combined methodologic error in the NIST LC measurements was <2.09% (% relative expanded uncertainty) for both levels and that the samples in both Level I and Level II were homogeneous. (d) The gravimetric addition and the LC analysis constituted the two independent methods of measurements, and they agreed (Table 3
). (e) A collaborative study was performed using a variety of analytical instruments and methods.
These homogeneous materials (Level I and Level II) contain (10.07 ± 0.21) and (30.57 ± 0.28) mmol AA/L of solution (expanded uncertainty), respectively. This was determined by statistically combining the NIST data for the gravimetric and LC-EC methods. These materials were also evaluated in an interlaboratory study that confirmed the NIST measurements. However, because the reproducibility was relatively large (RSD = 22% and 19%, respectively), these data were not included in the certification. The AA concentration in Level I lies at the first quartile in the distribution of AA concentrations in normal serum, and that of Level II lies at the third quartile, thus providing SRMs for the evaluation of clinical methods for the measurement of AA in serum. The lack of interlaboratory accuracy clearly demonstrates the need for these standards. The most likely source of bias appears to be the result of interlaboratory variation and may be attributable to the variety of methods used. In addition, several laboratories appeared to have some difficulty preparing the control sample.
|
Acknowledgments
|
|---|
We acknowledge and thank the laboratories that generously participated in the collaborative study. Certain commercial equipment, instruments, and materials are identified in this report to specify adequately the experimental procedure. Such identification does not imply recommendation or endorsement by NIST, nor does it imply that the equipment or material is necessarily the best for the purpose.
|
Footnotes
|
|---|
1 Nonstandard abbreviations: AA, ascorbic acid; RSD, relative standard deviation; TAA, total ascorbic acid; MPA, metaphosphoric acid; SRM, Standard Reference Material; and LC-EC, liquid chromatography–electrochemical detection. 
|
References
|
|---|
- Stahelin HB, Gey KF, Brubacher G. Plasma vitamin C and cancer death: the prospective Basel study. Ann N Y Acad Sci 1987;498:124-131.[CrossRef][Web of Science][Medline]
[Order article via Infotrieve]
- Sobala GM, Schorah CJ, Sanderson M, Dixon MF, Tompkins DS, Godwin P, et al. Ascorbic acid in the human stomach. Gastroenterology 1989;97:357-363.[Web of Science][Medline]
[Order article via Infotrieve]
- Correa P. The gastric precancerous process. Cancer Surv 1983;2:437-450.
- Romney SL, Basu J, Vermud S, Palan PR. Plasma reduced ascorbic acid and total ascorbic acid in human uterine cervix dysplasias and cancer. Ann N Y Acad Sci 1987;498:132-143.[CrossRef][Web of Science][Medline]
[Order article via Infotrieve]
- Block G. Vitamin and cancer prevention: the epidemiological evidence. Am J Clin Nutr 1991;53:270S-280S.[Abstract/Free Full Text]
- Pelletier O. Vitamin C (L-ascorbic acid and dehydro-L-ascorbic acids). Augustine J Klein BP eds. Methods of vitamin assay 1985:303-346 John Wiley and Sons New York. .
- Nicolson I, Macrae R. Comparative assessment of high performance liquid chromatographic methods for the determination of ascorbic acid and thiamin in foods. Analyst 1984;109:267-271.[CrossRef][Medline]
[Order article via Infotrieve]
- Bates CJ. Plasma vitamin C assays: a European experience. In J Vitam Nutr Res 1994;64:283-287.
- Margolis SA, Duewer DL. Measurement of ascorbic acid in human serum: stability, intralaboratory repeatability, and interlaboratory reproducibility. Clin Chem 1996;42:1257-1262.[Abstract/Free Full Text]
- May W, Paris R, Beck C, Fassett J, Greenberg R, Kramer G, Wise S, et al. Definitions of terms and modes used at NIST for value assignment reference materials of chemical measurement 2000 US Government Printing Office Washington, DC. .
- Margolis SA, Shapira RM. Liquid chromatographic measurement of L-ascorbic acid and D-ascorbic acid in biological samples. J Chromatogr 1997;690:25-33.
- Bauernfeind JC, De Ritter E. Properties of vitamins. Fasman GD eds. Handbook of biochemistry and molecular biology, physical and chemical data 1975;Vol. 2:293 CRC Press Cleveland, OH. .
- Levenson MS, Banks DL, Eberhardt KR, Gill LM, Guthrie WF, Liu HK, et al. An approach to combining results from multiple methods by the ISO GUM. J Res Natl Inst Stand Technol 2000;105:571-579.
- . ISO Guide 35. Certification of reference materials—general and statistical principles, 2nd ed 1989 International Organization for Standardization Geneva, Switzerland. .
- ISO guide to the expression of uncertainty on measurement, 1st ed 1993 International Organization for Standardization Geneva, Switzerland. .
- . Analytical Methods Committee, Royal Society of Chemistry, UK. Robust statistics—how not to reject outliers. Part 1. Basis concepts. Analyst 1989;114:1693-1697.[CrossRef]
- . Analytical Methods Committee, Royal Society of Chemistry, UK. Robust statistics—how not to reject outliers. Part 2. Inter-laboratory trials. Analyst 1989;114:1699-1702.[CrossRef]
- Youden WJ. Graphical diagnosis of interlaboratory test results. Ind Qual Control 1959;11:1-5.
- . ASTM E691-92.. Standard practice for conducting an interlaboratory study to determine the precision of a test method. Annual book of ASTM standards 1992;14.02:425-44 ASTM West Conshohocken, PA. .
The following articles in journals at HighWire Press have cited this article:
|
|
|
|
 
J. Lykkesfeldt
Ascorbate and Dehydroascorbic Acid as Reliable Biomarkers of Oxidative Stress: Analytical Reproducibility and Long-term Stability of Plasma Samples Subjected to Acidic Deproteinization
Cancer Epidemiol. Biomarkers Prev.,
November 1, 2007;
16(11):
2513 - 2516.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
