| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
CDC, National Center for Environmental Health, Division of Laboratory Sciences, Mail Stop F18, Inorganic Toxicology and Nutrition Branch, 4770 Buford Hwy, NE, Atlanta, GA 30341-3724;
aauthor for correspondence: fax 770-488-4139, e-mail RSchleicher{at}cdc.gov
The National Health and Nutrition Examination Survey (NHANES) laboratory at CDC has used a modification of methods (1)(2) with electrochemical detection for measurement of serum vitamin C for the past 9 years. The assay is relatively rapid, easy to perform, and gives good precision. Quality-control (QC) materials have been kept at –70 °C for more than 10 years without substantial degradation. A drawback of the method is the lack of an internal standard to correct for analyte degradation, procedural errors, and detector drift. Significant vitamin C degradation is intermittently encountered during the analytical process. Oxidation of ascorbic acid (AA) is accelerated by exposure to air, heat, light, and traces of copper and iron (3) and may be introduced through contact with unexpected sources, such as consumable supplies (4). Detector drift is a characteristic of electrochemical detection and has been noted by others performing the serum AA assay (5).
Our original HPLC assay used partition of largely un-ionized AA and amperometric detection. A 25-cm C18 reversed-phase column is equilibrated with a mobile phase consisting of monochloroacetic acid (pH 3.0) containing disodium EDTA and sodium octylsulfonate [originally used for ion pairing with catecholamines extracted from adrenal chromaffin cells in a mixture containing AA (1)]. The AA in serum is stabilized by addition of metaphosphoric acid (MPA), reduced by addition of dithiothreitol (DTT), and then oxidized at +650 mV referenced to an Ag/AgCl electrode. The working electrode is a thin-layer detector cell. When serum is treated as indicated, peaks for AA, uric acid (URIC), and DTT are resolved and detected at the applied potential within 
18 min. The run time for each sample is shortened by injecting the next sample before all peaks from the previous sample have eluted.
Several changes suggested themselves to modify this method: (a) Improved column technology would allow the use of a smaller column with smaller injection volumes and shorter retention times. (b) Sodium octylsulfonate could be eliminated because it does not effectively pair with ascorbate. (c) A small amount of methanol in the mobile phase would accelerate the elution and sharpen peaks. (d) Calibrators could be prepared and frozen to save daily preparation time and could be prepared in the same fashion as the samples with the addition of internal standard and other reagents to control for any errors in handling and/or analyte degradation. (e) Longer runs might be possible if an internal standard could be found to adequately correct for analyte degradation and detector drift.
We developed a revised method that uses an Agilent 1100 solvent delivery system connected in series to an 1100 diode array detector and a BAS electrochemical detector set at +650 mV. AA in serum was separated on a YMC ODS-AQS-3 column [15 cm x 3 mm (i.d.); 3-µm particle size (120 Å)] with a 10-µL injection volume of a 5-fold–diluted stabilized serum specimen. An Upchurch 0.5 mm stainless steel frit was used as a precolumn filter. A mixture of 0.15 mmol/L monochloroacetic acid, 0.2 mmol/L disodium EDTA, and 150 mL/L methanol at pH 3.0 was used as a mobile phase at a flow rate of 0.3 mL/min. Stock solutions of AA were prepared gravimetrically. Three concentrations of calibrators (1.42–28.39 µmol/L) representing 0.5, 3, and 10 ng on column were diluted in 60 g/L MPA–2.5 mL/L DTT at pH 1.8 and stored for up to 4 months at –70 °C with minimal change in values (0%–2% degradation). Once prepared, the highest calibrator was chromatographed, and a peak of 470 absorbance units ± 2% at 245 nm was used as an additional step to assess integrity. The internal standard, 1-methyluric acid (MURIC; Sigma Chemical Co), was added to all samples and calibrators to achieve a final concentration of 82.35 µmol/L. Assay calibration was performed for each run. NHANES serum specimens were prepared in the field as described previously (6). Peaks were integrated by use of peak-area ratios of AA to MURIC.
HPLC analysis using field-stabilized specimens showed no interfering peaks. Blanks (reagents only) in each run showed no interfering peaks. AA, DTT, URIC, and MURIC peaks in the specimens were identified by use of calibrators and inspection of their retention times (Fig. 1A
).
|
The optimum temperature to elute all constituents within 12 min was 30 °C. All peaks were baseline separated. Increasing the mobile phase pH from 2.0 to 4.5 lengthened peak retention times, whereas increasing the percentage of methanol shortened them. The retention times of the analytes were stable with CVs <12% over the course of 6 months on a single column. At the time of manuscript submission, 6 columns had been used in routinely performing this assay over 14 months; fusion of the AA peak with an earlier eluting peak was the primary reason for retiring columns. The mean number of injections per column was 1401 (range, 676–2196). On average, all reagents for this assay were prepared monthly.
Calibration curves were linear up to 28.39 µmol/L, which represents a final concentration of 141.95 µmol/L in serum specimens (mean of daily calibrations: y = 0.9897x – 0.0004; r2 = 1.0; SEregression = 0.0114; SEslope = 0.0018; SEintercept = 0.0006). The limit of detection, estimated as 3 times the SD of a near-zero sample, was 0.68 µmol/L, which represents 0.24 ng on column. The mean (SD) recovery of AA added to serum at final concentrations ranging from 38.4 to 116.1 µmol/L was 97 (2)% (Table 1 in the Data Supplement that accompanies the online version of this Technical Brief at http://www.clinchem.org/content/vol51/issue6/). Five specimens with serum AA concentrations >96.53 µmol/L were diluted 0- to 7-fold with 60 g/L MPA–2.5 mL/L DTT. The ratios of observed to expected results were 0.93–1.07.
The mean intraassay CVs for 5 samples (38.61–60.76 µmol/L) and 3 pools (24.42–55.08 µmol/L) processed in 10–20 replicates in 1 run ranged from 0.6% to 3.6% (Table 2 in the online Data Supplement). The mean interassay CVs for 5 samples (26.69–57.92 µmol/L), each processed in 1 replicate in 5 runs, ranged from 1.2% to 4.2% (Table 3 in the online Data Supplement). Over the course of 135 runs, 3 QC pools processed as duplicates in each run showed CVs of the run means of 8% (low; 13.06 µmol/L), 4% (medium; 60.76 µmol/L), and 4% (high; 119.24 µmol/L). The injection reproducibility was evaluated by use of 10 replicate injections using 3 separate QC pool preparations. The CV values of the means were 
0.7% for all pools.
The revised method was more accurate than our original method based on repetitive analysis of NIST standard reference material SRM 970 Level I and II (recertified in 2004). The mean (SD) results for Level I [8.57 (0.84) µmol/L] were 102% of the target value and for Level II [28.05 (1.34) µmol/L] were 100% of the target value during 13–14 runs, compared with 93% and 92% of the target values, respectively, obtained with the original method in a similar number of runs. The CVs for these results were also better with the revised method: 5%–10% vs 7%–12% for the old method. Participation in 2 NIST quality assurance exercises showed results in good agreement with consensus medians.
The mean number of injections per run (1 injection per sample) during the evaluation period was 57 (range, 19–94). Repeated analysis of low, medium, and high QC pools gave CVs of the individual results of 6% for AA when an internal standard was used. Quantification without an internal standard gave QC results with lower values and slightly higher CVs (8%). The internal standard increased the mean values for the QC pools by 3%–5%. Although the internal standard compensated for detector drift during runs, it was more stable than AA. The mean (SD) decrease in peak area in QC pools measured at the beginning and end of each run was 5 (6)% for AA and 0.5 (5)% for MURIC. Runs with significant drift (>10% fall-off of AA signal) were more likely to be in control when quantified with use of an internal standard.
Deming regression comparison between the original and revised assays for 308 specimens in 10 separate assays gave the following results: y = 1.06x – 1.9 µmol/L (Sy|x = 4.29 µmol/L; R2 = 1.00; n = 308). The 95% confidence interval for the y-intercept was –3.08 to –0.79 µmol/L, and the confidence interval for the slope was 1.04–1.07. Results in this data set spanned the reportable range (3.41–194.76 µmol/L). Bland–Altman analysis showed a small mean difference of 1.53 µmol/L (95% confidence interval, 1.02–2.04 µmol/L) for the revised method (Fig. 1B
). Regression analysis of the percentage difference between methods as a function of AA concentration showed that the difference increased with increasing concentration (y = 0.055x – 0.03; R2 = 0.13; P <0.0001 for the slope). We anticipated a shift toward more positive values with the revised method attributable to (a) the use of an internal standard to correct for detector drift and (b) processing of the calibrators as though they were unknowns. We did not expect a concentration-dependent difference from these 2 changes and do not have an explanation for the larger difference in the high concentration range.
Other assay conditions of interest were also investigated. A comparison set of 29 samples, selected because of substantially different values obtained with the 2 methods [mean (SD), 13 (1)% higher with the revised method], were separated chromatographically with and without the ion-pairing reagent in the mobile phase. The AA results differed by a mean (SD) of only 1 (2)%, demonstrating that the ion-pairing reagent provided no added specificity. Integration of the AA peak area gave more accurate and precise results than did peak height (data not shown). Because dilution of specimens led to losses of up to 30%, smaller sample volumes are recommended when re-measuring samples with unusually high AA values, i.e., greater than the NHANES III 99th percentile (123 µmol/L). Subjecting specimens to 1 freeze–thaw cycle generally did not lead to AA degradation. Only 1 in 34 sets of QC pools showed significant degradation of AA (>15% loss) after a single freeze–thaw cycle.
In summary, the development of an HPLC method that includes an internal standard improves the precision and accuracy of AA measurement and compensates for detector drift so that longer runs can be accommodated. Other changes have enhanced performance of the new assay.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
