Determination of Niacin in Infant Formula by Solid-Phase ... · determination of niacin in milk-based and soy-based infant formula. Analysis is in 3 steps: test sample digestion,

FOOD COMPOSITION AND ADDITIVES

Determination of Niacin in Infant Formula by Solid-PhaseExtraction and Anion-Exchange Liquid Chromatography

PVM 1:2000

METHOD AUTHORS:DENIS E. LACROIX and WAYNE R. WOLF

U.S. Department of Agriculture, Beltsville Human NutritionResearch Center, Food Composition Laboratory, 10300 Bal-timore Ave, Beltsville, MD 20705-2350, Tel:+1-301-504-8928; Fax: +1-301-504-8314; E-mail:[email protected], or [email protected]

SUBMITTING LABORATORY:U.S. Department of Agriculture, Beltsville Human NutritionResearch Center, Food Composition Laboratory, 10300 Bal-timore Ave, Beltsville, MD 20705-2350

PEERLABORATORY:EURAL PORTER and DENNIS CANTELLOPS

U.S. Food and Drug Administration, Atlanta Center for Nu-trient Analysis, 60 8th St, NE, Atlanta, GA 30309

ACKNOWLEDGMENT OFREVIEWERS:G. WILLIAM CHASE, JR

U.S. Food and Drug Administration, 22201 23rd Dr, SE,Bothwell, WA 98021

DAVID WOOLLARD

AgriQuality NZ, 131 Boundary Rd, PO Box 41, Auckland,NZ

Abstract

A peer-verified, solid-phase extraction (SPE)/anion ex-change liquid chromatographic method is presented for thedetermination of niacin in milk-based and soy-based infantformula. Analysis is in 3 steps: test sample digestion, extrac-tion/cleanup, and liquid chromatography (LC). Digestion usesa standard AOAC digestion procedure that involvesautoclaving at 121EC for 45 min in (1 + 1) H2SO4 to free en-dogenous niacin from protein and to convert addedniacinamide to niacin. The digest solution is adjusted topH 6.5 with 7.5M NaOH. Acidification to pH <1.0 with(1 + 1) H2SO4 precipitates the protein. The clarified solutionis then filtered, and the filtrate is brought to volume. SPE of ni-acin is accomplished by passing an aliquot of the digest solu-tion through an aromatic sulfonic acid–SPE (ArSCX–SPE)column. After the column is washed with methanol and waterto remove extraneous material, the niacin is eluted with 0.25Msodium acetate/acetic acid buffer at pH 5.6. An an-

ion-exchange polystyrene–divinylbenzene column with 0.1Msodium acetate/acetic acid buffer at pH 4.0 is used for LC. Ni-acin is determined by UV detection at 260 nm. A standardcurve is prepared by passing known amounts of niacinthrough the ArSCX–SPE columns used for niacin extraction.The following values for xand relative standard deviation(RSD) were obtained for National Institute of Standards andTechnology Standard Reference Material (NIST SRM) 1846Infant Formula with a certified value for niacin of 63.3 ±7.6 µg/g: Submitting laboratory.— x = 59.7 ± 4.0µg/g;RSD = >6.7%; confidence interval (CI) = ± 1.4µg/g;n = 27.Peer laboratory.—x = 56.6 ± 6.6µg/g; RSD = >11.7%;CI = ± 4.1µg/g;n = 8.

1 Summary of Results of Verification Study

1.1 Matrix

The matrix in this study was SRM 1846 Infant Formula(NIST, Gaithersburg, MD).

1.2 Number of Test Samples

The submitting laboratory analyzed 27 test samples; thepeer laboratory analyzed 8 test samples.

1.3 Precision of Submitting Laboratory

(1.3.1) Standard curve reproducibility.—Intralaboratorydata for the variation in 11 daily niacin standard curves ob-tained by the submitting laboratory over 3 months were usedto prepare the mean linear standard curve shown in Figure 1.Linear regression analysis of the mean peak areas at each levelof niacin represented in the standard curve shows excellentlinearity over the range of analysis. For each level of niacindetermined, the 95% confidence interval (CI) demonstratesexcellent reproducibility between daily runs. However, thesedata do show the necessity of preparing standard curves forevery series of determinations to account for daily fluctuationsin instrument response and chromatographic conditions.Values for the slope (m), the intercept (b), and the regressioncoefficients (R) of the mean standard curve are as follows:(peak area) m = 1449, b = +7936, R = 0.9997, andR2 = 0.9994; (peak height) m = 430.27, b = –2.34, R = 0.9998,and R2 = 0.9996.

(1.3.2) Test sample reproducibility.—Intralaboratory datafor the method show good reproducibility for the determinationof niacin in SRM 1846, and in commercial milk- and soy-basedliquid and powdered infant formulas (Tables 1–5).

LACROIX & WOLF: JOURNAL OF AOAC INTERNATIONAL V OL. 84, NO. 3, 2001 789

1.4 Precision of Peer Laboratory

(1.4.1) Standard curve repeatability.—Linear regressionanalysis of the composite niacin standard curve (Figure 2)shows excellent linearity and reproducibility of 4 daily stan-dard curves obtained by the peer laboratory.

(1.4.2) Test sample repeatability.—Average niacin valuesfor SRM 1846 obtained by the peer laboratory are given in Ta-ble 1. All analyzed matrixes gave RSDs consistent with the ac-ceptable value of 8.6% obtained for the 60µg/g range of thelevel of the analyte analyzed by using the modified Horwitzlimits of acceptability equation from the AOAC Peer-VerifiedMethods Program (1, 2).

1.5 Results

(1.5.1) Accuracy for reference material.—NIST SRM1846 Infant Formula, with a certified niacin value of 63.3 ±7.6µg/g (3), was used as a known reference matrix to establishthe accuracy of the method for milk-based infant formulaproducts. Values for the niacin content of SRM 1846 obtainedby the submitting and peer laboratories are summarized in Ta-ble 1. All niacin values from the submitting laboratory weredetermined by peak area calculation unless otherwise speci-fied. All niacin values from the peer laboratory were deter-mined by peak height calculation because its computer pro-gram for liquid chromatography (LC) malfunctioned. Niacin

790 LACROIX & WOLF: JOURNAL OF AOAC INTERNATIONAL V OL. 84, NO. 3, 2001

Figure 1. Mean linear standard curve, prepared from data obtained by the submitting laboratory, at the 95 %confidence interval ( n = 11 niacin standard curves).

Table 1. Niacin values obtained by submitting and peer laboratories for SRM 1846 Infant Formula

Lab (method of calculation) n Niacin found, µg/g SD, µg/ga RSD, %b CI, µg/gc % of certified valued

Submitting (std curve) 27 59.7 4.01 6.7 1.4 94.3

Submitting (MOSA)e 21 57.6 4.24 7.4 1.8 90.9

Peer (std curve) 8 56.6 6.61 11.7 4.1 89.4

Peer MOSA 8 56.17 6.13 10.9 4.3 88.7

a SD = standard deviation.b RSD = relative standard deviation.c CI = confidence interval.d Certified value for niacin in SRM 1846 = 63.3 ± 7.6 µg/g (3).e MOSA = method of standard additions.

values obtained by both laboratories were within the limits ofthe certified value for SRM 1846.

(1.5.2) Commercial products.—Niacin data for commer-cial milk- and soy-based liquid and powdered infant formulasobtained by the submitting laboratory are contained in Ta-ble 2. The range of the niacin values for these commercialproducts was higher than that for the declared label values.This is not surprising because these higher levels are routinelyprovided by manufacturers to ensure compliance with the nu-trition labeling requirements of the U.S. Food and Drug Ad-ministration (FDA; 4). For all the matrixes examined, theRSDs were well within the Horwitz (1) limits of acceptabilityfor the level of analyte determined.

2 Safety Precautions

Normal care should be exercised in the handling of the sol-vents, acids, bases, and buffers used in this procedure. Metha-nol is flammable. Use appropriate hazardous waste procedures.

3 Scope

3.1 Matrixes

Commercial milk- and soy-based powdered and liquid in-fant formulas.

(3.1.1) NIST SRM 1846.—Infant Formula (milk-based)was used as the test matrix to represent milk-based infant for-mulas for in-depth studies by the submitting laboratory (5)and for the interlaboratory peer study.

3.2 Number of Test Samples

For the interlaboratory collaboration, the submitting labo-ratory determined niacin in 27 unspiked test samples ofSRM 1846 and 10 niacin-spiked SRM 1846 test samples. Thepeer laboratory determined the niacin content of 10 test sam-ples of unspiked SRM 1846 and 6 test samples of spiked SRM1846. The submitting laboratory analyzed additional test sam-ples in order to link the interlaboratory study to a variety of


Table 2. Niacin values obtained by the submitting laboratory for commercial infant formulas

Sample Brand Niacin found, µg/g SD, µg/ga RSD, %b Label value, µg/g % of label value

Soy-based powder A 67.1 3.0 4.4 46.5 144

B 64.5 5.3 8.2 46.5 139

C 91.9 9.9 10.8 59.9 154

D 123.4 9.9 8.0 57.8 214

Soy-based liquid A 9.4 1.1 11.6 7.0 134

B 9.4 1.3 13.1 7.0 138

C 9.2 1.1 14.3 7.0 131

Milk-based powder A 126.4 16.3 12.9 91.0 140

B 65.0 3.0 4.7 46.5 138

C 65.2 3.8 5.9 46.5 139

a SD = standard deviation.b RSD = relative standard deviation.

Table 3. Determination of niacin in SRM 1846 Infant Formula by 5 calculation methods

Calculation method Mean, µg/ga SD, µg/gb RSD, %cLower confidence

limit, µg/gUpper confidence

limit, µg/g% of certified

valued

Peak area (unweighted) 58.0 3.9 5.9 53.6 62.5 91.7

Peak height (unweighted) 59.3 6.8 11.1 56.6 56.9 93.7

Peak area (weighted) 59.2 6.0 9.8 54.7 63.7 94.3

Peak height (weighted) 61.2 6.1 9.6 62.1 61.8 96.8

Method avg. (unweighted) 58.7 4.01 6.7 58.4 61.0 92.7

Method avg. (weighted) 60.2 5.8 9.4 61.0 64.0 95.1

MOSAe 59.8 2.05 3.4 58.1 61.4 94.4

a n = 24, except for MOSA mean (n = number of samples).b SD = standard deviation.c RSD = relative standard deviation.d Certified value for niacin in SRM 1846 = 63.3 ± 7.6 µg/g.e MOSA = method of standard additions.

commercial products. Twelve commercial milk- andsoy-based powdered and liquid infant formulas were analyzeda minimum of 3 times each. Four commercial soy-based in-fant formula powders were used for niacin recovery studies.

3.3 Applications

Solid-phase extraction (SPE) column cleanup of acid digestsof milk- and soy-based infant formula matrixes, followed byLC, provides for a simple, rapid analysis for niacin when com-pared with the laborious and time-consuming microbiologicalmethod (6). Naturally occurring compounds, which absorb at260 nm, are removed, resulting in the absence of interferingpeaks at the retention time of the niacin peak. Accuracy, preci-sion, and reproducibility are very good for the 25–100µg/grange usually found in these matrixes. An analyst should beable to process approximately 15 test samples per day, plus 4standards, assuming overnight, automated LC capabilities.

3.4 Limitations

This method is, at present, limited to milk- or soy-based in-fant formula matrixes, which are readily digestible in dilute

H2SO4. Our experience using high protein NIST standardssuch as Reference Material (RM) 8418 Wheat Gluten andRM 8414 Bovine Muscle resulted in low niacin recoveries,due mostly to poor acid digestion. Also, LC of niacin from ar-omatic sulfonic acid (ArSCX–SPE) extracts of these stan-dards and commercial household wheat flours showed inter-fering peaks at the retention time of niacin (7).

4 References

(1) Horwitz, W. (1982)Anal. Chem.54, 67a–76a

(2) AOAC Peer-Verified Methods Program (1993) AOAC IN-TERNATIONAL, Gaithersburg, MD

(3) Sharpless, K.E., Schiller, S.B., Margolis, S.A., Thomas, J.B.,Iyengar, G.V., Colbert, J.T., Wise, S.A., Tanner, J.T., &Wolf, W.R. (1997)J. AOAC Int.80, 611–622

(4) FDA Nutrition Labeling Manual (1993)A Guide for De-veloping a Labeling Database, U.S. Food and DrugAdministration, Washington, DC

(5) LaCroix, D.E., Wolf, W.R., & Chase, G.W., Jr (2001)J.AOAC Int. (in press)


Table 4. Comparison of niacin results by analysis of variance (ANOVA)

Source NDFa DDFb F-Value p-Value

SRM 1846 Infant Formulac,d

Std curve 1 86.1 3.31 0.0723

Peak type 1 86.1 4.08 0.0466

Std curve x peak type 1 86.1 0.22 0.6406

Four commercial soy-based infant formulas

Std curve 1 97.1 3.09 0.0239

Peak type 1 89.0 188.4 0.0001

Std curve x peak type 1 90.8 0.49 0.8806

a NDF = numerical degrees of freedom.b DDF = denominator degrees of freedom.

Table 5. Recovery of niacin added to SRM 1846 Infant Formula

Lab Niacin expected, µg/g Avg. niacin found, µg/ga SD, µg/gb RSD, %c CI, µg/gd Avg. rec., %

Submitting 226.9 223.6 10.0 4.48 4.71 98.6

276.9 285.6 11.8 4.14 5.54 103.2

326.9 341.4 30.4 8.90 14.28 104.4

Peer 134.2 139.8 8.48 6.07 5.88 104.2

172.9 181.5 12.33 6.79 8.54 105.0

250.3 252.2 8.43 3.34 5.84 100.8

a n = 6 for submitting laboratory; n = 7 for peer laboratory.b SD = standard deviation.c RSD = relative standard deviation.d CI = confidence interval.

(6) Gomori, G. (1955)Methods in Enzymology, Vol. 1, S.P.Colowick & N.O. Kaplan (Eds), Academic Press, Inc., NewYork, NY, pp 138–146

(7) LaCroix, D.E., Wolf, W.R., & Vanderslice, J.T. (1999)J.AOAC Int.82, 128–133

(8) Angyal, G. (1996) inMethods for the Microbiological Analy-sis of Selected Nutrients, AOAC INTERNATIONAL,Gaithersburg, MD

(9) DeVries, J.W. (1985) inMethods of Vitamin Assay, 4th Ed.,J. Augustin, B.P. Klein, D.A. Becker, & P.B. Venugopal(Eds), John Wiley & Sons, New York, NY, USA, pp 65–94

(10) Eitenmiller, R.R., & De Souza, S. (1985) inMethods of Vita-min Assay, 4th Ed., J. Augustin, B.P. Klein, D.A. Becker, &P.B. Venugopal (Eds), John Wiley & Sons, New York, NY,pp 385–398

(11) Finglas, G.A., & Faulks, R.M. (1987)J. Micronutr. Anal.3,251–283

(12) Rees, D.I. (1989)J. Micronutr. Anal. 5, 53–62

(13) Vidal-Valverde, C., & Reche, A. (1991)J. Agric. FoodChem.39, 116–121

(14) Ward, C.M., & Trenerry, V.C. (1997)Food Chem. 60,667–674

(15) Chase, G.W., Jr, Landen, W.O., Jr, Solman, A.M., &Eitenmiller, R.R. (1993)J. AOAC Int.76, 390–393

(16) Windahl, K.L., Trenerry, V.C., & Ward, C.M. (1998)FoodChem.65, 263–270

(17) Official Methods of Analysis(1990) 15th Ed., AOAC,Arlington, VA

(18) Cardone, M.J. (1983)J. Assoc. Off. Anal. Chem.66,1257–1282

(19) Wolf, W.R., & LaCroix, D.E. (1998)Fresenius Z. Anal.Chem. 360, 459–464

(20) Varian Separation Products Manual (1997)Solid-Phase Ex-traction: Method Development & Trouble Shooting, VarianSeparation Products, Harbor City, CA

(21) Zief, M., & Kiser, R. (1994)Solid-Phase Extraction for Sam-ple Preparation Manual, J.T. Baker, Inc., Phillipsburg, NJ

(22) Cardone, M.J. (1983)J. Assoc. Off. Anal. Chem.66,1283–1294

(23) Cardone, M.J. (1986)Anal. Chem. 58, 438–445

(24) Mishalanie, E.A. (1996)Intralaboratory Analytical MethodValidation Short Course Manual, AOAC INTERNA-TIONAL, Gaithersburg, MD

(25) SAS Institute, Inc. (1996)SAS/STAT Software: Changes andEnhancements Through Release 6.11, SAS Institute, Inc.,Cary, NC

(26) Dixon, W.J., & Massey, F.J. (1957)Introduction to Statisti-cal Analyses,McGraw-Hill, Inc., New York, NY

(27) Youden, W.J., & Steiner, E.H. (1975)Statistical Manual ofthe Association of Analytical Chemists, AOAC, Arlington,VA

(28) Atherton-Skaff, P., & Sloan, J. (1998) inDesign and Analysisof Equivalence Trials via the SAS System, Proceedings of the23rd Annual SAS Users Group International Conference,SAS Institute, Inc., Cary, NC, pp 1166–1171

(29) SAS Institute, Inc. (1990)SAS Procedures Guide, Version 6,3rd Ed., SAS Institute, Inc., Cary, NC, p. 705

(30) EPA QA/G4 (1994)Guidance for the Data Quality ObjectiveProcess, U.S. Environmental Protection Agency, Washing-ton, DC

(31) Mandel, J., & Linnig, F.J. (1957)Anal. Chem.29, 743–749(32) Chase, G.W., Jr, & Long, A.R. (1997)Food Test. Anal.3,

30–33(33) Mishalanie, E.A. (1997)Basic Statistics for Analytical Sci-

ence, Enigma Analytical, Breckenridge, CO(34) Wilson, A.L. (1970)Talanta17, 21–29(35) Taylor, J. (1990)Statistical Techniques for Data Analysis,

Lewis Publishers, Inc., Chelsea, MI(36) Wilson, A.L. (1973)Talanta20, 725–732

5 Definitions

(5.1.1) ANOVA.—Analysis of variance.(5.1.2) ArSCX–SPE.—Aromatic sulfonic acid–solid-

phase extraction.(5.1.3) CI.—Confidence interval.(5.1.4) DQO.—Data quality objective.(5.1.5) Excel.—Spreadsheet program.(5.1.6) LC.—Liquid chromatography.(5.1.7) LOQ.—Limit of quantitation.(5.1.8) MOSA.—Method of standard additions.(5.1.9) PC-1000.—Computer-generated LC program

used in this study.(5.1.10) RM.—Reference material.(5.1.11) R.—Correlation coefficient.(5.1.12) R2.—Coefficient of determination.


Figure 2. Composite niacin standard curve prepared from data obtained by the peer laboratory.

(5.1.13) RSD.—Relative standard deviation.(5.1.14) SD.—Standard deviation.(5.1.15) SPE.—Solid-phase extraction.(5.1.16) SRM.—Standard reference material.

6 Principle

6.1 Analytical Principles

Niacin determinations by microbiological methods (8) areboth laborious and time consuming. Several LC methods forniacin determination exist in the literature (9–14). However,LC analysis is complicated by interferences from coelutingendogenous substances that absorb at the 260 nm wavelengthconventionally used for UV detection of niacin. Cleanup pro-cedures to remove these interferences have been reported(15, 16). This new method is an adaptation of an open-columngravimetric cleanup procedure reported by Chase et al. (15).Modifications include use of a vacuum manifold system withSPE columns for simple, rapid extraction of a large number oftest samples.

According to the standard preparation procedure used forthe microbiological niacin assay, AOAC Method985.34(17),the test sample is dissolved in (1 + 1) H2SO4 and autoclavedfor 45 min at 121°C. To precipitate the protein, the acid digestis adjusted to pH 6.5 with 7.5M NaOH (8.1.3; ca 5 mL) andimmediately readjusted to pH <1.0 with (1 + 1) H2SO4 (ca1.5 mL). The digest solution is filtered, and the filtrate isbrought to a final volume of 35.0 mL. A 3.0 mL aliquot of thefiltered digest solution is passed through a prewashedArSCE–SPE column (8.1.10), the column is washed, and nia-cin is eluted by using 0.25M sodium acetate/acetic acid buffer(8.1.8). The column (8.1.10) is prewashed 3 times with metha-nol (8.1.1), and then 3 times with water, to remove potentialinterfering compounds and activate the column surface. Afterthe SPE extract is brought to final volume of 15 mL, aliquotsare then introduced into the LC system (9.1.1) for analysis onan anion-exchange polystyrene-divinylbenzene column(9.1.2) with an isocratic mobile phase of 0.1M sodiumacetate–acetic acid buffer (8.1.9), at a column temperature of35EC.

Niacin is determined by UV detection at 260 nm with avariable-wavelength detector or at 254 nm with afixed-wavelength detector (9.1.1).

6.2 Standard Curve Quantitation

Niacin levels in SRM 1846, as measured by peak area andpeak height and calculated by using both unweighted andweighted standard curves, were evaluated for differences byanalysis of variance (ANOVA), as shown in Tables 3 and 4.The results for 24 test samples, each calculated by 4 differentmethods (peak height, peak area, weighted, and unweighted),gave 96 data points for the overall ANOVA comparison. Theniacin data were analyzed as a 3-factor general linear modelby using SAS-PROC mixed with method (weighted vs un-weighted) and peak (height vs area) as the fixed effects anddate (of analysis) as the random effect. Data were calculated atthe 95% confidence level. ANOVA showed that no significant

differences in niacin values were obtained by using these 4 ad-aptations of methods for calculation of standard curves.

ANOVA of 4 commercial soy-based infant formulas, usingthe 4 calculation methods, showed a low degree of variability,the same as obtained for SRM 1846. Niacin values as deter-mined by the method of standard additions (MOSA; Tables 3and 4) were identical to those obtained by the 4 standard curvecalculation methods.

Regardless of the calculation method, niacin values werewithin the uncertainty limits of the certified values ofSRM 1846 (5).

6.3 Method of Standard Additions

Niacin concentrations in spiked test portions were deter-mined by MOSA as a means to evaluate and overcome matrixand/or preparation effects. The plotted MOSA line is the stan-dard curve reflecting the analytical response of the addedanalyte (niacin) in the presence of the matrix, offset by theamount of endogenous niacin in the test sample. Analyte (nia-cin) concentration in the test sample is obtained from theequationy= mx+ bby calculatingx= –b/m, aty= 0 (18). Lin-ear regression analysis of the MOSA data obtained by both thesubmitting (a typical curve) and peer laboratories (xof the8 determinations) gave linear responses with R2 = 0.9755(Figure 3). Table 1 shows the good agreement between the ni-acin levels obtained by both laboratories for SRM 1846.

7 Standards

(7.1.1) Niacin standard.—Nicotinic acid (Sigma Chemi-cal Co., St. Louis, MO).

(7.1.2) Stock standard solution, 10 mg niacin/mL.—Weigh

250 mg niacin (7.1.1), transfer to 25 mL volumetric flask, and

bring to volume with 0.25M H2SO4 (8.1.2). The stock solution

is stable for several weeks under refrigeration.(7.1.3) Working standard solution, 100 µg nia-

cin/mL.—Dilute stock standard solution (7.1.2) 1:100 with0.25M H2SO4.

(7.1.4) LC standards.—Into separate 150 mL Fleakers orbeakers (same type used for the digested test sample), pipet1000, 1200, 1600, 2000, and 2500µL niacin working standardsolution (7.1.3). FollowingProceduredirections in sections12.1–12.5, the standards are treated in the same manner as thetest samples. The working niacin standard curve represents100, 120, 160, 200, and 250µg niacin/column.

(7.1.5) Stability.—The stock solution is stable for severalweeks under refrigeration. LC standards are stable for 2 weeksin the SPE buffer extracts (8.1.8, 8.1.9) under refrigeration.Best results are obtained if the working standard is preparedfresh daily. Discard if turbid microbial growth is observed.

8 Reagents and Supplies

Disclaimer: Mention of a trademark or proprietary productdoes not constitute a guarantee or warranty of the product bythe U.S. Department of Agriculture and does not imply its ap-


proval to the exclusion of other products that may also be ap-plicable.

(8.1.1) Methanol.—LC grade.(8.1.2) Sulfuric acid, 0.25M.—ACS reagent grade. Add

14 mL H2SO4 and bring to 1 L with water.(8.1.3) Sodium hydroxide, 7.5M.—Prepare 300 g/L water.(8.1.4) Sodium acetate, 0.5M.—ACS grade. Prepare

68.0 g/L water.(8.1.5) Sodium acetate, 0.2M.—Prepare 27.2 g/L water.(8.1.6) Acetic acid, 0.5M.—Prepare 29.0 mL/L water.(8.1.7) Acetic acid, 0.2M.—Prepare 11.6 mL/L water.(8.1.8) SPE buffer, 0.25M, pH 5.6.—Mix 48 mL 0.5M

acetic acid and 452 mL 0.5M sodium acetate. Bring to finalvolume of 1.0 L. Buffers are stable for$3 weeks. Discard ifturbid microbial growth is observed.

(8.1.9) Mobile phase buffer, 0.1M, pH 4.0.—Mix 410 mL0.2M acetic acid and 90 mL 0.2M sodium acetate. Bring to fi-nal volume of 1.0 L. Microfilter before LC use. Buffers arestable for$3 weeks. Discard if turbid microbial growth is ob-served.

(8.1.10) SPE columns.—ArSCE–SPE, Bakerbond7090-7, 6 mL (Mallinckrodt-Baker Co., Phillipsburg, NJ);Varian Mega Bond Elut-SCX, 6 mL (Varian Separation Prod-ucts, Harbor City, CA).

(8.1.11) Nitric acid.—ACS grade.(8.1.12) Presoak detergents.—TERG-A-ZYME, HEMO-SOL,

or similar protein-cleaning detergents.(8.1.13) Vacuum manifold.—Baker SPE-12G or

SPE-24G column processor manifold system; Varian VacElut manifold system.


Figure 3. Method of standard additions curves from (A) submitting laboratory and (B) peer laboratory.

B

A

(8.1.14) Vacuum pump.—Benchtop minipump with vari-able needle-valve IN/OUT-flow control to control suctionvacuum.

(8.1.15) Filters.—Polyvinylidene fluoride membrane, hy-drophilic, 0.45µm pore size, with appropriate 47 mm glass fil-ter kit, for filtering LC mobile phase buffer. For test sampleacid digest, use pleated filter paper, 15 cm (S&S, Keene, NH,No. 605 or 588; Whatman International Ltd., Maidstone, UK,No. 40 or 41). Prewash the filter paper with 30 mL water be-fore use.

(8.1.16) Pipets.—10–1000 µL range, with disposablematching tips.

(8.1.17) Collection tubes.—50 mL, conical, graduated,screw-cap centrifuge tubes.

(8.1.18) Pasteur pipets.—Disposable, 5.75 in. (14.6 cm).(8.1.19) Vortex mixer.—Variable speed.(8.1.20) Glassware.—Watch glasses, 75 mm; Fleakers or

beakers, 150 mL; filter funnels, 80 mm; funnel rack; reagentstorage bottles; LC vials and caps.

(8.1.21) pH meter.—With standard reference electrodeand pH-calibrating buffers.

9 Apparatus

(9.1.1) LC system and detector.—Capable of producing ca3000 psi; autosampler desirable for overnight analysis of ex-tracts; variable-wavelength or fixed-wavelength UV detectorfor detection at 260 or 254 nm, respectively.

(9.1.2) LC column.—Anion-exchange PRP-X100 (Ham-ilton Co., Reno, NV), 250× 4.1 mm; in-line anion guard filter;column with oven heater at 35°C. Standard curve correlationcoefficient should be$0.995. An injection of 100µL solutioncontaining 1.0µg niacin/mL should give an area of$5 timesbaseline noise. Mobile phase flow rate is 1.5 mL/min. Discardmobile phase buffer if turbid microbial growth is evident.

(9.1.3) Recommended procedure for optimal LC perfor-mance.—Analysis of a large number of extracts results in adecrease in niacin retention time because of an accumulationof extracted materials on the LC column (9.1.2). The LC sys-tem (9.1.2) should be periodically scrubbed by passing150 mL nitric acid–methanol (1 + 99) through the system, fol-lowed by 100 mL water, at a flow rate of 1.0 mL/min. Our ex-perience was that niacin retention time decreased 4–5 min af-

ter analysis of ca 50–70 samples. Anion-exchange is then re-generated by passing LC mobile phase buffer (8.1.9) throughthe system for 3–4 h at 1.5 mL/min or until baseline stabilityreturns to the previous level. Repeated injections of niacinstandard (7.1.1) at 20µg/injection are made until the baselineand niacin peak retention times are reestablished to previousvalues.

Before a standard curve is prepared, the active sites of theanion-exchange column should be sorbed by injection of atleast one niacin standard at$20µg/injection (7.1.4). Comparebaseline, niacin peak shape, and reproducibility with those ofprevious LC runs. Confirm niacin retention time by spikingwith niacin standard at$2 times the niacin level of the positiveextract. Confirm niacin peak by matching increased peak ar-eas/heights of unspiked and spiked positive extracts.

LC resolution is affected by pH and to a lesser degree bymolarity. The niacin standard curve must be obtained by pass-ing standard niacin through the column (8.1.10). If computer-ized calculation software is used, evaluate slope, intercept,and correlation coefficient by using free-floating versuscurve-forced-through-origin. The niacin standard curveshould give the proper correlation coefficient and coefficientof determination if the above conditions are met. Our experi-ence with 100µL injection loops having 0.010 in. (0.025 cm)diameter periodically resulted in increased reproducibilities(RSDs) of >10% as well as increased variabilities. Substitu-tion of a 100µL loop having a 0.020 in. (0.051 cm) diameterresulted in a more consistent RSD of <10% and a significantdecrease in variability. Greater accuracy is obtained by brack-eting the test sample extracts with the niacin standards at thebeginning and the end of the LC run. The average of the 2 nia-cin standard curves is then used to calculate the niacin level inthe extract. If greater accuracy or precision is desired, 3 injec-tions per extract are recommended. Area counts should be suf-ficiently high that the lowest area count gives an RSD of<10% from multiple injections.

10 Sampling

10.1 Sampling Procedure

This method has been tested with milk- and soy-basedpowdered and liquid infant formulas.


Table 6. Comparison of niacin values obtained by submitting and peer laboratories by means equivalence analysis

Between labs Mean difference, µg/g Test statistic p-Value Decision

Standard curve 4.61 –5.330 0.000 Equivalent

MOSAa 1.37 –2.975 0.001 Equivalent

Between methods

Submitting lab 2.26 –3.129 0.001 Equivalent

Peer lab –0.98 –2.338 0.01 Equivalent

a MOSA = method of standard additions.

10.2 Test Sample Preparation

Once the test sample container has been opened, no specialprecautions are necessary for the powdered materials. Liquidinfant formulas are refrigerated. Powdered formula can be re-frigerated or kept at room temperature.

11 Controls Preparation

Reagent blanks and control reference materials are treatedexactly as laboratory samples.

12 Procedure

12.1 Test Portions

Accurately weigh test sample to give final niacin concen-tration of 100–180µg niacin/analysis. This would be ca 2–4 gfor powdered infant formulas. SRM 1846 sampling constantstudies showed that significant analytical errors can resultfrom using material containing niacin concentrations that arebelow these levels (19).

12.2 Digestion

(12.2.1) Quantitatively transfer milk powder or liquid to150 mL beaker by rinsing test portion into weighing boat with10.0 mL warm (70–90°C) water. Gently swirl mixture untiltest portion is solubilized (complete suspension), add 2.0 mL(1 + 1) H2SO4 (ca 1.25M in final volume), and cover withwatch glass. Autoclave at 121°C for 45 min and cool to room

temperature. Rinse underside of watch glass cover withdropwise additions of water (0.5 mL total), collecting rinsingsin the digestion vessel.

(12.2.2) Using a pH meter, carefully adjust pH to 6.5 with7.5M NaOH (ca 4.5–5.0 mL), followed immediately by1.5 mL (1 + 1) H2SO4 to adjust pH to 1.0.

(12.2.3) Filter pH-adjusted digestate through prewashedfilter paper (8.1.15), and collect filtrate in 50 mL screw-capconical centrifuge tube, being sure that all of the solutionpasses through the filter paper (30 min). Rinse the beaker with5 mL water, transferring the rinsings to the filter paper, allow-ing the rinsings to pass through and be collected with the fil-trate in the centrifuge tube. Repeat the 5 mL rinse a secondtime.

(12.2.4) Bring filtrate to a final volume of 35.0 mL (or an-other volume, marked on the side of the conical tube) with wa-ter. Treat working niacin standard digestate (7.1.3) in the sameway as the test portion digestate. Final volumes for the niacinstandard and test portion should be the same to minimize dilu-tion factor errors.

12.3 Solid-Phase Extraction

(12.3.1) Place the ArSCX–SPE column on the vacuummanifold. The ArSCX–SPE column is prewashed 3 timeswith 6 mL methanol, followed by 3 times with 6 mL water.Discard washings. This column pretreatment is necessary toremove potentially interfering compounds and to activate thecolumn surface (20, 21).


Figure 4. Liquid chromatograms obtained by the submitting laboratory for niacin (a) in unspiked SRM 1846 and (b)in spiked SRM 1846.

(12.3.2) With the manifold stopcocks closed, add 3.0 mLfiltrate to the activated ArSCX–SPE column. For the niacinstandard, add 3.0 mL aliquot to the activated ArSCX–SPEcolumn.

(12.3.3) Adjust vacuum gauge on pump (ca 10 psi) to ob-tain a slow-medium flow rate of ca 1.45 min for 6 mL buffer toflow through the ArSCX–SPE column.

(12.3.4) Open manifold stopcocks, vacuum-suction thefiltrates through the column into the system reservoir tank.Close individual column stopcocks when all liquid has just en-tered column. Wash excess fluid from the column by adding3.0 mL water to the ArSCX–SPE column reservoir, openstopcocks, flush liquid, and close stopcocks when all the washsolution has entered the column. Rinse the reservoir tank withwater and discard wash solution from the reservoir.

(12.3.5) Place tube rack containing 50 mL graduated cen-trifuge tubes into reservoir tank, and place manifold on top so

that the cannula of each column is aligned into the collectiontube.

12.4 Elution of Niacin from Digest Filtrate, UsingArSCX–SPE Column

(12.4.1) Place 3 mL of digest filtrate (12.2) on theArSCX–SPE column (for each of the ArSCX-SPE columnreservoirs, turn on vacuum to –10 psi on the manifold gauge,and open column stopcocks until the meniscus of the filtratereaches the top of the column packing).

(12.4.2) Add 3.0 mL water, and wash column until the me-niscus of the wash reaches the top of the column packing;close the column shut-off valve. Remove manifold and setaside, discard the wash eluate wastes, and rinse the collectionreservoir with water. Drain and towel dry.

(12.4.3) Insert the collection rack containing the labeled50 mL conical centrifuge tubes, and replace the manifold onthe collection reservoir. Add 6.0 mL sodium acetate/aceticacid elution buffer, and collect extraction buffer in the 50 mLconical tubes. Repeat niacin extraction by passing another6.0 mL sodium acetate/acetic acid elution buffer through theArSCX–SPE column, and collect extraction buffer in thetubes containing the first extract. When columns are dry aftersecond buffer extraction, shut off vacuum pump, and removecollection tube rack.

12.5 Niacin LC Analysis

Bring standard and sample extracts to final volume of15.0 mL with 0.5M sodium acetate/acetic acid buffer, usingscale on side of collection centrifuge tube. Vortex carefully,transfer 1.0 mL aliquot of each extract to LC vials, and cap vi-als. The vessel calibration scale is accurate for these measure-ments, as proven by linear regression analysis of the method.

13 Calculations

13.1 Standard Curve

A standard curve is generated by using$4 niacin levelscarried through the complete digestion and extraction proce-dure (12.4.1–12.4.5). Standard niacin levels should bracketthe expected amount present in the extract.

(13.1.1) Plot peak area or peak height response to dose ei-ther by using the computer generated standard curve or bygraphical plot. The standard curve must be obtained for eachextraction/LC series to compensate for daily instrumentalfluctuations.

13.2 Recovery of Niacin Added to SRM 1846

(13.2.1) To the same extract of SRM 1846, add the work-ing niacin standard. Pipet 1000, 1500, and 2000µL niacinworking standard solution to give a niacin curve of 0, 100,150, and 200µg/column. At least 3 levels of added niacin areneeded for recovery analysis.

(13.2.2) The values for total niacin obtained for the spikedextracts are compared with the expected amounts of niacinfrom the spike plus the endogenous amount in the extract. Theratio of niacin recovered to niacin expected is then plotted, and


Figure 5. Liquid chromatogram obtained by the peerlaboratory for niacin in unspiked SRM 1846.

the linear regression equation should approach m = 1.0 andb = 0.

(13.2.3) Another calculation method for recovery ofadded niacin spikes in a positive extract involves subtractionof the niacin found in the unspiked extract from that found inthe spiked extract. The percent recovery is calculated by di-viding the resulting observed increment due to the spike by theexpected added amount and multiplying the result by 100.This method accounts for the background level of the endoge-nous analyte in the extract (18, 22–24).

Ratio = niacin found/niacin expected

Niacin found =total niacin obtained – niacin per weight of test portion

Niacin expected =niacin per weight of test portion + added niacin spike

13.3 Method of Standard Additions

(13.3.1) Draw the dose-response curve through the y-axis.The negative intercept on the x-axis is the amount of niacin inthe test sample used.

(13.3.2) Another method of calculation is to use the linearregression equation,y = mx+ b. At y = 0, x = –b/m, the un-known analyte concentration in the unspiked extract is equalto the ratio of the unknown spiked extract response to theslope of the additions line (18).

13.4 Computer-Generated Standard Curve

Peak areas or peak heights are obtained by using the appro-priate LC computer program. Use the average of 2 peak areasfor each niacin standard if extracts are bracketed. The proce-dural blank data are not shown but are consistently a zerobaseline in the liquid chromatogram.

13.5 Excel-Generated Graphical Plot, y = mx + b

x (µg niacin found) =[y (peak area) –b (intercept)] ÷m (slope)

Niacin (µg/g) =(µg niacin found) ÷ weight of test portion

13.6 Recovery Calculations

For ratio plot.—Niacin (µg/test portion) found ÷ niacin(µg/test portion) expected. Linear regression equation:y = mx + b.

Positive test portion.—Niacin (µg/total test portion weight)found – niacin (µg/total test portion weight) in unspiked testportion = recovered niacin spike (µg/test portion).

14 Test Results Report

14.1 Data Results

Data obtained by the submitting and peer laboratories aresummarized in Tables 1–6 and Figures 1–9.

14.2 Statistical Analysis

Data were statistically analyzed with either Excel or ProcMixed SAS/STAT software (25). Where applicable, analyti-cal data measurements were statistically examined for ex-treme differences from the main body of measurements by us-ing the Dixon test for outliers (26, 27). Intra- andinterlaboratory precision were evaluated by testing for accept-able variability, using the limits of acceptability criteria ofHorwitz (1). All assumptions of linear regression analysiswere met.

14.3 Means Equivalence Analysis

Tests for niacin values were analyzed by the ratio of meanstechnique described by Atherton-Skaff and Sloan (28). Thedistribution of niacin values, obtained by the submitting andpeer laboratories, and the method of obtaining these niacinvalues by the standard curve and MOSA calculation methodswere tested for normality by using the Wilcox test in SASPROC UNIVARIATE (29). Because the hypothesis for nor-mality was not rejected, the niacin values and method of cal-culation were equivalent between calculation methods and be-tween the submitting and peer laboratories.

14.4 Interlaboratory Peer-Verified Study

The participants were 2 U.S. Government laboratory ana-lysts. The AOAC Peer-Verified Methods Program (2) wasused as the basis of this study. Test samples were NIST SRM


Figure 6. Liquid chromatogram, obtained by thesubmitting laboratory, showing that the nucleotides andnicotinamide are separated from niacin.

1846 Infant Formula. The integration of the Data Quality Ob-jective (DQO) Process (30) in conjunction with the use of ref-erence materials was followed for the application for valid sta-tistical procedures for data acceptance, data verification, andvalidation of method performance. All determinations by thepeer laboratory were in duplicate. No special handling precau-tions except those specified in the protocol were necessary.

14.5 Commercial Soy-Based Infant Formula

Individual niacin values obtained by the submitting and peerlaboratories for commercial soy-based infant formula powderswere higher than the label claim, which is not surprising.Higher levels of niacin are usually added so that compliancewith nutrition labeling requirements is satisfied (4). Variabilitywas consistent and within the limits of acceptability (1).

14.6 Liquid Chromatograms

Liquid chromatograms from the submitting laboratory forthe ArSCX–SPE extract of SRM 1846, unspiked and spikedwith niacin, are shown in Figure 4. The liquid chromatogramobtained by the peer laboratory for niacin in SRM 1846 isshown in Figure 5. Both laboratories obtained a clean, single,interference-free niacin peak. Figure 6 shows that niacin iscompletely separated from the 5 common nucleobases, whichabsorb at 260 nm with UV detection and may occur in thesematerials. The liquid chromatograms obtained for unspikedand niacin-spiked commercial soy-based infant formula pow-der are shown in Figure 7.

15 Quality Assurance

15.1 Dose-Response Curves

The dynamic range of 0.4–100µg/mL used in this studywas linear, and the detection limit was 0.2µg/mL. Peak arearesponse for the niacin standard dose-response curves was ob-tained by 3 different methods, i.e., computer-generated(PC-1000), Excel graphical plot, and MOSA. The actual lin-ear dynamic range is much higher than the range used in theseexperiments. MOSA was used to diagnose any matrix bias er-rors due to test portion extraction and complex matrix interfer-ences, and to verify and validate the values obtained by thePC-1000 and Excel graphical standard curves (Figure 8).

15.2 Reproducibility and Recovery

The niacin values obtained by the submitting and peer lab-oratories for niacin-spiked SRM 1846 Infant Formula areshown in Table 5. Values obtained by both laboratories showslightly high recoveries and RSDs that are well within the lim-its of acceptability (1).

15.3 Recovery as a Function of Concentration

The ratios of niacin found/niacin added (expected) thatwere obtained by the submitting and peer laboratories for atypical test portion of SRM 1846 are shown in Figure 9. Devi-ations from the theoretical slope (= 1) and intercept (= 0) val-ues of the response curves obtained are attributed to experi-mental variability (31) or to “corrigible systematic error,” as


A

Figure 7. Liquid chromatograms obtained by the submitting laboratory for niacin (A) in unspiked commercialsoy-based infant formula powder and (B) in spiked commercial soy-based infant formula powder.

B

defined by Cardone (18, 22, 23) as consisting of both constantand proportional errors. The intercept of the found/expectedline is an indication of the amount of niacin in the true testsample blank. In the absence of a true “zero control referencematerial” (32) in which the matrix contains no analyte, properblank correction for the matrix relative to the niacin standardcurve in buffer is not possible. Because of the background sig-nal, some discrepancy can be found in the estimate of recoveryfrom spiked matrixes as compared with the estimate of recov-ery from spiked chemical buffer solution blanks. Linear re-gression analyses of the found/expected ratio curves for boththe submitting and peer laboratories show that the amount ofniacin found is linear (R2 >0.99), with the slope of the ratio offound/expected amounts near unity, within experimental er-ror, for all test samples analyzed. This means that >98% of thevariation in the found value is accounted for only by the varia-tion in the expected niacin value over the concentration rangeanalyzed. Thus, the recovery of added niacin is consistent andcomplete over the dynamic range of added niacin that was in-vestigated (24, 33).

15.4 Dose-Response Curves for Niacin-SpikedPositive Test Sample

Figure 3 shows the dose-response curves for niacin and ni-acin-spiked SRM 1846 obtained by the submitting and peerlaboratories. The unweighted regression analyses of niacinadded versus instrument response for both peak area and peakheight curves indicate an absence of matrix bias because,within error, the slopes of the MOSA identity and standardcurves are nearly parallel, with an intercept near zero. MOSAcalibration results in normalization of proportional error fromthe point of the procedure where the analyte spike is intro-duced into the matrix. Also, LC/diode-array spectra of all ex-tracts showed a single peak, which is identical to that of thestandard niacin digest. For the 3 matrixes analyzed, these cri-teria indicate the absence of interferences under the niacinpeak (18, 23, 24). A linear regression analysis of the ratio offound/expected niacin from spiked SRM 1848 (positivespiked extract) shows a similar response with a slope nearunity. Thus, the background analyte level in the test sample isvery low and near or within the intended range of the method.


A

Figure 8. Niacin dose-response curves obtained by (A) the submitting laboratory and (B) the peer laboratory.

B

15.5 Test of Standard Curve Repeatability

For the 11 niacin standard curves obtained by the submit-

ting laboratory, over a 3-month period in 1997, by using peak

area calculations, all RSDs are <16% for each level of niacin

determined. Curves calculated from peak heights for these

data show an RSD of >16% for the 4µg niacin/column level.

All higher niacin levels were within the acceptable range. Be-

cause the 4µg niacin/column level is approaching the detec-

tion limit, the minimum amount of niacin used for the stan-

dard curve should be 6µg niacin/column. The peer laboratory

obtained 4 niacin standard curves by using peak height calcu-

lations. The RSDs for peak area and peak height calculations

by both the submitting and peer laboratories are under the ac-

ceptable 16% determined for the analyte level according to the

Horwitz criteria (1).

15.6 Test of Linearity

Laboratory data for the computer-generated niacin stan-dard curves in Figures 1 and 2 were used for the test of linear-ity. Linear regression analyses of the mean values for the datasets obtained by the submitting and peer laboratories showvery good linearity over the analytical range with R andR2 = 0.9990 for the data sets, obtained on separate days.

16 Footnotes and User Comments

16.1 Acknowledgments

(16.1.1) Eural Porter and Dennis Cantellops, Atlanta Cen-ter for Nutrient Analysis, U.S. Food and Drug Administration,


Figure 9. Ratios of niacin found/expected obtained by (A) the submitting laboratory and (B) the peer laboratory.

B

A

Atlanta, GA, for the collaborative analysis and their helpfulrecommendations and comments to improve this protocol.

(16.1.2) Mary Camp, Biometrical Consulting Service,U.S. Department of Agriculture, Beltsville, MD, for thePROC/MIXED SAS/STAT analysis of data, and the MeansEquivalence Analysis using the PROC UNIVARIATE/SASprogram.

(16.1.3) Raj Patel, Food Composition Laboratory, U.S.Department of Agriculture, Beltsville, MD, for developmentof the Excel program for data processing.

16.2 Performance Characteristics

The performance characteristics of an analytical methodare a set of quantitative and experimentally determined valuesfor parameters of fundamental importance in assessing thesuitability of a method for any given purpose (34). Thus, per-formance characteristics refer to the quality of the results ob-tained.

16.3 Data Quality Objectives

DQOs are quantitative specifications that provide for es-tablishing the minimum quality of data for a specific investi-gation. These DQOs also provide for establishing both quali-tative and quantitative method performance characteristicsand are used to evaluate decision criteria for data acceptance(24, 30).

Integration of the DQO process with the use of appropriatereference materials provides an additional mechanism for es-tablishing the potential of method performance for accurateanalytical results (1, 19). For this Peer-Verified Methodsstudy, the required DQOs include verification and validationof niacin data that are consistent and within the limits of thereference value for niacin in NIST SRM 1846 Infant Formula.Additional DQOs for variability of the analytical results areRSDs of <8.6% in accordance with the Horwitz (1) criteria forlimits of acceptability for analytes determined at the 60µg/glevel of the analyte tested.

For this niacin method, the best estimates of statistical pa-rameters resulting from intralaboratory and interlaboratorycollaborative studies are evaluated by application of valid sta-tistical procedures (5, 7, 35).

16.4 Recovery

(16.4.1) Response as a function of concentra-tion.—Dose-response curves obtained by the submitting andpeer laboratories are shown in Figure 8 for (i) MOSA spiked,(ii ) identity (MOSA spiked minus the intercept of theunspiked analyte), and (iii ) niacin standard curves. Figure 8Ais a typical set of curves obtained by the submitting laboratory.Figure 8B is the average of the 8 MOSA data sets (i, ii ), withthe average of the duplicate standard curves (iii ), obtained bythe peer laboratory. Linear regression analysis of thedose-response for the added niacin with peak area from thesubmitting laboratory shows a linear response with R2 close tounity. The slopes of the MOSA and identity curves are identi-cal by definition. Because the slopes of the identity and stan-dard curves are nearly identical, no significant matrix interfer-

ences are observed. These data show that within experimentalerror, the niacin-spiked test sample is chemically and physi-cally representative of the niacin analyte in the test samplematrix examined. Thus, the response is accounted for only bythe variation of the concentration of the analyte (niacin) beingmeasured (18, 23, 24, 36).

(16.4.2) Niacin recovery as a function of concentra-tion.—Data from the submitting and peer laboratories for nia-cin recovery as a function of expected concentration for a setof standard additions to SRM 1846 are shown in Figure 9. Lin-ear regression analysis shows that the response as a functionof concentration is linear with R2 = 0.9970 with slopes ofunity, within experimental error. These data demonstrate thatfor spiked positive test sample analysis, the background inter-ference level is very low and is not significant within the in-tended measurement range of the method. Thus, the recoveryof niacin is consistent and quantitative over the range investi-gated (18, 22–24, 33).

(16.4.3) Recovery of added niacin.—Results from thesubmitting and peer laboratories for recovery of niacin addedto SRM 1846 are given in Table 5. The average recoveries ob-tained by both laboratories range from 98 to 105%.

(16.4.4) Means equivalence statistical analysis.—Niacinvalues obtained between the 2 laboratories and between the2 calculation methods (standard curve and MOSA) were com-pared by the ratio of means technique described by Ather-ton-Skaff and Sloan (28). For this ratio test, it was assumedthat the ratio tested at the 10% level, with a significance of0.05. Therefore, when the decision is equivalence, the conclu-sion is that with 95% confidence, the 2 population means dif-fer by#10%. Results of this statistical analysis (Table 6) showthat both the submitting and peer laboratories obtained the sameniacin values by using both the standard curve and MOSA cal-culation methods. Because the hypothesis of normality was notrejected, the 2 data treatment means differ by#10%. Thus, thedata obtained by both laboratories are the same.

(16.4.5) Limit of detection.—The limit of detection (LOD)is the lowest content of the analyte that can be detected andidentified (38). For this method, the LOD, as defined by avalue of 3 times the SD of the lowest standard used in the anal-ysis, is 0.2µg/mL for the solution introduced into the liquidchromatograph.

(16.4.6) Lower limit of quantitation.—The limit ofquantitation (LOQ) is the smallest concentration for whichquantitative tests of the precision and bias of analytical resultscan be made (36). For this method, as defined by a value of10 times the SD of the lowest standard used in the analysis, thelowest concentration point at which niacin can be quantitatedis 0.7µg/mL, corresponding to a level of 20µg niacin per ana-lytical sample. Sampling constants (contribution to variationof 1% due to sampling) for niacin in SRM 1846 are in therange of 2–3 g. Use of subsample sizes lower than this amountof material can introduce significant variation when this SRMis used to validate analytical methodology (19).

(16.4.7) Mechanism of ArSCX–SPE column extrac-tion.—The mechanism for cleanup and extraction of niacin byArSCX–SPE from interfering compounds present in the acid



digest is by direct cationic exchange. The niacin forms ionicbonds with the polar SO3 group, and the niacin is removed bythe combination of pH and the Na+ counter ion. Additionalseparation of niacin from interfering compounds is obtainedby partitioning in the acetate buffer (21).

(16.4.8) Effect of pH.—Evaluation of niacin data and liq-uid chromatograms from acid digests that were pH-treatedand untreated in the preparation step did not show appreciabledifferences in niacin values obtained for SRM 1846. How-ever, all other matrixes examined must undergo thepH-adjusted procedure to obtain valid niacin results. Thus, thepH adjustment step should be included in preparing all ma-trixes for SPE and LC analyses. Also, to ensure complete ion-ization of the niacin moiety, the final pH should be <1.0.Buffer pH variations of ± 0.2 for both extraction of niacinfrom the ArSCX–SPE columns and LC separation do not af-fect results. However, greater variation in pH can influencereproducibility.

(16.4.9) Special glassware cleaning.—All glassware usedin the digestion/pH adjustment/filtration steps and digest stor-age vials should be presoaked for$2 h by using products suchas TERG-A-ZYME, HEMO-SOL to remove precipitatesfrom glassware before regular washing.

(16.4.10) Sources of ArSCX–SPE column.—TheMallinckrodt-Baker ArSCX–SPE columns were used exclu-

sively in this study. However, because of production and sup-ply problems, Varian Separation Products ArSCX–SPE col-umns (Mega Bond Elut-SCX, size 6 mL/1 g) were alsoinvestigated. Niacin values were the same as those obtainedby using the Bakerbond SPE column used in this study.

(16.4.11) Other food matrixes.—Niacin values obtainedfor commercial milk-based and soy-based infant formulaswere approximately 35% higher than label claim values. Thisis not unexpected because food fortification levels are usuallysomewhat higher than label claim to ensure compliance (4).Thus, this method can be applied to the determination of theniacin content of commercial milk-based and soy-based pow-dered and liquid infant formulas.

(16.4.12) Conclusions.—Use of an ArSCX–SPE systemis an improved refinement of the gravimetric sulfonatedFlorisil method. Besides being easier and less time consum-ing, the method increases test sample throughput by the use ofmany columns with a vacuum manifold system instead of asingle gravimetric column. LC preparations are clear of en-dogenous interfering components that absorb at 260 nm. Thedose-response of niacin is linear for both standard curves andMOSA, with a detection limit of 0.2µg/mL for solutions in-troduced into the liquid chromatograph. This method is appli-cable to the various infant formula matrixes tested in thisstudy.

Documents

Determination of Niacin in Infant Formula by Solid-Phase ... · determination of niacin in milk-based and soy-based infant formula. Analysis is in 3 steps: test sample digestion,