PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J AOAC Int. Author manuscript; available in PMC Sep 15, 2011.
Published in final edited form as:
J AOAC Int. 2011 May-Jun; 94(3): 803–814.
PMCID: PMC3173719
NIHMSID: NIHMS318251
Dietary Supplement Laboratory Quality Assurance Program: The First Five Exercises
Melissa M. Phillips, Catherine A. Rimmer, Laura J. Wood, Katrice A. Lippa, Katherine E. Sharpless, David L. Duewer, and Lane C. Sander
National Institute of Standards and Technology, Analytical Chemistry Division, Material Measurement Laboratory, Gaithersburg, MD 20899-8392
Joseph M. Betz
National Institutes of Health, Office of Dietary Supplements, Bethesda, MD 20892-7517
Corresponding author’s ; melissa.phillips/at/nist.gov
The National Institute of Standards and Technology (NIST) has established a Dietary Supplement Laboratory Quality Assurance Program (DSQAP) in collaboration with the National Institutes of Health Office of Dietary Supplements. Program participants measure concentrations of active and/or marker compounds as well as nutritional and toxic elements in food and dietary supplements distributed by NIST. Data are compiled at NIST, where they are analyzed for accuracy relative to reference values and concordance among the participants. Performance reports and certificates of completion are provided to participants, which can be used to demonstrate compliance with current Good Manufacturing Practices as promulgated by the U.S. Food and Drug Administration. The DSQAP has conducted five exercises to date, with total participation including more than 75 different laboratories and many more individual analysts.
The dietary supplement industry in the United States is booming, with 66% of adult Americans considering themselves to be supplement users (1, 2). Consumption of dietary supplements, including vitamin and mineral supplements, represents an annual U.S. expenditure of more than $25 billion (1). These figures represent an increasing American trend, as the dietary supplement industry experienced 10 to 15% growth in 2009, despite a sluggish economy (1, 3). As a result, the quality and safety of these products is critically important and must be both verified and maintained.
The Dietary Supplement Health and Education Act of 1994 (DSHEA) amended the Food, Drug, and Cosmetic Act to create the regulatory category called dietary supplements. The DSHEA also gave the U.S. Food and Drug Administration (FDA) authority to write current Good Manufacturing Practices (cGMPs) that require manufacturers to evaluate the identity, purity, and composition of their ingredients and finished products. To enable members of the dietary supplements community to improve the accuracy of the measurements made in compliance with these and other regulations, the National Institute of Standards and Technology (NIST) established a Dietary Supplement Laboratory Quality Assurance Program (DSQAP) in collaboration with the National Institutes of Health (NIH) Office of Dietary Supplements (ODS).
The DSQAP was originally established to improve the accuracy of measurements in the dietary supplement community. The program includes measurements of active or marker compounds, nutritional elements, contaminants (toxic elements, pesticides, and mycotoxins), and fat- and water-soluble vitamins in foods as well as botanical dietary supplement ingredients and finished products. The program also offers tools for demonstration of compliance with the cGMPs established by the FDA. In addition, the DSQAP assists the ODS Analytical Methods and Reference Materials program at NIH in supporting the development and dissemination of analytical tools and reference materials (4, 5). In the future, results from DSQAP exercises could be used by organizations such as AOAC INTERNATIONAL to identify problematic matrixes and analytes for which an Official Method of AnalysisSM would benefit the dietary supplement community.
NIST has experience in the area of QA programs, with longstanding efforts such as the Micronutrients Measurement Quality Assurance Program and the Organic Contaminants in the Marine Environment Quality Assurance Program, which have been in operation since 1984 and 1987, respectively (612). In addition, NIST and NIH began a Vitamin D Metabolites Quality Assurance Program in 2009 (13). In contrast to the above comparability programs, in which a set of analytes is measured repeatedly over time in the same or similar matrixes to demonstrate laboratory performance, the DSQAP takes a unique approach. Laboratories participating in the DSQAP typically do not have internationally accepted methodologies for the samples being analyzed, nor do the samples and matrixes of interest remain static over time. In a constantly changing marketplace, these laboratories must demonstrate their performance, including scientific validity and fitness for purpose, rapidly upon inspection by the FDA. One method for demonstration of competency is through comparison to performance of other laboratories, and the DSQAP provides a mechanism for such comparisons. Although the DSQAP currently functions as a survey program, some elements of the exercises, such as those involving inorganic analytes or vitamins, may have the potential to evolve into a true comparability program in the future.
In this paper, the purpose, goals, and function of the DSQAP are described, and the first five exercises of the DSQAP are summarized. In addition, the lessons learned from the initial stages of the program are reported to demonstrate the value of the DSQAP to participants and the dietary supplement community.
Samples
Samples and analytes for each study are selected based on polling of past participants, opinions of experts, and priority within regulatory and QC rankings. When possible, Standard Reference Materials (SRMs) developed and characterized by NIST are used as unknown samples to ensure the highest degree of material homogeneity and confidence in the target values. If a material with a certified value is unavailable, a material may be obtained from another agency or prepared at NIST specifically for use in the exercise. Unknown samples are paired with appropriate control materials whenever possible. Ideally, control materials are SRMs or certified reference materials from another national metrology institute with certified values for the analytes of interest. The control material is selected to have a similar concentration of the analyte in comparison to the unknown sample, to have a similar matrix, and to provide comparable or fewer challenges in sample preparation. Comparison of values obtained for the control material and unknown sample can also help identify potential analytical issues with calibration or sample preparation. Participants select the studies in which they are interested in participating for a given exercise, and samples are shipped directly to all participants. An overview of the samples used in Exercises A through E of the DSQAP is given in Table 1.
Table 1
Table 1
Summary of analytes and samples used in DSQAP Exercises A–Ea
Instructions to Participants
Participants are provided with storage and stability information for all samples. Laboratories are asked to use standard in-house methods of analysis for each sample, including their own calibrants and routinely applied methods of sample preparation, chromatographic separation, and detection. Participants are provided with the certified values for the control material when available, allowing participants to evaluate in-house analytical methodology prior to characterization of the unknown sample. Upon request, participants can also receive additional information such as target ranges for samples, analytical methods used at NIST, and troubleshooting assistance when problems are suspected.
Data Analysis
Results from each participating laboratory are compiled at NIST approximately 6 months after sample distribution. Data are assessed for accuracy (when possible) and concordance among participants. Certificates of participation and detailed reports, including target and consensus ranges, Z-scores, and a thorough discussion of trends, analytical challenges, and suggestions for improvement, are provided to participants. Full reports, including all data tables and discussion of each study, are available on the DSQAP website (14).
Program Goals
One important goal of the DSQAP is to provide information to participating laboratories regarding overall performance with respect to measurement of nutrients, contaminants, and marker compounds in various food and dietary supplement matrixes. The DSQAP is not a proficiency testing program, in that no pass/fail criteria exist or are provided to participants, and program coordinators work with participants to improve methodology and discuss analytical problems. Knowledge gained through participation in the DSQAP and documentation of participation may be used by laboratories to demonstrate appropriateness of analytical methodology and commitment to high quality measurements to regulatory agencies. To further improve performance, participant workshops are held to discuss results as well as methodological advancements in the characterization of dietary supplements.
Participating Laboratories
In February 2007, four sample sets were distributed to 18 laboratories in the first pilot exercise of the DSQAP. The fifth exercise of the DSQAP concluded in September 2010, with five sample sets distributed to 42 laboratories. In all, 75 laboratories have been involved in at least one DSQAP exercise (Figure 1). The program is free for participants, and to ensure anonymity, laboratories are assigned identification codes for each exercise known only to the laboratory itself and the program coordinators.
Figure 1
Figure 1
Distribution of participants in Exercises A–E of the DSQAP. In total, 75 laboratories have participated in at least one exercise.
Nutritional Elements
For determination of nutritional elements in foods and dietary supplements, methods of calibration and sample preparation are critically important. Exercises A, C, and E have included nutritional element studies and have focused on these two major analytical challenges. In the pilot exercise (Exercise A), Ca, Mg, Fe, and Zn were determined in SRM 3280 Multivitamin/Multielement Tablets, with SRM 2711 Montana Soil as a control (1535). Only four of the 18 laboratories participating returned data, but three of these laboratories reported values within the certified range for all four elements in the multivitamin tablet unknown. In nearly all cases, reported values were low for the control, where the levels for each element were two to three orders of magnitude lower than those in the unknown. Most likely the low levels for the control are a result of incomplete digestion, as several laboratories reported an inability to successfully digest the soil matrix. SRM 2711 was chosen for this study because it or SRM 2702 Inorganics in Marine Sediment was used at NIST in the certification of nutritional elements in SRM 3280.
In Exercise C, Ca, Na, P, and Zn were determined in fortified milk powder, with SRM 3244 Ephedra-Containing Protein Powder as a control (3638). The target values for nutritional elements in the fortified milk powder were determined at NIST as part of an international interlaboratory comparison exercise. Compared to Exercise A, the unknown and control were more appropriately matched in both concentration and potential matrix challenges. With 10 to 13 laboratories reporting, the consensus data encompassed the target value, with some outliers. Upon further examination, the data for Ca and Zn showed a trend in which laboratories reporting low values for the control also reported low values for the unknown. The same was true for laboratories reporting high values for the control. This type of trend suggested that many laboratories may have had difficulties with calibration for Ca and Zn, possibly indicating a calibration performed outside of the range necessary for these samples. In addition, some laboratories reported data for calcium that were much lower with respect to the reference value for the control than for the unknown, indicating that the analytical methods that were used may not have corrected for potential matrix effects. To isolate the source of these analytical problems, the similarity between the control and the unknown sample matrixes were more carefully considered in subsequent exercises.
In Exercise E, Ca, Fe, and Zn were determined in ground and flake fortified breakfast cereals of both wheat and rice origin. Twenty-three participants were provided with six samples, including candidate SRM 3233 Fortified Breakfast Cereal, ground and flake wheat cereal, ground and flake rice cereal, and a mixture of the wheat and rice flake cereals; 19 laboratories reported results. This study was designed to investigate the participants’ ability to sufficiently homogenize a sample and the effect that sampling-related inhomogeneity would have on the overall measurement uncertainty. As observed in Exercise C, the consensus data encompassed the target value for nearly all laboratories for candidate SRM 3233, with some outliers. The same trend discussed previously was also observed in the cereal matrix, with the results of a single laboratory compared for all samples confirming the conclusion that a calibration or digestion issue is occurring in many laboratories. As a result, upcoming exercises will address the calibration problem, perhaps by inclusion of calibration materials to be measured as part of the exercise.
Contaminants
Determination of contaminants in botanicals, dietary supplements, and foods is challenging because the levels of contaminants are often very low, approaching the LODs of many analytical methods. In addition, the cGMPs apply to all dietary supplement manufacturers and specify the necessity of testing for reasonably anticipated contaminants in addition to active or marker compounds named in a specification. Exercises A–E have all included a sample for contaminant determination, including toxic elements in Exercises A–D and organic contaminants in Exercise E. In the pilot exercise, participants determined Pb in two botanical dietary supplement preparations. The control tablet (SRM 3243) contained Ephedra sinica, and the unknown tablet (SRM 3248) contained Ginkgo biloba (3739). With only four laboratories reporting data, insufficient information was available to draw sound conclusions. Superficially, the data appeared to be scattered for both the unknown and control.
In Exercise B, the unknown and control materials were more closely matched when participants determined As in a commercial extract of E. sinica (SRM 3242) as the control and in the aerial parts of the E. sinica plant (SRM 3240) as the unknown (37, 38). The results are shown in a Youden plot in Figure 2 (40). The level of As in the control was about five times the level in the unknown. Overall the results were very good, with the data consensus encompassing the target value for both the control and the unknown with few outliers. As seen for the nutritional elements, however, laboratories reporting low values for the control also reported low values for the unknown, and the same was true for those reporting high values. This trend may indicate a possible calibration or digestion issue, and because As can easily be lost during some digestion processes, problematic digestion is more likely. In addition, interferences from ArCl may cause values to be biased high when As is determined by inductively coupled plasma/MS (ICP/MS).
Figure 2
Figure 2
Comparison of results from Exercise B for determination of arsenic in SRM 3240 Ephedra sinica Plant with SRM 3242 Ephedra sinica Extract as a control. The red box encompasses the NIST-certified values, and the blue box identifies the community consensus (more ...)
In Exercise C, the unknown and control materials were again closely matched as participants determined As and Cd in bitter orange fruit (SRM 3258) as the control and in a commercial extract of bitter orange (SRM 3259) as the unknown (41). The concentration of Cd was similar in both materials, while the As level was a factor of two higher in the unknown than in the control. For both toxic elements, the consensus of the results included the target value for both the unknown and the control with few outliers. Of the outliers, many values for Cd in the unknown were surprisingly high for several laboratories, despite accurate values for the control. Upon further examination, Sn and Mo are present in higher levels in the unknown than the control, which may lead to additional interferences in Cd measurement by ICP/MS for the unknown sample matrix. The certified values were determined at NIST by ICP/MS following a separation by anion exchange chromatography, so any potential interference from Sn was removed. Understanding that a separation may be too labor-intensive for routine QC analysis, participants were encouraged to perform a semiquantitative ICP/MS scan of the unknown and control prior to measurement in order to identify and correct for potential isobaric interferences.
In Exercise D, the examination of Pb in G. biloba products was revisited from the pilot exercise. In this exercise, SRM 3247 Ginkgo biloba Extract was used as the control, and SRM 3248 Ginkgo-Containing Tablets was used as the unknown (38). The concentration of Pb in the control was approximately five times the concentration in the unknown. The consensus of the results was within the target ranges for both the unknown and the control, and most laboratories made very precise measurements of Pb in these materials. A small group of participants reported results that were below the target value for both the control and the unknown. Because both the control and the unknown were relatively easy to digest, a calibration error is the most likely explanation for these low results. To demonstrate the importance of moisture corrections in the use of control materials, participants’ data were compared to NIST target values both with and without moisture correction (Figure 3). When the moisture content of the control (approximately 2% water) and unknown (approximately 5% water) materials is not considered, the data reported by participants appear to be precise but inaccurate, with a low bias. Participants were shown this data to encourage consideration of moisture when using certified control materials.
Figure 3
Figure 3
Comparison of participant data for Pb in SRM 3248 Ginkgo-Containing Tablets when the target value is corrected for moisture (top) and uncorrected (bottom). When uncorrected, the participant data are precise, but biased low.
In Exercise E, an organic contaminant was investigated for the first time. Participants determined aflatoxins B1, B2, G1, and G2, along with total aflatoxins in SRM 2387 Peanut Butter as the control and ground peanuts as the unknown (38, 42, 43). Ground whole peanuts were purchased from the Food and Environment Research Agency’s Food Analysis Performance Assessment Scheme (FAPAS). The peanut unknowns were analyzed by 65 laboratories between March and April 2009, as part of FAPAS to establish a target value for each of the aflatoxins. The levels of aflatoxins B1 and B2 in SRM 2387 were measured as part of an interlaboratory intercomparison exercise, while aflatoxins G1 and G2 were not measured in the SRM (42). For all aflatoxins, the concentrations in the unknown and control were within a factor of two. Only five laboratories reported data, but for all compounds, the reported values of all participants were within the target range (Figure 4). The success of this particular study indicates that the methods being used for aflatoxin determination are sufficient, and more participating laboratories are needed to draw further conclusions.
Figure 4
Figure 4
Participant data for determination of total aflatoxins in ground peanuts from DSQAP Exercise E. The consensus range is completely contained within the target range, and all laboratories performed well on this study.
Fat-Soluble Vitamins
The instability of fat-soluble vitamins makes their determination challenging for even the most experienced analyst. Analysis can be complicated when vitamins are encapsulated or if the food or supplement has a high fat content or is not fortified. In addition, the weak chromophores of vitamins D and K make their determination even more difficult at the low levels found in many foods and supplements. For this reason, Exercise A focused on determination of vitamins A and E in SRM 3280 Multivitamin/Multielement Tablets as the unknown and SRM 1849 Infant/Adult Nutritional Formula as the control material (4451). For both vitamins, the concentration in the control was two orders of magnitude lower than the concentration in the unknown. Participants reported retinol values that were both higher and lower than the target value for the infant formula control, while nearly all laboratories reported values lower than the target value for the multivitamin tablet unknown. These low values are likely related to the encapsulation used to stabilize retinol in the preparation of the multivitamin. In addition, retinol was present as retinyl acetate in the multivitamin and as retinyl palmitate in the infant formula, which may have resulted in calibration issues for some laboratories. Values for α-tocopherol, however, were more widely distributed for the control than for the unknown, with most participants reporting values within the target range for the unknown. The low levels of α–tocopherol in the control may have caused participants to use an inappropriate calibration range for the multivitamin unknown. Another possibility is that some laboratories may have had experience with either multivitamin analysis or infant formula analysis, but not both, making this control/unknown pair less than ideal.
In Exercise C, another study was conducted using retinol as the analyte. The unknown in this exercise was a fortified milk powder, the control SRM 1849 Infant/Adult Nutritional Formula (4548, 50, 51). The target value for retinol in the fortified milk powder was determined at NIST as part of an international interlaboratory comparison exercise. The retinol concentration in these two materials was the same, as were the potential extraction issues. Overall, the consensus of the results included the target value for both the control and the unknown. The trends in the data indicate a possible calibration issue, which is common in the quantitative measurement of retinol. The concentrations of the calibration solutions are traceable to a spectroscopic absorption coefficient, not to a gravimetric value, which requires careful purity analysis of the source material used to prepare the chemical calibrant. In addition, because 13-cis-retinol present in the unknown or calibration material may bias results, separation of these isomers or appropriate calculation adjustments must be made to remove this bias.
In Exercise D, β-carotene was determined in SRM 3276 Carrot Extract in Oil as the unknown and SRM 3251 Serenoa repens (Saw Palmetto) Extract as the control (38, 52, 53). The level of β-carotene in the control was approximately three times the level in the unknown. Both materials required minimal sample preparation as they could be diluted in organic solvent with no additional extraction required. Results from participants were significantly scattered, with values for total β-carotene ranging from 30 to 150% of the target value. Participants were also asked to report values for trans-β-carotene and cis-β-carotene; only half of the laboratories were able to report these values. The scatter in the total β-carotene results could be a result of improper calibration, including incorrect traceability as discussed above for retinol, or presence of β-carotene isomers for which concentrations were incorrectly accounted.
In Exercise E, β-carotene determination was revisited using SRM 3280 Multivitamin/Multielement Tablets as the unknown and SRM 3251 Serenoa repens (Saw Palmetto) Extract as the control (53). The amount of total β-carotene in the unknown was approximately 10 times the amount in the control material. The extraction of the β-carotene from the unknown was significantly more challenging than from the control material. Most laboratories reported moderate uncertainties for total β-carotene, and the consensus of the results fell within the target range (Figure 5). One group of laboratories reported correct values for the control material, but low values for the unknown, another group reported correct values for the unknown and high values for the control (Figure 5). This type of trend again points to a calibration issue. Fewer than half of the participating laboratories reported values for the isomers of β-carotene, again implying that many laboratories do not have the capability to measure each isomer specifically, which may contribute to the observed calibration issues. Another consideration is that the differences in matrixes between the unknown and control may have led some laboratories to use an improper extraction procedure for the unknown, in which β-carotene is encapsulated. In addition, isomerization or degradation of β-carotene may occur during many extraction procedures. Another concern was the uncertainty and potential errors associated with the spectrophotometric determination of the calibrant concentration for β-carotene. In future exercises, participants could be provided with a molar absorptivity value for β-carotene and be asked to use this value in addition to the in-house value when calculating final concentrations for comparative purposes. Participants should also be informed more completely as to the nature of the analyte within the matrix, and future control materials will be better matched with regard to extraction challenges.
Figure 5
Figure 5
Comparison of results for total β-carotene determination in SRM 3280 Multivitamin/ Multielement Tablets with SRM 3251 Serenoa repens Extract as the control. In the bottom graph, the red box encompasses the NIST certified values, and the blue box (more ...)
Water-Soluble Vitamins
Compared to measurement of fat-soluble vitamins, the determination of water-soluble vitamins has significantly fewer potential issues. Most of these vitamins are relatively stable and are fortified at high levels in foods and dietary supplements of interest to the DSQAP. In the pilot exercise, participants were asked to determine the level of folic acid in SRM 3280 Multivitamin/Multielement Tablets as the unknown and SRM 1849 Infant/Adult Nutritional formula as the control (46, 51, 5460). The level of folic acid was 200 times greater in the unknown than in the control. The consensus of the reported data included the target value, although results were scattered for the lower-level control. This scatter is likely the result of a calibration issue, in which the level of folic acid in the control was out of the linear range used for calibration of the method.
In Exercise B, participants analyzed SRM 3280 Multivitamin/Multielement Tablets for thiamine (vitamin B1) and riboflavin (vitamin B2) content without a control material. For both vitamins, the consensus of the data contained the NIST-certified range (57, 61). The spread in the data for both vitamins was greater than expected, with values ranging from 60 to 180% of the certified value for thiamine and from 70 to 170% of the certified value for riboflavin. Without a control material, the results from this exercise are difficult to interpret, but further investigation into the ability of the DSQAP participant community to accurately measure these compounds is necessary.
In Exercise C, the measurement of two different water-soluble vitamins was investigated. Participants determined the concentration of niacinamide (vitamin B3) and pyridoxine (vitamin B6) in a fortified milk powder as the unknown and in SRM 3244 Ephedra-Containing Protein Powder as the control (37, 38). The target values for niacinamide and pyridoxine in the fortified milk powder were determined at NIST as part of an international interlaboratory comparison exercise. The level of niacinamide in the control was approximately three times the level in the unknown, while the level of pyridoxine in the control was approximately two and a half times the level in the unknown. For niacinamide, the results for the unknown and control were more scattered than expected, and the scatter could not be correlated with the analytical method used. The scatter in the niacinamide unknown spanned a significant range, from less than 20% to more than 500% of the certified value in the unknown. The results for pyridoxine were also more scattered than expected, and as with niacinamide, the scatter was not correlated with the type of analytical method used. For both vitamins, the consensus of the reported data did encompass the target value; however, the large scatter was a concern. One concern was that the participating laboratories reported niacin (nicotinic acid) instead of niacinamide, or pyridoxine instead of pyridoxine hydrochloride. Another possibility is a calibration issue, which should be investigated further in future exercises with a more closely matched unknown and control.
In Exercise D, the measurement of niacinamide was revisited with fortified milk powder as the unknown and SRM 1849 Infant/Adult Nutritional Formula as the control (46, 51, 58, 6266). The target value for niacinamide in the fortified milk powder was determined at NIST as part of an international interlaboratory comparison exercise. The level of niacinamide in the control was very similar to the level of niacinamide in the fortified milk powder. To eliminate the possibility of calibration issues, a vial containing 500 mg U.S. Pharmacopeia niacinamide was provided to all participants for use in method calibration. Compared with the results from Exercise C, the results in Exercise D were significantly improved. The consensus of the data included the target value, and few laboratories reported significantly outlying data. The results from this exercise highlight the importance of proper selection and use of calibration materials. Reference materials must be appropriate (e.g., niacinamide, rather than nicotinic acid) and should be screened for purity for increased measurement comparability.
In Exercise E, niacinamide was measured again in a very different but well-matched unknown and control set. The control for this study was SRM 3233 Fortified Breakfast Cereal, and the unknown was a ground breakfast cereal containing a similar concentration of niacinamide. The target value for niacinamide in the ground breakfast cereal was determined at NIST in parallel to the certification of water-soluble vitamins in SRM 3233. The consensus of the reported data included the target value for both the unknown and the control, with few outliers. A comparison of the values that were reported for the unknown and control from each laboratory indicated a potential calibration issue, as discussed and solved in Exercise D. This again stressed the necessity of using well-characterized control materials in all measurements, not only when specifically asked to do so. Because participants performed so well on the niacinamide study of Exercise E, future exercises will consider more complex matrixes containing lower, nonfortified levels of niacinamide or perhaps unknown and control pairs that are less exactly matrix-matched. In addition, future exercises will be planned to expand to other water-soluble vitamins, building on the knowledge gained through this series of niacinamide studies.
Botanicals
The analysis of botanical dietary supplements is challenging, as these materials are complex natural products. In addition to the complexity of the sample matrix, reference standards are often unavailable for many compounds of interest. When reference standards are available, purity is often insufficient for true quantitative analysis, as the standards are often isolated from the plant materials and include isomers and other related compounds. A further complication lies in the lack of well-developed methods or a community consensus regarding the best analytical approach for new, uncharacterized materials. The DSQAP has focused several exercises on addressing some of these analytical challenges.
In the pilot exercise, participants measured caffeine in SRM 3260 Bitter Orange-Containing Solid Oral Dosage Form as the unknown and SRM 3243 Ephedra-Containing Solid Oral Dosage Form as the control (37, 67, 68). The consensus of the reported data included the target value for both the unknown and the control, with few outliers. Because participants performed so well on the measurement of caffeine, future exercises involved determination of more complex botanical analytes.
Exercise B included the determination of three botanical analyte classes: alkaloids, flavonols, and phytosterols. Participants measured synephrine in SRM 3258 Bitter Orange Fruit as the unknown and in SRM 3259 Bitter Orange Extract as the control (68, 69). The level of synephrine was approximately 10 times greater in the control than in the unknown. Overall, the consensus of the data for synephrine contained the target value for both the control and the unknown, with few outliers. However, the results for synephrine in the unknown and control were slightly low for all participants, indicating a potential extraction or calibration issue. Extraction issues are common for botanical matrixes when the compound of interest must be extracted from the natural matrix within plant tissues. Participants in Exercise B also measured the flavonols quercetin, kaempferol, and isorhamnetin in SRM 3248 Ginkgo-Containing Tablets as the unknown and SRM 3247 Ginkgo biloba Extract as the control (38, 7072). The level of each flavonol was approximately 5 to 10 times greater in the control than in the unknown. For all three compounds, the consensus of the data contained the target value for both the unknown and the control, with few outliers. As with synephrine, a potential calibration issue was apparent. Lastly, participants determined the levels of phytosterols campesterol, β-sitosterol, and stigmasterol in SRM 3250 Serenoa repens (Saw Palmetto) Fruit as the unknown and SRM 3251 Serenoa repens (Saw Palmetto) Extract as the control (53). The level of each flavonol was approximately 5 to 10 times greater in the control than in the unknown. For each of the phytosterols, only five laboratories reported data. The data reported, however, were very scattered, and the consensus of the data contained the target value for only one of the compounds (stigmasterol). Because the data were scattered and a limited number of laboratories reported data, further studies on phytosterols were planned for future exercises.
The phytosterols study from Exercise B was repeated in Exercise C, with the inclusion of a phytosterol solution. Participants measured campesterol, β-sitosterol, and stigmasterol in SRM 3250 Serenoa repens (Saw Palmetto) Fruit as the unknown and SRM 3251 Serenoa repens (Saw Palmetto) Extract as the control (53). At NIST, a solution of mixed phytosterols was gravimetrically prepared in chloroform at a concentration that would be appropriate for a calibration, requiring only derivatization prior to determination. The participants accurately measured the concentration of each phytosterol in the solution, although the measurements were less precise than expected. Participant questionnaires indicated that most laboratories utilized in-house methodology involving a hydrolysis step that was unnecessary for a simple solution. The additional sample processing associated with the hydrolysis step may lead to the increased uncertainty. Some participants also reported an inability to weigh the solution reproducibly, likely a result of solvent evaporation, which may have also contributed to the increased uncertainty. Most laboratories accurately determined the levels of the phytosterols in the control material, SRM 3251 Serenoa repens (Saw Palmetto) Extract, which required hydrolysis and derivatization prior to determination. Comparatively, most laboratories reported values significantly lower than the target values for the phytosterols in the unknown, SRM 3250 Serenoa repens (Saw Palmetto) Fruit, which required extraction, hydrolysis, and derivatization prior to determination. Taken together, these results indicated that the extraction of the phytosterols from the fruit was incomplete, and additional successive extraction steps may be necessary for quantitative analysis. In addition, most laboratories reported the addition of an internal standard immediately prior to the derivatization step, whereas ideally the internal standard would be added prior to the extraction step in order to reduce uncertainty and improve accuracy.
In Exercise D, participants measured organic acids (citric acid, malic acid, and quinic acid) in SRM 3291 Vaccinium myrtillus (Bilberry) Extract as the unknown and SRM 3281 Vaccinium macrocarpon (Cranberry) Extract as the control (73). The level of each acid was within a factor of two between the unknown and the control material. For quinic and malic acids, the consensus of the reported data included the target value for the unknown, while the reported data for citric acid were more scattered. The consensus ranges for each acid were very small, indicating excellent overall reproducibility. Surprisingly, the uncertainty for the determination of organic acids was greater in the control material than in the unknown. Several laboratories reported chromatographic interferences, particularly for quinic acid, in the control material that did not affect analysis of the unknown. In addition, several laboratories reported low values for both the unknown and the control material that may be related to the use of SPE for sample cleanup. A recommendation was made to participants to investigate recovery by using a calibration material prior to using a filter or SPE cartridge for quantitative analysis.
In Exercise E, participants measured seven catechins (catechin, epicatechin, epicatechin gallate, epigallocatechin, epigallocatechin gallate, gallocatechin, and gallocatechin gallate) in SRM 3256 Green Tea-Containing Tablets as the unknown and SRM 3255 Camellia sinensis Extract as the control. While participants were asked to report values for all of the catechins, not all chromatographic methods can separate all of the catechins, nor were calibration standards readily available for each of the analytes. Overall, the results for all catechins in both the unknown and the control material were extremely scattered. Relative to the values reported for the unknown, the values for the control material appeared to be less scattered, indicating a possible bias in the extraction step. One bias might be from epimerization occurring during the extraction process. When the reported values for each laboratory were summed to determine a value for total catechins, the results were much less scattered, and the consensus range of values was aligned with the target range, supporting the explanation of epimerization. In addition, laboratories and regulators might be more interested in a value of total catechins for labeling and QC than individual values for each catechin.
Fatty Acids
For the determination of fatty acids in foods, several sample preparation steps are required. A common approach involves transesterification of the existing triglycerides for measurement as fatty acid methyl esters. Although this procedure is well established and routine in many food and dietary supplement laboratories, the labor-intensive sample preparation can lead to issues with calibration and increased uncertainty. In the pilot exercise, participants measured fatty acids in SRM 3278 Tocopherols in Edible Oils as the unknown and SRM 3276 Carrot Extract in Oil as the control (38, 52). Only four laboratories participated in this study, making data interpretation difficult. Three of the four laboratories reported data that correlated well with the target values for each fatty acid as well as for total fatty acids. One laboratory reported values that were significantly lower than the target and consensus for each fatty acid and the total fatty acids. The results from this study were promising, and the determination of fatty acids was planned for a future study with the hope that more laboratories would participate.
In Exercise C, participants measured fatty acids in SRM 3274-1 Botanical Oils Containing Omega-3 and Omega-6 Fatty Acids (Borage, Borago officinalis) as the unknown and SRM 3274-2 Botanical Oils Containing Omega-3 and Omega-6 Fatty Acids (Evening Primrose, Oenothera biennis) as the control. For linoleic acid and γ-linolenic acid, the consensus of the reported data included the target value for the unknown, while the reported data for α-linolenic acid were significantly lower than the target value for the unknown. Because the consensus values for two of the three fatty acids of interest were acceptable while the value for the third was not, it is possible that the α-linolenic acid in these oils may be degrading over time. A comparison of the data from each laboratory for linoleic acid and γ-linolenic acid indicated that laboratories reporting low values for each fatty acid with respect to the target value in the control material also reported low values with respect to the target value in the unknown. The same was true for laboratories reporting high values with respect to the target values. This type of correlation indicated a potential issue in the derivatization process or the calibration procedure. As investigated with the phytosterols, inclusion of a calibration solution for fatty acid studies in future exercises may aid in troubleshooting problems of this nature.
Through five exercises of the DSQAP, many lessons have been learned by both the participants and the program coordinators. Overall, laboratories have performed as expected on most of the studies. When analytical issues were identified, they were attributed to common technical challenges, such as incomplete extraction or improper calibration. The DSQAP has identified these problems and encouraged participants to use these lessons to improve their measurement capabilities. Identification of these issues raises participants’ overall awareness and expands the benefits of the DSQAP to analytes not included in particular studies.
In several cases, follow-up studies have clarified an analytical issue and provided an insight into the real challenges facing analytical laboratories in the dietary supplement community. First, incomplete extraction and addition of unnecessary analysis steps complicated the determination of phytosterols in botanical oils. Troubleshooting of this problem involved eliminating potential matrix issues and providing participants with a calibration solution for comparative analysis. Second, participants demonstrated poor overall measurements for determination of niacinamide in infant formula. The study was repeated with inclusion of a common reference standard for calibration, and results improved significantly, demonstrating the need for complete characterization of all reference standards. Additionally, the determination of Pb in botanicals demonstrated the importance of moisture consideration in a reported value. Moisture present in the material being analyzed as well as in reference standards can affect determination of both organic and inorganic analytes and should always be considered.
Other common analytical problems have been identified through all of the exercises and studies. Some laboratories have incorrectly used the target value provided for the control material to determine a scaling factor for analysis of an unknown material. The matrixes of the control and unknown are often different, making a mathematical adjustment for recovery invalid. The target range for the control is provided only as a quality self-check. For organic analytes, recovery issues as a result of loss upon filtration or sample cleanup have been identified. Participants are encouraged to test all such filters and cartridges for recovery using a calibration solution to identify potential sources of loss and subsequent measurement bias.
An additional goal of the DSQAP is to identify upcoming problems and challenges in the dietary supplement analytical community. The highly innovative nature of the food and dietary supplement marketplace leaves the analytical methodology several steps behind in emerging markets. When new products and matrixes are introduced to the market, the DSQAP plans to organize exploratory exercises to help laboratories demonstrate performance within the community in the absence of a target value. These DSQAP studies can also identify analytes for which high-quality reference standards are needed, identify matrixes for which reference materials are needed, and assist groups such as AOAC INTERNATIONAL in determining for which analytes and matrixes official methods of analysis are needed.
As the DSQAP evolves, participants are polled regularly to identify types of samples and analytes that are of emerging interest. The variety of participants, from government laboratories to third-party analytical laboratories to dietary supplement manufacturers, leads to differing priorities and a continuing need for participant involvement and feedback. Additional guidance from regulatory agencies and trade associations assists the DSQAP coordinators in maintaining a relevant and useful program.
Acknowledgments
We acknowledge the extensive work of the many contributors on the measurements used in certification of the SRMs used in the DSQAP, as well as the continued efforts of the participants in the program. In addition, we authors acknowledge the continued support of many contributors to the program, including other U.S. government agencies, various trade organizations, and independent contractors and corporations.
(1) Dennis J. Nutraceuticals World. 2010;13:40–46.
(2) Hlasney E. Supplement Usage, Consumer Confidence Remains Steady. Council for Responsible Nutrition; Washington, DC: 2010.
(3) Rea P. Global Supplement and Nutrition Industry Report 2010. Nutritional Business Journal; Boulder, CO: 2010.
(4) Betz JM. In: Herbs: Challenges in Chemistry and Biology. Wang M, Sang S, Hwang LS, Ho C-T, editors. American Chemical Society; Washington, DC: 2006. pp. 2–13. (ACS Symposium Series 925).
(5) Betz J, Fisher K, Saldanha L, Coates P. Anal. Bioanal. Chem. 2007;389:19–25. [PubMed]
(6) Sharpless KE, Duewer DL. Anal. Chem. 1995;67:4416–4422. [PubMed]
(7) Duewer DL, Thomas J.M. Brown, Kline MC, MacCrehan WA, Schaffer R, Sharpless KE, May WE, Crowell JA. Anal. Chem. 1997;69:1406–1413. [PubMed]
(8) Duewer DL, Kline M, Sharpless KE, Thomas J.M. Brown, Gary KT. Anal. Chem. 1999;71:1870–1878. [PubMed]
(9) Duewer DL, Kline M, Sharpless KE, Thomas J.M. Brown, Stacewicz-Sapuntzakis M, Sowell AL. Anal. Chem. 2000;72:3611–3619. [PubMed]
(10) Duewer DL, Kline MC, Sharpless KE, Thomas JB. Anal. Chem. 2000;72:4163–4170. [PubMed]
(11) Parris RM, Schantz MM, Wise SA. Polycycl. Aromat. Comp. 1996;9:249–256.
(12) [accessed on October 18, 2010];Micronutrients Measurement Quality Assurance Program. http://www.nist.gov/mml/analytical/organic/mmqap.cfm.
(13) [accessed October 18, 2010];Vitamin D Metabolites Quality Assurance Program. http://www.nist.gov/mml/analytical/vitdqap.cfm.
(14) [accessed October 18, 2010];Dietary Supplement Laboratory Quality Assurance Program. http://www.nist.gov/mml/analytical/organic/dsqap.cfm.
(15) Rybak ME, Salin ED. Appl. Spectrosc. 2001;55:816–821.
(16) Sutherland RA, Tack FMG. Adv. Environ. Res. 2003;8:37–50.
(17) Rosende M, Savonina EY, Fedotov PS, Miro M, Cerda V, Wennrich R. Talanta. 2009;79:1081–1088. [PubMed]
(18) Hall GEM, Gauthier G, Pelchat JC, Pelchat P, Vaive JE. J. Anal. Atom. Spectrom. 1996;11:787–796.
(19) Melucci D, Torsi G, Locatelli C. Ann. Chim.-Rome. 2007;97:141–151. [PubMed]
(20) Avula B, Wang YH, Smillie TJ, Duzgoren-Aydin NS, Khan IA. J. Agric. Food Chem. 2010;58:8887–8894. [PubMed]
(21) Arslan Z, Tyson JF. Talanta. 1999;50:929–937. [PubMed]
(22) Tyson JF. J. Anal. Atom. Spectrom. 1999;14:169–178.
(23) Smichowski P, Polla G, Gomez D, Espinosa AJF, Lopez AC. Fuel. 2008;87:1249–1258.
(24) Locatelli C. J. Sci. Food Agric. 2007;87:305–312.
(25) Tuzen M, Soylak M, Elci L, Dogan M. Anal. Lett. 2004;37:1185–1201.
(26) Soylak M, Tuzen M, Narin I. Quim. Nova. 2006;29:203–207.
(27) Kubova J, Stresko V, Bujdos M, Matus P, Medved’ J. Anal. Bioanal. Chem. 2004;379:108–114. [PubMed]
(28) Li XD, Coles BJ, Ramsey MH, Thornton I. Analyst. 1995;120:1415–1419.
(29) Moon JH, Kim SH, Chung YS, Lee YN, Kim HJ, Kim YE. Anal. Chim. Acta. 2008;619:137–142. [PubMed]
(30) Arslan Z, Tyson JF. Microchim. Acta. 2008;160:219–225.
(31) Kartal S, Aydin Z, Tokalioglu S. J. Hazard. Mater. 2006;132:80–89. [PubMed]
(32) Lee YL, Chang CC, Jiang SJ. Spectrochim. Acta B. 2003;58:523–530.
(33) Sutherland RA, Tack FMG. Anal. Chim. Acta. 2002;454:249–257.
(34) Lu HH, Jiang SJ. Anal. Chim. Acta. 2001;429:247–255.
(35) Chen M, Ma LQ. J. Environ. Qual. 1998;27:1294–1300.
(36) Mackey EA, Paul RL, Lindstrom RM, Anderson DL, Greenberg RR. J. Radioanal. Nucl. Chem. 2005;265:273–281.
(37) Sharpless KE, Anderson DL, Betz JM, Butler TA, Capar SG, Cheng J, Fraser C, Gay ML, Gardner G, Howell DW, Ihara T, Lam JW, Long SE, McCooeye M, Mackey EA, Mindak WR, Mitvalsky S, Murphy KE, Phinney KW, NguyenPho A, Porter BJ, Roman MC, Sander LC, Satterfield MB, Scriver C, Sturgeon RE, Thomas JB, Vocke RD, Jr, Wise SA, Wood LJ, Yang L, Yen JH, Ziobro GC. J. AOAC Int. 2006;89:1483–1495. [PubMed]
(38) Sharpless KE, Thomas J.M. Brown, Christopher SJ, Greenburg RR, Sander LC, Schantz M, Welch MJ, Wise SA. Anal. Bioanal. Chem. 2007;389:171–178. [PubMed]
(39) Castro J, Spraul JC, Marcus RK. Anal. Methods. 2009;1:188–194.
(40) Youden WJ. Ind. Qual. Control. 1959;15:24–28.
(41) Sander LC, Putzbach K, Nelson BC, Rimmer CA, Bedner M, Thomas JB, Porter B, Wood LJ, Schantz MM, Murphy KE, Sharpless KE, Wise SA, Yen JH, Siitonnen PH, NguyenPho A, Roman MC, Betz JM. Anal. Bioanal. Chem. 2008;391:2023–2034. [PubMed]
(42) Sharpless KE, Phinney CS, Wood LJ, Yen JH, Howell DW. J. Agric. Food Chem. 2003;51:6745–6751. [PubMed]
(43) Sharpless KE, Greenberg RR, Schantz MM, Welch MJ, Wise SA, Ihnat M. Anal. Bioanal. Chem. 2004;378:1161–1167. [PubMed]
(44) Bedner M, Schantz M, Sander LC, Sharpless KS. J. Chromatogr. A. 2008;119:74–80. [PubMed]
(45) Heudi O, Trisconi MJ, Blake CJ. J. Chromatogr. A. 2004;1022:115–123. [PubMed]
(46) Sharpless KE, Lindstrom RM, Nelson BC, Phinney KW, Rimmer CA, Sander LC, Schantz MM, Spatz R, Thomas J.M. Brown, Turk GC, Wise SA, Wood LJ, Yen JH. J. AOAC Int. 2010;93:1262–1274. [PubMed]
(47) Huang T, Zhang W, Liu J, Tian Y, Yang GX. J. Liq. Chromatogr. Relat. Technol. 2009;32:712–722.
(48) Sharpless KE, Margolis SA, Thomas J.M. Brown. J. Chromatogr. A. 2000;881:171–181. [PubMed]
(49) Rodrigo N, Alegria A, Barbera R, Farre R. J. Chromatogr. A. 2002;947:97–102. [PubMed]
(50) Chase GW, Eitenmiller RR, Long AR. J. Liq. Chromatogr. Relat. Technol. 1997;20:3317–3327.
(51) Sharpless KE, Schiller SB, Margolis SA, Thomas JB, Iyengar GV, Colbert JC, Gills TE, Wise SA. J. AOAC Int. 1997;80:611–621. [PubMed]
(52) Sharpless KE, Thomas J.M. Brown, Duewer DL, Putzbach K, Rimmer C, Sander LC, Schantz M, Wise SA, Yarita T, Yen JH. Anal. Bioanal. Chem. 2007;389:207–217. [PubMed]
(53) Schantz MM, Bedner M, Long SE, Molloy JL, Murphy KE, Porter BJ, Putzbach K, Rimmer CA, Sander LC, Sharpless KE, Thomas JB, Wise SA, Wood LJ, Yen JH, Yarita T, NguyenPho A, Sorenson WR, Betz JM. Anal. Bioanal. Chem. 2008;392:427–438. [PubMed]
(54) Nelson BC, Sharpless KE, Sander LC. J. Chromatogr. A. 2006;1135:203–211. [PubMed]
(55) Chen L, Eitenmiller RR. J. Food Sci. 2007;72:C243–C247. [PubMed]
(56) Baffi C, Bettinelli M, Beone GM, Spezia S. Chemosphere. 2002;48:299–306. [PubMed]
(57) Chen P, Atkinson R, Wolf WR. J. AOAC Int. 2009;92:680–687. [PubMed]
(58) Goldschmidt RJ, Wolf WR. Anal. Bioanal. Chem. 2010;397:471–481. [PubMed]
(59) Chun J, Martin JA, Chen LW, Lee J, Ye L, Eitenmiller RR. J. Food Compos. Anal. 2006;19:182–187.
(60) Pawlosky RJ, Flanagan VP. J. Agric. Food Chem. 2001;49:1282–1286. [PubMed]
(61) Chen P, Ozcan M, Wolf WR. Anal. Bioanal. Chem. 2007;389:343–347. [PubMed]
(62) LaCroix DE, Wolf WR, Vanderslice JT. J. AOAC Int. 1999;82:128–133. [PubMed]
(63) LaCroix DE, Wolf WR, Hindsley TM. Anal. Lett. 2002;35:2173–2184.
(64) LaCroix DE, Wolf WR, Porter E, Cantellops D, Chase GW, Woollard D. J. AOAC Int. 2001;84:789–804. [PubMed]
(65) LaCroix DE, Wolf WR. J. AOAC Int. 2002;85:654–664. [PubMed]
(66) Wolf WR, LaCroix DE. Fresenius J. Anal. Chem. 1998;360:459–464.
(67) Thomas JB, Sharpless KE, Mitvalsky S, Roman M, Yen J, Satterfield M. J. AOAC Int. 2007;90:934–940. [PubMed]
(68) Putzbach K, Rimmer CA, Sharpless KE, Sander LC. J. Chromatogr. A. 2007;1156:304–311. [PubMed]
(69) Nelson BC, Putzbach K, Sharpless KE, Sander LC. J. Agric. Food Chem. 2007;55:9769–9775. [PubMed]
(70) Mustafa O, Brendan M, Pei C. J. Food Drug Anal. 2007;15:55–62.
(71) Gray D, LeVanseler K, Pan MD. J. AOAC Int. 2005;88:692–702. [PMC free article] [PubMed]
(72) Gray D, LeVanseler K, Pan M, Waysek EH. J. AOAC Int. 2007;90:43–53. [PMC free article] [PubMed]
(73) Phillips MM, Case RJ, Rimmer CA, Sander LC, Sharpless KE, Wise SA, Yen JH. Anal. Bioanal. Chem. 2010;398:425–434. [PubMed]