Official websites use .gov
A .gov website belongs to an official government organization in the United States.
Secure .gov websites use HTTPS
A lock (
) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.
and USDA
The second national study was initiated to track how label and ingredient levels for Adult MVMs changed over time. The second study (products purchased in 2011) was of adult MVMs; dietary supplements containing three or more vitamins with or without minerals or other bioactive components) estimated the associations between label values and analytical values for 21 vitamins and minerals.
A national sampling plan for adult MVMs was developed to identify and purchase dietary supplement product samples that represent the US market. A national sampling of adult MVMs was conducted for two purposes:
Representative adult MVM products were identified using weighted frequency data from the NHANES 2007-08, a 2010 DS use survey by an independent marketing firm, and 2010 market share information from the Nutrition Business Journal (3). To attain geographically diverse sample sources across different market channels and to combine precise estimates of mean content with reliable assessments of product and lot variability, a statistical plan was developed for DS purchase. In 2011, 124 products were purchased. Retail products were purchased in 6 representative U.S. counties in 6 states, with 64 (three lots each) from the mass market channel (e.g., Safeway, Target, Sam’s Club), and 30 (three lots each) from the natural health channel (e.g., Whole Foods, GNC, organic markets). Thirty (30; two lots each) were purchased from direct channels (products sold exclusively on-line or from multi-level marketers like Amway).
The purchased products were sent to MAFCL for processing. Relevant information on each product purchased (e.g., ingredient names and levels, lot number, purchase location and date, and expiration date) was recorded in MAFCL’s in-house database. Samples were repackaged and sent for laboratory analysis in defined batches. Each product sample sent to labs contained at least 30 units (tablets, capsules or liquid serving amounts) of the MVM product. Labs were instructed to homogenize at least 30 sample units before sub-sampling for analysis (per the United States Pharmacopeia (USP) recommendations for the analysis of dietary supplements).
Qualified analytical contract laboratories analyzed the sample sets using validated sample-handling protocols and appropriate methods to obtain analytical information about ingredient levels (Table 1). The major components of vitamin A (retinol and betacarotene) were measured separately, converted to international units (IU) and combined to calculate total vitamin A for comparison to label levels.
| Nutrients | Analytical Method Used |
|---|---|
Calcium Copper Iron Magnesium Manganese Phosphorus Potassium Zinc | Multi-element inductively coupled plasma spectrometry (ICP) with wet ashing methodology |
Selenium | Hydride generation/atomic absorption spectroscopy |
Iodine | ICP- mass spectroscopy (MS) with wet ashing |
Chromium | ICP- mass spectroscopy (MS) with wet or dry ashing |
Beta-carotene Retinol | High-performance liquid chromatography (HPLC) with ultraviolet detection at 542 nm |
Niacin Riboflavin Thiamin Vitamin B6 | HPLC with ultraviolet detection at 210 nm |
Vitamin C | HPLC with ultraviolet detection at 254 nm |
Folic acid | Microbiological method using the bacteria, Enterococcus hirae |
Vitamin B-12 | Microbiological method using the bacteria, Lactobacillus delbrueckii |
Vitamin E | HPLC with fluorescence detection |
Vitamin D | HPLC-MS-MS |
Quality control (QC) materials were analyzed with samples of adult MVM products to evaluate laboratory precision and accuracy. NIST Standard Reference Material (SRM) 3280, an MVM matrix with certified values for vitamins and minerals, was sent in each batch to monitor laboratory measurement accuracy. In addition, each batch included a set of product duplicates (two sets of 30 tablets/capsules of the same MVM product with different test sample identification numbers) that were analyzed for all ingredients in the study, and at least two in-house control materials. For each in-house control material, a case of a single lot of an adult MVM product was purchased and samples were sent with each batch to evaluate the precision of laboratory methods over time in a matrix similar to the study products.
Analytical retests for ingredients in specific products were conducted to check unusually high or low results, high variability among product lots, and questionable data in batches where QC results showed a bias. For each sample analyzed, laboratory results reported in mg/g or µg/g were compared to label levels and a percent difference from the label levels was calculated.
Ingredient data from laboratory analyses were prepared for weighted regression analysis by applying market share estimates as product weights. Market share estimates were based on data from NHANES and from an independent marketing firm, as previously discussed in Section 3. To identify overly influential supplement observations, a jackknife technique was used to calculate Cook’s distances and restricted likelihood distances.
Relationships between the label and percent difference from label were estimated by regression with SAS® mixed model procedures. For each supplement ingredient, the label value was the independent variable and the percent difference from the label level (based on the laboratory analysis) was the dependent variable. Percent differences from label were calculated: ((analytical value − label value)/label value) × 100%. Three models (mean, linear and quadratic) were used to fit the data for all ingredients, and the most complex and statistically significant model was selected. Lab, supplement within label level and lot within supplement were modeled as random sources of variation. The accuracy of the models’ predictions was assessed with validation techniques.
The selected regression equations were used to predict mean analytical levels for each ingredient in adult MVMs: label value × (1 + predicted percent difference/100). In the DSID-4 files, these mean predictions are shown in data tables as predicted percent differences from the label levels or as predicted mean values in international units (IU), mg, or µg per serving or per day.
In addition, the standard error of the mean (SEM), 95% confidence intervals (CI) for the mean, and the standard error (SE) of an individual observation were calculated at each label level. Because the regression equation could be used to predict ingredient values of independent supplement samples, SE were adjusted to reflect this expected greater prediction variability.
Detailed results for this study, including regression equation parameters and predicted values, are listed in the data files released in DSID-4. Regression results are reported for 21 vitamins and minerals: folic acid, niacin, riboflavin, thiamin, vitamin A, vitamin B12, vitamin B-6, vitamin C, vitamin D, vitamin E, calcium, chromium, copper, iodine, iron, magnesium, manganese, phosphorus, potassium, selenium, and zinc. Regression results for mean predicted percent differences from the label amount and the associated SE and CI varied by ingredient and, in some cases, by ingredient level.
The regression results and SEM for the most common labeled level for each ingredient in the adult MVM-2 study are summarized in Tables 2 and 3 below. Table 2 lists the predicted mean percent differences from labeled levels for vitamins, and Table 3 does the same for minerals. If a linear or quadratic regression model was selected, a range of label levels was predicted. If a means model was selected, the predicted mean percent difference was not dependent on the label level.
| Ingredient | Range of Predicted Mean Percent Differences from Label Levels | Most Common Label Level per Serving | Predicted Mean Percent Difference at Most Common Label Level | Predicted SEM at Most Common Label Level |
|---|---|---|---|---|
| Folic Acid | -15.4% to 24.0% | 400 µg | 23.6% | 1.9% |
| Niacin | 4.48% | 20 mg | 4.48% | 0.96% |
| Riboflavin | 1.95% to 18.7% | 1.7 mg | 17.4% | 2.4% |
| Thiamin | -3.52% | 1.5 mg | -3.52% | 1.1% |
| Vitamin A | -8.52% to 49.3% | 3500 IU | 25.6% | 2.6% |
| Vitamin B-12 | 21.8% | 6 µg | 21.8% | 2.2% |
| Vitamin B-6 | 9.08% | 2 mg | 9.08% | 1.6% |
| Vitamin C | 5.08% | 60 mg | 5.08% | 1.4% |
| Vitamin D | 19.8% to 45.5% | 400 IU | 40.5% | 1.4% |
| Vitamin E | 9.36% | 30 IU | 9.36% | 1.9% |
| Ingredient | Range of Predicted Percent Differences from Label Levels | Most Common Label Level per Serving | Predicted Percent Difference at Most Common Label Level | Predicted SEM at Most Common Label Level |
|---|---|---|---|---|
| Calcium | 8.09% | 200 mg | 8.09% | 2.0% |
| Chromium | 9.67% to 29.4% | 120 µg | 20.4% | 2.0% |
| Copper | -1.55% to 15.4% | 2 mg | 6.42% | 2.3% |
| Iodine | 20.2% | 150 µg | 20.2% | 2.8% |
| Iron | 0.84%* | 18 mg | 0.84%* | 1.6% |
| Magnesium | -0.23%* | 50 mg | -0.23%* | 1.4% |
| Manganese | 4.42% to 13.0% | 2 mg | 7.43% | 0.79% |
| Phosphorus | -3.90% to 15.2% | 20 mg | 15.2% | 2.3% |
| Potassium | -1.70% to 5.29% | 80 mg | 2.50% | 0.55% |
| Selenium | 10% to 26.4% | 55 µg | 23.9% | 1.7% |
| Zinc | 3.92% to 13.9% | 15 mg | 4.55% | 0.70% |
For two minerals, iron and magnesium, the mean predicted amounts were not significantly different from label claims. For all vitamins, the mean predictions were significantly different from label. Five ingredients (niacin, vitamin B-6, vitamin C, vitamin E, calcium) had predicted mean percent differences from label between 0% and 10% above label at the most common label level and across the entire regression range (mean models). Seven ingredients (chromium, iodine, selenium, folic acid, vitamin A, vitamin B-12 and vitamin D) had predicted mean percent differences from label > 20% above label level at the most common label level. Folic acid, vitamin A and vitamin D had the largest ranges of percent differences from label (up to 69% for retinol) followed by selenium, chromium, copper, and phosphorus (up to 19% for chromium). One ingredient, thiamin, had predicted means slightly below label level for the entire regression range.
In both adult MVM-2009 and adult MVM-2017 studies, overages in mean ingredient content, at the product level, were found for most vitamins (with thiamin a consistent exception) and minerals (all but magnesium in adult MVM-2017). An evaluation of changes in the percent differences from label, variability and in regression models are pending. The information obtained from this monitoring study will be used to plan the frequency and scope of updates to the DSID in order to provide up-to-date tools for nutrient intake assessment from MVM and other DS.
We now provide estimates for chromium and vitamins A and D in adult MVMs, which were found to have mean percent differences from label ranging from 20-40% above label. More detailed results for the adult MVM-2017 study are available on the “Data Files” page of the DSID website. DSID application tables and linking codes are also provided for the 2009-2014 NHANES DS files.
The regression equations for the adult MVM-2017 study released in DSID-4 (Table 1) predict the mean percent differences from label levels for 21 ingredients in dietary supplements consumed in the United States. The predicted amounts are linked to label levels for each ingredient (Table 2 and adult MVM-2017 calculator) and are not specific to any brand or supplement. These predictions (predicted mean values) are intended for research purposes and are not meant to provide analytical estimates for ingredients in individual products.
Measures of variability are reported with predicted means, as discussed previously. The SE for an individual observation is much larger than the SEM because it represents the error of prediction for an individual sample vs. the error of prediction of a mean value for many products.
Results predicted by regression for mean percent differences from label level and SE have been assigned linking codes that may be applied to NHANES DS data files or used for other studies of DS intake. The predicted analytical content from the DSID can be used to replace label ingredient information to more accurately assess ingredient intakes from dietary supplements in large population surveys.
Documentation about the DSID-4 data files and instructions for appropriate use of the files are described in the report, DSID-4 Data File Documentation, available on the “Data Files” page of the website. Please refer to that report for additional information on how best to interpret and use each data file.
An online, interactive, Adult MVM-2017 Calculator has been released with DSID-4. This calculator allows the user to enter ingredient information from MVM labels and generate the appropriate predicted mean values, SE and CI for those labeled levels.
The adult MVM-2017 calculator is based on the chemical analysis of products purchased in 2011, and so provides more current data than the calculator with results from the original adult MVM (adult MVM-2009) study (products purchased in 2006-07). The calculator that is based on the adult MVM-2009 study is also available on the DSID website and can be used for historical data or trends analysis.
DSID pilot studies are underway to evaluate ingredient quantities in prescription prenatal MVMs and green tea dietary supplements. In addition, a study evaluating the phytochemical content of botanical dietary supplements containing turmeric/curcumin is planned.
