Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Dec 1.
Published in final edited form as: J Food Compost Anal. 2017 Jul 12;64(Pt 1):18–26. doi: 10.1016/j.jfca.2017.07.024

Development of a food composition database to monitor changes in packaged foods and beverages

Jennifer M Poti 1,*, Emily Yoon 1, Bridget Hollingsworth 1, Jessica Ostrowski 1, Julie Wandell 1, Donna R Miles 1, Barry M Popkin 1
PMCID: PMC5721674  NIHMSID: NIHMS894596  PMID: 29230079

Abstract

In order to monitor nutritional changes in the US food supply and assess potential impact on individual dietary intake, an approach was developed to enhance existing standard food composition tables with time-varying product- and brand-specific information for barcoded packaged foods. A “Crosswalk” was formed between barcoded products and USDA foodcodes in a time-specific manner, such that sales-weighted average nutritional profiles were generated for each foodcode based on corresponding products (275,000 to 350,000 per 2-year cycle). This Crosswalk-enhanced food composition table was applied to dietary intake data from the National Health and Nutrition Examination Survey (cycles 2007–2008, 2009–2010, and 2011–2012). Total energy density of foods consumed by Americans from stores/vending was stable over time and differed by <5 kcal/100g using the Crosswalk-enhanced vs standard database. However, changes in the energy density of food groups were found utilizing the Crosswalk that were not detected using the standard database. Likewise, significant declines in energy intake from beverages among children (288±7.3 to 258±6.8 kcal/d) were found using the Crosswalk-enhanced database but were non-significant using the standard database. The Crosswalk approach can potentially augment national nutrition surveys by utilizing commercial food purchase and nutrient databases to capture changes in the nutrient content of packaged foods.

Keywords: Food composition, Food analysis, Product reformulation, Branded food database, Packaged food, Nutrition facts panels, Dietary intake, NHANES

1 Introduction

Packaged foods and beverages represent a major segment of the US food supply. Previous studies estimate that retail food stores, including grocery stores and supermarkets, provide approximately two-thirds of calorie intake for US children and adults (Poti and Popkin, 2011; Ng et al., 2014). For these store-bought products, data from a nationally representative sample of US households indicate that packaged foods and beverages comprise about 80% of store expenditures (Slining et al., 2013b; Stern et al., 2016) and that US households purchase about 1200–1300 calories of packaged products per capita each day from retail food stores (Poti et al., 2015). Despite the dominant role of these packaged foods and beverages in the purchases and diets of Americans, existing food composition databases are limited in their ability to capture accurate nutrient content information for these products due to the complex, dynamic nature of this important part of the US food supply (Leclercq et al., 2001; Pennington et al., 2007; Ng and Dunford, 2013).

The packaged food sector is highly complex, including a wide array of diverse products. Between 2000 and 2012, US households purchased over 1.2 million barcoded products, yet most food composition databases include aggregate nutrition information for only a limited number of items (Ng and Dunford, 2013; Poti et al., 2015). For example, nutrition researchers rely on What We Eat in America (WWEIA), the dietary intake component of the National Health and Nutrition Examination Survey (NHANES), to study dietary intake among Americans. Nutrient information for all food and beverages reported by NHANES participants comes from the United States Department of Agriculture (USDA) Food and Nutrient Database for Dietary Studies (FNDDS) for the corresponding survey cycle (USDA Agricultural Research Service, 2010, 2012, 2014). While FNDDS 2011–2012 includes nutrient values for 7,618 foods and beverages, over 300,000 packaged foods were purchased in this time span (Ng and Dunford, 2013; USDA Agricultural Research Service, 2014).

The packaged food and beverage segment of the food supply is also dynamic, characterized by continuous change and turnover as new products are introduced to and less-favored products are removed from the retail market (Pennington et al., 2007; Ng and Dunford, 2013). Existing products may be reformulated, particularly to reduce trans fat, added sugar, or sodium content (Mozaffarian et al., 2010; Jacobson et al., 2013; Otite et al., 2013; Briguglio et al., 2015). Indeed, several major food manufacturers and retailers have made pledges to improve the nutritional quality of their products (Ng and Popkin, 2014; Taillie et al., 2015, 2016). These constant changes require food composition databases to be continuously updated (Pennington et al., 2007; Ng and Dunford, 2013). However, the USDA lacks the resources needed to update nutrient values for all foods in each NHANES cycle, and only specific food categories are selected for review (National Research Council, 2005; USDA Agricultural Research Service, 2010). For example, pizza, ready-to-eat cereal, and cereal/snack bars were among the limited set of categories reviewed for FNDDS 4.1 (USDA Agricultural Research Service, 2010).

To address these research gaps, we aimed 1) to develop an approach for enhancing the existing FNDDS food composition table with time-varying product- and brand-specific nutrition information for the diverse array of packaged foods and beverages in the US food supply and 2) to illustrate advantages of using this novel approach for monitoring changes in the nutrient content of packaged foods over time and assessing corresponding changes in individual dietary intake of American children and adults. To the best of our knowledge, no previous studies have examined the impact of changes in the nutrient content of packaged foods and beverages available in the US food supply on what is actually consumed by Americans.

2 Materials and Methods

2.1 Data sources

Our approach builds a “Crosswalk” between packaged foods and beverages purchased by US households and foods reported as consumed by NHANES participants. Our food composition database relies on data from three primary sources: 1) purchases of packaged foods and beverages by a nationally representative sample of US households; 2) Nutrition Facts Panel (NFP) data for barcoded packaged products; and 3) FNDDS for agricultural commodities, single ingredient foods, and home-prepared recipes.

2.1.1 Commercially available food purchase data

A vast number of products are available in the US market, but not all are widely purchased; this uneven sales distribution (i.e., higher sales volume of market leaders and lower sales of less-popular products) must be taken into account in nutrition surveillance (Ni Mhurchu et al., 2011; Korosec and Pravst, 2014). To generate sales-weighted nutrient profiles for packaged foods, we used data from the 2007–2012 Nielsen Homescan consumer panel, a nationwide study of packaged food and beverage purchases by US households (The Nielsen Company). The Homescan study design has been described in detail previously (Ng and Popkin, 2012; Poti et al., 2015). Briefly, panel households record all purchases from retail food stores using a handheld Universal Product Code (UPC) barcode scanner, as well as the date of each shopping episode, number of units purchased, weight (g), and price paid for each product. Data include product attributes for each barcoded item, including characteristics such as flavor (e.g., vanilla or chocolate ice cream), formula (e.g., low-fat or regular cheese), type (e.g., instant or regular oatmeal), or salt content (e.g., regular or low-sodium). Approximately 60,000 households from 76 geographic markets throughout the US participated each year. These data were used to quantify household purchases of packaged foods and beverages with barcodes. Foods and beverages without a barcode (e.g., unpackaged fresh fruits and vegetables) and products sold by variable weight that are assigned a barcode by the store where they are packaged (e.g., fresh meats sold by weight, meats cut to order then wrapped in-store by a butcher, or bulk foods with UPC printed at the weigh station) cannot be linked to nutrient composition data and were not included in the Crosswalk database. Fresh fruits and vegetables with a barcode (e.g., bagged salad, baby carrots, and bags of apples) were linked to nutrient composition data.

2.1.2 Commercially available NFP food composition data

Our research team linked each barcoded product to a corresponding NFP, obtained from various sources including the Mintel Global New Products Database, as described in detail in earlier publications (Ng and Popkin, 2012; Slining et al., 2013b). As required by the US Food and Drug Administration (FDA), NFP data includes the product’s serving size and total energy, total fat, saturated fat, trans fat, total sugar, carbohydrate, protein, dietary fiber, sodium, cholesterol, vitamin A, vitamin C, calcium, and iron per serving (US FDA, 2013). The label data also provides each product’s ingredient list and all information included on the product’s package. All NFP records include the date of data collection. Our Crosswalk-enhanced 2007–2012 FNDDS database includes NFP data collected between 1996 and 2013.

2.1.3 Publically available FNDDS food composition data

FNDDS is the database of foods/beverages and their nutrient values that is used to process dietary recalls collected from participants in NHANES (USDA Agricultural Research Service, 2010, 2012, 2014). FNDDS provides nutrient values for approximately 7,200–7,600 foodcodes per 2-year NHANES cycle and is based on the National Nutrient Database for Standard Reference (SR) (USDA Agricultural Research Service). Our database uses FNDDS 4.1 (SR Release 22), FNDDS 5.0 (SR Release 24), and FNDDS 2011–2012 (SR Release 26) which correspond to NHANES 2007–2008, 2009–2010, and 2011–2012, respectively.

2.1.4 Publically available dietary intake data

To illustrate potential advantages of using NFP data to enhance FNDDS when studying dietary intake, we used data from three cycles of NHANES, collected in 2007–2008, 2009–2010, and 2011–2012 (USDA and CDC/National Center for Health Statistics, 2010, 2012, 2014). NHANES is a cross-sectional survey that uses a complex, stratified, multistage probability cluster sampling design to select a nationally representative sample of the civilian non-institutionalized US population. One 24-hour dietary recall was collected in-person by trained interviewers using the USDA’s 5-step Automated Multiple-Pass Method, and a second recall was collected by phone 3–10 days later. Our analysis used only the first recall, as recommended by NHANES analytic guidelines in order to generate group mean intake (USDA and CDC/National Center for Health Statistics).

For each food or beverage reported, the participant provided the location where the item was obtained. Because our database focuses on packaged foods and beverages, our study included only foods reported from stores or vending machines and excluded items reported from restaurants, fast food establishments, and all other away-from-home food sources. All reported items were aggregated into 59 mutually exclusive food and beverage groups based on nutritional content and eating behaviors, as described in Supplementary Table S1 and elsewhere (Popkin et al., 1999; Slining et al., 2013a).

Our study included children aged ≥2 y (n=8,974) and adults (n=16,267) with one 24-hour dietary recall deemed reliable by study administrators.

2.2 Linkage of barcoded products to FNDDS foodcodes

These sources of data were used to create “Crosswalk-enhanced” FNDDS food composition tables for 2007–2008, 2009–2010, and 2011–2012 in a multi-step process described in detail below. The goal was to generate a time-specific database including mean nutrient profiles for each FNDDS foodcode by survey cycle using sales-weighted NFP data from corresponding packaged foods and beverages.

2.2.1 Identification of FNDDS foodcodes for packaged foods from stores

First, all FNDDS foodcodes reported by NHANES participants during the first 24-hour recall and obtained from stores or vending machines were identified and were eligible for linking to NFP data; identification of foodcodes for linking was conducted separately for each 2-year NHANES cycle. Because the purpose of our database is to better understand packaged foods, foodcodes for home-cooked dishes were not linked to NFP data; these foodcodes were identified by food descriptions including terms such as “homemade,” “home-made style,” or “from home recipe.” Although NFP data are advantageous for capturing product- and brand-specific variation in nutrient content as well as product reformulations and turnover, FNDDS is the gold-standard for raw agricultural commodities and single-ingredient foods (USDA Agricultural Research Service, 2010, 2012, 2014). For these items, FNDDS foodcodes link directly to single foods in SR that have nutrient content primarily derived analytically by the USDA Nutrient Data Laboratory. Because these items are not industrial formulations that might be modified over time, our database retains nutrient values from FNDDS for these items, including fresh fruits and vegetables, meats, eggs, dried beans, oil, and sugar.

2.2.2 Linkage of barcoded products to FNDDS foodcodes

After FNDDS foodcodes for packaged products were identified, a team of registered dietitians (RDs) linked UPCs for packaged products to each FNDDS foodcode in a time-specific manner. RDs made all linkages manually after reviewing the UPC’s item description, brand name, attributes, ingredients, and marketing, as described previously (Slining et al., 2015). To standardize the linkages, the research team jointly determined the decision rules to apply when matching UPCs to each FNDDS foodcode. For each foodcode, however, a single RD performed the linking process after documenting the rationale for matching specific products to the foodcode. A separate database of UPC-foodcode links was created for each NHANES cycle. Multiple NFP records may exist for a given UPC if the product has been reformulated over time; the NFP record(s) dated closest to the time of reported consumption was selected for each cycle. If a given UPC was reformulated and multiple formulations were purchased within one NHANES cycle, both UPC-NFP versions were linked. Barcoded products were linked to one or more corresponding FNDDS foodcodes for each cycle, as appropriate; for example, a UPC for 2% fat chocolate milk would be linked to the FNDDS foodcode for “Milk, chocolate, reduced fat milk-based, 2%” as well as to the foodcode for “Milk, chocolate, NFS” (Not Further Specified).

2.2.3 Conversion of nutrient information as purchased to as consumed

FNDDS provides nutrient values for foods in the form in which they are consumed, which may differ from the form of the food when purchased for products such as dry/powdered drink mixes, baking mixes, dried hot cereal or grains, dry pudding or gelatin mixes, condensed soup, and mixed dish/meal kits or mixes. Methodology used to convert nutrition information from “as purchased” to “as consumed” form is described elsewhere and in Supplementary Appendix A (Slining et al., 2015).

2.3 Estimation of sales-weighted nutrient profiles for FNDDS food codes

For each 2-year NHANES cycle, nutrient values per 100g for all UPC-NFP records linked to a given FNDDS foodcode were weighted by sales (g) within that 2-year period in order to calculate the weighted average nutrient profile per 100g for that foodcode in that survey cycle.

2.4 Application of quality control measures

A series of quality control checks were used to monitor the accuracy of Crosswalk-generated profiles, as described in Supplementary Appendix B.

2.5 Creation of Crosswalk-enhanced FNDDS Versions 1 and 2

For many FNDDS foodcodes, the food description does not specify whether the item is home- prepared from scratch, home-prepared from packaged ingredients, or industrially pre-prepared. For example, the foodcode “Pancakes, plain” could represent pancakes prepared from a home-recipe, pancakes prepared from a boxed dry mix, or frozen ready-to-heat pancakes; however, the Crosswalk approach can only represent the latter two packaged options. To incorporate this uncertainty and to avoid the assumptions about home- vs industrial-food preparation, we created two alternate versions of our food composition database. For any foodcode that could be used to report a food prepared from a home-recipe or an industrially-prepared packaged food, the Crosswalk-enhanced FNDDS Version 1 used the FNDDS nutrient profile based on a “recipe” of SR food ingredients, whereas Version 2 used the Crosswalk-based nutrient values. However, if the FNDDS nutrient profile was based on a “recipe” of SR codes including industrial ingredients (e.g., “modified food starch”) or based on a single SR code for a commercially prepared item (e.g., the foodcode “Pie, pumpkin” uses SR nutrient values for “Pie, pumpkin, commercially prepared”), both Versions 1 and 2 used Crosswalk-based nutrient profiles because nutrient values for a home-prepared food were not available in FNDDS.

2.6 Application of Crosswalk-enhanced FNDDS to dietary intake

To evaluate the utility of our Crosswalk database for studying dietary intake, nutrient intake from each food reported from stores/vending by NHANES participants was calculated using three alternate food composition tables: FNDDS only, Crosswalk-enhanced FNDDS Version 1, and Crosswalk-enhanced FNDDS Version 2. First, nutrient intake from each foodcode was determined using nutrient values from FNDDS alone for all foodcodes, which is the standard approach for analyzing NHANES data. Next, nutrient intake from each foodcode from stores/vending was calculated using nutrient values from the Crosswalk-enhanced FNDDS Version 1 (which reflects a scenario with minimal levels of industrially-prepared packaged food consumption) and additionally using nutrient values from Crosswalk-enhanced FNDDS Version 2 (which reflects a scenario with maximal levels of industrially-prepared packaged food consumption and minimal consumption of home-prepared foods). For each approach, nutrient intake from each foodcode was calculated as grams reported consumed by the participant multiplied by the nutrient value per 100g for each foodcode; calculations were made separately for each of the three food composition tables.

Some foodcodes can be reported using modification codes to indicate adjustment to recipe ingredients, such as the type of milk used in food preparation or type of fat used in cooking, to more closely match the food/beverage reported. Methods used to account for reporting with modification codes are described in Supplementary Appendix C. Foods consumed together as part of a single dish may be reported as individual foodcodes linked together using a combination code variable. We used combination codes to sum nutrient values for all components into a single food item as described previously (Slining et al., 2013a).

2.7 Use of the Crosswalk-enhanced FNDDS for monitoring trends in energy density and dietary intake

To illustrate the use of the Crosswalk-enhanced FNDDS for monitoring changes in the US food supply, we used NHANES 2007–2012 data to examine trends in the energy density of food groups and trends in dietary intake reported from stores and vending machines calculated using the three approaches described above: FNDDS only, Crosswalk-enhanced FNDDS Version 1, and Crosswalk-enhanced FNDDS Version 2.

To determine whether the energy density of packaged foods changed between 2007–2008 and 2011–2012, we calculated mean energy density for all items reported by an individual, both overall and by food group. Survey-weighted unadjusted total mean energy density was calculated across all individuals separately for each survey cycle; mean energy density of food groups was calculated among food group consumers only. Trends over time were tested using post-estimation Wald tests to compare estimated energy density in 2009–2010 and 2011–2012 to that in 2007–2008 for FNDDS only, Crosswalk Version 1, and Crosswalk Version 2.

To determine whether trends in dietary intake differed when estimated using the standard FNDDS food composition table compared with the Crosswalk-enhanced FNDDS Versions 1 or 2, survey-weighted unadjusted mean total energy intake and per capita food group intakes were determined for each NHANES cycle using FNDDS only, Crosswalk Version 1, and Crosswalk Version 2. Post-estimation Wald tests were used to compare mean intake in 2009–2010 and 2011–2012 with intake in 2007–2008 for each of the three approaches. Intake trends were examined separately for children aged 2–18 y and for adults >18 y.

Database creation and calculation of Crosswalk-based weighted nutrient profiles were conducted using Excel and SAS version 9.4 (SAS Institute Inc., Cary, NC). All statistical analyses were conducted using survey commands in Stata 14 (StataCorp LP, College Station, TX) to incorporate survey weights and account for complex survey design. This secondary data analysis was deemed exempt from review by the University of North Carolina Institutional Review Board.

3 Results

The Crosswalk-enhanced FNDDS food composition table included NFP-based nutrient profiles for n=4,872 (Version 1) or n=6,032 (Version 2) foodcodes reported from stores or vending machines by NHANES participants in any survey cycle between 2007–2008 and 2011–2012. For Crosswalk-enhanced FNDDS Version 1, 73% of store/vending energy intake among children and 62% of intake among adults was derived from Crosswalk-based nutrient profiles, with the remaining intake derived from FNDDS. For Version 2, 81% of store/vending energy intake among children and 71% of intake among adults was derived from Crosswalk-based nutrient profiles.

For foods and beverages obtained from stores or vending machines during the first dietary recall among NHANES participants, FNDDS provides nutrient values for 3,509 foods/beverages in 2007–2008, 3,669 in 2009–2010, and 3,738 in 2011–2012 (Table 1). Nutrient values for these foodcodes in the Crosswalk-enhanced FNDDS Versions 1 and 2 were based on 276,056 – 291,884 UPCs in 2007–2008, 315,355 – 329,384 in 2009–2010, and 328,330 – 345,899 in 2011–2012. Similar patterns of >150-fold difference in the number of UPCs and corresponding FNDDS foodcodes were observed for food groups including cheese, yogurt, nuts and nut butters, salad dressing, dips and spreads, granola bars and other bars, dairy-based desserts, salty snacks, water, sugar sweetened beverages (SSBs), fruit juice, and milk. For example, approximately 13,000 – 14,000 UPCs for SSBs were represented by ~50 FNDDS foodcodes.

Table 1.

Counts of foodcodes and barcoded food products contributing to nutrient profiles in the Crosswalk-enhanced FNDDS food composition table for use with NHANES 2007–2008, 2009–2010, 2011–2012 consumption from stores/vendinga

# FNDDS codes # Barcodesb
Food group 2007–2008 2009–2010 2011–2012 2007–2008 2009–2010 2011–2012
All foods and beverages 3,509 3,669 3,738 276,056 – 291,884 315,355 – 329,384 328,330 – 345,899
Foods
 All food groups 3,182 3,336 3,406 234,282 – 249,863 270,944 – 284,646 279,982 – 297,161
 Cheese 67 70 66 26,763 32,751 24,102
 Yogurt 16 17 20 4,121 7,889 8,246
 Other dairy products 39 39 38 4,300 4,691 4,886
 Meat 123 133 121 149 – 454 184 – 550 189 – 403
 Meat dishes with starch/veg 66 74 71 62 – 647 0 – 603 0 – 591
 Poultry 119 125 114 407 – 563 445 – 640 510 – 722
 Poultry dishes with starch/veg 56 66 62 83 – 352 92 – 424 96 – 431
 Processed meat 114 114 100 8,186 – 8,312 10,969 – 11,122 12,189 – 12,347
 Processed meat dishes with starch/veg 20 26 16 56 – 79 57 – 95 92 – 129
 Fish & shellfish 86 90 116 512 – 942 372 – 857 564 – 1078
 Fish & shellfish dishes with starch/veg 30 36 35 0 – 183 0 – 180 0 – 182
 Processed fish & shellfish 12 17 12 1,069 1286 1166
 Meat substitutes 13 18 19 448 – 455 469 541
 Eggs & egg dishes 48 57 77 144 – 275 180 – 281 732 – 829
 Legumes & legume dishes 55 55 70 1,845 – 2,103 1,840 – 2,132 1,870 – 2,210
 Nuts, nut butters, & seeds 48 49 56 10,661 11,395 14,012
 Breads 112 122 116 15,624 – 16,574 21,014 – 21,244 22,258 – 23,473
 Quick breads 59 61 56 4,238 – 4,699 4,672 – 5,174 4,501 – 5,011
 Ready-to-eat cereals 144 148 151 13,161 12,907 17,431
 Fruit 116 129 121 9,780 10,865 10,315 – 10,316
 Vegetables 307 310 337 15,668 – 16,718 19,745 – 20,271 16,959 – 18,324
 Non-starchy vegetable dishes 13 14 15 626 – 677 617 – 710 777 – 834
 Starchy vegetables & dishes 146 130 148 10,643 – 12,006 9,570 – 10,620 9,456 – 10,576
 Fried potatoes 10 11 13 1,340 – 1,377 1,112 – 1,149 1,756 – 1,807
 Fats & oils 32 36 33 1,087 1,302 1,147
 Sweeteners, syrups, & toppings 48 47 52 4,209 4,029 4,289
 Salad dressings 33 34 28 5,460 – 5,469 6,010 – 6,039 6,418
 Sauces & condiments 56 65 66 4,801 – 5,279 5,573 – 6,020 6,147 – 6,995
 Dips & spreads 19 16 17 2,575 – 3,750 2,818 – 4,086 2,877 – 4,378
 Grain-based desserts 275 253 214 16,702 – 20,527 17,213 – 19,827 17,684 – 21,059
 Granola bars & other bars 33 37 36 6,038 9,746 10,256
 Candy & sweet snacks 117 121 130 18,062 – 18,139 18,037 – 18,125 19,361 – 19,425
 Dairy-based desserts 88 97 85 13,137 – 13,297 15,117 – 15,311 13,373 – 13,540
 Salty snacks 84 91 97 14,026 – 14,198 15,247 – 15,452 15,964 – 16,181
 Soups & stews 166 180 172 6,192 – 7,520 8,000 – 9,181 13,636 – 14,646
 Frozen meals 46 50 42 1,373 1,801 503
 Sandwiches 28 18 45 255 – 574 273 – 347 354 – 412
 Hamburgers 3 3 . 11 80 0
 Tortilla- & corn-based dishes 61 61 84 122 – 407 305 – 580 155 – 590
 Pasta & pasta dishes 81 99 103 3,044 – 3,994 3,689 – 5,366 4,091 – 5,794
 Rice, grains, & rice/grain dishes 144 157 192 5,512 – 5,930 6,629 – 7,096 8,944 – 9,746
 Pizza 32 45 32 1,679 1,771 1,909
 Baby food & formula 16 14 24 111 182 226
Beverages
All beverage groups 327 333 332 41,774 – 42,021 44,411 – 44,738 48,348 – 48,738
 Water 9 10 10 2,632 3,198 3190
 Coffee & tea 60 59 62 4,752 – 4,874 6,195 – 6,335 6,111 – 6,256
 Sugar sweetened beverages 48 50 51 13,078 13,199 14,329
 Fruit juice 44 47 40 8,512 – 8,589 7,924 – 8,051 9,076 – 9,253
 Vegetable juice 10 9 9 532 570 563
 Milk 59 59 61 9,242 10,621 10,643
 Soy-, yogurt-, & milk-based  beverages 16 18 16 1,063 – 1,080 1,195 – 1,226 1,379 – 1,412
 Meal replacement beverages 19 20 27 36 66 35
 Sports drinks 5 7 6 516 556 642
 Energy drinks 10 10 11 634 704 803
 Other beverages 16 14 14 777 – 808 183 – 212 1,577 – 1,612
Open in a new tab

Note: FNDDS, Food and Nutrient Database for Dietary Studies (FNDDS); NHANES, National Health and Nutrition Examination Survey; UPC, Universal Product Code

a

Crosswalk-enhanced FNDDS nutrient profiles were created for all FNDDS foodcodes reported by participants aged 2 years and older during the first day of 24-hour dietary recall and obtained from stores or vending machines in the 2007–2008, 2009–2010, and 2011–2012 NHANES.

b

Number of unique UPC barcodes used to generate the sales-weighted nutrient profile for all FNDDS food codes in the specified food group for the Crosswalk-enhanced FNDDS Version 1 and Version 2. Version 1 defaults to the FNDDS nutrient profile for FNDDS food codes that may be home-prepared/not packaged and for which FNDDS is based on nutritional composition for a “recipe” of basic foods. Version 2 uses the Crosswalk-based nutrient profile for all FNDDS food codes for which a packaged/barcoded version of that food is available. Unique UPCs may or may not represent unique product formulations.

Trends in energy density and dietary intake were similar when estimated using the Crosswalk-enhanced FNDDS versions 1 and 2; thus, we focus on version 2 and have provided version 1 results as Supplementary Material.

Mean energy density of all foods and beverages, foods only, or beverages only differed by <5 kcal/100g when estimated using these alternate food composition tables for intake from stores/vending (Table 2 and Supplementary Table S2). No significant trends in energy density between 2007–2008 and 2011–2012 were observed for total, food, or beverage intakes from stores/vending. However, differences were observed within food groups. Estimated mean energy density of yogurt was 8–14 kcal/100g lower using the Crosswalk-enhanced FNDDS vs standard FNDDS; for example in 2011–2012, the Crosswalk-enhanced energy density was 79 ± 1.5 kcal/100g compared to 92 ± 0.9 kcal/100g in FNDDS. Significant declines in the energy density of fried potatoes were found when estimated using the Crosswalk-enhanced FNDDS, but not using the standard FNDDS; moreover, energy density for this food group was lower in 2011–2012 when estimated using the Crosswalk-enhanced (192 ± 7.8) vs standard (214 ± 5.5) FNDDS.

Table 2.

Trends in mean energy density of foods consumed from stores/vending using FNDDS and Crosswalk-enhanced FNDDS food composition tables, NHANES 2007–2008 to 2011–2012a

Mean energy density (kcal/100g)b
FNDDSc Crosswalk-enhanced FNDDSd Difference, FNDDS and Crosswalk-enhanced FNDDS
Food group 2007-2008 2009-2010 2011-2012 Time trende 2007- 2008 2009- 2010 2011- 2012 Time trend 2007- 2008 2009- 2010 2011- 2012 Time trend
All foods and beverages 213 213 214 1 212 211 212 0 −1 −2 −2 −1
Foods
 All food groups 280 275 278 −2 277 273 276 −1 −3 −2 −2 1
 Yogurt 92 92 92 0 78 84* 79 1 −14 −8 −13 1
 Tortilla- & corn-based
 dishes 241 235 240 −1 235 235 232 −3 −6 0 −8 −2
 Fried potatoes 226 235 214 −12 221 208 192* −29* −5 −27 −22 −17
 Grain-based desserts 406 413* 422* 16* 418 419 421 3 12 6 −1 −13
 Granola bars & other bars 412 406 408 −4 385 392* 395* 10* −27 −14 −13 14
 Candy & sweet snacks 398 395 392 −6 394 389 386 −8 −4 −6 −6 −2
 Dairy-based desserts 202 205 207 5 214 211 211 −3 12 6 4 −8
 Salty snacks 492 484* 485 −7 487 484 486 −1 −5 0 1 6
Beverages
All beverage groups 38 39 40 2 39 39 40 1 1 0 0 −1
 Sugar sweetened beverages 30 30 31 1 31 31 31 0 1 1 0 −1
 Milk 53 52 54 1 56 53* 54 −2 3 1 0 −3
 Sports drinks 26 26 27 1 21 21 20 −1 −5 −5 −7 -2
Open in a new tab

Note: FNDDS, Food and Nutrient Database for Dietary Studies (FNDDS); NHANES, National Health and Nutrition Examination Survey

*

Indicates mean is different from 2007–2008 at P<0.05.

a

Crosswalk-enhanced FNDDS nutrient profiles were created for all FNDDS food codes reported by participants aged 2 years and older during the first day of 24-hour dietary recall and obtained from stores or vending machines in the 2007–2008, 2009–2010, and 2011–2012 NHANES. Crosswalk-enhanced nutrient profiles are derived from Nutrition Facts Panel information on packaged foods and beverages purchased from grocery stores, supermarkets, convenience stores, and all other retail food stores by US households participating in the Nielsen Homescan Consumer Panel.

b

Values are the survey-weighted unadjusted mean energy density (kcal/100g) among consumers of any foods included in the specified food group. Combination codes were not applied. All estimates are weighted to be nationally representative and take into account complex survey design. Selected example food groups shown; all food groups shown in Supplemental Table 2.

c

Energy density of foods obtained from stores and vending machines was calculated using FNDDS.

d

Energy density of foods obtained from stores and vending machines was calculated using the Crosswalk-enhanced FNDDS Version 2, which defaults to the Crosswalk-based profile for mixed dish recipe foods.

e

Values are the difference in energy density between 2007–2008 and 2011–2012.

The mean decrease in energy intake among children between 2007–2008 and 2011–2012 was slightly larger when estimated using the Crosswalk-enhanced FNDDS compared with using the standard FNDDS (-35 vs -20 kcal/d); however, the decrease was not statistically significant using either approach (Table 3 and Supplementary Table S3). A significant decline in total beverage intake among children was found using the Crosswalk-enhanced FNDDS but not using the standard FNDDS, primarily because of slightly larger estimated declines in SSB and milk intakes when estimated using the Crosswalk-enhanced vs standard FNDDS.

Table 3.

Trends in mean per capita energy intake from stores/vending among US children using FNDDS and Crosswalk-enhanced FNDDS food composition tables, NHANES 2007–2008 to 2011–2012a

Mean energy intake (kcal/d) from stores/vendingb
FNDDSc Crosswalk-enhanced FNDDSd Difference, FNDDS and Crosswalk-enhanced FNDDS
Food group 2007- 2008 2009–2010 2011–2012 Time trende 2007–2008 2009–2010 2011–2012 Time trend 2007–2008 2009–2010 2011–2012 Time trend
All foods and beverages 1260 1320* 1240 −20 1256 1309* 1221 −35 −4 −11 −19 −15
Foods
All food groups 980 1034* 978 −2 967 1022* 963 −4 −13 −12 −15 −2
 Yogurt 7 12* 12* 5* 6 11* 11* 5* −1 −1 −1 0
 Tortilla- & corn-based
 dishes 36 48 57* 21 35 47 57* 22* −1 −1 0 1
 Fried potatoes 6 7 6 0 6 7 6 0 0 0 0 0
 Grain-based desserts 92 95 113* 21* 95 96 111 16 3 1 −2 −5
 Granola bars & other bars 8 8 12 4 7 8 11 4 −1 0 −1 0
 Candy & sweet snacks 54 43* 35* −19* 53 42* 34* −19* −1 −1 −1 0
 Dairy-based desserts 34 30 27 −7 37 31 28 −9 3 1 1 −2
 Salty snacks 95 98 91 −4 93 98 90 −3 −2 0 −1 1
Beverages
All beverage groups 279 286 262 −17 288 287 258* −30* 9 1 −4 −13
 Sugar sweetened beverages 86 86 74 −12 90 88 73 −17 4 2 −1 −5
 Milk 120 124 117 −3 126 125 118 −8 6 1 1 −5
 Sports drinks 7 8 7 0 5 7 6 1 −2 −1 −1 1
Open in a new tab

Note: FNDDS, Food and Nutrient Database for Dietary Studies (FNDDS); NHANES, National Health and Nutrition Examination Survey

*

Indicates mean is different from 2007–2008 at P<0.05.

a

Data from children aged 2–18 years participating in the 2007–2008 (n=2930), 2009–2010 (n=3097), and 2011–2012 (n=2947) NHANES.

b

Values are survey-weighted unadjusted mean per capita energy intake (kcal/d) from the specified food group for foods reported during the first day of 24-hour dietary recall and obtained from stores or vending machines. All estimates are weighted to be nationally representative and take into account complex survey design. Selected example food groups shown; all food groups shown in Supplemental Table 3.

c

Energy intake from foods obtained from stores and vending machines was calculated using the reported amount consumed and kcal/100g from FNDDS.

d

Energy intake from foods obtained from stores and vending machines was calculated using the reported amount consumed and kcal/100g from the Crosswalk-enhanced FNDDS Version 2, which defaults to the Crosswalk-based profile for mixed dish recipe foods.

e

Values are the difference in energy intake between 2007–2008 and 2011–2012.

Among adults, mean energy intake overall and for food groups was similar when estimated using the Crosswalk-enhanced or standard FNDDS (Table 4 and Supplementary Table S4). Significant declines in SSB intake were identified using both food composition tables.

Table 4.

Trends in mean per capita energy intake from stores/vending among US adults using FNDDS and Crosswalk-enhanced FNDDS food composition tables, NHANES 2007–2008 to 2011–2012a

Mean energy intake (kcal/d) from stores/vendingb
FNDDSc Crosswalk-enhanced FNDDSd Difference, FNDDS and Crosswalk-enhanced FNDDS
Food group 2007–2008 2009–2010 2011–2012 Time trende 2007–2008 2009–2010 2011–2012 Time trend 2007–2008 2009–2010 2011–2012 Time trend
All foods and beverages 1444 1501 1458 14 1448 1498 1447 −1 4 −3 −11 −15
Foods
All food groups 1123 1168 1151 28 1116 1160 1141 25 −7 −8 −10 −3
 Yogurt 10 11 12 2 8 10 10 2 −2 −1 −2 0
 Tortilla- & corn-based
 dishes 48 40 60 12 46 40 59 13 −2 0 −1 1
 Fried potatoes 10 13 9 −1 9 10 7 −2 −1 −3 −2 −1
 Grain-based desserts 90 94 102 12 92 95 101 9 2 1 −1 −3
 Granola bars & other bars 12 10 10 −2 11 10 10 −1 −1 0 0 1
 Candy & sweet snacks 43 37 34 −9 42 36 33 −9 −1 −1 −1 0
 Dairy-based desserts 40 40 33 −7 42 41 34 −8 2 1 1 −1
 Salty snacks 79 71 77 −2 78 71 77 −1 −1 0 0 1
Beverages
All beverage groups 321 333 307 −14 332 338 305 −27 11 5 −2 −13
 Sugar sweetened beverages 90 82 70* −20* 96 86 72* −24* 6 4 2 −4
 Milk 63 65 59 −4 67 67 60 −7 4 2 1 −3
 Sports drinks 7 5 7 0 6 4 6 0 −1 −1 −1 0
Open in a new tab

Note: FNDDS, Food and Nutrient Database for Dietary Studies (FNDDS); NHANES, National Health and Nutrition Examination Survey

*

Indicates mean is different from 2007–2008 at P<0.05.

a

Data from adults aged >18 years participating in the 2007–2008 (n=5516), 2009–2010 (n=5873), and 2011–2012 (n=4878) NHANES.

b

Values are survey-weighted unadjusted mean per capita energy intake (kcal/d) from the specified food group for foods reported during the first day of 24-hour dietary recall and obtained from stores or vending machines. All estimates are weighted to be nationally representative and take into account complex survey design. Selected example food groups shown; all food groups shown in Supplemental Table 4.

c

Energy intake from foods obtained from stores and vending machines was calculated using the reported amount consumed and kcal/100g from FNDDS.

d

Energy intake from foods obtained from stores and vending machines was calculated using the reported amount consumed and kcal/100g from the Crosswalk-enhanced FNDDS Version 2, which defaults to the Crosswalk-based profile for mixed dish recipe foods.

e

Values are the difference in energy intake between 2007–2008 and 2011–2012.

4 Discussion

We developed a multi-step approach for incorporating time-varying product- and brand-specific nutrition information from NFP data for >300,000 packaged foods and beverages purchased by US households into time-specific food composition tables for use with nationally representative dietary intake surveys. We demonstrate how this “Crosswalk” between packaged products in the US food supply and food consumption data can be used to study trends in the nutrient content of foods consumed by Americans and the impact of these changes on estimated dietary intake trends. The large number of barcoded products (275,000 to 350,000 per 2-year cycle) that were linked to foods consumed by US children and adults confirms the need for food composition databases to account for the complex, diverse nature of the US food supply. Although we found that the overall energy density of foods consumed by Americans from stores/vending was stable between 2007–2008 and 2011–2012, our approach was able to identify changes in the energy density of foods groups that were not detected using the standard FNDDS. Likewise, declines in energy intake among US children and adults during this time span were small, yet utilizing the Crosswalk-enhanced FNDDS enabled us to detect significant declines in energy intake from beverages among children that were non-significant with the standard approach. Further studies are needed to determine whether our approach can identify more nutritionally meaningful changes in the nutrient content of store-bought foods and corresponding changes in dietary intake, which may be more readily apparent for other nutrients such as sugar or sodium than shown here for calories.

Many scholars have recognized the need for branded food databases as well as the challenges of creating such tools (National Research Council, 2005; Pennington et al., 2007; Ng and Dunford, 2013). To the best of our knowledge, no other existing US database has integrated comprehensive sales-weighted barcode-specific nutrition information into a time-specific food composition table for use with dietary intake data (Ng and Dunford, 2013; The Food Monitoring Group, 2013; Slining et al., 2015). By identifying differences in nutrient content estimated using the Crosswalk-enhanced vs standard FNDDS, our database can help identify food groups to target for updating in FNDDS; examples identified in our study include yogurt and fried potatoes. In addition, the Crosswalk can identify newly emerging products that are not yet included in FNDDS, such as Greek yogurt, and could thereby help in updating FNDDS to capture these newer products. Our Crosswalk-enhanced database also revealed significant declines in total beverage intake among children. As currently only ~50 FNDDS foodcodes are available to represent ~14,000 barcoded SSBs purchased by US households, brand-specific foodcodes for these beverages may be advantageous.

4.1 Potential applications of the Crosswalk-enhanced FNDDS food composition database

An important application of the Crosswalk-enhanced FNDDS food composition database is to examine changes in the saturated fat, sugar, and sodium content of foods reported from stores/vending. In particular, due to recent efforts to reduce the sodium content of packaged foods, such as the National Salt Reduction Initiative, we expect that changes in sodium captured by our Crosswalk database will be more nutritionally meaningful than the minor differences we observed for calories (New York City Department of Health and Mental Hygiene, 2010; Levings et al., 2012; Curtis et al., 2016). Extension of our database to future years can facilitate a comprehensive evaluation of sodium reduction in packaged foods in response to voluntary sodium reduction targets recently proposed by the FDA (US FDA, 2016b). Further, we developed an approach to estimate the added sugar content of industrially formulated packaged foods using linear programming, and Crosswalk-based added sugar profiles can be incorporated into our database (Ng et al., 2015). Thus, the Crosswalk can provide an essential tool for monitoring changes in the added sugar content of packaged foods in response to new regulations requiring inclusion of added sugar on NFPs and corresponding changes in dietary intake among Americans (US FDA, 2016a).

Another potential application of the Crosswalk approach is to generate subpopulation-specific food composition tables to reflect the product or brand preferences of groups by race, ethnicity, income, region, or presence of children in the household. For example, scholars have noted limitations of dietary assessment when foods familiar to or preferred by different racial and ethnic groups are not adequately captured (Signorello et al., 2009; Mossavar-Rahmani et al., 2013). Yet only one prior study has developed race-specific food composition tables; this novel approach revealed that black-white differences in nutrient intakes were underestimated among women using standard vs population-specific food databases (Signorello et al., 2009). In our nationally representative sample of US households, we can determine the exact brands and products purchased by each subpopulation, weight nutrient profiles of products accordingly, and thus generate Crosswalk databases specific for each population.

4.2 Limitations

Several major limitations of our approach are related to the proprietary nature of NFP databases and household purchase data. Commercial datasets are costly to obtain, update, and maintain in order to accurately capture variation in products available and purchased over time. Because this proprietary data are licensed from commercial vendors, our Crosswalk-enhanced FNDDS cannot be shared with other researchers, limiting transparency and replication of our findings. Moreover, this restriction limits the utility of this approach for improving public health.

Although use of NFP data was a key innovation of our database that provided up-to-date product-specific nutrient information, it was also a key limitation. The accuracy of NFP labels is limited because federal regulations require that nutrient content on the label only be within 20% of the actual value, and substantial rounding of nutrient values is allowed (US FDA, 2011). Previous studies have found that the stated energy content of packaged foods such as frozen meals or snack foods was not consistently accurate (Urban et al., 2010; Jumpertz et al., 2013). Although data quality and missing nutrients are additional concerns, our team of RDs does review NFP labels to screen out implausible data.

Our approach is time- and labor-intensive. Thus, because of the large number of UPCs included in our database, we did not have sufficient resources to use a double entry method to link UPCs to FNDDS foodcodes. Another important limitation is that we do not have a system to check the validity and reliability of the RDs’ coding; as described in Appendix B, our quality control measures include checks of the nutrient profile distribution for each foodcode to inspect for potential extreme values and to verify the plausibility of estimates, but do not include checks of the linking of UPCs to FNDDS foodcodes. Further research is needed to develop a method to validate our database and evaluate its accuracy. Conversion of nutrition information from “as purchased” to “as consumed” is a major advantage of our packaged food database, yet this process is complex and cannot account for consumer-level variation in food preparation. Changes in energy density of food groups over time might reflect product reformulations, introduction and removal of products from the market, or shifts in the types of foods purchased and consumed; however, the Crosswalk approach cannot distinguish changes resulting from these different explanations. WWEIA does not collect information about whether a consumed food was home-prepared or an industrially-prepared packaged food; this lack of information limits our ability to determine when use of our NFP-derived nutrient profiles is appropriate.

4.3 Strengths

Our study has important strengths. We used data from two nationally representative samples of the US population to develop our food composition database and demonstrate its application to monitor dietary intake of individuals. The Crosswalk-enhanced FNDDS was developed using an unprecedented amount of time-matched brand- and product-specific food composition data, as it includes ~300,000 branded products per NHANES cycle. A unique advantage of our approach is the review and correction of NFP label data by our team of RDs. Another major innovation of our packaged food database is the conversion of nutrient content from “as purchased” to “as consumed” form to match how foods are reported in WWEIA.

5 Conclusion

Our Crosswalk approach can potentially augment national nutrition surveys by utilizing commercial food purchase and food composition databases to capture changes in the nutrient content of packaged foods and beverages. This system has the potential to advance our understanding of the packaged food sector of the US food system and the impact of product reformulations, introduction of new products, and shifts in purchasing patterns on human health.

Supplementary Material

1

Appendix: Supplementary Material (as Appendices A, B, C), Tables S1-S4.

Highlights.

  • A standard food composition database was enhanced using Nutrition Facts Panel data.

  • A “Crosswalk” was formed between USDA foodcodes and time-matched barcoded products.

  • Foodcode nutrient values were generated by sales-weighting ~300,000 barcoded foods.

  • Trends in dietary intake were detected using Crosswalk-enhanced nutrient profiles.

  • Our approach revealed changes in energy density of select food groups from stores.

Acknowledgments

We thank the Robert Wood Johnson Foundation (Grants 67506, 68793, 70017, 71837), NIH (R01DK098072; DK56350) and the Carolina Population Center (P2C HD050924) for financial support. We also wish to thank Meghan Slining for exceptional earlier research support, Shu Wen Ng and Emily Busey for contributions to review and improve this manuscript, and Phil Bardsley and Kuo-Ping Li for outstanding programming assistance. Authors have no conflicts of interest to declare.

Funding: This work was supported by the Robert Wood Johnson Foundation (grant numbers 67506, 68793, 70017, 71837), NIH (grant numbers R01DK098072; DK56350) and the Carolina Population Center (NIH grant number P2C HD050924).

Footnotes

This paper was originally presented as an oral presentation at the 39th National Nutrient Databank Conference (NNDC), held May 16–18, 2016 in Alexandria, VA (USA)

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Briguglio S, Chirtel S, Spungen J, Rabbani PI, Hoffman-Pennesi D, Wirtz M. Sodium trends in selected US Total Diet Study foods, 2003–2011. Procedia Food Sci. 2015;4:27–38. [Google Scholar]
  2. Curtis CJ, Clapp J, Niederman SA, Ng SW, Angell SY. US food industry progress during the National Salt Reduction Initiative: 2009–2014. Am J Public Health. 2016;106(10):1815–1819. doi: 10.2105/AJPH.2016.303397. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Jacobson MF, Havas S, McCarter R. Changes in sodium levels in processed and restaurant foods, 2005 to 2011. JAMA Intern Med. 2013;173(14):1285–1291. doi: 10.1001/jamainternmed.2013.6154. [DOI] [PubMed] [Google Scholar]
  4. Jumpertz R, Venti CA, Le DS, Michaels J, Parrington S, Krakoff J, Votruba S. Food label accuracy of common snack foods. Obesity. 2013;21(1):164–169. doi: 10.1002/oby.20185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Korosec Z, Pravst I. Assessing the average sodium content of prepacked foods with nutrition declarations: the importance of sales data. Nutrients. 2014;6(9):3501–3515. doi: 10.3390/nu6093501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Leclercq C, Valsta LM, Turrini A. Food composition issues--implications for the development of food-based dietary guidelines. Public Health Nutr. 2001;4(2B):677–682. doi: 10.1079/phn2001153. [DOI] [PubMed] [Google Scholar]
  7. Levings J, Cogswell M, Curtis CJ, Gunn J, Neiman A, Angell SY. Progress toward sodium reduction in the United States. Rev Panam Salud Publica. 2012;32(4):301–306. doi: 10.1590/s1020-49892012001000009. [DOI] [PubMed] [Google Scholar]
  8. Mossavar-Rahmani Y, Tinker LF, Huang Y, Neuhouser ML, McCann SE, Seguin RA, Vitolins MZ, Curb JD, Prentice RL. Factors relating to eating style, social desirability, body image and eating meals at home increase the precision of calibration equations correcting self-report measures of diet using recovery biomarkers: findings from the Women's Health Initiative. Nutr J. 2013;12:63. doi: 10.1186/1475-2891-12-63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Mozaffarian D, Jacobson MF, Greenstein JS. Food reformulations to reduce trans fatty acids. N Engl J Med. 2010;362(21):2037–2039. doi: 10.1056/NEJMc1001841. [DOI] [PubMed] [Google Scholar]
  10. National Research Council. Improving Data to Analyze Food and Nutrition Policies. The National Academies Press; Washington, DC: 2005. [Google Scholar]
  11. New York City Department of Health and Mental Hygiene. National Salt Reduction Initiative. 2010 Retrieved July 30, 2016 from https://www1.nyc.gov/site/doh/health/health-topics/national-salt-reduction-initiative-packaged-food.page.
  12. Ng SW, Bricker G, Li KP, Yoon EF, Kang J, Westrich B. Estimating added sugars in US consumer packaged goods: An application to beverages in 2007–08. J Food Compost Anal. 2015;43:7–17. doi: 10.1016/j.jfca.2015.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Ng SW, Dunford E. Complexities and opportunities in monitoring and evaluating US and global changes by the food industry. Obes Rev. 2013;14(Suppl 2):29–41. doi: 10.1111/obr.12095. [DOI] [PubMed] [Google Scholar]
  14. Ng SW, Popkin BM. Monitoring foods and nutrients sold and consumed in the United States: dynamics and challenges. J Acad Nutr Diet. 2012;112(1):41–45. e44. doi: 10.1016/j.jada.2011.09.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Ng SW, Popkin BM. The Healthy Weight Commitment Foundation pledge: calories purchased by U.S. households with children, 2000–2012. Am J Prev Med. 2014;47(4):520–530. doi: 10.1016/j.amepre.2014.05.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Ng SW, Slining MM, Popkin BM. Turning point for US diets? Recessionary effects or behavioral shifts in foods purchased and consumed. Am J Clin Nutr. 2014;99(3):609–616. doi: 10.3945/ajcn.113.072892. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Ni Mhurchu C, Capelin C, Dunford EK, Webster JL, Neal BC, Jebb SA. Sodium content of processed foods in the United Kingdom: analysis of 44,000 foods purchased by 21,000 households. Am J Clin Nutr. 2011;93(3):594–600. doi: 10.3945/ajcn.110.004481. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Otite FO, Jacobson MF, Dahmubed A, Mozaffarian D. Trends in trans fatty acids reformulations of US supermarket and brand-name foods from 2007 through 2011. Prev Chronic Dis. 2013;10:E85. doi: 10.5888/pcd10.120198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Pennington JA, Stumbo PJ, Murphy SP, McNutt SW, Eldridge AL, McCabe-Sellers BJ, Chenard CA. Food composition data: the foundation of dietetic practice and research. J Am Diet Assoc. 2007;107(12):2105–2113. doi: 10.1016/j.jada.2007.09.004. [DOI] [PubMed] [Google Scholar]
  20. Popkin BM, Haines PS, Siega-Riz AM. Dietary patterns and trends in the United States: the UNC-CH approach. Appetite. 1999;32(1):8–14. doi: 10.1006/appe.1998.0190. [DOI] [PubMed] [Google Scholar]
  21. Poti JM, Mendez MA, Ng SW, Popkin BM. Is the degree of food processing and convenience linked with the nutritional quality of foods purchased by US households? Am J Clin Nutr. 2015;101(6):1251–1262. doi: 10.3945/ajcn.114.100925. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Poti JM, Popkin BM. Trends in energy intake among US children by eating location and food source, 1977–2006. J Am Diet Assoc. 2011;111(8):1156–1164. doi: 10.1016/j.jada.2011.05.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Signorello LB, Munro HM, Buchowski MS, Schlundt DG, Cohen SS, Hargreaves MK, Blot WJ. Estimating nutrient intake from a food frequency questionnaire: incorporating the elements of race and geographic region. Am J Epidemiol. 2009;170(1):104–111. doi: 10.1093/aje/kwp098. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Slining MM, Mathias KC, Popkin BM. Trends in food and beverage sources among US children and adolescents: 1989–2010. J Acad Nutr Diet. 2013a;113(12):1683–1694. doi: 10.1016/j.jand.2013.06.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Slining MM, Ng SW, Popkin BM. Food companies' calorie-reduction pledges to improve U.S. diet. Am J Prev Med. 2013b;44(2):174–184. doi: 10.1016/j.amepre.2012.09.064. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Slining MM, Yoon EF, Davis J, Hollingsworth B, Miles D, Ng SW. An approach to monitor food and nutrition from "factory to fork". J Acad Nutr Diet. 2015;115(1):40–49. doi: 10.1016/j.jand.2014.09.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Stern D, Ng SW, Popkin BM. The Nutrient Content of U.S. Household Food Purchases by Store Type. Am J Prev Med. 2016;50(2):180–190. doi: 10.1016/j.amepre.2015.07.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Taillie LS, Ng SW, Popkin BM. Gains Made By Walmart's Healthier Food Initiative Mirror Preexisting Trends. Health Aff (Millwood) 2015;34(11):1869–1876. doi: 10.1377/hlthaff.2015.0072. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Taillie LS, Ng SW, Popkin BM. Walmart and Other Food Retail Chains: Trends and Disparities in the Nutritional Profile of Packaged Food Purchases. Am J Prev Med. 2016;50(2):171–179. doi: 10.1016/j.amepre.2015.07.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. The Food Monitoring Group. Progress with a global branded food composition database. Food Chem. 2013;140(3):451–457. doi: 10.1016/j.foodchem.2012.10.065. [DOI] [PubMed] [Google Scholar]
  31. The Nielsen Company, Nielsen Homescan Consumer Panel. University of North Carolina calculation based in part on data reported by Nielsen through its Homescan Services for all food categories, including beverages and alcohol for the 2000–2012 periods, for the U.S. market. The Nielsen Company; Copyright © 2014. [Google Scholar]
  32. Urban LE, Dallal GE, Robinson LM, Ausman LM, Saltzman E, Roberts SB. The accuracy of stated energy contents of reduced-energy, commercially prepared foods. J Am Diet Assoc. 2010;110(1):116–123. doi: 10.1016/j.jada.2009.10.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. US FDA. Title 21 - Food and Drugs; Chapter 9 - Federal Food, Drug, and Cosmetic Act. 21 U.S.C. 2011. [Google Scholar]
  34. US FDA. A Food Labeling Guide: Guidance for Industry. FDA; College Park, MD: 2013. [Google Scholar]
  35. US FDA. Changes to the Nutrition Fact Label. 2016a Retrieved July 15, 2016 from http://www.fda.gov/Food/GuidanceRegulation/GuidanceDocumentsRegulatoryInformation/LabelingNutrition/ucm385663.htm.
  36. US FDA. Sodium Reduction. 2016b Retrieved July 20, 2016 from http://www.fda.gov/Food/IngredientsPackagingLabeling/FoodAdditivesIngredients/ucm253316.htm.
  37. USDA. CDC/National Center for Health Statistics, NHANES Dietary Web Tutorial. Retrieved March 2012 from http://www.cdc.gov/nchs/tutorials/Dietary/Basic/PopulationMeanIntakes/intro.htm.
  38. USDA, CDC/National Center for Health Statistics. National Health and Nutrition Examination Survey 2007–2008. 2010 Retrieved May 2013 from http://wwwn.cdc.gov/Nchs/Nhanes/Search/Nhanes07_08.aspx.
  39. USDA, CDC/National Center for Health Statistics. National Health and Nutrition Examination Survey 2009–2010. 2012 Retrieved May 2013 from http://wwwn.cdc.gov/Nchs/Nhanes/Search/Nhanes09_10.aspx.
  40. USDA, CDC/National Center for Health Statistics. National Health and Nutrition Examination Survey 2011–2012. 2014 Retrieved September 2014 from http://wwwn.cdc.gov/Nchs/Nhanes/Search/Nhanes11_12.aspx.
  41. USDA Agricultural Research Service. National Nutrient Database for Standard Reference. Retrieved June 2016 from the Nutrient Data Laboratory Home Page http://www.ars.usda.gov/ba/bhnrc/ndl.
  42. USDA Agricultural Research Service. USDA Food and Nutrient Database for Dietary Studies 4.1. 2010 Retrieved June 2013 from the Food Surveys Research Group Home Page http://www.ars.usda.gov/Services/docs.htm?docid=12089.
  43. USDA Agricultural Research Service. USDA Food and Nutrient Database for Dietary Studies 5.0. 2012 Retrieved August 2014 from the Food Surveys Research Group Home Page http://www.ars.usda.gov/Services/docs.htm?docid=12089.
  44. USDA Agricultural Research Service. USDA Food and Nutrient Database for Dietary Studies 2011–2012. 2014 Retrieved October 2014 from the Food Surveys Research Group Home Page http://www.ars.usda.gov/Services/docs.htm?docid=12089.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1

Appendix: Supplementary Material (as Appendices A, B, C), Tables S1-S4.

NIHMS894596-supplement-1.docx (79.8KB, docx)

RESOURCES