We reviewed dietary intake data from 9255 respondents who completed an in-person 24-h dietary recall as part of What We Eat in America (WWEIA), NHANES 2007–2008 (22). WWEIA, NHANES is an ongoing dietary intake survey of the nationally representative, noninstitutionalized US population of all ages. Sodium density (mg/100 g of food), frequency of consumption by respondents in the survey, and percentage of contribution to sodium intake of commercially processed and restaurant foods were carefully evaluated to determine the list of 125 sentinel foods. Table 1 shows examples of the sentinel foods, by food type [adapted from the WWEIA food categories (23)]. Supplemental Table 1 lists all of the foods. Approximately half of the sentinel foods are in the 10 food categories that contribute the most sodium to the US diet according to the CDC (4). Other sentinel foods, such as catsup (“condiments and sauces”) and French fries (“potato products”), are foods with high sodium density and/or that are very popular.

Approximately three-fourths of the sentinel foods are commercially processed (92 of 125), and the rest come from fast-food or restaurant chains (33 sentinel foods). Sentinel foods account for approximately one-third of the total sodium intake of all individuals, excluding breastfed infants, in WWEIA 2007–2008. The USDA validated the accuracy of mean dietary sodium intake estimates in the survey by comparing these data to the results of 24-h urinary sodium excretion tests (24).

The NDL developed a sampling plan for each food item that it used to represent a given sentinel food. The sampling design used a hierarchical, 3-stage, probability-proportional-to-size sample selection process to ensure the selection of a nationally representative sample of food products that were geographically dispersed across the United States. The 3 stages of this selection process were as follows (25): 1) county and/or city (based on population density from US Census data), 2) retail locations (e.g., supermarkets and restaurants) within the counties and cities (based on annual sales from Nielsen data and Trade Dimensions data), and 3) food brands based on market shares of units sold (derived by the NDL from Nielsen data).

The NDL selected 12 counties across the United States during stage 1 on the basis of the most recent US Census data available. The sampling frame for foods sampled before summer 2013 was based on data from the 2000 Census, which was subsequently updated with data from the 2010 US Census (25, 26). At stage 2, for commercially processed foods, the NDL selected 1 retail outlet and 2 alternates in each of these 12 counties and cities. The retail outlets that we used to purchase sentinel foods included Walmart and grocery stores with gross sales >$2 million. In stage 3, we determined the market shares of food brands using product-level point-of-sales Nielsen data (27). These data captured unit sales from the vast majority of US grocery stores that have annual sales of at least $2.0 million. However, they did not include sales from warehouse-type stores (e.g., Costco), Walmart, and stores with gross sales of <$2 million. We assumed that consumer shopping behavior for types of brands purchased was similar across store types for these analyses. We used 2009 Nielsen sales data for foods sampled before summer 2013, which we subsequently updated with Nielsen sales data from 2012.

The NDL identified major national and private brands of sentinel foods on the basis of units sold in a calendar year for these foods. When the market share of the private brands collectively was high (i.e., the brand was among the top 3 or 4 brands), we purchased the private brands of the retail location that it had identified for sampling because the Nielsen data do not identify stores associated with private brands. We sampled private brands for two-thirds of the commercially processed sentinel foods.

The number of brands sampled varied by food item, but most foods had a market leader or a few prominent brands that were selected. For example, for the sentinel food “meat and poultry hot dog,” there are 274 universal product codes and 74 brands in the Nielsen data. However, only 5 national and private brands account for >80% of sales (see Figure 2). Some foods had no prominent brands. For example, for “cheese pizza, thin crust, frozen,” 20 brands represented 70% of the market. In such cases, the brands sampled were limited to the top 2 or 3 because of cost constraints of nationwide sampling and laboratory analysis.

We reviewed the labels for the universal product codes with the highest market share within each brand, when available. In most cases, we selected the universal product codes with the highest sales within the brand for purchase on the basis of the assumption that different package sizes within a brand have similar nutrient values per serving.

Limited analysis of the Nielsen data shows that, for many food categories, a small number of products represent a major proportion of that market. For example, among ready-to-eat cereals, there are ∼ 3000 unique universal product codes and 1300 unique cereals. Of these, only 127 cereals account for 88% of total sales of all cereals. Hence, it is cost-effective to focus on the items with the highest sales (28).

For fast-food and restaurant items, we primarily used QSR magazine’s list of top 50 restaurants and their menus to select samples of the relevant sentinel food (29). We also used information on the most common fast-food restaurants reported by WWEIA 2007–2008 respondents (obtained from WWEIA 2007–2008 nonpublic internal files) to purchase relevant sentinel foods. We sampled local, independent restaurants for Asian mixed dishes and a few other selected ethnic foods. We used industry publications and/or contracted data analyses to identify prominent brands before 2011 (i.e., before NDL purchased Nielsen data).

An NDL-directed professional product-purchasing company purchased sample units of each of the major brands and shipped it under optimum conditions (30) to the Food Analysis Laboratory Control Center at Virginia Tech (Blacksburg, VA) or the Texas Tech Department of Animal and Food Sciences (Lubbock, TX). The scientists at these universities processed and prepared the samples for chemical analyses. They randomly mixed the 12 sample units for each brand to yield 6 composites for cost-efficiency. Sometimes, low-sales regional or national brands were composited with private brands to reduce laboratory costs. The composites were then shipped to prequalified commercial and university laboratories for chemical analyses. Figure 2 shows the steps in the sampling and analysis of the sentinel food “meat and poultry hot dog.”

The laboratories analyzed most foods to develop a full nutrient profile of up to 133 nutrients comprising macronutrients, minerals, vitamins, fatty acids, and amino acids. The laboratories analyzed sodium by inductively coupled plasma atomic emission spectroscopy by using the Association of Analytic Chemists method 985.01 (3.2.06) + 984.27 (50.1.15) (20). Details on the methods used for other nutrients are available in the SR documentation (20). We used a rigorous quality-control program that included analysis of in-house control materials and standard reference materials, blinded with food samples to monitor the accuracy of the chemical analyses. A panel of food specialists and chemists reviewed the nutrient data obtained from the laboratories to verify the accuracy and precision of the analytic results (31). We weighted the analytic nutrient values by the market share of the selected brands of commercially processed foods to generate nationally representative values for the SR. For fast-food restaurant items, we weighted nutrient values by the market share of the restaurant on the basis of total US dollar sales data from QSR magazine (29). For family-style restaurant foods, we equally weighted all brands because market share data on brands were not available. For some foods, when the new analytic values for sodium and related nutrients were not much different from old analytic data, we combined the new analytic data with the old.

We will resample and analyze the sentinel foods every 4–8 y using the same methods as discussed above. To determine how frequently to resample the foods, we divided the foods into 4 groups on the basis of these criteria:

Frequency of consumption of the sentinel food based on WWEIA 2009–2010 dietary intake data (32).

We created a summated scale using the shared variance contributions of the 3 variables based on the principal axis factoring (factor analysis) method. On the basis of the scale, we divided the sentinel foods into 4 groups that provide a framework for the frequency with which we will resample sentinel foods. Foods that WWEIA respondents reported consuming most frequently have the greatest percentage difference between the baseline and proposed targets from the New York National Salt Reduction Initiative, and whose values have changed most frequently in the SR, represent group 1. Supplemental Table 3 gives examples of sentinel foods categorized into the 4 groups, along with the values for the 3 criteria above. We will resample foods in group 1, such as “American cheese” and “white bread,” every 4 y and foods in group 4, such as “canned tuna, in water” and “tomato juice,” every 8 y. However, we may adjust the sampling frequency for logistical reasons, such as to equally distribute food sampling over the years for resource allocation, to analyze foods using similar quality-control materials, and to sample foods from similar locations at the same time for costefficiency. For example, we are likely to sample pizza items in groups 2 and 3 in the same pick-up to ensure efficient resource use. We will use the newest market share data available at that time to select brands to sample and then to weight the analytic nutrient values.

We will track changes in sodium content by comparing baseline analytic values (2010–2013) with sodium content from resampling (2014 onward). We will statistically analyze changes of ≥10% for significance. We deem 10% to be an appropriate change in nutrient values to justify further review. The CV is ∼ 6% for the sodium analyses of duplicate samples of matrixmatched food materials from the different laboratories that the NFNAP uses. This value has been no higher than 6% for most control materials over the past several years (K Patterson, NDL, personal communication, December 2013).

In this report, we compare the new analytic data that the NDL has generated since 2010 with older SR values to identify changes in values of sodium of at least 10%. For these foods, we identified changes in one or more related nutrients of at least ±10%. We did not statistically analyze differences in sodium and related nutrients for this report, because some of the older SR values were based on label data, analytic data from literature, or not based on nationally representative sampling plans.

We selected the P2Fs on the basis of careful examination of dietary intake data from the WWEIA 2009–2010. Food descriptions, frequency of consumption, sodium density (mg/100 g of food), and source of sodium values were reviewed to identify ∼ 1200 commercially processed and restaurant foods other than sentinel foods that contained added sodium and, hence, their sodium content had the potential to be modified. The P2Fs exclude foods with naturally present sodium: milk, yogurt, natural spices and herbs, fruit, infant formulas, most beverages, legumes, unprocessed vegetables, fresh meats and poultry, cereal grains, nuts, seeds, and fish. Foods with sodium values < 50 mg/100 g were also excluded for resource efficiency, resulting in ∼ 1100 foods in the list of P2Fs. The list includes brand-name items, such as Kellogg’s Froot Loops, and generically described items, such as sour cream. We will update the P2F list subsequent to the release of dietary intake data from WWEIA 2011–2012 and so on.

We use the following procedures to monitor sodium values for the P2Fs:

Identification of brands. For generically described P2Fs, such as sour cream or ranch dressing, we identified brands with the highest market share. We used the same procedures as described above for sentinel foods to identify brands for sampling.

After the SR release in 2013, we reviewed changes in sodium values to determine their potential priority for laboratory analysis. We selected foods for possible analysis that WWEIA respondents frequently reported, that had changes in sodium content >10%, and that we had not analyzed for the longest periods.

We sampled all sentinel foods and disseminated updated data for 73 foods in SR releases 23–26 (2010–2013). Similarly, we reviewed all P2Fs in 2010–2013.

Sodium values changed by at least ±10% for 43 of the 73 sentinel foods on the basis of a comparison of the new analytic data in SR 23 (2010) to SR 26 (2013) with older SR data. Table 2 lists the sodium values for the 43 foods, along with the sources of these data (analytic, label, or manufacturer), data points, and the year of the most recent and previous nutrient values in the SR. The content of one or more related nutrients (potassium, total and saturated fat, total sugar, or total dietary fiber) changed by at least ±10% for most sentinel foods. Supplemental Table 4 shows the changes in the related nutrients for these 43 sentinel foods. For ∼70% of the sentinel foods (31 of 43 foods with a ±10% change), the newer sodium values were lower than previous SR values. These foods include commonly consumed foods, such as bread, tomato catsup, French fries and chicken tenders from fast-food restaurants, packaged macaroni and cheese, and potato chips. The updated sodium values were >10% higher than the previous SR values for the remaining 12 sentinel foods, including commonly consumed foods, such as salsa, ham, mayonnaise, and fast-food cheese pizza.

Sodium values changed by at least ±10% for 328 P2Fs, as listed in Supplemental Table 5, from SR 23 (2010) to SR 26 (2013). Sodium values were lower in the SR from previous values for ∼60% of the 328 P2Fs, including frequently consumed foods such as sour cream, butter, frankfurters, selected cheeses, salad dressings, nuts, and selected sweet bakery products. The updated sodium values were >10% higher than previous values for the remaining 40% of P2Fs, including shrimp, selected ready-to-eat cereals, and baked products. One or more related nutrients (potassium, total and saturated fat, total sugar, or total dietary fiber) changed in the SR by at least ±10% for approximately one-quarter of the P2Fs (identified in Supplemental Table 5). We may not have updated data on some of these nutrients in the SR because of the lack of reliable data.

We have scheduled many P2Fs for laboratory analysis on the basis of reviews of changes in their sodium values between SR 23 (2010) and SR 26 (2013). Supplemental Table 6 lists the P2Fs that the NDL analyzed in fiscal year 2014. This table lists the percentage of change in sodium values, frequency of reported intakes of foods based on WWEIA, NHANES 2009– 2010, and the year the food was last analyzed. The results of sodium reduction efforts, based on resampling of the sentinel foods or re-review of P2Fs, will become available beginning in 2015 in a separate “Sodium Monitoring” section on the NDL website.