Scope and principles of the Nestlé Nutritional Profiling System (NNPS)
Our goal was to develop a nutrient profiling system to help guide the formulation of new foods and beverages or the reformulation of existing ones. This methodology is being implemented for the majority of the Nestlé food and beverage portfolio. Infant formulas and medical nutrition products are regulated separately and are beyond the scope of the NNPS, as are food products for children <4 years of age. The NNPS was only applied to the brands owned by Nestlé; the products sold from Cereal Partners Worldwide (a joint venture between Nestlé and General Mills) were not profiled with the NNPS but with the specific profiling system used by the joint venture .
The outcome of the NNPS is dichotomous, i.e., “YES” or “NO,” and relies on pre-defined target values for a set of nutritional factors. To achieve a positive outcome, all target values must be met, i.e., the algorithm is non-compensatory. Four guiding principles were used to define targets for the nutritional factors: (1) nutritional factors include energy, nutrients to limit and nutrients to encourage; (2) the set of nutritional factors and associated target values is category specific; (3) target values are age specific, depending on the products; and (4) target values are defined per serving of the product as consumed.
Pure roast and ground coffee, soluble coffee, pure tea, plain water and products consisting of >95 % whole milk with no added energy-providing nutrients were identified as not in the scope of reformulation for nutrients to limit and are therefore scored as “YES.”
The system, originally developed in 2004, has a dynamic design that allows for the product categories, nutrient targets and reference values to be modified to incorporate the latest evidence in the area of food technology and public health nutrition. Here, we present the latest version of the system as modified in March 2014.
Development of product categories
The NNPS food categorization was decided by a cross-functional team that included food technologists and nutritionists and taking into account the renovation history of the food category as well as latest innovations in food science. The food categorization process was structured in two steps: (1) identifying the product role in the diet and (2) grouping products based on similar nutrient composition and/or challenges in the reformulation process.
In the first step, the role of the product in the daily diet was considered. The total daily energy intake was divided into the different eating occasions. Based on the results of a literature review of dietary guidelines and dietary intake surveys , the NNPS assumes an energy repartition pattern of three main meal occasions, each accounting for 20–35 % of daily energy intake, and 1–2 snacking occasions, each accounting for 5–10 % of daily energy intake. Three main product roles were identified from these eating occasions: larger meal components, smaller meal components/snacks, and accessories. Larger meal components represent foods that provide the main source of energy during one of the three main meals. Smaller meal components and snacks are products either consumed in-between meals or as part of one of the main meals, but in the latter case they are normally consumed as an appetizer, dessert or individual element of a dish providing less than one-third of the overall energy from the meal. Accessories are products consumed as a minor component of a meal (cold sauces, dressings) or between meals (e.g., hard candies) and share the common characteristic of very small energy contribution to the daily diet.
The second step in defining food categories aimed at pooling together products based on those parameters significantly influencing a product’s nutrient composition. This allowed for the establishment of challenging yet achievable targets. Such grouping was either:
Ingredient based, i.e., based on the presence of a dominant raw material or of a specific combination of raw materials (cereal- based, milk- based, cocoa- based - e.g., chocolate - , vegetable based - e.g., soups-, or a specific combination - e.g., pizzas)
Process based, i.e., products sharing the same unique production process impacting the nutritional composition (e.g., ice creams, cheeses). In the case of ice cream and cheese the key ingredient is dairy milk/cream in both, but technological constraints are different due to the role of salt in cheese for ripening and preservation versus the role of sugar in the freezing of ice cream.
Combination of the above, i.e., water ice and sorbet share common production considerations with ice cream (freezing), but they differ in the principal ingredient (dairy vs. water/juice based).
Categories relevant to the Nestlé food and beverage portfolio were considered; this included the majority of manufactured foods and beverages available in the food supply. The categories were named in such a way that food developers could easily identify which Nestlé products would fall into a specific category (e.g., dairy-based accessories were named as sweetened condensed milk based on the products available in the Nestlé portfolio). The resulting food categories are listed in Table 1.
Selection of nutritional factors
Based on the World Health Assembly Global Strategy on Diet, Physical Activity and Health , saturated fatty acids (SFA), trans-fatty acids (TFA), added sugars (AS) and sodium (Na) were designated as compulsory nutrients to limit across all categories. The NNPS is a system designed to be equally applicable for use by food developers and nutritionists; to facilitate optimal guidance for the end user, additional factors were incorporated into the system. Under this scope, three additional factors were added for these respective reasons: total fat (TF) to ensure the overall reduction in fat content alongside the removal of SFA and TFA, energy (E) to guide portion size optimization and fructose (Fr) to ensure that fructose or high-fructose corn syrup (HFCS) (ingredients with higher sweetening power) were not used excessively in light of any possible health effects of high-fructose consumption  (Table 1). Only TFA originating from partially hydrogenated oils were considered in the calculation of nutrient profiles, in line with the company’s policy on trans-fat . Added sugars were defined as all mono- and disaccharides added during the manufacturing or preparation of a product. Naturally occurring sugars (lactose from milk/dairy fractions, mono- and disaccharides from unsweetened fruit ingredients) were excluded provided that the unsweetened fruit ingredient was not added for sweetening purposes. Based on the molecular structure, any added sucrose was counted as 50 % fructose and 50 % glucose. For the calculation of sodium, both added sodium and sodium from natural sources were considered.
Alongside the nutrients to limit, a list of nutrients to encourage was identified (e.g., protein, calcium, fiber). Targets for nutrients to encourage were included in product categories where this was considered sensible and where the nutrients supported the role of the product in the diet (Table 1). All products falling under the larger meal components group were required to include at least one nutrient to encourage, as they are the main energy sources of the diet. Due to their importance in the diet, the categories complete meals, and side dishes & center of plate food items (i.e., food items acting as the principal protein source in a main meal) needed to be profiled against a minimum of two nutrients to encourage, including protein. In such cases, the second compulsory nutrient to encourage had to be selected from a list of nutrients/ingredients. For milk and dairy products, protein and calcium were considered nutrients to encourage. Fiber was considered a nutrient to encourage for cereal-based foods. The list of other nutrients/ingredients to encourage included fruits & vegetables, whole grains, vitamins, minerals and essential fatty acids all selected according to the relevance for the specific food product (Table 1). No nutrients to encourage were applied to products classified as accessories, because their contribution to the diet was considered too small.
Selection of age-appropriate daily reference values
Daily reference values (DVs) of various nutrients to encourage and various nutrients to limit are presented in Table 2. Consumers were split into three age groups: adults, children 4–8 years old and children 9–11 years old (children >11 years old were considered as “adults”). Public health nutritional recommendations for each age group were applied to define what constitutes a nutritionally adequate diet.
The United States Food and Drug Administration (USFDA) DVs  in combination with the Institute of Medicine (IOM) recommendations for total fat intake  were used as the initial basis to set the NNPS global DVs for a 2000 kcal daily diet for adults. In the absence of specific US FDA values for trans-fat and added sugars, the WHO recommendations were followed [23, 24]. In each NNPS update the latest recommendations of the WHO and other bodies were considered in order to adjust the DVs. DVs for children were adopted accordingly for a 1700 kcal daily diet for children 4–8 years old based on the IOM recommendation . For children, dietary fiber intake followed the European Food Safety Authority (EFSA) guidance , while for all micronutrients, except sodium, adult DVs followed CODEX recommendations  and children DVs followed recommendations of the Joint FAO/WHO Expert Consultation . NNPS is applied globally through a dedicated proprietary software which allows local DVs to be applied as the basis of the profiling system instead of using international DVs, when the former are available. This decision was made to allow the system to address local dietary needs.
Selection of the base for calculation
Nutritional factor targets were set based on the assumption that food products are consumed as part of a nutritionally adequate diet . As a result, the base for calculation was decided to be one serving of the final product as consumed after reconstitution and/or cooking. Consumer behavior data from local market research were used to define the eating habits and estimate amounts customarily consumed as portion sizes in each geographical region separately. To help guide the reformulation of serving, maximum energy targets per serving were developed for each food category.
Development of nutrient and energy targets per category
Following the principles described below, target values were set for energy, nutrients to limit and nutrients to encourage (Tables 3, 4).
Energy targets were developed for each product category based on the energy repartition pattern of three main meals (20–35 % DV for energy each) and two snacks (5–10 % DV for energy each) (Table 3): for larger meal components, the maximum energy target was set between 15 and 30 % of total energy depending on whether the components were the sole source of energy in the meal or not; for smaller meal components and snacks, the energy target was set to ≤10 %; for accessories, the target was set to ≤5 %.
The first approach toward setting targets for nutrients to limit was to set up baseline targets aligned with the energy targets for the product category. For example, product categories delivering ≤10 % of daily energy per serving should not deliver more than 10 % of the DV of a given nutrient to limit per serving. Alternatively, for larger meal components delivering ≥10 % of energy per serving, nutrient targets were aligned as closely as possible with inter- or- national dietary recommendations. For example, the WHO recommends that total fat intake should not exceed 30 % of daily energy, and therefore, the target for total fat in cereal-based foods (as a main meal) was set accordingly. As a result, targets for nutrients to limit were expressed in two ways, either as % of energy (e.g., total fat ≤30 % energy) or as % DV (e.g., sodium ≤10 % DV/serving). The reference unit as % of energy was used for product categories with a higher contribution to the daily energy intake (i.e., larger meal components) for nutrients providing energy; the reference unit as a % DV was used in all product categories delivering ≤10 % of recommended daily energy per serving, for all nutrients.
Category-specific targets were then set up starting from the above-defined “baseline” targets. The setting-up of category-specific targets considered technological and organoleptic aspects by a cross-functional team including nutritionists and food technologists. The following elements were considered: nutrient content of similar foods in the food supply and existing reformulation targets from other profiling systems (Table 2). As an example, although ice creams are not expected to contribute to more than 10 % of the daily energy, the saturated fat and sodium targets were set to ≤20 % DV/serving (allowing for dairy ingredients, including milk fat) and ≤5 % DV/serving (due to a lower intrinsic concentration of sodium in the raw ingredients), respectively, instead of ≤10 % DV/serving. Other examples of category-specific targets include products with a preferred savory or sweet taste, such as salty and savory biscuits (in the case of sodium), or ice cream, water ice and dairy desserts (products where sugar has technical and taste properties) in the case of added sugars. In the case of sodium targets, they were aligned with technical considerations and local profiling systems (e.g., sodium target for soups in the Heart Foundation Tick Australia) or as described in other nutrient profiling systems .
In the case of added sugars, a maximum value applying to all category-specific added sugars target was set at ≤25 % DV per serving for the smaller meal components/snacks and accessories product roles and ≤25 % of total energy for larger meal components. This decision was made based on the WHO recommendation of limiting free sugar intake over a maximum of four “sugary” eating occasions  and the US (IOM) recommendation that ≤25 % of energy from added sugars should be consumed to ensure adequate micronutrient intake . Unlike other nutrients, TFA were profiled against one common threshold across all categories (Table 3).
Targets were set in a similar way for nutrients to encourage. For protein, the WHO guidance of 10–15 % of total energy  was translated into a target of ≥12 % of energy per serving (Table 3). The nutrient target for calcium was set to ≥111 mg/100 kcal in alignment with the protein target, based on an average calcium-to-protein ratio of 37 mg/g found in cow’s milk. For dairy desserts, the calcium target was set to ≥5 % DV/100 kcal based on the CODEX minimum level to claim “source of calcium” for adult products. For children products, the calcium target was set to ≥6 % DV/100 kcal adapted from the target for adults, taking into consideration the lower total energy intake of children (1700 kcal vs. 2000 kcal). The nutrient target for fiber was aligned with the USFDA criterion for “source of fiber” and set to ≥10 % DV per serving (Table 4) . The micronutrient targets were set in alignment with the local requirements for a product to be classified as “source of” (Table 2).
In relevant product categories, targets for fruit and vegetables and whole grain were set at ≥1/2 to ≥1 portion per serving of food or beverage product depending on the product role. A portion of fruit and vegetables was defined as 80 g of fruit and vegetables. The WHO population nutrient intake goals recommend the consumption of at least 400 g of fruit and vegetables per day . This recommendation, combined with local campaigns for the promotion of consumption of fruit and vegetables (“5-a-day” campaigns), allows defining one portion of fruit or vegetable at 80 g. For whole grain, the US dietary guidelines  recommend consuming at least 3 oz equivalent of whole grain per day. One ounce corresponds to 28 g of whole grain product, corresponding to 16 g of pure whole grain which was selected as a portion of whole grain.
For the specific case of milk and dairy products for children as well as cereal consumed with milk, the decision was made for the system to allow for whole milk to be used; as a result, the nutrient targets for total fat and saturated fat were adjusted accordingly.
Validation and testing
A scientific advisory board was put in place to provide guidance and to ensure the validity of the system. In the interest of testing the consistency and magnitude of nutritional changes associated with the application of the NNPS, a case study was conducted on eight product categories in the USA and France: pizza, milk-based beverages (as a snack), water ice and sorbet, and complete meals for the USA; and children’s ice cream, center of plate food items (main protein carriers), soups and cold sauces for France.
The USA and France were selected, being two major countries for Nestlé in terms of sales. Product categories were selected based on the respective volume of sales (covering 40–50 % of total sales) and data completeness of the nutritional information at the time of analysis (nutritional information for 2014–2015). Categories selected were also representative of the three main product roles as defined by the NNPS (i.e., larger meal components, smaller meal components/snacks, and accessories). For each product category, the 15 most widely purchased Nestlé products were identified, unless a product category included less than 15 products in which case all products available were analyzed. Nutrient profiles were retrieved for the years 2009–2010 (as reference year) and 2014–2015 in a paired manner, based on the availability of a complete dataset, so that longitudinal changes after the NNPS application could be analyzed. For the purpose of the manuscript, analyses were focused on total sugars rather than added sugars. Total sugars content is a regulatory requirement for on-pack labeling, and hence, the accuracy of retrospective data is higher than for added sugars. That in turn ensures higher accuracy when studying longitudinal changes. Added sugars are the only fraction of total sugars that is amenable to reformulation; therefore, the reported changes in added sugars reflect directly changes in total sugars. There was no attempt to reduce e.g., naturally occurring sugars in milk. Portion sizes as indicated on the labels were used.
Given that the sample selected for the analysis is a fixed sample of products and not meant to represent all products under the same category, only descriptive statistics were used to identify changes in key nutrients per product category. All reductions were compared against an arbitrary 10 % cutoff point, with this cutoff being used as a measure of how extensive the reformulation has been. As the same parent recipe may be used to produce more than one product, the decision was made to analyze changes in the composition of the most widely purchased products and not the composition of the parent recipes (i.e., not to merge products based on the parent recipe). This strategy focuses on the impact at the consumer level and not on the number of recipes reformulated, plus it accounts for differences in packaging and usage, which can have an impact on the final amount of product (serving) consumed.