The concept of nutrient profiling systems has recently aroused renewed interest from scientific and regulatory authorities dealing with nutrition issues1. Indeed, such tools can be useful in helping decision-makers on topics such as allowing a food to bear a nutrition or health claim, restricting television advertisements for certain foods or limiting their occurrence in vending machines. These profiling systems could also help educate consumers more effectively on nutritional matters and facilitate nutrition-oriented innovation and improvements in the food industry.
However, there is currently still some controversy as to the relevance of such tools, which are sometimes thought to be incompatible with, or even to jeopardise, health education programmes. To our mind, the major issue is the characteristics, adequacy and performance of the systems which are proposed. Very briefly, the existing systems can be divided into two broad categories.
Some systems favour an ‘across the board’ approach, in which every food is positioned using the same nutritional criteria. We have recently analysed four of these tools by comparing their performance in classifying a series of 125 food items and, although some tools seem more accurate than others, there are still difficulties in reaching full consistencyReference Azaïs-Braesco, Goffi and Labouze2. An additional pitfall of this approach is its propensity to heap opprobrium on some food categories as a whole, such as fats, which nearly always appear among the least favourable food products, whatever their quality. This is in conflict with the usual, and still not debated, need for a balanced but varied diet in which any food can find its place, provided that the amount and frequency of consumption are related to its nutritional characteristics.
Another set of tools includes a variety of systems, most of the time unpublished, which take into account food groups but do not consider a given food item with the same nutritional criteria or thresholds, depending on the food group it belongs to. Such systems have been developed in Sweden3 and The Netherlands4, for example. The concept is interesting and probably more in line with the issue of a wholesome diet, but an appropriate methodology and sometimes the scientific justification for the choice of criteria and thresholds are lacking. Moreover, these systems remain relatively rigid, and cannot easily be adapted to various contexts (e.g. food for adults or children) or uses (e.g. regulatory, educational or help in formulation).
We describe here a nutrient profiling system, named Nutrimap®, that aims to position food items in relation to others within the same food category, and which pays special attention to flexibility and pragmatism. The principles, scientific background and implementation of the system are described and some uses are presented in more detail.
The position of a food product within the Nutrimap® system is defined according to its nutritional composition, the food category it belongs to, the nutritional needs of the consumer, available consumption data and major public health objectives for nutrition. Of these five items, the last two are clearly country-dependent. Nutrimap® is described here in its French/healthy adult version, but the flexibility of the system enables it to be adapted easily to other situations.
Development of Nutrimap®
Nutritional criteria
The selection of nutritional criteria has been driven by public health objectives, as detailed in several reports by the World Health Organization5, the Eurodiet task forceReference Ferro-Luzzi, Gibney and Sjöström6 or, in France, the PNNS7 (National Programme for Nutrition and Health). Nutrimap® uses a set of 15 nutritional criteria (Table 1), chosen because of the nutritional issues they raise. Lipids are considered both for their quantity (% of lipid energy) and their quality (saturation level of fatty acids), as are carbohydrates (quantity as % of carbohydrate energy; quality as % of sugars). Fibre, vitamins (folic acid, C, D, E), iron, calcium and magnesium are considered because their intake in France is below the recommended levels in some adult population groups. Sodium is taken into account because of the current excessive intake by the French population8.
The nutritional criteria taken into account in Nutrimap® and thresholds of recommended intake and current consumption for healthy French adults, where relevant (i.e. not for vitamins and minerals, see text)
Although other choices could have been made, energy has been chosen as the reference basis because it seems consistent with the increasingly consensual concepts of nutritional and/or energy density. Nutritional criteria are thus expressed in weight units per 100 kcal of food.
Scoring
Each of the criteria is then allocated a score between − 1 and +1. The score depends on the amount of the nutrient present in 100 kcal of the product, as illustrated in Fig. 1. For a nutrient whose intake should be limited (fat, saturated fats, sodium and sugars), the score will be − 1 if the food under study contains more than the actual recorded French intake. This would mean that the considered product increases the imbalance already observed. The score will be +1 if the considered food contains less than the recommended maximum intake, meaning that the product is able to rebalance the diet for this nutrient. The score will develop in linear fashion between − 1 and +1 if the value for this nutrient is between the recommended maximum intake and the actual intake. For a nutrient whose intake should be increased (carbohydrates, fibre, polyunsaturated fats, vitamins and minerals), the argument is reversed. Nutrimap® thus uses two thresholds for each nutrient: its recommended intake and its current consumption. These thresholds, as determined for a population of French adults, are shown in Table 1. At this stage it is possible to use the system for specific purposes; for example, considering the recommended values for children and their recorded intake will address the nutritional relevance of a given food item for this age group.
Scoring mechanism for each nutrient whose intake is not consistent with recommendations
Lipid quality is estimated as the mean of the scores for saturated, mono- and polyunsaturated fats (or monounsaturated fats if only these are available) and only considered if more than 10% of energy is of lipid origin.
Although vitamins and minerals have been chosen in relation to a recorded gap between recommendation and current supply, this gap does not exist for the whole adult population and is different for men, women, young adults and people over 50 years of age. For this reason, the ‘level of current consumption’ threshold is not activated for vitamins and minerals (except sodium), and the scoring allocates a mark of ‘+1’ when the supply exceeds the daily recommended intake as expressed per 100 kcal, a mark of ‘0’ when the supply is below half this level and a mark which develops in a linear fashion between 0 and +1 otherwise; vitamin and mineral criteria cannot have a negative score. All vitamin marks are summed. In order to avoid an over-representation of vitamins, which could lead to vitamin supplementation where this is not in fact needed, the maximum score for vitamins is set at +2. A similar rule applies to minerals, apart from sodium.
Food categories
The definition of food groups is always a matter of debate. This is a pragmatic concept which has been developed to make it easier to vary diet and whose purpose is primarily to ensure an overall balanced diet. We tried here to avoid the occurrence of too many food groups, which would have been difficult to handle and interpret correctly. We have thus defined seven food categories (Table 2) in accordance with the groupings most often encountered in the literatureReference Gatenby, Hunt and Rayner9 and with the dietary guidelines of various public health authorities. Five to nine food groups are usually identified. These are usually defined, based on recommendations, according to the nutrient(s) they provideReference Gatenby, Hunt and Rayner9, Reference Gibney and Sandström10. Carbohydrates are thus supposed to be mostly provided by the ‘cereals, potatoes, pulses and derived products’ group; proteins by the ‘meat, egg, fish’ group; calcium by the ‘milk and dairy products’ group, and so on. One of course knows that proteins can also be provided by dairy products or that fresh potatoes contain a lot of vitamin C, but this approach remains pragmatic and is not confusing with the current food groups. However, we chose to keep a specific category for composite dishes, whose allocation to a nutrient-based group may largely depend on the recipe. We included a category for sugary foods, which also includes soft drinks (with caloric or non-caloric sweeteners) but we did not consider alcoholic and energy drinks or water.
The proportion of nutrients supplied by different food categories, using data from the INCA surveyReference Volatier8. Total of percentages is not 100, because alcoholic drinks, energy drinks and water are not considered. This concerns less than 3% of most nutrients, except for minerals for which 10.8% are supplied by water
* The contribution of each food category to the lipid quality of the whole diet is estimated to be the highest value of the three fatty acid categories (saturated, monounsaturated and polyunsaturated), standardised by dividing by the sum of the maximum values obtained for each food category. A similar procedure is used to obtain a global weight for minerals and vitamins.
Defining food categories is a difficult and critical issue nowadays. It appears that the answer will hardly be a straightforward scientifically based solution, because numerous other aspects such as consumer perceptions, industrial innovations and current legislative definitions also come into play. It is not our objective to address this question; however, Nutrimap® is able to adapt to any kind of categorisation.
Weighting
This step aims to take into account the relative importance of the various nutritional criteria when estimating the nutritional value of a given product. This might seem of little use and unnecessarily complex, but one should keep in mind that not considering this issue means in fact that each nutrient is allocated the same weighting and impacts the nutritional quality equally, whatever the food category. This statement is not really scientifically justified and is not consistent with the generally acknowledged nutrient-based approach. We propose here to set the weighting allocated to each nutrient in a food category in relation to the amount of this nutrient in this category. The method consists of: (1) considering the proportion of each nutrient added to the diet by each of the seven categories defined above, which is illustrated in Table 2; (2) standardising the values obtained so that, within a food category, the most abundant nutrient is allocated a value of ‘100’ and the score allocated to each of the other nutrients remains proportional to its supply by this food category (Table 3); and (3) reducing the range to a scale of 1 to 3 (Table 3). This last step retains the relative levels between the weightings while restricting the scale and setting a minimum weighting of 1 for nutrients usually poorly supplied by products in the category yet which may appear in specific or newly developed food products, such as a butter-enriched vegetable purée or fibre-rich yoghurt.
Weighting coefficients allocated to each group of nutritional criteria according to the food category or moment of consumption. For the seven food categories, the values are obtained by standardising the figures shown in Table 2: the value of ‘100’ is allocated to the nutrient for which the category is the highest contributor, and then other nutrients are allocated proportional values (figure into brackets). The final weighting is established when the scale is reduced from 1 to 3
It is generally not easy to obtain reliable data about the amounts of nutrients supplied at different meals, thus making it difficult to use the same method to allocate weightings to nutrients when questioning the nutritional values of composite meals. The weightings proposed for children's breakfasts in Table 3 are based only on the expertise of paediatric nutritionists and can therefore be challenged.
Final calculations and mapping
Separate totals are then compiled for the positive scores (corresponding to the nutritional benefits of the product) and the negative scores (nutritional weaknesses). Both these scores will characterise the product and are not further aggregated. These scores are then mathematically standardised to a scale of 100 (100 corresponding to the theoretical maximum positive or negative score in the food group considered). A graph can then be plotted on which one can easily visualise the position of a given food product, both individually and in comparison to others belonging to the same food group.
Summary
To summarise, the following steps are used to evaluate the nutritional quality of an individual product:
1 Assignment of the food to a category. Nutrimap® does not define categories, but rather adapts to any given category definition.
2 Calculation of the 15 nutrient values, for 100 kcal of the product and according to available food composition tables or specific analyses.
3 Separate scoring of each nutrient, which is allocated a value between − 1 and +1, according to the mechanism detailed above and in Fig. 1. The thresholds for recommended and actual consumption depend on the country and the population group (see Table 1 for figures in a French context).
4 Weighting of each score, using a coefficient which differs from one food category to another, according to the contribution of the food category to the intake of this nutrient by the considered population (Tables 2 and 3).
5 Separate additions of negative and positive scores, and standardisation of the values on a 100-scale.
An example is detailed in Fig. 2. These five steps could, in theory, be carried out by hand, but software performs all the calculations quickly and easily. Instructions and detailed data are available from the corresponding author.
Step-by-step positioning and the resulting diagram of muesli, Swiss type
Comparison with other nutrient profiling systems
The best way to really validate a nutrient profiling system would probably be to demonstrate that the long-term preferential intake of foods which are positively ranked by the system is significantly associated with a lower incidence of nutrition-related diseases, or at least with positive changes to validated biomarkers. This is a very challenging objective requiring extensive and specifically designed studies, which were not undertaken here. However, we did make some attempts to compare the final position of some foods with the results provided by other systems and especially by the ones developed by the UK Food Standards Agency (FSA)Reference Rayner, Scarborough, Boxer and Stockley11 and by the Dutch Vovo system4. These systems classify foods as healthy/less healthy respectively with an ‘across the board’ approach or as preferable/medium course/exceptional with criteria depending on the category. For the purpose of this comparison, and from the Nutrimap® classification, threshold values for nutritional weaknesses and benefits have been arbitrarily determined to split foods into three categories: ‘healthy’, ‘intermediate’ and ‘less healthy’ (Table 4).
Results
The data used to assess the nutritional quality of foods within the French context of healthy adults come from the McCance & WiddowsonReference Paul and Southgate12 and CIQUAL13 food composition tables, from Eurodiet and the French nutritional recommendationsReference Martin14 and from the French food consumption surveyReference Volatier8 for the current nutrient intake information.
Example 1: Products in the same category
A mapping of products belonging to the carbohydrate-rich food category, which includes cereals, pulses, potatoes and derived products, is shown in Fig. 3, along with the final positive and negative scores. Figure 4 is the mapping of fruit and vegetables, with derived products, including juices. The mappings are highly discriminative, with lentils and muesli in the upper right part (more benefits than weaknesses) whereas wafers and potato crisps are in the lower left part (more weaknesses than benefits). If such a sophisticated tool does not seem necessary to distinguish between the nutritional quality of boiled potatoes and wafer biscuits or between tomato ketchup and green beans, it becomes more useful when comparing the overall nutritional quality of croissants and cookies, or spaghetti and oven potato chips.
Mapping of selected cereals, potatoes and derived products and table of nutritional assets and weaknesses
Mapping of selected fruits, vegetables and derived products and table of nutritional assets and weaknesses
Example 2: Meal and time of consumption
We have compared individual products regularly eaten for breakfast (Fig. 5) and whole breakfast meals composed of these products (Fig. 6). The nutritional composition of each breakfast is calculated by adding the contribution of each food relative to its amount and the calculations are then made considering 100 kcal of the meal as a whole. The corresponding weighting of criteria, with a specific focus given to lipid quality and quantity and to sugars, is shown in Table 4 and aims at being a translation of usual dietary recommendations for breakfasts, although these recommendations are not really precise: this mapping should thus be primarily regarded as an illustration of the ability of Nutrimap® to address whole meals as well as individual foods. Nutrimap® makes it easy to differentiate between different types of breakfasts and to evaluate the nutritional consequences of substituting orange juice with an orange (breakfasts 1 and 2), or bread and butter with a croissant (breakfasts 10 and 11). It is interesting to note that the distribution obtained by mapping the individual food items is no broader than the distribution obtained by mapping whole breakfasts. This is somewhat contrary to the generally accepted idea that a varied dietary supply is more balanced than a unique foodstuff, which should have resulted in a grouping of the various breakfasts in the central region of the map. Indeed, when examining the composition of these breakfasts, it can be seen that combining cereal-based, dairy-based and fruit-based products is not in itself sufficient to ensure the adequate balance of the resulting meal: the nutritional characteristics of each of the individual products remain key factors.
Mapping of selected products consumed at breakfast and table of nutritional assets and weaknesses
Mapping of composite breakfasts and table of nutritional assets and weaknesses
Comparison of the categorisation of 40 individual foods belonging to three different food groups by Nutrimap®, the UK Food Standards Agency (FSA) system and the Dutch tripartite system
UHT – ultra heat-treated; PUFA – polyunsaturated fatty acids. H – healthy/preferable; I – intermediate/medium course; LH – less healthy/exceptional; NE – not eligible (no criteria proposed by the Dutch system for this category of food). Discrepancy is mentioned by an ‘X’ when a food classified as healthy by Nutrimap® is not classified as such by each one of the other systems and when a food classified as healthy by these systems is not classified as such by Nutrimap®.
Comparison with other nutrient profiling systems
The comparison has been made for 98 foods, but Table 4 shows data for 40 food items only, belonging to three different categories, according to Nutrimap®, the FSA and Dutch tripartite systems. Although a straightforward comparison remains hazardous, it shows that it is possible with Nutrimap® to define thresholds of benefits and weaknesses which classify foods as more or less healthy. The resulting classification is highly consistent with the one provided by the FSA system, except for the fat group and composite food groups (not shown). Indeed, only 10 foods out of 81 (12%) are classified differently by the two systems when not considering fats and composite foods. This proportion reaches 21% of discrepancies for all the food products assayed. The discrepancies observed for fats can be explained by the lipid quantity criteria which cannot be adapted in the FSA scheme. Comparison with the Dutch tripartite system cannot be made on the same basis, since there is more possibility of discrepancies with three categories than two. However, if we focus only on food products which are ‘strongly misclassified’ (‘healthy’ instead of ‘less healthy’ and ‘less healthy’ instead of ‘healthy’), there are only eight misclassifications for 98 food products (8%). Although these comparisons are indicative only, they confirm that Nutrimap® has a comparable level of performance to tools promoted by official agencies. A case-by-case examination of discrepancies would provide additional insights.
Discussion
Nutrimap® is technically a very simple system. It comes as an Excel-type spreadsheet and can be used without any specific technical skill, except knowledge and expertise in nutrition and dietetics. It is a powerful and innovative tool which efficiently describes the nutritional quality of foods or meals by focusing on the concept of rebalancing non-optimal diets. For this purpose, it integrates dietary consumption data – which are not usually taken into account in such systems – and develops an original method for measuring the balancing potential of a food in relation to both nutritional recommendations and the reality of nutrient supplies. Nutrimap® considers both the benefits and weaknesses of each food, and keeps this duality until the final step; this positions foods clearly while avoiding a compensation of nutritional weaknesses by benefits.
Nutrimap® is probably one of the most objective nutrient profiling systems available: significant efforts have been made to base calculations on data that come from analyses (food composition), surveys (food consumption data) or strong quantitative scientific consensus (nutritional recommendations). These factual data drove the choice of nutrients, the nutrient scoring method and the weighting of nutrients within a food category. However, we should recognise that some decisions are not fully justified from a scientific point of view, such as the weighting of nutrients when addressing specific meals, or score corrections, or even choices made following recommendations that are not always based on very reliable scientific evidence. Although some of these flaws can be sorted out methodologically, it is likely that a subjective dimension will remain present in these tools. The important point here is to remain aware of what comes from reliable data and what comes from human expertise.
Nutrimap® is also extremely flexible: via simple and rapid changes to the scoring scale, and provided that the required data exist, it can adapt to various population groups (recommended intake levels), different geographic conditions (intake levels and nutrient weighting) or changing official dietary advice. One of the strongest advantages of Nutrimap® lies in this flexibility, which allows it to meet several goals while keeping a strong overall consistency because the principles and the methods remain unchanged. It can thus be used in a wide variety of contexts and for different purposes, bearing in mind that such a tool is intended only to help with decisions. Nutrimap® provides a positioning of foods (or meals) and decision-makers still have to set the limits for categorisation according to their own objectives. We have seen above that Nutrimap® can be of assistance for meal designers: by comparing the nutritional quality of various proposals, they can decide if the requested change in their habits and suppliers is worthwhile or not. Another potential use of Nutrimap® can be to assist in the development or revival of industrial food products. In most food companies, the nutritional consequences of changes in recipe are very seldom considered; a system such as Nutrimap® could be a very rapid and convenient way to simulate or monitor the effect on nutritional characteristics of the product of any change in the ingredient list, enabling nutrition to be integrated as a quantitative, and therefore easily measurable, item in industrial decisions.
Nutrimap® can also help public health authorities to take up consistent positions concerning legal limitations for some products, such as those appearing in television advertisements targeting children, or being available in vending machines, or bearing nutrition or health claims. The final decision about where the border lies would remain difficult yet would be founded on scientific and consistent bases. On Nutrimappings, it is possible to set limits for nutritional flaws (a horizontal line on the mapping) and qualities (vertical line) that can be more or less severe according to the final purpose.
Nutrimap® can of course be improved still further, for example by introducing n–3 fatty acids or by addressing more specifically the issue of drinks, which should probably not be considered as solid foods. Nutrimap® does not consider energy as a criterion; however, by introducing the quantities of lipids, sugars and carbohydrates, energy is duly taken into account. This is shown in Fig. 5, where the breakfasts are placed on the mapping diagonal from the least (upper right) to the most (lower left) energetic meals. Another limitation, which cannot be attributed to the system itself, is the existence and accuracy of nutritional composition data; this is the pragmatic reason for limiting the number of criteria to 15 nutrients in the usual version of Nutrimap®.
Although Nutrimap® will probably not end the controversy surrounding the intrinsic principles of nutrient profiling systems, the tool addresses some of the criticisms levelled at these methods. First, most of the decisions made when developing the model are justified, by referring to the soundest available data or consensus, thus making the system less dependent on subjective opinions. Second, it does not classify the food products strictly, which can rapidly lead to the concept of ‘good foods and bad foods’, but rather gives separate information on the nutritional benefits and weaknesses of the foods. Third, it can handle not only individual foods but also composite meals, and even whole diets, and so it can be a real help in improving dietary management, especially for people in charge of planning meals for canteens or restaurants.
Nutrimap® is a nutrient profiling system which considers each food in its own category and is consistent with food-based dietary guidelines that recommend consuming a given number of servings from each category each day. Nutrimap® is indeed the tool needed to complement these approaches, by giving information about the best choice that can be made within a food group; this seems to be a key issue in the worldwide challenge of fighting the dramatic increase in nutrition-linked pathologies.
Acknowledgements
Sources of funding: We thank the Danone group for its support during the first step of development of the tool.
Conflict of interest declaration: The Nutrimap® tool described here has been developed within the company BIO intelligence service. Each author received a salary from the enterprise and declares to have no conflict of interest.
Authorship responsibilities: The Nutrimap® tool has been developed and improved by E.L.V.A.-B. worked on the practical examples for specific food categories and prepared the manuscript. C.G. carried out the analyses with the software and designed the tables and figures. L.M. collected the nutritional information on products and gave advice on the manuscript.
