Food composition data
Despite their widespread use and the wealth of information they provide, food composition databases have their limitations. Recognition of the potential sources of error in food composition values is important. Errors may be random or systematic and all such errors add additional uncertainty to the calculated nutrient intakes (Gibson, 2005). Limitations in the use of food composition databases can be summarised as follows:
- Bioavailability: Food composition values are figures representing the total amount of the constituent in the food, rather than the amount actually absorbed in the body. As a consequence, the potential bioavailability of nutrients in the local diet should be considered when nutrient intake data are assessed.
- Natural variability in the composition of foods: Foods exhibit natural variation in the amount of nutrients they contain. Nutrient concentration can be affected by different methods of plant husbandry, storage, soil, weather and ultimately transport and storage conditions. For example, in contrast to many European countries, the selenium content of foods is higher in the United States and Canada due to the higher selenium levels in the soil and levels are particularly low in New Zealand.
- The natural nutrient variation also varies within food type:
Meats - The nutrient composition of meat products can vary greatly depending on the proportion of lean to fat tissue. The ratio between the two also affects levels of most other nutrients.
Fruit & vegetables - Storage conditions can affect the water content of plant foods. Changes in water content are associated with changes in all other constituents mainly as a result of changes in nutrient density. Trace elements in particular are affected by husbandry conditions, soil composition and fertiliser use.
Cereal - The natural variation in nutrient content of flours and grains is less than that in fruits and vegetables. However, as with other foods, fertiliser and soil type produce some variation in mineral content. Different fortification practices in some countries markedly affect micronutrient levels such as B vitamins, folate, iron and calcium. For example, in the US mandatory fortification of grain with folic acid was introduced in 1998. In Canada, mandatory fortification of white flour and pasta has been carried out since 1998. In the UK, agreement has yet to be fully reached as to whether flour should be fortified with folic acid (UK folic acid fortification).
- Processed foods and composite dishes: The nutritional content of processed foods and composite dishes varies greatly. Processed foods change in formulation and production and the introduction of reduced-energy, saturated fat and salt versions of standard foods mean that databases need to be continuously updated. Common foods such as mayonnaise, muesli and sausages vary by brand and many supermarkets have introduced their own brands with different formulations and fortification levels. Composite dishes show great variation mostly due to differences in recipes and actual cooking method.
- Limited coverage of food items: It is unlikely that a database can be comprehensive for more than a short period. New foods are constantly introduced to the market and although modern databases can hold information on a large number of different foods, only a limited number of foods can be practically included in a database.
- Coverage of nutrients: Ideally, food composition databases should include complete data on all nutrients known or thought to be important to human nutrition. However, this can rarely be achieved. Factors such as the availability of reliable analytical methods, the availability of existing data, health concerns in a given country (and hence priorities given to certain nutrients) as well as national and international labelling regulations are all determinants of the coverage of nutrients in a database.
- Labelling: It should be taken into account that the values declared on food labels include a tolerable margin (Food labelling). As a consequence, using values obtained from food packaging may add a degree of error to the food composition databases. Furthermore, some countries have different guidelines. In the UK the acceptable level of tolerance permitted for macronutrients decreases as the level of the macronutrient in a product increases (Food Label Tolerance Levels).
- Others: It should be noted that there are a number of variations in terminology and analysis between countries which can result in different values for nutrients for the same food and hence total intakes. The most notable of these are carbohydrate and dietary fibre.
- Carbohydrate – In many countries, including the US and most countries in Europe, carbohydrate is determined ‘by difference’ as the remainder after subtraction of the values for protein, fat, water and ash from the weight of a food. In other countries, such as the UK, carbohydrate is the sum of the components of carbohydrate subtypes, namely starch and sugars, added together. These two approaches give different values for many foods particularly those containing dietary fibre and more complex types of starch. Moreover, the energy values applied to carbohydrate also varies, in most countries being a value of 4 kcal/g, whereas in the UK the value applied is the monosaccharide equivalent of 3.75kcal/g. Hence the same value of the carbohydrate content of a food will have a different energy value, and hence diets will have different energy values from countries using one method versus the other. These differences are often not appreciated in international comparisons.
- Dietary fibre – is now analysed largely by the Association of Official Analytical Chemists (AOAC) (Prosky) gravimetric method to give a value for Total Dietary Fibre (TDF) whereas in the UK it is measured chemically as non-starch polysaccharide (NSP) according to the method of Englyst. These give different values for most foods, those for TDF being generally higher than NSP. The fibre method used should always be specified in publications reporting dietary intakes.
Summary
Accurate and reliable estimation of nutritional intake is key to nutritional epidemiology. Food composition data provide an approximation of the energy and nutrient composition of foods. Calculating nutrient intake involves matching as closely as possible the foods and drinks recorded in a diet record with the items included in the database of the dietary analysis program. This can be a time consuming and lengthy process and is not always achievable without some error. Although modern databases contain information on a large number of foods it is not possible to maintain a database that incorporates all the different foods consumed by a population. Vast changes in the food supply add to the difficulties encountered with estimating calculated nutrient intakes. The range of ‘ready meals’ has increased and new foods such as low fat, and reduced sugar items have entered the market place as have consumables fortified with additional nutrients. Data entry programs that allow new foods, and portion sizes to be added to the existing database are essential if research into diet-disease relationships is to be carried out accurately and reliably. Figure 1 highlights some of the factors influencing the accuracy of nutrient intake estimation. Even though nutrition analysis software is designed to reduce reliance on nutritional knowledge and efforts are taken to standardize coding, skilled diet coders will still be required to make judgement decisions when coding diet records. Moreover, although values in food composition databases can be updated, it should be remembered that food composition data are, by their very nature, quite variable.