Why do nutrition databases give different values for the same food?

The food itself is different

"Bread" in a US database and "bread" in an Australian one were baked from different flour, to different recipes, under different food laws. The labels match and the loaves are different.

• Varieties and farming. Lamb, beef and dairy vary with breed and feed. Fruit and vegetables vary with variety, season and ripeness.
• Fortification. Australia and New Zealand require iodised salt in many commercially prepared breads, with specified exceptions. Fortification practices for breakfast cereals, plant-based milks and spreads can also differ between countries.
• Composite foods. A meat pie, a muffin or a "chicken curry" is a recipe, and every country's standard recipe is its own.
• Age of the data. Some entries were analysed recently and some decades ago. Foods change, and so do the lab methods. The publication or update date of a database entry does not necessarily show when the underlying food sample was analysed.

Carbohydrate, fibre and energy are defined differently

This catches out more practitioners than anything else. USDA FoodData Central commonly reports carbohydrate "by difference", which means whatever is left after water, protein, fat and ash are subtracted, so dietary fibre is inside it. Some datasets also provide carbohydrate by summation or other carbohydrate measures. UK and Australian and New Zealand data generally report available carbohydrate, with dietary fibre listed separately.

Consider a hypothetical food containing 20 g of available carbohydrate and 8 g of dietary fibre per 100 g. A carbohydrate measure that includes fibre might report approximately 28 g, whereas an available-carbohydrate measure would report 20 g. Actual differences can also reflect analytical methods and other components, so the figures are not always directly interchangeable.

Food composition databases generally estimate metabolizable energy using nutrient composition and appropriate energy conversion factors, rather than directly measuring the usable energy from every food. Databases may use general or food-specific Atwater factors, and may account differently for components such as fibre, polyols and organic acids. For example, CoFID documents factors for protein, fat, available carbohydrate and alcohol. These differences can move the energy of some foods, particularly high-fibre ones, by a few per cent.

Generic, branded and user-added data

Generic entries come from lab analysis of samples, or are calculated from other entries when a food wasn't analysed. Branded entries come from the manufacturer's label, which means rounded values and only the nutrients the label has to show. A branded entry with no vitamin C doesn't mean the food has none. It means nobody reported it.

User-added entries vary in quality. Some accurately reproduce product labels, while others may contain transcription errors, incomplete nutrient information or duplicate entries. Where possible, check the source and completeness of the data before using it for professional meal planning. For a comparison of the main sources, see best nutrition databases for meal planning.

Which one should you trust?

• Choose entries that best match the food's origin, brand, preparation and composition, using a relevant national database as a starting point. A client in Auckland eating New Zealand bread and milk is often well described by New Zealand data, but no national database is always the best fit.
• For packaged products, prioritise accurate, current product-specific nutrition information.
• Maintain a consistent approach to nutrient definitions and database selection, while allowing appropriate exceptions. Switching halfway can add noise that looks like a change in eating.
• When values disagree substantially, check the food's preparation state, edible portion, nutrient definitions, units and data source.

If an entry is simply wrong, don't fight it. Make your own corrected copy, as described in how to add a food that isn't in the database.