Why Nutrient Databases Disagree About The Same Food

Looking up the iron in a cup of lentils produces different answers depending on where the question is asked. The variation is real and has identifiable causes.

Databases report samples, not foods

A composition value is derived from laboratory analysis of specific purchased samples. Those samples came from particular farms, varieties, seasons and processing conditions.

The United States Department of Agriculture maintains the reference database most American sources draw from, built through sampling programs and contributed industry data.

Other databases derive from different national programs, from manufacturer submissions, or from older editions of the same source. Each traces to its own set of samples.

Growing conditions move the numbers

Mineral content in plants depends on soil composition, which varies regionally. Selenium is the starkest example, differing widely across American growing regions.

Variety matters as much. Different cultivars of the same vegetable can differ substantially in vitamin and mineral content, and databases usually average across whatever was sampled.

Ripeness at harvest, storage duration and transport all shift values further, particularly for the less stable vitamins. A single printed number conceals all of it.

Method choices affect the result

Analytical methods have improved over time, and older values were sometimes generated by techniques that measured a nutrient less specifically than current ones do.

Folate is a well-known case, since older methods and newer ones can report different amounts for the same food. Revisions in databases reflect method changes as much as food changes.

Protein values are typically calculated from measured nitrogen using a conversion factor, and which factor was applied changes the printed figure without any change in the food.

Cooked versus raw creates further divergence

Databases list many foods in both raw and cooked states, and cooked values may be measured directly or calculated from raw values with retention factors.

Water gain or loss during cooking changes concentration per gram, so comparing a raw and a cooked entry without accounting for that shift produces misleading conclusions.

Tracking apps compound the problem by mixing entries from multiple sources, including user-submitted ones that were never analyzed at all.

How to use the numbers anyway

Composition data is well suited to comparing foods and identifying strong sources. Lentils are a far better iron source than rice regardless of which database answers.

It is poorly suited to precise personal accounting, since the error on any individual figure is not small and absorption is not modeled at all.

Where a genuine question about adequacy exists, laboratory testing arranged through a physician measures the outcome directly, which no amount of database arithmetic substitutes for.