The CheckYourFood Group (CYFG) approach food composition data.
All food composition data on our platform is taken from the UK Quadram Institute, (formerly the UK Institute of Food research), publication – McCance and Widdowson, the US Department of Agriculture dataset and various other world food composition datasets (usually for very localised ingredients). The data is entered into the CheckYourFood Group (CYFG) content management system manually and then checked on an ongoing basis by CYFG staff and specific users.
Gaps in the data
All ‘N’ values in the UK data, which are given when a nutrient is present in an ingredient, but the amount has not been established, are replaced with data from another dataset usually the USDA. This gives a complete nutrient profile for accurate analysis and labeling.
When entering data from the USDA into an ‘N’ value gap, growing and feeding practices are taken into consideration – for example the selenium in US pork is often double that of UK pork due to the inclusion of selenium in the feed. Therefore a missing selenium value in a cut of pork from the UK data is replaced with 50% of the US selenium amount
Choline & Iodine
Choline figures are exclusive to the US dataset and are applied to UK ingredients to give a fuller nutritional picture. Similarly iodine figures are not present in the US dataset and are imputed into for example a US Pacific mackerel from the UK mackerel iodine data.
Disparities between data sources
Any glaring differences between datasets for the same nutrient in an ingredient is double checked with the Quadram Institute. An average of other datasets is then taken if the Institute feel their data is in error – see ‘Broad bean case study’ and USDA example below.
If entering ingredients using solely USDA data or filling in an ‘N’ value gap all carbohydrate figures are reset to the net figure (total carbs minus fibre), for accurate analysis and labeling.
Using the most accurate data
Where the phrase ‘Literature sources’ is cited in the UK data we use the USDA data for that particular ingredient as the strong likelihood is that this will confer greater accuracy for analysis
UK and US data relationship
We note that the UK dataset also references the USDA dataset for various ingredients – see bay leaf data in UK 13-806 and compare to US data 02004 and Thyme dried UK 13-883, US 02042. The datasets for these are identical with the UK citing literature sources, also the US data is for fresh bay which has been added to the UK data as dried bay leaf.
Vitamin A equivalents (RAE)
For Vitamin A equivalents (RAE) we use the more up to date US conversion factors of beta carotene at 12:1 and other carotenes @ 24:1 (double the figures used by the UK dataset giving half the final vitamin A figure entered into our dataset). This ensures that consumers are given the most accurate nutritional picture possible for personalised nutrition and claims and labeling.
In general we do not include the data for fortified foods, as from the information we have, we are unclear about the benefits/dangers of adding synthetic vitamins and ‘free’ minerals in high amounts to food and may give a confused nutritional profile to consumers/users.
Accessible ingredient naming
All ingredients are re-named to be easily recognisable and accessible both for consumers and food professionals. For example would a UK user know that the phrase ‘MATURE SEEDS’ in the US dataset refers to dried pulses.
For niacin (B3) figures CYFG use the preformed data amounts (as does the US dataset) as the tryptophan equivalent (60:1) only ‘may’ be converted to niacin and the requisite health benefits conferred. Again giving the consumer/user the fairest and clearest nutritional picture possible using established food composition data.
Data for cooked ingredients is only entered into the CYFG dataset if the data is from the actual analysis of a cooked ingredient, rather than data arrived at by a calculation from raw data.
The CYFG recipe analyser
All recipe data and some cooked ingredient data in the CYFG dataset has come from using the CYFG recipe analyser which factors in cooking losses by a combination of specific ingredient and the cooking method applied. This confers the greatest accuracy possible of the likely nutrient profile of cooked ingredients and meals.
Automatic weight change calculations
For nutrition and health claims accuracy CYFG have entered weight change calculations again by ingredient and cooking method, thus ‘automating’ the final nutrient amounts in 100g/portion of a cooked food product.
Broad bean case study
We noticed a big disparity between the pantothenic acid amounts in broad beans between the UK and US data which prompted the correspondence below.
Email 1 from the CheckYourFood group
I hope you are well.
This is Matt from checkyourfood with small question relating to the IFR composition data for broad beans specifically the pantothenic acid content of 4.94mg per 100g. The USDA data for this ingredient is 0.97mg for mature seeds and 0.086mg for immature seeds.
Is the reason for this difference because the IFR includes the skin around the bean (not the pod) and thus the pantothenic acid may be contained in this skin?
This is one of the only ingredients where I have found this margin of difference between the IFR and USDA datasets, (the others are the selenium in lentils and the folate in asparagus) and so I am looking for a reason, as your data for pantothenic acid is consistent between cooked and uncooked broad beans.
Any help you can offer would be greatly appreciated
Reply from the Quadram Institute
Nice to hear from you, hope things are going well.
I’ve looked into our broad bean data. Originally the pantothenic acid data originated in MW3 (1960) for raw broad beans (5.4mg). The cooked value of 3.8mg appears in the next edition (1978) and it looks as if it has been calculated from the raw value using a vitamin loss factor. As it is so old, we do not have any documentation to know exactly where it has come from. I’ve just had a look in some other datasets and the pantothenic acid data in those is as follows:
Japan – 0.39mg
France – 1.34mg
I think, based on the above, this particular UK value does not look correct and it would be better to estimate it from the USDA data. Hopefully broad beans can be included in future PHE analytical surveys.
Another example for the USDA dataset
The vitamin K for Amaranth leaves is very high in the USDA data = 1,140mcg per 100g with the proviso – “Taken from another source–other tables of food composition “
We double checked with the Indian composition dataset, (as the Amaranth leaf is native to India), which gives a figure of 280mcg per 100g which appears much more likely amount especially when comparing with similar green leaf content.