In this blog, Dr Maria Traka from the Food and Nutrition NBRI at the Quadram Institute and Professor Michelle Morris from the University of Leeds discuss the need to change the way we work to raise the quality of food composition data and realise the benefits of doing this for everyone across the food industry, as well as consumers.

What was the Food Composition Roundtable event?

It is not often that a range of stakeholders with a shared passion for a better food composition database get together with a sole purpose to think about food composition data.

On Friday 12 December, 27 of us convened at Caxton House in London, in an event hosted by the Quadram Institute.

With delegates from different parts of industry, government, academia, and the third sector we had wide representation, expertise, experience and passion for food composition data.

We spent some time setting the scene before discussing the value, opportunities and risks in doing things differently. Opportunities were highlighted through short talks and after a coffee break we reconvened into group activities before reporting back and thinking about next steps.

Where are we now?

Food composition data currently exists in silos, in a fragmented way and joining these diverse sources together is challenging and sometimes impossible.

We have the McCance and Widdowson’s Composition of Foods database (CoFID), which provides high quality composition data with over 3000 generic foods, each with detailed information on over 280 nutrients, including macro- and micronutrients. These data are open access and freely downloadable online.

CoFID aims to give an in-depth nutritional composition for foods that represent what people in the UK eat. However, with tens of thousands of foods on the supermarket shelves, the coverage of CoFID is limited.

Commercial data sources on food composition exist to fill these gaps and provide nutrient information on individual branded products that we typically buy. These all contain the eight nutrients that we can see on the back of a packet in supermarkets, mandated by labelling legislation, and micronutrient information for fortified products (micronutrients added).

Such primary data from labels can then be extended to fill in missing micronutrient values based on, primarily, CoFID, and to categorise foods, but without a standardisation framework.

Examples of primary and extended data sources include: NIQ Brandbank, Syndigo and myfood24, Nutritics. However, these commercial sources exist behind a paywall which means that not everyone can access them with costs often prohibitive. These companies may additionally have ingredient data, packaging information and derived metrics like Nutrient Profile Model (NPM) scores and High in Fat, Salt and Sugar (HFSS) status that are generated by teams within that company.

This means that there is no single source of truth for food composition data in the UK.

Different versions of food composition data mean that food composition data is not standardised. This can become a problem when being used to implement policy as it means that different versions of the data are being used by stakeholders during implementation and then in turn a potentially different version for evaluation and for enforcement. Policy enforcement is devolved to local authorities with small budgets who potentially can’t access the detailed commercial data at all.

Why do we need to work differently?

The food environment needs to change to support healthier lives. We are already seeing these changes happen, but more still needs to be done. To monitor this, trustworthy and consistent reporting or benchmarking against agreed and meaningful metrics is essential.

Quality food composition data underpins all metrics and benchmarking.

An example metric is the percentage of ‘healthy food and drink’ sold by supermarkets, restaurants and takeaways, all calculated using a Nutrient Profile Model. However, if the source of data used to calculate the Nutrient Profile Model is different for each company, differences in classification of ‘healthy food and drink’ can occur. When considering difference at a product level this might not seem so bad, but when we consider the millions of sales of that product over the course of a year, the problem is amplified and becomes huge.

Thinking back to the ‘why’ – the pressing need to support healthier lives – it is critical to have accurate data so that research can unlock the true relationships between food and health, not something observed as a result of low quality or inconsistent data.

Lack of standardisation in micronutrient gap filling can also have a huge impact in how we assess nutrient intakes in the UK, and has a risk of eroding public trust in digital and AI solutions for dietary recommendations and personalised health.

This is not only a problem for health research or policy evaluation, it can additionally be a problem for businesses. If companies all use different data sources and there are inconsistencies between these data, then they are not operating on a level playing field. Considering that these data underpin metrics that inform business decisions and are reported to investors and other stakeholders this could be misleading, at best. There could also be economic impacts.

Conversely, quality food composition data and consistent reporting metrics could be an enabler of economic development and productivity across the food system. Better data can provide better customer experience, enabling easy access to a variety of information about the food they are buying, which is then better for businesses too.

At the moment we could miss the real impact of innovation to enhance the nutrient content of our home-grown foods and crops.

Within CoFID (which is used by many companies) the nutrient composition of a tomato is the same for all tomatoes, but in reality they are not all equal. Another example that we see often on our supermarket shelves are high Vitamin D mushrooms. These are not captured in CoFID as they were not popular at the time of the database creation. But this is also true for ingredients under development, like high fibre white flour.

If the same food composition data is used for all products, then true nutrient benefit that we could potentially gain by different agricultural innovations and practices is not captured, and we are not able to measure the true impact of making better choices within a type of food. It is therefore not possible to evaluate the return on investment of these innovations – tracking all the way from the farm to the impact on human health.

What is needed to fix this problem?

There is always a cost to doing things differently. However, the cost of not making a change is also expensive. As things stand every business that uses food composition data have to develop their own bespoke solution and approach which is expensive and when multiplied across the food industry and to other sectors requiring this data (national and local government, charities and research institutions), sums to tens of millions very quickly and potentially billions in the medium term.

We need to work more efficiently.

Bring together high-quality data when they exist and step-up the quality of other data when they don’t. The former requires agreeing a common ‘language’ for data connectivity. In the data science world this is termed ‘interoperability’’ and consists of well-defined ontologies, data models and terminologies for foods and their related food composition data, which is currently lacking or not up to par.

The latter requires data standardisation frameworks agreed and implemented either at source (when data are captured from labels or generated through lab analysis) or when the data are transformed for policy implementation.

Finally, we need primary data, underpinning further data transformations, to be current and their updates to target the gaps that will amplify and raise the quality of the extended food composition data for all.

What did our roundtable event cover?

Judith Batchelar OBE welcomed delegates to Caxton house and set out our mission. Dr Maria Traka took over setting the scene with detailed examples from McCance and Widdowson’s CoFID and beyond.

Professor Michelle Morris ran an interactive session to explore the perceived value, opportunities and risks in doing things differently with questions such as: What do you see at the main opportunities for a centralised food composition database? And what are the risks of doing nothing, maintaining status quo?

In response to these questions, consistency and transparency were key opportunities roundtable members identified in centralising food composition data. While fragmentation was seen as the main risk of doing nothing, along with other risks such as inconsistence data and lack of relevance.

A series of short talks highlighting opportunities followed from:

• Ann Godfrey (GS1) talking about QR codes, the Digital Product Passport and the future potential of carrying substantial nutritional information on product labels
• Matthew Gilmour (Quadram) talking about the Food-Microbiology Intelligence Network and a paradigm for sharing sensitive microbiology safety data for public/private benefit
• Michelle Morris (University of Leeds) with an example using Supermarket sales data from four major retailers to evaluate the impact of national legislation

Fuelled by tea and coffee, we finished the day with a series of working groups discussing:

• Database content, including quality and scope
• Technical requirements from a data and digital perspective
• Membership models, including the financial/business model

We closed out the morning agreeing some key next steps, before sharing lunch. We are now working on a full report form the meeting, but for now we will leave you with our key take home messages.

Take home messages

• There is a shared will to work differently to improve the quality, coverage, consistency and access to world leading food composition data in the UK
• There is a cost to maintaining the status quo and to changing the way we work – but the costs from changing will improve data quality and benefit all stakeholders considerably in the longer term.
• We need more primary data, we need standardisation of primary data transformation, and we need better access to the data
• Strong leadership and governance are essential
• Clear uses cases are essential
• We can learn from other sectors – this has been achieved before
• With more comprehensive data consumers can make more informed choices and debunk misinformation (e.g. via a QR code on foods)
• We will continue this journey together