article

Should we trust nutrition calculators?

by  Dr Rupy Aujla27 Oct 2022

We recently added a nutrition calculator to the Doctor's Kitchen app to automatically calculate the nutrient content of our recipes. Nutritional values are widely used to guide food choices. But they are inherently inaccurate. So why are nutrition calculators flawed and how can we use them despite their limitations?

This week, we released a new feature in the doctor’s kitchen app - our nutrition calculator and a new visual representation of daily value percentages to learn more about your nutrient intake. If you use the app, you will notice a new design for the nutrition information under each recipe. More than that, the nutrient values of our recipes – protein, carbohydrates, fat, calories, fibre, sugar and micronutrients – slightly changed. That’s because they are now based on our own food dataset.

If you are not an app user, this article might still be of interest. Curating a food dataset with nutrition values for over 700 ingredients raised interesting questions about nutrition facts in the wider sense. There are numerous recipe calculators online used by millions of people to find the nutrient content of meals. However, for the same meal, they often calculate different nutrition values. And ours will too. So why is nutrition information mostly wrong and how should we use it?

Building our nutrition calculator


The nutrition information displayed in the app before this week was sourced from external online calculators such as Whisk and MyFitnessPal. We would manually input each value for every recipe in the app. Other than being time-consuming, this process relied on data we knew little about. So we decided to curate our food dataset and build a nutrition calculator in the app to:

(1) Automatically and rapidly calculate nutrition values for each recipe added to the app, and

(2) Improve the accuracy, reliability and transparency of the data we use

How our nutrition calculator works

Nutrition calculators work by breaking down the amounts and nutrient values for each ingredient in a recipe and calculating the total nutrient values, including how much macronutrients, minerals and vitamins are in a meal. It requires two elements (1) quality nutrient data for each ingredient and (2) an appropriate calculation method.

Our first step was to methodologically curate a food dataset collecting nutrient values for the 600+ ingredients used in the Doctor’s Kitchen recipes. Then, our development team built the calculation program in the backend of the app.

We started with the goal of quickly creating a high-quality food composition dataset to accurately calculate the nutrient content of our recipes. Early on, we faced huge variability in the nutrient values of the same food, leading to the realisation that nutrient data is inherently inaccurate and limited.

The issue with nutrition values


Nutrient data is detailed information about the composition of foods typically expressed per 100g. It includes:

  • Macronutrients – energy, fat, carbohydrate, sugar, fibre and protein
  • Micronutrients – vitamins and minerals
  • Other compounds like phytochemicals and antinutrients

These nutrient values per 100 grams of food are used to calculate the overall nutrient content of meals and inform major areas of nutrition. However, nutrient values can vary significantly for the same food depending on where the information comes from. For example, raw fennel can be attributed 1.8, 3.3 or 7.3 grams of carbohydrates per 100g depending on the food database used. Processed foods like sauces and cereals show even larger variations depending on the brand chosen.

1/ There is no single source of truth for nutrient values

Nutrition data is provided by a variety of sources including commercial nutrition apps & websites, national organisations, supermarkets and brands. There is huge variability between sources in how complete, up-to-date and rigorous the data provided is. This makes it difficult to determine a global source of truth for nutrient values, even for whole foods. A good source should typically provide complete and detailed data documentation, have the expertise required and apply international standards. However, data provided by national organisations like the U.S Department of Agriculture can be outdated and incomplete.

2/ Data is subjective – it depends on who needs it

Collecting nutrient data requires constantly making choices about what data is good enough to be included. When we search for whole foods like potatoes or branded products like granola, we need to choose between dozens to thousands of results across different databases. This means the data we select is reflective of our priorities and judgment. So, nutrition data is at least partially subjective – apps and websites control which data they select to best meet their needs and objectives. Scaled up to thousands of recipes, the nutrition information provided by calculators is variable and subjective.

But why are there so many differences between different data sources? Should we not have a single nutrient value for foods, especially whole foods like fruits and vegetables?

3/ Food is not just “food” – nutrient content varies

The nutrient composition of food changes depending on a variety of factors. So the food on your plate is not the same as the food analysed to determine nutrient values. No two foods are the same – they differ based on:

  • How the food was grown including plant variety, soil, farming practices and climate
  • Manufacturing practices like product formulation and processing methods
  • Food policies like fortification and preparation methods
  • Cooking and recipe preparation – cooking influences the water, fat and nutrient content of foods

4/ Different methods are used to calculate the nutrient content of foods

In addition to variations in food composition, the way nutrient values are derived also varies. Food composition databases are created using different methodologies, including:

  • Direct chemical analyses
  • Derived from scientific literature and reports from laboratories
  • Calculated values – for cooked foods, calculated based on preparation factors such as loss or gains of weight and estimate micronutrient changes
  • Estimated values – derived from analytical data available for similar foods
  • Borrowed values – taken from other food composition databases

Even if the source is reputable and reliable, nutrient values can be approximations based on averages of similar products. Food companies and nutrition databases are not required to verify every product and are allowed a 20% margin of error. So 150 calories actually means between 120 and 180 calories.

5/ We all respond differently to food

Besides variations in the analysis methods and the food itself, the way our body responds to food varies depending on our unique biology and lifestyle. A series of studies showed that nutrients in the lab are not the same as how they are absorbed by the body, which means that calorie labelling may not reflect the number of calories absorbed. (Baer et al. 2019)

👉 Nutritional values are imprecise and vary widely depending on:

  • The source of information and method used
  • The food itself – where and how it was grown or manufactured
  • How we prepare the food
  • Our unique body and physiology

A single value cannot encompass all the variables attached to food. So nutrition calculators are an estimate, not a measure and should not be relied upon with 100% certainty. The nutritional information attributed to a meal is a rough guideline of what you are eating and how your body will respond to it.

However, variability does not condemn nutrition calculations as useless.

Nutrient values can be useful despite their limitations


In 1940 was published the first edition of a piece of work called The Chemical Composition of Foods, which detailed the nutritional content of thousands of foods. Still considered a standard reference, the authors emphasised that “a knowledge of the chemical composition of foods is the first essential in the dietary treatment of disease or in any quantitative study of human nutrition”. (McCance & Widdowson 1940)

Nutrient values, despite their limitations, remain essential guidelines. For individuals, information about the nutrient content of a recipe highlights key nutrients in food that can support health. It can help you develop a better understanding of food, what makes a meal healthy, and how to diversify your weekly intake to meet your daily needs. Nutrition information guides a personalised approach to eating where you can support your individual needs and increase the diversity of nutrients you get in a week.

For the doctor’s kitchen team, nutrition information guides recipe creation and whether we are including the right building block ingredients to create nutritious meals that fit your health goals and provide a healthy range of nutrients.

In the wider sense, food composition data is the basis for important areas of nutrition including food labelling, nutrient requirements, consumer information, dietary guidelines and nutrition research investigating the links between nutrient intake and health.

👉 Nutrient values help us better understand food to create balanced nutritious meals that support our health. At the wider level, they are the basis for major areas of nutrition like dietary guidelines and nutrition research. Quality data is important to ensure we do not introduce mistakes in our understanding of foods and the nutritional values of meals.

Better food data – How we’re working to improve the reliability of our nutrition information


When we started selecting nutrient values, we were rapidly faced with the huge variability between different sources, preparation methods, food composition and brands. We are continuously working to consolidate a trusted set of food data to provide growing value on health and nutrition.

How can we improve the reliability of nutrition information?

Because food data is inherently inaccurate, we believe better data starts with:

1/ Transparency about how the data is selected, used and its limitations,

2/ Continuous testing and improvements

3/ Open data freely available to all

We cannot pretend our data is optimum. Instead, we are choosing to remain critical of food data in general and we are open to discussion and improvements.

👉 Download the Doctor’s Kitchen app for free here and let us know what you think

Our sources

We compiled our food dataset using a combination of sources, including:

  • U.S Department of Agriculture (USDA) – Food Data Central
  • Public Health England – the Composition of Foods Integrated Dataset (CoFID)
  • UK product labels

Using these three sources, we created the Doctor’s Kitchen food Dataset which provides nutrition information on the macronutrient, vitamin and mineral content of the 700+ foods used in the Doctor’s Kitchen recipe app. We are on the lookout for other sources to improve and complete our data.

How we selected nutrient data

We manually selected nutrient data for each food aiming for the highest quality possible, based on 4 principles:

1/ Precooked raw values

We are using values for raw foods to account for differences in cooking times that can result in different weights. This principle excludes food items that are specifically cooked like pre-cooked rice or typically bought cooked in the UK like crab meat.

2/ Similarity to UK products

We compared a large proportion of our data with products found in common UK supermarkets, choosing the healthier option with lower sugar and sodium content. This principle ensures that our data reflects ingredients app users are most likely to cook with.

3/ Most complete data, including micronutrient values

Some sources only provide macronutrient values for a given food which only partially represent its contribution to a meal. Where available, we chose the most complete data with no or few missing values to appropriately assess nutrient intake.

4/ Most recent update

Some food data was analysed in the 1980s with no recent update, which increases the chance of inaccurate data. Where available, we chose the most recently updated data.

5/ Consistency across the dataset

Using multiple sources may reduce consistency. Therefore, when choosing data, we compare food by categories – grains, pasta, root vegetables, sauces, etc – and aim for consistency.

We are in the process of actualising our data. You can find our current detailed sources here.

Data evaluation and improvements

We are reviewing and improving our data every few months by comparing our values to other sources, online calculators, and common UK products and seeking better ways to manage our food data.

Where we’re heading

Exploring the field of food composition data expands to other areas that concern you directly. Some ideas on where we might be heading next:

  • Improving the quality and reliability of the data we use for recipe calculations
  • Adding more complexity to food data – phytochemicals, plant points, food matrix, blood glucose response, culture, enjoyment
  • Comparing food items and guiding more informed food purchases
  • Supporting a personalised approach to nutrition for health and wellbeing
  • Developing more integrative ways to think about food and its effects on the body

👉 Download the Doctor’s Kitchen app for free here

References


I was inspired by the perspectives of:

What Makes a Food Composition Database (FCDB)?

How Do We Go From Food Data to FCDBs?

  • Research papers including:

Maintaining and updating food composition datasets for multiple users and novel technologies: Current challenges from a UK perspective – published in 2020 Nutrition Bulletin by Traka and colleagues – DOI: https://doi.org/10.1111/nbu.12433

The unmapped chemical complexity of our diet – published in 2020 in Nature Food by Barabási and colleagues – DOI: https://doi.org/10.1038/s43016-019-0005-1

  • Food composition data documentation – The Composition of Foods Integrated Dataset 2021: User guide
by Dr Rupy Aujla
Print
Share:

SUBSCRIBE

Free 7 day meal plan

Sign up to our newsletter and receive a free copy of our breakfast, lunch and dinner plan to kick start your healthy eating.