AI reveals how much hidden sugar is in packaged foods worldwide

AI reveals how much hidden sugar is in packaged foods worldwide

A groundbreaking algorithm reveals how much hidden sugar is lurking in your foods - and shows which countries and products hit the mark for healthy carbs.

Carbohydrates contribute approximately 70% of daily energy intake in the average human diet worldwide. However, the importance of carbohydrate quality is often overshadowed by their quantity. In a study recently published in the journalLimits in nutritionA European research team developed an algorithm to predict free sugar content in packaged foods, providing insights into carbohydrate quality on a global scale.

Carbohydrates in the diet

Carbohydrates are an important source of energy and play a crucial role in global nutrition. While discussions about nutrition often focus on the quantity of carbohydrates, the quality of carbohydrates is equally important for maintaining good health. Scientific evidence shows that the quality of carbohydrates influences metabolic function and the risk of chronic diseases.

One tool for assessing carbohydrate quality is the carbohydrate quality ratio (CQR), which evaluates the balance of total carbohydrates, fiber, and free sugars in foods. This ratio gives at least 1 gram of dietary fiber per 10 grams of total carbohydrates and no more than 2 grams of free sugars per 1 gram of fiber. This ratio helps distinguish nutritional foods from those that may contribute to poor health outcomes.

However, accurately determining the free sugar content in packaged foods remains a challenge. Only a few countries require explicit labeling of additional sugars, limiting transparency for consumers and researchers. Free sugars defined by the World Health Organization (WHO) include added sugars as well as naturally occurring sugars in honey, syrups and fruit juices, while the FDA defines added sugars as only those introduced during processing. This lack of information hinders efforts to effectively assess carbohydrate quality, making it difficult to make informed dietary decisions and study the effects of carbohydrate consumption on health.

About the study

The algorithm prioritized the ingredients listed first on labels because food manufacturers often order ingredients by weight, providing clues about sugar dominance in a product's composition.

In the present study, researchers developed an algorithm to predict free sugars in packaged foods worldwide, addressing a critical knowledge gap in carbohydrate quality. They used data from the Mintel Global New Products Database (GNPD), which contains extensive information on packaged foods from 86 countries, including nutritional composition and ingredient lists.

Before analysis, the team meticulously cleaned and standardized the data to ensure consistency. A crucial step involved manually curating and labeling ingredients with regular expressions to classify them as added or naturally occurring sugars - a distinction that was essential to accurately estimating free sugar content.

To build predictive models, researchers used machine learning techniques. They trained their models using data from the United States (US) and formally tested their performance in 14 selected countries, while applying the models to products from 81 additional countries. The models analyzed product labels taking into account the first six ingredients classified as added sugars, fruits or dairy, as well as detailed nutritional information such as energy content, fats, carbohydrates, fiber, protein, sugar and sodium.

The pipeline included three binary classifiers to detect the presence of additional sugars and tree-based stacked regression models to estimate their quantity. In addition, predicted added sugar levels were used as estimates of free sugars, with the exception of certain food categories such as juice drinks and sugar confections, where total sugars were used directly due to their unique sugar profiles.

Finally, the models were applied to products without explicit additional sugar explanations to predict carbohydrate composition. Carbohydrate quality was assessed using a predefined 10:1 to 1:2 ratio of carbohydrates, fiber, and free sugars.

Key Findings

Plant-based milk alternatives (e.g., oat or almond drinks) demonstrated surprisingly high compliance with carbohydrate quality standards, outperforming many dairy-based products worldwide.

The study found that the machine learning models demonstrated a high level of accuracy in predicting free sugar content in packaged foods. The mean absolute error for the test set was calculated to be 0.96 g/100 g, indicating a relatively small average difference between the predicted and declared values.

In addition, the model achieved a high R² of 0.98 between predicted and declared values, outperforming previous models such as K-Nearest Neighbors, which had a much higher error rate, confirming the reliability of the predictions. Notably, the model's predictive capabilities were not limited to the United States. The researchers found that the model performed accurately when formally tested in 14 countries and applied in an additional 81 countries, highlighting its global applicability.

The study also examined the proportion of food products that met the quality ratio of target carbohydrates and showed significant variation across both foods and food countries. In the United States, products hitting the carbohydrate quality ratio varied significantly, ranging from a relatively high 60% for hot cereals to 0% for flavored milks and malt beverages. This wide range demonstrated the diversity of carbohydrate quality even in a single country.

Chocolate-flavored products (like cereal or snack bars) were among the worst offenders, with 95% failing to meet the target ratio due to excessive free sugars and low fiber.

When looking at all food categories, the percentage of products meeting the target ratio ranged from 67% in the UK, representing a relatively high level of compliance to the quality standard, to 9.8% in Malaysia, indicating a significantly lower proportion of products meeting the desired carbohydrate quality.

Notably, plant-based beverages—unlike most beverage categories—induced relatively high adherence to carbohydrate quality ratios across countries due to their higher fiber content and lower sugar content.

However, the researchers acknowledged that the accuracy of the predictions for specific countries may be limited to some extent by small sample sizes, which could potentially affect the generalizability of the results to these specific regions.

In addition, the authors conducted Z-tests comparing predicted and declared free sugar levels across 18 food categories in the United States and found statistically significant differences, confirming the robustness of the model.

Diploma

In summary, the study successfully developed and validated a machine learning-based method for predicting free sugar content in packaged foods using a large-scale global database. This fully automated and scalable approach demonstrated strong accuracy across countries and food categories and can be extended to other databases and nutrient metrics that require free sugar estimates.

The predicted free sugar levels could also improve nutrient profiling systems such as Nutri-Score, which currently rely on total sugars due to limited labeling requirements.

This innovative methodological approach provided a valuable and powerful tool for monitoring and assessing carbohydrate quality in the global food supply and provided important insights for public health and nutritional advice initiatives.

Sources: