An innovative algorithm exposes how much hidden sugar is hidden in your food – and shows which countries and products meet the signal for healthy carbohydrates.
Study: Providing the quality of carbohydrates in a global database of packed food. Credit Picture: New Africa / Shutterstock
Carbohydrates contribute about 70% of daily energy intake in the middle human diet worldwide. However, the importance of carbohydrate quality is often overshadowed by its quantity. In a recent study published in the magazine Borders in dietA European research team has developed an algorithm to predict the free sugar content in packaged foods, providing information on the quality of carbohydrates worldwide.
Carbohydrates in diet
Carbohydrates are a vital source of energy and play a decisive role in the world diet. While diet discussions often focus on carbohydrates, carbohydrate quality is equally essential for maintaining good health. Scientific evidence shows that the quality of carbohydrates affects metabolic function and the risk of chronic diseases.
A tool used to assess the quality of carbohydrates is the ratio of carbohydrate quality (CQR), which evaluates the balance of total carbohydrates, dietary fibers and free sugars in food. This ratio determines at least 1 gram of dietary fibers per 10 grams of total carbohydrates and not more than 2 grams of free sugars per 1 gram of fiber. This ratio contributes to the distinction of nutritional beneficial foods from those that can contribute to the bad health results.
However, the determination of the sugar content accurately in packaged foods remains a challenge. Few countries require an explicit labeling of added sugars, limiting transparency for consumers and researchers. Free sugars, as defined by the World Health Organization (WHO), include additional sugars as well as natural sugars in honey, syrups and fruit juices, while the FDA defines added sugars such as those introduced during processing. This lack of information prevents efforts to effectively evaluate the quality of carbohydrates, making it difficult to carry out documented dietary choices and studying the impact of eating carbohydrates on health.
For the study
In the present study, the researchers developed an algorithm to predict free sugars in packaged foods worldwide, facing a critical gap in carbohydrate quality. They used data from the MINTEL GLOBAL NEW PRODUCTS (GNPD) database, which contains extensive information on packaged foods from 86 countries, including the synthesis of nutrients and lists of ingredients.
Before the analysis, the team cleans meticulously and standardized the data to ensure consistency. A crucial step involved manual care and labeling of the ingredients using regular expressions to classify them as added or natural sugars – a distinction necessary to accurately assess free sugar content.
To build prognostic models, the researchers used mechanical learning techniques. They were trained their models using data from the United States (USA) and officially examined their performance in 14 selected countries, applying models to products from 81 additional countries. The models analyzed product labels, taking into account the first six ingredients categorized as additional sugars, fruits or dairy products, along with detailed nutritional information, such as energy content, fats, carbohydrates, fibers, proteins, sugars.
The pipeline included three binary classifiers to detect the presence of added sugars and stacked trees -based reflux models to estimate their quantity. In addition, the predicted added sugar values were used as free sugar estimates, in addition to specific food categories, such as juices and pastry drinks, where total sugars were used directly due to the unique sugar profiles.
Finally, the models were applied to products without explicit added sugar statements to predict carbohydrate synthesis. The quality of carbohydrates was evaluated using a predetermined ratio of carbohydrates, fibers and free sugars.
Basic findings
The study found that mechanical learning models showed a high degree of accuracy in predicting the free sugar content of packaged food products. The average absolute error for the test set was estimated to be 0.96 g/100g, indicating a relatively small average difference between the predicted and stated values.
In addition, the model achieved a high R2 0.98 among the predicted and stated prices and exceeded previous models, such as the neighbors K, who showed a much higher error rate, confirming the reliability of forecasts. Specifically, the possibilities of predicting the model were not limited to the US. Researchers found that the model accurately performed when it was officially tested in 14 countries and implemented in additional 81 countries, underlining its worldwide possibility.
The study also examined the percentage of food products that met the quality of target carbohydrates, revealing significant fluctuations in both food and countries. In the US, products that meet the quality ratio of carbohydrates vary significantly, ranging from relatively high 60% for hot cereals in a particularly low 0% for flavored milk and malt drinks. This wide range emphasized the diversity in carbohydrate quality even in a single country.
When examining all food categories, the percentage of target products ranges from 67% in the United Kingdom, representing a relatively high quality attachment to the quality standard, at 9.8% in Malaysia, indicating a significantly lower percentage of products that meet the desired quality carbohydrate.
Specifically, vegetable drinks-contrary to most categories of drinks-have led to a relatively high attachment to the quality of carbohydrates among countries, due to the higher fiber content and lower levels of sugar.
However, the researchers acknowledged that the accuracy of forecasts for some countries may be limited to some extent by small sample sizes, which could possibly affect the generality of the findings for these specific areas.
In addition, the authors made z-tests that compare the intended and declared free sugar prices in 18 food categories in the US and did not find statistically significant differences, confirming the model of the model.
Conclusion
In summary, the study successfully developed and endorsed a mechanical learning method to predict the free sugar content in packaged foods using a large -scale global database. This fully automated and gradual approach has shown strong accuracy between countries and food categories and can be expanded to other databases and nutrient measurements that require free sugar assessment.
Sugar-free free prices could also boost nutrient profile systems such as Nutri-Score, which are currently based on total sugars due to limited labeling requirements.
This innovative methodological approach provided a valuable and powerful tool for monitoring and evaluating the quality of carbohydrates to the global food supply, offering critical knowledge of public health initiatives and nutritional guidance.
