Explore the vast range of titles available.

To assess time trends in the contribution of processed foods to food purchases made by Brazilian households and to explore the potential impact on the overall quality of the diet. Application of a new classification of foodstuffs based on extent and purpose of food processing to data collected by comparable probabilistic household budget surveys. The classification assigns foodstuffs to the following groups: unprocessed/minimally processed foods (Group 1); processed culinary ingredients (Group 2); or ultra-processed ready-to-eat or ready-to-heat food products (Group 3). Eleven metropolitan areas of Brazil. Households; n 13,611 in 1987-8, n 16,014 in 1995-5 and n 13,848 in 2002-3. Over the last three decades, the household consumption of Group 1 and Group 2 foods has been steadily replaced by consumption of Group 3 ultra-processed food products, both overall and in lower- and upper-income groups. In the 2002-3 survey, Group 3 items represented more than one-quarter of total energy (more than one-third for higher-income households). The overall nutrient profile of Group 3 items, compared with that of Group 1 and Group 2 items, revealed more added sugar, more saturated fat, more sodium, less fibre and much higher energy density. The high energy density and the unfavourable nutrition profiling of Group 3 food products, and also their potential harmful effects on eating and drinking behaviours, indicate that governments and health authorities should use all possible methods, including legislation and statutory regulation, to halt and reverse the replacement of minimally processed foods and processed culinary ingredients by ultra-processed food products.

To investigate the concordance between Australian government guidelines for classifying the healthiness of foods across various public settings. Commonly available products in Australian food service settings across eight food categories were classified according to each of the seventeen Australian state and territory food classification guidelines applying to public schools, workplaces and healthcare settings. Product nutrition information was retrieved from online sources. The level of concordance between each pair of guidelines was determined by the proportion of products rated at the same level of healthiness. Australia. No human participants. Approximately half (56 %) of the 967 food and drink products assessed were classified as the same level of healthiness across all fifteen 'traffic light'-based systems. Within each setting type (e.g. schools), pairwise concordance in product classifications between guidelines ranged from 74 % to 100 %. 'Vegetables' (100 %) and 'sweet snacks and desserts' (78 %) had the highest concordance across guidelines, while 'cold ready-to-eat foods' (0 %) and 'savoury snacks' (23 %) had the lowest concordance. In addition to differences in classification criteria, discrepancies between guidelines arose from different approaches to grouping of products. The largest proportion of discrepancies (58 %) were attributed to whether products were classified as 'Red' (least healthy) or 'Amber' (moderately healthy). The results indicate only moderate concordance between all guidelines. National coordination to create evidence-based consistency between guidelines would help provide clarity for food businesses, which are often national, on how to better support community health through product development and reformulation.

The food industry prioritizes quality control and product knowledge, emphasizing factors like quantity, freshness, and color. This research addresses Sustainable Development Goals (SDGs) focused on controlling food consumption, promoting health, reducing energy usage, and minimizing environmental impact. The primary objective was to utilize machine vision and deep learning to identify consumed food products. The study categorizes food into 32 classes, divided into three main categories, and includes the documentation of images and videos captured during consumption across various situations. Initially, the dataset comprised 12,000 images in 16 classes and 24,000 images in 32 classes, which were subsequently augmented to yield 60,000 and 120,000 images, respectively. The augmented datasets were then processed through nine popular deep learning architectures, identifying ResNet50, EfficientNetB5, B6, and B7 as the most effective architectures. An essential step involved updating hyperparameters, including image size, batch size, learning rate, and optimizer settings, to enhance convergence rates and accuracy. The EfficientNetB7 model was adapted for further testing and compared against two prominent optimizers, Adam and Lion. Ultimately, the EfficientNetB7 model with the Lion optimizer was chosen for the dataset. The results of this deep learning algorithm demonstrated remarkable performance, achieving 100% accuracy in identifying images of food-consumed products within 16 classes when using EfficientNetB7 and the Lion optimizer. For the 32-class case, the accuracy reached 99%, with the mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE) recorded at 0.0079, 0.035, and 0.18, respectively. These findings illustrate the robustness of the adjusted dataset in alignment with the designed deep learning architecture.

Nutrient-based indices are commonly used to assess the health potential of individual foods for nutrition policy actions. This study aimed to evaluate the nutrient profile-informed Australian Health Star Rating (HSR), against NOVA and an index informed by the Australian Dietary Guidelines (ADGs), to determine the extent of alignment. All products displaying an HSR label in the Australian marketplace between June 2014 and June 2019 were extracted from the Mintel Global New Product Database, and classified into one of four NOVA categories, and either as an ADG five food group (FFG) food or discretionary food. Of 4451 products analysed, 76.5% were ultra-processed (UP) and 43% were discretionary. The median HSR of non-UP foods (4) was significantly higher than UP foods (3.5) (p < 0.01), and the median HSR of FFG foods (4) was significantly higher than discretionary foods (2.5) (p < 0.01). However, 73% of UP foods, and 52.8% of discretionary foods displayed an HSR ≥ 2.5. Results indicate the currently implemented HSR system is inadvertently providing a ‘health halo’ for almost ¾ of UP foods and ½ of discretionary foods displaying an HSR. Future research should investigate whether the HSR scheme can be reformed to avoid misalignment with food-and diet-based indices.

Objectives. To identify the effect of a Breakfast in the Classroom (BIC) initiative on the foods and drinks students consume in the morning. Methods. Sixteen public schools in Philadelphia, Pennsylvania, that provide universal breakfast participated in a group randomized trial to examine the effects of BIC with complementary nutrition promotion between 2013 and 2016. Control schools (n = 8) offered breakfast in the cafeteria before school. Baseline data were collected from 1362 students in grades 4 to 6. Endpoint data were collected after 2.5 years. Students self-reported the foods and drinks they consumed in the morning. Results. At endpoint, there was no effect of the intervention on breakfast skipping. Nearly 30% of intervention students consumed breakfast foods or drinks from multiple locations, as compared with 21% of control students. A greater proportion of intervention students than control students consumed 100% juice, and a smaller proportion consumed sugar-sweetened beverages and foods high in saturated fat and added sugar. Conclusions. A BIC initiative led to improvements in the types of foods and drinks students consumed in the morning. However, the program did not reduce breakfast skipping and increased the number of locations where students ate.

The objective of this study was to inform consumer-facing dietary guidance by (1) adapting the current University of North Carolina at Chapel Hill (UNC) food processing framework to include a home processing (HP) component and (2) pilot testing the adapted version using a nationally representative sample of foods consumed in the U.S. The UNC framework was adapted to include guidelines for categorizing home-prepared (HP) foods. The original UNC and adapted HP frameworks were used to code dietary recalls from a random sample of National Health and Nutrition Examination Survey (2015–2016 cycle) participants (n = 100; ages 2–80 years). Percent changes between the UNC and HP adapted frameworks for each processing category were calculated using Microsoft Excel, version 16.23. Participants were 56% female, 35% non-Hispanic white (mean age = 31.3 ± 23.8). There were 1,376 foods with 651 unique foods reported. Using the HP compared to the UNC framework, unprocessed/minimally processed foods declined by 11.7% (UNC: 31.0% vs. HP: 27.4%); basic processed foods increased by 116.8% (UNC: 8.2% vs. HP: 17.8%); moderately processed foods increased by 16.3% (UNC: 14.2% vs. HP: 16.6%); and highly processed foods decreased by 17.8% (UNC: 46.5% vs. HP: 38.2%). Home-prepared foods should be considered as distinct from industrially produced foods when coding dietary data by processing category. This has implications for consumer-facing dietary guidance that incorporates processing level as an indicator of diet quality.

There is growing evidence linking the consumption of ultra-processed foods (UPFs) to adverse health outcomes. Accurate classification of foods according to the extent and purpose of industrial processing is therefore essential for improving dietary assessment and public health strategies. This study aimed to evaluate and compare the performance of two large language models (LLMs), DeepSeek-R1 and ChatGPT o1, in classifying foods according to the NOVA classification system. Both LLMs were tasked with categorizing a standardized list of 1,168 food items obtained from the Brazilian Food Composition Table (TBCA, version 7.0). The classifications generated by the models were compared with a reference list manually classified by a trained researcher. Quantitative analyses included the calculation of unweighted Cohen’s kappa between the LLMs, as well as accuracy, sensitivity, specificity, precision, and F1 score for each model. Qualitative analyses were conducted to explore discrepancies in food classification. ChatGPT o1 demonstrated superior performance across all evaluated metrics, achieving an accuracy of 98.0%, sensitivity of 94.7%, specificity of 99.0%, and an F1 score of 95.6%. In comparison, DeepSeek-R1 achieved an accuracy of 92.6%, sensitivity of 69.8%, specificity of 99.3%, and an F1 score of 81.1%. ChatGPT o1 also produced substantially fewer misclassifications than DeepSeek-R1 (23 versus 86, respectively). The findings highlight the potential of large language models to support dietary assessment and nutrition research. The development of an automated tool based on the NOVA food classification framework is recommended to assist nutritionists and researchers, enabling faster and more consistent food classification in both clinical and research settings. •ChatGPT o1 outperformed DeepSeek-R1 in NOVA food classification.•ChatGPT o1 scored: Accuracy 98.0%, Sensitivity 94.7%, Specificity 99.0%, F1 95.6%.•DeepSeek-R1 scored: Accuracy 92.6%, Sensitivity 69.8%, Specificity 99.3%, F1 81.1%.•ChatGPT o1 offered substantially fewer misclassifications (23) than DeepSeek-R1 (86).•Qualitative analysis showed five patterns caused 67% of model discrepancies.•This study highlights LLMs’ potential to aid dietary and nutrition research.•A NOVA-based automated tool is recommended to assist nutrition professionals.

Background Food manufacturers have made public statements and voluntary commitments, such as the Healthier Australia Commitment (HAC), to improve the nutritional quality of foods. However, limited information about the nutritional quality or healthfulness of new products makes it difficult to determine if manufacturers are doing this. The purpose of this study was to assess the healthfulness of new food products released into the Australian retail market in 2015, and whether those companies who were HAC members released healthier food options compared to non-HAC members. Methods This cross-sectional study assessed the healthfulness of all new retail food products launched in Australia in 2015 as indexed in Mintel’s Global New Products Database. Healthfulness was assessed using three classification schemes: Healthy Choices Framework Victoria, Australian Dietary Guidelines and NOVA Food Classification System. Descriptive statistics and chi-squared tests described and compared the number and proportions of new foods falling within each of the food classification schemes’ categories for companies that were and were not HAC members. Results In 2015, 4143 new food products were launched into the Australian market. The majority of new products were classified in each schemes’ least healthy category (i.e. red, discretionary and ultra-processed). Fruits and vegetables represented just 3% of new products. HAC members launched a significantly greater proportion of foods classified as red (59% vs 51% for members and non-members, respectively) discretionary (79% vs 61%), and ultra-processed (94% vs 81%), and significantly fewer were classified as green (8% vs 15%), core foods (18% vs 36%) and minimally processed (0% vs 6%) (all p  < 0.001). Conclusions This study found that the majority of new products released into the Australian retail food market in 2015 were classified in each of three schemes’ least healthy categories. A greater proportion of new products launched by companies that publicly committed to improve the nutritional quality of their products were unhealthy, and a lower proportion were healthy, compared with new products launched by companies that did not so commit. Greater monitoring of industry progress in improving the healthfulness of the food supply may be warranted, with public accountability if the necessary changes are not seen.

This study investigates diet quality across four popular dietary patterns: Ketogenic Diet, Low‐Carbohydrate Healthy‐Fat, Vegetarian, and Vegan, employing the NOVA and Human Interference Scoring System (HISS) classification systems. Utilizing a modified Food Frequency Questionnaire (FFQ) and analyzing 168 participants' dietary habits, the research identifies notable differences in dietary quality among the dietary patterns. While all groups reported lower consumption of UPFs than the general population, plant‐based diets demonstrated higher UPF consumption than ketogenic and low carbohydrate diets. The study reveals that both NOVA and HISS effectively identify UPFs, with significant differences observed at various processing levels, except for UPFs where both systems showed similarity. This research contributes to the detailed understanding of diet quality within popular dietary patterns, highlighting the importance of considering food processing in dietary choices and the need for ongoing research to further elucidate the health implications of different types of UPFs. This study evaluates the diet quality of four popular dietary patterns—Ketogenic Diet, Low‐Carbohydrate Healthy‐Fat, Vegetarian, and Vegan—using the NOVA and Human Interference Scoring System (HISS). Analyzing the dietary habits of 168 participants through a modified Food Frequency Questionnaire, the research found that while all dietary groups consumed fewer UPFs compared to the general population, plant‐based diets had higher UPF consumption than ketogenic and low carbohydrate diets. The study underscores the effectiveness of NOVA and HISS in distinguishing UPFs, highlighting the need for further research on the health implications of UPFs and the role of food processing in dietary choices.

To consider the use of systematic methods for categorising foods according to their nutritional quality ('nutrient profiling') as a strategy for promoting public health through better dietary choices. We describe and discuss several well-developed approaches for categorising foods using nutrient profiling, primarily in the area of food labelling and also with respect to advertising controls. The best approach should be able to summarise and synthesise key nutritional dimensions (such as sugar, fat and salt content, energy density and portion size) in a manner that is easily applied across a variety of products, is understandable to users and can be strictly defined for regulatory purposes. Schemes that provide relative comparisons within food categories may have limited use, especially for foods that are not easily categorised. Most nutrient-profiling schemes do not clearly identify less-healthy foods, but are used to attract consumers towards products with supposedly better profiles. The scheme used in the UK to underpin the colour-coded 'traffic light' signalling on food labels, and the one used by the UK broadcasting regulator Ofcom to limit advertising to children, together represent the most developed use of nutrient profiling in government policy-making, and may have wider utility. Nutrient profiling as a method for categorising foods according to nutritional quality is both feasible and practical and can support a number of public health-related initiatives. The development of nutrient profiling is a desirable step in support of strategies to tackle obesity and other non-communicable diseases. A uniform approach to nutrient profiling will help consumers, manufacturers and retailers in Europe.