Current food labelling practices in online supermarkets in Australia.

BACKGROUND: Food product labelling can support consumer decision-making. Several food product labels (nutrition information panels (NIPs), ingredients lists, allergen declarations and country-of-origin) are mandated for physical product packaging in Australia, with a voluntary front-of-pack nutrition labelling system, Health Star Ratings (HSRs), also available. However, labelling requirements are not explicitly extended to online settings and the extent to which this information is available in these increasingly important food environments has not been assessed. METHODS: Data from all individual food product pages was collected from the online stores of the two dominant supermarket retailers in Australia using automated web scraping in April-May 2022 (n = 22,077 products collected). We assessed the proportion of pages displaying NIPs, ingredients, allergens, country-of-origin and HSRs after excluding products ineligible to display the respective label. We also assessed whether HSRs were differentially available for higher- (healthier) and lower-scoring (less healthy) products, with HSR scores drawn from a comprehensive Australian food composition database, FoodSwitch. A manual inspection of randomly selected product pages (n = 100 for each label type per supermarket), drawn from products displaying the relevant label, was conducted to assess whether the labels were immediately visible to users (i.e. without scrolling or clicking). Differences in labelling prevalence and visibility were compared using chi-squared tests. RESULTS: Across both supermarkets, country-of-origin labelling was almost complete (displayed on 93% of food product pages), but NIPs (49%), ingredients (34%) and allergens (53%) were less frequently displayed. HSRs were infrequently displayed (14% across both supermarkets) and more likely to be applied to higher-scoring products (22% on products with ≥ 3.5HSR v 0.4% on products with < 3.5HSR, p < 0.001). One supermarket was far more likely to make NIPs (100% v 2%, p < 0.001), ingredients (100% v 19%, p < 0.001) and allergens (97% v 0%, p < 0.001) information immediately visible, though the other made HSRs more apparent (22% v 75%, p < 0.001). Both supermarkets displayed country-of-origin labels prominently (100% v 86%, p < 0.001). CONCLUSIONS: Food product labelling varies in online supermarkets in Australia overall and between supermarkets, while the design of online stores resulted in differences in labelling visibility. The near-complete display of country-of-origin labels and differential application of HSRs to higher-scoring products may reflect their use as marketing tools. Our findings highlight an urgent need for food labelling regulations to be updated to better account for online retail food environments.

Packaged foods purchased on price promotion in Australia.

We examined the prevalence and magnitude of price promotions among purchases of packaged foods and beverages in Australia, as well as the contribution of price-promoted foods and beverages to apparent energy intake. We utilized grocery purchase data from a nationally representative panel of 10 000 households in 2019 (NielsenIQ Homescan panel), combined with a food nutrition dataset (FoodSwitch). Nutritional quality was defined using the Australian and New Zealand Health Star Rating (HSR), where products with an HSR <3.5 were classified as 'less healthy' and products with an HSR ≥3.5 were classified as 'healthy'. Apparent energy intake was expressed as the total energy content of all purchased products per day per capita. Price promotions were claimed by panel members. Overall, four-in-ten packaged products (41%) were purchased on price promotion. Compared to 'healthy' products, 'less healthy' products were more frequently purchased on price promotion (33% vs 48%, respectively, p < 0.001), but had a similar mean magnitude of price discount (both 22%). Low socio-economic status (SES) households consumed 18% more energy from 'less healthy' packaged products on price promotion than high SES households (1141 vs 970 kJ/day/capita, p < 0.001). In conclusion, restricting price promotions for 'less healthy' packaged foods and beverages could potentially improve diet quality and dietary inequalities in Australia.

A Machine Learning Approach to Predict the Added-Sugar Content of Packaged Foods.

BACKGROUND: Dietary guidelines recommend limiting the intake of added sugars. However, despite the public health importance, most countries have not mandated the labeling of added-sugar content on packaged foods and beverages, making it difficult for consumers to avoid products with added sugar, and limiting the ability of policymakers to identify priority products for intervention. OBJECTIVE: The aim was to develop a machine learning approach for the prediction of added-sugar content in packaged products using available nutrient, ingredient, and food category information. METHODS: The added-sugar prediction algorithm was developed using k-nearest neighbors (KNN) and packaged food information from the US Label Insight dataset (n = 70,522). A synthetic dataset of Australian packaged products (n = 500) was used to assess validity and generalization. Performance metrics included the coefficient of determination (R2), mean absolute error (MAE), and Spearman rank correlation (ρ). To benchmark the KNN approach, the KNN approach was compared with an existing added-sugar prediction approach that relies on a series of manual steps. RESULTS: Compared with the existing added-sugar prediction approach, the KNN approach was similarly apt at explaining variation in added-sugar content (R2 = 0.96 vs. 0.97, respectively) and ranking products from highest to lowest in added-sugar content (ρ = 0.91 vs. 0.93, respectively), while less apt at minimizing absolute deviations between predicted and true values (MAE = 1.68 g vs. 1.26 g per 100 g or 100 mL, respectively). CONCLUSIONS: KNN can be used to predict added-sugar content in packaged products with a high degree of validity. Being automated, KNN can easily be applied to large datasets. Such predicted added-sugar levels can be used to monitor the food supply and inform interventions aimed at reducing added-sugar intake.

Uptake of mandated pregnancy warnings in the Australian alcoholic ready-to-drink beverage market.

INTRODUCTION: A mandatory pregnancy warning was introduced in Australia 2020 to advise the public of the potential harms of prenatal alcohol exposure. Due to industry pressure, a 3-year implementation period was granted. The aim of this study was to analyse the extent to which the mandatory warning had been applied to ready-to-drink (RTD) alcohol product labels almost 2 years into the implementation period. METHODS: The sample included 491 RTD products sold in three alcohol stores in Sydney, Australia in March-May 2022. Identified warnings were categorised as a mandated warning, a DrinkWise warning (an industry-developed option) or 'Other' warning. Analyses were conducted overall and by RTD type. RESULTS: Almost all (94%) of the sampled RTD products had some form of pregnancy warning, but only 36% displayed the mandatory version. Of the non-mandatory warnings, 74% were DrinkWise warnings (42% of total sample) and 27% were 'Other' warnings (15% of total sample). There was no apparent relationship between alcohol content and likelihood of displaying a mandatory warning. DISCUSSION AND CONCLUSIONS: Two years into the three-year implementation period for the mandatory pregnancy warning, only around one-third of the assessed RTD products exhibited compliance. Uptake of the mandatory pregnancy warning appears to be slow. Continued monitoring will be required to determine whether the alcohol industry meets its obligations within and beyond the implementation period.

The absence of mandatory pregnancy warning labels in online alcohol purchasing contexts.

INTRODUCTION: As people increasingly migrate to online shopping platforms, hard-won improvements in requirements for consumer information provision at the point of sale are being eroded. An example is the alcohol pregnancy warning label for packaged alcoholic beverages that has been recently introduced in Australia and New Zealand. The aim of the present study was to assess the extent to which the pregnancy warning was visible at the online point of sale when the requirement became mandatory in August 2023. METHODS: Data for alcohol products sold on the websites of the two largest alcohol retailers in Australia were web-scraped from 1 to 3 August 2023. The captured data for 8343 alcoholic beverages were inspected to determine whether the pregnancy warning was visible. RESULTS: Virtually no products (0.1%) had the mandatory warning visible on the main sales page, and only 7% enabled visibility of the warning via optional product image rotation functionality. DISCUSSION AND CONCLUSIONS: The almost complete absence of the mandatory pregnancy warnings on the main product pages of major alcohol retailers' websites highlights the regulatory problems posed by the emerging shift to online shopping. The very low prevalence of visible pregnancy warnings is likely to be an overestimate of the extent to which consumers would be exposed to warnings due to images being counted as being present regardless of their quality or readability. New regulation is needed to ensure that mandatory information requirements for harmful products are applied to online shopping contexts.

Assessing the Policy Implications of Different Definitions for Added Sugars: An Analysis Across the Australian Packaged Food and Beverage Supply.

Background: In recent years, various definitions of "added sugars" have emerged across jurisdictions. Although it is clear how these definitions differ, there is limited understanding of the policy implications associated with these variations. Objective: To test the potential policy implications of different definitions of "added sugars" on the Australian packaged food supply, we developed a method to estimate the content of "added sugars" in packaged foods and applied this to 3 different definitions of "added sugars": (i) United States Food and Drug Administration (US FDA) added sugar definition, (ii) the World Health Organization (WHO) free sugar definition, and (iii) a comprehensive definition that was developed from a review of the evidence on "added sugars." Methods: Using a representative sample of 25,323 Australian packaged foods, the "added sugar" content and proportion of products that contain "added sugar" under the 3 definitions were estimated. In addition, a comparative analysis exploring the impact of the US FDA definition (least comprehensive) vs. the comprehensive definition was conducted to understand potential implications of adopting different regulatory definitions in Australia. Results: The US FDA definition identified the lowest number and proportion of products with any "added sugars" at 14,380 products (representing 56.8% of all products), followed by the WHO free sugar definition at 15,168 products (59.9%) and the comprehensive definition at 16,260 products (64.2%). The mean estimates for "added sugars" were 8.5 g/100 g, 8.7 g/100 g, and 9.6 g/100 g for the US FDA, WHO, and comprehensive definitions, respectively. Compared with the US FDA definition, the comprehensive definition captured an additional 7.4% of products, largely driven by nonalcoholic beverages, special foods and fruit, vegetables, nuts, and legumes. Conclusions: Despite small variations in different "added sugars" definitions, their application has some significant policy implications. Findings highlight the importance of applying a comprehensive regulatory definition that adequately captures all sugars that have been linked to poor health.

Deconstructing the Supermarket: Systematic Ingredient Disaggregation and the Association between Ingredient Usage and Product Health Indicators for 24,229 Australian Foods and Beverages.

Unhealthy diets are underpinned by the over-consumption of packaged products. Data describing the ingredient composition of these products is limited. We sought to define the ingredients used in Australian packaged foods and beverages and assess associations between the number of ingredients and existing health indicators. Statements of ingredients were disaggregated, creating separate fields for each ingredient and sub-ingredient. Ingredients were categorised and the average number of ingredients per product was calculated. Associations between number of ingredients and both the nutrient-based Health Star Rating (HSR) and the NOVA level-of-processing classification were assessed. A total of 24,229 products, listing 233,113 ingredients, were included. Products had between 1 and 62 ingredients (median (Interquartile range (IQR)): 8 (3-14)). We identified 915 unique ingredients, which we organised into 17 major and 138 minor categories. 'Additives' were contained in the largest proportion of products (64.6%, (15,652/24,229)). The median number of ingredients per product was significantly lower in products with the optimum 5-star HSR (when compared to all other HSR score groups, p-value < 0.001) and significantly higher in products classified as ultra-processed (when compared to all other NOVA classification groups, p-value < 0.001). There is a strong relationship between the number of ingredients in a product and indicators of nutritional quality and level of processing.

An Innovative Machine Learning Approach to Predict the Dietary Fiber Content of Packaged Foods.

Underconsumption of dietary fiber is prevalent worldwide and is associated with multiple adverse health conditions. Despite the importance of fiber, the labeling of fiber content on packaged foods and beverages is voluntary in most countries, making it challenging for consumers and policy makers to monitor fiber consumption. Here, we developed a machine learning approach for automated and systematic prediction of fiber content using nutrient information commonly available on packaged products. An Australian packaged food dataset with known fiber content information was divided into training (n = 8986) and test datasets (n = 2455). Utilization of a k-nearest neighbors machine learning algorithm explained a greater proportion of variance in fiber content than an existing manual fiber prediction approach (R2 = 0.84 vs. R2 = 0.68). Our findings highlight the opportunity to use machine learning to efficiently predict the fiber content of packaged products on a large scale.