Packaged food now predominates food purchases in most countries [1
], and the consumption of packaged foods in developing countries is rapidly increasing [3
]. Technological developments in food processing along with advancements in food distributions systems have seen traditional wet markets for fresh foods replaced with supermarkets and hypermarkets which are abundant in packaged foods [5
]. Usually, packaged foods are manufactured or processed before reaching the consumer and are more likely to have salt, sugar and fat added [5
]. The increasing availability of packaged foods plays a key role in the nutrition transition, whereby nations move away from a traditional diet based on wholefoods to a diet of convenience [3
]. Packaged food sales grew by 3% in Australia in 2016, with 58.8% of food and beverage products sold from retail and foodservice outlets being packaged foods [11
The over-consumption of energy-dense, nutrient poor foods is a well-established risk factor for obesity and diet-related non-communicable diseases (NCDs) [4
]. In 2013, the World Health Organization (WHO) developed a global NCD monitoring framework that included food environment-related risk factors such as the sale of unhealthy packaged foods high in salt, saturated fat and trans
]. The framework identified approaches to monitoring national system responses to address these risks [15
], but data on the healthiness of the food supply at the country-level are few [17
]. Household expenditure surveys and national nutrition surveys, the most common measures of national dietary behaviours, capture important information but they do not provide adequate data to characterise and monitor the impact of actions of government and the food industry affecting the healthiness of the food supply [17
INFORMAS (International Network for Food and Obesity/NCD Research, Monitoring and Action Support)—a global network of public-interest organisations and researchers—developed a framework to monitor, benchmark and support public and private sector actions to create healthy food environments [19
]. Specific to food composition, the monitoring approach identified was to systematically collect information about nutrient composition of the food supply and to assess the energy density, salt (sodium), saturated fat, sugar, trans fat and portion sizes of packaged foods and beverages as well as their summary nutrient profile [17
The objective of this research was to use INFORMAS measures to examine the nutrient composition of the Australian packaged food supply and explore additional measures based on dietary guidelines, extent of food processing and progress towards established reformulation targets.
This was a cross-sectional examination of packaged foods available in supermarkets in Australia between August 2016 and August 2017. Assessment measures proposed by INFORMAS [17
] were used and supplemented by methods based upon dietary guidelines, extent of processing and nutrient reformulation targets (Table 1
). Specifically, we applied the Australian Health Star Rating (HSR) methodology based on nutrient profiling [21
], the classification of foods as core or discretionary according to the Australian Dietary Guidelines (ADGs) [22
], the classification of foods based upon their level of processing [24
] and the proportion of foods that meet established reformulation targets for sodium, saturated fat and sugars based upon national or international recommendations (Table 1
2.1. Data Source
Packaged food data (nutrient information, serving size and food category) were obtained from the 2016/17 Australian FoodSwitch database [25
]. Briefly, the collection and processing of packaged food data from four large supermarkets follows standard processes [27
] and is supplemented with crowd-sourced information for additional products [25
]. Packaged food product data were only included in the analysis if nutrient information was available in either per 100 g or per 100 mL format.
2.2. Food Categorisation
Products in the FoodSwitch database are assigned to one of 1271 food categories that group through a tree structure into 17 major food categories [25
]. The major foods categories ‘Foods for specific dietary use’, ‘Vitamins and minerals’ and ‘Alcoholic beverages’ were excluded, leaving 14 major food categories for evaluation. Analysis of first order subcategories of the 14 major categories ‘Dairy’ and ‘Non-alcoholic beverages’ was also undertaken to illustrate specific points. The ‘Herbs and spices’ and ‘Variety packs’ first order subcategories were excluded, leaving 55 for evaluation.
2.3. Measures of Nutritional Quality
—Selected risk-associated nutrients known to have adverse effects on an individual’s health (energy density, saturated fat, total sugars, sodium), as well as declared serving size were assessed. Trans
fat and added sugars were excluded because labelling is not required under Australian law and most food products did provide this information. Targets for nutrient composition were identified [28
] and nutrient levels in Australian packaged foods were compared against applicable Australian Government Food and Health Dialogue [29
], New Zealand Heart Foundation [31
] and United Kingdom [32
] food reformulation targets. Where there were multiple reformulation targets for the same food category, the most recent target was used. If both average and maximum targets were available, the maximum target was used.
Nutrient profiling summary score
—The nutrient profiling method used for this analysis was that underlying the Australian Health Star Rating (HSR) front of pack labelling system, a government-endorsed nutrient profiling method. The HSR system assigns packaged food products a rating between 0.5 (least healthy) and five stars (healthiest) in ten half-star increments based on the nutritional composition of the product [20
]. The Australian FoodSwitch database captures the HSR when labelled on pack, and for instances when HSR is not available on pack, calculates this using energy, protein, saturated fat, total sugar and sodium values from product labelling. The HSR algorithm also requires information that is not always available on pack (fruit, vegetable, nut and legume (FVNL), concentrated FVNL and fibre). Where required nutrition data were absent, proxy values were estimated using information drawn from the back-of-pack ingredients list, generic food composition databases, or similar products using methods described previously [26
]. We classified a product as ‘unhealthy’ if the product had a HSR < 3.5. This cut-off point is based on previous research indicating that healthy core foods with a HSR of ≥3.5 can be confidently promoted in public settings [35
Guidelines—Foods were classified as core or discretionary based on definitions within the Australian Dietary Guidelines [22
]. Core foods are foods that form the basis of a healthy diet (e.g., fruit, vegetables, lean meats, milk, yoghurts, cheeses, grains). Discretionary foods are energy-dense and nutrient-poor foods not necessary for providing the nutrients the body needs (e.g., sweetened soft drinks/cordials/waters, biscuits, chocolate, meat pies, butter, salty snacks). The relevant assignment was applied to each of the 1271 food subcategories.
processing—Foods were classified as ‘less processed’, ‘moderately processed’ or ‘highly processed’ using an adapted version of the NOVA classification framework [24
]. This classification framework was used as previous research indicated it was easiest to apply [38
]. The relevant assignment was applied to each of the 1271 food sub-categories and primary reporting was of the proportion ‘highly’ processed.
Median and interquartile range (IQR) was determined for each measure of nutrient composition; mean and standard deviation (SD) for the measure of HSR; and proportions (%) for the measures based upon HSR < 3.5, classification of foods as either core or discretionary, extent of processing, and proportions of foods meeting established reformulation targets. Estimates were made overall, for each major food category and for illustrative sub-categories.
Coherence of the different measures of nutritional quality was assessed by ranking each of the 14 major food categories by each measure—lower energy density, saturated fat, total sugar, sodium, serving size, proportion HSR < 3.5, proportion discretionary and proportion highly processed, and higher mean HSR, being better. To enable a visual comparison, the upper five ranks were coloured light grey, the lower five dark grey and the middle four grey, with the data plotted as a heat map. Major food categories on the heat map were ordered for listing according to average rank across proportion HSR < 3.5, proportion discretionary and proportion highly processed. Quantitative analyses were done using Minitab™ 17 Statistical Software (Minitab Inc., State College, PA, USA).
Broadly consistent assessments of the nutritional quality of the Australian food supply were obtained using methods based upon nutrient profiling, dietary guidelines and extent of food processing. The observation that simpler methods based upon dietary guidelines and extent of food processing [7
] provided comparable high-level findings to the more intensive nutrient profiling process is important because these methods may be more feasible in settings where detailed nutritional data and resources are limited [42
]. Methods based upon dietary guidelines and extent of food processing are, however, unlikely to provide the detailed insight provided by nutrient profiling [38
], which has a demonstrated capacity to discriminate between the healthiness of products within fine subcategories. For example, processed foods do not all have a poor nutrient composition with products suchas hummus being classified as highly-processed despite a substantially better nutrient composition than many other highly processed foods such as processed meats or sugar-sweetened beverages.
The assessments based upon individual nutrients or serving size, by contrast, did not provide assessments of nutritional quality that were coherent across food categories or nutrients. Neither were the rankings of the nutritional quality of the food categories based upon individual nutrients consistent with those obtained using nutrient profiling, dietary guidelines or extent of processing. The nutrient values did identify specific aspects of concern about food categories that might be targeted for action.
Assessments based upon achievement of nutrient reformulation targets are currently of limited value because targets have been set for only the minority of foods, and the foods with targets set are unlikely to be representative sample of all foods. The proportion of foods with a target set, and progress against a full set of targets might, however, be another useful measure by which to hold government and the food industry accountable for progress towards public health nutrition initiatives [17
]. Consistency of targets across jurisdictions would simplify policy development, the implementation of interventions, and program monitoring. If targets are applied at a fine sub-category level this is a highly plausible approach and while some limited tailoring to local context may be required, most problems with the packaged food supply are not bounded by geography, and there is a strong argument for greater standardisation and global coordination.
A key strength of these analyses is the large and diverse dataset upon which the assessments were made, and the ability to calculate and compare multiple different measures of nutritional quality. This study reports findings for more complex approaches such as nutrient profiling which quantitatively summarise multiple dimensions of nutritional quality, as well as simpler and more pragmatic methods such as dietary guidelines and extent of processing. In addition to the practicalities of data collection, it is also possible that different measures will be of value for the assessment of different aspects of food policies seeking to improve the nutritional quality of the food supply.
Important weaknesses are the restriction of the data to Australian products, which raises uncertainties about the broader generalisability of the findings to other settings. For every measure, weighting of estimates for sales volumes would enhance the insight provided [45
] although crude estimates have been shown previously to capture key characteristics of the packaged food supply [37
]. In the absence of sales data, an alternate approach would be to utilise data from national dietary surveys to examine the proportion of daily energy that comes from the main food categories. For example, according to the 2011–12 Australian Health Survey, 35% of an adult Australian’s energy intake comes from discretionary foods. The assessment methods reported here are an incomplete set of all possible methods but cover the main types of approaches possible. The World Health Organization’s has criteria that define eligibility for marketing food products to children [47
] and the Food Standards Australia New Zealand has another set of Nutrient Profiling Score Criteria [2
] that determine the eligibility for products to carry health and nutrient content claims, that might be studied in future. Similarly, there are tens of other nutrient profiling methods [48
] and a range of different systems for defining the extent of food processing [24
] that could also be evaluated.
Another important weakness is the absence of data on trans
fat because trans
fat is a strong determinant of disease risk that can be addressed by proven policy interventions [56
]. Likewise, the reliance on data describing total sugar is sub-optimal because it is free sugars that are the primary health concern and the target of interventions. Legislation of added sugar labelling has been passed in the United States [58
] and should be adopted in Australia. It should also be a part of the standard Codex recommendation [60
] for nutrient declarations. This paper has reported only on packaged foods but many of the methods could be extended and applied to assess foods served in restaurants or other foodservice establishments. Application of the assessment process to encompass the entire food environment might provide for a more complete assessment and more holistic decision-making.
Data of this type have multiple potential uses. The comparisons made against reformulation targets provide a direct assessment of government action and highlight areas for food industry action. Between country comparisons of the nutritional quality of foods in conjunction with data describing food environment features might also provide insight into the food system interventions most likely to be effective. More granular reporting of comparisons between individual products and major companies would be highly newsworthy, and has the potential to drive corporate product ranging and formulation decisions [17