1. Introduction
Over the past 30–40 years, the portion-size of many packaged food and beverage products has increased significantly [
1,
2,
3,
4,
5,
6]. Package and portion size are known to influence the quantity of food an individual selects and consumes [
5,
6,
7]. When offered larger packages or portions of food or beverages, individuals are known to consume more and are unlikely to compensate by increasing their physical activity or reducing the quantity of other foods and beverages eaten at the same sitting or later in the day [
6,
8,
9,
10,
11,
12]. The sustained provision of large portion sizes of nutrient poor but energy dense foods and beverages may be an important contributor to obesity and non-communicable diseases (NCD) [
8,
11].
Initiatives targeting portion and package size (henceforth referred to as portion size) have been identified as a promising approach to reduce obesity and obesity-related NCDs [
13,
14,
15]. To-date, the majority of initiatives targeting portion size have been voluntary, with uptake at the discretion of the food industry [
15]. These interventions have targeted portion size in two main ways: (i) through a reduction in the quantity of the product provided, predominantly through a change in package size; or (ii) by reducing the energy density per serve of the food or beverage product (reformulation) [
15]. The proposed New York City (NYC) ban of very large servings of sugar-sweetened beverages (SSBs) is the most high profile example of the first approach, although its implementation was unsuccessful [
16]. The pledge by sugary drink companies to reduce the sugar content and energy density of their beverages as a part of the UK’s Public Health Responsibility Deal [
17], is an example of the second approach.
In 2011–2012, in Australia, 63% of adults and 25% of children were classified as overweight or obese, with overweight and obesity deemed the second highest contributor to the burden of disease in Australia [
18]. Similar to other countries, many Australians currently consume poor quality diets with nearly a third of energy coming from discretionary foods [
19]. The latest dietary survey results showed that 34% of Australians consume SSBs, which contributed to 4% of total energy consumed and 17% of total sugars consumed for individuals aged two years and older [
19]. Furthermore, the survey highlighted that consumption of SSBs was higher for children aged 2–18 years than adults (47% and 31%, respectively) [
19]. The Australian government has identified tackling obesity as a priority action [
20,
21], and changes in portion size as a key target [
21]. Despite strong interest by government and the food industry, the potential effectiveness and cost-effectiveness of implementing such strategies to reduce obesity at a population level has not been investigated. We therefore estimated the potential cost-effectiveness of implementing: (i) a package size cap on single-serve products (
package size cap); and (ii) product reformulation to lower energy density (
energy reduction) on packaged SSBs available for sale in Australia.
3. Modelling Cost-Effectiveness
Changes in energy intake (kJ) at the population level (by age and sex) were converted to changes in body weight (kg) using validated energy balance equations for children (aged 2–19) and adults (aged ≥ 20) [
43,
44]. These changes in weight were converted to changes in BMI using average Australian height and weight by sex, for single-year age groups (children) and five-year age groups (adults) from the AHS 2011 [
25].
The altered distribution of BMI as a result of the interventions was applied to the Obesity model to estimate lifetime HALYs. The Obesity model quantifies changes in the total mortality and morbidity of the 2010 Australian population resulting from changes in the epidemiology of obesity-related diseases (i.e., incidence, prevalence and mortality) and the independent impact of non-disease obesity on quality of life. Nine causally obesity-related diseases were included: ischaemic heart disease, hypertensive heart disease, ischemic stroke, diabetes, colorectal cancer, kidney cancer, breast cancer, endometrial cancer and osteoarthritis [
45]. Total cost offsets were the result of health care cost savings attributable to the intervention as a result of reduced incidence of obesity related diseases. Total HALYs and costs were estimated comparing: (i) a reference population that represents the current BMI distribution and disease patterns of a cohort of the 2010 Australian population; and (ii) an intervention population that is identical to the reference population but includes the impact of the intervention on the BMI distribution of the population. The modelling was conducted for the Australian population aged 2–100 years, over their lifetime.
All costs and benefits were discounted at 3% and are expressed in 2010 values. The Health Price Index from the Australian Institute for Health and Welfare (AIHW) [
46] was used to adjust costs to 2010 AUD as required. The Incremental Cost Effectiveness Ratio (ICER) was calculated by dividing the incremental net costs by incremental health benefits of the intervention compared to current practice. Further details on the model can be found elsewhere [
47].
Uncertainty Analysis
The estimates for each cost element and the changes in weight, BMI and HALYs resulting from the interventions were estimated as means with 95% uncertainty intervals (UIs). Monte Carlo simulation (2000 iterations) was used to estimate parameter uncertainty using Ersatz (version 1.35) software—an Excel add-in [
48]. ICERs are presented on a cost-effectiveness plane, which demonstrates the range of plausible ICERs for each intervention and associated scenarios. Interventions that are both cost saving and increase health benefits are considered “dominant”. If the intervention is more costly and more effective than current practice, a willingness to pay threshold of AUD 50,000 per HALY gained [
49] is used to determine cost-effectiveness.
5. Discussion
This cost-effectiveness analysis showed that both the
package size cap and
energy reduction interventions were likely to be “dominant” (both effective and cost saving) in the Australian context under current modelling assumptions (
Figure 2). This modelling exercise suggests that policy-based population wide interventions such as these are likely to offer excellent “value for money” as obesity prevention measures, especially if implemented on a mandatory basis.
Results from the modelled scenarios confirm the current consensus that population-level portion size interventions are a promising approach to addressing obesity and obesity-related NCDs [
13,
14,
15] and therefore should be considered by policy makers. There is a lack of cost-effectiveness studies that investigate the impact of population-based interventions that change the food environment. Such studies are needed for policy makers to make informed decisions on how to spend limited resources [
50]. To provide further context to policy makers as to whether these proposed interventions are a worthwhile investment, the results of these analyses should be interpreted in relation to other comparable cost-effectiveness studies in the Australian setting. To-date, only one other study has used a similar standardised methodology to evaluate the cost-effectiveness of active transport on obesity-related health outcomes [
47]. This study found that active transportation interventions result in fewer health benefits relative to their cost. It follows that interventions targeting positive change in the food environment may be more impactful than those seeking to increase levels of active transportation. Another cost-effectiveness study from the broader ACE-Prevention study [
51], evaluated the implementation of traffic light labelling (TLL) and a junk food tax targeting at Australian adults [
29]. This study and other interventions from ACE-Prevention which sought to implement changes to the food environment were found to be dominant compared to “program-based” food interventions [
51]. The results of our study were comparable to a more recent study which modelling the impact of a tax on SSBs on the 2010 Australian adult population [
52].
The primary strength of this analysis is the policy relevance of the interventions chosen. In Australia in 2015, the Healthy Food Partnership (the Partnership) was established as the primary government-led initiative to address food reformulation in relation to NCD prevention. The Partnership consists of the Australian Government, food industry bodies and leading public health groups that have agreed to work cooperatively to tackle obesity, encourage health eating and empower food manufacturers to make positive changes to their products [
21]. The interventions modelled in this research strongly align to two of the three core objectives of the Partnership (portion size and reformulation) [
53]. Using multiple scenarios and varying realistic assumptions, this research has provided the opportunity to assess the potential impacts of such public health initiatives on the Australian population.
The limitations of this study centre around the quality of evidence relating to intervention effectiveness. Direct evidence supporting the likely impact of the interventions on consumer behaviour is weak. To counter this uncertainty, we made conservative assumptions to estimates of change in energy consumption; in addition to modelling multiple scenarios of intervention effectiveness. Another limitation involved the estimation of consumption of single-serve SSBs. As data on SSB consumption by package size were not available from nationally representative dietary surveys, we had to use a calculated estimate using sales data from a relatively small intervention trial. This may have underestimated the actual consumption of single-serve SSBs. As more evidence of these effects of these types of interventions becomes available, these assumptions can be revisited.
Furthermore, as definitive data on the costs for changes in packaging and reformulation could not be sourced, estimated costs to industry are not well substantiated. However, for the package size cap intervention, as food manufacturers are reducing their products to a package size that already exists (375 mL), we think the estimates are likely to be conservative. Additionally, although it is acknowledged that implementing a package size cap or energy reduction may be more complex than estimates of implementing changes to front of pack labelling (from which the cost estimates were derived), we believe that if the food industry were given sufficient lead time, the costs to changing packaging and reformulation would be significantly reduced because of changes in product packaging and formulation that occur as a part of natural product-lifecycles. Our model did not take in to account the potential loss of revenue to the food industry and its subsequence impact on consumers. It is likely that these costs would be recovered from re-distribution of sales (e.g., sales lost from 600 mL SSBs would be replaced from sales from 375 mL), however it is also possible that some costs may be passed on to the consumer, especially for the small-to-medium sized manufacturers. Additionally, as with all population-based models, our results represent a simplified version of reality. However, input parameters included 95% uncertainty intervals determined by the Monte Carlo simulation (2000 iterations) using Ersatz (version 1.35).
The implications of this study are that both a
package size cap and
energy reduction on SSBs are likely to be highly cost-effective and have sizeable effects on population health. Despite the degree of uncertainty around the size of benefits, both interventions should be considered for implementation in Australia. Interventions chosen to be modelled were based on a global review of current literature and policies [
15], and reflected currently implemented Calorie Reduction Pledges as proposed in the United Kingdom’s Responsibility Deal [
17]. Furthermore, our chosen interventions also reflected portion size related public health initiatives in the United States, such as the proposed New York City ban on 16oz SSBs [
54] and the Health Weight Commitment pledge. Comparisons of government implemented legislative scenarios and voluntary scenarios highlighted the importance of food industry adherence if these interventions were implemented in a real-world setting. Both HALYs gained and healthcare system savings were higher in the government legislated scenarios, than the voluntary scenarios (
Supplementary Table S1). The role of government support is thus of great importance to the success of these interventions [
55].
Further research in this area could be undertaken to investigate other scenarios for the design and implementation on both interventions modelled in this analyses. For example, this model could be used to analyse the effects of a
package size cap or
energy reduction targeted at different food categories such as confectionery, ready meals (including those sold at Quick Service Restaurants), snack foods and foods specifically designed for school canteens. Furthermore, this model could be used to look at the potential impact of pricing food items proportionate to package size. Many countries are taking action around SSBs, including SSB taxes, and it would also be valuable to model the impact of combined interventions targeting SSBs. Importantly, it would also be imperative to perform research into consumer acceptability of portion size interventions, particularly because consumer acceptance studies have indicated that such initiatives may be perceived by consumers to restrict their freedom of choice [
32,
56].