Impact simulation of portion guidance on commonly consumed packaged food to reduce discretionary energy and public health sensitive nutrients in the United States, United Kingdom, and France

Abstract

Purpose

Excessive consumption of discretionary food is a global concern linked to obesity prevalence. Providing reference portion sizes and reducing portion size is a potential solution to tackle this issue. The aim of this study was to model the impact of systematically established portion values, without changing food choices, on discretionary energy and public health-sensitive nutrients using nationally representative datasets from the United States (NHANES), United Kingdom (NDNS), and France (INCA3).

Methods

Diet modeling was used to adjust large portions to reference portion sizes per eating occasion for selected discretionary, energy-dense food categories. Nutritional quality and adherence to World Health Organization guidelines were compared before and after substitutions. Linear regressions assessed whether reductions in daily total energy and added/free sugar were associated with participants’ demographic and anthropometric characteristics.

Results

After substitution, daily energy from discretionary foods decreased by 5.4 percentage points for NHANES (mean ± SD from 20.7 ± 13.1 to 15.3 ± 9.8%, p < 0.001), 3.1 percentage points for NDNS (from 15.6 ± 9.8 to 12.5 ± 8.0%, p < 0.001), and 3.6 percentage points for INCA3 (from 16.1 ± 10.7 to 12.5 ± 8.1%, p < 0.001) data. Following portion guidance improved diet quality, reduced public health-sensitive nutrients (e.g., saturated fat, added sugar, sodium) and enhanced adherence to WHO guidelines. Older age was associated with smaller reductions in discretionary energy and added/free sugar among participants in UK and France.

Conclusion

Adjusting portion sizes without altering food preferences may decrease energy and public health-sensitive nutrients intakes. Further research is needed to explore these findings in other regions and their potential health outcomes.

Introduction

Discretionary foods refer to food items that are typically energy-dense and nutrient-poor with a minor role in the diet but can provide dietary diversity [1]. Foods in this category are not a necessary part of the diet but can be consumed occasionally in small amounts by those who are physically active [1]. In publications, these foods have also been described as “discretionary choices” [1, 2] or “Calories For Other Use” in the 2020–2025 Dietary Guidelines for Americans [3]. Because portion sizes all over the world have increased in the past decades [4], the associated increase in energy intake and intake of health-sensitive nutrients (added sugars, saturated fat, and sodium) from discretionary foods is thought to contribute to the high prevalence of obesity in western countries and across the globe [5–7].

Some countries have established specific upper daily limits for discretionary energy, which generally range from 4 to 18% of total daily energy. For example, in the US, up to 12% of energy (based on a standardized 2000 kcal diet) can be allocated for discretionary food consumption [3]. Denmark allows a maximum of 4–6% of energy intake from nutrient-poor and energy-dense foods in a recommended diet [8]. Other countries, such as Australia and Ghana, have limits of 9% and 5% of energy intake from discretionary foods, respectively [1, 2]. However, national nutrition surveys in some countries show that actual consumption of discretionary foods is much higher, with energy intakes from these foods ranging from 19% of daily energy intake in Scotland [9], 23% in US [10], 26% in Mexico [11], to 33% in Australia [12].

Portion size refers to the amount of food an individual chooses to consume at one time, whereas serving sizes are reference amounts used for food labelling purposes in some countries [13]. For example, serving sizes in the U.S. are defined by the Food and Drug Administration (FDA) and listed on the Nutrition Facts label. Interventions on portion size reduction have been shown to be effective in reducing energy intakes, thereby offering potential strategies to mitigate the ongoing epidemic of obesity and cardiometabolic diseases [14, 15]. It is well documented that people eat more when served larger portions, the so-called “portion-size effect” [16, 17]. In addition, reference portion sizes are known to have an impact on actual intakes without modifying overall dietary pattern, highlighting the potential effectiveness of this approach as a valuable public health tool for providing guidance on healthy eating habits [18–20]. Regulations and official guidance on portion sizes, in countries where they are implemented, are meant to guide consumers towards a more reasonable intake of discretionary foods and lead to decreased intakes of public health sensitive nutrients.

In countries where portion or serving sizes are not regulated, guidance towards responsible portions could come from reference portion values, such as those recently developed [21]. These values, called global portion values (GPV), were established based on regulated serving sizes around the world, portion recommendations from Food-Based Dietary Guidelines (FBDG), and typical consumption habits. The potential effects of adhering to these portion guidelines on nutritional intake and overall diet quality have not been previously investigated.

The aim of this study was to model the effects of following recently established portion guidelines (country specific regulations where applicable and GPV for the remainder) without changing food choices on discretionary energy intake, public health sensitive nutrient intakes, and overall diet quality using nationally representative datasets from the United States, United Kingdom, and France.

Methods

Cohorts included

Data from three nationally representative population surveys were accessed and analyzed: 1) The US National Health and Nutrition Examination Survey (NHANES) wave 2017–2018; 2) the UK National Diet and Nutrition Survey (NDNS) years 9–11 (2016/17 to 2018/19); and 3) the Third French Individual and National Food Consumption (INCA3) Survey, 2014–2015 (Fig. 1). All three studies have been approved by the appropriate ethics committee and have therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

Fig. 1.

NHANES, National Health and Nutrition Examination Survey (wave 2017–2018); NDNS, The UK National Diet and Nutritional Survey (years 9–11, 2016/17 to 2018/19); INCA3, The Third French Individual and National Food Consumption (2014–2015). Flowchart of the three national survey cohorts included and the exclusion criteria for the analytic sample

These studies are described in detail elsewhere [22–24]. Briefly, NHANES is an ongoing survey conducted by the National Center for Health Statistics, which is part of the Centers for Disease Control and Prevention to monitor the health, dietary, and nutritional status of the noninstitutionalized civilian population in the United States. The survey employs a complex sampling design, involving stratification and multistage cluster sampling. Participants are interviewed in their homes and then undergo a comprehensive examination of their diet and health in a mobile examination center. Detailed information about subject recruitment, survey design, and data collection procedures can be found online [25]. The NHANES protocols have been approved by the National Center for Health Statistics Ethics Review Board. All participants provided written informed consent. In this study, data from NHANES wave 2017–2018 were included (publicly available at: http://www.cdc.gov/nchs/nhanes/).

NDNS is a UK government-funded survey (2008–2019) providing a representative snapshot of dietary and nutritional intake and status in the population. Details of the study design and data collection procedures were recently published [24]. The survey collected detailed, quantitative information on the food consumption, nutrient intake and nutritional status of the general population living in private households in the UK. The NDNS operates under the approval of the UK National Health Service Health Research Authority Research Ethics Committee (Years 6–10 and 11–15, East of England-Cambridgeshire South REC, REF 13/EE/0016), adhering to the principles outlined in the Declaration of Helsinki. This study included NDNS data from years 9–11 (2016/17 to 2018/19), data access is approved by UK Data Service.

The French INCA3 food consumption survey is a national cross-sectional survey carried out by The French Agency for Food, Environmental and Occupational Health & Safety (ANSES) in mainland France between February 2014 and September 2015 [23]. One of its main objectives was to update the data on the foods and beverages consumed and the nutrient intakes for individuals living in mainland France, following as much as possible with methodology set by the European Food Safety Authority (EFSA) [26]. Data from INCA3 used in the current study is publicly available at: https://www.data.gouv.fr/fr/.

Inclusion/exclusion criteria

From NHANES cycle 2017–2018, NDNS years 9–11 (2017–2019), and INCA3 (2014–2015), participants (Fig. 1) who were under the age of 18 years (y) and who were pregnant were excluded. All other adults 18y and older were included in the analysis. The final analytical sample consisted of 8285 adults (4273 for NHANES, 1912 for NDNS, 2100 for INCA3).

Dietary assessment

NHANES

Dietary information in NHANES is assessed using the Automated Multiple Pass Methodology (AMPM) and included two 24-h recalls collected by trained interviewers 3–10 days apart. Participants indicated all foods and beverages consumed during the previous day, providing detailed food descriptions and quantitative intakes per eating occasion. Energy and nutrient intakes were then assessed by merging intake information to the Food and Nutrient Database for Dietary Studies (FNDDS). The FNDDS is a US food composition table, containing energy, macro- and micronutrient contents for 7083 items.

NDNS

Dietary information in NDNS was collected every year from food diaries over four randomly assigned non-consecutive days to be representative of all days of the week [24]. Food diaries were self-reported and portion size of the foods/beverages consumed were estimated to reduce participant burden from reporting intakes. Quantitative and detailed dietary intakes collected enabled estimation of energy and nutrient intake at an individual level, per eating occasion, by linkage with the NDNS food composition table.

INCA3

Data on food and beverage consumption was collected over three non-consecutive days using the GloboDiet program, a standardized 24-h recall method [27–30]. The GloboDiet methodology has been proven to produce reliable dietary data for epidemiological studies [31] and food consumption surveys [32, 33]. Participants reported their intake for ten eating occasions, including main meals and snacks. Portion sizes were estimated using pictures, household measures, or weight/volume measurements. A picture book, validated on 509 individuals, aided participants in accurately reporting their food intake.

Added sugars were estimated with a semi-automated method developed by one of the authors [34]. First, foods were manually annotated as belonging to one of three groups: 1) foods not containing any added sugars (e.g., whole fruit), 2) foods where all sugar is added (e.g., candies), 3) foods with a mixture of added and naturally occurring sugars. The proportion of added sugars in the third group was estimated in a data-driven way using a machine-learning based predictive model [34]. The regression model used protein, carbohydrates, fiber, total fat, saturated fat and sodium as predictors. Although the model was trained on US and Australian datasets, it generalized well to data from other countries [34].

Definition of discretionary food

Discretionary foods in the current study were defined as energy-dense, nutrient-poor food items with minor role in the diet. This definition was modified from the Australian Dietary Guidelines and generally includes energy-dense foods that are high in saturated fat, added sugars, or salt [1]. No quantitative cut-off values were used in this study to identify discretionary foods based on their nutrient content. Rather, their role in the diet was assessed, and the discretionary food groups were assigned accordingly. Table 1 provides the list of food groups considered as discretionary foods and their corresponding portion values used as reference in this analysis.

Table 1.

Reference portion values for discretionary food categories

	NHANES	NHANES	NDNS and INCA3
Food categories	Description	(RACCs)	(GPV)
All soft drinks (ml)	Soft drinks, sugar- and/or artificially sweetened	360	250
Chocolate (g)	Candies, confectioneries and chocolates	30	30
Cookies or biscuits (g)	Sweet biscuits, cookies	30	30
Dairy desserts (g)	Pudding, custards	120	120
Dressings for salads (g)	Excludes single oils (e.g. olive oil, sunflower oil etc.)	30	30
Gummies and jelly sweets (g)	Soft sugar including gummies and jelly sweets	30	30
Hard sugar and chewy candies (g)	Toffees, caramels, fruit flavored sweets	15	15
Ice-cream (g)	All types of ice-creams (includes tubs, bars, cones and sandwiches)	160	60
Main dish sauces small (g)	Includes gravy, cheese sauce, pizza sauce, pesto, alfredo, white sauce, cocktail sauce, etc	60	60
Major condiments (g)	Includes ketchup, barbecue sauce, soy sauce, oyster sauce, steak sauce, teriyaki sauce, marinades	15	15
Mayonnaise (g)	Includes sandwich spreads, mayonnaise-based dressings, tartar	15	15
Minor condiments (ml)	Includes liquid seasonings, fish sauces, horseradish, hot sauce, Worcestershire sauce	5	5
Mints and rolled candies (g)	Includes breath mints and other pressed candies	5	5
Mustard (ml)		5	10
Prepared condiments (ml)		30	30
Salted crackers or biscuits (g)	Savory crackers, snack	30	30
Sweetened condensed milk (ml)		30	20
Cakes and pies (g)		80	80
Doughnuts, sweet rolls, and pastries (g)		55	55
Popcorn (g)	Both salted and sweetened	30	30
Potato chips, tortilla, corn, other chips (g)		30	30
Pretzels/snack mix (g)		30	30

NHANES, National Health and Nutrition Examination Survey (wave 2017–2018); NDNS, The UK National Diet and Nutritional Survey (years 9–11, 2016/17 to 2018/19); INCA3, The Third French Individual and National Food Consumption (2014–2015); RACCs, Reference Amounts Customarily Consumed; GPV, Global Portion Value

Global portion value (GPV)

GPVs were derived using a stepwise systematic approach incorporating portion values from regulated serving sizes around the world, official guidance documents portion recommendations from quantitative FBDG, and typical consumption amounts from national nutrition surveys. The detailed methodology to establish globally relevant portion recommendations for a variety of packaged foods and beverages was recently published [21]. Briefly, for each food group, the GPVs are calculated as the median of all regulated serving size values (from Canada, Colombia, Indonesia, Brazil, Uruguay, Paraguay, Argentina, Mexico, South Korea, Turkey, USA, South Africa, and Thailand), and 90 FBDG aggregated into six regions (Africa, Europe, Western Pacific, South-East Asia, Americas, and Eastern Mediterranean). The calculated medians were then compared with intake data (from Europe and Australia) to ensure they were reasonable and credible in the context of a normal diet and rounded to a meaningful measure.

Substituting consumed portion size from discretionary food items with reference portion sizes

Diet modeling was used to substitute any amount above the reference portion with the reference portion value for selected discretionary food groups, without altering food choices. If the amount consumed per eating occasion was less than the reference portion, no changes were made. In the US, regulated serving sizes defined in Reference Amounts Customarily Consumed (RACCs) were used as the reference amounts for NHANES. GPVs were used as the reference amounts for NDNS and INCA3 when the portion consumed (per eating occasion) exceeded the reference portion size (Table 1). The choice of reference portion size in each cohort is in line with recent work on the methodology to define GPV [21], which specifies that regulated serving size must be applied in countries where they exist (e.g., the US in this study).

Indicators of nutritional quality

The nutritional quality of the diet of participants was comprehensively assessed using various indicators, such as the total intake of each single nutrient, percentage of energy from discretionary energy, adherence to the WHO dietary guidelines regarding sodium, fat, saturated fatty acids (SFAs), and sugar, and the nutrient density. These nutritional quality indicators were compared before and after substitution.

Nutritional content of the diet

This indicator assesses energy (in kcal/day) and nutrients (in grams/day), i.e., total fats, SFAs, polyunsaturated fatty acids (PUFAs), carbohydrate, total sugars, added sugar or free sugar, protein, fiber, and sodium (in mg/day). The assessment of these individual nutrients provides a comprehensive understanding of the nutritional content of the diet and allows to identify potential deficiencies or excesses in specific nutrients.

Adherence to WHO guidelines

The adherence to the World Health Organization dietary guidelines enabled us to evaluate diet quality based on the recommendations set by the World Health Organization. Adherence to the guidelines for total fat (≤ 30% of total energy intake) [35], saturated fat (≤ 10% of total energy intake) [36], free sugar (≤ 10% of total energy intake) [37], and sodium (≤ 2000 mg/day) [38]. For NHANES and INCA3, added sugar was used as a proxy for free sugar. The percentage of participants following these guidelines at baseline and after substitution were compared.

Nutrient density standardized to 2000 kcal diet

To assess changes in nutrient density before and after diet modeling, intakes of nutrients at baseline and after modeling for a 2000 kcal diet were estimated, including total fat, SFAs, PUFAs, carbohydrates, sugars, added/free sugars, fiber, protein, and sodium. This standardization allows for a comparison of the nutrient content of the diet across different energy levels and provides insights into the overall nutrient density of the diet.

Sociodemographic and anthropometric characteristics

Sociodemographic variables (age, sex, education level, and income level) were collected using a standardized approach in each dataset.

In NHANES, data were gathered via in-person interviews at participants' homes and mobile examination centers [39, 40]. Age was reported as a continuous variable, and sex as a categorical variable (male, female). Education level were coded separately for participants aged 6–19 and those aged 20 and older, then classified into four categories: “Less than high school graduate,” “High school graduate,” “Some college,” and “College graduate.” Income level was assessed using the family income-to-poverty ratio (PIR), classified as low (PIR < 1), medium (PIR 1–4), and high (PIR > 4).

In NDNS, data were collected through face-to-face interviews and self-completed questionnaires [41]. Age was coded as a continuous variable and sex as a categorical variable (male, female). Education level were coded into eight categories that were aligned to the four classes used in NHANES. Income level was categorized into tertiles of household income (lowest, middle, highest), which was adjusted for the number of people per household.

In INCA3, data was collected via structured interviews at participants' homes [23]. Age was categorized into three groups (18–44y, 45–64y, 65–79y) and sex as male or female. Education level were categorized into four classes: “Primary + middle school,” “High school,” “1–3 years of post-secondary education,” and “4 or more years of post-secondary education.” Income was reported per month per consumption unit and categorized into four groups: < 900 €, [900–1340[ €, [1340–1850[ €, and ≥ 1850 €.

Anthropometric measurements (height in cm and weight in kg) were primarily obtained through direct physical examinations by trained health professionals. In NHANES, measurements were taken at mobile examination centers [42]. Height was measured using a stadiometer, and weight using a calibrated digital scale, with participants barefoot and in light clothing. In NDNS, measurements were taken by trained interviewers during home visits using portable stadiometers and scales [24]. For participants aged 65 and older, or those without reliable measures, demi-span was used. In INCA3, measurements were taken during home visits [43]. Participants refusing to be measured had to self-report their measurements. Weight was measured to the nearest 0.1 kg using an electronic scale, and height to the nearest cm using a tape measure and set square.

BMI (kg/m²) was calculated as weight (kg) divided by height squared (m²). BMI categories were defined according to WHO guidelines [44]: “Underweight: < 18.5 kg/m²”, “Normal weight: 18.5–24.9 kg/m²”, “Overweight: 25–29.9 kg/m²”, and “Obese: ≥ 30 kg/m²”.

Statistical analyses

Participants’ demographic and anthropometric characteristics were described using means, standard deviations, and percentages. Statistical differences before and after portion substitution were tested using paired t-tests for total intakes, adherence to WHO guidelines, nutrient density standardized to a 2000 kcal diet, and percentage of energy from discretionary foods.

A series of linear regressions were conducted separately for each study cohort to determine if difference in daily total energy (energy after – energy before substitution) and difference in daily added/free sugar (added/free sugar after – added/free sugar before substitution) was associated with participants’ demographic and anthropometric characteristics. As age was only available as categorical variable in INCA3, age (continuous variable) in NHANES and NDNS was transformed to match the age groups in INCA3. For each outcome of interest and for each cohort, a model was fitted and adjusted for age, sex, Body Mass Index (BMI), level of education, income, and baseline daily energy intake or baseline added/free sugar intake. The results of linear regression analyses are presented as beta-coefficient (β) and 95% confidence interval (CI). In terms of results interpretation, a more positive β-coefficient means there is a “smaller reduction,” and a more negative β-coefficient implies “greater reduction” after portion substitution.

Statistical significance was set at p < 0.05 and all analyses were performed using R statistical software (version 4.4.0., Vienna Austria). To ensure reliable statistics representative of each country's population, appropriate survey weights, strata, and clusters were applied in all analyses to adjust for the complex sampling design of each cohort.

Results

Mean age of the NHANES was 47.6 ± 17.8y with 51.5% female (n = 4273); mean age of NDNS was 48.7 ± 18.1y with 51.2% female (n = 1912), and age distribution of INCA3 was 45.4% in 18–44, 36.8% in 45–64, 17.8% in 65-79y with 51.0% female (n = 2100). Among the three cohorts, NHANES had the highest proportion of individuals living with obesity (43.4%), followed by NDNS (24.4%) and INCA3 (17.1%). A quarter of participants in NHANES had normal weight (25.3%), while the corresponding proportion in NDNS and INCA3 were 36.5% and 44.9%, respectively. In NHANES and NDNS, almost half of participants received some education above high school (43.6% and 51.6%), while 33.8% of participants in INCA3 received at least 1 year of post-secondary education (Table 2).

Table 2.

Baseline sociodemographic and anthropometric characteristics of participants by study cohort

NHANES (n = 4273)	NHANES (n = 4273)	NDNS (n = 1912)	NDNS (n = 1912)	INCA3 (n = 2100)	INCA3 (n = 2100)
Age (years), mean (SD)	47.6 (17.8)	Age (years), mean (SD)	48.7 (18.1)	Age groups, %
				18–44 years	45.4
				45–64 years	36.8
				65–79 years	17.8
Female sex, %	51.5	Female sex, %	51.2	Female sex, %	51.0
Height (cm), mean (SD)	168.2 (10.0)	Height (cm), mean (SD)	168.9 (9.8)	Height (cm), mean (SD)	168.6 (9.5)
Weight (kg), mean (SD)	84.4 (22.6)	Weight (kg), mean (SD)	78.0 (18.2)	Weight (kg), mean (SD)	73.7 (15.4)
BMI (kg/m2), mean (SD)	29.7 (7.2)	BMI (kg/m2), mean (SD)	27.1 (5.3)	BMI (kg/m2), mean (SD)	25.9 (4.8)
Weight status, %		Weight status, %		Weight status, %
Underweight, < 18.5 kg/m2	1.5	Underweight, < 18.5 kg/m2	1.0	Underweight, < 18.5 kg/m2	3.1
Normal weight, 18.5 24.9 kg/m2	25.3	Normal weight, 18.5 24.9 kg/m2	36.5	Normal weight, 18.5 24.9 kg/m2	44.9
Overweight, 25–29.9 kg/m2	29.8	Overweight, 25–29.9 kg/m2	38.1	Overweight, 25–29.9 kg/m2	34.9
Obese, ≥ 30 kg/m2	43.4	Obese, ≥ 30 kg/m2	24.4	Obese, ≥ 30 kg/m2	17.1
Education level, %		Education level, %		Education level, %
Less than high school graduate	15.7	Less than high school graduate	14.9	Primary and middle school	48.1
High school graduate	40.6	High school graduate	33.5	High school	18.1
Some college	27.3	Some college	11.7	1–3y of post-secondary education	17.1
College graduate	16.3	College graduate	39.9	≥ 4y of post-secondary education	16.7
Income (Ratio of family income to poverty, PIR), %		Income (Equalized household tertiles), %		Income (per month, per consumption unit), %
Low income (PIR < 1)	13.2	Lowest tertile	27.5	< 900 €	24.6
Middle income (PIR between 1–4)	48.4	Middle tertile	33.9	[900–1 340[ €	23.1
High income (PIR > 4)	38.4	Highest tertile	38.7	[1 340–1 850[ €	23.6
				≥ 1 850 €	28.8

NHANES, National Health and Nutrition Examination Survey (wave 2017–2018); NDNS, The UK National Diet and Nutritional Survey (years 9–11, 2016/17 to 2018/19); INCA3, The Third French Individual and National Food Consumption (2014–2015)

In NHANES, NDNS, and INCA3, the average number of portion reductions per participant per day was 1.4 (SD 1.2), 1.3 (SD 1.2), and 1.0 (SD 0.9) times, respectively. Overall, 83.5%, 89.9%, and 84.0% of participants from NHANES, NDNS, and INCA3, respectively, had at least one of their discretionary food intakes substituted for the recommended portion. Results from diet modeling (i.e., substitution of individual intakes of discretionary foods which exceeded recommendations to the recommended portion size) are summarized in Table 3, 4, 5. Baseline daily energy intakes were 2073 ± 852 kcal/d among US participants, 1782 ± 546 kcal/d among UK participants, and 2058 ± 715 kcal/d among French participants. Overall, statistically significant reductions after portion substitution were observed in daily energy and individual nutrients in all three cohorts. Daily energy decreased by 7.8 percentage points (122 g/d, 161 kcal/d) in NHANES, by 4.0 percentage points (86 g/d, 75 kcal/d) in NDNS, and by 4.9 percentage points (46 g/d, 102 kcal/d) in INCA3. Notably, the proportion of daily energy from discretionary foods decreased by 5.4 percentage points in NHANES (mean ± SD from 20.7 ± 13.1 to 15.3 ± 9.8%, p < 0.001), by 3.1 percentage points in NDNS (from 15.6 ± 9.8 to 12.5 ± 8.0%, p < 0.001), and 3.6 percentage points in INCA3 (from 16.1 ± 10.7 to 12.5 ± 8.1%, p < 0.001). These reductions represent 26.0%, 19.9%, and 22.3% of the initial percentage of discretionary energy for NHANES, NDNS, and INCA3, respectively. In NHANES, added sugar decreased by 18.6% (from 67.3 ± 58.7 to 54.8 ± 42.8 g, p < 0.001), while in INCA3 it decreased by 14.2% (from 42.6 ± 31.1 to 36.6 ± 25.3 g, p < 0.001). NDNS observed an 11.6% reduction in free sugar (from 47.4 ± 33.3 to 41.9 ± 27.5 g, p < 0.001). Saturated fat significantly decreased by 7.5%, 4.9%, and 6.1% in NHANES, NDNS, and INCA3, respectively. Additionally, sodium levels decreased by 6.4%, 3.9%, and 3.0% after the substitution in NHANES, NDNS, and INCA3, respectively.

Table 3.

Diet modeling results—NHANES, substitution with RACCs in discretionary food categories

	Baseline	Baseline	After substitution	After substitution	Difference	Difference	Difference
	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total intake/weight (g/d)	3337	1348	3214	1316	− 122	− 3.7	< 0.001
Energy intake (kcal/d)	2073	852	1912	739	− 161	− 7.8	< 0.001
Energy from discretionary foods (%E/d)	20.7	13.1	15.3	9.8	− 5.4	− 26.0	< 0.001
Total fat (g/d)	84.6	40.6	77.4	35.6	− 7.2	− 8.5	< 0.001
Saturated fat (g/d)	27.6	14.3	25.5	12.8	− 2.1	− 7.5	< 0.001
Polyunsaturated fat (g/d)	19.9	11.0	17.7	9.3	− 2.2	− 11.0	< 0.001
Total carbohydrate (g/d)	239.1	109.2	215.9	89.8	− 23.1	− 9.7	< 0.001
Total sugar (g/d)	102.7	65.8	89.0	50.4	− 13.7	− 13.3	< 0.001
Added sugar (g/d)	67.3	58.7	54.8	42.8	− 12.5	− 18.6	< 0.001
Total protein (g/d)	80.4	35.1	78.4	34.4	− 1.9	− 2.4	< 0.001
Total fiber (g/d)	16.5	9.0	15.6	8.6	− 0.9	− 5.4	< 0.001
Sodium (mg/d)	3391	1486	3173	1356	− 218	− 6.4	< 0.001
Adherence to recommendations	%		%				P-value
< 30% of total fat (N, %)	18.8		19.5				0.176
< 10% of saturated fat (N, %)	29.0		28.7				0.682
< 10% of added / free sugars (N, %)	45.3		50.7				< 0.001
< 2000 mg of sodium (N, %)	15.2		18.1				< 0.001
Nutritional density (/2000 kcal)	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total fat (g/2000 kcal)	80.9	16.9	80.3	16.9	− 0.6	− 0.7	< 0.001
Saturated fat (g/2000 kcal)	26.3	7.2	26.4	7.3	0.1	0.3	0.135
Polyunsaturated fat (g/2000 kcal)	19.1	6.9	18.5	6.7	− 0.6	− 3.1	< 0.001
Carbohydrate (g/2000 kcal)	232.0	49.0	228.6	48.9	− 3.5	− 1.5	< 0.001
Total sugar (g/2000 kcal)	98.6	43.2	94.5	40.6	− 4.1	− 4.2	< 0.001
Added / Free sugar (g/2000 kcal)	62.9	42.6	57.3	38.2	− 5.5	− 8.8	< 0.001
Fiber (g/2000 kcal)	16.5	7.6	16.8	7.7	0.3	1.7	< 0.001
Protein (g/2000 kcal)	79.4	21.9	83.0	21.8	3.6	4.6	< 0.001
Sodium (mg/200 kcal)	3336	846	3368	826	31.5	0.9	< 0.001

Table 4.

Diet modeling results—NDNS, substitution with GPVs in discretionary food categories

	Baseline	Baseline	After substitution	After substitution	Difference	Difference	Difference
	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total intake/weight (g/d)	2893	959	2807	937	− 86	− 3.0	< 0.001
Energy intake (kcal/d)	1782	546	1710	512	− 75	− 4.0	< 0.001
Energy from discretionary foods (%E/d)	15.6	9.8	12.5	8.0	− 3.1	− 19.9	< 0.001
Total fat (g/d)	68.3	25.9	64.9	24.1	− 3.5	− 5.1	< 0.001
Saturated fat (g/d)	25.1	11.1	23.9	10.5	− 1.2	− 4.9	< 0.001
Polyunsaturated fat (g/d)	16.0	10.2	15.3	9.8	− 0.7	− 4.4	< 0.001
Total carbohydrate (g/d)	212.5	73.4	202.7	67.5	− 9.8	− 4.6	< 0.001
Total sugar (g/d)	84.7	40.4	78.6	35.7	− 6.1	− 7.2	< 0.001
Free sugar (g/d) UK	47.4	33.3	41.9	27.5	− 5.5	− 11.6	< 0.001
Total protein (g/d)	73.8	23.9	72.9	23.7	− 0.9	− 1.3	< 0.001
Total fiber (g/d)	19.2	8.0	18.8	7.9	− 0.4	− 2.2	< 0.001
Sodium (mg/d)	1969	764	1892	730	− 77	− 3.9	< 0.001
Adherence to recommendations	%		%				P-value
< 30% of total fat (%)	23.4		25.4				< 0.001
< 10% of saturated fat (%)	23.0		24.6				< 0.001
< 10% of added / free sugars (%)	53.9		60.4				< 0.001
< 2000 mg of sodium (%)	57.2		61.2				< 0.001
Nutritional density (/2000 kcal)	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total fat (g/2000 kcal)	76.0	14.4	75.2	14.4	− 0.8	− 1.0	< 0.001
Saturated fat (g/2000 kcal)	27.9	7.9	27.7	8.0	− 0.2	− 0.8	< 0.001
Polyunsaturated fat (g/2000 kcal)	17.4	8.1	17.4	8.2	0.0	− 0.1	0.458
Carbohydrate (g/2000 kcal)	239.8	40.7	238.7	40.7	− 1.1	− 0.4	0.001
Total sugar (g/2000 kcal)	94.9	32.4	92.2	31.5	− 2.7	− 2.8	< 0.001
Added / Free sugar (g/2000 kcal)	51.7	28.9	48.2	26.5	− 3.5	− 6.8	< 0.001
Fiber (g/2000 kcal)	22.1	7.7	22.4	7.7	0.3	1.5	0.001
Protein (g/2000 kcal)	84.8	21.3	87.0	21.4	2.1	2.5	< 0.001
Sodium (mg/200 kcal)	2230	616	2230	616	0.3	0.0	0.916

Table 5.

Diet modeling results—INCA3, substitution with GPVs in discretionary food categories

	Baseline	Baseline	After substitution	After substitution	Difference	Difference	Difference
	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total intake/weight (g/d)	2963	863	2917	852	− 46	− 1.5	< 0.001
Energy intake (kcal/d)	2058	715	1956	655	− 102	− 4.9	< 0.001
Energy from discretionary foods (%E/d)	16.1	10.7	12.5	8.1	− 3.6	− 22.3	< 0.001
Total fat (g/d)	76.0	30.4	71.2	27.2	− 4.8	− 6.3	< 0.001
Saturated fat (g/d)	32.0	14.3	30.1	13.0	− 1.9	− 6.1	< 0.001
Polyunsaturated fat (g/d)	9.9	4.8	9.2	4.3	− 0.6	− 6.5	< 0.001
Total carbohydrate (g/d)	231.0	91.4	218.3	83.7	− 12.8	− 5.5	< 0.001
Total sugar (g/d)	93.6	43.0	86.3	37.5	− 7.3	− 7.8	< 0.001
Added sugar (g/d)	42.6	31.1	36.6	25.3	− 6.1	− 14.2	< 0.001
Total protein (g/d)	81.1	27.6	79.6	27.1	− 1.5	− 1.9	< 0.001
Total fiber (g/d)	19.9	7.4	19.2	7.1	− 0.7	− 3.3	< 0.001
Sodium (mg/d)	3141	1318	3045	1288	− 95	− 3.0	< 0.001
Adherence to recommendations	%		%				P-value
< 30% of total fat (%)	30.5		32.9				< 0.001
< 10% of saturated fat (%)	11.7		12.5				0.011
< 10% of added / free sugars (%)	71.6		77.0				< 0.001
< 2000 mg of sodium (%)	17.9		20.2				< 0.001
Nutritional density (/2000 kcal)	Mean	SD	Mean	SD	Mean dif	% dif	P-value
Total fat (g/2000 kcal)	73.8	13.7	73.0	13.8	− 0.8	− 1.1	< 0.001
Saturated fat (g/2000 kcal)	30.9	7.7	30.6	7.7	− 0.3	− 1.0	< 0.001
Polyunsaturated fat (g/2000 kcal)	9.7	3.5	9.6	3.4	− 0.1	− 1.4	< 0.001
Carbohydrate (g/2000 kcal)	223.6	36.2	222.4	36.7	− 1.1	− 0.5	< 0.001
Total sugar (g/2000 kcal)	91.6	28.9	89.4	28.8	− 2.2	− 2.4	< 0.001
Added / Free sugar (g/2000 kcal)	40.2	23.3	37.0	21.5	− 3.3	− 8.1	< 0.001
Fiber (g/2000 kcal)	19.8	5.7	20.1	5.7	0.3	1.3	< 0.001
Protein (g/2000 kcal)	81.2	18.9	83.3	18.8	2.1	2.6	< 0.001
Sodium (mg/200 kcal)	3099	925	3143	926	44.0	1.4	< 0.001

In general, statistically significant improvements were observed in adherence to WHO guidelines after substitution (i.e., proportion of participants adhering to each of the WHO guidelines increased after portion reduction to the reference amounts) (Tables 3, 4, 5).

The nutritional density of a 2000 kcal diet was compared before and after substitution, revealing a significant decrease in public health sensitive nutrients such as total fat, total sugar, and added/free sugar. The most notable decrease was observed for added/free sugar, with reductions of 8.8%, 6.8%, and 8.1% among NHANES, NDNS, and INCA3 participants, respectively. On the other hand, the nutritional density of fiber and protein significantly increased in all three cohorts after portion substitution. Specifically, when daily energy intake was standardized to 2000 kcal, there were 1.7%, 1.5%, and 1.3% increases in fiber, and 4.6%, 2.5%, and 2.6% increases in protein among NHANES, NDNS, and INCA3 participants, respectively.

The association between the difference in daily total energy and the difference in daily added/free sugar (before and after substitution) with demographic and anthropometric characteristics is presented in Table 6. Since the substitutions were performed with lower portion sizes of discretionary foods, the daily amounts of total energy and added/free sugar cannot increase after these substitutions. Consequently, the regression outcome will always be less than or equal to zero. Older age was associated with smaller reduction in energy in the fully-adjusted model for NDNS (45–64y vs. 18–44y: β-coefficient (95% CI) = 16.1 (4.6, 27.5), p = 0.006; 65–79y vs. 18–44y: 35.9 (24.5, 47.4), p < 0.001; 80 + y vs. 18–44y: 35.4 (19.0, 51.9), p < 0.001) and INCA3 (45–64y vs. 18–44y: 51.2 (30.5, 71.8), p < 0.001; 65–79y vs. 18–44y: 62.5 (42.2, 82.7), p < 0.001). Females exhibited greater reductions in daily energy in INCA3 (−28.4 kcal (−48.1, −8.7), p = 0.005) compared to males. The percentage of variation in reduction of total energy (R²) explained by the independent variables ranged between 0.2 and 0.4.

Table 6.

Association between the difference in daily total energy intake and the difference in daily added/free sugar intake (before and after substitution) with demographic characteristics

		NHANES (n = 4273)			NDNS (n = 1912)	INCA3 (n = 2100)
Difference in		β-coefficient (95% CI)	P-value	R2	β-coefficient (95% CI)	P-value	R2	β-coefficient (95% CI)	P-value	R2
Energy	Age groupsa (Ref. 18–44)
	45–64	− 17.6 (− 52.8, 17.5)	0.163		16.1 (4.6, 27.5)	0.006		51.2 (30.5, 71.8)	< 0.001
	65–79	− 30.4 (− 92.4, 31.7)	0.170		35.9 (24.5, 47.4)	< 0.001		62.5 (42.2, 82.7)	< 0.001
	80 +	− 29.5 (− 103.0, 44.0)	0.227		35.4 (19.0, 51.9)	< 0.001		NA
	Sex (female vs male)	− 30.5 (− 62.4, 1.3)	0.054		0.5 (− 8.9, 9.9)	0.915		− 28.4 (− 48.1, − 8.7)	0.005
	Baseline energy intake	− 142 (− 178.5, − 105.4)	< 0.001		− 35.7 (− 42.2, − 29.2)	< 0.001		− 81.2 (− 95.9, − 66.6)	< 0.001
				0.4			0.2			0.3
Added/Free sugarb	Age groupsa (Ref. 18–44)
	45–64	1.0 (− 3.5, 5.6)	0.433		1.0 (0.1, 1.9)	0.024		1.3 (0.1, 2.5)	0.029
	65–79	− 0.4 (− 5.7, 4.9)	0.782		3.2 (2.3, 4.1)	< 0.001		1.2 (− 0.1, 2.5)	0.063
	80 +	− 0.9 (− 5.1, 3.4)	0.475		3.9 (2.6, 5.2)	< 0.001		NA
	Sex (female vs male)	0.5 (− 2.3, 3.2)	0.533		0.0 (− 0.8, 0.7)	0.954		− 0.6 (− 1.5, 0.2)	0.139
	Baseline free/added sugar intake	− 19.4 (− 28.8, − 10.0)	0.013		− 5.0 (− 5.7, − 4.3)	< 0.001		− 7.4 (− 8.7, − 6.0)	< 0.001
				0.6			0.4			0.1

All models are additionally adjusted with BMI, education level, and income status

^aAge range in INCA3was 18–79y

^2bAdded sugar in NHANES and INCA3, Free sugar in NDNS

In NDNS, older age was associated with smaller reduction in free sugar when adjusted for baseline free sugar (45–64y vs. 18–44y: β-coefficient (95% CI) = 1.0 (0.1, 1.9), p = 0.024; 65–79y vs. 18–44y: 3.2 (2.3, 4.1), p < 0.001; 80 + y vs. 18–44y: 3.9 (2.6, 5.2), p < 0.001), and the same was observed in INCA3 among middle-aged participants (45-64y old) when compared to younger participants (18–44y old): 1.3 (0.1, 2.5), p = 0.029. The percentage of variation in reduction of added/free sugar explained by the independent variables (R²) ranged between 0.1 and 0.6.

Discussion

Adherence to reference portion guidance for discretionary foods—whether derived from regulatory standards or calculated GPVs—was associated with improved overall diet quality. This approach also led to reductions in nutrients of public health concern, including saturated fat, free/added sugars, and sodium, and enhanced alignment with WHO dietary recommendations. Notably, older participants in the UK and France exhibited smaller reductions in discretionary energy and free/added sugars after substitution.

In the current study, baseline discretionary energy accounted for 15.6–20.7% of daily energy intake depending on the country, which is slightly lower when compared to values reported from Australia (32.5%) [12], Mexico (26%) [11] and earlier waves of US NHANES (2003–2012) (23%) [10]. One reason could be the slightly different classification of discretionary foods in each study. For example, processed meat and alcohol were not included as discretionary foods in this study, but other studies did include these items [12, 45]. However, there are currently no reproducible and quantitative rules to identify discretionary foods and no consensus on their definition [5], so this variation across studies is expected. With the food categories that were included, substantial reductions in discretionary energy were observed among three countries (ranging from −19.9 to −26.0%), implying potentially more significant reduction if a wider range of discretionary food and beverage categories were to be included in other studies.

In line with a recent meta-analysis from 14 experimental studies, current results showed that adopting smaller portions was associated with reduction in daily energy. However, the magnitude of reduction is smaller (NHANES: 161 kcal/d; NDNS: 75 kcal/d; INCA3: 102 kcal/d) compared to their average reduction of approximately 235 kcal per day in intervention studies [14]. The reason for this difference could be due to inclusion of studies with diverse participant groups and various study designs (e.g., within-subjects design and between-subjects design). In comparison, the current study went further by providing specific results related to the reduction of discretionary foods and the consequent reduction of public health sensitive nutrients without changing individual food choices.

While it is well-documented that portion reduction can contribute to reduction in daily energy and nutrient intakes [14], the utilization of a diet modeling approach in this context has been limited. In a 2021 study, diet modeling methods were employed to assess strategies for reducing discretionary food and beverage consumption among 12,153 Australians aged 2–85y [45]. These strategies, encompassing 26 scenarios, targeted reductions in amount, frequency, or variety of discretionary items. Each scenario estimated the percentage reduction in energy and nutrient intakes. When discretionary items were limited to a maximum portion size of 143 kcal (aligned with Australian dietary guidelines [1]), significant reductions in public health sensitive nutrients were observed (−11.3% for sugar, −15.0% for saturated fat, and −11.6% for sodium). Even greater reductions were observed when individual intakes of discretionary items were halved (−22.8% for sugar, −19.4% for saturated fat, and −16.8% for sodium). While these findings align with the current analyses, specific portion sizes for each discretionary food and beverage category were employed in the current study, making it more realistic and less restrictive.

Implications

Results from the current study suggest that adhering to RACCs in the US and GPVs for UK and France would lead to significant reductions of discretionary energy to a reasonable level (between 12.5 and 15.3% of daily energy). Given the “relatively small behavioral change” applied (i.e., only portion modification for discretionary foods without changing food choices), the observed effect of reduction in discretionary energy and increased adherence to WHO guidelines for added sugar and sodium at a population level shows potential for public health impact.

Few countries provide quantitative upper limits on discretionary energy intake. In most EU countries, qualitative recommendations are provided in FBDGs to limit the consumption of discretionary foods, including crisps, cakes and sweets, or limiting public health sensitive nutrients, such as free sugars to 10% of total energy intake [46]. However, at the consumer level, effectively monitoring and restricting the total intake of discretionary energy remains challenging. This emphasizes the importance of consistent on-pack information on portion sizes which enable consumers to make healthier portion choices [47]. Additionally, harmonizing portion sizes across regions can improve nutritional education and facilitate better communication of dietary information [20]. Standardized portion sizes play a vital role in promoting consumer well-being and balanced diets [20].

Older age was associated with smaller reductions in discretionary energy and added/free sugar among participants in UK and France even after adjustment for corresponding baseline intakes. This can be explained by the smaller number of discretionary foods consumed in older age compared to younger adults (18–44y). Similarly, a recent study including 235,203 Australian adults found that young to middle-aged adults (18–50y) consume more discretionary food and beverages compared to older adults (51 + y) [48]. In addition, in all three cohorts, the frequency of exceeding reference portion sizes per day was less for 65 + y (0.78 times/d) compared to 18–44y (0.90 times/d) and 45–64y (0.85 times/d).

Strengths and limitations

The current study has several strengths. It is the first to apply substitution analyses to evaluate the effect of adhering to food-category-specific portion guidance for discretionary foods. The study is based on three large-scale population-based cohorts, with a total of over 8000 participants. The inclusion of one country with serving size regulations and two non-regulated countries in the analyses adds to the robustness of the sample. Survey weighting factors were applied to ensure the results are representative of the general population in each country. Furthermore, the study considered various indicators of diet quality, providing a comprehensive overview of the diet and how it would evolve after following portion guidance for discretionary foods.

While this study provides valuable insights, the following limitations should be considered when interpreting the results of this study. The dietary intakes from the NDNS are likely to be underestimated due to the method used to assign intake amounts, resulting in an attenuated improvement of diet quality in this population. However, this also suggests that in the overall population, the effects would be even more important in terms of public health, with a greater decrease in saturated fat, sodium, and free sugar, among other factors. Another limitation is that the study only captures the excessive intake of discretionary foods at the occasion of consumption level, meaning that if a participant has multiple intakes of a discretionary food under the portion limit, it would not be substituted. Additionally, the models used in the study were aimed to simulate the effect of portion size reduction but did not account for possible compensatory effects of consuming other foods following this procedure, which may occur in practice. However, it is important to note that the study is based on a theoretical model and the reason for not choosing an isocaloric model is to evaluate the maximum potential effect of a relatively small behavioral change (i.e. food category-specific portion reduction on discretionary foods without changing food choices) on overall nutrient intake and diet quality. To address this limitation, the nutrient density was considered by standardizing to a 2000kcal diet.

Conclusions

This study found that implementing portion size guidance for discretionary foods resulted in a significant reduction in public health sensitive nutrients and improved adherence to dietary guidelines. These findings show that modifying portion sizes without altering food preferences holds promise for improving intakes of public health sensitive nutrients. Future research is warranted to explore the applicability of these findings to other geographical regions and to investigate potential associations with health outcomes. This would contribute to a more comprehensive understanding of the impact of portion size guidance on dietary habits and overall health.

Author contributions

ALE, LTT, CD, FV, and MM designed research; LTT, CD, ES, and FM analyzed data; and LTT, CD, ALE wrote the paper; FV, MM reviewed the manuscript. LTT had primary responsibility for final content. All authors read and approved the final manuscript.

Funding

This research was funded by Nestlé Research, Société des Produits Nestlé, Vevey, Switzerland.

Data availability

Data described in the manuscript are freely accessible (NHANES, INCA3) or needed to be requested directly to the data owner (NDNS). Analytic code will be made available upon request from the authors.

Declarations

Conflict of interest

LTT, CD, ES, FM, ALE are employees of Nestlé Research. FV and MM received a consultancy fee from Nestlé Research.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the authors used NesGPT (OpenAI’s GPT-4 language model for Nestlé corporate use) in order to finetune R codes in data analysis and to improve readability of some sentences of the manuscript. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.