The Effect of Eating Rate of Ultra‐Processed Foods on Dietary Intake, Eating Behaviour, Body Composition and Metabolic Responses—Rationale, Design and Outcomes of the Restructure Randomised Controlled Trial

ABSTRACT

Extensive epidemiological research and the findings from one randomised controlled feeding trial (RCT) have shown associations between the consumption of ultra‐processed foods (UPFs) and higher energy intakes. To date the specific properties of UPF foods and diets that may be responsible for driving higher energy intakes remain unclear. A comparison of the diets in the single RCT to date showed a significantly higher eating rate (g/min, ER) for meals in the UPF diet compared to those in the minimally processed diet. Numerous acute feeding trials have shown that foods with textures that promote a slower ER tend to be consumed in smaller portions compared to those consumed at a faster ER. Here, we describe the rationale and approach for the Restructure RCT with the primary aim to determine the effect of meal texture‐derived ER (g/min) of UPF diets (‘UPF Slow ER’ vs. ‘UPF Fast ER’) on daily ad libitum energy intake (kcal/day) across a 14‐day period. The secondary objectives of the Restructure RCT are to compare changes in body composition and metabolic markers following the same two diets. We hypothesise that texture‐derived differences in ER will moderate food and energy intakes from UPF diets such that participants will consume less when they encounter textures that promote a slower eating rate. The outcomes of the Restructure RCT aim to provide new insights on the proposed mechanisms by which UPF are thought to promote greater energy intakes, and aid in the development of food‐based strategies to moderate dietary energy intakes from processed foods.

Trial Registration: clinicaltrials.gov identifier: NCT06113146

1. Introduction

A growing body of dietary epidemiological studies and the results from a single randomised controlled trial (RCT) have demonstrated an association between diets dominated by foods that are classified as ultra‐processed (UPF) and greater energy intake, body weight gain and a broad range of diet‐related diseases (Askari et al. 2020; Hall et al. 2019). The Nova scheme defines UPF as ‘industrial formulations that are made entirely from food derivatives, chemical substances and sequence of processes, that bears little resemblance to the original food material’ (Monteiro et al. 2019). The single inpatient RCT to date showed that over 2 weeks, there was a net difference of 508 kcal/day greater energy intake between the UPF diet compared to the minimally processed diet, resulting in a 1.8 kg change in body weight (Hall et al. 2019). Participants ate less and lost weight on the minimally processed diet arm, and ate more and gained weight on the UPF diet. Despite extensive profiling of metabolic parameters, no significant changes in fasting metabolic markers of health, taste threshold or preferences were observed following the UPF diet compared to the minimally processed diet (Jaime‐Lara et al. 2023). The lack of differences in reported taste perception, food preferences and fasting metabolic markers between minimally processed and UPF diets suggests that the observed increases in energy intake may be driven by sensory and eating behaviours, particularly the physical properties of food—such as texture and eating rate—rather than underlying metabolic dysregulation. The large differences in energy intake and lack of a clear metabolic explanation for the observed differences in the trial has stimulated widespread debate and research prioritisation on the putative mechanisms specific to UPF that underpin the observed differences in energy intake (Forde 2023; O'Connor et al. 2023; Trumbo et al. 2024; Valicente et al. 2023).

The mechanism(s) underlying the observed differences in energy intake of UPF are difficult to elucidate due to the broad nature of the Nova category as it captures a wide diversity of foods, nutrients and processes (Gibney and Forde 2022; Monteiro et al. 2019). It is currently not known whether there are particular nutritive, non‐nutritive or sensory properties commonly shared by UPF that may drive higher energy intakes. The Nova scheme initially proposed that the health effects of consuming UPF were independent of their nutrient properties and were instead driven by the degree to which a food is processed (Monteiro 2009). However, concerns have been raised that by combining formulation and food processing, the scheme may attribute observed effects to food processing when they are more likely driven by differences in nutrient and energy content (i.e., formulation) (Levine and Ubbink 2023; Rolls et al. 2020). To date, there has not been a clear demonstration that processing has an independent impact on health outcomes for foods of identical nutrient quality (Robinson and Jones 2024). Increasingly, observational association studies suggest that the link between UPF intake and health is driven by the consumption of specific food groups, such as processed and preserved meats and sugar‐sweetened beverages and desserts (Kliemann et al. 2023). In addition to nutrient properties of food, the sensory properties of a food can be considered non‐nutrient factors that influence eating behaviours and moderate meal size (Forde and de Graaf 2022, 2023). Two of such properties that have been extensively studied in relation to energy intake are food texture‐derived differences in eating rate (g/min, ER) and energy density (kcal/g) (Robinson et al. 2014). Energy‐dense (kcal/g) and softly textured foods that require little oral processing (oral break down of food through chewing and mixing food particles with saliva) and can be consumed quickly have been shown to promote higher energy intakes through an increased energy intake rate expressed as calories consumed per minute (kcal/min, EIR) (de Graaf 2011; McCrickerd et al. 2017; Teo et al. 2021). UPF and minimally processed diets in the previous RCT by Hall and colleagues differed considerably in their observed ER and EIR, with meals in the UPF diet arm consumed approximately 50% faster in terms of their EIR (kcal/min) compared to those of the minimally processed diet (48 vs. 31 kcal/min) (Hall et al. 2019). A higher energy intake rate can be achieved by preferentially eating foods with soft textures that can be consumed at a fast ER and have a high energy density, over hard textured‐lower energy dense meal components, and has been proposed as a possible mechanism that underlies the sustained higher energy intakes on the UPF diet in that study (Gibney and Forde 2022).

The role of energy density in promoting higher intake has been well established. Evidence accumulated over many years of controlled research studies has demonstrated both an acute and prolonged effect of higher energy density on increased energy intakes, across single meals, meal occasions, food forms and formats, in adults, children, across weight classes and genders, at the level of the meal, the day and the diet; this effect is shown to be sustained (Burns et al. 2022; Raynor et al. 2012; Robinson et al. 2022; Rolls 2009; Rolls et al. 2024, 2007). Less is understood about the sustained effect of food texture‐based differences in ER on food intake. Two recent acute controlled feeding trials have compared the relative influence of degree of processing and ER on ad libitum energy intake for meals that vary in food texture (hard texture, slow ER vs. soft texture, fast ER) and degree of food processing (minimally vs. UPF) (Lasschuijt et al. 2023; Teo et al. 2022). Comparison of ad libitum energy intake for a single meal showed that food texture‐based differences in ER had a significant effect on food (g, decrease of 21%) and energy intake (kcal, decrease of 26%) for both minimally processed and UPF meals (Teo et al. 2022). A follow up trial showed similar results across meals over the course of a single test day, where food texture‐derived reductions in eating speed significantly reduced amount (g, decrease of 14%) and energy (kcal, decrease of 33%) consumed, independently of processing level (Lasschuijt et al. 2023). To date, only acute studies have demonstrated the impact of texture‐based differences in ER on meal and day‐level differences in intake, and longer‐term trials are now needed to understand whether texture‐based differences in ER could have a sustained impact on energy intakes (Hall et al. 2019).

The current protocol describes the rationale, design and experimental approach for a controlled feeding trial to test the sustained impact of ER on energy intakes from UPF diets at the level of the meal, the day and cumulatively over a 14‐day intervention period. The primary objective of the study is to determine whether texture‐derived differences in ER (g/min) of ultra‐processed diets (‘UPF Fast ER’ vs. ‘UPF Slow ER’) has a significant impact on daily ad libitum energy intakes (kcal/day), across a 14‐day intervention period. To address potential metabolic targets linking UPF consumption to ill health (Fazzino et al. 2019; Juul et al. 2021; Leo and Campos 2020; Martins et al. 2022; Mignogna et al. 2022; Tobias and Hall 2021), the secondary objectives of the Restructure RCT compare body composition and pre‐specified metabolic changes when on a 14‐day UPF diet consisting of food textures that promote either a slow or a fast ER.

2. Objectives and Hypotheses

The primary objective of the Restructure RCT is to determine the effect of meal texture‐derived differences in ER (g/min) on daily ad libitum energy intakes (kcal/day) across a 14‐day period for diets composed of ultra‐processed foods (UPFs). We hypothesise that the energy intake from UPF diets is moderated by eating speed such that when meals are consumed at a slower ER, they will lead to a lower daily energy intake compared to UPF diets consumed with a faster ER.

The secondary objectives of the Restructure RCT are to compare body composition and pre‐specified metabolic changes when on a 14‐day diet of UPF consisting of food textures that promote either a slow or a fast ER. Our hypothesis is that a diet of UPF with a slow ER will lead to sustained reductions in energy intake and support a reduction in body weight and adiposity compared to a UPF diet with a faster ER. We hypothesise that there will be no differences in metabolic outcomes between the two diets.

3. Materials and Methods

3.1. Controlled Feeding Study Design

The Restructure RCT is a single‐blind, block randomised cross‐over trial with two diet periods, that is conducted at the Health Research Unit of Wageningen University and Research, The Netherlands. The two diet periods include a 14‐day ultra‐processed diet with a texture‐derived slow ER (‘UPF Slow ER’) and a 14‐day ultra‐processed diet with a texture‐derived fast ER (‘UPF Fast ER’). Participants have a one‐week baseline period, during which we determine their habitual dietary habits and collect baseline measures, followed by a 14‐day diet period on one intervention diet, a washout period of 14‐days between the two diets to reduce carry‐over effects, and finally a second 14‐day diet period on the second intervention diet (Figure 1). All participants receive both diets (within‐subject design) and comparisons are drawn between their food and energy intake, changes in body composition and changes in metabolic markers of health after each 14‐day diet period.

FIGURE 1.

Overview of study design. UPF Slow ER, diet consisting of ultra‐processed foods with a texture‐derived slow eating rate. UPF Fast ER, diet consisting of ultra‐processed foods with a texture‐derived fast eating rate.

During the 14‐day diet period, energy intake at every meal and snack is measured and body composition is measured at baseline (T1), after intervention period 1 (T2), at the end of the washout prior to diet period 2 (T3) and after the second diet period intervention on completion of the study (T4) (Figure 2). Metabolic responses are measured at T1, T2 and T4. To reduce participant burden, the decision was taken not to re‐test at T3 after the 2‐week washout. The duration of the washout was based on previous studies with the same primary outcome (ad libitum intake) (Griffioen‐Roose et al. 2012; Rolls et al. 1999, 2007). The study is approved by the Medical Ethical Committee ‘East‐Netherlands’, The Netherlands (ABR: NL83462.091.23) and complies with the Declaration of Helsinki for Medical Research involving Human Subjects. The data analyses plan is pre‐registered at Open Science Framework (https://osf.io/afmkh).

FIGURE 2.

Overview of study design and measures. Blue dots in the faecal sample collection indicate the blue muffin test (blue dye method). The weighed food intake weekdays 1–5 and 8–12 participants consumed all of their main meals in the eating behaviour lab (dark blue). On weekend days 6, 7, 13 and 14 participants were provided with all the meals and snacks to consume them at home, participants were asked to return all leftovers in the original package for post‐weighing to determine the amount consumed (mint green). EOS, End of study; FFQ, Food Frequency Questionnaire.

3.2. Sample Size Calculation

The sample size estimation is based on the primary outcome which is the between‐diet difference in ad libitum average daily energy intake (kcal/day) across the 14 days. The estimated effect size (130 ± 280 kcal/day, dz. of 0.54) is determined based on a previous study with 1‐day diets differing in ER (Lasschuijt et al. 2023) and a previous randomised controlled 2‐week trial comparing ultra‐processed vs. minimally processed diets (Hall et al. 2019). We anticipate a smaller effect size with more variation around the mean since texture‐derived differences in eating rate are a behavioural manipulation shown to produce consistent effects on intake, but known to be variable between people (Hall et al. 2019; McCrickerd and Forde 2017). Based on a power analysis in G*Power (Windows version 3.1.9.7; Heinrich Heine Universität Düsseldorf, Düsseldorf, Germany), 39 participants are needed to achieve a 1‐β power of 80% to detect differences (α = 0.05, two tailed).

3.3. Study Participants

To be included in the trial participants must be aged between 21 and 50 years old and have a BMI between 21 and 27 kg/m², report having a normal appetite (self‐reported, yes/no) and generally eat three meals a day around the same times, and self‐reported to be in good general and mental health ascertained using the following set of questions: ‘Do you consider yourself healthy at the moment?’—if not please indicate why; ‘Do you have trouble chewing, swallowing or eating’—is so indicate why/what; ‘Do you have any smell or taste problems’. If yes please indicate what ‘Do you suffer/have you suffered from hormonal disorders/eating disorders such as anorexia nervosa or bulimia/chronic illness of the gastrointestinal tract/thyroid disease/diabetes or respiratory diseases’ (asked individually), as well as objective measures detailed below. Participants are excluded if they are on a vegetarian or vegan diet, are allergic or intolerant to any of the test foods, are smokers, use any type of chronic medication other than contraceptives, consume more than 21 units/week (men) or more than 14 units/week (women) of alcohol, gained or lost ≥ 5 kg of body weight over the last half year or if they do more than 4 h/week moderate‐to‐vigorous physical activity.

Blood samples are drawn after an overnight fast to determine resting glucose (inclusion range; 3.5–8 mmol/L, based on the values physicians in the Netherlands take to classify blood glucose levels as either hypoglycaemic or hyperglycaemic) and haemoglobin (inclusion range; 7.5–11.0 mmol/L female and 8.5–11.0 mmol/L male) (DutchDiabetesFoundation, n.d.; Homephysician.nl, n.d.). As a screening for overall health, blood pressure (Omron Healthcare Europe B) is measured to ensure participants blood pressure was not below the normal range (low blood pressure increases fainting risk during blood withdrawal [below 90 and/or below 60 mm hg]) (DutchHeartFoundation, n.d.). Additionally, body weight (grams) (SECA 635 platform scale) and height (mm) (SECA stadiometer) are measured in duplicate to calculate BMI (kg/m²) and participants are excluded if this was outside the inclusion range (21–27 kg/m²).

To evaluate participants habitual consumption of UPF a newly developed UPF food frequency questionnaire (UPFFQ) was administered to determine amount (gram), energy (kcal) and % macronutrient intakes and % energy consumed from UPF (van Bruinessen et al. 2025, under review). Based on two previously conducted cohort studies the median daily energy % derived of UPF in the Netherlands was estimated to be 50% (Duan et al. 2022; Pinho et al. 2021). To ensure a homogeneous study population, participants are excluded if their UPF intake exceeds 50% of their total dietary food intake.

Participants' natural eating rate is determined using behavioural annotation of video recordings when consuming four (two hard raw, two soft boiled) 15 g carrot sticks (width and depth of 13 mm; 1 in.) (Tang et al. 2024). The carrot test is widely used in different fields as a standard measure of mastication efficiency and eating behaviour. The raw carrot was chosen as it is easy to standardise in terms of shape, weight and texture. Participants are excluded if they exhibit unusually high or low eating rates for the hard carrot stick (i.e., below 10th percentile (5 g/min) or above the 90th percentile (35 g/min)) or if the eating rate of the two carrot samples differs by less than 20%. These cut‐offs were determined based on a normal distribution of previously collected eating rate data in a similar population (n = 69) (Lasschuijt MPH et al., n.d. in preparation). Participants are also asked to taste samples (servings of 30g) of foods and meals representative of the two diet periods and make a series of hedonic and familiarity ratings on a 9‐point Likert Scale; participants are excluded if they dislike > 20% of the meals (score of ≤ 4). Given the potential for systematic subjective differences in liking to confound the comparison of the main intervention, it was necessary to ensure the ad libitum meals and menus were generally rated within the acceptable range (Blundell et al. 2010).

The study population is further profiled for their stimulated salivary flow rates and salivary α‐amylase activity (Goh et al. 2021) and a number of psychometric questionnaires are completed to profile participants' eating behaviours. This includes measures of interoceptive awareness, Multi‐dimensional Assessment of Interoceptive Awareness version 2 (MAIA‐2) (Mehling et al. 2018); appetitive and eating behaviour traits, Adult Eating Behaviour Questionnaire (AEBQ) and the Dutch Eating Behaviour Questionnaire (DEBQ) (Hunot et al. 2016; van Strien et al. 1986); and a trait measure of reasons to stop eating, Reasons Individuals Stop Eating Questionnaire (RISE‐Q15) (Chawner et al. 2022).

3.4. Randomisation Procedure and Concealment of Study Aim

Participants are randomised to test blocks that last 7 weeks from baseline to study completion. Each block (max size: n = 8) consists of two groups, with one on each diet order (UPF Slow ER diet > UPF Fast ER diet or UPF Fast ER diet > UPF Slow ER diet). Eligible participants are randomly allocated to a diet order (1:1) by means of randomisation envelopes stratified by block. Participant enrolment and allocation sequence are done by two researchers, both responsible for study oversight.

Each diet consists of seven unique daily menus (fixed paired breakfast, lunch, and dinner meals and snacks) that are presented twice within each diet period. To prevent bias of menu sequence, the order of the menus is semi‐randomised such that we ended with 14 unique orders (fixed weekend meals) that are randomly allocated to each group within a block. The study data is stored under codes (blinded) assigned by an independent researcher not involved in the analyses of the data (K.P.). Unblinding will occur after the primary analyses have been conducted. The diets cannot be administered in a fully blinded manner due to the nature of the study meals.

To conceal the true aim of the study, participants are told that the study is about different protein sources and body composition changes and that this is not a weight‐loss study. Participants are debriefed about the study goals on study completion.

3.5. Dietary Intervention

The diets comprise familiar Dutch meals that consist of commercially available food ingredients. Familiarity with the study foods is screened during recruitment. During the diet interventions, all foods are provided, and participants attend all meal sessions (three main meals per day) at the Health Research Unit at Wageningen University on Monday–Friday for each 14‐day diet period (semi‐residential). On weekends, participants receive breakfast, lunch and dinner pre‐made and packed to consume at home. On both weekends and weekdays, participants receive snack packages to‐go after breakfast, lunch and dinner. Both diets consist of foods classified as UPF (UPF Slow ER diet 97% and UPF Fast ER diet 94%) based on the Nova categorisation approach (Nova category 4) (Monteiro et al. 2019) and Nova attributions are made by two independent coders based on a standardised definition of each category (Table 1). An overview of all meals and snacks can be found in Table S1.

TABLE 1.

Average daily energy and (macro) nutrient composition of the ultra‐processed slow eating rate and ultra‐processed fast ER intervention diets.

	UPF fast ER	UPF slow ER	Average
Amount (g)	3823	3821	3822
Energy (kcal)	5835	5831	5833
Energy density (kcal/g)	1.53	1.53	1.53
Fat (EN%)	33	22	28
Saturated fat (g/100 kcal)	1.52	0.64	1.08
Saturated fat (EN%)	14	6	10
Carbohydrates (EN%)	47	53	50
Mono and disaccharides (g/100 kcal)	4.8	3.9	4.4
Mono and disaccharides (EN%)	19	15	17
Protein (EN%)	16	21	18
Fibre (g/100 kcal)	1.5	1.5	1.5
Salt (NaCl) (mg/100 kcal)	407	408	407
Ultra‐processed foods a (EN%)	94	97	95
Carbohydrate‐to‐fat ratio b	0.58	0.71	0.65
Costs per day (€)	35.85	36.61

^a Nova 4 Classification was determined by two independent coders. There was disagreement for 7% of the breakfast ingredients, 2% of the lunch ingredients, and 9% of the dinner ingredients. These ingredients were classified by a third coder to reach consensus.

^b This is a part‐to‐whole ratio, calculated as: carbohydrate energy/(carbohydrate energy + fat energy).

The food textures of the products and meals in both diets were selected to influence eating rate (ER), with one diet promoting slower consumption rates and the other encouraging faster consumption (Slow vs. Fast ER, g/min). Texture is a perceptual property that cannot be described accurately by one single term due to the multitude of textural properties (i.e., geometrical, material, absorptive, dynamic‐mechanical properties) that combined produce a slower or faster eating rate. For the purpose of simplicity and ease of interpretation, we summarise the texture manipulation as ‘Fast’ or ‘Slow’ diets to represent the combinations of food textures used in each diet arm to deliver this behavioural intervention. The eating rate of the meals are manipulated based on previous experience with texture and eating rate measurements (Bolhuis and Forde 2020; Forde and Bolhuis 2022), and is measured as a study outcome to confirm whether significant differences in diet ER was achieved. In the ‘UPF Slow ER’ diet, meals include familiar textural properties such as ‘hard’ and ‘chewy’ that are known to slow down ER. The ‘UPF Fast ER’ diet period consists of soft and lubricating textures known to lead to a relative fast ER. Meals are developed based on previous research and through extensive pre‐testing and based on this estimated to be categorised as a relatively fast or slow meal or snack (Forde et al. 2017, 2020, 2013b; Heuven et al. 2024, 2023; Lasschuijt et al. 2023; Teo et al. 2022; Van den Boer et al. 2017). All meals are served as three times a standard Dutch portion to ensure ad libitum meal consumption (Donders‐Engelen et al. 2003). Beverages are not included as part of the dietary interventions (with exception of two cans of non‐alcoholic drinks in the weekends). On all intervention days participants are only allowed to drink water, tea and coffee without sugar or milk (unlimited).

Diets are developed such that they should differ in ER, but are matched on the level of the meal for portion size served (gram), variety (number of meal components, for example a carb component and a protein component or a mixed meal such as spaghetti), total energy served (kcal), non‐beverage energy density (kcal/g), visual volume, liking and familiarity.

At the level of the week, the diets are matched as closely as possible on fibre, sodium and macronutrient content but giving priority to the matching of the variables at the meal level. As a result, the UPF Slow versus the UPF Fast diet differs in the proportion of fat (22 EN% vs. 33 EN%, respectively), carbohydrates (53 EN% vs. 47 EN%, respectively) and protein 21 EN% vs. 16 EN%, respectively served.

4. Overview of Study Measures

A comprehensive overview of all study measures, including the specific days of the trial on which they are conducted, is shown in Figure 2 and Table 2.

TABLE 2.

Overview of all measurements of the restructure RCT.

	Methods	Study outcome	Study timepoints
			BL	DP1	WO	DP2	EOS
Dietary intake and eating behaviour	Food served ad libitum, pre–post‐weighing of all meals and snacks	Intake per meal, day, week and intervention period (g, kcal) Average daily energy intake (kcal/day) Average daily food intake (g/day)		●		●
	Video recordings	Oral processing behaviour (meal duration, eating rate (g/min), number of bites and chews, average bite size)		●		●
	Hidden automated weighing tray Lasschuijt et al. (2021)	Cumulative eating rate over time, bite size (g)		●		●
	Food Frequency Questionnaire	Estimated habitual dietary intake	●
	Food diary (3 days)	Reported dietary intake	●		●
	Appetite questionnaire pre‐ and post‐main meals	Hunger, fullness, thirst, desire to eat, prospective consumption		●		●
	Pre‐meal liking of the meal	Liking and familiarity ratings for each meal on 100 mm line scale		●		●
	24 h urine	Compliance: nitrogen and sodium content		●		●
	Carrot test Tang et al. (2024)	Person trait: oral processing behaviour	●
	Saliva collection	Salivary flow rate, α‐amylase concentration	●
	Multi‐dimensional Assessment of Interoceptive Awareness version 2 (MAIA‐2)	Interoceptive awareness	●
	Adult Eating Behaviour Questionnaire (AEBQ)	Appetitive traits	●
	Dutch Eating Behaviour Questionnaire (DEBQ)	Restrained eating Emotional eating	●
	Reasons Individuals Stop Eating Questionnaire (RISE‐Q15)	Reasons for eating cessation	●
	Diet Satisfaction questionnaire	Diet satisfaction		●		●
Body weight and body composition	Electronic weighing scale	Body weight (weight not visible to the participant)	●	●	●	●	●
	Measuring tape	Waist‐to‐hip ratio	●		●		●
	Dual‐energy X‐ray absorptiometry (DEXA)	Changes in fat and fat‐free mass	●		●		●
	Bio‐impedance analysis (BIA)	Changes in water retention	●		●		●
	Accelerometery	Energy expenditure, exercise	●	●	●	●	●
	The International Physical Activity Questionnaire‐Short Form (IPAQ‐SF)	Exercise		●		●
Acute post‐prandial markers of metabolic health	Mixed meal tolerance test	Post‐prandial hormone responses, subjective appetite (hunger, fullness, thirst, desire to eat, prospective consumption)	●	●		●
	Indirect calorimetry with ventilated hood	Resting energy expenditure Respiratory quotient Diet induced thermogenesis	●	●		●
Other markers of metabolic health	Continuous glucose monitoring (CGM)	Glycaemic variability (continuous measure, 15 min time interval)		●		●
	Fasted blood plasma samples	Ghrelin	●	●		●
		HbA1c	●	●		●
		Total cholesterol, HDL, LDL, free fatty acids	●	●		●
		Inflammatory markers	●	●		●
		Leaky gut markers	●	●		●
	Blood pressure monitor	Changes in blood pressure (mmHg)	●		●		●
	Faecal collection	Faecal metabolite profile Microbiota composition Biomarkers of intestinal barrier function	●	●	●	●	●
	24 h urine	C‐peptide creatinine ratio Urinary metabolites	●	●		●
	IBS‐SSS questionnaire	Gastrointestinal symptoms	●	●		●
	Bristol stool scale	Self‐reported stool consistency	●	●		●
	Blue dye method	Gut transit time	●		●		●

Abbreviations: BL, baseline; DP1, diet period 1; DP2, diet period 2; EOS, end of study; WO, washout.

4.1. Food and Energy Intake

To determine ad libitum food (grams) and energy (kcal) intake, the weight of the food as served and consumed is measured. Energy and nutrient intake are calculated using packaging information. If packaging information is unavailable, the data are supplemented using the current Dutch Food Composition Table (NEVO table 2019, Version 6). Besides the weighed ad libitum food intake, the disappearance of food from the plate over the course of consumption (meal duration) of each main meal is measured using continuous weighing sensors. The sensors are hidden in a regular dining tray on which the food is served. This scale is not visible to participants, and the device is an advanced version of the previously published eating pattern monitor (Lasschuijt et al. 2021).

Participants consume their main meals in individual cubicles in a common dining room at the eating behaviour laboratory in the Health Research Unit, Wageningen University. All meals and snacks are offered ad libitum and participants are instructed to eat in their normal way and are free to consume each meal until they feel comfortably full. If a participant finishes their first portion of a meal, a second portion of the meal is provided without request. All meals are served with a single glass of water (240 mL) and at breakfast participants also receive coffee or tea (130 mL) without sugar or milk. Throughout the day, participants can drink water from a water bottle (provided by the researchers; refill themselves) or coffee or tea (without sugar or milk) and this intake is recorded. Participants are free to season their evening meals to their personal taste, and these meals are served with two sachets of salt (2 × 1 g) and two sachets of pepper (2 × 0.2 g). Participants are provided with a snack in between main meals and are asked to note the time they consume the snack and return the (empty) packaging and leftovers to record snack intake. In addition, participants could choose one piece of fruit (apple or orange) a day (the same for both diet periods).

There are two weekends in each 14‐day diet period, and participants are asked to continue their meal protocols at home over these weekends. To facilitate this, a researcher will complete a home visit at the start of the study to give instructions about meal‐setting and meal storage and to provide standardised plates, bowls, glasses and cutlery. Each Friday within the intervention periods, participants receive a package with fully prepared and packaged meals and snacks together with beverages (two cans of soda or 0.0% alcohol beer; the same for both diet periods). Participants are asked to take pictures of their meals pre and post consumption and to return their weekend meal packages, complete with all leftovers and empty packaging, to determine intake over the course of the weekend.

Dietary intake during the baseline and washout periods is estimated using a validated food diary application (Traqq) on two weekdays and one weekend day, where participants are asked to record their food and beverage intakes via their smartphone (Lucassen et al. 2021).

4.2. Eating Behaviours and Appetite

Eating behaviour is measured during consumption of each meal for each diet arm as a manipulation check to confirm the differences in ER are significant between each condition. On weekdays, participants are video recorded when eating all meals using an integrated webcam camera (ThinkPad) and Action Camera (EKEN H9R, Action cam). Participants know they are being recorded but cannot see themselves and are instructed to look forward when eating. Eating rate is determined from the start (first bite) and stop (swallow last bite) time of the meal based on videos recordings. Detailed micro‐structural patterns of eating behaviours are extracted from video recording through manual behavioural annotation by trained video coders using a coding scheme developed previously (Forde et al. 2013a). Behaviours are coded using the behavioural annotation software (ELAN version 6.0 Max Planck Institute for Psycholinguistics, the Language Archive, Nijmegen, The Netherlands) and include bites, chews, sips and swallows and oro‐sensory duration (bite duration, seconds), total meal duration, active eating time (s) and used to derive the average measures; bite size, chews per bite, chewing frequency?, oro‐sensory exposure (s/g)? and ER (g/min). The eating rate of meals consumed outside the lab (during weekends) will be extracted from the integrated time measure in an online appetite questionnaire participants were asked to fill in on their smartphone. To obtain an estimate of subjective satiation and satiety of the diets, we will compare subjective appetite ratings before and after eating each meal. Participants rate their current ‘hunger’, fullness', ‘thirst’, ‘desire to eat’ and prospective consumption on a 100 mm line scale anchored from ‘not at all’ to ‘extremely’ in an online questionnaire (Flint et al. 2000). For each test meal, participants are asked to consume a single bite and rate their ‘liking’ and ‘familiarity’ of the meal on the same line scale. This questionnaire was presented on a laptop during week days and participants filled out the questionnaire on their smartphone during weekends to be able to upload meal images directly from their smartphone. At the end of each diet period, participants complete a diet satisfaction questionnaire to assess their overall satisfaction with the intervention diets (James et al. 2018).

4.3. Anthropometric Outcomes

To be able to compare body weight changes between the two diets, body weight is measured every Monday, Wednesday and Friday at the Health Research Unit after an overnight fast and before breakfast. Participants could not see their own weight as a platform weighing scale was used, and the weight was depicted on a screen visible to only the researcher.

Body composition measures are taken before and after each diet period (Figure 2). This includes changes in fat mass, fat‐free mass, whole body and regional lean and fat mass, measured by dual‐energy X‐ray absorptiometry (DEXA) scan (Lunar, USA) for each tissue. The ratio of the attenuation 35 keV/61 keV is calculated, and bone, fat and fat‐free tissue mass is calculated using Prodigy Pro package with 39ncore V18 SP4.1 software. Waist to hip circumference measures in duplicate using a measuring tape (SECA 201, Germany). Waist‐to‐hip ratio is calculated by dividing mean waist values by the mean hip value, and water retention is measured by bio‐impedance analysis (BIA) (Fresenius Medical Care, Netherlands). Fat‐free mass and fat mass will be corrected for water retention should water retention values differ between end points.

4.4. Physical Activity

To limit variations in food intake and appetite due to exercise, participants are asked to maintain their usual physical activity routine of less than 4 h/week (inclusion criteria) and keep this routine consistent over the two diet periods. To ensure compliance, physical activity is monitored throughout both diet periods. Participants are asked to wear an Actigraph on their hip and an ActivPAL monitor on their leg during baseline, both diet periods, and during the washout period to measure sedentary energy expenditure and monitor physical activity expenditure to compare active energy expenditure (kcal/day) between the diets. Participants also complete the International Physical Activity Questionnaire‐Short Form (IPAQ‐SF) at baseline and at the end of each diet period. The questionnaire records the activity of the last 7 days on four intensity levels: (1) vigorous‐intensity activity such as aerobics, (2) moderate‐intensity activity such as leisure cycling, (3) walking and (4) sitting (Lee et al. 2011).

4.5. Acute Post‐Prandial Metabolic Markers of Health

To determine acute post‐prandial metabolic markers of health, a mixed meal tolerance test (MMTT) is performed at baseline and at the end of both diet periods. Resting energy expenditure (REE) is measured in the morning after an overnight fast and 25 min of bed rest, using an air‐exchange ventilated hood (MAX‐IIa indirect calorimeter, USA). Additionally, ventilated hood measures are done to calculate the respiratory quotient and diet‐induced thermogenesis based on the air carbon‐dioxide ratio after the MMTT (Compher et al. 2006). This same MMTT is used to measure changes in post‐prandial hormone responses in response to the diets (Figure 3).

FIGURE 3.

Overview of the mixed meal tolerance test performed at baseline and after each intervention period. Participants are asked to consume the entire meal (202 g rice‐based porridge and 200 mL chocolate milk) at their own pace but without any breaks within 10 min. Fasting and post‐ingestive blood samples are drawn, see indicated time points. Outcomes of interest are: Glucose, Insulin, Ghrelin, GLP‐1, PP, PYY, Glucagon, C‐peptide, energy expenditure, subjective satiety, inflammatory markers: Interleukin‐6, interleukin‐8, tumour necrosis factor‐α and C‐reactive protein and leaky gut markers: Bactericidal increasing protein, soluble CD14, lipopolysaccharide‐binding protein, gamma‐glutamyl transferase, peptidoglycan and immunoglobulin A.

The MMTT consists of a fixed portion of a rice‐based porridge (202 g) and 200 mL of chocolate milk (total meal: 402 g, 552 kcal, 77 g CHO, 15 g fat, 25 g protein), both of which are categorised as UPF based on the Nova scheme (Monteiro et al. 2019). Participants are cannulated (t = −60) and blood samples are collected at t = −10 and t = −5 (baseline samples) and at t = 10 (end of the meal), 15, 30, 45, 60, 90, 120 and 180 min to measure changes in blood glucose, insulin, ghrelin (total), glucagon, glucagonlike‐peptide 1 (GLP‐1), pancreatic polypeptide (PP), pancreatic peptide YY (PYY) and c‐peptide (Meso Scale Discovery, Rockville, MD, USA). Before and after the MMTT, participants rate their subjective appetite (hunger, fullness, prospective consumption, thirst, desire to eat) using the same approach described earlier. To profile differences in bolus properties (i.e., particle number and size, total surface area and saliva uptake) a bolus sample is collected from each participant by asking them to chew and spit out two bites of rice porridge (10 g each) at the point they would normally swallow.

4.6. Metabolic Markers of Health After 14‐Days on a UPF Diet

To measure glucose responses to the diet interventions, participants are asked to wear a continuous glucose monitor (CGM) device (Freestyle Libre Pro iQ continuous glucose monitor, Abbot Diabetes Care, CA) to estimate plasma glucose levels every 15 min during each 14‐day intervention period in order to track trends in terms of time spent within and outside range (based on data modelling) and variability (% coefficient of variation 100× SD/Mean glucose within 24 h and over 2 weeks) on each diet (Battelino et al. 2023; Chimene et al. 2024).

To measure changes in gut, coronary health and appetite hormones in response to the diets, blood samples are collected at baseline and post both diet periods (Figure 3). These samples are analysed for inflammatory markers: interleukin‐6, interleukin‐8, tumour necrosis factor‐α and C‐reactive protein, and leaky gut markers: bactericidal increasing protein, soluble CD14, lipopolysaccharide‐binding protein, gamma‐glutamyl transferase, peptidoglycan and immunoglobulin A. To determine changes in levels of long‐term appetite hormones, fasted ghrelin levels are measured and to determine changes in cardiovascular biomarkers, total, HDL and LDL cholesterol are determined. Additionally, plasma lipidomic metabolite profiling is done using the high‐throughput nuclear magnetic resonance (NMR) metabolomics (Nightingale Health Ltd., Helsinki, Finland).

To assess endogenous insulin secretion, participants are asked to collect 24 h urine samples in 3 L containers during baseline and post interventions to determine the c‐peptide/creatinine ratio. Urine collection starts with the first voiding each morning and is completed 24 h later. Participants store containers at refrigeration temperature (4°C) and each participant's urine is weighed, stirred, aliquoted and stored at −80°C for later analysis. Urine collection completeness is assessed by comparing urinary creatinine to expected reference values based on body weight, age and gender.

To explore the effect of UPF fast ER and slow ER diets on changes in faecal microbiome and metabolite profiles, participants collect faecal samples (one sample weekly) at baseline, at the end of each week during the diet periods, and at the end of the washout period. Faecal microbiota composition is measured by shotgun metagenomics (8 Gb/sample PE150) to explore bacterial communities at taxonomic and functional levels of all samples. Raw reads deriving from shotgun sequencing will be mapped to bacterial databases to analyse taxonomic and functional traits of the bacteriome. Briefly, metagenomic reads will be processed using different pipelines (Biobakery pipeline tools, MetaPhAn4 and HUMAnN3) to generate taxonomic (family‐level relative abundance, specific species, alpha/beta diversity) and microbial pathway abundance profiles (CAZyme abundance, bile salt hydrolases) complemented by MelonnPan (predicting metabolite) and METABOLIC (metabolic functional). Changes over the course of the intervention will be determined by comparing MaAsLin2, zero‐inflated models and fast zero‐inflated negative binomial mixed modelling.

To assess the influence of UPF fast ER and slow ER diets on gut microbiome functionality and gut health, the irritable bowel syndrome severity scoring system (IBS‐SSS) is used to assess gastrointestinal symptoms (Francis et al. 1997), faecal biomarkers of intestinal barrier function are measured during baseline and post each intervention. Additionally, gut transit time is assessed at baseline and post each intervention period, via the blue dye method (Asnicar et al. 2021) and stool consistency is rated by participants using The Bristol Stool Scale (Blake et al. 2016).

5. Protocol Compliance Measures

The 7‐week study protocol may present challenges for continued adherence due to the burden on participants, and the semi‐residential nature of the intervention can offer opportunities for participants to deviate from pre‐specified procedures. In the design of the diets and measures, emphasis has been placed on protecting the key intervention variables (UPF diet, meal texture) and primary outcome (measured energy intake at each meal and snack) by providing each meal and measuring intake through pre‐ and post‐meal weighing. This is considered the most accurate way of measuring dietary intake (Tien et al. 2024). The non‐residential aspect of the protocol retains greater ecological validity in mimicking daily life compared to full residential trials, but has a disadvantage: the risk of protocol deviations that may confound some of the outcomes of interest.

To enhance the likelihood of adherence and reduce the risk of non‐compliance, a series of additional measures have been implemented to encourage participants to adhere to the protocol and monitor compliance. All food is provided to the participants, which enhances adherence (Tien et al. 2024). Participants wear a CGM and have their physical activity monitored throughout the diet periods, both as a measure of compliance and as a psychological motivation to adhere to the study protocol. Before each main meal in the trial, participants are encouraged to report any non‐study foods or beverages consumed during the inter‐meal period (i.e., between the main meals). As there is less control over measured food intake at weekends, diet adherence is measured by measuring urinary nitrogen and sodium concentrations with the 24 h urine samples and titrated against reported protein and salt intake during the weekend period (Bingham 2003). Participants are encouraged throughout the trial to communicate any specific challenges they encounter in adhering to the protocol, and adjustments or accommodations are made on a case‐by‐case basis. Although we are confident that this promotes good adherence during the diet periods, it is acknowledged that there may be deviations that will not be captured by these various compliance measures. Participant raw data is periodically monitored for unusual values and is completed by designated research team members to ensure values are within the expected range.

6. Statistical Analyses Plan

Statistical analyses are performed using SAS (SAS version 9.4; SAS Institute, Cary, NC, USA) and p‐values < 0.05 are considered as statistically significant. Prior to data analyses, normality of the data is visually inspected (histogram). If outcome measures are non‐normally distributed, data will be (Log) transformed or analysed using non‐parametric tests.

The main outcome analysis to determine whether average daily energy intake (kcal/day‐ averaged across 14 days) differs between the two diets will be conducted on a per protocol basis. All data of a participant will be excluded if a participant is deemed non‐compliant or if primary outcome data (energy intake) is missing for four or more study intervention days within one of the diet arms. Participants who report consuming foods which combined lead to more than 350 kcal/day consumed outside of the study protocol are excluded (on day‐level) in the per protocol analyses. This will be determined during a blind review of the data. The secondary outcome analyses will be conducted on an intention‐to‐treat (ITT) basis, which includes all valid (within range) data available for all subjects. In the event of missing data, the outcome variables that have not been measured will be treated as missing data and will not be modelled or interpolated.

We will test whether average daily energy intake (kcal/day‐averaged across 14 days) differs between study periods (primary outcome) using a repeated measures mixed model to test for a main effect of diet period (PROC MIXED). The model will include diet period (UPF Fast ER, UPF Slow ER) and diet‐day (1–14) and their interaction as fixed factors and participant, block, diet order and day menu as random variables. The covariate structure (compound symmetry [CS], variance components [VC], autoregressive [AR(1)] and unstructured [UN]) is selected based on the log likelihood difference (chi‐square test). If the main effect or interaction effect is significant, a post hoc t‐test will be performed when comparing diets with Tukey correction for multiple comparisons when looking at the diet*diet‐day interaction applying a contrast such that only the same day on both diets is compared (i.e., day 1 Slow ER diet vs. day 1 Fast ER diet). The secondary objectives of this study are analysed using the same model, with a few adaptations due to time points at which the data was collected or by adding covariates. For example, intake at the level of the meal, day, week and diet will be analysed using the same model but with added covariates including participant characteristics (age, gender and eating behaviour) and subjective reporting of liking and familiarity of the meals. For the metabolic secondary outcomes (body weight, fat‐free mass, fat mass, hormones, glucose and blood lipids) we will add baseline differences as covariates. A detailed statistical analysis plan including data analysis script is pre‐registered online (https://osf.io/afmkh).

7. Discussion

The Restructure RCT aims to determine whether the observed differences in ad libitum energy intakes from UPF diets are driven by meal texture‐derived differences in ER. Previous research has demonstrated a consistent effect of meal texture‐derived differences in ER on energy intake from minimally processed and UPFs within a meal (Teo et al. 2022) and over the course of a day (Lasschuijt et al. 2023). The current trial will extend this further and test whether these differences are sustained over a 14‐day period, to potentially explain the previously observed differences in energy intake from UPF diets. The intervention diets are designed and pre‐tested to have texture combinations that have a direct effect on ER while maintaining cultural appropriateness and being within an acceptable hedonic range. To determine the effect of ER of UPF on energy intake independently, the two diets are matched on other established factors known to moderate energy intake such as meal energy density, food volume, portion size, variety and meal palatability (English et al. 2015; McCrickers 2015; Rolls 2009; Rolls et al. 1998). The scientific and societal debate on the role of processed foods in promoting diet‐related chronic diseases is predominantly based on observational studies, with a general lack of understanding of potential causal mechanisms (Askari et al. 2020; Juul et al. 2021; Lane et al. 2024; Srour et al. 2022; Tobias and Hall 2021; Valicente et al. 2023). Understanding possible mechanisms is challenging due to the wide diversity of ingredients, nutrient and sensory properties in industrially processed foods (Gibney and Forde 2022). Numerous hypotheses have been proposed to explain the association; however, none are supported by experimental data (O'Connor et al. 2023). The findings of this RCT will provide insight on the importance of food texture‐derived ER in moderating dietary energy intake. Additionally, the study investigates the effects of UPF texture on body composition, metabolic health, appetite and the gut microbiome, in an attempt to further elucidate potential underlying metabolic mechanisms.

The primary outcome of the trial will be food and energy consumed from each diet arm. To further explore other possible underlying mechanisms that may influence the reported health effects of consuming a diet dominated by UPF, a series of secondary and exploratory outcomes have been pre‐specified. We aim to compare changes in body weight and explore acute post‐prandial and longer‐term metabolic markers of health, gut microbiome and intestinal metabolites over a 14‐day diet comprising over 90% of daily energy from UPF with either a Slow texture‐derived eating rate or a Fast texture‐derived ER. The study is powered on the primary outcome (energy intake) and not on the secondary outcomes, nor is the duration of the trial sufficient to make conclusive statements on these secondary outcomes. The exploratory outcomes will enable a preliminary comparison of a wide range of metabolic outcomes that have been speculated to drive the associations between dietary UPF consumption and negative health outcomes (O'Connor et al. 2023; Trumbo et al. 2024).

Within our protocol we implement a 14‐day washout period between the two diet periods to reduce potential carry‐over effects on body composition, eating behaviours, metabolic outcomes and microbiome changes between the two diets. Including a washout period of this length carries the risk of higher attrition rates and potential dropouts; however, we have prioritised minimising any confounding from carry‐over effects between the two diets to ensure a clean comparison of outcomes from each diet arm. To ignore a washout would be convenient from the perspective of participant burden, but requires definitive evidence that the residence time or half‐life of UPF ingredients consumed in each diet arm is no more than a few days, and evidence that any residual effects of prolonged UPF consumption on our secondary outcomes such as gut microbiome are not relevant. In the absence of this information, we chose not to a priori assume there is no carry‐over, and instead pursue a cautious approach that gives confidence that there is minimal carry‐over in metabolic or microbiome measures between each diet period, and the impact of carry‐over from a 14‐day diet is mitigated by an equivalent washout period.

To ensure a homogeneous study population, participants are excluded if their habitual eating rate falls at the extremes of the ER distribution (an eating rate slower than 5 g/min or an eating rate faster than 35 g/min), which is based on data from a previous similar cohort (Lasschuijt et al. 2025, under review) or if their dietary energy intake from UPF is high, based on their habitual consumption patterns, in comparison to the average Dutch consumer. To assess UPF intake we have developed a novel UPF Frequency Questionnaire (UPFFQ) to capture the frequency of intake across food categories that have been coded into their respective Nova category. Preliminary data using this approach was collected in a previous study, and has led to further refinement of questions to ensure sufficient information is collected on each food item to enable accurate classification into their respective Nova category (van Bruinessen et al. 2025 under review). The assignment of the Nova category to each item in the UPFFQ has been completed by independent classification of each branded food item by trained nutrition researchers and dieticians within the metabolic unit at the Division of Human Nutrition and Health.

Within each diet period, all foods are provided by the study team, and a clear protocol has been implemented to ensure consistency in food preparation and texture, to ensure compliance with the intervention guidelines, and to encourage participants to adhere to each diet. We carefully selected foods and menu items that are familiar, and participants' familiarity with the meals and menu items was checked during screening. As a semi‐residential trial, all weekday meals are consumed under supervision in a dining room at the Health Research Unit to facilitate precise measurement of food intake and eating behaviours. UPF snacks matching the texture manipulation of each diet are also provided ad libitum to minimise potential risks of consuming non‐study foods when away from the laboratory. Weekend meals are delivered fully prepared for reheating, and leftovers are returned by the participants on Monday. Participants are instructed to record meals with before and after pictures to enable monitoring of remote food intakes alongside a range of metabolic and behavioural compliance measures. Rather than a restrictive in‐patient approach, we have opted for a protocol that better mimics food consumption behaviour in a free‐living environment to prioritise ecologically valid consumption patterns alongside comprehensive compliance measures and continuous data checking to ensure the reliability of our findings. Through this, we aim to ensure control and accuracy in the measures, but also not to interfere unduly with participants' normal routines. Although controlled feeding trials of this nature offer excellent control in accurately measuring intake and metabolic outcomes, it should be acknowledged that the nature of the study design is such that serving large portions of meals in an unfamiliar environment may impact participants' behaviours, and caution should be exercised when extrapolating these outcomes to everyday eating behaviour in a free‐living context.

Epidemiological evidence increasingly shows that the association of UPF intakes with health‐related outcomes varies greatly for different UPF food groups, with a rising number of publications highlighting the positive association between the consumption of processed and preserved meats and sugar‐sweetened beverages and desserts, and a neutral or even negative association for all of the other UPF food groups (Chen et al. 2023; Cordova et al. 2023; Duan et al. 2022; Mendoza et al. 2024; Osté et al. 2022). Despite this, the current trial has chosen to manipulate the ER of a full UPF diet rather than focusing only on specific food groups, which may result in including foods that offset the impact of these two specific UPF food groups. In doing so, we place our emphasis on the importance of sensory‐mediated eating behaviours at the level of the diet, rather than linked only to specific food groups within the diet.

Through the restructure RCT we aim to test the sustained contribution of meal eating speed to energy intakes from diets dominated by UPF. We acknowledge that observed consistent effects of UPFs on energy intake and health may not be driven by a single factor, but aim that with our trial, we can help better clarify the role of food texture derived eating behaviours in moderating energy consumed from prolonged consumption of diets dominated by UPF.

8. Conclusions and Future Directions

The Restructure RCT will test whether meal texture‐derived differences in eating rate moderate energy intake from UPF diets and cast new light on one of the potential mechanisms underpinning the observed differences in energy intake. A better understanding of the influence of sensory cues on the behavioural and energy intake differences observed can inform food‐based strategies to moderate dietary energy intake and mitigate diet‐related health risks. Our secondary comparison of body composition changes and exploratory comparison of metabolic changes will further clarify the potential underlying causal link between UPF consumption and adverse health outcomes.

Author Contributions

C.G.F. designed the research (project conception), C.G.F. and K.G. acquired the funding, C.G.F. and M.P.L. developed the overall research plan and wrote the manuscript, and M.P.L., L.A.J.H., M.B., Z.L., J.R., M.S. and C.G.F. were responsible for designing (part of) the study protocol and manuscript editing. All authors contributed to the article and approved the submitted version of the manuscript.

Ethics Statement

This study is ethically approved by the Medical Ethical Committee East‐Netherlands (Dutch: Oost‐Nederland), the Netherlands (NL83462.081.23). Approval obtained on September 29th, 2023, recruitment began in October 2023, and study measurements started from January 2024 onwards. Estimated end of data collection is November 2024.

Conflicts of Interest

C.G.F. reports travel reimbursements from Kerry Taste and Nutrition, USDA, International Life Sciences Institute, Ajinomoto Co. Inc., British Nutrition Society, Nestlé Nutrition Institute, AB Mauri, and the Institute for the Advancement of Food and Nutrition Sciences. He also reports speaking and lecture fees, as well as travel reimbursements, from the World Sugar Research Organisation and Northern Irish Dairy Council, and speaking and lecture fees from Ferrero, PepsiCo Inc., General Mills Inc., and Mondelez International Inc. M.P.L. reports speaking and lecture fees, as well as travel reimbursement, from Nestlé Nutrition. K.G. is a member of the Global Independent Nutrition Advisory Board of the Mars company. All other authors declare no conflicts of interest.

Supporting information

Data S1: nbu70027‐sup‐0001‐DataS1.docx.

Data Availability Statement

Study data and analyses script will be made available after publication of the manuscripts.