The effects of nudging and pricing strategies on the availability and purchases of ultra-processed foods: A secondary analysis of the Supreme Nudge trial

Regular consumption of ultra-processed foods (UPF) is a risk factor for morbidity and mortality. UPF are widely available in supermarkets. Nudging and pricing strategies are promising strategies to promote healthier super-market purchases and may reduce UPF purchases. We investigated whether supermarket nudging and pricing strategies targeting healthy foods, but not specifically discouraging UPF, would change UPF availability, price, promotion and placement (UPF-APPP) in supermarkets and customer UPF purchases. We used data from the Supreme Nudge parallel cluster-randomized controlled trial, testing the effect of a combined nudging and pricing intervention promoting healthy products. The Dutch Consumer Food Environment Score (D-CFES) was used to audit 12 participating supermarkets in terms of UPF-APPP. We used customer loyalty card data of the first twelve intervention weeks from 321 participants to calculate the proportion of UPF purchases. Descriptive statistics were used to assess differences in D-CFES between supermarkets. Mixed model analyses were used to assess the association between the D-CFES and UPF purchases and the effect of the intervention on UPF purchases. No difference in the D-CFES between intervention and control supermarkets were found. No statistically significant association between the D-CFES and UPF purchases ( β = (cid:0) 0.00, 95%CI: (cid:0) 0.02, 0.01) and no significant effect of the intervention on UPF purchases ( β = 0.02, 95%CI: (cid:0) 0.07, 0.12) was observed. Given the significant proportion of unhealthy and UPF products in Dutch supermarkets, nudging and pricing strategies aimed at promoting healthy food purchases are not sufficient for reducing UPF – APPP nor purchases, and nationwide regulation may be needed.Trial registration number: Dutch Trial Register ID NL7064, May 30, 2018, https://trial search.who.int/Trial2.aspx?TrialID = NTR7302.


Introduction
Non-communicable diseases (NCDs) such as type 2 diabetes, several types of cancer and cardiovascular diseases are the main causes of death worldwide (Afshin et al., 2019).It is estimated that 74% of all deaths per year globally are caused by NCDs (World Health Organization, 2022).In the Netherlands this percentage is even higher: in 2016, 90% of all deaths were caused by NCDs (World Health Organization, 2020).Diets high in sugar, salt and saturated fat are major contributors to the development of these diseases (Bouvard et al., 2015;Imamura et al., 2015;Mozaffarian et al., 2014;Wang et al., 2015) (Bouvard et al., 2015;Imamura et al., 2015;Mozaffarian et al., 2014;Wang et al., 2015).More recently, ultra-processed foods and drinks (UPF) have been proposed as dietary risk factor for NCDs (Canhada et al., 2020;Chen et al., 2020;Fiolet et al., 2018;Levy et al., 2021;Pagliai et al., 2021) (Canhada et al., 2020;Chen et al., 2020;Fiolet et al., 2018;Levy et al., 2021;Pagliai et al., 2021).The level of food processing is typically defined based on the NOVA classification.The NOVA classification describes unprocessed and minimally processed foods (e.g., cucumbers) as group 1, culinary foods (e.g., butter) as group 2, processed foods (e.g., milk) as group 3 and UPF make up group 4 (Monteiro et al., 2016).UPF, as defined by this classification, are "formulations of ingredients, most of exclusive industrial use, that result from a series of industrial processes"( (Monteiro et al., 2019), p. 937).A recent meta-analysis demonstrated convincing evidence that higher UPF consumption is directly associated with risk of type 2 diabetes, cardiovascular-related mortality and mental disorders (Lane et al., 2024).The mechanisms of the harmful effects of UPF are yet to be elucidated.Many UPF are high in salt, sugar and fat, but contaminants from packaging materials, food additives and emulsifiers, contaminants from processing and changes in the food matrix structure may also affect health via speed of eating, satiety, glycemic response, the composition of the gut microbiota and inflammation (Hall et al., 2019;Juul et al., 2021;Srour et al., 2022).
UPF are widely available in supermarkets (Baker et al., 2020).Supermarkets are an important source of product sales for at-home food preparation in Europe, of which 80% take place in supermarkets (Wetenschappelijke et al., 2016).The way that products are presented, priced, promoted and placed in these supermarkets affects what consumers buy (Glanz et al., 2005).Supermarkets often use the four Ps of marketing, known as Product, Place, Price and Promotion to persuade -or 'nudge'-consumers to buy specific products or brands (Goi, 2009).Similar marketing techniques can be used by health promotion professionals.Nudges are a way to address Product, Place and Promotion as a health promotion strategy.More specifically, 'health nudges' are defined as subtle changes in the environment used to make the healthier choice the easier choice without removing any unhealthy options (Thaler & Sunstein, 2008).Pricing strategies are a way to address Price as a health promotion strategy.These strategies include subsidies that make healthy options more affordable or taxes that make unhealthy options more costly (Afshin et al., 2017).
Previous studies from the Netherlands and New Zealand showed that nudging and pricing strategies in a virtual supermarket can be effective to improve healthy food purchases (Hoenink et al., 2020, pp. 1-12;Stuber et al., 2021;Vellinga, Eykelenboom, et al., 2022;Waterlander et al., 2012Waterlander et al., , 2019) ) (Hoenink et al., 2020, pp. 1-12;Stuber et al., 2021;Vellinga, Eykelenboom, et al., 2022;Waterlander et al., 2012Waterlander et al., , 2019)).Nudging and pricing interventions in real-life supermarket settings are also promising for increasing healthy food purchase behaviours (Huitink et al., 2020;Vogel et al., 2021;Waterlander et al., 2013) (Huitink et al., 2020;Vogel et al., 2021;Waterlander et al., 2013).In this context, the Supreme Nudge trial was developed with the primary aim to promote healthier food intake and with the secondary aim to increase healthy food purchases and improve cardio-metabolic health through nudging and pricing strategies (Lakerveld et al., 2018).The nudging and pricing strategies applied in this trial were mainly focused on healthy products according to the Dutch dietary guidelines.The level of processing of products is currently not taken into account in the Dutch dietary guidelines, as in the majority of other countries (Koios et al., 2022).Yet, UPF constitute 29% of daily food consumption in grams in the Netherlands, which together represents 61% of daily energy intake (Vellinga, van Bakel, et al., 2022).To the best of our knowledge, no nudging and pricing studies have focused on UPF outcomes specifically.Thus, we investigated whether a supermarket intervention focused on the availability, price, promotion and placement of healthy products according to the Dutch dietary guidelines would affect UPF availability, price, promotion and placement (UPF-APPP) and purchases as secondary analysis in the Supreme Nudge trial (Stuber et al., 2020(Stuber et al., , 2023a)).This was based on the hypothesis that increasing the availability, promotion and placement of healthy products may lead to reduced UPF-APPP.
More specifically, we aimed to assess: (1) the difference between control and intervention supermarkets in terms of UPF-APPP; (2) the association between UPF-APPP and the proportion of UPF purchased in relation to all food and drink purchases by participants in the Supreme Nudge trial, and whether this association differs between the intervention and control supermarkets; and (3) the effect of nudging and pricing strategies on the amount of UPF purchased, as a proportion of the total count of food and drink products purchased by participants in the Supreme Nudge trial.

Supreme Nudge trial study design
This study is a secondary analysis of the Supreme Nudge parallel cluster-randomized controlled trial (Lakerveld et al., 2018;Stuber et al., 2020).A detailed description of this 12-month nudging and pricing experiment can be found in the study protocol and paper reporting on the main results (Stuber et al., 2020(Stuber et al., , 2023b)).Six control supermarkets were randomized to no intervention and six to receiving both nudging and pricing strategies across 13 food groups (e.g., vegetables, bread, fish).A total of 421 registered participants signed informed consent.Briefly, 73% of participants were female, the average age was 58 years and their average diet quality was 105 on a scale from 0 to 150 (Stuber, Mackenbach, et al., 2023).The nudges consisted of altering the proximity and access to healthy foods, combined with visual attractiveness enhancements and highlighting the taste, convenience, or popularity of healthy foods (Stuber et al., 2020(Stuber et al., , 2023a)).Nudges targeted 11% of the supermarket assortment and pricing strategies 3% (Stuber, Mackenbach, et al., 2023).The pricing strategies entailed a 25% price reduction on healthy foods or whenever possible a 25% price difference between healthy and unhealthy products.Price increases were not actively highlighted, but price reductions were highlighted using promotional signs.Nudging and pricing strategies mainly focused on promoting healthy products, which were defined as products recommended by the Dutch dietary guidelines.The recruitment of participants as well as the inclusion criteria for participants in the Supreme Nudge trial are described elsewhere (Stuber, van Hoek, et al., 2023).The Supreme Nudge trial was approved by the Medical Ethics Review Committee of VU University Medical Centre (reference number: 2019.334).Participants provided informed consent before participating in the study and were not actively informed of the intervention condition.
For the current study we made use of data collected at baseline via an online questionnaire, and of purchase data of participants collected via the supermarket's loyalty card.This loyalty card was part of an existing loyalty program of the collaborating supermarket chain and allowed us to track individual purchases.This purchase data was tracked for the complete duration of the 12-month intervention.Due to the time investment associated with classifying each individual product based on its level of processing, we used the first 14 weeks of purchase data for the current study.

Audit of in-store ultra-processed foods availability, price, promotion and placement
UPF-APPP was evaluated with an adapted version of the validated Consumer Food Environment Healthiness Score (CFEHS) as developed by Borges et al. (Borges et al., 2021).The CFEHS is a tool used to score the healthiness of food stores regarding UPF-APPP based on the NOVA classification.It considers two domains: the availability of certain foods (dimension 1) and the promotion of certain foods (dimension 2).The CFEHS tool was originally developed for the Brazilian context.The food dimension included the most consumed foods in Brazil divided into four groups of the NOVA classification.The environment dimension (dimension 2) included questions about food placement and advertising for each food product.Both dimensions got scored by assigning positive points to products in groups 1, 2 and 3 of the NOVA classification and negative points to group 4 products.An average of both scores results in the final CFEHS (Borges et al., 2021).
The food dimension of the CFEHS tool included products that are not usually consumed in the Netherlands.It was therefore adapted based on the top 5% consumed foods within each food group according to the latest Dutch National Food Consumption Survey (van Rossum et al., 2020).We also made two changes to the structure of the tool.Firstly, the group Beans and Rice in the original tool was changed to Potatoes and Grains because there were no beans included in the list of the most consumed foods in the Netherlands.Brown rice and whole-wheat pasta were added to this group in addition to refined grain pasta or rice.Secondly, a new food group was added that contained mineral water, coffee and tea.These foods were not included in the CFEHS, but constituted a large part of the 5% most consumed foods in the Netherlands.The environmental dimension remained unchanged.During the adaptation process, we kept in close contact with the developers of the original CFEHS tool to ensure that adaptation still reflected the purpose of the tool.We labelled the adapted tool the Dutch Consumer Food Environment Score (D-CFES).Both dimensions received a total score that was standardized to 0-100 such that higher scores reflected lower UPF-APPP.An average of the standardized food and environmental dimension scores was considered the final D-CFES of the supermarket.
After the adaptation of the tool to the Dutch environment, a pilot test was performed in one supermarket that was not included in the Supreme Nudge trial to check if the tool was practical in use.In addition, we checked how customers and supermarket managers would react to the supermarket audit, how long an audit would take, and if any barriers needed to be addressed before the participating supermarkets could be audited.Small adaptations were made according to the findings of this pilot.The final D-CFES can be found in Supplementary Table 1.
The D-CFES was used in audits of the participating supermarkets, which took on average 2 h per supermarket.The audits were done in a timeline of two weeks in March 2022 (week 12 and 13 of the intervention) by one auditor (JR) that was blinded from the intervention condition of the supermarkets.During the audit, pictures were taken when the auditor was in doubt about the scoring of certain parts of the supermarket and these pictures were discussed with the other researchers to make a final decision for the score.

Purchase data
Purchase data based on customer loyalty cards contained the name, quantity, price and weight of each purchased product.All individual products purchased by the participants were classified according to the NOVA classification (UPF; category 4 vs. non-UPF; category 1-3) by a team of three trained researchers.One researcher with experience in classifying foods into the NOVA classification trained the other two researchers.A random sample of 10% of food products was selected and classified by the three researchers, after which disagreements were checked and a final consensus was reached for these food items.Classification was done by analysing the product's ingredient list, which was found on Dutch supermarkets' websites.UPF purchase was operationalized as the count of UPF products as a proportion of the total count of food and drink products purchased.For the cross-sectional association between the D-CFES and the UPF purchase, this proportion was calculated as an average across four weeks in March 2022 (weeks 11, 12, 13 and 14 of the intervention) in which the audits took place.This was done because not all study participants purchased their groceries during the two weeks that the audit was performed.For the intervention effect on UPF purchase, the proportion was calculated for every week during the first 12 weeks of the intervention.See section 2.3 for a further explanation about the different samples used.As part of the main aim of the Supreme Nudge trial, purchased products were also classified into healthy (if recommended according to the Dutch dietary guidelines of 2015) or unhealthy (if not recommended according to these guidelines) (Stuber et al., 2020(Stuber et al., , 2023a)).

Participant characteristics
Participants reported their age, sex, educational level (low, medium or high), household composition (one adult, two or more adults or one or more adults with children living at home), sale proneness and health goals.Sale proneness was measured as the mean of five 7-point Likert scale items on sale proneness such as "The odds of me buying brands that are on promotion are high" (Stuber et al., 2020).Health goals was measured with the question: "I think it is important to eat healthy" on a 7-point Likert scale (Stuber et al., 2020).

Implementation fidelity scores
During the Supreme Nudge trial, the implementation fidelity was regularly measured by a trained researcher (Lakerveld et al., 2018;Middel et al., 2023-a).Implementation fidelity was measured by scoring correct place/position, correct price, correct products targeted and readability/visibility.The implementation fidelity was scored on a scale from 0 to 5 in which 0 reflected that the intervention was not started or temporarily discontinued and 5 reflected implementation according to protocol (Middel et al., 2023-a;Stuber, Mackenbach, et al., 2023).The results of the implementation fidelity for the first 12 weeks of the intervention and for the weeks around the time of the audit are presented in this study to facilitate the interpretation of the results for the three sub questions.

Data analysis
We used three different samples to answer the three research questions.First, to assess the difference between control (n = 6) and intervention (n = 6) supermarkets in terms of UPF-APPP we present the median, interquartile range (IQR), maximum and minimum values of in the food dimension, the environment dimension and total D-CFES of intervention and control supermarkets.We also present the implementation fidelity scores of intervention supermarkets (n = 6) by using the median, IQR, minimum and maximum.Finally, we describe the interrelation between the healthiness of products purchased (n = 9453 unique products) in the 12 supermarkets according to the Dutch dietary guidelines and level of processing using a cross tabulation.
Second, to assess the association between UPF-APPP and the proportion of UPF purchased, and whether this association differed between the intervention and control supermarkets, we used data from n = 231 participants with purchasing data available for the four weeks around which the D-CFES audit was conducted.We used mixed model analyses with a random intercept at supermarket-level to test the association between the D-CFES of the supermarkets and the proportion of UPF purchased.This analysis was adjusted for age, sex, educational level, household composition, sale proneness and health goals (a priori selected confounders).To test for the difference in association between the intervention and control condition, an interaction term for D-CFES and the intervention arm was included in a second model.
Third, to assess the effect of nudging and pricing strategies on the proportion of UPF purchased, we used data from n = 321 participants of which purchasing data were available between week 1-12 of the intervention.This sample was larger than described above because there were more participants who had any purchasing data available in these first 12 weeks compared to week 11-14.We conducted mixed model analysis with a random intercept at supermarket-level and at participant-level to test the effect of the intervention on UPF purchases according to the intention-to-treat principle.
We report descriptive statistics for baseline characteristics of the 321 participants in a table using percentages, means and standard deviations or medians and IQR in case of non-normally distributed variables.For comparison, we also report descriptive statistics for the smaller sample of 231 participants used to test the association between the D-CFES of the supermarkets and the proportion of UPF purchased in text.
Missing data were not imputed since mixed models use the maximum available data.Beta-coefficients and 95% confidence intervals (CI) are presented.For all statistical tests, a p-value <0.05 was considered significant.Statistical analyses were performed with IBM SPSS statistics 26.

Characteristics of participants
The main study sample included 321 participants with n = 2630 purchasing data points.Their mean age was 57.6 (SD = 10.8) and most of the participants were female (72%).The majority of the participants had a medium (35%) or high (42%) educational level (Table 1).Most households consisted of two or more adults without children living at home (44%).The mean score on sale proneness was 4.7 (SD = 1.2) and the mean on health goals was 6.3 (SD = 0.9).A smaller sample was used to test the association between the D-CFES and the UPF purchase (n = 231).Their age was 58.1 (1.3) years on average (SD) and 73% was female.Forty-two percent of these participants had a high educational level.

Characteristics of supermarkets in terms of UPF-APPP, implementation fidelity and product healthiness
The median (IQR) of the D-CFES food dimension for the intervention and control supermarket respectively was 46.4 (5.1) and 46.4 (5.3).The median (IQR) for the environment dimension for the intervention and control supermarkets was respectively 44.1 (13.1) and 46.4 (8.9).The median (IQR) for the total D-CFES of the intervention supermarkets was 45.2 (6.3) and of the control supermarkets was 45.8 (3.8).Visual inspection indicated no difference in D-CFES between intervention and control supermarkets.
For first 12 weeks of the intervention the mean implementation fidelity scores ranged from 2.5 to 4.1 with a median (IQR) of 3.7 (3.1-4.0).Around the time of the D-CFES supermarket audit the implementation fidelity scores ranged from 2.9 to 4.3 with a median (IQR) of 3.2 (2.9-3.8).
Of the 9453 unique food and beverage products sold in the 12 participating supermarkets, 69.0% was classified as UPF and 21.7% was classified as healthy (Table 2).Of all unhealthy products, 82.5% was classified as UPF.Of all healthy products, 22.7% was classified as UPF.

The association between UPF-APPP and UPF purchase
The association between the D-CFES and the proportion of UPF purchased in March 2022 was small and not statistically significant (β = − 0.00, 95%CI: − 0.02, 0.01).There was also no significant difference in the association of D-CFES and UPF purchase between the intervention and control supermarkets (p = 0.33).

The effect of the Supreme Nudge intervention on UPF purchase
There was no significant effect of nudging and pricing strategies on the proportion of UPF purchased during the first 12 weeks of the intervention (β = 0.02, 95%CI: − 0.07, 0.12).

Interpretation
We investigated whether supermarket nudging and pricing strategies targeting healthy foods, but not specifically discouraging UPF, would lower UPF-APPP and UPF purchases.No difference in UPF-APPP between the intervention supermarkets and the control supermarkets was observed and no effect of the intervention on UPF purchases was found.In addition, there was no evidence for a cross-sectional association between supermarket-level UPF-APPP score and proportion of UPF purchases from these supermarkets.
The use of the D-CFES as supermarket audit tool may be a methodological explanation for the lack of difference in UPF-APPP between intervention and control supermarkets.The original CFEHS tool was designed to audit different types of food stores, such as supermarkets, convenience stores, and bakeries.While the study by Borges et al. (Borges et al., 2021) showed very similar CFEHS scores for Brazilian supermarkets and grocery stores compared to the current study, the tool may not be sensitive enough to detect differences between comparable stores (i.e., supermarkets).This may be partly due to the fact that the most consumed foods in the Netherlands are typically available in all supermarkets leading to little variation in the score.This may in part explain that we could not detect any associations.Indeed, the scores for the food dimension, i.e., demonstrating the availability of most consumed foods, did not differ between intervention and control supermarkets at all while there was a small difference in the environment dimension between intervention and control supermarkets.However, the modest implementation fidelity, and the fact that the Supreme Nudge interventions mostly targeted healthy products, are likely important alternative explanations for the comparable UPF-APPP between intervention and control supermarkets.
We did not observe a significant association between UPF-APPP and UPF purchases.A previous cross-sectional study concluded that availability and promotion of healthy options was associated with an increase in the purchase of healthy options in small/non-traditional food stores (Caspi et al., 2017).In addition to the lack in variation in UPF-APPP between supermarkets, another methodological explanation for our null-results may be that the UPF-APPP data was collected in two of four  Note.UPF = ultra-processed foods, according to NOVA category 4.Not UPF = foods classified as NOVA categories 1-3.Healthy = as recommended by the Dutch dietary guidelines.Unhealthy = not recommended by the Dutch dietary guidelines.
weeks and linked to the average proportion of UPF during four weeks which may have diluted any association present.However, with a small number of supermarkets and very little variation in UPF-APPP, our study may have been underpowered to detect this association.
The fact that the nudging and pricing strategies did not affect UPF purchases may have a range of explanations.First, they also did not affect the healthiness of participants' supermarket purchases, their diet quality or cardiometabolic risk markers, as reported elsewhere (Stuber, Mackenbach, et al., 2023).The co-created nudging and pricing strategies focused mostly on promoting healthy products.Since 80% of the products in Dutch supermarkets are unhealthy [ref] and a considerable proportion UPF (Srour et al., 2022), only a small segment of the assortment was eligible for nudging and price promotion strategies.Considering their competitive position, the supermarket chain allowed for 11% of the supermarket assortment to be nudged, and 3% to be priced.This intervention dosage was not strong enough to influence the proportion of healthy foods and beverages (Stuber, Mackenbach, et al., 2023).In addition, unhealthy and UPF products are more heavily promoted (Hendriksen et al., 2021).As such, the interventions may simply not have been powerful enough to change purchasing and consumption behaviors.Some other studies have shown that supermarket nudging and pricing strategies can increase the healthiness of purchases and consumption (Vellinga, Eykelenboom, et al., 2022;Waterlander et 2012Waterlander et , 2013Waterlander et , 2019) ) (Vellinga, Eykelenboom, et al., 2022;Waterlander et al., 2012Waterlander et al., , 2013Waterlander et al., , 2019)).However, no previous study has focused on UPF.As we show in this study, products classified as healthy are not necessarily non-UPF.In fact, 69% of all food and beverage products sold was classified as UPF, and no less than 23% of products recommended by the Dutch dietary guidelines was classified as UPF.Previous studies have also shown that many dietary guidelines do not take into account levels of processing (Koios et al., 2022).Another explanation is the poor to moderate implementation fidelity (Middel et al., 2023-a;Stuber, Mackenbach, et al., 2023).Imperfect implementation fidelity has likely reduced any effects of the intervention on UPF purchases and will also have reduced differences in UPF-APPP between the intervention and control supermarkets.

Strengths and limitations
Strengths of this study included the use of a context-specific version of the D-CFES, the objective measurement of food purchasing through a loyalty card and use of existing data to answer innovative research questions around an emerging nutritional topic, i.e., UPF.We provided training and a classification guideline to all involved UPF researchers which improved the internal validity of the UPF classification.Some limitations also need to be acknowledged.Both the number of supermarkets and number of participants was relatively low to detect small effects.Even though the D-CFES was locally adapted, it may not be sensitive enough to detect differences between comparable stores.Another limitation is that the supermarket audits were performed by one researcher only.However, any doubts about product of promotion score were discussed with the broader research team.The same is true for the NOVA classification: unique products were only scored by one researcher and because the scoring is not standardized, bias may have occurred.While we used objective data to measure UPF purchases, it was unknown whether participants had always used their loyalty cards, so we may have missed purchase occasions.Finally, the participants included in the Supreme Nudge trial are comparable with the general Dutch population in terms of educational attainment, but women were overrepresented and participants had a relatively high diet quality (Stuber, van Hoek, et al., 2023), limiting the generalizability of the current results.

Implications for practice and suggestions for future research
One recommendation is to further develop the D-CFES.The tool is promising in scoring the availability, price, advertising and placement strategies of UPF, but may be adapted to increase sensitivity to detect differences in UPF-APPP between the same supermarket types.This would also allow an objective assessment of the current availability of UPF in Dutch supermarkets and potentially automated monitoring of this availability over time.The second recommendation would be to design strategies directly targeting the availability and promotion of UPF in the supermarket.However, it should be considered that UPF are often highly profitable branded products, made from low-cost ingredients and with a long shelf-life, resulting in market advantages compared to unprocessed foods (Monteiro et al., 2019).Supermarkets are therefore unlikely to collaborate with researchers to decrease the amount of UPF in their supermarkets, especially if competing supermarkets are not joining.As such, national level policies may be needed to restrict the UPF-APPP in supermarkets.
While there is increasing evidence for the effects of UPF on the development of several severe health conditions (Lane et al., 2024), there is also criticism on the NOVA classification, namely that the definition of the NOVA groups is not clear enough leading to inconsistency in classifying some products to one of the four NOVA groups (Braesco et al., 2022;Jones, 2019).This problem mostly arises in the classification of generic (food frequency) questionnaire items, while we were able to classify unique products.Yet, future research could further validate the NOVA classification and develop a detailed guidance protocol for researchers.

Conclusion
There is evidence that nudging and pricing strategies can promote healthier food choices.However, such strategies focused on promoting healthy products do not seem to have an effect on UPF availability and purchases.Given the overwhelming availability and promotion of unhealthy foods and UPF in Dutch supermarkets, stronger measures such as restriction the sales and price promotions of UPF should be considered to help decrease UPF intake and ultimately non-communicable disease incidence.

Ethics statement
Ethics approval: The Supreme Nudge trial was approved by the Medical Ethics Review Committee of VU University Medical Centre (reference number: 2019.334).Participants provided informed consent before participating in the study and were not actively informed of the intervention condition.

Funding
This study was funded through a 'Sustainable Prevention of Cardiometabolic Risk through Nudging Health Behaviours' (Supreme Nudge) talent fellowship awarded to MGMP.The Supreme Nudge project is funded by the Dutch Heart Foundation (grant number CVON 2016-04) and the Netherlands Organization for Health Research and Development (ZonMw) (531003001).Funders had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript, nor have they authority on the decision to submit the manuscript for publication.

Availability of data and materials
The data analyzed during the current study are not publicly available as it will violate participant consent (individual-level data) or confidentially agreements (supermarket-level data).The analyses plans and analytical codes are available from the corresponding author on reasonable request.

Table 1
Characteristics of the Supreme Nudge trial participants with loyalty card purchase data available between weeks 1 and 12 of the intervention (N = 321).
Note.SD = Standard deviation.aFour participants had missing values for this variable.

Table 2
Cross tabulation of level of processing and healthiness of n = 9453 products purchased in the supermarkets participating in the Supreme Nudge trial.