The impact of a low-carbohydrate nutrition education program on food preferences: The correspondence between self-report consumption and supermarket purchases

Monteiro, Sofia; Pujol-Busquets, Georgina; Smith, James; Larmuth, Kate Article — Published Version The impact of a low-carbohydrate nutrition education program on food preferences: The correspondence between self-report consumption and supermarket purchases PLOS ONE Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Monteiro, Sofia; Pujol-Busquets, Georgina; Smith, James; Larmuth, Kate (2025) : The impact of a low-carbohydrate nutrition education program on food preferences: The correspondence between self-report consumption and supermarket purchases, PLOS ONE, ISSN 1932-6203, Public Library of Science (PLoS), San Francisco, CA, Vol. 20, Iss. 4, pp. 1-16, https://doi.org/10.1371/journal.pone.0319503 This Version is available at: https://hdl.handle.net/10419/319531 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 1 / 16 OPEN ACCESS Citation: Monteiro S, Pujol-Busquets G, Smith J, Larmuth K (2025) The impact of a lowcarbohydrate nutrition education program on food preferences: The correspondence between self-report consumption and supermarket purchases. PLoS ONE 20(4): e0319503. https://doi.org/10.1371/journal.pone.0319503 Editor: Mehran Rahimlou, Zanjan University of Medical Sciences, Iran, Islamic Republic Of Received: May 23, 2024 Accepted: February 3, 2025 Published: April 8, 2025 Copyright: © 2025 Monteiro et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data availability statement: All data are within the paper and its Supporting Information files, as well as at the following Harvard Dataverse link: Monteiro, Sofia; PujolBusquets, Georgina; Smith, James; Larmuth, Kate, 2025, "Replication Data for: The impact of a low-carbohydrate nutrition education program on food preferences: The correspondence between self-report consumption RESEARCH ARTICLE The impact of a low-carbohydrate nutrition education program on food preferences: The correspondence between self-report consumption and supermarket purchases Sofia Monteiro 1*, Georgina Pujol-Busquets 2,3, James Smith4,5,6, Kate Larmuth 4,5,6 1 Global Health Research Group, Kiel Institute for the World Economy, Kiel, Schleswig-Holstein, Germany, 2 Faculty of Health Sciences, Universitat Oberta de Catalunya (Open University of Catalonia, UOC), Barcelona, Catalonia, Spain, 3 MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, Gauteng, South Africa, 4 Health through Physical Activity Lifestyle and Sport Research Centre, Sports Science Institute of South Africa, University of Cape Town, Cape Town, Western Cape, South Africa, 5 International Federation of Sports Medicine (FIMS) Collaborative Centre of Sports Medicine, HPALS, University of Cape Town, Cape Town, Western Cape, South Africa, 6 Division of Physiological Sciences, Department of Human Biology, University of Cape Town, Cape Town, Western Cape, South Africa * sofia.monteir[email protected] Abstract There is reasonable concern that self-reported nutrition assessments do not reflect actual food choices. Yet, a correspondence between both is imperative to evaluate any intervention on food preferences. This paper makes such a comparison. It provides evidence from a low-carbohydrate nutrition education program, which is assessed with both surveys and an incentivized behavioral measure of food choice. The main result is that there is a large correspondence between survey and behavioral measures for our sample of 95 women from two historically underprivileged communities in the Western Cape, South Africa. Compared to the control, the treatment group reported a 35% lower intake from the high-carbohydrate/ ultra-processed food Red List and 60% higher intake from the low-carbohydrate whole foods Green List. The treatment group was also 40% less likely to buy anything from the Red List with a supermarket voucher. In terms of the Green List, the treatment group was significantly more likely to buy eggs, organ meat, traditional fats, avocado and fish but there was no difference in red meat and chicken, non-starchy vegetables and full cream dairy. Low-cost incentivized measures of revealed preferences can be designed to validate subjective habits, increasing confidence in the quality of evidence from nutrition intervention studies. Introduction The burden of non-communicable diseases (NCDs) such as type 2 diabetes is driven to a great extent by what people eat and drink. Hence, influencing dietary behavior for disease prevention and management has become a key challenge for policymakers [1]. In response, PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 2 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences behavioral economists have explored “How do we persuade people to eat better and lose weight?” [2] and “How do we make the healthier choice the easier choice?” [3]. There is arguably a need for evidence-based, sustainable dietary interventions that consider how people allocate their scarce resources (money, time and attention) to improve health [4]. Yet, there is currently still no consensus on how to measure dietary change accurately (in a real-world setting) adding to the challenge of improving people’s dietary preferences. While this is typically performed with self-reported surveys, there is reason to believe that what people say is not always what they do – as is, well known regarding physical activity [5], but debated with respect to dietary intake [6,7]. Our study contributes to an emerging literature on the behavioral economics of food choice [8–14]. Epidemiological observational studies, such as the much-cited Nurses’ Health Study, have significantly influenced public health policy and practice globally [15–18]. Such studies of dietary risk factors use assessments that require subjective responses (namely the Food Frequency Questionnaire (FFQ), 24-hour food recalls and food diaries) to make conclusions about the optimal diet for populations. It is understood that subjective measures of food preferences may suffer from hypothetical bias and a lack of incentive for accuracy [13,19], and in FFQs, there may be noise due to people’s inattention to what they typically eat, inability to recall fully, and social desirability bias [6,7]. We collaborated with the Eat Better South Africa (EBSA) program, a community-based non-profit organization which runs nutrition education programs in Ocean View and Atlantis; two under-resourced communities in the Western Cape. Past programs were six weeks long and involved weekly two-hour educational sessions at a central community hall with a group limited to about 30 women. These sessions aim to teach participants about nutrition, NCDs, shopping on a budget, cooking and how to access healthier foods. The program uses the Noakes Foundation’s traffic lights lists of foods which are available to the public and free to download (alternatively, see S1 Fig). Peer support via the instant messaging group is central to the program, and engagement in the group chat continues after the six-week course ends. In our previous qualitative study, the nutrition education program was evaluated using a different methodology, namely focus group discussions on women’s perceptions of the program, lifestyle choices and shopping habits [20]. Various diets are touted both in the nutrition literature and in public opinion, but the quality of evidence for them varies [21]. We posit that the behavioral measurement of food choice is a complementary tool that could increase methodological credibility of nutrition intervention studies. There is, however, no gold standard for directly assessing the validity of the FFQ [22]. Moreover, an evidence base founded on self-report survey instruments is arguably problematic [6]. Therefore, we pose the question, do revealed food preferences corroborate self-report responses in the FFQ? This is an open empirical question since in another health-related behavior, physical activity, no correlation between self-report responses and behavior was found [5]. The present cross-sectional study compares the food choices of 44 women who had taken part in a low-carbohydrate high-fat nutrition education program to 51 similar women who had not yet taken part but qualified for the program, using two complimentary diet assessment tools: purchases made with a supermarket voucher and an FFQ. Materials and methods Study design We conducted a non-randomized cross-sectional quantitative study with two groups: a treatment group of women who had already completed the EBSA program and a control group consisting of women who are eligible for future programs in the same communities. and supermarket purchases", https://doi. org/10.7910/DVN/G36VF6, Harvard Dataverse, V1, UNF:6:MZOKGXndixByE85oFJUuvg== [fileUNF] Funding: This research was funded by SM’s research budget during a Research Fellowship at The Max Planck Institute for Research on Collective Goods, Bonn. SM’s contribution was also supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC 2167-390884018. KL’s salary was funded by the National Research Foundation of South Africa. GPG is a recipient of the Tim and Marilyn Noakes postdoctoral fellowship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. GPG is a recipient of the Tim and Marilyn Noakes postdoctoral fellowship. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 3 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences The design allowed us to identify the impact of participating in the program by controlling for observable individual characteristics. Our control group provides a convincing counterfactual to the treatment group, since the women share many similar observable characteristics. This means that, if found, a significant difference in revealed food preferences between the two groups can be attributed with a high degree of confidence to the program. Fig 1 provides a flow diagram of the study. Study procedures Participation in the study took approximately 90 minutes. Each participant completed: (a) informed consent; (b) a questionnaire on socioeconomic challenges, medical conditions, food insecurity and shopping habits; (c) a task, in which they purchased food and (non-alcoholic) drink items in a local supermarket with a retail voucher and photographed their groceries and receipt; (d) a Food Frequency Questionnaire; and (e) a feedback questionnaire about their Fig 1. Flow diagram. https://doi.org/10.1371/journal.pone.0319503.g001 PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 4 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences experience. Interview 1 comprised (a) and (b), while Interview 2 later in the week comprised (d) and (e). Ethical compliance The study was conducted in accordance with the Declaration of Helsinki, and approved by the Human Research Ethics Committee at the University of Cape Town (HREC REF 295/2019) on 5 August 2019, as well as by the University of Cologne Faculty of Management Economics and Social Sciences Ethics Committee on 25 November 2019 (Reference: 19024SM) and the protocol amendment making the entire study remote to be compliant with Covid-19 protocols was approved 5 June 2020. Informed consent Informed consent was obtained from all subjects involved in the study. Participants received study information depending on whether they were in the treatment or control group and infographics explaining the steps involved in the study via WhatsApp message. To get informed consent via telephone interview, the research assistant explained the study, sent the participant the Qualtrics survey via WhatsApp, ended the call for a couple of minutes while the participant completed it, then called them back straight away to continue the interview. If they were not able to submit the online consent form because of technological or literacy difficulties, the study information was read out loud to them, the research assistant filled out the form, and an audio recording of the participant’s verbal consent was recorded (22/95 participants). Inclusion and exclusion criteria Eligible participants were: adult women; 18-69 years old; capable of providing informed consent; able to understand and speak English or Afrikaans, and not using private healthcare or private health insurance. To be eligible as a previous program participant, they had attended at least four out of six weekly sessions from one of the four previous Western Cape EBSA programs. Controls were drawn from similar communities through nomination to the study by a participant or program community coach/ambassador. This is typically how the program recruits future program participants who are relatively naïve to its Low-Carbohydrate HighFat (LCHF) dietary advice. Recruitment The recruitment period began 7 July 2020 and ended 28 August 2020. The educators and community coaches were asked to assist with recruiting women who were enrolled in previous programs, either using the WhatsApp group or by telephone. For the non-program participants, a person identified in the community was asked to help recruit women and participants were also invited to nominate a non-program woman to take part in the study. Those women interested in taking part were given the appropriate participant information sheet (i.e., treatment or control) electronically. Women were not necessarily expected to read it but were asked if they would like a researcher to contact them to explain the study. Women who communicated their interest in receiving more information about the study were contacted by a member of the research team. An infographic explaining the 10 steps to complete the study was sent electronically. A researcher explained the study, answered questions, screened for eligibility and invited interested women to take part. PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 5 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences Data sources and processing Supermarket purchases The participants each received a ZAR 250 (~13 USD) voucher via SMS. At the time of the study, this amount equated to about a week’s typical food shopping in our sample population. The shopping activity was held at a major local supermarket. Participants could choose from a selection of fresh produce and cupboard groceries (at current retail prices). A member of the research team explained the activity and conducted a demographic questionnaire. Each participant had a week to purchase her groceries at her convenience. An online shopping account with which the vouchers were purchased allowed the investigators to track which vouchers had been spent. Purchasing decisions were made privately in the store. The participant had access to her phone to keep track of the total cost of her shopping. When she had finished selecting items, she proceeded to the checkout. At home, she took a photo of both her groceries and the till slip, and sent them to the researcher. The researcher ticked off on a checklist with the participant’s code the types of foods she chose to buy and noted the total spent on the receipt. Nutrition survey When the shopping activity was completed, the follow-up interview was booked, usually on the same day or the following day. Participants completed an interviewer-administered FFQ on a phone call with a member of the research team and were asked about foods eaten in the past four weeks. The FFQ was developed according to guidelines [22] and the South African Medical Research Council’s FFQ was adapted to include food items frequently eaten by people following a Low-Carbohydrate High-Fat (LCHF) diet and foods reportedly eaten by previous program participants. It also included standard portion sizes and frequency options. It underwent a period of pilot testing in volunteers that habitually followed a LCHF diet and was modified accordingly [23]. The FFQ data were primarily used to assess the types of foods eaten. Data cleaning and variable transformation The data cleaning and analysis was conducted using Stata 15. To measure average reported dietary intake, the FFQ asked respondents how frequently they had eaten a standard portion of a particular food in the past four weeks. For example, “In the past 4 weeks, how often did you eat a slice of white bread?” to which the participant could choose one of nine fixed responses, e.g., “None”, “1 or less per month”, “2-3 per month, “1-2 per week”, 3-4 per week”, “5-6 per week”, “1 per day”, “2-3 per day” or “3 or more per day”. For the analysis, categorical variables were converted into continuous numerical variables, specifically, an estimate of the number of standard portions eaten per week. e.g., if a participant responded “1 per day”, this was converted to “7 per week”. The food purchases that the sample of women made with a retail voucher were coded as a list of dummy variables. If a food item (e.g., bread) was purchased, this choice was captured as “1”, and “0” otherwise. The selected list of foods corresponded to the EBSA program’s traffic lights list of foods (S1 Fig) as well as the FFQ items for ease of comparison. Presentation of the reported food intake and food choices according to the traffic lights lists makes it straightforward to infer visually whether the sample of women are eating according to the program’s recommendations. Five of the control group’s observations were excluded from the analysis because these non-program women were part of a gym group run by an EBSA community coach (Fig 1) and already eating according to the program’s dietary guidelines (S1 Fig). PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 6 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences A RED index score was created that added up the binary food choice variables, with a maximum score of 8 if all Red List foods were purchased. An ORANGE index score added up the relevant binary food choice variables, with a maximum score of 3 if all three Orange List foods were purchased. A GREEN index score added up the relevant binary food choice variables, with a maximum score of 8 if all eight Green List foods were purchased. Tests and regression models Individual characteristics were examined to motivate that the control group can be used as a valid counterfactual to evaluate the impact of the program. The non-parametric Kruskal-Wallis H test was conducted to test for significant differences (at the conventional significance threshold of 5%) in the distributions of individual and household characteristics. The self-report dietary intake of participants was categorized into the program’s Red, Orange and Green traffic-light lists. To evaluate the impact of the EBSA program with women presumably attempting to follow the recommended LCHF diet, the types of foods eaten by the control and treatment groups were compared. If the program influenced eating habits according to its recommendations, one would expect to see fewer Red List food items and more Green List items consumed by the treatment group compared to the control group. We controlled for observable characteristics in the Ordinary Least Squares (OLS) regression analysis with robust standard errors (which also used the conventional threshold of 5% to evaluate significance). The types of food purchases of the treatment and control groups were compared. In the Linear Probability Model (LPM) regression analysis, we controlled for observable characteristics. Finally, we considered the correspondence between surveyed responses and purchasing behavior, which allowed us to make an evaluation of the validity of the FFQ. If the FFQ responses of our sample correlated well with the incentivized behavioral measure of food preferences, this would support its use as a valid instrument to assess the impact of the nutrition education program on women’s food preferences, while if the two instruments diverged this would undermine the FFQ’s credibility. Results Sample characteristics The treatment and control groups were balanced in terms of household characteristics: number of household members, number of employed household members, shopping frequency and food insecurity (S1 Table). About half of all participants worried they would not have enough food because of lack of money in the preceding four weeks. Table 1 shows that the two groups were well balanced and similar on a number of observable individual characteristics as well. According to the most recent census conducted in 2022 the sample is representative of the ethnicity and language of our two Western Cape communities [24]. It should be noted, however, that the treatment group was significantly older (χ2(1) = 10.760, p = 0.001). The mean age for the control group was about 44 years old, while the mean age for the program group was about 51 years old. In the regression analysis, we controlled for observable characteristics. Almost all the women had children (mode: 2; max: 5), and 75% of the women shared the household with a partner. The women were usually responsible for groceries (χ2(1) = 1.639, p = 0.2005). On average, households had four members, with one or two members employed. A higher proportion of the control group were employed (53% compared to 32% (χ2(1) = 4.251, p = 0.0392) but there was no significant difference between the two groups at the 5% level in the proportion of working-age individuals employed, χ2(1) = 2.833, p = 0.0924. In the pooled sample, 43% had completed primary school only, 35% had completed high school and 21% PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 7 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences higher education. The treatment group also had more participants whose highest level of education was primary school, and this difference was significant at the 5% level, χ2(1) = 5.226, p = 0.0222. However, there was no significant difference at the 5% level between the groups in basic education attainment (where primary school and high school are pooled), χ2(1) = 3.378, p = 0.0661. In South Africa, the returns to education are convex, in that the marginal rate of return is extremely high for tertiary levels of education and approaches zero for lower levels of education [25]. Thus, completing school versus not is arguably the more relevant characteristic with which to consider differences in life chances between the two groups. There were no significant differences between the groups in non-communicable disease risk factors at the 5% level, however, the control group had a marginally lower incidence of diagnosed high blood pressure at the 10% level. The positive effect of the program on reported food intake in the past four weeks Fig 2 shows histograms of total weekly consumption of standard portions of foods organized by the program’s traffic lights lists of foods (S1 Fig) where, according to the EBSA program, the Red List is to be avoided, the Orange List for occasional consumption, and the Green List to be eaten liberally. The Red List includes foods high in carbohydrates and seed oils such as fast food, sugar, bread, pasta, potatoes. The Orange List includes fruit, such as bananas and apples. The Green List includes eggs, unprocessed meat, unsweetened full cream dairy, green leafy vegetables. In each histogram, lighter bars indicate the distribution of the control group’s responses, darker bars indicate the distribution of the treatment group’s responses, and darkest bars show where the distributions overlap. The treatment’s total weekly consumption from Table 1. Sample characteristics. Treatment Group Control (n = 51) Treatment (n = 44) Age (years) 43.76 50.84 1 Lives with partner 75% 75% Has children 96% 93% Buys household groceries 94% 86% Work status Employed 53% 32% 1 Employed & working age 43% 38% Education No formal schooling 0 2% Primary school 37% 55% 1 High school 37% 32% College or University 27% 9% Post-graduate degree 2% 2% Health risk factors Smoker (daily) 12% 7% Alcohol (past year) 39% 41% Elevated blood pressure 33% 52% Blood pressure medication 16% 32% Diabetes 12% 7% CVD event or chest pain 4% 14% 1Significant p-value at the 5% level in Kruskal-Wallis H tests, which were conducted as a balance check between control and treatment groups., i.e., groups were unbalanced on this characteristic. https://doi.org/10.1371/journal.pone.0319503.t001 PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 8 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences the Red List is clearly left-shifted compared to the control (i.e., fewer standard portions of Red List items were consumed, K-Smirnov test: Treatment <Control, p < 0.001). In contrast, for the Green List, the distributions are reversed in favor of the program (Control <Treatment, p < 0.001). In general, for the Orange List, there is not much interesting variation between treatment and control. As we did not have particular hypotheses for the consumption from the Orange List, we focus on the Green and Red Lists in the main paper and report the Orange List results in S1 File supplementary text, for completeness. Fig 3 shows coefficient plots from OLS regressions of the number of standard portions per week on a binary variable = 1 if treatment, and 0 if control, as well as explanatory variables age and highest education. Each regression coefficient illustrated originates from a separate model. We modelled the effect of participating in the program on total weekly intake of Red, Orange and Green List foods, respectively. The effect on reported consumption of particular foods should be interpreted with caution because of multiple hypothesis testing concerns with a sample of 95 women. The OLS models in Fig 3 considered only the FFQ data. The corresponding regression results are reported in S2 Table. The treatment group reported significantly fewer total Red List portions per week, i.e., 56 in treatment versus 86 in control, a 35% reduction in unhealthy food by the treatment group, from the perspective of the program. Fig 2. Overview of reported food intake in the past four weeks by the treatment and control group. Histograms indicate the distribution of food intake by a sample of women who have completed the EBSA program (n = 44) versus similar women who are eligible for future programs in the community (n = 51). The data are categorized by the low-carb program’s traffic lights lists of foods. See S1 Fig for detailed Green, Orange and Red Lists. Darker colored bars indicate the treatment group, while lighter bars indicate the control group. The treatment group is significantly different on Red and Green List, but not Orange List (non-parametric K-Smirnov tests for equality of distribution: p < 0.001). https://doi.org/10.1371/journal.pone.0319503.g002 PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 15 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences Author contributions Conceptualization: Sofia Monteiro, Georgina Pujol-Busquets, James Smith, Kate Larmuth. Data curation: Sofia Monteiro. Formal analysis: Sofia Monteiro. Funding acquisition: Sofia Monteiro. Investigation: Sofia Monteiro, Kate Larmuth. Methodology: Sofia Monteiro, Georgina Pujol-Busquets, James Smith, Kate Larmuth. Project administration: Sofia Monteiro, Kate Larmuth. Resources: Sofia Monteiro. Software: Sofia Monteiro. Supervision: Kate Larmuth. Visualization: Sofia Monteiro. Writing – original draft: Sofia Monteiro. Writing – review & editing: Sofia Monteiro, Georgina Pujol-Busquets, Kate Larmuth. References 1. Ong KL, Stafford LK, McLaughlin SA, Boyko EJ, Vollset SE, Smith AE, et al. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2023;402(10397):203–34. https://doi. org/10.1016/S0140-6736(23)01301-6 PMID: 37356446 2. Roberto C, Kawachi I, editors. Behavioral economics and public health. New York: Oxford University Press; 2016. 3. Hallsworth M. Rethinking public health using behavioural science. Nat Hum Behav. 2017;1(9):612. https://doi.org/10.1038/s41562-017-0188-0 PMID: 31024132 4. Cawley J. An economic framework for understanding physical activity and eating behaviors. Am J Prev Med. 2004;27(3 Suppl):117–25. https://doi.org/10.1016/j.amepre.2004.06.012 PMID: 15450622 5. Prince SA, Adamo KB, Hamel ME, Hardt J, Connor Gorber S, Tremblay M. A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act. 2008;5:56. https://doi.org/10.1186/1479-5868-5-56 PMID: 18990237 6. Archer E, Lavie CJ, Hill JO. The failure to measure dietary intake engendered a fictional discourse on diet-disease relations. Front Nutr. 2018;5:105. https://doi.org/10.3389/fnut.2018.00105 PMID: 30483510 7. Medic N, Ziauddeen H, Forwood SE, Davies KM, Ahern AL, Jebb SA, et al. The presence of real food usurps hypothetical health value judgment in overweight people. eNeuro. 2016;3(2): ENEURO.0025-16.2016. https://doi.org/10.1523/ENEURO.0025-16.2016 PMID: 27280152 8. Ginon E, Combris P, Lohéac Y, Enderli G, Issanchou S. What do we learn from comparing hedonic scores and willingness-to-pay data?. Food Quality and Preference. 2014;33:54–63. https://doi. org/10.1016/j.foodqual.2013.11.003 9. List J, Samek A. The behavioralist as nutritionist: Leveraging behavioral economics to improve child food choice and consumption. J Health Econ. 2017 [cited 2 Aug 2019]. Available from: https://www. sciencedirect.com/science/article/pii/S0167629614001398 10. Crosetto P, Muller L, Ruffieux B. Helping consumers with a front-of-pack label: Numbers or colors?: Experimental comparison between Guideline Daily Amount and Traffic Light in a diet-building exercise. J Econ Psychol. 2016;55:30–50. 11. Müller L, Prevost M. What cognitive sciences have to say about the impacts of nutritional labelling formats. Journal of Economic Psychology. 2016;55:17–29. 12. VanEpps EM, Downs JS, Loewenstein G. Calorie label formats: using numeric and traffic light calorie labels to reduce lunch calories. Journal of Public Policy & Marketing. 2016;35(1):26–36. https://doi. org/10.1509/jppm.14.112 13. Alemu MH, Olsen SB. Can a repeated opt-out reminder mitigate hypothetical bias in discrete choice experiments? An application to consumer valuation of novel food products. European Review of Agricultural Economics. 2018;45(5):749–82. https://doi.org/10.1093/erae/jby009 PLOS ONE | https://doi.org/10.1371/journal.pone.0319503 April 8, 2025 16 / 16 PLOS ONE The impact of a low-carbohydrate nutrition education program on food preferences 14. Charness G, Cobo-Reyes R, Eyster E, Sánchez Á, Sutter M. Improving healthy eating in children: Experimental evidence. MPI Collective Goods Discussion Paper. 2020;1. 15. Liu S, Stampfer MJ, Hu FB, Giovannucci E, Rimm E, Manson JE, et al. Whole-grain consumption and risk of coronary heart disease: results from the Nurses’ Health Study. Am J Clin Nutr. 1999;70(3):412– 9. https://doi.org/10.1093/ajcn/70.3.412 PMID: 10479204 16. Colditz GA, Hankinson SE. The nurses’ health study: lifestyle and health among women. Nat Rev Cancer. 2005;5(5):388–96. https://doi.org/10.1038/nrc1608 PMID: 15864280 1 7. Colditz GA, Philpott SE, Hankinson SE. The impact of the nurses’ health study on population health: prevention, translation, and control. Am J Public Health. 2016;106(9):1540–5. https://doi.org/10.2105/ AJPH.2016.303343 PMID: 27459441 18. Crous-Bou M, Fung TT, Prescott J, Julin B, Du M, Sun Q, et al. Mediterranean diet and telomere length in Nurses’ health study: population based cohort study. BMJ. 2014;349:g6674. https://doi. org/10.1136/bmj.g6674 PMID: 25467028 19. Lusk J, Schroeder T. Are choice experiments incentive compatible? A test with quality differentiated beef steaks. American Journal of Agricultural Economics. 2004;86:467–82. https://doi. org/10.2307/30139569 20. Pujol-Busquets G, Smith J, Larmuth K, Fàbregues S, Bach-Faig A. Exploring the perceptions of women from under-resourced south african communities about participating in a lowcarbohydrate high-fat nutrition and health education program: a qualitative focus group study. Nutrients. 2020;12(4):894. https://doi.org/10.3390/nu12040894 PMID: 32218249 21. Freire R. Scientific evidence of diets for weight loss: different macronutrient composition, intermittent fasting, and popular diets. Nutrition. 2020;69:110549. https://doi.org/10.1016/j.nut.2019.07.001 PMID: 31525701 22. Cade J, Thompson R, Burley V, Warm D. Development, validation and utilisation of food-frequency questionnaires - a review. Public Health Nutr. 2002;5(4):567–87. https://doi.org/10.1079/PHN2001318 PMID: 12186666 23. Webster CC, Murphy TE, Larmuth KM, Noakes TD, Smith JA. Diet, diabetes status, and personal experiences of individuals with Type 2 diabetes who self-selected and followed a low carbohydrate high fat diet. Diabetes Metab Syndr Obes. 2019;12:2567–82. https://doi.org/10.2147/DMSO.S227090 PMID: 31827331 24. Statistics South Africa. Census 2022. Pretoria; 2023 Oct. Available from: https://census.statssa.gov. za/#/ 25. Keswell M, Poswell L. Returns to education in South Africa: a retrospective sensitivity analysis of the available evidence. South African J Economics. 2004;72(4):834–60. https://doi. org/10.1111/j.1813-6982.2004.tb00136.x