Barriers and Facilitators of Healthy Eating in Adults

Research Collection

Doctoral Thesis

Burger versus broccoli - Barriers and facilitators of healthy eating in adults

Author(s): Hagmann, Désirée

Publication Date: 2019-10

Permanent Link: https://doi.org/10.3929/ethz-b-000372537

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library

Diss. ETH NO. 26255

Burger versus broccoli – Barriers and facilitators of healthy eating in adults

Dissertation for the degree of Doctor of Science of ETH Zurich

Désirée Hagmann

DISS. ETH NO. 26255

BURGER VERSUS BROCCOLI – BARRIERS AND FACILITATORS OF HEALTHY EATING IN ADULTS

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH (Dr. sc. ETH Zurich)

presented by DÉSIRÉE ASTRID HAGMANN MSc in Psychology, University of Basel born on 22.01.1984 citizen of Olten SO

accepted on the recommendation of Prof. Dr. Michael Siegrist Dr. Christina Hartmann Prof. Dr. Joachim Westenhöfer

2019

This thesis was written in the context of the project Swiss Food Panel 2.0 which is funded by the Department of Health Sciences and Technology, ETH Zurich.

TABLE OF CONTENTS

TABLE OF CONTENTS III

LIST OF TABLES VI

LIST OF FIGURES VIII

SUMMARY X

ZUSAMMENFASSUNG XII

1 General Introduction 14

1.1 Introduction 15

1.2 Healthy eating – definitions and evaluation methods 17

1.3 Determinants of food choices 20

1.4 Individual determinants of food choices 21

1.5 Environmental determinants of food choices 25

1.6 The Swiss Food Panel 2.0 27

1.7 Research questions and overview of the thesis 30

References 32

2 Acquisition of cooking skills and their importance for healthy eating 39

2.1 Introduction 41

2.2 Methods 43

2.3 Results 47

2.4 Discussion 53

References 56

3 Motives for meat avoidance and reduced meat intake 60

3.1 Introduction 62

3.2 Methods 64

3.3 Results 69

III

3.4 Discussion 78

References 81

4 Intuitive eating and food choices 85

4.1 Introduction 87

4.2 Methods 89

4.3 Results 99

4.4 Discussion 105

References 112

5 Self-control as a moderator of the effect of hedonic hunger on overeating and snacking 115

5.1 Introduction 117

5.2 Methods 119

5.3 Results 123

5.4 Discussion 130

References 132

6 Public acceptance of interventions aimed at reducing sugar intake 135

6.1 Introduction 137

6.2 Methods 140

6.3 Results 145

6.4 Discussion 156

References 160

7 Nutrition labels and their effect on healthiness perception of salty snack food 165

7.1 Introduction 167

7.2 Materials and Methods 171

7.3 Results 179

7.4 Discussion 185

7.5 Supplementary Materials 189

References 191

8 General Discussion 195

8.1 Introduction 196

8.2 Central findings and implications for research and practice 196

8.3 Limitations of the studies 204

8.4 Final conclusions 208

References 209

DANKSAGUNG 213

CURRICULUM VITAE 214

V LIST OF TABLES

Table 1.1 Classification of weight status in adults based on the Body Mass Index ([BMI]; WHO, 2008)...... 16 Table 1.2 Dietary recommendations for healthy adults (Swiss Society for Nutrition, 2011)...18 Table 1.3 Socio-demographic characteristics of the Food Panel 2.0 study sample...... 29 Table 1.4 Overweight and obesity in males and females of the Swiss Food Panel 2.0 compared to the general Swiss population...... 29 Table 1.5 Overview of the chapters, topics and research questions included in the thesis....31 Table 2.1 Description of the study sample...... 43 Table 2.2 Associations of cooking skills with food intake, diet quality, and eating habits in male and female participants of the Swiss Food Panel survey 2018...... 48 Table 2.3 Predictors of cooking skills among men and women...... 52 Table 3.1 Items assessing motives for no or low-meat consumption...... 65 Table 3.2 Foods and beverages included in the diet quality index...... 68 Table 3.3 Perceived difficulty of low or no meat intake, diet quality, and weight of adult males and females separated by self-reported diet styles...... 71 Table 3.4 Consumption of animal- and plant-based proteins (weekly portions) by adult males...... 73 Table 3.5 Consumption of animal- and plant-based proteins (weekly portions) in adult females...... 74 Table 3.6 Hierarchical multiple regression analysis predicting total meat consumption in adults who reported eating little or no meat...... 76 Table 3.7 Correlations of total meat consumption and predictors. Calculations based on the subsample of participants who reported eating little or no meat...... 77 Table 4.1 Sociodemographic characteristics of the study population...... 90 Table 4.2 Food group compositions, standard portion definitions, mean values and standard deviations for the number of standard portions consumed per week of the different food groups assessed in the semiquantitative food frequency questionnaire. .... 93 Table 4.3 Rotated component matrix for the IES-2 for different subsamples of the Swiss Food Panel...... 95 Table 4.4 Cronbach’s alpha for the Intuitive Eating Scale-2 (IES-2) total score and the four subscales separately for gender and the two language versions...... 99 Table 4.5 Rotated component matrix of the IES-2 (varimax rotation), data from the Swiss Food Panel...... 100

Table 4.6 Means, standard deviations and inter-correlations for the Intuitive Eating Scale-2 (IES-2) total score and the four subscales...... 103 Table 4.7 Pearson correlations for the associations between Intuitive eating (IES-2), food intake, physical activity level and BMI...... 104 Table 4.8 Pearson correlations for the associations between Intuitive eating (IES-2) and participants’ perceptions of their diet and diet-related health consciousness. ... 107 Table 5.1 Sociodemographic characteristics, hedonic hunger, and self-control in males and females of the study sample...... 119 Table 5.2 Mean values (SD) and correlations for Power of Food Scale scores, overeating and snacking variables...... 121 Table 5.3 Results of four moderation analyses with self-control as moderator for the relationship between Power of Food and overeating frequency as well as snacking behavior...... 124 Table 6.1 Items used to assess acceptance of different interventions to reduce sugar intake (English translation)...... 143 Table 6.2 Pearson correlations between socio-demographic, dietary, and health-related variables...... 146 Table 6.3 Mean acceptance of different interventions for participants from the German- and French-speaking parts...... 150 Table 6.4 Hierarchical regression analysis predicting general acceptance of interventions aimed at reducing sugar intake in the population...... 151 Table 6.5 Mean acceptance ratings of the four clusters for each intervention strategy.....153 Table 6.6 Description of the four clusters by means of different characteristics...... 154 Table 7.1 Characteristics of the salty snack products used in the choice task...... 172 Table 7.2 Study participants in each condition: Recruited sample, excluded participants and demographic characteristics...... 178 Table 7.3 Exploratory analysis of the perceived usefulness of different types of nutrition information...... 183 Table 7.4 Public support of a mandatory implementation of the MTL and Nutri-Score labels...... 184

VII LIST OF FIGURES

Figure 1.1 An ecological framework of the determinants of food choices (Story et al., 2008)...... 20 Figure 1.2 Flow chart of the Swiss Food Panel 2.0 study sample...... 28 Figure 2.1 Boxplots of cooking skills among men and women...... 47 Figure 2.2 Gender differences in the importance of various sources for the acquisition of cooking skills...... 49 Figure 2.3 Involvement in cooking activities during childhood/adolescence at different ages among men and women...... 50 Figure 3.1 Motives for vegetarianism and low-meat consumption in different diet styles, separated by gender...... 70 Figure 5.1 Proposed model explaining overeating frequency, high sugar foods consumption, high fat salty snack consumption and overall snacking frequency by power of food scale (PFS) scores and self-control...... 118 Figure 5.2 Association between Power of Food and overeating frequency for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control...... 126 Figure 5.3 Association between Power of Food and frequency of high fat salty snack foods intake for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control...... 127 Figure 5.4 Association between Power of Food and frequency of high sugar foods intake for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control...... 128 Figure 5.5 Association between Power of Food and frequency of snacking for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control...... 129 Figure 6.1 Gender differences in the consumption frequencies of foods and beverages...... 145 Figure 6.2 Mean acceptance ratings for different interventions and 95% confidence intervals in the study sample...... 147 Figure 6.3 Distribution of the acceptance scores in the sample for each intervention type...149 Figure 7.1 Product examples for the experimental conditions...... 175 Figure 7.2 Boxplots of the proportion of correct choices in the five conditions. The objective healthiness of the snack products was determined on the basis of the UK Ofcom/FSA nutrient profiling model (Food Standards Agency, 2011)...... 180

VIII

Figure 7.3 Boxplots of the average weighted inaccuracy per comparison in the five conditions. The objective healthiness of the snack products was determined on the basis of the UK Ofcom/FSA nutrient profiling model (Food Standards Agency, 2011). .. 181 Figure 7.4 Boxplots of the proportion of correct choices in the five conditions. The objective healthiness of the snack products was determined on the basis of the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014)...... 190 Figure 7.5 Boxplots of the average weighted inaccuracy per comparison in the five conditions. The objective healthiness of the snack products was determined on the basis of the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014)...... 191

IX SUMMARY

SUMMARY

Along with other risk factors, dietary behaviour plays a central role in the development of becoming overweight and other preventable lifestyle diseases. Similar to most industrial nations, many peoples’ diets in Switzerland are not as balanced as recommended by official dietary guidelines. In the present doctoral thesis, several individual and environmental factors have been investigated that are associated with healthy eating. Most results are based on data from the Swiss Food Panel 2.0 study, a longitudinal study on the dietary behaviours of adults in Switzerland (N = 5,586 in 2017). In addition, to investigate another research question, a randomised online experiment was conducted. On the level of individual determinants, it was shown that better cooking skills were associated with a higher diet quality, particularly with a higher consumption of vegetables, and among women only, also with a lower consumption of snacks and sugar-sweetened beverages. Mothers represented the most important source of learning for the acquisition of cooking skills. Males also learn from their spouses/partners, whereas females make stronger use of cooking courses and various media. Stronger involvement in cooking activities during childhood and adolescence has a positive effect on the development of cooking skills. Participants from younger generations, on average, reported that they had been involved in these activities more often. Low meat consumption is desirable for a balanced as well as for an environmentally sustainable diet. Many consumers perceive their meat consumption as reduced even though it is higher than the recommended amounts. For vegetarians, vegans and pescatarians, ethical reasons (animal welfare, environment) and taste preferences are more important motives than for omnivores, who stated that they deliberately reduce their meat consumption. Female participants from the latter group avoid meat more frequently because they want to reduce their weights. Concerns regarding animal welfare and a preference for vegetarian meals promoted a lower meat consumption, whereas perceived difficulties in renouncing meat in everyday life and abstaining for reasons of weight regulation were associated with a higher meat intake. An intuitive eating style is propagated as a promising alternative to dieting. Cross- sectionally, intuitive eaters exhibit a lower BMI and a lower tendency to overeat; however, for most aspects of intuitive eating, there are hardly any associations with healthier food choices. The ‘unconditional permission to eat’, on the contrary, is associated even with a lower diet quality. Hedonic hunger, or the tendency to eat for reasons of pleasure without an existing energy deficit, does not necessarily lead to overeating and increased snack consumption. High self-control regarding food intake can be protective in individuals with a strong tendency to hedonical hunger and can weaken this association.

X SUMMARY

On the level of the environmental determinants, it could be concluded that nutrition labelling with the Nutri-Score label and the MTL label leads to a somewhat higher accuracy in identifying healthier snack options compared to the mandatory nutrition facts table or the absence of nutrition information on product packaging. For the recently introduced Nutri-Score, it could be concluded that such a label only contributes to more accurate healthiness evaluations when all available products have the label and can be compared by consumers accordingly. Another study investigated the public acceptance of various governmental interventions aimed at reducing sugar intake. The results showed a clear preference of the public for interventions that sensitise and inform regarding this issue (information campaigns, symbol on high-sugar foods), whereas the majority was against the implementation of a sugar tax. Study participants from the German-speaking part, males, and some risk groups (overweight individuals, participants with a high consumption of sugar-sweetened beverages) rather oppose such public health interventions. Higher health consciousness, particularly higher consciousness regarding individual sugar consumption, is associated with a stronger acceptance of these measures. Healthy food choices are influenced by various individual and environmental determinants. Public health interventions should therefore both strengthen individual competences and health-promoting attitudes and optimise the conditions of the food environment to promote healthy eating in the population.

XI ZUSAMMENFASSUNG

ZUSAMMENFASSUNG

Das Ernährungsverhalten spielt neben anderen Risikofaktoren eine zentrale Rolle bei der Entstehung von Übergewicht und weiteren vermeidbaren Zivilisationskrankheiten. Ähnlich wie in den meisten Industrienationen ernähren sich auch in der Schweiz viele Menschen nicht so, wie es von den offiziellen Ernährungsrichtlinien empfohlen wird. In der vorliegenden Dissertation wurden verschiedene individuelle und umweltbezogene Faktoren untersucht, welche mit einer gesunden Ernährungsweise in Zusammenhang stehen. Die meisten Ergebnisse basieren auf Daten des Ernährungspanels Schweiz 2.0, einer Längsschnittstudie zum Ernährungsverhalten von erwachsenen Schweizern (N = 5,586 in 2017). Für eine weitere Fragestellung wurde zusätzlich ein randomisiertes Online-Experiment durchgeführt. Auf der Ebene der individuellen Einflussfaktoren konnte gezeigt werden, dass bessere Kochfertigkeiten mit einer höheren Ernährungsqualität einhergehen, insbesondere mit einem höheren Konsum an Gemüse und bei Frauen auch mit einem geringeren Konsum an Snacks und Süssgetränken. Mütter stellen die wichtigste Quelle für den Erwerb von Kochfertigkeiten dar. Männer lernen ausserdem von ihren Partnerinnen, während Frauen vermehrt Kochkurse und diverse Medien nutzen. Ein stärkerer Einbezug in Kochaktivitäten im Kindes- und Jugendalter wirkt sich positiv auf die Entwicklung von Kochfertigkeiten aus. TeilnehmerInnen jüngerer Generationen berichteten, im Schnitt häufiger in diese Tätigkeiten einbezogen worden zu sein. Ein geringer Fleischkonsum ist sowohl für eine ausgewogene als auch für eine nachhaltige Ernährung wünschenswert. Viele Konsumenten schätzen ihren Fleischkonsum als reduziert ein, obwohl dieser deutlich über den empfohlenen Mengen liegt. Für Vegetarier, Veganer und Pesketarier spielen ethische Gründe (Tierwohl, Umwelt) und Geschmackspräferenzen für den Fleischverzicht eine stärkere Rolle als für Omnivoren, welche ihren Fleischkonsum bewusst reduzieren. Weibliche Teilnehmer der letzteren Gruppe hingegen verzichten häufiger auf Fleisch, weil sie ihr Gewicht reduzieren möchten. Bedenken bezüglich des Tierwohls sowie eine Präferenz für vegetarische Gerichte sind förderlich für einen geringen Fleischkonsum, während wahrgenommene Schwierigkeiten im Alltag auf Fleisch zu verzichten und ein Verzicht aus Gewichtsgründen eher mit höherem Konsum in Zusammenhang stehen. Ein intuitiver Ernährungsstil wird als vielversprechende Alternative zu Diäten propagiert. Intuitive Esser weisen querschnittlich zwar einen geringeren BMI und eine geringere Tendenz zum Überessen auf, jedoch bestehen für die meisten Teilaspekte von intuitvem Essen kaum Zusammenhänge mit einer gesünderen Lebensmittelwahl. Die ‘bedingungslose Erlaubnis zu essen’ hängt im Gegenteil sogar mit einer geringeren

XII ZUSAMMENFASSUNG

Ernährungsqualität zusammen. Hedonischer Hunger, oder die Tendenz aus Lust zu essen ohne ein vorhandenes Energiedefizit, muss nicht zwingend zu Überessen und erhöhtem Snackkonsum führen. Eine hohe Selbstkontrolle in Bezug aufs Essen kann in Individuen mit starker Neigung zu hedonischem Hunger schützend wirken und diesen Zusammenhang abschwächen. Auf der Ebene der Umweltfaktoren konnte gezeigt werden, dass die Nährwertkennzeichnung mittels Nutri-Score oder Lebensmittelampel die Fähigkeit von Konsumenten im Vergleich zur obligatorischen Nährwerttabelle auf der Verpackungsrückseite oder fehlender Information etwas erhöht, gesündere Snackalternativen zu identifizieren. Am Beispiel des kürzlich in Frankreich eingeführten Nutri-Score konnte jedoch gezeigt werden, dass ein solches Label nur dann zu einer genaueren Gesundheitseinschätzung des Lebensmittels beiträgt, wenn alle verfügbaren Produkte dieses aufweisen und vom Konsumenten entsprechend verglichen werden können. Eine weitere Studie zur Akzeptanz verschiedener staatlicher Interventionen zur Reduktion des Zuckerkonsums ergab eine klare Präferenz der Bevölkerung für Interventionen, die für das Thema sensibilisieren und informieren (Informationskampagnen, Symbol auf stark zuckerhaltigen Lebensmitteln), während eine Mehrheit gegen eine Zuckersteuer ist. Studienteilnehmer aus der Deutschschweiz, Männer sowie gewisse Risikogruppen (Übergewichtige, Personen mit hohem Süssgetränkekonsum) lehnen staatliche Interventionen tendenziell eher ab. Ein höheres Gesundheitsbewusstsein, insbesondere ein höheres Bewusstsein über den eigenen Zuckerkonsum, hängt mit stärkerer Akzeptanz zusammen. Eine gesunde Lebensmittelwahl wird von verschiedenen individuellen und umweltbezogenen Faktoren beeinflusst. Staatliche Gesundheitsmassnahmen sollten daher sowohl an der Stärkung individueller Kompetenzen und gesundheitsförderlicher Einstellungen ansetzen, als auch an der Optimierung der essensbezogenen Umweltbedingungen, um ein gesundes Ernährungsverhalten in der Bevölkerung zu fördern.

XIII

Chapter 1

General Introduction 1 General Introduction

GENERAL INTRODUCTION

1.1 Introduction

Eating is essential to providing the body with the energy and nutrients needed to survive and to master the demands of everyday life. In modern food environments in which highly palatable foods are easily available, the challenge is to avoid an excessive consumption of calories rather than to suffer from nutrient deficiencies (Harris & Mattes, 2008). The World Health Organization (WHO) classified unhealthy diets along with physical inactivity, smoking and alcohol use as the major preventable risk factors for health (WHO, 2003). However, although a healthy diet would substantially reduce the risk for various diseases and premature death, adherence to dietary guidelines is in general rather low in many countries, including Switzerland (Schneid Schuh, Campos Pellanda, Guessous, & Marques-Vidal, 2018). In most industrialised countries, unhealthy dietary patterns are highly prevalent. The average energy intake of adults has increased over the last decades, and the portion sizes of foods and beverages offered by food suppliers have steadily become larger (Duffey & Popkin, 2011). Many adults also do not consume sufficient amounts of fruits and vegetables, which are important sources of nutrients and have a satiating effect due to their high content of dietary fibre (Boeing et al., 2012; Mytton, Nnoaham, Eyles, Scarborough, & Mhurchu; Slavin & Lloyd, 2012). Rather, a large proportion of energy is consumed from animal-based fat and protein, refined grains and added sugar (Malik, Willett, & Hu, 2013). In many countries, including Switzerland, the consumption of meat, processed meat, dairy products and eggs is higher than recommended by official dietary guidelines (FAO, 2018; Swiss Society for Nutrition, 2011; Westhoek et al., 2014; World Cancer Research Fund International, 2018). Similarly, the amount of energy consumed from dietary sugars is considered critically high among Europeans (Azais-Braesco, Sluik, Maillot, Kok, & Moreno, 2017). The consumption of sugar- sweetened beverages (SSBs) constitutes a particular problem because they supply a considerable amount of ‘empty’ calories but do only have a weak satiating effect and therefore may lead to even greater energy intake compared to sugar consumption from solid foods (Johnson et al., 2009). Similarly, energy consumption from alcohol increased in the world’s population with the highest per capita alcohol consumption observed in the European region (WHO, 2018a). Regular physical activity (PA) is important for maintaining good health and as a key determinant of energy expenditure essential for weight control (WHO, 2010). Worldwide, many adults are not physically active to a sufficient extent (Bauman et al., 2012). Generally, levels of PA carried out for work, transportation and household chores have declined, whereas sedentary behaviours, such as time spent in front of the television or computer, have substantially increased (Malik et al., 2013). For a sufficient PA, WHO recommends for adults aged 18–64 years to do moderate-intensity PA at least 150 minutes/week or vigorous-intensity

15 GENERAL INTRODUCTION

PA at least 75 minutes/week or an equivalent combination of both (WHO, 2010). PA beyond these recommendations is considered to provide additional health benefits (WHO, 2010). According to the Swiss Federal Statistical Office (2018a), in 2017, around a quarter of the Swiss population aged from 15 years up (around 22% of males and 26% of females) was below the minimum PA level recommended by health authorities. An overconsumption of calories combined with a lack of PA over a longer period results in a positive energy balance, which is the main cause of overweight and obesity. In 2016, worldwide, almost two billion adult people had excessive body weight, with around 39% being overweight and around 13% being obese (WHO, 2018b). In Switzerland, around 42% of the adult population is either overweight or obese, with higher rates observed in males and with increasing age up to 74 years (Swiss Federal Statistical Office, 2018b). Evidence from numerous epidemiological studies suggests that excessive body weight increases the risk for various noncommunicable diseases, including cardiovascular diseases, type 2 diabetes and some forms of cancer (WHO, 2018b). In addition, excessive body weight is frequently associated with psychosocial problems, such as depression, body dissatisfaction or social stigma (Chu et al., 2019; Puhl & Heuer, 2010). For the determination of weight status in adults, the Body Mass Index (BMI) is the most commonly used and simple method (WHO, 2008). For the calculation of BMI, the person’s body weight (in kg) is divided by his/her squared body height (in m2). According to this method, adults are overweight if their BMI is greater than or equal to 25 kg/m2 and obese if their BMI is greater than or equal to 30 kg/m2 (see Table 1.1). BMI was also used in the Swiss Food Panel 2.0 study because it represents the most feasible and economic method for use in a large population study.

Table 1.1 Classification of weight status in adults based on the Body Mass Index ([BMI]; WHO, 2008).

BMI Nutritional status < 18.5 Underweight 18.5–24.9 Normal weight 25.0–29.9 Pre-obesity (Overweight) 30.0–34.9 Obesity class I 35.0–39.9 Obesity class II ³ 40 Obesity class III

16 GENERAL INTRODUCTION

1.2 Healthy eating – definitions and evaluation methods

Searching for the term ‘healthy eating’ on Google delivers millions of hits with various sometimes contradictory information regarding which foods and diets are more beneficial or less beneficial for achieving good health. Numerous nutritional epidemiological studies have investigated associations of dietary intake with morbidity and mortality risks (Willett & Stampfer, 2013), and they have contributed to a better understanding of the role of nutrition in health promotion and to the current definition of what constitutes a healthy diet. In the scientific literature, different approaches are used to determine the healthiness of populations’ diets and of single food products. In the following sections, some approaches that are relevant to the studies included in the present thesis are addressed.

1.2.1 Adherence to recommendations from dietary guidelines

In public health communication and nutrition counselling, recommendations regarding the elements of a healthy diet are mainly made for food groups rather than for nutrients because this is easier for consumers to understand and to implement in their daily food choices (Willett & Stampfer, 2013). Accordingly, in many dietary studies, the consumption of these food groups is measured to determine adherence to existing dietary guidelines. Dietary recommendations have dramatically changed throughout history (Ogden, 2010) with a shift of the focus away from the prevention of deficiency diseases to the contemporary goal of promoting long-term health (Willett & Stampfer, 2013). Although some aspects are controversly discussed in the literature, there is largely a consensus among nutritionists regarding the main components of a healthy diet (Ogden, 2010). Based on solid evidence from nutritional epidemiological research, diets are generally considered beneficial for health when they contain plenty of fruits and vegetables, plant-based sources of fat and protein, fish, nuts and whole grains (Willett & Stampfer, 2013), whereas the intake of meat (especially red and processed meat) and sources of refined carbohydrates and saturated fat should be limited (Willett & Stampfer, 2013). Despite this, several national and international dietary guidelines slightly vary in their specific recommendations, particularly regarding the recommended serving sizes. In the following section, a short summary of the current recommendations for the intake of the most important food groups and nutrients is provided, oriented specifically on the dietary guidelines of the Swiss Society for Nutrition (2011), which were the basis for the studies presented in this thesis. The recommendations for a balanced diet for healthy adults do not differ between males and females. Table 1.2 provides a brief summary of the most important recommendations for a balanced diet in healthy adults (Swiss Society for Nutrition, 2011).

17 GENERAL INTRODUCTION

Table 1.2 Dietary recommendations for healthy adults (Swiss Society for Nutrition, 2011).

Fruits and Should be consumed in plenty, at least 5 a day in different colours vegetables (3 portions of vegetables, 2 portions of fruits). Cereal products Whole-grain products should be preferred over products made of refined grains (e.g., for bread, rice, pasta) due to their higher content of dietary fibre. Meat and meat Should be consumed in moderation, no more than 2–3 portions of products 100–120 g per week. Dairy products A sufficient consumption of these products is recommended in order to ensure a sufficient intake of protein, calcium and vitamin D. Fish 1–2 portions of 100–120 g per week (dependent on the type of fish; particularly sea fish is a good source of Omega-3 fatty acids; however, for sustainability reasons not a more frequent consumption is recommended) Fats Vegetable fats should be preferred over animal fats. Sweets, sugar- Should be enjoyed in moderation and in small portions. For sugar, sweetened the WHO (2015a) additionally recommends that the sugar intake beverages, and should not exceed 10%, or better, 5%, of daily energy intake. For an salty snacks average adult person with a daily caloric intake of 2000 calories, this corresponds to a maximum of 50 g (or 25 g) of sugar per day.

Alcohol Should be enjoyed in moderation. Salt (sodium) No more than 5 g of salt per day. Beverages It is recommended to drink least 1–2 litres per day, preferably water or unsweetened beverages.

Some of the recommendations for a healthy diet (particularly a reduced consumption of meat and an increased consumption of plant-based sources of protein and fat) similarly apply to recommendations for a more sustainable diet (Tukker et al., 2011). Considering the large impact food production has on the environment and climate change (Aiking, 2011; Westhoek et al., 2014), this constitutes an additional reason that people should be motivated to meet these dietary guidelines.

18 GENERAL INTRODUCTION

1.2.2 Diet quality indices to evaluate overall diet quality

In nutritional epidemiological research, diet quality indices are frequently used as tools to evaluate populations’ dietary intake, combining evaluations of the consumption of specific food/beverage groups in one indicator of overall diet quality (Alkerwi, 2014; Lassale et al., 2016). There is no consensus regarding the exact definition of diet quality (Alkerwi, 2014). Rather, numerous indices have been developed and validated, including a different number of dietary components, and being based on different dietary guidelines and scoring systems (Lassale et al., 2016). For the calculation, some of these indices use predefined cut-off values to evaluate adherence to specific dietary guidelines (e.g., Healthy Eating Index, HEI-2015; Krebs-Smith et al., 2018; WHO Healthy Diet Indicator, WHO HDI, Berentzen et al., 2013), whereas others use sample-based cut-off values, which provide a relative evaluation of a diet or a combination of both (e.g., Mediterranean Diet Score, MDS; Trichopoulou, Costacou, Bamia, & Trichopoulos, 2003). Previous research indicates that these types of indices in general have good predictive power for morbidity and mortality risk (Lassale et al., 2016; Schwingshackl, Bogensberger, & Hoffmann, 2018; Wirt & Collins, 2009).

1.2.3 Nutrient profiling to determine the healthiness of single food products Recommendations for a healthy diet from dietary guidelines are usually made for whole groups of foods or beverages. Nutrient profiling is a different approach that classifies single foods and beverages based on their nutritional composition and that allows for a ranking of products regarding their healthiness (Rayner, 2017; WHO, 2017). There are several nutrient profiling systems in use, such as the UK Ofcom/FSA nutrient profile model (Food Standards Agency, 2011), the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014), or the WHO-Euro models (Rayner, 2017; WHO, 2015b). These systems are based on a different number of nutrients, and therefore the evaluation and ranking of the healthiness of specific products can vary depending on the system that is used (Poon et al., 2018). However, most of these models have not yet been fully validated (Cooper, Pelly, & Lowe, 2016; Poon et al., 2018). Nutrient profiling models have so far been used for different regulatory purposes. Originally, these systems were developed to decide upon the foods for which TV marketing activities directed to children should be restricted (Rayner, 2017). Furthermore, different nutrition labels are based on nutrient profiling models, such as the UK traffic light label, the French Nutri-Score and the Nordic keyhole health logo used in Sweden, Norway and Denmark (Food Standards Agency, 2011; Julia & Hercberg, 2017; Rayner, 2017; Roodenburg, 2017). In recent years, simple nutrient profile models are also increasingly considered a basis for the determination of taxes on unhealthy foods and beverages (Rayner, 2017).

19 GENERAL INTRODUCTION

1.3 Determinants of food choices

Making food choices might initially seem to be a rather simple process. However, according to the findings of numerous studies from different disciplines, it is assumed that food choices are the result of a complex interplay of multiple biological, psychological and environmental influential factors. There are a couple of models that attempt to summarise and categorise all these various influence factors on food choice. Figure 1.1 shows the multi-level model suggested by Story, Kaphingst, Robinson-O'Brien, and Glanz (2008). In their ecological framework, they distinguish four levels of determinants influencing food choices: individual factors and factors of the social, physical and macro-level environment (Figure 1.1).

Figure 1.1 An ecological framework of the determinants of food choices (Story et al., 2008).

20 GENERAL INTRODUCTION

Individual factors include biological, demographic and psychological determinants of eating behaviours as well as a person’s knowledge and skills (Story et al., 2008). The influence of these individual factors on food choices are in turn moderated by factors such as an individuals’ motivation, self-efficacy or outcome expectations. Some individual determinants potentially can be modified through interventions (e.g., knowledge, skills, attitudes), whereas others cannot be changed (e.g., gender, age, genes). Social environments (networks) refer to influences of family, friends, peers and other communities, such as through social support, rule modelling and social norms (Story et al., 2008). Physical environments (settings) include factors of the food environment in which a person lives, such as conditions at an individual’s home and work site or restaurants and supermarkets in the neighbourhood (Story et al., 2008). These different physical settings determine which foods are principally available and therefore may facilitate or hinder healthy eating. Macro-level environments include factors on a higher societal level, such as influences from food policies, culture or food marketing (Story et al., 2008). The following sections focus on selected individual and environmental influence factors on food choices that were investigated in the studies of this present doctoral thesis.

1.4 Individual determinants of food choices

1.4.1 Cooking skills Poor cooking skills are considered a barrier for the preparation of healthy home-cooked meals (Soliah, Walter, & Jones, 2011). It is assumed that this makes people more dependent on meals prepared away from home and ready-meals, which frequently are less healthy regarding their nutritional composition and calorie content (Kanzler, Manschein, Lammer, & Wagner, 2015; Lin, Guthrie, & Frazão, 1999; van der Horst, Brunner, & Siegrist, 2010). It is also argued that a lack of knowledge regarding how to prepare certain foods (e.g., vegetables or legumes) may make the implementation of recommendations from dietary guidelines more difficult (Caraher, Dixon, Lang, & Carr-Hill, 1999). In recent years, cooking skills have increasingly been the focus of nutritional studies, and many cooking interventions have been done to promote healthy eating and to prevent obesity (Hartmann, Dohle, & Siegrist, 2013; Hollywood et al., 2018; Reicks, Kocher, & Reeder, 2018). To design effective public health interventions, it is important to know more about the determinants of cooking skills, how these skills are acquired and if there are possibly underutilised sources of learning. Several studies consistently identified mothers as the primary source from whom people learn to cook (Caraher et al., 1999; Lavelle et al., 2016; Wolfson, Frattaroli, Bleich, Smith, & Teret, 2017). However, there are concerns that the transmission from cooking skills from parents to their offspring will be insufficient in the future and similarly the knowledge dissemination of these skills in schools (Jomori, Vasconcelos, Bernardo,

21 GENERAL INTRODUCTION

Uggioni, & Proença, 2018; Lavelle et al., 2016; Lyon et al., 2011). This could contribute to a further decline in the frequency and time spent for cooking meals at home with fresh ingredients as observed in different populations (McGowan et al., 2017; Möser, 2010). It is assumed, that people spend increasingly less time on home cooking due to increased perceived time pressure (Jabs & Devine, 2006) and thus also spend less time for cooking together with their children. In Chapter 2 of this thesis, whether there were changes in the time people were involved in cooking activities with their parents over the last generations was investigated. The influence of an early involvement in cooking activities during childhood/adolescence and the importance of various sources for the acquisition of cooking skills were also investigated. Furthermore, associations of adults’ cooking skills with their food intake, dietary quality, and eating habits were further investigated.

1.4.2 Motives for meat avoidance

There are many factors other than health considerations that influence consumers’ food choices, such as sensory appeal, convenience, price and ethical concerns (Steptoe & Pollard, 1995). Wanting to change dietary behaviours in a healthier direction makes it necessary to understand the main motives behind specific food choices. Meat is probably one of the most controversial food categories, with many different motives underlying both passionate meat consumption and convinced meat avoidance. A moderate consumption of meat, especially of red and processed meat, is considered an important aspect of a healthy diet (Swiss Society for Nutrition, 2011). A reduction in meat consumption in favour of more plant-based diets would have positive effects on health, and at the same time, the environmental burden caused by meat production could be reduced (Aiking, 2011; Tukker et al., 2011; Westhoek et al., 2014). In the last decade or so, the number of people following a vegetarian or vegan lifestyle has increased, and there is also a trend towards more conscious meat consumption among certain omnivores (Derbyshire, 2017; Swissveg, 2017). However, at the same time, in many industrialised countries, the consumption of meat and processed meat products is much higher in a large portion of the population than recommended by official dietary guidelines (FAO, 2018; Swiss Society for Nutrition, 2011; World Cancer Research Fund International, 2018), and it is expected to further increase in the future (Henchion, McCarthy, Resconi, & Troy, 2014). In Switzerland, for example, the average meat consumption of adults was around 111 g per day in 2014/15 (most recent available consumption data), which is more than three times as high as the recommended quantity (Swiss Federal Food Safety and Veterinary Office, 2017; Swiss Society for Nutrition, 2011). Regarding the design of possible intervention measures aiming at promoting moderate meat consumption by the public, it is helpful to know the most important

22 GENERAL INTRODUCTION

reasons motivating consumers to follow more plant-based diets. In Chapter 3, consumer groups with different self-declared diet-styles related to meat (including vegetarians/vegans, pescatarians and low-meat consumers) were compared in relation to their motives to avoid meat. These groups were also compared with regular meat consumers regarding their consumption of meat, meat substitutes and other protein sources and regarding diet quality and weight status. Lastly, independently of the specific diet style, which motives among all meat avoiders best predict low meat consumption and may therefore be the most effective in promoting such diets was investigated.

1.4.3 Intuitive eating

There are innumerable diets available on the market promising weight loss, but many of them fail to show long-term success, frequently resulting in increased psychological distress and unhealthy eating behaviours (Schaefer & Magnuson, 2014). In response, intuitive eating was suggested as an alternative strategy to lose weight without the need to deliberately restrict food intake (Tribole & Resch, 1995). The concept can be understood as an eating style or philosophy including the following principles: (1) intuitive eating is eating in response to body signals of hunger and satiety (2) rather than due to emotional or situational reasons, and in contrast to dieting, there are no ‘forbidden’ foods or restrictions but rather (3) an unconditional permission to eat when hungry and to eat those foods one desires (Tribole & Resch, 2012; Tylka, 2006). Later on, a further dimension was added to the construct: (4) the Body-Food Choice congruence, or the tendency to choose foods according to one’s bodily needs, such as eating foods that make the body perform efficiently (Tylka & Kroon Van Diest, 2013). Principles of intuitive eating have increasingly been included in intervention programmes, and evaluations of these programmes showed some benefits for psychological well-being and reductions in dieting behaviour, restrained eating and eating disorders (Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). However, concerning the association of intuitive eating with BMI, the results are mixed (Mensinger, Calogero, Stranges, & Tylka, 2016; Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). Only few studies have examined the associations with food intake, and the results are inconclusive (Carbonneau et al., 2017; Mensinger et al., 2016). Moreover, existing studies on intuitive eating are mostly conducted with women only. In Chapter 4, whether intuitive eating is associated with healthier food choices and overall diet quality in males and females and whether intuitive eaters have a lower incidence to overeat were investigated. All four dimensions of intuitive eating were examined separately to evaluate the efficiency of each of these single principles in promoting healthy eating.

23 GENERAL INTRODUCTION

1.4.4 Hedonic hunger and self-control

In the literature on human eating behaviours, there is a distinction between two major eating motives: Physical (or homeostatic) hunger, which is the result of a current or anticipated state of energy deficit, and hedonic (or psychological) hunger, which refers to eating driven by pleasure and not only by the need for calories (Harris & Mattes, 2008; Heshmat, 2011; Lowe & Butryn, 2007). It is assumed that hedonically driven hunger, in contrast to physical hunger, is much more influenced by the food environment, i.e., the availability and palatability of the foods individuals are exposed to (Lowe & Butryn, 2007). Thus, hedonic hunger is more likely to occur and is appropriate only in individuals who live in food-abundant environments. It is assumed that individuals substantially differ in their susceptibility to respond to temptations of the food environment and in their mental occupation with food even when not currently eating (Lowe & Butryn, 2007). These individual differences in appetitive responsiveness might be an explanation that some individuals face more difficulties in resisting temptations of the food environment and thus eat beyond their bodily needs, while others have less difficulties. Lowe et al. (2009) developed a measure for the assessment of individual differences in appetitive responsiveness to the food environment, the Power of Food Scale (PFS). In their psychometric validation study, the PFS showed high reliability and significantly predicted related constructs, such as the tendency for external eating, emotional eating and disinhibition with regard to eating behaviours (Lowe et al., 2009). Some previous studies found that individuals with higher levels of hedonic hunger are more likely to experience cravings for high-calorie foods and choose unhealthy snack foods (van Dillen & Andrade, 2016). However, a recent review concluded that hedonic hunger is not necessarily related with higher intake of palatable foods and that there was not a consistent association between hedonic hunger and BMI in previous studies (Espel-Huynh, Muratore, & Lowe, 2018). There is some evidence from the literature indicating that hedonic hunger might only lead to overconsumption of palatable foods in individuals who also exhibit low levels of self-control or high levels of impulsivity, respectively (Appelhans et al., 2011; Espel-Huynh et al., 2018). In Chapter 5 of this thesis, the aim was to answer the question regarding whether self-control moderates the association of hedonic hunger with overeating frequency and snacking behaviour and therefore whether self-control may be a protective factor in individuals with a high susceptibility to temptations in the food environment.

24 GENERAL INTRODUCTION

1.5 Environmental determinants of food choices

1.5.1 Nutrition labels

The provision of nutrition information on pre-packaged foods and beverages, especially in the form of front-of-package (FOP) labels, is a strategy intended to increase healthy food choices of consumers. Nutrition labels are assumed to increase the likelihood of healthy eating in several ways, primarily through improving consumer knowledge and awareness about the healthiness of products, and they are also expected to motivate food producers to improve the nutritional composition of the products they offer (Roodenburg, 2017). In many countries, the provision of detailed nutrition information through the nutrition facts table on the back of the package is now mandatory (European Food Information Council, 2018). Because nutrition information in this form is difficult to understand for many consumers, a variety of FOP nutrition labels has been proposed as well (Campos, Doxey, & Hammond, 2011; Roberto & Khandpur, 2014), and these labels are frequently implemented on a voluntary basis in many countries (Kanter, Vanderlee, & Vandevijvere, 2018). The available FOP labels vary in terms of complexity, the type and number of nutrients they focus on, used reference values, and whether they provide any interpretive guidance to the consumer (e.g., by using colour codes, verbal cues, shapes) (Kanter et al., 2018; Roodenburg, 2017). Many studies investigated consumer preferences, understanding and use of these labels (e.g., Borgmeier & Westenhoefer, 2009; Egnell, Talati, Hercberg, Pettigrew, & Julia, 2018; Hersey, Wohlgenant, Arsenault, Kosa, & Muth, 2013; Julia & Hercberg, 2017; Siegrist, Hartmann, & Lazzarini, in press). The multiple traffic light (MTL) label developed in the UK (Department of Health/Food Standards Agency, 2016), which uses a colour coding system to highlight the content of fat, saturated fat, sugar and salt, has been considered the most effective FOP nutrition label in terms of increasing consumers’ accuracy in evaluating the healthiness of pre-packaged foods (Borgmeier & Westenhoefer, 2009; Roberto et al., 2012). Recently, an additional label was developed in France, the Nutri-Score, and initial studies indicate even larger effects on consumers’ healthiness evaluation and on the healthiness of foods they added to their shopping carts in experiments conducted with virtual and real-world supermarkets (Crosetto, Lacroix, Muller, & Ruffieux, 2017; Julia & Hercberg, 2017). In Switzerland, the food company Danone recently began introducing the Nutri-Score on all its dairy products (Danone, 2019). Moreover, a nationwide mandatory implementation of the label is currently discussed (Public Health Schweiz, 2019). In Chapter 7, a randomised controlled online experiment is presented in which the Nutri-Score and the MTL nutrition label were compared with the nutrition facts table, and no information on the product packaging regarding their effect on consumers’ evaluation of the healthiness of salty snack products. An additional goal of this study was to investigate the effect of such a label when it is only present on some of the products

25 GENERAL INTRODUCTION

(incomplete labelling). This may be the case in real-world situations when the use of nutrition labels is voluntary.

1.5.2 Public health interventions and their acceptance in the population

There are several possible public health strategies to tackle the obesity problem, including mandatory nutrition labels on pre-packaged foods, taxes on unhealthy foods and beverages, regulations for food producers and portion size restrictions. These interventions are based on different approaches used to encourage health behaviour changes, including the information approach, nudging or incentivising/penalising (Guthrie, Mancino, & Lin, 2015; Michie, van Stralen, & West, 2011). In addition to the available evidence on the effectiveness of these strategies and cost- benefit considerations, public support for these interventions is a key criterion that is considered by policymakers prior to their implementation (Diepeveen, Ling, Suhrcke, Roland, & Marteau, 2013; Sekhon, Cartwright, & Francis, 2017). Public support, or the individual agreement regarding the implementation of intervention measures in the population (Hilbert, Rief, & Braehler, 2007), is important not only because governments do not want to act against the will of the public but also because a lack of support can negatively affect the successfulness of interventions, failing to lead to the intended behaviour change (Diepeveen et al., 2013; Stok et al., 2016). Previous research suggests that acceptance for public health interventions varies considerably depending on several aspects, including 1) the type of behaviour that is the focus of the intervention (e.g., more support for smoking prevention than prevention related to diet, PA or alcohol consumption); 2) the type of intervention (higher readiness for interventions intending behaviour change on the individual level, such as increasing consumer awareness about health issues through information provision, than regulations); 3) more support for interventions aimed at children and 4) for interventions already implemented and 5) interventions are also rather accepted when they are directed towards behaviours people do not engage in themselves, e.g., more support for smoking cessation interventions among non- smokers (Diepeveen et al., 2013; Hilbert et al., 2007). Currently, one of the most hotly discussed regulatory public health interventions is the implementation of excise taxes on sugar-sweetened beverages. So far, there is some first evidence from Mexico and the US after the introduction of an SSB tax suggesting that taxation may reduce sales for taxed beverages, while at the same time, it may increase sales for untaxed beverages, especially in households with a low socioeconomic status (Chaloupka, Powell, & Warner, 2019; Colchero, Molina, & Guerrero-Lopez, 2017). However, as mentioned, taxation of unhealthy products is among the most unpopular interventions (Diepeveen et al., 2013). Thus, the question arises regarding whether this intervention would similarly work in

26 GENERAL INTRODUCTION

high-income countries or in countries with different conditions. In Switzerland, there are currently no taxes on unhealthy foods or beverages. Similarly, there are no mandatory FOP labels implemented on pre-packaged products so far (see also section 1.5.1). In Chapter 6 of this thesis, current public support of various public health measures aimed at reducing sugar intake among consumers in Switzerland was investigated. Moreover, determinants of general acceptance of such interventions were evaluated, including socio-demographic characteristics, diet-related and health-related risk factors (high intake level of sugar and SSBs, overweight, low health consciousness).

1.6 The Swiss Food Panel 2.0

Five of the six studies presented in this thesis are based on data from the first and second waves of the Swiss Food Panel 2.0 study, which was launched in 2017. The main purpose of this longitudinal study was to investigate determinants of (un)healthy eating behaviours in the adult Swiss population of 20 years old and above. Furthermore, from the findings, recommendations for health promotion and interventions have been derived. The study consisted of a paper and pencil questionnaire, which is annually sent to the participants’ households. The study participants did not receive a monetary incentive for participation in the study. Instead, an information flyer with results for selected topics from the latest survey was sent to all participants some months after each round of data collection. The Swiss Food Panel 2.0 follows a similar preceding study, the Swiss Food Panel, conducted by the group from 2010 to 2014 (Hartmann, Dohle, & Siegrist, 2014), includes new constructs and research questions and a modified food frequency questionnaire for the dietary assessment.

1.6.1 Sample recruitment and exclusion criteria Most of the study participants were randomly selected from the phonebook. Additional addresses from younger people aged 20–30 years were purchased from an address company because many younger people are not registered in the phonebook. Residents living in the German and the French areas of Switzerland were included. Figure 1.2 shows the survey flow of the first two waves of the Food Panel 2.0 with the number of persons contacted, those who responded, those who were excluded and the final sample size for each round of data collection. In the first wave in 2017, the response rate was about 25%, which is typical for surveys (Budowski & Scherpenzeel, 2005; Lipps, 2007) and dietary surveys in Switzerland for which people receive no incentive for participation (Hartmann et al., 2013). In 2018, the response rate was considerably higher (73.5%) because the questionnaire was only sent to those who had participated in the first wave already. Participants with missing gender and age,

27 GENERAL INTRODUCTION

and those who completed less than 50% of the questionnaire were excluded. In the longitudinal sample, additional participants were excluded when there was a mismatch between reported gender, year of birth and/or body height (tolerance range of +/- 5 cm) between the first and the second wave (indicating that a different person filled in the questionnaire). Between the two waves, n = 70 participants dropped out because they died, were not willing or able to participate again or because of missing address details.

Baseline survey 2017: Contacted N = 23,002

Participated N = 5781 (RR* 25.1%) Excluded Final sample 2017: N = 5586 n = 195

Second survey 2018: Contacted N = 5,516

Participated N = 4056 (RR* 73.5%) Excluded Final sample 2018: N = 3681 n = 375

Further annual data collections planned until 2021

Figure 1.2 Flow chart of the Swiss Food Panel 2.0 study sample.

*RR = Response rate.

1.6.2 Sample characteristics and representativeness Table 1.3 shows the socio-demographic characteristics of the final study samples of 2017 and 2018. Roughly the same number of males and females took part in the panel. Compared to the general Swiss population, the proportion of young adults aged between 20–39 years in the sample of 2017 (17.0%) and 2018 (12.9%) was lower (census: 33.4%; percentage relative to the adult population from 20 years of age, Swiss Federal Statistical Office, 2017). Concerning the educational level, the proportion of highly educated people in the study sample (Table 1.3) was higher compared to the census (43.7%; Swiss Federal Statistical Office, 2019).

28 GENERAL INTRODUCTION

Table 1.3 Socio-demographic characteristics of the Food Panel 2.0 study sample.

2017 2018 Sample size [N] 5,586 3,681 Gender Females (%) 51.9 53.2 Males (%) 48.1 46.8 Mean Age (SD) [years] 56.5 (17.1) 58.6 (16.1) Age groups [years] 20–39 (%) 17.0 12.9 40–64 (%) 47.2 47.5 65–79 (%) 28.2 31.5 80+ (%) 7.6 8.1

Educational level Low1 (%) 7.5 6.6 Middle2 (%) 38.4 37.7

High3 (%) 52.6 55.0 Not indicated (%) 1.4 0.3

Region

German-speaking part (%) 71.9 73.7

French-speaking part (%) 28.1 26.3 1Low = compulsory school or no graduation; 2Middle = vocational or middle school; 3High = higher secondary school, college or university.

The proportion of overweight and obese males and females in the study sample is shown in Table 1.4. Compared to the general Swiss population in 2017 (Swiss Federal Statistical Office, 2018b), the percentage of overweight and obese women was slightly lower, and the percentage of overweight men was slightly higher in the study sample (see Table 1.4).

Table 1.4 Overweight and obesity in males and females of the Swiss Food Panel 2.0 compared to the general Swiss population.

2017 2018 Census 2017 Males Overweight (%) 42.1 40.8 38.7 Obese (%) 12.8 12.4 12.3 Total (%) 54.9 53.2 51.0 Females Overweight (%) 19.6 19.2 22.8 Obese (%) 8.8 8.6 10.2 Total (%) 28.4 27.8 33.0

29 GENERAL INTRODUCTION

1.7 Research questions and overview of the thesis

The central aim of the present doctoral thesis was to investigate individual and environmental factors that facilitate or hinder healthy eating in adults with the wider aim to gain new insights that can be used for health promotion and interventions. Furthermore, in one study, people’s acceptance of different types of public health interventions was explored in the case of measures aimed at reducing sugar consumption (see Chapter 6). Most studies presented in this thesis are based on cross-sectional data from the first and second waves of the Swiss Food Panel 2.0 (see Section 1.6; Chapters 2–6). In one additional online study conducted with a different sample of Swiss adults, an experimental design was used (see Chapter 7). Table 1.5 provides an overview of the chapters, the studies and the central research questions included in this thesis. The present thesis concludes with a general discussion and outlook on the studied topics (Chapter 8).

Table 1.5 Overview of the chapters, topics and research questions included in the thesis.

Chapter Topic Research question(s) 1 General Introduction – 2 Acquisition of cooking skills and their • Which are the most important sources of learning to cook for males and females? importance for healthy eating • Does stronger involvement in cooking activities at a younger age predict cooking skills in adulthood? • Are cooking skills associated with healthier food choices? 3 Motives for meat avoidance and • How do individuals with different self-declared diet styles (vegetarians/vegans, reduced meat intake pescatarians, low and regular meat consumers) differ in terms of their motives to avoid meat, protein consumption, diet quality, and weight status? • Do socio-economic factors, motives and consumption of meat alternatives predict meat consumption in individuals who report eating little or no meat? • How large is the proportion of individuals who meet the dietary recommendations for meat intake in groups with different self-declared diet styles? 4 Intuitive eating and food choices • Is intuitive eating associated with healthier food choices? 5 Self-control as a moderator of the effect • Does self-control moderate the effect of hedonic hunger on overeating and snacking? of hedonic hunger on overeating and snacking 6 Public acceptance of interventions • Which type of intervention aimed at reducing sugar intake is most/least accepted by aimed at reducing sugar intake the public? • Which factors predict acceptance of those interventions in general? 7 Nutrition labels and their effect on • Do the Nutri-Score and the MTL nutrition labels lead to greater accuracy in identifying healthiness perception of salty snack healthier snack options, compared to the nutrition facts table or no nutrition information food for consumers? • How does the effectiveness of the Nutri-Score label change when it appears on only some of the products? 8 General Discussion, –

GENERAL INTRODUCTION

References

Aiking, H. (2011). Future protein supply. Trends in Food Science & Technology, 22, 112-120. doi:10.1016/j.tifs.2010.04.005 Alkerwi, A. (2014). Diet quality concept. Nutrition, 30(6), 613-618. doi:10.1016/j.nut.2013.10.001 Appelhans, B. M., Woolf, K., Pagoto, S. L., Schneider, K. L., Whited, M. C., & Liebman, R. (2011). Inhibiting food reward: delay discounting, food reward sensitivity, and palatable food intake in overweight and obese women. Obesity (Silver Spring), 19(11), 2175- 2182. doi:10.1038/oby.2011.57 Azais-Braesco, V., Sluik, D., Maillot, M., Kok, F., & Moreno, L. A. (2017). A review of total & added sugar intakes and dietary sources in Europe. Nutrition Journal, 16(1), 6. doi:10.1186/s12937-016-0225-2 Bauman, A. E., Reis, R. S., Sallis, J. F., Wells, J. C., Loos, R. J. F., & Martin, B. W. (2012). Correlates of physical activity: why are some people physically active and others not? The Lancet, 380(9838), 258-271. doi:10.1016/s0140-6736(12)60735-1 Berentzen, N. E., Beulens, J. W., Hoevenaar-Blom, M. P., Kampman, E., Bueno-de-Mesquita, H. B., Romaguera-Bosch, D., . . . May, A. M. (2013). Adherence to the WHO's healthy diet indicator and overall cancer risk in the EPIC-NL cohort. PloS One, 8(8), e70535. doi:10.1371/journal.pone.0070535 Boeing, H., Bechtold, A., Bub, A., Ellinger, S., Haller, D., Kroke, A., . . . Watzl, B. (2012). Critical review: vegetables and fruit in the prevention of chronic diseases. European Journal of Nutrition, 51, 637-663. doi:10.1007/s00394-012-0380-y Borgmeier, I., & Westenhoefer, J. (2009). Impact of different food label formats on healthiness evaluation and food choice of consumers: a randomized-controlled study. BMC Public Health, 9, 184. doi:10.1186/1471-2458-9-184 Budowski, M., & Scherpenzeel, A. (2005). Encouraging and maintaining participation in household surveys: the case of the Swiss household panel. ZUMA-Nachrichten, 56(29), 10-36. Campos, S., Doxey, J., & Hammond, D. (2011). Nutrition labels on pre-packaged foods: a systematic review. Public Health Nutrition, 14(8), 1496-1506. doi:10.1017/S1368980010003290 Caraher, M., Dixon, P., Lang, T., & Carr-Hill, R. (1999). The state of cooking in England: the relationship of cooking skills to food choice. British Food Journal, 101(8), 590-609. doi:10.1108/00070709910288289 Carbonneau, E., Bégin, C., Lemieux, S., Mongeau, L., Paquette, M.-C., Turcotte, M., . . . Provencher, V. (2017). A Health at Every Size intervention improves intuitive eating and diet quality in Canadian women. Clinical Nutrition, 36, 747-754. doi:10.1016/j.clnu.2016.06.008 Chaloupka, F. J., Powell, L. M., & Warner, K. E. (2019). The use of excise taxes to reduce tobacco, alcohol, and sugary beverage consumption. Annual Review of Public Health, 40, 187-201. doi:10.1146/annurev-publhealth- Chu, D. T., Minh Nguyet, N. T., Nga, V. T., Thai Lien, N. V., Vo, D. D., Lien, N., . . . Pham, V. H. (2019). An update on obesity: Mental consequences and psychological interventions. Diabetes & Metabolic Syndrome, 13(1), 155-160. doi:10.1016/j.dsx.2018.07.015

32 GENERAL INTRODUCTION

Colchero, M. A., Molina, M., & Guerrero-Lopez, C. M. (2017). After Mexico Implemented a Tax, Purchases of Sugar-Sweetened Beverages Decreased and Water Increased: Difference by Place of Residence, Household Composition, and Income Level. Journal of Nutrition, 147(8), 1552-1557. doi:10.3945/jn.117.251892 Cooper, S. L., Pelly, F. E., & Lowe, J. B. (2016). Construct and criterion-related validation of nutrient profiling models: A systematic review of the literature. Appetite, 100, 26-40. doi:10.1016/j.appet.2016.02.001 Crosetto, P., Lacroix, A., Muller, L., & Ruffieux, B. (2017). Modification des achats alimentaires en réponse à cinq logos nutritionnels. Cahiers de Nutrition et de Diététique, 52(3), 129- 133. doi:10.1016/j.cnd.2017.04.002 Danone. (2019). Nutri-Score. Retrieved January 24 2019, from http://www.danone.ch/unsere- ueberzeugung/nutri-score/ Department of Health/Food Standards Agency. (2016). Guide to creating a front of pack (FoP) nutrition label for pre-packed products sold through retail outlets. Retrieved, from https://www.gov.uk/government/publications/front-of-pack-nutrition-labelling-guidance Derbyshire, E. J. (2017). Flexitarian Diets and Health: A Review of the Evidence-Based Literature. Front Nutr, 3, 55. doi:10.3389/fnut.2016.00055 Diepeveen, S., Ling, T., Suhrcke, M., Roland, M., & Marteau, T. M. (2013). Public acceptance of government intervention to change health-realated behaviours: a systemativ review and narrative synthesis. BMC Public Health(13). doi:10.1186/1471-2458-13-756 Duffey, K. J., & Popkin, B. M. (2011). Energy Density, Portion Size, and Eating Occasions: Contributions to Increased Energy Intake in the United States, 1977–2006. PLoS Medicine, 8(6). doi:10.1371/journal.pmed.1001050 Egnell, M., Talati, Z., Hercberg, S., Pettigrew, S., & Julia, C. (2018). Objective Understanding of Front-of-Package Nutrition Labels: An International Comparative Experimental Study across 12 Countries. Nutrients, 10(10). doi:10.3390/nu10101542 Espel-Huynh, H. M., Muratore, A. F., & Lowe, M. R. (2018). A narrative review of the construct of hedonic hunger and its measurement by the Power of Food Scale. Obes Sci Pract, 4(3), 238-249. doi:10.1002/osp4.161 European Food Information Council. (2018). Global update on nutrition labelling. Retrieved from https://www.eufic.org/images/uploads/healthy-living/Executive-Summary-GUNL- 2018-V2.pdf FAO. (2018). Food Balance Sheets. Retrieved July 2018), from http://www.fao.org/faostat/en/#data/FBS Food Standards Agency. (2011). Nutrient Profiling Technical Guidance. Retrieved January 2018, from https://www.gov.uk/government/publications/the-nutrient-profiling-model Guthrie, J., Mancino, L., & Lin, C.-T. J. (2015). Nudging Consumers toward Better Food Choices: Policy Approaches to Changing Food Consumption Behaviors. Psychology & Marketing, 32(5), 501-511. doi:10.1002/mar.20795 Harris, R. B. S., & Mattes, R. D. (2008). Appetite and Food Intake. Behavioral and Physiological Considerations. Boca Raton, Florida.: CRC Press. Hartmann, C., Dohle, S., & Siegrist, M. (2013). Importance of cooking skills for balanced food choices. Appetite, 65, 125-131. doi:10.1016/j.appet.2013.01.016 Hartmann, C., Dohle, S., & Siegrist, M. (2014). Time for change? Food choices in the transition to cohabitation and parenthood. Public Health Nutrition, 17(12), 2730-2739. doi:10.1017/S1368980013003297

33 GENERAL INTRODUCTION

Health Canada. (2014). The Development and Use of a Surveillance Tool: The Classification of Foods in the Canadian Nutrient File According to Eating Well with Canada’s Food Guide. Retrieved from http://publications.gc.ca/collections/collection_2014/sc- hc/H164-158-2-2014-eng.pdf Henchion, M., McCarthy, M., Resconi, V. C., & Troy, D. (2014). Meat consumption: trends and quality matters. Meat Sci, 98(3), 561-568. doi:10.1016/j.meatsci.2014.06.007 Hersey, J. C., Wohlgenant, K. C., Arsenault, J. E., Kosa, K. M., & Muth, M. K. (2013). Effects of front-of-package and shelf nutrition labeling systems on consumers. Nutrition Reviews, 71(1), 1-14. doi:10.1111/nure.12000 Heshmat, S. (2011). Eating Behavior and Obesity. Behavioral Economics Strategies for Health Professionals. New York: Springer Publishing Company. Hilbert, A., Rief, W., & Braehler, E. (2007). EVIDENCE BASED PUBLIC HEALTH POLICY AND PRACTICE: What determines public support of obesity prev. Journal of Epidemiology and Community Health, 61, 585-590. doi:10.1136/jech.2006.050906 Hollywood, L., Surgenor, D., Reicks, M., McGowan, L., Lavelle, F., Spence, M., . . . Dean, M. (2018). Critical review of behaviour change techniques applied in intervention studies to improve cooking skills and food skills among adults. Critical Reviews in Food Science and Nutrition, 58(17), 2882-2895. doi:10.1080/10408398.2017.1344613 Jabs, J., & Devine, C. M. (2006). Time scarcity and food choices: an overview. Appetite, 47(2), 196-204. doi:10.1016/j.appet.2006.02.014 Johnson, R. K., Appel, L. J., Brands, M., Howard, B. V., Lefevre, M., Lustig, R. H., . . . Prevention. (2009). Dietary sugars intake and cardiovascular health: a scientific statement from the American Heart Association. Circulation, 120(11), 1011-1020. doi:10.1161/CIRCULATIONAHA.109.192627 Jomori, M. M., Vasconcelos, F. d. A. G. d., Bernardo, G. L., Uggioni, P. L., & Proença, R. P. d. C. (2018). The concept of cooking skills: A review with contributions to the scientific debate. Revista de Nutrição, 31(1), 119-135. doi:10.1590/1678-98652018000100010 Julia, C., & Hercberg, S. (2017). Development of a new fron-of-pack nutrition label in France: the five-color Nutri-Score. Public Health Panorama, 3(4), 537-820. Kanter, R., Vanderlee, L., & Vandevijvere, S. (2018). Front-of-package nutrition labelling policy: global progress and future directions. Public Health Nutrition, 21(8), 1399-1408. doi:10.1017/S1368980018000010 Kanzler, S., Manschein, M., Lammer, G., & Wagner, K. H. (2015). The nutrient composition of European ready meals: protein, fat, total carbohydrates and energy. Food Chemistry, 172, 190-196. doi:10.1016/j.foodchem.2014.09.075 Krebs-Smith, S. M., Pannucci, T. E., Subar, A. F., Kirkpatrick, S. I., Lerman, J. L., Tooze, J. A., . . . Reedy, J. (2018). Update of the Healthy Eating Index: HEI-2015. Journal of the Academy of Nutrition and Dietetics, 118(9), 1591-1602. doi:10.1016/j.jand.2018.05.021 Lassale, C., Gunter, M. J., Romaguera, D., Peelen, L. M., Van der Schouw, Y. T., Beulens, J. W., . . . Tzoulaki, I. (2016). Diet Quality Scores and Prediction of All-Cause, Cardiovascular and Cancer Mortality in a Pan-European Cohort Study. PloS One, 11(7), e0159025. doi:10.1371/journal.pone.0159025 Lavelle, F., Spence, M., Hollywood, L., McGowan, L., Surgenor, D., McCloat, A., . . . Dean, M. (2016). Learning cooking skills at different ages: a cross-sectional study. The International Journal of Behavioral Nutrition and Physical Activity, 13(1), 119. doi:10.1186/s12966-016-0446-y

34 GENERAL INTRODUCTION

Lin, B.-H., Guthrie, J., & Frazão, E. (1999). Away-From-Home Foods Increasingly Important to Quality of American Diet. Washington, DC: United States Department of Agriculture. Economic Research Service, Agriculture Information Bulletin No. 749. Lipps, O. (2007). Attrition in the Swiss Household Panel. Methoden, Daten, Analysen, 1(1), 45-68. Lowe, M. R., & Butryn, M. L. (2007). Hedonic hunger: a new dimension of appetite? Physiology and Behavior, 91(4), 432-439. doi:10.1016/j.physbeh.2007.04.006 Lowe, M. R., Butryn, M. L., Didie, E. R., Annunziato, R. A., Thomas, J. G., Crerand, C. E., . . . Halford, J. (2009). The Power of Food Scale. A new measure of the psychological influence of the food environment. Appetite, 53(1), 114-118. doi:10.1016/j.appet.2009.05.016 Lyon, P., Mattsson Sydner, Y., Fjellström, C., Janhonen-Abruquah, H., Schröder, M., & Colquhoun, A. (2011). Continuity in the kitchen: how younger and older women compare in their food practices and use of cooking skills. International Journal of Consumer Studies, 35(5), 529-537. doi:10.1111/j.1470-6431.2011.01002.x Malik, V. S., Willett, W. C., & Hu, F. B. (2013). Global obesity: trends, risk factors and policy implications. Nature Reviews: Endocrinology, 9(1), 13-27. doi:10.1038/nrendo.2012.199 McGowan, L., Caraher, M., Raats, M., Lavelle, F., Hollywood, L., McDowell, D., . . . Dean, M. (2017). Domestic cooking and food skills: A review. Critical Reviews in Food Science and Nutrition, 57(11), 2412-2431. doi:10.1080/10408398.2015.1072495 Mensinger, J. L., Calogero, R. M., Stranges, S., & Tylka, T. L. (2016). A weight-neutral versus weight-loss approach for health promotion in women with high BMI: A randomized- controlled trial. Appetite, 105, 364-374. doi:10.1016/j.appet.2016.06.006 Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6. Möser, A. (2010). Food preparation patterns in German family households. An econometric approach with time budget data. Appetite, 55(1), 99-107. doi:10.1016/j.appet.2010.04.008 Mytton, O. T., Nnoaham, K., Eyles, H., Scarborough, P., & Mhurchu, C. N. Systematic review and meta-analysis of the effect of increased vegetable and fruit consumption on body weight and energy intake. BMC Public Health, 14. doi:10.1186/1471-2458-14-886 Ogden, J. (2010). The psychology of eating. From healthy to disordered behavior. Chichester, West Sussex, UK: Wiley-Blackwell. Poon, T., Labonte, M. E., Mulligan, C., Ahmed, M., Dickinson, K. M., & L'Abbe, M. R. (2018). Comparison of nutrient profiling models for assessing the nutritional quality of foods: a validation study. British Journal of Nutrition, 120(5), 567-582. doi:10.1017/S0007114518001575 Public Health Schweiz. (2019). Medienmitteilung zum Thema Nutri-Score [Media release on the Nutri-Score]. Retrieved May 24, 2019, from https://www.public-health.ch/de/news- de/medienmitteilung-zum-thema-nutri-score/ Puhl, R. M., & Heuer, C. A. (2010). Obesity Stigma: Important Considerations for Public Health. American Journal of Public Health, 100, 1019-1028. doi:10.2105/AJPH.2009.159491 Rayner, M. (2017). Nutrient profiling for regulatory purposes. Proceedings of the Nutrition Society, 76(3), 230-236. doi:10.1017/S0029665117000362

35 GENERAL INTRODUCTION

Reicks, M., Kocher, M., & Reeder, J. (2018). Impact of Cooking and Home Food Preparation Interventions Among Adults: A Systematic Review (2011-2016). Journal of Nutrition Education and Behavior, 50(2), 148-172 e141. doi:10.1016/j.jneb.2017.08.004 Roberto, C. A., Bragg, M. A., Seamans, M. J., Mechulan, R. L., Novak, N., & Brownell, K. D. (2012). Evaluation of consumer understanding of different front-of-package nutrition labels, 2010-2011. Preventing Chronic Disease, 9, E149. doi:10.5888/pcd9.120015 Roberto, C. A., & Khandpur, N. (2014). Improving the design of nutrition labels to promote healthier food choices and reasonable portion sizes. International Journal of Obesity (2005), 38 Suppl 1, S25-33. doi:10.1038/ijo.2014.86 Roodenburg, A. J. C. (2017). Nutrient profiling for front of pack labelling: how to align logical consumer choice with improvement of products? Proceedings of the Nutrition Society, 76(3), 247-254. doi:10.1017/S0029665117000337 Schaefer, J. T., & Magnuson, A. B. (2014). A Review of Interventions that Promote Eating by Internal Cues. Journal of the Academy of Nutrition and Dietetics, 114(5). doi:10.1016/j.jand.2013.12.024 Schneid Schuh, D., Campos Pellanda, L., Guessous, I., & Marques-Vidal, P. (2018). Trends and determinants of change in compliance to dietary guidelines in a Swiss community- dwelling sample. Preventive Medicine, 111, 198-203. doi:10.1016/j.ypmed.2018.03.008 Schwingshackl, L., Bogensberger, B., & Hoffmann, G. (2018). Diet Quality as Assessed by the Healthy Eating Index, Alternate Healthy Eating Index, Dietary Approaches to Stop Hypertension Score, and Health Outcomes: An Updated Systematic Review and Meta- Analysis of Cohort Studies. Journal of the Academy of Nutrition and Dietetics, 118(1), 74-100 e111. doi:10.1016/j.jand.2017.08.024 Sekhon, M., Cartwright, M., & Francis, J. J. (2017). Acceptability of healthcare interventions: an overview of reviews and development of a theoretical framework. BMC Health Services Research, 17(1), 88. doi:10.1186/s12913-017-2031-8 Siegrist, M., Hartmann, C., & Lazzarini, G. (in press). Healthy choice label does not substantially improve consumers’ ability to select healthier cereals: Results of an online experiment. British Journal of Nutrition, 1-24. doi:10.1017/S0007114519000448 Slavin, J. L., & Lloyd, B. (2012). Health benefits of fruits and vegetables. Advances in Nutrition, 3(4), 506-516. doi:10.3945/an.112.002154 Soliah, L. A. L., Walter, J. M., & Jones, S. A. (2011). Benefits and Barriers to Healthful Eating. American Journal of Lifestyle Medicine, 6(2), 152-158. doi:10.1177/1559827611426394 Steptoe, A., & Pollard, T. M. (1995). Development of a Measure of the Motives Underlying the Selection of Food: the Food Choice Questionnaire. Appetite, 25, 267-284. Stok, F. M., de Ridder, D. T., de Vet, E., Nureeva, L., Luszczynska, A., Wardle, J., . . . de Wit, J. B. (2016). Hungry for an intervention? Adolescents' ratings of acceptability of eating- related intervention strategies. BMC Public Health, 16, 5. doi:10.1186/s12889-015- 2665-6 Story, M., Kaphingst, K. M., Robinson-O'Brien, R., & Glanz, K. (2008). Creating healthy food and eating environments: policy and environmental approaches. Annual Review of Public Health, 29, 253-272. doi:10.1146/annurev.publhealth.29.020907.090926 Swiss Federal Food Safety and Veterinary Office. (2017). Fachinformation - Fleischkonsum [Technical information - meat consumption]. Retrieved July 15, 2019, from

36 GENERAL INTRODUCTION

https://www.blv.admin.ch/blv/de/home/lebensmittel-und- ernaehrung/ernaehrung/menuch/menu-ch-ergebnisse-ernaehrung.html Swiss Federal Statistical Office. (2017). Population. Retrieved May 4, 2019, from https://www.bfs.admin.ch/bfs/en/home/statistics/population/effectif- change/population.html Swiss Federal Statistical Office. (2018a). Körperliche Aktivität [Physical activity]. Retrieved May 22, 2019, from https://www.bfs.admin.ch/bfs/de/home/statistiken/gesundheit/determinanten/koerperli che-aktivitaet.html Swiss Federal Statistical Office. (2018b). Übergewicht [overweight]. Retrieved April 28, 2019, from https://www.bfs.admin.ch/bfs/de/home/statistiken/gesundheit/determinanten/ueberge wicht.html Swiss Federal Statistical Office. (2019). Bildungsstand [Educational level]. Retrieved May 3, 2019, from https://www.bfs.admin.ch/bfs/de/home/statistiken/bildung- wissenschaft/bildungsindikatoren/bildungssystem- schweiz/themen/wirkung/bildungsstand.html Swiss Society for Nutrition. (2011). Schweizer Lebensmittelpyramide [Swiss food pyramid]. Retrieved April 17, 2019, from http://www.sge-ssn.ch/ich-und-du/essen-und- trinken/ausgewogen/schweizer-lebensmittelpyramide Swissveg. (2017). Veggie survey. Retrieved July 2018), from https://www.swissveg.ch/veggie_survey?language=en Tribole, E., & Resch, E. (1995). Intuitive eating: A recovery book for the chronic dieter. New York: St. Martin’s Press. Tribole, E., & Resch, E. (2012). Intuitive eating. New York: St. Martin’s Press. Trichopoulou, A., Costacou, T., Bamia, C., & Trichopoulos, D. (2003). Adherence to a Mediterranean Diet and Survival in a Greek Population. The new england journal of medicine, 348(26), 2599-2608. Tukker, A., Goldbohm, R. A., de Koning, A., Verheijden, M., Kleijn, R., Wolf, O., . . . Rueda- Cantuche, J. M. (2011). Environmental impacts of changes to healthier diets in Europe. Ecological Economics, 70, 1776-1788. doi:10.1016/j.ecolecon.2011.05.001 Tylka, T. L. (2006). Development and psychometric evaluation of a measure of intuitive eating. Journal of Counseling Psychology, 53(2), 226-240. doi:10.1037/0022-0167.53.2.226 Tylka, T. L., & Kroon Van Diest, A. M. (2013). The intuitive eating scale–2: Item refinement and psychometric evaluation with college women and men. Journal of Counseling Psychology, 60(1), 137-153. van der Horst, K., Brunner, T. A., & Siegrist, M. (2010). Ready-meal consumption: associations with weight status and cooking skills. Public Health Nutrition, 14(2), 239-245. doi:10.1017/S1368980010002624 van Dillen, L. F., & Andrade, J. (2016). Derailing the streetcar named desire. Cognitive distractions reduce individual differences in cravings and unhealthy snacking in response to palatable food. Appetite, 96, 102-110. doi:10.1016/j.appet.2015.09.013 Van Dyke, N., & Drinkwater, E. J. (2014). Relationships between intuitive eating and health indicators: literature review. Public Health Nutrition, 17(8), 1757-1766. doi:10.1017/S1368980013002139 Westhoek, H., Lesschen, J. P., Rood, T., Wagner, S., De Marco, A., Murphy-Bokern, D., . . . Oenema, O. (2014). Food choices, health and environment: Effects of cutting Europe’s

37 GENERAL INTRODUCTION

meat and dairy intake. Global Environmental Change, 26, 196-205. doi:10.1016/j.gloenvcha.2014.02.004 WHO. (2003). Diet, nutrition and the prevention of chronic diseases. Report of the joint WHO/FAO expert consultation. WHO Technical Report Series 916 (TRS 916). Geneva: World Health Organization. WHO. (2008). Mean Body Mass Index (BMI). Retrieved May 9, 2019, from https://www.who.int/gho/ncd/risk_factors/bmi_text/en/ WHO. (2010). Global recommendations on physical activity for health. In. Retrieved from https://apps.who.int/iris/bitstream/handle/10665/44399/9789241599979_eng.pdf?seq uence=1 WHO. (2015a). Guideline: Sugars intake for adults and children. Geneva: World Health Organization. WHO. (2015b). WHO Nutrient Profile Model for the Western Pacific Region. A tool to protect children from food marketing. Retrieved May 14, 2019, from https://iris.wpro.who.int/bitstream/handle/10665.1/13525/9789290617853-eng.pdf WHO. (2017). Nutrient profiling. Retrieved May 14, 2019, from https://www.who.int/nutrition/topics/profiling/en/ WHO. (2018a). Global status report on alcohol and health 2018. Retrieved from https://apps.who.int/iris/bitstream/handle/10665/274603/9789241565639- eng.pdf?ua=1 WHO. (2018b). Obesity and overweight. Retrieved May 21, 2019, from https://www.who.int/en/news-room/fact-sheets/detail/obesity-and-overweight Willett, W. C., & Stampfer, M. J. (2013). Current evidence on healthy eating. Annual Review of Public Health, 34, 77-95. doi:10.1146/annurev-publhealth-031811-124646 Wirt, A., & Collins, C. E. (2009). Diet quality--what is it and does it matter? Public Health Nutrition, 12(12), 2473-2492. doi:10.1017/S136898000900531X Wolfson, J. A., Frattaroli, S., Bleich, S. N., Smith, K. C., & Teret, S. P. (2017). Perspectives on learning to cook and public support for cooking education policies in the United States: A mixed methods study. Appetite, 108, 226-237. doi:10.1016/j.appet.2016.10.004 World Cancer Research Fund International. (2018). Recommendations and public health and policy implications. Retrieved November 2018), from https://www.wcrf.org/sites/default/files/Cancer-Prevention-Recommendations- 2018.pdf

Chapter 2

Acquisition of cooking skills and their importance for healthy eating 2 Acquisition of cooking skills and their importance for healthy eating

A study based on the Swiss Food Panel 2.0 Désirée Hagmann, Michael Siegrist, Christina Hartmann ETH Zurich

Manuscript submitted for publication Hagmann, D., Siegrist, M., Hartmann, C. Acquisition of cooking skills and associations with healthy eating in Swiss adults.

ACQUISITION OF COOKING SKILLS

Abstract

Objective: To evaluate the influence of early involvement in cooking activities, the importance of different sources for the acquisition of cooking skills, and the association of cooking skills with healthy eating. Design: Cross-sectional survey. Participants: A random sample of 3,659 Swiss adults (47% men; average age = 58.8 years). Main Outcome Measure: Self-perceived cooking skills. Analysis: Analyses of variance, correlations, t-tests, and multiple regressions. Results: Stronger involvement in cooking activities during childhood predicted better cooking skills in adulthood in both genders (P < .001). Females were more involved than males in most age groups (P < .001). Women learned most about cooking from their mothers, cooking courses, and self-study from different media. Men indicated their partners/spouses and mothers as the two most important sources. The study found associations between cooking skills and diet quality in men (r = .11, P < .001) and women (r = .12, P < .001). Conclusions and Implications: Learning to cook should be encouraged at all ages. In children and adolescents, frequent involvement in cooking and food preparation at home or in cooking classes or cooking interventions at school may promote developing these skills. Similarly, effective strategies are needed to increase cooking skills in adults.

40 ACQUISITION OF COOKING SKILLS

2.1 Introduction

Knowing how to prepare and cook food is an important life skill to encourage because better cooking skills and more frequent home cooking have repeatedly been associated with a healthier diet in the short- and long-term (Hartmann, Dohle, & Siegrist, 2013; McGowan et al., 2017; Mills et al., 2017; Utter, Larson, Laska, Winkler, & Neumark-Sztainer, 2018; Wolfson & Bleich, 2015). For example, in a study with Swiss consumers, people with better cooking skills reported consumption of more vegetables and fewer convenience foods (Hartmann et al., 2013) In another study, conducted in the US, a high frequency of cooking dinner was associated with lower consumption of energy, fat, and sugar in adults (Wolfson & Bleich, 2015). Further studies found that better cooking abilities were also linked with better emotional wellbeing and lower levels of depression in adolescents (Utter, Denny, Lucassen, & Dyson, 2016). Furthermore, a growing number of interventions focusing on improved cooking skills as a measure to promote healthy eating have shown a positive impact on food choices, some diet-related diseases, including obesity, and psychosocial outcomes (Farmer, Touchton- Leonard, & Ross, 2018; Hollywood et al., 2018; Overcash et al., 2018; Reicks, Kocher, & Reeder, 2018). Today, it is often assumed that a chronic lack of time is a significant factor contributing to a steady decrease in the time adults spend preparing meals at home (Jabs & Devine, 2006), and meals are consumed more frequently away from home (Lin, Guthrie, & Frazão, 1999) and outside the family (Jabs & Devine, 2006). Furthermore, the range of convenience foods, including ready-meals, available on the market has greatly increased, and more frequent ready-meal consumption has been associated with poorer cooking skills (van der Horst, Brunner, & Siegrist, 2010). Thus, the consumption of ready-meals and meals away from home is considered problematic. An earlier study concluded that meals Americans consume away from home on average contain more fat but fewer favorable nutrients, such as calcium, fiber, and iron, compared to self-prepared meals (Lin et al., 1999). Similarly, many ready-meals are rich in calories, fat, salt, and sugar and do not contain the recommended quantity of vegetables (Kanzler, Manschein, Lammer, & Wagner, 2015; Remnant & Adams, 2015; van der Horst et al., 2010). Beyond these changes in lifestyle, a lack of cooking skills constitutes a major barrier for frequent home cooking (Soliah, Walter, & Jones, 2011). As reported in previous studies, cooking skills seem to decline in some population subgroups (Hartmann et al., 2013) due to several possible reasons. For example, many countries do not have cooking lessons and home economics as compulsory subjects in school (Stitt, 1996). Numerous studies have also supposed that successful parent-offspring knowledge transfers of cooking skills are no longer

41 ACQUISITION OF COOKING SKILLS guaranteed, which might become even more problematic in the future, given the current populations’ decreased cooking skills (Lavelle et al., 2016; Lyon et al., 2011). To design effective interventions to improve cooking skills, it is important to know the determinants of cooking skills and particularly the means by which these skills are acquired. A few previous studies have specifically investigated the importance of various sources for the acquisition of cooking skills, with mothers consistently reported as the primary source (Martin Caraher, Dixon, Lang, & CarrHill, 1999; Lavelle et al., 2016; Wolfson, Frattaroli, Bleich, Smith, & Teret, 2017). Autodidactic learning by trial and error, cookbooks, and recipe websites were also found to be important sources used to develop cooking skills; however, formal cooking classes at school are now rarely reported (Wolfson et al., 2017). This demonstrates a change when compared to an earlier study from the 1990s, in which cooking classes were named as the second most important source (Martin Caraher et al., 1999). In terms of intervention, it is also important to know whether childhood involvement in cooking and food preparation activities positively contributes to the development of cooking skills. A previous study suggested that learning cooking skills earlier in life (i.e., childhood or adolescence) positively affects the development of cooking skills and other related skills, such as cooking creativity and food safety, and it found a positive effect on diet quality, particularly among those who learned to cook during childhood (Lavelle et al., 2016). An Australian study found that boys with greater involvement in food-related household activities (e.g., meal planning, food shopping, meal preparation, and cleaning after the meal) showed more favorable dietary patterns (Leech et al., 2014). However, the study found no longitudinal effect nor did these activities have an effect on girls’ dietary patterns. The present study had three main objectives. The first objective was to investigate whether stronger involvement in cooking activities during childhood/adolescence was associated with better cooking skills in adulthood. The second objective was to evaluate the importance of various information sources for the acquisition of cooking skills. The third objective was to explore the link between cooking skills and healthy eating behavior. The present study used a large random sample, which included an equal number of men and women, different age groups, and diverse socio-economic subgroups. Furthermore, compared to some previous research (Lavelle et al., 2016), the present study placed a stronger emphasis on potential gender differences.

42 ACQUISITION OF COOKING SKILLS

2.2 Methods

2.2.1 Participants and procedure The study was based on cross-sectional data from the second survey of the Swiss Food Panel 2.0 longitudinal study, which aimed to examine different aspects of eating behavior in the adult Swiss population to identify the psychological determinants of (un)healthy eating behaviors. Most of the participants (91.7%, n = 3,356) were recruited by random selection from the phonebook. An additional number of participants, aged 20–30 years (8.3%, n = 303), were recruited through an address company to increase the percentage of younger participants, who are less often registered in the phonebook. In 2018, a mail survey was sent to all panel participants (n = 5,516) who had already completed the first survey in 2017, except for those who had died, moved away, or were unwilling or unable to participate again. Of those contacted, 4,056 participants returned the completed questionnaire (response rate of 73.5%). Several participants (n = 375) were excluded because they did not indicate their gender or age, because they filled in less than 50% of the questionnaire, or because there was a discrepancy between the reported gender, year of birth, and/or height (tolerance range of +/- 5 cm) from the first to the second survey, indicating that the respondents for the first and second surveys were not the same individuals. Women who reported pregnancy at the time of the survey (n = 22) were also excluded from the present analysis. The final sample used for the present study was 3,659 respondents. Table 2.1 provides an overview of the most important characteristics of the study sample.

Table 2.1 Description of the study sample (N = 3,659).

Mean (SD) or % Males 47.1% Age [years] 58.8 (16.0) Language region German-speaking 73.7% French-speaking 26.3% Educational level Low (compulsory school or no education) 6.0% Medium (vocational or middle school) 38.0% High (higher vocational education or university) 54.8% Missing information 1.1%

43 ACQUISITION OF COOKING SKILLS

Compared to the general Swiss population, the percentage of young adults aged between 20– 39 years was lower (12.5%; census: 33.4%; Swiss Federal Statistical Office, 2017) and the percentage of highly educated people was higher (55.0%; census: 43.7%; Swiss Federal Statistical Office, 2018). The Ethics Committee of ETH Zurich approved the study. The participants were informed about the purpose and conditions of the study. No explicit written consent form was obtained. The participants did not receive any monetary incentive for their participation.

2.2.2 Measures related to cooking

Cooking skills. The participants’ self-perceived cooking skills were assessed with seven items from a previously published scale (Hartmann et al., 2013): 1) “I consider my cooking skills as sufficient”; (2) “I am able to prepare a hot meal without a recipe”; (3) “I am able to prepare gratin”; (4) “I am able to prepare soup”; (5) “I am able to prepare sauce”; (6) “I am able to bake cake”; (7) “I am able to bake bread.” Each item was evaluated on a 6-point scale from 1 (“Does not apply at all”) to 6 (“Applies completely”). For the cooking skills variable, the mean of the seven items was calculated. The internal consistency of the scale was high (Cronbach’s α coefficient = .91 in both the present and the original reliability study sample; Hartmann et al., 2013). The items’ test-retest reliabilities were high as well (between r = 0.6 and r = 0.8; Hartmann et al., 2013). All cooking-related measures were included only cross-sectionally in the 2018 survey.

Sources of learning to cook. On a 6-point scale—from 1 (“Nothing”) to 6 (“Very much”) — the participants were asked how much of their cooking skills they learned from the following persons or information sources: mother; father; grandparents; partner/spouse; another family member; friends and acquaintances; cooking class at school; cooking courses; books, magazines, internet tutorials, videos, or apps; and cooking shows / TV cooks.

Involvement in cooking activities in childhood and adolescence. The participants were asked to retrospectively evaluate the frequency with which they had been involved in cooking activities at home during childhood and adolescence (one question): “daily” (coded as 7), “4– 6 times/week” (coded as 5), “1–3 times/week” (coded as 2), “1–3 times/month” (coded as 0.5), “several times/year” (coded as 0.1), and “never” (coded as 0).

44 ACQUISITION OF COOKING SKILLS

Responsibility for home cooking. The survey included two questions about responsibilities for home cooking during the week and the weekends: “Who (in your household) usually prepares the main meal during the week [on the weekends]?” For both questions, the response options were: “me”, “my partner”, “both” or “another person.”

2.2.3 Dietary behavior Self-reported food intake. A semiquantitative food frequency questionnaire (sFFQ) was used to assess the usual average consumption of 47 foods and beverages during the previous year. Similarly to the Nurses’ Health Study questionnaire (Hu et al., 2016), standard portion sizes for every food item were provided, e.g., “1 handful or 120 g of cooked vegetables (e.g., broccoli, carrots)” or “100–120 g of pork”, and the participtants were asked to indicate how many of these portions they usually consumed, with the following nine response options (recoded as weekly portions for the analysis): “4 or more/day” (coded as 28), “3/day” (coded as 21), “2/day” (coded as 14), “1/day” (coded as 7), “5–6/week” (coded as 5.5), “2–4/week” (coded as 3), “1/week” (coded as 1), “1–3/month” (coded as 0.5), and “seldom/never” (coded as 0). The present study focused on foods/beverages previously associated with negative health outcomes when consumed in excess or insufficient quantities (Lassale et al., 2016; Malik, Willett, & Hu, 2013). The study also focused on foods requiring some cooking or preparation before they can be consumed (e.g., potatoes, legumes, fish).

Eating habits. This study also assessed how often the participants eat a hot homemade meal, eat away from home (e.g., at a restaurant, take-away from a non-fast food restaurant), and eat something from a fast food restaurant (e.g., McDonald’s, Burger King, etc.). In each case, the response options were: “daily” (coded as 7), “4–6 times/week” (coded as 5), “1–3 times/week” (coded as 2), “1–3 times/month” (coded as 0.5) and “seldom/never” (coded as 0). Snacking frequency was assessed with four items (using the same response options as with the other eating habits): the participants reported how frequently they consume something between meals (i.e., a snack) in the (1) morning, (2) afternoon, (3) evening, and (4) independently of a specific time of day. These four items were totaled to get an overall weekly snacking frequency. These items’ test-retest reliabilities were between r = 0.6 and r = 0.8.

45 ACQUISITION OF COOKING SKILLS

Diet quality index. A diet quality index was calculated as an indicator of the overall healthiness of the participant’s diet. The index was based on self-reported consumption frequencies of five groups of foods/beverages that have been associated with an increased risk for modern lifestyle diseases when consumed in excessive or insufficient amounts (Lassale et al., 2016; Malik et al., 2013): fruit (excluding fruit juice); vegetables and salad (raw and cooked); whole- grain products (e.g. bread, rice, and pasta); meat and meat products; and sweets, salty snacks, sugar-sweetened beverages (SSBs), and alcohol. The diet quality index ranged from 0 (rather unhealthy diet) to 5 (rather healthy diet). A previous study described the index calculation in greater detail (Hagmann, Siegrist, & Hartmann, 2019).

2.2.4 Data analysis

The statistical analysis was conducted with IBM SPSS Statistics version 25, IBM, Armonk, NY, 2017. Pearson correlations, t-tests, Chi-square tests, and two-way ANOVAs were used to investigate gender and age differences in cooking skills, involvement in cooking activities during childhood/adolescence, used sources to learn cooking, and responsibilities for home cooking. Pearson correlations were calculated separately for gender to analyze the associations of cooking skills with diet quality, food intake, and eating habits. Multiple regression analyses were conducted to investigate whether involvement in cooking activities in childhood/adolescence and the amount of cooking skills learned from different sources predict cooking skills in adulthood (dependent variable). These analyses were run separately for gender. Normality testing was done for all continuous variables by visual inspection of histograms and Q-Q plots and by conducting Kolmogorov-Smirnov tests. None of the variables were normally distributed. Nevertheless, parametric tests were used because with large sample sizes, the violation of the normality assumption can be ignored (Ghasemi & Zahediasl, 2012). Given the large sample size, the significance level was set at 1% for all analyses.

46 ACQUISITION OF COOKING SKILLS

2.3 Results

In the present sample, women indicated being mainly responsible for meal preparation during the week (75.1%) and the weekends (64.3%) more often than men (26.7% during the week; c2 (1) = 840.74, P < .001; 28.9% during the weekends; c2 (1) = 448.85, P < .001). Women self- reported better cooking skills (M = 5.38, SD = 0.77, range = 1–6) than men (M = 4.05, SD = 1.45, range = 1–6), t(3,588) = –34.72, P < .001 (Figure 2.1). For men, cooking skills were negatively correlated with age (r = –.23, P < .001), whereas a small positive correlation was found with age for women (r = .06, P = .006).

6 s 5 ill sk

4 ing k oo

C 3

Men Women n = 1691 n = 1899

Figure 2.1 Boxplots of cooking skills among men and women. The line dividing the box indicates the median. The lower and the upper tilts of the box represent the 25th and the 75th percentile, respectively. The two lines outside the box (whiskers) extend to the highest and lowest observations that are not outliers.

Cooking skills and diet quality were positively correlated in both genders (Table 2.2). With single food groups, the strongest correlations were observed for total vegetable consumption and, separately, for cooked and raw vegetables/salad consumption. Small negative correlations with cooking skills were observed with the reported intake frequencies of unhealthy food groups in women: SSBs, sweet and salty snacks, and sausages/cold cuts (Table 2.2). Men with better cooking skills consumed more fish and seafood. Concerning the assessed eating habits, in women, cooking skills were negatively correlated with the frequency of fast food consumption and away-from-home eating and positively correlated with hot homemade meal consumption (Table 2.2).

47 ACQUISITION OF COOKING SKILLS

Table 2.2 Associations of cooking skills with food intake, diet quality, and eating habits in male and female participants of the Swiss Food Panel survey 2018 (N = 3,659).

Variables Men Women (n = 1722)a (n = 1937)a r P r P Diet quality [index based on sFFQ] .11 < .001 .12 < .001 Food intake [portions/week] Vegetables (total) .12 < .001 .16 < .001 Salad (incl. raw vegetables) .08 .001 .16 < .001 Cooked vegetables .13 < .001 .12 < .001 Fruit .05 .06 .06 .02 Legumes –.00 .91 .01 .63 Whole-grain products .05 .06 .05 .05 Cooked potatoes –.02 .55 .03 .23 Pork –.03 .22 –.03 .24 Beef/veal –.00 .96 .00 .94 Poultry .05 .06 –.05 .02 Sausages and cold cuts –.03 .17 –.07 .002 Fish and sea food .10 < .001 –.04 .06 Dairy products –.05 .05 .01 .83 Sweet and salty snacks –.03 .23 –.10 < .001 SSBs –.06 .01 –.14 < .001 Eating habits [frequency/week] Hot homemade meal consumption .05 .06 .15 < .001 Away-from-home eating .08 .001 –.11 < .001 Fast food consumption .02 .56 –.17 < .001 Snacking frequency .05 .04 –.05 .06

48 ACQUISITION OF COOKING SKILLS

Figure 2.2 shows the self-reported importance of various sources for the acquisition of cooking skills separately by gender. Women indicated a higher importance for most of these sources, except for their partners/spouses, which was a more important source of learning for men (P < .001). For women, mothers were the most important source, followed by cooking courses and self-learning from books, magazines, internet tutorials, videos, and apps. Men reported having learned most of their cooking skills from their mothers and partners/spouses.

Concerning your cooking skills, how much did you learn from the following persons or sources?

nothing learnt from the source (% men/women)

Mother 17% 4% ***

Father 60% 57% ***

Grandparents 64% 53% ***

Partner/spouse 14% 44% *** Men

Another family member 59% Women 50% ***

Friends and acquaintances 43% 22% *** 68% Cooking class at school 25% ***

Cooking courses, books, 30% *** magazines, internet, videos, apps 6% 62% Cooking shows, TV cooks 51% ***

1 2 3 4 5 6 nothing very much Figure 2.2 Gender differences in the importance of various sources for the acquisition of cooking skills. Means, standard deviations, 99% confidence intervals, and the percentages of men/women that indicated having learned nothing from a source are shown. All gender differences were significant (*** P < .001).

49 ACQUISITION OF COOKING SKILLS

Figure 2.3 shows the amount of time participants were involved in cooking activities during childhood/adolescence (retrospective estimation). The two-way ANOVA revealed a main effect for gender: F(1,3601) = 107.13, P < .001. Post hoc tests (multiple t-tests with Bonferoni correction) showed that, in most age groups (except age 30–39 years), women were involved more often in cooking activities at a younger age (for all P < .001). Involvement was also higher in younger than in older age groups (F(2,3601) = 105.24, P < .001).

Figure 2.3 Involvement in cooking activities during childhood/adolescence at different ages among men and women. Means and 99% confidence intervals are shown. Non-overlapping whiskers (the two lines outside the box that extend to the highest and lowest observations that are not outliers) indicate that the means are significantly different from each other.

50 ACQUISITION OF COOKING SKILLS

Table 2.3 shows the results of the multiple regression analyses predicting cooking skills, with the various sources of learning and the frequency of involvement in cooking activities during childhood/adolescence as predictors. The regression model was significant: F(11,1607) = 142.21, P < .001, and it explained 49% of the variance in cooking skills among men. For women, the model was significant as well: F(11,1796) = 20.03, P < .001, and it explained 11% of the variance in their cooking skills. Stronger involvement in cooking activities at a younger age predicted better cooking skills in adulthood in both genders (P < .001). The amount of cooking skills learned from cooking courses, books, the internet, etc. was the strongest predictor of better cooking skills for both genders (P < .001).

Table 2.3 Predictors of cooking skills among men and women in the Swiss Food Panel survey 2018 (N = 3,427).

Men (n = 1,619) Women (n = 1,808) B SE B 99% CI β P B SE B 99% CI β P Constant 1.37 .17 [.94, 1.81] < .001 3.94 .12 [3.63,4.26] < .001 Age [years] < –.01 < .01 [–.01,<.01] –.02 .32 .01 < .01 [<.01,.01] .13 < .001 Involvement in cooking .13 .02 [.08, .18] .13 < .001 .05 .01 [.02,.07] .12 < .001 activities during childhood/adolescence [days/week] Sources of learning to cook [Reported importance] Mother .23 .02 [.18, .28] .27 < .001 .07 .01 [.04, .10] .13 < .001 Father .09 .02 [.03, .14] .08 < .001 .02 .01 [–.01, .06] .05 .06 Grandparents .06 .02 [.01, .12] .06 .005 .03 .01 [<–.01,.06] .05 .02 Partner/spouse .09 .02 [.04, .13] .10 < .001 < .01 .01 [–.03, .03] < –.01 .97 Another family member -.03 .02 [–.09, .03] –.03 .18 .01 .01 [–.02, .04] .02 .39 Friends and acquaintances .13 .02 [.08, .19] .13 < .001 –.01 .01 [–.05, .02] –.02 .33 Cooking classes at school .03 .02 [-.03, .08] .02 .23 .02 .01 [–.01, .04] .03 .16 Cooking courses, books, .27 .02 [.23, .32] .33 < .001 .12 .01 [.09, .15] .22 < .001 internet, etc. Cooking shows .05 .02 [–.01, .11] .05 .03 < .01 .01 [–.03, .03] < .01 .91 R2 .49 .11

ACQUISITION OF COOKING SKILLS

2.4 Discussion

Cooking skills may facilitate the implementation of dietary guidelines in daily food consumption (Martin Caraher et al., 1999), and they decrease dependency on meals prepared by others or the industry. Therefore, the acquisition of cooking skills is important. One of the study’s main aims was to evaluate whether more frequent involvement in home cooking during childhood/adolescence was associated with better cooking skills in adulthood. The results support the assumption that both males and females profit from early involvement in food preparation for the development of their cooking skills, giving them more opportunities for observational learning and for becoming familiar with these activities. Another previous study similarly found that adults who had acquired most of their cooking skills as children or teenagers exhibited better cooking and food skills compared to adult learners (Lavelle et al., 2016). Women in most age groups reported stronger involvement in cooking activities at home than men did. This may indicate that a gender gap persists in the way children are introduced to cooking, and it may to some extent explain why men still exhibit poorer cooking skills later in life than women do. However, it is possible that children and adolescents spending more time with their parents in the kitchen reflects an already higher interest in, and enjoyment of, cooking and food preparation activities. Greater enjoyment of cooking, in turn, is one of the strongest predictors for cooking skills in both genders (Hartmann et al., 2013). Possibly, children who enjoyed this activity more may also be more likely to seek opportunities to cook more often with their parents. A generation effect regarding how frequently people were involved in cooking activities during their childhood/adolescence at home was observed. Younger participants between about 20 and 39 years, on average, reported more frequent involvement in these activities in their childhood compared to older participants. At first, this might be surprising, given the frequently reported general lack of time for cooking activities today (Jabs & Devine, 2006), which is expected to result in less time for parents to spend in such activities with their children. However, today’s parents in America and in European countries generally spend more time with their children compared to 50 years ago (Dotti Sani & Treas, 2016). There is much research supporting the view that the amount of parental time and interaction is important for children because they can benefit from the acquisition of skills in academic, linguistic, and other behavioral domains (Dotti Sani & Treas, 2016). A second aim of the study was to evaluate the different sources individuals use to develop their cooking skills. Mothers were mentioned as the primary source, which was consistent with previous research in other populations (Martin Caraher et al., 1999; Lavelle et al., 2016; Wolfson et al., 2017). For women, a second important source was cooking courses

53 ACQUISITION OF COOKING SKILLS and self-study from different media. Approximately 94% of the women (and 70% of the men) reported using books, magazines, the internet, apps, etc. to learn new cooking skills. Furthermore, learning from cooking courses and different media was the strongest predictor for cooking skills among the investigated predictors in both genders. Using these channels to improve knowledge and skills probably best reflects a voluntary interest in cooking. Digitalization and the increased availability of recipes and cooking tutorials on the internet might also contribute to the increased importance of this channel compared to the findings in earlier studies (Caraher et al., 1999). A study with Australian adults found that television, newspapers, and cooking magazines were among the most preferred and popular sources of learning to cook and that the majority of participants were interested in learning how to prepare ethnic dishes (e.g., Asian cuisine; Worsley, Wang, Ismail, & Ridley, 2014). This increased interest in ethnic dishes, however, may also make it necessary to use sources such as the internet and other media because traditional sources of learning may not necessarily provide the requisite knowledge to prepare these kinds of meals. Men reported that their spouses/partners were about as important as their mothers were when learning to cook. Furthermore, men seemed to learn more from their partners/spouses than the other way around. This might indicate that many men acquire their cooking skills later in life than women do. A study conducted in Ireland (Lavelle et al., 2016), comparing the impact of learning to cook at different ages on cooking skills and related skills, found a relatively low percentage of men had learned to cook during their childhood (19%) and a higher percentage had learned to cook in adulthood (53%). The study showed a link between later cooking acquisition and poorer cooking skills and practices in adulthood. Contrary to women, men also seemed to profit from friends and acquaintances in developing their cooking skills. As earlier studies suggested, men more often than women perceive cooking as a fun/leisure activity, particularly if they are not primarily responsible for domestic cooking (Szabo, 2013). Possibly, people who perceive cooking as a hobby rather than as an obligation may also be more likely to share this leisure activity with their friends, enabling them to learn new things. Similarly, men who had learned to cook from their fathers had better cooking skills. This indicates that having a male role model may positively affect the development of cooking skills among boys. Considerably more women reported having learned from cooking classes at school (75%) than men did (32%). This might be because in Switzerland, cooking classes were mandatory for girls long before they became available to boys in the 1980s (Hartmann et al., 2013). Cooking classes are no longer a compulsory subject in many countries’ school curricula, which is probably why cooking classes have become less important compared to other sources. Because of concerns about missing basic cooking skills and the possible consequences on diet and health, some countries, such as the UK (Department of Education,

54 ACQUISITION OF COOKING SKILLS

2019), have made cooking classes at school compulsory again. Furthermore, many school- based interventions and programs focusing on improved cooking skills have been developed (Caraher, Wu, & Seeley, 2010; Hersch, Perdue, Ambroz, & Boucher, 2014). However, many of these studies have qualitative issues and do not show the long-term effect of such interventions on food consumption (Caraher et al., 2010). On average, women in the present study rated their cooking skills higher than men did, which was also consistent with previous research (Hartmann et al., 2013; McGowan et al., 2016). This may be because women are still more likely to be responsible for cooking and food preparation in their family (Caraher et al., 1999; Flagg, Sen, Kilgore, & Locher, 2014; Mills et al., 2017), making them obliged to develop and practice their cooking skills. Likewise, in the study sample, 75% of the women (but only 27% of the men) reported being responsible for meal preparation during the week (similarly on weekends). The third aim of this study was to investigate the associations between cooking skills and dietary behavior, finding that cooking skills were linked with higher dietary quality, particularly with a higher consumption of healthier food groups (vegetables and fish). Women also had lower intake levels of unhealthy food groups (snacks, SSBs, and fast food) and more frequent consumption of hot homemade meals. These findings are in line with previous research that suggested a link between cooking and a healthy diet (Hartmann et al., 2013; McGowan et al., 2017; Wolfson & Bleich, 2015). A previous study with Swiss consumers reported largely similar associations between cooking skills and intake levels of specific food groups, and they also found a stronger link between cooking skills and healthy eating in women (Hartmann et al., 2013). This study has some limitations that must be addressed. The first limitation is related to the cooking-related measures. The frequency of involvement in cooking activities during childhood/adolescence was assessed subjectively and retrospectively; therefore, the values could be subjected to distortion, particularly in elderly people for whom this life period was farther in the past. Furthermore, involvement in cooking activities was assessed in an extremely broad way, not distinguishing between specific tasks (e.g., meal planning, chopping, and seasoning), and the complexity of the performed tasks might make a difference in the prediction of cooking skills. The questions used to assess involvement and the sources of learning to cook have been qualitatively pre-tested in our research group, consisting of psychologists and food scientists. However, similar sources were included as those in previous studies (Lavelle et al., 2016). Cooking skills were assessed with a validated measure (Hartmann et al., 2013); however, since it is based on a self-report, it may not perfectly reflect the participants’ real abilities. Finally, the study sample was not fully representative regarding age and educational level, which to a certain degree limits the generalizability of the results to the general Swiss population.

55 ACQUISITION OF COOKING SKILLS

Implications for research and practice The acquisition of cooking skills should be encouraged at all ages because it may contribute to healthier food consumption. The study suggests that the early involvement of children and adolescents in cooking activities and food preparation may promote the development of these skills. Since there might be reasons why this involvement cannot be perfomed in certain homes, complementary measures, such as mandatory cooking classes or interventions that give all children the opportunity for observational learning and for becoming familiar with these tasks, may be necessary. Furthermore, relatively few participants (particularly men) mentioned cooking classes at school as a learning source, suggesting that this may be a relatively underutilized learning source. There are currently no official published interventions in Switzerland that specifically target the promotion of cooking skills in adults. Further studies may be needed to evaluate the different reasons for the lack of cooking skills in adults, such as a lack of motivation, capabilities, or opportunities to acquire these skills. Strong evidence is still missing for how cooking skills interventions targeted to adults should best be designed (Reicks et al., 2018). Based on the study results, parent-child cooking interventions might be effective, encouraging the cooking skills in both adults and children and at the same time to promote the frequency of shared cooking and family meals at home. The motivation to cook could be increased with interventions targeting self-learning through the use of new media such as apps or e-mail interventions. In the present study, the scale that was used only measured basic cooking skills. Future studies should also investigate if knowledge about specific food preparation methods (e.g., healthy ways to prepare food, such as steaming and low-fat preparation) have even stronger links with healthy food consumption.

References

Caraher, M., Dixon, P., Lang, T., & Carr-Hill, R. (1999). The state of cooking in England: the relationship of cooking skills to food choice. British Food Journal, 101(8), 590-609. doi:10.1108/00070709910288289 Caraher, M., Wu, M., & Seeley, A. (2010). Should we teach cooking in schools? A systematic review of the literature of school-based cooking interventions. Journal of the Home Economics Institute of Australia, 17(1), 10-18. Department of Education. National curriculum in England: design and technology programmes of study. Cooking and nutrition (2019). Retrieved from https://www.gov.uk/government/publications/national-curriculum-in-england-design- and-technology-programmes-of-study/national-curriculum-in-england-design-and- technology-programmes-of-study#cooking-and-nutrition

56 ACQUISITION OF COOKING SKILLS

Dotti Sani, G. M., & Treas, J. (2016). Educational Gradients in Parents' Child-Care Time Across Countries, 1965-2012. Journal of Marriage and Family, 78(4), 1083-1096. doi:10.1111/jomf.12305 Farmer, N., Touchton-Leonard, K., & Ross, A. (2018). Psychosocial Benefits of Cooking Interventions: A Systematic Review. Health Education & Behavior, 45(2), 167-180. Flagg, L. A., Sen, B., Kilgore, M., & Locher, J. L. (2014). The influence of gender, age, education and household size on meal preparation and food shopping responsibilities. Public Health Nutrition, 17(9), 2061-2070. doi:10.1017/S1368980013002267 Ghasemi, A., & Zahediasl, S. (2012). Normality tests for statistical analysis: a guide for non- statisticians. Int J Endocrinol Metab, 10(2), 486-489. doi:10.5812/ijem.3505 Hagmann, D., Siegrist, M., & Hartmann, C. (2019). Meat avoidance: motives, alternative proteins and diet quality in a sample of Swiss consumers. Public Health Nutrition, 22(13), 2448-2459. doi:10.1017/S1368980019001277 Hartmann, C., Dohle, S., & Siegrist, M. (2013). Importance of cooking skills for balanced food choices. Appetite, 65, 125-131. doi:10.1016/j.appet.2013.01.016 Hersch, D., Perdue, L., Ambroz, T., & Boucher, J. L. (2014). The impact of cooking classes on food-related preferences, attitudes, and behaviors of school-aged children: a systematic review of the evidence, 2003-2014. Preventing Chronic Disease, 11, E193. doi:10.5888/pcd11.140267 Hollywood, L., Surgenor, D., Reicks, M., McGowan, L., Lavelle, F., Spence, M., . . . Dean, M. (2018). Critical review of behaviour change techniques applied in intervention studies to improve cooking skills and food skills among adults. Critical Reviews in Food Science and Nutrition, 58(17), 2882-2895. doi:10.1080/10408398.2017.1344613 Hu, F. B., Satija, A., Rimm, E. B., Spiegelman, D., Sampson, L., Rosner, B., . . . Willett, W. C. (2016). Diet Assessment Methods in the Nurses' Health Studies and Contribution to Evidence-Based Nutritional Policies and Guidelines. American Journal of Public Health, 106(9), 1567-1572. doi:10.2105/AJPH.2016.303348 Jabs, J., & Devine, C. M. (2006). Time scarcity and food choices: an overview. Appetite, 47(2), 196-204. doi:10.1016/j.appet.2006.02.014 Kanzler, S., Manschein, M., Lammer, G., & Wagner, K. H. (2015). The nutrient composition of European ready meals: protein, fat, total carbohydrates and energy. Food Chemistry, 172, 190-196. doi:10.1016/j.foodchem.2014.09.075 Lassale, C., Gunter, M. J., Romaguera, D., Peelen, L. M., Van der Schouw, Y. T., Beulens, J. W., . . . Huybrechts, I. (2016). Diet quality scores and prediction of all-cause, cardiovascular and cancer mortality in a pan-european cohort study. PLoS One, 11(7), e0159025. Lavelle, F., Spence, M., Hollywood, L., McGowan, L., Surgenor, D., McCloat, A., . . . Dean, M. (2016). Learning cooking skills at different ages: a cross-sectional study. The International Journal of Behavioral Nutrition and Physical Activity, 13(1), 119. doi:10.1186/s12966-016-0446-y Leech, R. M., McNaughton, S. A., Crawford, D. A., Campbell, K. J., Pearson, N., & Timperio, A. (2014). Family food involvement and frequency of family dinner meals among Australian children aged 10-12years. Cross-sectional and longitudinal associations with dietary patterns. Appetite, 75, 64-70. doi:10.1016/j.appet.2013.12.021 Lin, B.-H., Guthrie, J., & Frazão, E. (1999). Away-From-Home Foods Increasingly Important to Quality of American Diet. Washington, DC: United States Department of Agriculture. Economic Research Service, Agriculture Information Bulletin No. 749.

57 ACQUISITION OF COOKING SKILLS

Lyon, P., Mattsson Sydner, Y., Fjellström, C., Janhonen-Abruquah, H., Schröder, M., & Colquhoun, A. (2011). Continuity in the kitchen: how younger and older women compare in their food practices and use of cooking skills. International Journal of Consumer Studies, 35(5), 529-537. doi:10.1111/j.1470-6431.2011.01002.x Malik, V. S., Willett, W. C., & Hu, F. B. (2013). Global obesity: trends, risk factors and policy implications. Nature Reviews: Endocrinology, 9(1), 13-27. doi:10.1038/nrendo.2012.199 McGowan, L., Caraher, M., Raats, M., Lavelle, F., Hollywood, L., McDowell, D., . . . Dean, M. (2017). Domestic cooking and food skills: A review. Critical Reviews in Food Science and Nutrition, 57(11), 2412-2431. doi:10.1080/10408398.2015.1072495 McGowan, L., Pot, G. K., Stephen, A. M., Lavelle, F., Spence, M., Raats, M., . . . Dean, M. (2016). The influence of socio-demographic, psychological and knowledge-related variables alongside perceived cooking and food skills abilities in the prediction of diet quality in adults: a nationally representative cross-sectional study. The International Journal of Behavioral Nutrition and Physical Activity, 13(1), 111. doi:10.1186/s12966- 016-0440-4 Mills, S., White, M., Brown, H., Wrieden, W., Kwasnicka, D., Halligan, J., . . . Adams, J. (2017). Health and social determinants and outcomes of home cooking: A systematic review of observational studies. Appetite, 111, 116-134. doi:10.1016/j.appet.2016.12.022 Overcash, F., Ritter, A., Mann, T., Mykerezi, E., Redden, J., Rendahl, A., . . . Reicks, M. (2018). Impacts of a Vegetable Cooking Skills Program Among Low-Income Parents and Children. Journal of Nutrition Education and Behavior, 50(8), 795-802. doi:10.1016/j.jneb.2017.10.016 Reicks, M., Kocher, M., & Reeder, J. (2018). Impact of Cooking and Home Food Preparation Interventions Among Adults: A Systematic Review (2011-2016). Journal of Nutrition Education and Behavior, 50(2), 148-172 e141. doi:10.1016/j.jneb.2017.08.004 Remnant, J., & Adams, J. (2015). The nutritional content and cost of supermarket ready-meals. Cross-sectional analysis. Appetite, 92, 36-42. doi:10.1016/j.appet.2015.04.069 Soliah, L. A. L., Walter, J. M., & Jones, S. A. (2011). Benefits and Barriers to Healthful Eating. American Journal of Lifestyle Medicine, 6(2), 152-158. doi:10.1177/1559827611426394 Stitt, S. (1996). An international perspective on food and cooking skills in education. British Food Journal, 98(10), 27-34. doi:10.1108/00070709610153795 Swiss Federal Statistical Office. Bildungsstand [Educational attainment] (2018). Retrieved from https://www.bfs.admin.ch/bfs/de/home/statistiken/wirtschaftliche-soziale- situation-bevoelkerung/gleichstellung-frau-mann/bildung/bildungsstand.html Swiss Federal Statistical Office (2017). Population. Retrieved from https://www.bfs.admin.ch/bfs/en/home/statistics/population/effectif- change/population.html Szabo, M. (2013). Men nurturing through food: Challenging gender dichotomies around domestic cooking. Journal of Gender Studies, 23(1), 18-31. doi:10.1080/09589236.2012.711945 Utter, J., Denny, S., Lucassen, M., & Dyson, B. (2016). Adolescent Cooking Abilities and Behaviors: Associations With Nutrition and Emotional Well-Being. Journal of Nutrition Education and Behavior, 48(1), 35-41 e31. doi:10.1016/j.jneb.2015.08.016 Utter, J., Larson, N., Laska, M. N., Winkler, M., & Neumark-Sztainer, D. (2018). Self-Perceived Cooking Skills in Emerging Adulthood Predict Better Dietary Behaviors and Intake 10

58 ACQUISITION OF COOKING SKILLS

Years Later: A Longitudinal Study. Journal of Nutrition Education and Behavior, 50(5), 494-500. doi:10.1016/j.jneb.2018.01.021 van der Horst, K., Brunner, T. A., & Siegrist, M. (2010). Ready-meal consumption: associations with weight status and cooking skills. Public Health Nutrition, 14(2), 239-245. doi:10.1017/S1368980010002624 Wolfson, J. A., & Bleich, S. N. (2015). Is cooking at home associated with better diet quality or weight-loss intention? Public Health Nutrition, 18(8), 1397-1406. doi:10.1017/S1368980014001943 Wolfson, J. A., Frattaroli, S., Bleich, S. N., Smith, K. C., & Teret, S. P. (2017). Perspectives on learning to cook and public support for cooking education policies in the United States: A mixed methods study. Appetite, 108, 226-237. doi:10.1016/j.appet.2016.10.004 Worsley, A., Wang, W., Ismail, S., & Ridley, S. (2014). Consumers' interest in learning about cooking: the influence of age, gender and education. International Journal of Consumer Studies, 38(3), 258-264. doi:10.1111/ijcs.12089

Chapter 3

Motives for meat avoidance and reduced meat intake 3 Motives for meat avoidance and reduced meat intake

A study based on the Swiss Food Panel 2.0 Désirée Hagmann, Michael Siegrist, Christina Hartmann ETH Zurich

Published manuscript

Hagmann, D., Siegrist, M., Hartmann, C. (2019). Meat avoidance: Motives, alternative proteins and diet quality in a sample of Swiss consumers. Public Health Nutrition, 22(13), 2448-2459. doi: 10.1017/S1368980019001277

MEAT AVOIDANCE

Abstract

Objective: Diets lower in meat are considered both highly beneficial for human health and more environmentally friendly. The present study compared consumer groups with different self-declared diet styles regarding meat (vegetarians/vegans, pescatarians, low and regular meat consumers) in terms of their motives, protein consumption, diet quality, and weight status. Design: Cross-sectional data from the Swiss Food Panel 2.0 (survey 2017). Setting: Switzerland, Europe. Participants: Data of 4,213 Swiss adults (47.4% females) from a nationally representative sample living in the German- and French-speaking regions of Switzerland (mean age 55.4 years). Results: For vegetarians, vegans, and pescatarians, ethical concerns about animal welfare and environmental friendliness, as well as taste preferences are stronger reasons to avoid meat consumption. Female low-meat consumers are more likely to be motivated by weight regulation. Only 18% of the sample and 26% of self-declared low-meat consumers met the official dietary recommendations for meat intake. Concerns about animal welfare and taste preferences predicted lower meat intake, whereas perceived difficulty of practising a low-meat diet and weight loss motives were associated with higher meat consumption in consumers who reported eating little or no meat. Conclusions: Our study demonstrates that there can be large discrepancies between consumers’ self-perception and their actual meat consumption. This must be taken into account when designing public health interventions. Addressing ethical concerns about animal welfare (e.g., through awareness campaigns), further improving the range of vegetarian options, and increasing consumers’ knowledge about the dietary recommendations may be ways to promote diets lower in meat.

61 MEAT AVOIDANCE

3.1 Introduction

Meat production and consumption have a large impact on human health and the environment (Aiking, 2011; Burkholder, Rajaram, & Sabaté, 2016; Tukker et al., 2011; Westhoek et al., 2014). Evidence from epidemiological studies suggests that opting for more plant-based over meat-based diets would have a positive effect on human longevity and decrease the risk for several modern lifestyle diseases such as diabetes type 2, cardiovascular disease, obesity, and some types of cancer (Appleby, Thorogood, & Key, 1998; Battaglia Richi et al., 2015; Burkholder et al., 2016; Hever & Cronise, 2017). Additionally, the adoption of such diets would substantially reduce the environmental burden caused by the production of animal proteins (Aiking, 2011; Tukker et al., 2011; Westhoek et al., 2014) and the animal suffering caused by many common practises in livestock production (de Jonge, van der Lans, & van Trijp, 2015; Rothgerber, 2015). Increased awareness of the issues associated with meat consumption may be a cause of the increasing popularity of vegan, vegetarian, and flexitarian lifestyles (Derbyshire, 2016; Pollan, 2006; Swissveg, 2017). At the same time, however, global meat consumption is continuously increasing or, in some countries stagnating at a high level of per capita consumption (FAO, 2018; Henchion, McCarthy, Resconi, & Troy, 2014). Official dietary guidelines such as those of the Swiss Society for Nutrition (Swiss Society for Nutrition, 2017) or the World Cancer Research Fund International (World Cancer Research Fund International, 2018) strongly recommend moderate consumption of meat, especially of red and processed meat. These guidelines also encourage individuals to frequently replace meat with other sources of animal- and plant-based proteins. Moreover, public organisations are attempting to increase public awareness of the issues associated with high levels of meat consumption (Apostolidis & McLeay, 2016) in an effort to direct consumer food choices in a more healthy, sustainable direction. The transition to more plant-based diets, however, is challenging, (Apostolidis & McLeay, 2016) especially because meat is very popular and is positively regarded in terms of taste, tradition, and beliefs about healthiness by a large portion of society (Ruby, 2012). According to the OECD (2019), European consumers eat about 70 kg of meat (i.e., beef/veal, pork, poultry, and sheep; retail weight) per year and thus clearly exceed official recommendations for meat intake (Swiss Society for Nutrition, 2017; World Cancer Research Fund International, 2018). Because meat consumption can be considered a highly habituated behaviour (Rees et al., 2018), it is rather difficult for people to change their intake levels. It is thus important to understand what motivates people to consume less or no meat in order to develop strategies that can effectively motivate people who consume excessive amounts of meat to shift their habits towards more plant-based diets.

62 MEAT AVOIDANCE

Previous research has emphasised that the reasons for choosing a vegetarian or semivegetarian lifestyle are be diverse and vary widely cross cultures (Ruby, Heine, Kamble, Cheng, & Waddar, 2013). The most commonly reported reasons for vegetarianism and reduced meat consumption in Western countries are ethical concerns about the rearing and slaughtering of animals for the purpose of meat production, concerns about personal health, the environment, and disgust (Ruby, 2012). Other reasons to adopt a diet that excludes some or all types of meat are religious in nature (De Backer & Hudders, 2014), based on attempts to save money (Mullee et al., 2017) or a result of seeking for variety beyond the traditional meat-based menu composition (Forestell, Spaeth, & Kane, 2012). Previous research has suggested that vegetarians and nonvegetarians basically have similar motives but that some of them are mentioned more frequently in one or the other group (Ruby, 2012). For vegetarians, the most frequently reported motives are ethical concerns about animal welfare followed by personal health, environmental concerns and disgust towards meat (Ruby, 2012). In semivegetarians, however, health motives, especially the benefits of a vegetarian diet and the attempt to lose weight by lowering meat intake, are more common (De Backer & Hudders, 2014; de Boer, Schosler, & Aiking, 2017; Forestell et al., 2012) whereas ethical reasons and environmental concerns are less frequently reported in this diet group and thus seem to play a secondary role (Ruby, 2012). The motives for following a vegetarian diet may also change over time, because the longer a certain diet style is practised the more knowledge people gain about the diet and related issues (Ruby, 2012).

The aim of the present study is threefold. First, it explores differences in motives for low- or no-meat consumption among vegetarians/vegans, pescatarians, and self-declared low- meat consumers. Second, intake levels of both animal- based and plant-based proteins are investigated in order to determine whether and with which alternative protein sources low/no- meat consumers may compensate for meat avoidance. The study also compares diet quality and weight status between self-declared diet groups as indicators of the healthiness of more plant-based diets. Third, our study seeks to explain which factors predict variance in meat consumption among consumers who reported deliberately eating little or no meat. More specifically, it investigates which motives are linked to lower meat consumption and whether the consumption of plant-based alternatives is associated with lower meat intake. A large sample of Swiss adults from different sociodemographic groups was used to achieve a high level of generalisability. The results of this study are relevant for public health organisations’ efforts to support people keep their meat consumption within a healthy range.

63 MEAT AVOIDANCE

3.2 Methods

3.2.1 Participants The present study used data from the Swiss Food Panel 2.0, a longitudinal study of the eating behaviours of the Swiss population. The first wave of data was collected during the spring of 2017. The survey consisted of a paper-and-pencil questionnaire which was sent to a random sample of residents in the German- and French-speaking regions of Switzerland. Additional addresses from people aged between 20 and 30 years were purchased from an address company to ensure the presence of a sufficient proportion of younger adults in the sample. The questionnaires were completed and returned by 5,781 people, representing a response rate of 25.1%. Participants who did not indicate their gender or age and those who completed less than 50% of the questionnaire were excluded (n = 195). Women who were pregnant at the time (n = 348) were also excluded, because analyses including body mass index (BMI) were conducted. After these exclusions, 5,238 participants remained in the final sample. Of these, 48.7% were females and 73.8% were German speakers. The mean age of the sample was 56.5 years (SD = 17.3, range 20–100). Compared to the census (33.4%), young adults 20 to 39 years old were underrepresented (17%). For the present analysis, only participants with no missing answers to questions on the relevant variables (diet style and meat reduction) were considered (N = 4,213; 47.4% females; average age = 55.4 years).

3.2.2 Individual measures Motives for low or no meat consumption. First, participants were asked a filter question, ‘Do you intentionally eat little or no meat?’ Participants who answered ‘yes’ were then asked about the strength of a variety of motives for low or no meat consumption. Motives were assessed with 12 statements and a Likert-type response scale (e.g., ‘I eat little or no meat because I want to eat healthily’; see Table 3.1). Response options ranged from 1 (‘totally disagree’) to 7 (‘totally agree’).

Perceived difficulty of low or no meat consumption. The perceived difficulty of practising a diet containing little or no meat was assessed with one item: ‘How difficult is it for you to practise a diet with little or no meat?’. The response scale varied from 1 (‘not at all difficult’) to 7 (‘very difficult’).

Diet style. Participants were asked to indicate whether they defined themselves as omnivore (a typical Western diet including meat and other animal-based foods), pescatarian (no meat but fish and seafood), vegetarian (no meat, fish, or seafood but other animal-based foods such as dairy products or eggs), or vegan (no animal-based foods).

64 MEAT AVOIDANCE

Table 3.1 Items assessing motives for no or low-meat consumption.

Motive Cronbach’s Motive category alpha I eat little or no meat because… 1. …it helps me regulate my weight.† 2. …I want to eat healthily. 3. …I frequently don’t like the taste.† Taste .68 4. …I prefer the taste of vegetarian dishes.†

5. …it is better for the environment.† Environment .89 6. …I want to eat in an environmentally friendly way. al concerns …meat production has negative effects on animal 7. welfare. Animal .81 …I consider it unethical that animals are killed to welfare 8. produce meat. 9. …my doctor recommended it to me. 10 …my religion forbids me to eat (certain types of) . meat.† 11 …I was brought up that way. . 12 …my social environment expects it from me. . Note. † Item was adapted from De Backer and Hudders (2014)

Diet styles were predefined in the questionnaire according to the definitions presented in parentheses above. Based on that self-declaration and independent of their self-reported dietary behaviour as measured with a food frequency questionnaire (FFQ, see next section), participants were classified into the different diet styles. Of all the participants in the sample, 82.5% (n = 4,321) indicated their diet style. Of those participants, 93% (n = 4,019) were self- defined omnivores, 2.9% (n = 127) pescatarians, 3.6% (n = 156) vegetarians, and 0.4% (n = 19) vegans. Due to the small group size of vegans, vegans and vegetarians were collapsed into one group (n = 175). In addition, omnivores were split into two groups based on the filter question presented above (‘Do you intentionally eat little or no meat?’). Those who indicated that they deliberately ate little meat were labelled as (subjective) low-meat consumers (n = 1,296) and those who did not as (subjective) regular meat consumers (n = 2,615). All participants who failed to respond to any of the questions about diet style or meat consumption were excluded from the analyses, because they could not be clearly classified.

Self-reported food consumption. The survey included a semiquantitative food frequency questionnaire (sFFQ) which consisted of a subset of food items and nine response options adapted from the Nurses’ Health Study questionnaire (Hu et al., 2016). Typical consumption

65 MEAT AVOIDANCE frequencies of 47 types of food and beverage during the past year were assessed. As in the original questionnaire (Hu et al., 2016), for each item a standard portion was specified (e.g., “100–120 g of beef/veal” or “one handful or 120 g of fruit”), and the participants were asked to indicate the amount of standard portions they usually consumed. For the analysis, the number of portions per week was calculated by recoding the response options as follows: ‘4 or more per day’ (= 28 portions per week), ‘3 per day’ (= 21), ‘2 per day’ (= 14), ‘1 per day’ (= 7), ‘5–6 per week’ (= 5.5), ‘2–4 per week’ (= 3), ‘1 per week’ (= 1), ‘1–3 per month’ (= 0.5) and ‘seldom/never’ (= 0). Consumption frequencies of four types of unprocessed meat (including pork, beef/veal, poultry, and other types of meat such as venison and lamb) and two types of processed meat (including sausages and cold cuts such as salami or bacon) were assessed. The predefined portion of unprocessed meat was 100–120 g; for processed meat it was three slices for cold cuts and one piece for sausages. Total meat consumption was calculated by summing up the weekly portions of all six meat items. A plausibility check of the calculated total meat consumption revealed implausibly high total amounts of meat consumed per week for some participants. Therefore, an upper limit of 35 portions per week (five portions per day) was defined, and all values above this limit were corrected down and set to 35 portions per week. The sFFQ also asked for information regarding the consumption frequency of other animal-based proteins, including fish and seafood, eggs, and dairy products ([cow’s] milk, quark, cheese, yoghurt, and cottage cheese), as well as of plant-based proteins including meat substitutes such as tofu and vegetarian cold cuts, legumes, and soy products (soy milk and soy yoghurt). The pre-specified portion size was 100–120 g for fish and seafood; one egg; one glass or 2dl for cow and soy milk; 150–200 g or one bowl for quark, yogurt, and cottage cheese; 30–60 g for cheese; and 2 dl or one cup for (cooked) legumes.

Diet quality index. As an indicator of the healthiness of the participants’ diet a diet quality index was calculated based on the self-reported consumption frequencies of the following five food/beverage groups: fruit (excluding fruit juice); vegetables and salad (raw and cooked); whole-grain products (bread, rice, and pasta); meat and meat products; sweets, salty snacks, sugar-sweetened beverages, and alcohol. These foods and beverages were selected based on official dietary guidelines (Swiss Society for Nutrition, 2017) and because previous studies have shown that their consumption either positively or negatively affects health (Malik, Willett, & Hu, 2013). For each food group, the recommended minimum or maximum amount of weekly intake was used as the cutoff value (see Table 3.2). For foods with a positive impact on health (fruit, vegetables and salad, and whole-grain products), one point was assigned if a participant’s consumption was equal to or higher than the recommended minimum. For all other indicators, one point was assigned if consumption was equal to or below the recommended maximum. A summary score for these points was calculated to create the diet

66 MEAT AVOIDANCE quality index. Possible scores ranged from 0 (rather unhealthy diet) to 5 (rather healthy diet). Research has shown that diet quality indices based on dietary recommendations are valid measures with good predictive power for morbidity and all-cause mortality when combined with other risk factors such as age, gender, smoking, or BMI (Lassale et al., 2016; McNaughton, Ball, Crawford, & Mishra, 2008).

Weight status. BMI was calculated as the quotient of self-reported body weight (in kg) divided by height squared (in m2). Participants with a BMI ≥ 25 kg/m2 were categorised as overweight. In the present sample, the mean BMI of the participants was 25.9 kg/m2 (SD = 4.00) for males and 23.7 kg/m2 (SD = 4.42) for females. More males in the sample (54.9% versus 50% in the census) and fewer females (28.3% versus 32% in the census) were overweight than in the general Swiss population (Swiss Federal Statistical Office, 2012).

Sociodemographic characteristics. The questionnaire included questions about the participants’ gender, age, monthly net household income in Swiss francs, and education, among other sociodemographic variables. Six response options were used for household income, ranging from ‘less than 3,000 Swiss francs’ to ‘more than 11,000 Swiss Francs’.

3.2.3 Data analysis The statistical analyses were conducted using the IBM SPSS Statistics software. One-way analyses of variance (ANOVAs) using the Games-Howell post hoc test were conducted to compare the groups with the diverse diet styles. This post hoc test is adequate for comparing groups of very different group sizes and if the homogeneity of variances is not given (Field, 2013). For categorical variables, c2 tests were calculated. T-tests for independent samples were used to compare men and women with respect to several variables. A principal component analysis with Varimax rotation was used to analyse whether some of the assessed motives for low/no-meat consumption could be combined into one component. Finally, a three- step hierarchical regression analysis based on 1,000 bootstrapped samples was used to identify predictors of total meat consumption among those participants who reported deliberately eating little or no meat (bootstrapping was used to deal with outliers). Sociodemographic characteristics (step 1), motives (step 2), the perceived difficulty of practising a diet with little or no meat consumption, and the consumption of plant-based proteins (step 3) were included in the regression model as predictors. Due to the large sample size, only p-values below .01 were considered significant in all statistical tests.

Table 3.2 Foods and beverages included in the diet quality index.

Food group Official dietary recommendation of the Scoring Percentage of Swiss Society for Nutrition(Swiss Society participants in the for Nutrition, 2017) Swiss Food Panel 2.0 study meeting the recommendations

Fruits Minimum 2 per day ³ 14 portions/week = 1 32.6% (= 14 per week) < 14 portions/week = 0

Vegetables and salad Minimum 3 per day ³ 21 portions/week = 1 23.5% (= 21 week) < 21 portions/week = 0

Whole-grain products Should be preferred compared to refined ³ 1 portion/week = 1 82.8% grain products (no specific portion < 1 portion/week = 0 recommendation)

Meat and meat products Maximum 2–3 portions per week £ 3 portions/week = 1 18.0% > 3 portions/week = 0 Sweets, salty snacks, With moderation; either a small portion of £ 7 portions/week = 1 22.2% SSB, alcohol sweets or salty snacks or sugar-sweetened > 7 portions/week = 0 beverage, or alcohol per day (= 7 per week)

MEAT AVOIDANCE

3.3 Results

The proportion of females in the four diet styles differed substantially, c2(3) = 263.31, p < .001 (vegetarians/vegans = 76.6%, pescatarians = 71.7%, subjective low-meat consumers = 60.0% and regular meat consumers = 38.0%). Therefore, all group comparisons were conducted separately for males and females. The groups did not differ significantly in age with the exception of the vegetarians/vegans, who were significantly younger (M = 49.22 (SD = 17.87) years) than the pescatarians (M = 55.83 (SD = 16.34) years), the low-meat consumers (M = 56.28 (SD = 15.99) years), and the regular meat consumers (M = 55.30 (SD = 16.87) years).

3.3.1 Motives for consuming little or no meat Table 3.1 shows the items used to assess the participants’ motives for following a diet with low or no meat consumption. Based on the results of the principal component analysis (scree plot, eigenvalue > 1, content-based considerations), two items were combined in each case for the motive categories of taste, environmental concerns, and animal welfare. Table 3.1 shows the Cronbach’s alphas for those items which were combined into one motive category. Mean values for the motives separated by diet style (i.e., vegetarian/vegan, pescatarian, low-meat consumer) and gender are graphically displayed in Figure 3.1. The regular meat consumers were not included in this comparison, because it would have been nonsensical to ask them to indicate their motivation for meat avoidance. One-way independent ANOVAs revealed significant differences in the motivation for low- or no-meat consumption between the three diet styles. In particular, environmental and animal welfare concerns were highly endorsed by vegetarians/vegans and pescatarians, but less so by low-meat consumers. Taste-related motives were a stronger motivation for low/no meat consumption for vegetarians/vegans and pescatarians than for low-meat consumers. Health motives did not differ between the diet styles. Lastly, low-meat consumers (females only) were more strongly motivated by weight- related motives than were vegetarians/vegans and pescatarians. There were no differences between diet styles observed in the other motives assessed. In particular, no differences were observed for the motives related to medical advice, religious rules, upbringing, or social expectations. Subjective low-meat consumers considered it more difficult to maintain a low level of meat consumption compared to vegetarians/vegans but not compared to pescatarians (see Table 3.3). There was also a gender difference: males reported more difficulties than females with practising a low-meat or meatless diet.

69 MEAT AVOIDANCE

7 *** ** ab ns a a 6 ab *** a a 5 b ns b Males 4

3 b

2 Strength of the motive the of Strength 1 Animal Environmental Taste Healthy Weight welfare concerns eating regulation *** 7 a *** ns a a ab *** 6 a b a 5 b Females 4 b

3 *** b

2 a a Strength of the motive the of Strength 1 Animal Environmental Taste Healthy Weight welfare concerns eating regulation

Vegetarians/vegans Pescatarians Low meat consumers Males, n = 41 Males, n = 36 Males, n = 519 Females, n = 134 Females, n = 91 Females, n = 777

Figure 3.1 Motives for vegetarianism and low-meat consumption in different diet styles, separated by gender (n = 1,598). Means and 99%-confidence intervals are shown. One-way independent ANOVAs were used to investigate differences in motives between diet styles. Mean values with no common superscript letter are significantly different (according to the Games-Howell post hoc test, p £ .01). The two aspects of taste motives (i.e., not liking the taste of meat and preferring vegetarian foods) were analysed together here because the results for both aspects were the same. n per group vary due to missing values. ** p < .01, *** p £ .001.

Table 3.3 Perceived difficulty of low or no meat intake, diet quality, and weight of adult males and females separated by self-reported diet styles (N = 4,213).

Self-declared Self-declared Vegetarians Pescatarians low-meat regular meat & vegans consumers consumers M or M or M or SD SD SD M or % SD % % % F(df1,df2) or c2(df) Males n = 41 n = 36 n = 519 n = 1,620 Age 54.1 17.7 64.6 14.3 59.6 16.1 58.2 16.1 F(3,2212) = 3.67 Diet quality † 2.84a 1.10 2.32ab 1.16 1.73b 1.03 1.26c 0.85 F(3,2046) = 74.80*** % Meeting the dietary 79.2 - 54.8 - 19.5 - 3.6 - c2(3) = 462.86*** recommendations for meat intake‡ Perceived difficulty § 1.39a 0.79 2.19ab 1.73 2.47b 1.51 - - F(2,567) = 9.63*** BMI 25.1ab 4.1 25.9ab 4.3 24.9a 3.4 26.1b 3.9 F(3,2191) = 12.73*** % overweight (BMI 25–29.9) 30.0 - 20.0 - 36.6 - 44.0 - c2(3) = 17.84*** % obese (BMI ³ 30) 12.5 - 22.9 - 5.8 - 12.8 - c2(3) = 23.73*** Females n = 134 n = 91 n = 777 n = 995 Age 47.7a 17.7 52.4ab 15.8 54.1b 15.5 50.6c 17.0 F(3,1993) = 9.60*** F(3,1831) = Diet quality † 3.49a 1.11 3.22a 1.20 2.33b 1.12 1.77c 1.05 133.31*** % Meeting the dietary 96.9 - 78.7 - 30.8 - 7.0 - c2(3) = 728.03*** recommendations for meat intake‡ Perceived difficulty § 1.36a 0.96 1.73ab 1.39 2.02b 1.28 - - F(2,971) = 16.87*** BMI 21.7a 2.9 22.1a 3.3 23.4b 4.1 24.2c 4.7 F(3,1965) = 19.68*** % overweight (BMI 25–29.9) 10.8 - 14.4 - 17.8 - 20.9 - c2(3) = 10.09* % obese (BMI ³ 30) 2.3 - 2.2 - 6.9 - 10.9 - c2(3) = 20.76*** Note. One-way ANOVAs and c2 tests were used to investigate differences between the groups. Mean values within rows with no common superscript letter differ significantly (based on the Games-Howell post hoc test, p £ .01). † Diet quality index based on recommendations of the Swiss Society for Nutrition. ‡ Percentage of individuals per group who did not eat more than the maximum recommended amount of three portions of meat per week (Swiss Society for Nutrition). Based on self-reported meat consumption on the food frequency questionnaire. § Perceived difficulty of practising a diet with little or no meat. n per group varies due to missing values. ** p <. 01, *** p < .001.

MEAT AVOIDANCE

3.3.2 Dietary behaviour, diet quality, and weight status Table 3.4 (males) and Table 3.5 (females) show the mean consumption frequencies of different protein sources for each diet style. In males, consumption of processed, unprocessed, and thus total meat did not differ between vegetarians/vegans, pescatarians, and self-declared low- meat consumers. In females, however, the lowest consumption of processed, unprocessed, and total meat was in fact reported by vegetarians/vegans. Among all diet style groups, the highest consumption of fish was reported by male and female pescatarians. The weekly intake of eggs and dairy products did not differ between any of the groups in either males or females. Male vegetarians reported a considerably higher total meat intake (Table 3.4) than did vegetarian females (Table 3.5). With regard to meat substitutes, male and female vegetarians/vegans and pescatarians reported the highest consumption. Self-declared low-meat consumers also ate more of these foods than self-declared regular meat consumers. For the males, consumption frequency of other plant-based proteins did not differ among the groups. For females, the differences were more pronounced: Vegetarians and vegans consumed more of all kinds of plant proteins than did the two meat consumer groups (regular and reduced). For both genders, the regular meat consumers had the lowest diet quality of all groups, followed by the self-declared low-meat consumers (Table 3.3). Further, the percentage of overweight people (i.e., BMI 25–29.99), and in females the percentage of obese people (i.e., BMI ³ 30), was considerably higher among self-declared regular and low-meat consumers than in the other groups (Table 3.3). However, in males the prevalence of obesity was highest in the pescatarian group.

Adherence to dietary recommendations for meat intake A relatively high percentage of individuals (82.0%) exceeded the weekly recommended maximum of three portions of meat and meat products. This percentage was higher among males (90.3%) than females (73.1%). Only 4.5% of the sample indicated never eating meat or meat products in the sFFQ, and 55.7% reported eating one portion of meat or even more per day. The percentage of participants in the sample that met the official dietary recommendation for meat intake separated by diet style is shown in Table 3.3. Among self-declared low-meat consumers, the percentage of participants whose meat consumption fell within the recommended amount was rather low (19.5% in men, 30.8% in women).

Table 3.4 Consumption of animal- and plant-based proteins (weekly portions) by adult males (n = 2,216).

Diet style Self- Self-declared Vegetarians & declared low-meat vegans regular meat consumers (n = 41) Pescatarians consumers (n = 519) (n = 36) (n = 1,620) Protein source M SD M SD M SD M SD F(df1,df2) Animal-based Unprocessed meat† 2.02a 5.14 3.39a 4.56 3.99a 2.78 6.37b 5.35 F(3,2186) = 23.58*** Processed meat (sausages, cold cuts) 1.56a 4.11 2.94abc 6.56 2.45ab 2.58 4.25c 4.58 F(3,2190) = 27.71***

a a a 10.23 Total meat‡ 3.24 6.79 4.69 5.03 6.42 4.17 b 5.81 F(3,2175) = 85.48*** Fish and seafood 0.45a 1.20 1.58b 1.60 0.89a 1.02 0.89a 1.32 F(3,2199) = 4.96** Eggs 2.07 2.11 2.08 2.84 2.00 2.22 2.00 2.43 F(3,2195) = 0.02ns Dairy products 12.43 9.54 11.43 6.18 13.24 8.19 12.71 9.56 F(3,2170) = .71ns Plant-based Meat substitutes (e.g., tofu, Quorn, 1.67a 1.87 0.87abc 1.90 0.26b 0.64 0.13c 0.98 F(3,2197) = 41.92*** seitan) Vegetarian cold cuts 0.59a 1.36 0.13a 0.54 0.05a 0.32 0.05a 0.77 F(3,2188) = 7.69*** Soy products (milk, yoghurt) 1.55a 3.45 0.55a 2.09 0.29a 1.42 0.12a 1.38 F(3,2182) = 14.30*** Legumes (e.g., lentils, peas, beans) 3.85a 5.30 2.13a 2.94 1.92a 2.32 1.71a 2.49 F(3,2194) = 10.22*** Note. One-way ANOVAs were used to investigate differences in the consumption of protein sources. Mean values within rows with no common superscript letter differ significantly (based on the Games-Howell post hoc test, p £ .01). † Including beef, pork, poultry and other types of meat such as venison or lamb. ‡ Total meat consumption differs slightly from the sum of the single meat items, because implausible extreme values in this variable were corrected and set to a maximum of 35 portions per week; this number also includes other types of meat (e.g., lamb, venison). n per group varies due to missing values. ** p < .01, *** p £ .001.

Table 3.5 Consumption of animal- and plant-based proteins (weekly portions) in adult females (n = 1,997).

Diet style Self- Self-declared declared Vegetarians & low-meat meat

vegans Pescatarians consumers consumers (n = 134) (n = 91) (n = 777) (n = 995) Protein source M SD M SD M SD M SD F(df1,df2) or c2 (df) Animal-based Unprocessed meat † 0.32a 0.71 2.09b 5.09 3.41c 2.27 5.49d 3.52 F(3,1959) = 159.11*** Processed meat (sausages, cold cuts) 0.14a 0.46 0.54a 1.21 1.88b 1.98 2.95c 2.67 F(3,1978) = 93.39*** Total meat ‡ 0.46a 1.07 2.53b 4.82 5.28c 3.35 8.41d 4.68 F(3,1956) = 221.99*** Fish and seafood 0.17a 0.37 1.43c 1.54 0.83b 1.03 0.87b 1.30 F(3,1971) = 22.25*** Eggs 2.16 2.73 2.26 2.01 2.14 2.14 2.01 2.05 F(3,1967) = 0.81ns Dairy products 11.18 10.43 12.58 9.64 13.17 8.13 12.45 7.52 F(3,1952) = 2.71ns Plant-based Meat substitutes (e.g., tofu, Quorn, 1.75a 2.17 1.09a 1.44 0.37b 0.75 0.12c 0.39 F(3,1975) = 171.16*** seitan) Vegetarian cold cuts 0.51a 1.36 0.25ab 0.77 0.03b 0.20 0.03b 0.26 F(3,1965) = 52.49*** Soy products (milk, yoghurt) 1.47a 3.23 1.05ab 2.25 0.49bc 1.91 0.26c 1.56 F(3,1966) = 19.18*** Legumes (e.g., lentils, peas, beans) 3.15a 4.30 2.86ab 4.09 1.59bc 2.06 1.39c 2.06 F(3,1972) = 28.70** Note. One-way ANOVAs were used to investigate differences in the consumption of protein sources. Mean values within rows with no common superscript letter differ significantly (based on the Games-Howell post hoc test, p < .01). † Including beef, pork, poultry and other types of meat such as venison or lamb. ‡ Total meat consumption differs slightly from the sum of the single meat items, because implausible extreme values in this variable were corrected and set to a maximum of 35 portions per week; this number also includes other types of meat (e.g., lamb, venison). n per group varies due to missing variables. ** p £ .01, *** p < .001.

MEAT AVOIDANCE

Predictors of meat consumption among consumers who reported eating little or no meat Self-declared low-meat consumers varied considerably in their meat intake levels (Tables 3.4 & 3.5). A three-step hierarchical regression analysis was conducted to identify factors associated with higher total meat intake in these participants (Table 3.6). The correlations between the predictors and total meat consumption are shown in Table 3.7. The final regression model including all variables was significant with F(17,1595) = 23.16, p < .001) and explained about 20% of the variance. Motives alone accounted for 11% of the variance in total meat consumption (Model 2); the sociodemographic variables (Model 1) and consumption of plant-based meat alternatives with the perceived difficulty of practising a low/no-meat diet (Model 3) explained an additional 4% and 5%, respectively. Female gender was a significant predictor of lower total meat consumption in models 1 and 2 (see Table 3.6) but no longer significant when the perceived difficulty of practising a low/no-meat diet and plant-based protein consumption were additionally included in model 3; age and income did not predict meat consumption. Stronger motives regarding animal welfare and a stronger preference for vegetarian foods were associated with lower meat consumption, and stronger weight loss motivation was associated with higher meat consumption. All the other motives included were not significant predictors. Participants who perceived following a low-meat or meatless diet as more difficult consumed more meat. None of the plant-based alternative sources of protein were associated with lower meat intake.

Table 3.6 Hierarchical multiple regression analysis predicting total meat consumption in adults who reported eating little or no meat (N = 1,596).

Predictors Model 1 Model 2 Model 3 B (SE B) 99% CI B (SE B) 99% CI B (SE B) 99% CI Constant 5.36*** (.56) [4.02, 6.89] 7.33*** (.63) [5.73, 9.00] 5.44*** (.65) [3.92, 7.12] Gender -1.66*** (.23) [-2.23, -1.09] -.70** (.22) [-1.30, -.11] -.52 (.22) [-1.08, .10] Age .01 (.01) [-.01, .03] .01 (.01) [-.01, .02] .01 (.01) [-.01, .03] Income .05 (.07) [-.13, .24] .09 (.07) [-.08, .27] .08 (.06) [-.07, .25] Animal welfare -.37*** (.06) [-.54, -.20] -.39*** (.07) [-.56, -.20] Environmental concerns -.10 (.06) [-.26, .05] -.13 (.06) [-.29, .03] Weight regulation .26** (.07) [.08, .43] .20** (.07) [.01, .38] Health/healthy diet -.04 (.05) [-.18, .09] -.04 (.05) [-.18, .09] Medical advice .19 (.10) [-.04, .43] .13 (.09) [-.09, .35] Disliking the taste of meat -.05 (.06) [-.20, .10] -.001 (.05) [-.14, .15] Preferring vegetarian dishes -.32*** (.05) [-.47, -.19] -.28*** (.06) [-.42, -.12] Religious rules -.20 (.13) [-.53, .19] -.20 (.12) [-.51, .10] Been brought up that way .04 (.06) [-.11, .18] .10 (.06) [-.06, .24] Social expectations .24 (.11) [-.05, .53] .13 (.11) [-.18, .38] Perceived difficulty .64*** (.09) [.41, .84] Meat substitutes .02 (.24) [-.54, .31] Soy products .05 (.08) [-.15, .20] Legumes .13 (.06) [-.02, .30] Note. Results are based on 1,000 bootstrap samples. Total R2 = .20, R2 = .04 for Step 1, DR2 = .11 for Step 2, DR2 = .05 for Step 3. B, unstandardised beta; SE B, standard error; 99% CI, 99% bias corrected accelerated confidence interval (predictors are significant if the BCa does not include 0). Gender: 1 = female, 0 = male. *** p £ .001, ** p < .01.

Table 3.7 Correlations of total meat consumption and predictors. Calculations based on the subsample of participants who reported eating little or no meat (n = 1,596).

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 Age - 2 Income -.20*** - 3 Animal welfare ns ns - 4 Environmental concerns -.09*** ns .55*** - 5 Weight regulation .14*** -.07** - -.07** - .09*** 6 Health/healthy diet .14*** ns .20*** .25*** .25*** - 7 Medical advice .15*** -.08** ns ns .28*** .10*** - 8 Disliking the taste of ns ns .15*** ns ns ns ns - meat 9 Preferring vegetarian ns ns .27*** .17*** ns .14*** ns .51*** - dishes 10 Religious rules ns -.08** ns ns .09*** ns .13*** .08** .08** - 11 Been brought up that ns -.07** ns .09*** .14*** .07** .13*** .07** ns .20*** - way 12 Social expectations .11*** ns ns .14*** .17*** ns .22*** ns ns .28*** .42*** - 13 Perceived difficulty† ns ns ns ns .14*** ns .10*** - - ns ns .14*** - .20*** .21*** 14 Meat substitutes -.17*** ns .20*** .15*** ns ns ns ns .15*** ns ns ns - - consumption‡ .10*** 15 Soy products -.11*** ns .12*** .10*** ns .09*** ns ns .08*** ns ns ns ns .52*** - consumption 16 Legumes consumption -.10*** ns .12*** .11*** .06** .10*** ns ns ns ns ns ns ns .29*** .21*** - 17 Total meat consumption .06** ns - - .16*** ns .12*** - - ns ns .09*** .26*** ns ns ns .26*** .16*** .15*** .24*** Note. ns not significant. n vary due to missing values. † Perceived difficulty of practising a diet with little or no meat. ‡ Including meat substitutes such as tofu, Quorn, seitan, and vegetarian cold cuts (in portions per week). ***p < .001 (two-tailed), **p < .01 (two-tailed).

MEAT AVOIDANCE

3.4 Discussion

The motives for pursuing a vegetarian or semivegetarian diet are diverse. The present study assessed the prevalence of commonly mentioned motives for following a vegetarian diet among a large sample of self-declared low-meat consumers as well as vegetarians/vegans and pescatarians. Our study confirmed the finding that most individuals are motivated by more than one reason but that certain motives differ in importance depending on the diet style with which people identify (De Backer & Hudders, 2014; Forestell et al., 2012). Although vegetarians and pescatarians indicated largely similar motives for their dietary choices regarding meat, compared to low-meat consumers these two groups are more strongly driven by concerns about animal welfare and environmental issues associated with meat consumption, as well as by taste preferences. Regarding health motives, no differences were observed between vegetarians and nonvegetarians in the motivation to eat healthily, but weight loss turned out to be a stronger motivation in female low-meat consumers than in female vegetarians. Previous studies have reported similar motivational differences regarding weight regulation, especially in young women (Forestell et al., 2012; Gilbody, Kirk, & Hill, 1999). However, the absolute values in our sample (see Figure 3.1) indicate that the weight-loss motivation was much less prevalent across all the diet groups than was the motivation to eat healthily. A possible explanation for why the groups did not differ regarding their motivation to eat healthily is that believes about the healthiness of meat vary among consumers, some of whom associate it with negative health outcomes and some of whom perceive meat as an important source of essential nutrients such as protein and iron, and therefore as a component of a healthy diet (Verbeke, Perez-Cueto, Barcellos, Krystallis, & Grunert, 2010). In a recent Belgian study, only around 22% of the respondents believed that meat consumption is unhealthy (Mullee et al., 2017). Moreover, around 24% even viewed eating vegetarian food frequently as unhealthy (Mullee et al., 2017). As suggested by previous research, consumers’ awareness of the environmental impact of meat consumption is generally rather low (Hartmann & Siegrist, 2017) and that they tend to underestimate it compared to other product characteristics, such as the packaging (Tobler, Visschers, & Siegrist, 2011). Because both the vegetarians and semivegetarians in our study reported a relatively high environmental motivation for their dietary choices regarding meat, we suggest that people practising these diet styles are more aware of and knowledgeable about the environmental issues associated with meat production than are regular meat consumers. However, this inference requires to be further exploration. Similarly to earlier studies (Rothgerber, 2014; Ruby, 2012; Vinnari, Montonen, Harkanen, & Mannisto, 2009), we observed an inconsistency between people’s self- declaration as vegetarian and their self-reported meat consumption in the food frequency

78 MEAT AVOIDANCE questionnaire. On average, vegetarians and pescatarians reported consuming meat occasionally or even regularly. There are several possible explanations for this discrepancy. Although there are many vegetarian and vegan alternatives to meat (Apostolidis & McLeay, 2016), vegetarians may still face situations in which where these foods are unavailable, leaving them no meat-free options (Ruby, 2012). Furthermore, in certain social situations occasional meat intake may be necessary to avoid embarrassment (Ruby, 2012). Interestingly, in our study the meat consumption of vegetarian males was considerably higher than that of vegetarian females. Apart from males’ generally higher food intake (Peter Herman & Polivy, 2010), a possible explanation for this finding is that males face more social pressure to eat meat in certain situations. Moreover, meat is often associated with typical masculine attributes (e.g., power and virility; Ruby & Heine, 2011), and vegetarian men are perceived as weaker and less masculine compared to omnivores (Ruby & Heine, 2011), especially by other men. This may also explain why males perceived it as more difficult to consistently follow a vegetarian diet in our sample. Nevertheless, even though the questionnaire in this study explicitly defined the diet styles, the self-evaluation of diet style seems to be problematic. People may have different ideas of what constitutes a vegetarian diet, and the social consensus regarding the definition of such a diet may not always be sufficiently clear (Rothgerber, 2014). Moreover, self-identified vegetarians may differ in how strictly they follow their diet, to what degree they allow themselves to make exceptions, and how strongly they experience cognitive dissonance or inner conflict when consuming meat. Apparently, although a substantial number of meat eaters view their meat intake as rather low, this is not in fact the case when comparing their consumption level with official dietary recommendations. In our sample, about 81% of the male and 69% of the female self- declared low-meat consumers exceeded the recommended upper limit of three portions per week (Swiss Society for Nutrition, 2017). This finding has implications for the efforts of public health organisations to bring peoples’ meat consumption within a healthy range. It demonstrates the need for further awareness campaigns to disseminate appropriate knowledge about the dietary recommendations for meat and to enable an accurate self- evaluation of one’s meat intake. Our analyses support the assumption that both vegetarian and semivegetarian diets are beneficial for health (Derbyshire, 2016; McEvoy, Temple, & Woodside, 2012), contributing to a better diet quality and to a healthy body weight. However, for obese males, these associations were not shown in our sample. Based on our results, vegetarians and pescatarians seemed not to compensate for the absence of meat in their diets by eating more other animal-based proteins such as eggs or dairy products. And for plant-based proteins, gender differences seem to exist. Whereas vegetarian women reported higher intake of all kinds of plant-based proteins, vegetarian males hardly differed from their omnivore

79 MEAT AVOIDANCE counterparts in this respect. Moreover, among low/no-meat consumers, higher consumption frequencies of plant-based protein sources were not associated with lower meat intake, which may indicate that the available products are not perceived as equivalent alternatives for meat. Nevertheless, regular and low-meat eaters also seem to consume meat substitutes on occasion. The low consumption frequencies reported, however, suggest that the products currently on the market may not be sufficiently satisfying, especially for avid meat eaters. Hence, these omnivores may represent an interesting market segment, especially considering their size and the potential market share of total consumption. Therefore, industries should focus their efforts on using new food technologies to improve the taste and texture of meat substitutes in order to increase their similarity to real meat. This may help to convince omnivores that such products are viable alternatives to meat and may encourage them to replace meat with plant-based protein sources more often. However, our results equally suggest that different target groups in the food market have opposite needs from meat substitutes. On the one hand, there is a large group of omnivores with a low willingness to give up meat consumption, creating a need for meat substitutes that are very similar to meat. On the other hand, as we have seen, many vegetarians do not like the taste of meat or prefer vegetarian options, creating a need for meat alternatives that do not taste like meat but have the same functionality.

Strengths and limitations The large sample size of randomly selected participants and the equal inclusion of males and females represent clear strengths of our study. However, there are also some limitations that should be mentioned. The first is related to the assessment of BMI. Self-reported body weight can be subject to underreporting, especially in women and heavier individuals (Gunnare, Silliman, & Morris, 2013) and thus might lead to less accurate estimations of weight status. However, the use of BMI based on self-reported data is very common in large nutritional studies and direct measurements would not be feasible. Moreover, it has been shown that the correlation between self-reported and objectively measured weight is relatively high (Luo et al., 2018). Another limitation is the way we asked participants whether or not they viewed their usual meat consumption as low. Future studies should provide a clearer definition of what is understood under a low consumption. Otherwise, participants’ understanding of low consumption will remain unclear as will the standard to which it is related, such as official dietary recommendations, the consumption level of friends and relatives, or a participant’s own consumption at an earlier time. We also did not assess consumers’ familiarity with the official dietary recommendations for meat, which is relevant.

80 MEAT AVOIDANCE

Conclusion and implications for public health This study supports the assumption that ethical motives along with taste preferences are more prevalent in consumers identifying with a vegetarian lifestyle, whereas weight-loss motivation plays a more important role for low-meat consumers, especially in women. It also contributes new evidence that vegetarian and semivegetarian diets may be associated with better diet quality and a lower prevalence of overweight, even though these associations should be further investigated in long-term studies. Further, the study reveals that a substantial number of consumers view their meat consumption as low even though this is not the case when comparing their intake levels to the official dietary recommendations. This has implications for public health organisations’ efforts to promote healthy levels of meat consumption in the public. Our study provides some hints for where to start. First, there seems to be a lack of knowledge about what is considered a low meat intake. Many people seem to have incorrect reference standards. This problem could be addressed with public awareness campaigns in order to improve consumers’ knowledge about and ability to appropriately evaluate their own meat intake. Second, our results suggest that the perceived difficulties of practising a diet with little or no meat were associated with higher meat consumption and thus constitute a barrier to reaching a healthy meat-intake level. This highlights the need for public health programmes which provide strategies to support people in eating more plant-based diets and to break their ‘bad’ habits regarding meat. Enhancing the familiarity with and preference for vegetarian alternatives, for example by launching campaigns for meatless days (e.g., ‘veggie day’ (de Boer, Schösler, & Aiking, 2014; Mullee et al., 2017)), implementing campaigns for portion size reduction(de Boer et al., 2014), or addressing concerns about animal welfare, may be effective ways to promote healthier and more sustainable diets in the public without suggesting that consumers must completely stop meat consumption.

References

Aiking, H. (2011). Future protein supply. Trends in Food Science & Technology, 22, 112-120. doi:10.1016/j.tifs.2010.04.005 Apostolidis, C., & McLeay, F. (2016). Should we stop meating like this? Reducing meat consumption through substitution. Food Policy, 65, 74-89. doi:10.1016/j.foodpol.2016.11.002 Appleby, P., Thorogood, J. I., & Key, T. J. (1998). Low body mass index in non-meat eaters: the possible roles of animal fat, dietary fibre and alcohol. International Journal of Obesity, 22, 454-460. Battaglia Richi, E., Baumer, B., Conrad, B., Darioli, R., Schmid, A., & Keller, U. (2015). Health Risks Associated with Meat Consumption: A Review of Epidemiological Studies.

81 MEAT AVOIDANCE

International Journal for Vitamin and Nutrition Research, 85(1-2), 70-78. doi:10.1024/0300-9831/a000224 Burkholder, N., Rajaram, S., & Sabaté, J. (2016). Vegetarian Diets. Encyclopedia of Food and Health, 401-410. doi:10.1016/B978-0-12-384947-2.00722-4 De Backer, C. J., & Hudders, L. (2014). From meatless Mondays to meatless Sundays: motivations for meat reduction among vegetarians and semi-vegetarians who mildly or significantly reduce their meat intake. Ecology of Food and Nutrition, 53(6), 639-657. doi:10.1080/03670244.2014.896797 de Boer, J., Schosler, H., & Aiking, H. (2017). Towards a reduced meat diet: Mindset and motivation of young vegetarians, low, medium and high meat-eaters. Appetite, 113, 387-397. doi:10.1016/j.appet.2017.03.007 de Boer, J., Schösler, H., & Aiking, H. (2014). "Meatless days" or "less but better"? Exploring strategies to adapt Western meat consumption to health and sustainability challenges. Appetite, 76, 120-128. doi:10.1016/j.appet.2014.02.002 de Jonge, J., van der Lans, I. A., & van Trijp, H. C. M. (2015). Different shades of grey: Compromise products to encourage animal friendly consumption. Food Quality and Preference, 45, 87-99. doi:10.1016/j.foodqual.2015.06.001 Derbyshire, E. J. (2016). Flexitarian Diets and Health: A Review of the Evidence-Based Literature. Front Nutr, 3, 55. doi:10.3389/fnut.2016.00055 FAO. (2018). Food Balance Sheets. Retrieved from http://www.fao.org/faostat/en/#data/FBS Field, A. (2013). Discovering statistics using IBM SPSS statistics. (4 ed.). London: SAGE. Forestell, C. A., Spaeth, A. M., & Kane, S. A. (2012). To eat or not to eat red meat. A closer look at the relationship between restrained eating and vegetarianism in college females. Appetite, 58, 319-325. doi:10.1016/j.appet.2011.10.015 Gilbody, S. M., Kirk, S. F., & Hill, A. J. (1999). Vegetarianism in Young Women: Another Means of Weight Control? International Journal of Eating Disorders, 26, 87-90. Gunnare, N. A., Silliman, K., & Morris, M. N. (2013). Accuracy of self-reported weight and role of gender, body mass index, weight satisfaction, weighing behavior, and physical activity among rural college students. Body Image, 10(3), 406-410. doi:10.1016/j.bodyim.2013.01.006 Hartmann, C., & Siegrist, M. (2017). Consumer perception and behaviour regarding sustainable protein consumption: A systematic review. Trends in Food Science & Technology, 61, 11-25. doi:10.1016/j.tifs.2016.12.006 Henchion, M., McCarthy, M., Resconi, V. C., & Troy, D. (2014). Meat consumption: trends and quality matters. Meat Sci, 98(3), 561-568. doi:10.1016/j.meatsci.2014.06.007 Hever, J., & Cronise, R. J. (2017). Plant-based nutrition for healthcare professionals: implementing diet as a primary modality in the prevention and treatment of chronic disease. J Geriatr Cardiol, 14(5), 355-368. doi:10.11909/j.issn.1671-5411.2017.05.012 Hu, F. B., Satija, A., Rimm, E. B., Spiegelman, D., Sampson, L., Rosner, B., . . . Willett, W. C. (2016). Diet Assessment Methods in the Nurses' Health Studies and Contribution to Evidence-Based Nutritional Policies and Guidelines. American Journal of Public Health, 106(9), 1567-1572. doi:10.2105/AJPH.2016.303348 Lassale, C., Gunter, M. J., Romaguera, D., Peelen, L. M., Van der Schouw, Y. T., Beulens, J. W., . . . Tzoulaki, I. (2016). Diet Quality Scores and Prediction of All-Cause, Cardiovascular and Cancer Mortality in a Pan-European Cohort Study. PloS One, 11(7), e0159025. doi:10.1371/journal.pone.0159025

82 MEAT AVOIDANCE

Luo, J., Thomson, C. A., Hendryx, M., Tinker, L. F., Manson, J. E., Li, Y., . . . Margolis, K. L. (2018). Accuracy of self-reported weight in the Women's Health Initiative. Public Health Nutrition, 1-10. doi:10.1017/S1368980018003002 Malik, V. S., Willett, W. C., & Hu, F. B. (2013). Global obesity: trends, risk factors and policy implications. Nature Reviews: Endocrinology, 9(1), 13-27. doi:10.1038/nrendo.2012.199 McEvoy, C. T., Temple, N., & Woodside, J. V. (2012). Vegetarian diets, low-meat diets and health: a review. Public Health Nutrition, 15(12), 2287-2294. doi:10.1017/S1368980012000936 McNaughton, S. A., Ball, K., Crawford, D., & Mishra, G. D. (2008). An Index of Diet and Eating Patterns Is a Valid Measure of Diet Quality in an Australian Population. The Journal of Nutrition, 138, 86-93. Mullee, A., Vermeire, L., Vanaelst, B., Mullie, P., Deriemaeker, P., Leenaert, T., . . . Huybrechts, I. (2017). Vegetarianism and meat consumption: A comparison of attitudes and beliefs between vegetarian, semi-vegetarian, and omnivorous subjects in Belgium. Appetite, 114, 299-305. doi:10.1016/j.appet.2017.03.052 OECD. (2019). Meat consumption (indicator) (Publication no. 10.1787/fa290fd0-en). Retrieved from https://data.oecd.org/agroutput/meat-consumption.htm Peter Herman, C., & Polivy, J. (2010). Sex and Gender Differences in Eating Behavior. In J. Chrisler & D. McCreary (Eds.), Handbook of Gender Research in Psychology (Vol. 1, pp. 455-469): Springer. Pollan, M. (2006). The omnivore’s dilemma. The search for a perfect meal in a fast-food world. London: Bloomsbury. Rees, J. H., Bamberg, S., Jäger, A., Victor, L., Bergmeyer, M., & Friese, M. (2018). Breaking the Habit: On the Highly Habitualized Nature of Meat Consumption and Implementation Intentions as One Effective Way of Reducing It. Basic and Applied Social Psychology, 40(3), 136-147. doi:10.1080/01973533.2018.1449111 Rothgerber, H. (2014). A comparison of attitudes toward meat and animals among strict and semi-vegetarians. Appetite, 72, 98-105. doi:10.1016/j.appet.2013.10.002 Rothgerber, H. (2015). Can you have your meat and eat it too? Conscientious omnivores, vegetarians, and adherence to diet. Appetite, 84, 196-203. doi:10.1016/j.appet.2014.10.012 Ruby, M. B. (2012). Vegetarianism. A blossoming field of study. Appetite, 58(1), 141-150. doi:10.1016/j.appet.2011.09.019 Ruby, M. B., & Heine, S. J. (2011). Meat, morals, and masculinity. Appetite, 56(2), 447-450. doi:10.1016/j.appet.2011.01.018 Ruby, M. B., Heine, S. J., Kamble, S., Cheng, T. K., & Waddar, M. (2013). Compassion and contamination. Cultural differences in vegetarianism. Appetite, 71, 340-348. doi:10.1016/j.appet.2013.09.004 Swiss Federal Statistical Office. (2012). Schweizerische Gesundheitsbefragung (SGB) [Swiss Federal Statistical Office: Swiss Health Survey (SHS)]. Neuchâtel, Switzerland: BFS Bundesamt für Statistik. Swiss Society for Nutrition. (2017). Schweizer Lebensmittelpyramide [Swiss food pyramid]. Retrieved from http://www.sge-ssn.ch/ich-und-du/essen-und- trinken/ausgewogen/schweizer-lebensmittelpyramide Swissveg. (2017). Veggie survey. Retrieved from https://www.swissveg.ch/veggie_survey?language=en

83 MEAT AVOIDANCE

Tobler, C., Visschers, V. H., & Siegrist, M. (2011). Eating green. Consumers' willingness to adopt ecological food consumption behaviors. Appetite, 57(3), 674-682. doi:10.1016/j.appet.2011.08.010 Tukker, A., Goldbohm, R. A., de Koning, A., Verheijden, M., Kleijn, R., Wolf, O., . . . Rueda- Cantuche, J. M. (2011). Environmental impacts of changes to healthier diets in Europe. Ecological Economics, 70, 1776-1788. doi:10.1016/j.ecolecon.2011.05.001 Verbeke, W., Perez-Cueto, F. J., Barcellos, M. D., Krystallis, A., & Grunert, K. G. (2010). European citizen and consumer attitudes and preferences regarding beef and pork. Meat Sci, 84(2), 284-292. doi:10.1016/j.meatsci.2009.05.001 Vinnari, M., Montonen, J., Harkanen, T., & Mannisto, S. (2009). Identifying vegetarians and their food consumption according to self-identification and operationalized definition in Finland. Public Health Nutrition, 12(4), 481-488. doi:10.1017/S1368980008002486 Westhoek, H., Lesschen, J. P., Rood, T., Wagner, S., De Marco, A., Murphy-Bokern, D., . . . Oenema, O. (2014). Food choices, health and environment: Effects of cutting Europe’s meat and dairy intake. Global Environmental Change, 26, 196-205. doi:10.1016/j.gloenvcha.2014.02.004 World Cancer Research Fund International. (2018). Recommendations and public health and policy implications. Retrieved from https://www.wcrf.org/sites/default/files/Cancer- Prevention-Recommendations-2018.pdf

Chapter 4

Intuitive eating and food choices 4 Intuitive eating and food choices

A study based on the Swiss Food Panel 2.0 Caroline Horwatha, Désirée Hagmannb, Christina Hartmannb a Department of Human Nutrition, University of Otago, Dunedin, New Zealand b ETH Zurich

Published manuscript Horwath, C., Hagmann, D., Hartmann, C. (2019). Intuitive eating and food intake in men and women: Results from the Swiss food panel study. Appetite 135, 61-71. doi: 10.1016/j.appet.2018.12.036

INTUITIVE EATING AND FOOD CHOICES

Abstract

This Although intuitive eating (IE) interventions have consistently shown benefits for psychological wellbeing and some have shown improvements in physical wellbeing, there is scarce information on the relationship between IE and food intake. Given the popularity of IE as an alternative to dieting, it's important to explore its relationship with food intake. The relationships between IE, Body Mass Index (BMI), diet quality, self-evaluation of dietary intake and physical activity were investigated. A randomly selected sample of adults from the German and French-speaking parts of Switzerland (N=5,238, 51% men, 20–100 years, BMI 15–62 kg/m2) completed a self-administered questionnaire comprising measures of a diverse range of eating related variables. Intuitive Eating was assessed with the IES-2. Food intake was measured with a semiquantitative food frequency questionnaire. Pearson correlations between the IES-2 and variables of interest were calculated for men and women separately. Although total IES-2 scores showed moderate negative correlations with BMI in men and women, the four IES-2 subscales showed different relationships with food intake. In contrast to the other subscales, unconditional permission to eat moderately correlated with poorer diet quality scores, and consistently showed associations with a more negative self-evaluation of eating behavior. The other three IES-2 subscales showed a few small positive and negative correlations with food intake, including small positive associations of diet quality scores in women, but not men, with eating for physical rather than emotional reasons and reliance on hunger and satiety cues. Further studies are needed to determine the impact of IE interventions on food intake.

86 INTUITIVE EATING AND FOOD CHOICES

4.1 Introduction

In response to the limited long-term success of dieting in achieving weight loss, the concept of intuitive eating (IE) was proposed as a promising alternative to the deliberate restriction of food intake (Tribole & Resch, 2012). It is generally defined as eating in response to physical hunger and satiety signals rather than emotional or external cues, while also giving oneself unconditional permission to eat when hungry and to eat those foods one desires (Tribole & Resch, 2012; Tylka & Wilcox, 2006). The growing literature on IE interventions has consistently revealed benefits for psychological wellbeing (e.g. depression, anxiety, body satisfaction, self- acceptance, quality of life), and a limited number of studies suggest physical health indicators other than body mass index (e.g. blood pressure, total and LDL cholesterol) may also be improved (Mensinger, Calogero, Stranges, & Tylka, 2016; Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). Although cross-sectionally IE is consistently associated with lower BMI, in randomized trials of IE interventions body weight has been either maintained or reduced (Mensinger et al., 2016; Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). As expected given that IE encourages individuals to let go of diet plans or rules regarding what, when and/or how much to eat, interventions show reductions in dieting behaviours and restraint (Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). Participants in IE interventions also experience reductions in disordered eating behaviours such as binge eating and disinhibition (Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). There is scarce information, however, on the relationship between IE and food intake, either from cross-sectional or intervention studies, and findings are mixed (Carbonneau et al., 2017; Mensinger et al., 2016). Such information is important for health professionals teaching the key messages of the IE approach. A potential barrier to adopting a more intuitive way of eating may be the concern that letting go of diet rules and lists of ‘forbidden’ foods may lead to increased consumption of such foods along with weight gain. In fact, some nutritionists may be reluctant to promote IE due to a concern that if individuals allow themselves to eat what they desire they will consume high levels of high fat or high sugar foods (Smith & Hawks, 2013). However, since IE involves the idea of ‘body wisdom’ (i.e. if we pay attention to our body signals to guide us regarding what, when and how much to eat, we'll consume foods that support our wellbeing) (Tribole & Resch, 2012), intuitive eating would be expected to lead to patterns of food intake that honour health (Carbonneau et al., 2017). One principle of IE described by Tribole and Resch (2012) specifically refers to making “food choices that honour your health and tastebuds while making you feel well”. Furthermore, since restrained eaters tend to overeat (Herman & Polivy, 1983) and emotional eaters tend to consume more fatty or sweet foods (Oliver, Wardle, & Gibson, 2000), reductions in restraint and emotional eating may also lead intuitive eaters to have lower intakes of such foods.

87 INTUITIVE EATING AND FOOD CHOICES

The most widely used measure of IE, the Intuitive Eating Scale (IES) (Tylka, 2006), has been improved upon with the new IES-2 (Tylka & Kroon Van Diest, 2013). The IES-2 encompasses four dimensions: eating for physical rather than emotional reasons (EPR), reliance on hunger and satiety cues (RHSC), unconditional permission to eat when hungry and those foods desired (UPE), and body-food choice congruence (BFCC). The latter assesses the ‘gentle nutrition’ component articulated by Tribole and Resch (2012), which refers to making food choices that honour health and body functioning (e.g. foods that promote body performance, energy or stamina) in addition to tasting good. When teaching IE, the B-FCC aspect is introduced after the other three aspects have been learned so that a “healthy relationship with food” is first established and health-based choices do not feel like dieting or deprivation (Tribole & Resch, 2012). In some cross-sectional studies, IE has been found to be largely unrelated to self- reported food intake (Madden, Leong, Gray, & Horwath, 2012; Smith & Hawks, 2013). For example, in a nationwide random sample of middle-aged women Madden et al. (2012) found total IE scores were unrelated to the frequency of intake of high fat/ high sugar foods or fruit intake, and in a student sample Smith and Hawks (2013) reported no association with junk food intake or breakfast consumption. Although total IE scores were associated with higher vegetable intake, the effect was too small to be of practical significance (Madden et al., 2012). To date, only one study has explored the relationships between food intake and the different dimensions of IE in a large population sample of men and women (Camilleri et al., 2017). This study explored the first three dimensions of the IES-2 scale, and reported that the different aspects of IE may relate to food intake in different ways. Few studies have explored the impact of IE interventions on food intake and findings are mixed. Three studies reported a positive impact on diet quality scores (Carbonneau et al., 2017; Hawley et al., 2008; Mensinger et al., 2016), while others showed no improvements in food intake (Cole & Horacek, 2010; Leblanc et al., 2012). There is a scarcity of studies that examined the relationship between IE and food intake patterns. Therefore, the aim of the present study was to examine the associations of all 4 dimensions of the IES-2 (EPR, RHSC, UPE and B-FCC), with both self-reported food intake and subjective perceptions of eating behaviour in men and women from a large population- based sample in Switzerland.

88 INTUITIVE EATING AND FOOD CHOICES

4.2 Methods

4.2.1 Participants and procedure

This study is based on data from the first wave of the Swiss Food Panel 2.0, a longitudinal study of the eating behavior of the Swiss population. The Food Panel 2.0, which follows Food Panel 1.0 (2010–2014), involves a new set of participants and new research questions. Data collection for the Swiss Food Panel 2.0 started in spring 2017 and will continue with annual follow-up surveys. The primary aim of this study is to investigate dietary and weight management behaviors, physical activity, and factors influencing these behaviors, in addition to exploring changes over time. For the present study, only crosssectional data from 2017 were analyzed. Participants completed a paper-and-pencil questionnaire that gathered information about food intake, socio-demographic characteristics, and various psychological constructs related to eating behavior. Only those constructs relevant to the present study are described below. Questionnaires were mailed to a random sample of residents in the German- and French-speaking parts of Switzerland (N=23,002). Most were randomly selected from the phone book, but in order to increase the percentage of younger people who are often not registered in the phone book, some extra addresses of people aged between 20 and 30 were bought from an address company. Questionnaires were returned by 5781 people, giving a response rate of 25% for the telephone book and 20% for the address company. Participants who did not indicate their sex or age, those who completed less than 50% of the questionnaire and pregnant women were excluded. The final sample consisted of 5238 participants (51% men) with a mean age of 56.5 years (SD=17.3; range=20–100); 74% were from the German- speaking part of Switzerland (see Table 4.1). Compared to the general Swiss population (Swiss Federal Statistical Office, 2016), adults aged 20–39 years were underrepresented (17% versus census 33%). Percentage of individuals with a higher educational level (higher professional school or university degree) was higher (53.9%) compared to the census (30%) (Federal Statistical Office, 2016, p. 25). The study was approved by the Ethics Committee of ETH Zurich (EK 2017-N-19).

89 INTUITIVE EATING AND FOOD CHOICES

Table 4.1 Sociodemographic characteristics of the study population (N = 5,238).

Mean (SD) or % Men 51% Age [years] 56.48 (17.25) French-speaking part 26.2% Body Mass Index (BMI) 24.84 (4.35) Underweight (BMI < 18.5) Men 0.5% Women 5.5% Normal weight (BMI 18.5 - 24.9) Men 44.7% Women 66.1% Overweight (BMI 25-29.9) Men 42.1% Women 19.3% Obesity (BMI ≥ 30) Men 12.8% Women 9.0% Last finished school Primary and lower secondary school 5.4% Vocational education, middle school 38.1% Higher vocational education 36.9% University 16.2% No education/missing 3.5%

4.2.2 Measures Intuitive Eating. The validated German translation (van Dyck, Herbert, Happ, Kleveman, & Vogele, 2016) of the 23-item Intuitive Eating Scale (IES-2, Tylka & Kroon Van Diest, 2013) was used to measure intuitive eating. The IES-2 consists of four subscales: Unconditional permission to eat (UPE), Reliance on hunger and satiety cues (RHSC), Eating for physical rather than emotional reasons (EPR), and Body-Food-Choice Congruence (B-FCC). The response options ranged from 1 (“Do not agree at all”) to 5 (“Fully agree”). The subscales can be combined to a total score indicating the overall measure of intuitive eating. For the German version of the IES-2, the wording of one item was slightly adapted so that it was closer to the original English version. The validated French translation of the IES-2 was used in the French-

90 INTUITIVE EATING AND FOOD CHOICES

speaking part of Switzerland (Camilleri et al., 2015). Camilleri et al. (2015) only reproduced three factors in a French adult sample. The three-item B-FCC subscale was not reproduced in the French data and two items from the UPE subscale were omitted because of high crossloadings. Therefore, for the present study the five items missing from the French version were added. Dietary assessment. Participants completed a semiquantitative food frequency questionnaire (sFFQ) which consisted of a subset of food items and nine response options from the Nurses' Health Study questionnaire (Hu et al., 2016). Usual consumption frequencies in the last year for 47 food items and food groups were assessed. As in the original dietary questionnaire, standard portions for every food were provided (e.g., “1 glass/2 dl of milk” instead of “milk”) and participants were asked to indicate how many standard portions of the food they usually consumed. The nine response options were recoded to reflect the number of standard portions consumed per week: 4 or more per day (coded 28 portions per week), 3 per day (coded 21), 2 per day (coded 14), 1 per day (coded 7), 5–6 per week (coded 5.5), 2– 4 per week (coded 3), 1 per week (coded 1), 1–3 per month (coded 0.5) and seldom/never (coded 0). Higher frequency response options of more than 4 portions per day used in the original Nurses’ Health Study questionnaire were omitted since they were not appropriate for those food groups assessed in this study. Foods analysed in the present study are displayed in Table 4.2. These core foods were chosen because their over- or underconsumption are potential risk factors for the development of non-communicable diseases (Lassale et al., 2016; Malik, Willett, & Hu, 2013) and they are part of national nutrition recommendations and dietary guidelines for adults. Single food items were collapsed into food groups. A plausibility check indicated that for total meat consumption some participants (n=47) reported implausibly high consumption which was consequently set to an upper limit of 35 portions per week. Participants indicated on a five-point response scale ranging from daily (coded 7), 4–6 times per week (coded 5), 1–3 times per week (coded 2), 1–3 times per month (coded 0.5) to less or never (coded 0), how often they eat something from a fast food restaurant (e.g., McDonald's, Burger King). Participants were additionally asked how often they consume breaded/fried meat and fish. Response categories varied from daily (coded 7), several times per week (coded 3), once per week (coded 1), 1–3 per month (coded 0.5) and seldom/never (coded 0). Diet quality index. Nine items from the semi-quantitative FFQ were used to create a diet quality index by applying sample-derived median cut-off values. The diet quality index was constructed based on food groups and specific foods that are typically included in such diet quality indices, because they are linked to human health and dietary guidelines (Kourlaba & Panagiotakos, 2009; Waijers, Feskens, & Ocke, 2007; Wirt & Collins, 2009). Following the

91 INTUITIVE EATING AND FOOD CHOICES

scoring scheme from previous publications (Mötteli & Dohle, 2017), participants received one point if their consumption of fruits, raw and cooked vegetables as well as whole grains was at or above the study sample median. For sweet and salty snacks, sugar-sweetened beverages, meat and processed meat, breaded/fried meat and fish, alcohol and fast foods, participants received one point for each food if their consumption (portions per week) was at or below the study sample median. The food groups and median intake levels included in the index were: weekly consumption of fruit (Mdn=7.00, IQR=11.00), raw and cooked vegetables (Mdn=14.00, IQR=11.00), sweet and salty snacks (Mdn=6.75, IQR=7.50), sugarsweetened beverages (Mdn=0.00, 55% consumed zero portions), meat and processed meat (Mdn=7.00, IQR=6.50), whole grains (Mdn=4.50, IQR=8.00), breaded/fried meat and fish (Mdn=0.00, 51% zero portions), alcohol (Mdn=3.00, IQR=6.50), and the frequency of eating something from a fast food restaurant (Mdn=0.00, 83% reported “rarely/never”). Finally, the score was built by summing up the points for all nine food groups. The index ranged from 0 (lower diet quality) to 9 (higher diet quality) and had a bell-shaped distributed (M=5.04, SD=2.05). Subjective perception of eating behavior. Participants evaluated their eating behavior with regard to six attributes on a semantic differential scale ranging from −3 (coded 1) to 3 (coded 7). Participants were asked: “Please indicate with the terms below how you evaluate your eating behavior.” The terms (attributes) were unhealthy – healthy, low in calories – high in calories, low in fat – high in fat, low in sugar –high in sugar, mostly small portions – mostly large portions, low selfcontrol when eating – high self-control when eating. In order to have consistency in the direction of the coding, the attributes calories, fat, sugar and portion size were recoded so that high values indicated low in fat, low in calories, low in sugar and small portions. Participants’ selfperception of the frequency with which they overeat was assessed with the item “How often does it happen to you that you eat too much or overeat?” Five response options ranged from “daily” to “rarely/never”. The item was recoded so that high values indicate high overeating frequencies per week. Weight status. As an indicator of the participants’ weight status, body mass index (BMI) was calculated by dividing the self-reported body weight (in kg) by the squared height

(in m2). The average BMI in the study sample was 23.7 (SD=4.4; range=15–55) for women and 25.9 (SD=4.0; range 16–62) for men. Average BMI for men and women in the present study was similar to the average BMI observed in the Swiss National Nutrition Survey (based 2 on anthropometric measures by trained personnel): women 24.1 kg/m2, men 25.9 kg/m (Bochud, Chatelan, Blanco, & Beer-Borst, 2017, p. 33). Participants with a BMI between 25 and 29.9 kg/m2 were classified as overweight and those with a BMI ≥30 kg/m2 as obese. Prevalence of overweight in the study sample (see Table 4.1) was comparable to the prevalence observed in the Swiss National Nutrition Survey. However, slightly fewer men

Table 4.2 Food group compositions, standard portion definitions, mean values and standard deviations for the number of standard portions consumed per week of the different food groups assessed in the semiquantitative food frequency questionnaire (N = 5’238).

Mean (SD)a Statistic Food group Food items included Standard portion Women Men t(df) p

Meat (6 items) Pork, beef/veal, poultry, other 100-120 g or piece or 3 6.33 (4.85) 9.25 (6.01) 19.13 (5122) < .001 types of meat, sausages, cold slices/30 g cuts Vegetables (2 items) Salad/vegetables (raw), handful/120g 16.62 (9.59) 12.65 (7.64) -16.51 (5175) < .001 vegetables (cooked) Fruit (1 item) Fruits (e.g., apples, grapes) handful/120g 10.29 (6.99) 7.08 (6.09) -17.68 (5181) < .001 Salty snacks (1 item) Salty snacks (e.g., chips, salted handful 0.93 (1.58) 1.03 (1.74) 2.18 (5190) n.s. nuts) Sweets (6 items) Candies/gummy bears, cookies, handful or 3 pieces or 1 7.62 (7.43) 7.57 (8.55) -.19 (5036) n.s. chocolate, sweet pastries, piece or teaspoon or chocolate spread, milk desserts cup/150-200g Dairy products (4 Milk, hard and soft cheese, Glass/2dl or Piece/30-60g 12.71 (8.43) 13.07 (9.56) 1.45 (5111) n.s. items) quark, yogurt, cottage cheese, or Cup/2dl or Bottle/2.5dl yogurt drink Whole grains (3 item) Cereals, bread, pasta/rice Cup/2dl or Slice 7.37 (7.14) 6.18 (7.34) -5.83 (5038) < .001

SSB (1 item) Sugar-sweetened beverages Glass/2dl 1.18 (3.57) 2.13 (4.88) 7.92 (5148) < .001 (e.g., ice tea, coke, energy drinks) Alcohol (4 items) Red wine, white wine, beer, Glass/1dl or Glass/3dl or 3.63 (5.27) 8.03 (9.68) 20.05 (5117) < .001 liquor 4cl a for consumption frequency of standard portions

INTUITIVE EATING AND FOOD CHOICES

(12.8% vs. 13.9%) and fewer women (9.0% versus 11.3%) taking part in the present study were classified as obese compared to the National Survey (Bochud et al., 2017). Physical activity was measured with a brief instrument published and validated by Johansson and Westerterp (2008) which assesses recreational and occupational physical activity. Occupational physical activity was assessed with the question “Please describe your physical activity at work (even work at home, sick leave at home and studying, for instance in a university)”. The response options were: (1) very light, e.g., sitting at the computer most of the day or sitting at a desk; (2) light, e.g., light industrial work, sales or office work that comprises light activities; (3) moderate, e.g., cleaning, staffing at kitchen or delivering mail on foot or by bicycle; and (4) heavy, e.g., heavy industrial work, construction work or farming. In addition, recreational physical activity was assessed using the question “Please describe your physical activity at leisure time. If the activities vary between summer and winter, try to give a mean estimate.” The respondents were offered five response options: (1) very light: almost no activity at all; (2) light, e.g., walking, nonstrenuous cycling or gardening approximately once a week; (3) moderate: regular activity at least once a week, e.g., walking, bicycling, or gardening or walking to work 10–30 min a day; (4) active: regular activities more than once a week, e.g., intense walking or bicycling or sports; (5) very active: strenuous activities several times a week. Based on the indicated activity level in these two domains, overall activity level can be categorized into very light, light, moderate, active and very active (Johansson & Westerterp, 2008).

4.2.3 Statistical analysis A principal component analysis (PCA) with orthogonal rotation (varimax) was 1 conducted with the IES-2 data . The Kaiser's criterion of 1 and the screeplot were used to determine the number of components to retain. In addition, parallel analysis was run utilizing the parallel.sps script developed by O'Connor (2000) in order to determine the number of components to be retained. Moreover, the rotated component matrix was inspected for 1) low factor loadings (<0.4), 2) for high crossloadings on different factors, and 3) for items that loaded on the wrong factor. Reliability of the scale was assessed using Cronbach's Alpha. Pearson correlations were calculated for the various eating behavior related variables with all four subscales of the IES-2 and the total IES-2 score. Statistical analyses were performed with IBM SPSS Statistics 25 software. Only correlations of at least small magnitude (i.e. r≥0.1) (Cohen, 1988) and p < 0.001 have been interpreted.

1 For separate PCAs for the French- and German-speaking subsamples as well as for men and women see Table 4.3. For Cronbach's Alphas for the subsamples see Table 4.4.

Table 4.3 Rotated component matrix for the IES-2 for different subsamples of the Swiss Food Panel (varimax rotation, only factor loadings >.4 are displayed).

German-version French-version Men Women Label Item 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Reliance on hunger and satiety cues

IE6 I trust my body to tell me when .705 .735 .705 .726 to eat.

IE7 I trust my body to tell me what .664 .707 .659 .691 to eat.

IE8 I trust my body to tell me how .785 .798 .783 .786 much to eat.

IE21 I rely on my hunger signals to .678 .705 .671 .701 tell me when to eat.

I rely on my fullness (satiety) IE22 signals to tell me when to stop .764 .711 .738 .751 eating.

IE23 I trust my body to tell me when .807 .764 .786 .796 to stop eating.

German-version French-version Men Women Label Item 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Eating for physical rather than emotional reasons

I find myself eating when I’m IE2_R feeling emotional (e.g., .791 .781 .761 .784 anxious, depressed, sad), even when I’m not physically hungry.

I find myself eating when I am IE5_R lonely, even when I’m not .775 .770 .773 .776 physically hungry.

IE10_R I use food to help me soothe .793 .850 .805 .819 my negative emotions.

I find myself eating when I am IE11_R stressed out, even when I’m .810 .852 .818 .825 not physically hungry.

I am able to cope with my IE12 negative emotions (e.g., .686 .486 .505 .682 .498 .503 anxiety, sadness) without turning to food for comfort.

IE13 When I am bored, I do NOT eat .766 .859 .781 .823 just for something to do.

German-version French-version Men Women Label Item 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

IE14 When I am lonely, I do NOT .766 .861 .798 .779 turn to food for comfort. I find other ways to cope with IE15 stress and anxiety than by .432 .560 .550 .618 .539 .443 eating. Unconditional permission to eat I try to avoid certain foods high IE1_R in fat, carbohydrates, or .649 .660 .664 .643 calories.

IE3 If I am craving a certain food, I .590 .497 .561 .590 allow myself to have it.

IE4_R I get mad at myself for eating .519 .464 .507 .449 .478 .462 .468 something unhealthy.

IE9_R I have forbidden foods that I .614 .642 .626 .635 don’t allow myself to eat.

IE16 I allow myself to eat what food .686 .579 .662 .661 I desire at the moment.

I do NOT follow eating rules or IE17 dieting plans that dictate what, .502 .604 .510 .530 when, and/or how much to eat.

German-version French-version Men Women Label Item 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Body-Food choice congruence

IE18 Most of the time, I desire to .455 .477 .562 .541 .417 .464 eat nutritious foods. I mostly eat foods that make IE19 my body perform efficiently .887 .829 .876 .878 (well).

IE20 I mostly eat foods that give my .887 .861 .880 .879 body energy and stamina.

Note. R = recoded

INTUITIVE EATING AND FOOD CHOICES

Table 4.4 Cronbach’s alpha for the Intuitive Eating Scale-2 (IES-2) total score and the four subscales separately for gender and the two language versions (N = 5’238).

Whole German French sample Men Women version version

IES-2 Total score .82 .80 .84 .83 .81

UPE subscale .65 .65 .65 .66 .62

EPR subscale .85 .81 .87 .85 .83

RHSC subscale .85 .83 .86 .85 .85

B-FCC subscale .64 .67 .62 .64 .65

IES-2 = Intuitive Eating Scale-2, UPE = Unconditional permission to eat, EPR= Eating for physical rather than emotional reasons, RHSC = Reliance on hunger and satiety cues, B-FCC = Body-food choice congruence.

4.3 Results

4.3.1 Factor structure of the IES-2 and gender effects Parallel analyses based on principal component analyses yielded five components to be retained. The scree-plot indicated a five-factor solution and five components had eigenvalues over Kaiser's criterion of 1 and in combination they explained 58.6% of the variance. Inspection of the rotated component matrix revealed that all items loaded on the right factor with factor loading of 0.4 or higher, and only the negatively worded items of the EPR subscale loaded on a separate factor2 (Table 4.5). However, combining all items of the EPR subscale revealed good alpha reliabilities of 0.85. The item “I get mad at myself for eating something unhealthy” loaded on two factors; one representing the UPE subscale and one representing the EPR subscale. Another cross-loading was observed for the item “Most of the time I desire to eat nutritious foods” on the UPE subscale and the B-FCC subscale. Cronbach's alpha for the UPE subscale was low (0.65) for the whole sample; however, all item total correlations for the UPE items were above 0.3 and thus acceptable. The three-item B-FCC subscale also had a low alpha reliability (0.64), but acceptable item-total correlations of > 0.3. Alpha reliabilities for the

2 Difficulties in understanding negatively worded items can be experienced by participants, especially when the described state is not in accordance with the actual state of the respondents (van Sonderen, Sanderman, & Coyne, 2013). This can lead to the common observation that the negatively worded items load on a separate factor. Nevertheless, we kept the same subscales as in the original publication (Tylka & Kroon Van Diest, 2013) in order to make results comparable between studies.

99 INTUITIVE EATING AND FOOD CHOICES

RHSC subscale, the EPR subscale and the total score were very good at 0.85, 0.85 and 0.82, respectively (Table 4.6).

Table 4.5 Rotated component matrix of the IES-2 (varimax rotation), data from the Swiss Food Panel.

Components Label Item 1 2 3 4 5 Reliance on hunger and satiety cues IE6 I trust my body to tell me when to eat. .712 IE7 I trust my body to tell me what to eat. .675 IE8 I trust my body to tell me how much to eat. .788 IE21 I rely on my hunger signals to tell me when to eat. .684 I rely on my fullness (satiety) signals to tell me when to IE22 .752 stop eating. IE23 I trust my body to tell me when to stop eating. .797

Eating for physical rather than emotional reasons I find myself eating when I’m feeling emotional (e.g., IE2_R anxious, depressed, sad), even when I’m not .782 physically hungry. I find myself eating when I am lonely, even when I’m IE5_R .774 not physically hungry. IE10_R I use food to help me soothe my negative emotions. .810 I find myself eating when I am stressed out, even when IE11_R .824 I’m not physically hungry. I am able to cope with my negative emotions (e.g., IE12 .635 anxiety, sadness) without turning to food for comfort. When I am bored, I do NOT eat just for something to IE13 .804 do. IE14 When I am lonely, I do NOT turn to food for comfort. .806 I find other ways to cope with stress and anxiety than IE15 .409 .567 by eating. Unconditional permission to eat I try to avoid certain foods high in fat, carbohydrates, IE1_R .651 or calories.

100 INTUITIVE EATING AND FOOD CHOICES

Components Label Item 1 2 3 4 5 IE3 If I am craving a certain food, I allow myself to have it. .569 IE4_R I get mad at myself for eating something unhealthy .514 .462 IE9_R I have forbidden foods that I don’t allow myself to eat. .620 IE16 I allow myself to eat what food I desire at the moment. .668 I do NOT follow eating rules or dieting plans that IE17 .526 dictate what, when, and/or how much to eat. Body-Food Choice congruence E18 Most of the time, I desire to eat nutritious foods. .414 .502 I mostly eat foods that make my body perform IE19 .875 efficiently (well). I mostly eat foods that give my body energy and IE20 .879 stamina.

Note. R = items recoded, only factor loadings >.4 are displayed.

Mean values for the IES-2 for men and women separately and intercorrelations between subscales are displayed in Table 4.6. Men scored higher (p < .001) on the UPE and the EPR subscales, but not on the RHSC subscale. No sex difference was observed for the B- FCC subscale. The B-FCC subscale correlated either comparatively weakly or not at all with the other IES-2 subscales in men and women.

4.3.2 Intuitive eating and food intake The IES total score did not show meaningful correlations with most of the sFFQ variables (Table 4.7). The strongest correlational relationships were observed for the unconditional permission to eat subscale (UPE) and food intake; however, effect sizes were small (rs=0.10 to 0.21, all p < 0.001). In particular, high scores on the UPE subscale were positively correlated with higher consumption frequencies of salty snacks, SSB, sweets and meat, as well as lower consumption frequencies of vegetables, fruits and whole grains. The UPE subscale also showed small positive correlations with frequency of fast food consumption. The same correlational pattern was observed for men and women, even though the strengths of the relationships varied slightly for the different food groups. The UPE subscale was the only subscale that was moderately negatively correlated with diet quality, indicating that in both men (r=−0.33, p < .001) and women (r=−0.29, p < .001) those scoring high on UPE had poorer

101 INTUITIVE EATING AND FOOD CHOICES

diet quality. For the eating for physical rather than emotional reasons subscale (EPR), small inverse associations were found with the number of standard portions of sweets consumed (r=−0.13, p < .001 for both sexes) and frequency of fast food consumption (men r=−0.11, p < .001, women r=−0.10, p < .001) per week. With regard to the other food groups, no meaningful correlations were observed. For reliance on hunger and satiety cues (RHSC), there were no noteworthy correlations with consumption frequencies for individual food groups. Only diet quality showed small positive associations with both RHSC in women (r=0.12, p < .001) and EPR in women (r=0.14, p < .001), but not men. Interestingly, the body-food choice congruence subscale was unrelated to diet quality. For B-FCC, there were small positive associations with intake of dairy products (r=0.11, p < .001) and meat (r=0.13, p < .001) in men, and with whole grain intake (r=0.11, p < .001) in women.

102

Table 4.6 Means, standard deviations and inter-correlations for the Intuitive Eating Scale-2 (IES-2) total score and the four subscales.

Cronbach’s Men Women

Alpha Mean SD Mean SD t(df) 1 2 3 4 5

1 IES-2 Total score .82 3.71 0.47 3.62 0.53 6.24(4888)*** – .59 .76 .68 .30

2 UPE subscale .65 3.55 0.70 3.44 0.71 5.87(5110)*** .56 – .22 .17 n.s.

3 EPR subscale .85 4.06 0.76 3.84 0.88 9.57(5062)*** .80 .18 – .27 n.s.

4 RHSC subscale .85 3.54 0.78 3.63 0.81 -4.02(5086)*** .74 .24 .38 – .18

5 B-FCC subscale .64 3.40 0.72 3.36 0.71 1.89(5147) .29 n.s. n.s. .21 –

Note. p £ .001 for all correlations displayed, n.s. = not significant (p > .001). IES-2 = Intuitive Eating Scale-2, UPE = Unconditional permission to eat, EPR= Eating for physical rather than emotional reasons, RHSC = Reliance on hunger and satiety cues, B-FCC = Body-food choice congruence. Possible range for all IES-2 variables 1-5.

Table 4.7 Pearson correlations for the associations between Intuitive eating (IES-2), food intake, physical activity level and BMI (N = 5’238).

Men (n = 2’685)a Women (n = 2’553)a

IES-2 IES-2 total total score UPE EPR RHSC B-FCC score UPE EPR RHSC B-FCC Food intake [number of standard portions per week] Vegetables n.s. -.12 n.s. n.s. .08 n.s. -.15 n.s. .08 .08 Fruits n.s. -.19 n.s. n.s. .07 n.s. -.18 n.s. n.s. n.s. Whole grains n.s. -.15 n.s. n.s. .09 n.s. -.14 n.s. .10 .11 Dairy products n.s. n.s. n.s. n.s. .11 n.s. n.s. n.s. n.s. n.s. Meat n.s. .18 n.s. n.s. .13 n.s. .11 -.08 -.07 n.s. Sweets n.s. .13 -.13 n.s. n.s. n.s. .10 -.13 n.s. n.s. Salty snacks n.s. .13 -.08 n.s. n.s. n.s. .12 -.09 n.s. n.s. SSBs .11 .21 n.s. n.s. .07 n.s. .16 n.s. n.s. n.s. Alcohol n.s. .07 n.s. n.s. -.10 n.s. .11 n.s. n.s. n.s. Diet quality index -.10 -.33 n.s. n.s. n.s. n.s. -.29 .14 .12 .07 Fast food [frequency per week] n.s. .11 -.11 -.07 n.s. n.s. .12 -.10 n.s. n.s. Physical activity level .11 n.s. n.s. .08 .24 .10 n.s. .12 .07 .12 BMI [kg/m2] -.30 -.13 -.25 -.21 -.08 -.34 -.11 -.32 -.26 n.s. Note. Only significant correlations on a significance level of p £ .001 are displayed, n.s. = not significant (p > .001). a n varies due to missing values. BMI = Body mass index, SSB = Sugar-sweetened beverages, IES-2 = Intuitive Eating Scale-2, UPE = Unconditional permission to eat, EPR = Eating for physical rather than emotional reasons, RHSC = Reliance on hunger and satiety cues, B- FCC = Body-food choice congruence.

INTUITIVE EATING AND FOOD CHOICES

4.3.3 Intuitive eating and self-evaluation of diet Correlational patterns between the IES-2 subscales and participants' evaluation of their dietary behavior on a list of attributes were also examined (Table 4.8). High scores on the UPE subscale were consistently associated with a more negative evaluation of one's diet; however, effect sizes were small (rs=−0.10 to −0.28, all p < 0.001) with the exception of moderate inverse associations with diet-related health consciousness. Those scoring higher on UPE perceived their diet to be less healthy, high in calories, high in fat and high in sugar. They also reported larger portions and lower self-control capacities with regards to eating behavior. The UPE subscale was the only subscale of the IES-2 that was consistently associated with a negative evaluation of one's diet. For the other subscales, to varying degrees, small to moderate positive associations were found with participants' perceptions of their diet. For example, EPR and RHSC, respectively, were both negatively correlated with perceived frequency of overeating in men (r=−0.23 and −0.27, p < .001) and women (r=−0.43 and −0.36, p < .001). In addition, among women EPR was moderately associated with the perception of higher self-control regarding eating (r=0.30, p < .001). In contrast to the findings for the UPE subscale, the other subscales showed either no relation or a small positive association with diet-related health consciousness.

4.3.4 Intuitive eating, BMI and physical activity The UPE, EPR and RHSC subscales were found to have small to moderate negative associations with BMI (rs=−0.11 to −0.32, all p≤0.001) (Table 4.7), with the strongest associations observed for the EPR subscale and total IES-2 scores, respectively (men: r=−0.25 and −0.30, women: r=−0.32 and −0.34, all p < .001). Lastly, small positive associations with physical activity were observed for the B-FCC subscale in both sexes (men: r=0.24, women: r=0.12, both p < .001), and among women only for the EPR subscale (r=0.12, p < .001).

4.4 Discussion

This study shows that although total IES-2 scores had a moderate inverse correlation with BMI and three IES-2 subscales (UPE, EPR, RHSC) were inversely related with BMI in both men and women, the four aspects of IE showed different relationships with food intake. The EPR, RHSC and B-FCC aspects of intuitive eating showed only a few small positive correlations with food intake, including small positive associations of diet quality scores with EPR and RHSC in women but not men. The EPR, RHSC and B-FCC subscales were unrelated to key aspects of

105 INTUITIVE EATING AND FOOD CHOICES

dietary guidelines such as fruit and vegetable consumption. The finding in both sexes that eating for physical rather than emotional reasons showed small inverse associations with consumption of sweets and fast foods is consistent with the findings of a large French study (Camilleri et al., 2017) and also with research showing emotional eaters tend to consume more fatty or sweet foods (Oliver et al., 2000). However, reliance on hunger and satiety cues was unrelated to the consumption of sweets and fast foods, nor did it show an association with dairy or meat consumption as reported in the large French study (Camilleri et al., 2017). In contrast, in men and women unconditional permission to eat scores were moderately correlated with poorer diet quality scores. UPE also showed small associations with lower intakes of fruits, vegetables, and wholegrains, and higher intakes of sweets, fast foods, salty snacks, and, in women, alcohol. This is consistent with the findings of Camilleri et al. (2017) (despite their use of only a 4-item UPE subscale) who, in addition, reported higher energy intakes in both men and women. In the present study, unconditional permission to eat also consistently showed small negative correlations with a range of subjective perceptions of eating behavior. If a conscious choice is made to allow oneself to enjoy a wide range of foods without guilt, it might be expected that the UPE subscale would not be related to negative perceptions of one's eating. However, it is possible that although people scoring higher on UPE may show some tendency to describe their eating as higher in fat/sugar/large portions, they may not criticize themselves for this or judge this as wrong. The recently reported finding of a strong positive correlation between UPE and the acceptance subscale of the Mindful Eating Scale (r=0.59, p < 0.01) lends support to this interpretation (Kerin, Webb, & Zimmer- Gembeck, 2018). The acceptance subscale includes reverse scored items such as “I criticize myself for the way I eat”. Being more accepting of one's perceived larger portions or higher fat/sugar intake is not necessarily maladaptive given that rigid restraint of food intake is associated with greater psychological distress, including disordered eating (Tylka, Calogero, & Danielsdottir, 2015). Qualitative research is needed among those scoring higher on unconditional permission to eat in order to explore their motivations for allowing themselves to eat whatever they desire rather than limiting certain foods, and to better understand how this eating style is paired with other aspects of IE. Unconditional permission to eat is strongly negatively correlated with restraint (Camilleri et al., 2015; Kerin et al., 2018), and the present study confirms that those scoring higher on UPE also perceive themselves to be consuming a higher calorie intake and larger portions. The observation that higher unconditional permission to eat was moderately associated with poorer diet quality scores but also with lower BMI may at first appear puzzling. However, inverse associations with BMI were reported in cross-sectional studies (Madden et al., 2012;

106

Table 4.8 Pearson correlations for the associations between Intuitive eating (IES-2) and participants’ perceptions of their diet and diet-related health consciousness (N = 5’238).

Men (n = 2’685) b Women (n = 2’553) b IES-2 IES-2 Total score UPE EPR RHSC B-FCC Total score UPE EPR RHSC B-FCC Attributesa Unhealthy - healthy n.s. -.21 .12 .12 .11 .22 -.16 .28 .21 .14 High in calories - low in calories -.09 -.25 n.s. n.s. -.14 n.s. -.26 .17 n.s. -.07 High in fat - low in fat n.s. -.28 .09 .07 -.08 n.s. -.25 .17 .07 n.s. High in sugar - low in sugar n.s. -.27 .13 n.s. -.09 .07 -.25 .21 .08 n.s. Large portions - small portions n.s. -.17 .09 .17 -.14 .15 -.10 .21 .21 -.10 Low self-control - high self-control .12 -.23 .17 .21 .10 .21 -.20 .30 .21 .08 Overeating [frequency per week] -.22 n.s. -.23 -.27 n.s. -.41 n.s. -.43 -.36 n.s. Diet-related health consciousness n.s. -.38 n.s. .11 .18 . n.s. -.34 .09 .16 .20 Note. Only significant correlations on a significance level of p £ .001 are displayed, n.s. = not significant (p > .001). aParticipants evaluated their own diet on a bipolar adjective scale ranging from -3 to +3 (e.g. unhealthy – healthy, high in calories – low in calories). The attributes calories, fat, sugar and portion size were recoded so that high values indicated low in fat, low in calories, low in sugar, small portions and high self-control capacities related to eating. bn varies due to missing values. IES-2 = Intuitive Eating Scale-2, UPE = Unconditional permission to eat, EPR= Eating for physical rather than emotional reasons, RHSC = Reliance on hunger and satiety cues, B-FCC = Body-food choice congruence.

INTUITIVE EATING AND FOOD CHOICES

Tylka & Kroon Van Diest, 2013; Tylka & Wilcox, 2006), and the findings of poorer diet quality suggest that these cross-sectional associations with BMI reflect that people with lower BMIs feel comfortable to allow themselves to eat whatever they desire, including foods higher in sugars, salt or fats. To our knowledge, no previous published study has examined the association between the body-food choice congruence dimension and food intake. It is noteworthy that B-FCC, the tendency to choose foods that honour health and body functioning, was largely unrelated to food intake. In women, B-FCC showed a small positive correlation with wholegrain consumption. Amongst men, the strongest, albeit small, correlation with B-FCC was found for physical activity level (r=0.24, p < 0.001), suggesting that B-FCC measure may reflect a tendency to choose higher food intakes to support physical performance and stamina in activity or sports, rather than choosing foods to support general wellbeing. In men, the small positive correlations of B-FCC with intakes of protein sources (meat and dairy) and associations with perceptions of higher calorie intake and larger portion size are consistent with this interpretation. These findings are somewhat less surprising when it is noted that two of the three items in the B-FCC subscale focus on body performance, energy and stamina. Only one item in this subscale focuses on the “desire to eat nutritious foods”; however, in the German translation (van Dyck et al., 2016) the word “nutritious” is translated into a word that conveys that foods are rich not only in micronutrients but also in energy, further emphasizing the “energy” focus of this subscale. The original conceptualization of this aspect of IE by Tribole and Resch (2012) refers to making “food choices that honour your health and tastebuds while making you feel well”. The diet-related health consciousness of people scoring higher on B- FCC may be related primarily to a consciousness of foods needed to support physical activity. Further research is needed to explore whether following our ‘body wisdom’ (i.e. following our body signals to guide us in our food choices) leads to patterns of food intake that honour health. IE intervention studies should explore the impact of teaching a more intuitive eating style not only on all four IE subscales, but on food consumption. Investigation of the latter should not be restricted to frequency of food intake since the present study suggests that people scoring higher on reliance on hunger and satiety signals perceive that they consume smaller portions. Thus, it is possible that intuitive eating may affect portion sizes to a greater degree than frequency of intake. Intervention studies which explore the impact of eating more intuitively on food intake are scarce, and one study has reported that dietary quality is improved in the short-term but not long-term (Carbonneau et al., 2017). However, a randomized trial in overweight and obese women that compared three different nondieting interventions teaching eating in response to hunger and satiety showed that significant improvements in a dietary quality score were maintained at 2 years (Hawley et al., 2008). In another randomized trial, a

108 INTUITIVE EATING AND FOOD CHOICES

weight-neutral IE intervention produced sustained improvements at 2 years in fruit and vegetable consumption and dietary risk scores (Mensinger et al., 2016). It is possible that a higher degree of awareness may be needed to notice more subtle body signals (i.e. I feel well after consistently eating plenty of fresh fruit and vegetables) than to know that one feels completely full as opposed to hungry. Many people may lack the body awareness to notice these more subtle body signals that, if followed, may be associated with patterns of food intake that honour wellbeing. In the absence of a high degree of body awareness, B-FCC scores might be expected to show little relationship with food intake. The aim of the present study was to utilize the published Frenchand German translations of the IES-2 in the Swiss population rather than to comprehensively test their psychometric properties as has already been reported by previous researchers (Camilleri et al., 2015; Carbonneau et al., 2016; Ruzanska, Alexandra, & Warschburger, 2017; van Dyck et al., 2016). It is common that questionnaires and their translations perform less well in other cultural contexts than those in which they were developed, which might change some of their psychometric properties (van Dyck et al., 2016). Therefore, some concerns with the IES-2 scale in the present study must be acknowledged. Two items cross-loaded on two factors. The item “I get mad at myself for eating something unhealthy” (reverse coded) is expected to load only on the UPE subscale, but also loaded on the EPR subscale, perhaps because it contains emotional aspects of self-criticism. The item “Most of the time I desire to eat nutritious foods” was expected to load on the B-FCC subscale, but loaded on the UPE subscale as well, which may reflect that the German translation of the word “nutritious” (“nahrhaft”) has an ambivalent meaning (i.e. energy dense as well as full of vitamins and minerals). Difficulties with the latter item were also observed in the validation study of the German translation of the IES-2 (van Dyck et al., 2016). The IES-2 total score had a good Cronbach's alpha (α=0.82). However, lower alpha values were observed for the UPE (α=0.65) and B-FCC (α=0.64) subscales compared with the other subscales (α=0.85). These observations are in line with results from other studies. For instance, in another large population study using the validated French version of the IES-2 a similarly low α-value of 0.67 for the UPE subscale and higher α′s for the EPR and RHSC subscales (B-FCC was not measured) were reported (Camilleri et al., 2017). In a Finish study of obese adults utilizing the original IE scale, alpha of .69 was observed for the UPE subscale (B-FCC was not included in this version of the IE scale) (Järvelä-Reijonen et al., 2016). Somewhat higher alpha reliabilities for the UPE subscale were reported in other studies. In a large New Zealand population study, the original English-language IES (Tylka & Wilcox, 2006) showed a Cronbach's Alpha of .77 for UPE (Madden et al., 2012), but also substantially lower than the 0.88-0.89 (IES) reported by Tylka (2006) in the US. An alpha value of 0.77 for the

109 INTUITIVE EATING AND FOOD CHOICES

UPE subscale was also observed among female college students in the US (Tylka & Kroon Van Diest, 2013). Again, one out of the six items of the UPE subscale crossloaded on a separate factor in the present study. Removing that item did not increase alpha reliabilities. Because of the lower alpha values, we cannot rule out that we might have underestimated the true effects. In addition, one item out of the three items of the B-FCC subscale loaded on the UPE subscale, which points to potential deficiencies in capturing the concept of body-food choice congruence. In general, the B-FCC subscale consists of only three items and short scales very often suffer from low alpha reliabilities. A refinement of the two subscales, the UPE and the B-FCC, in order to fully capture these concepts in different populations seems indicated. The study therefore highlights the challenges in translating scales for use in different cultures. For example, there may be cultural differences in the degree to which people allow themselves to experience the pleasure of eating desired foods. Considering sex differences, the present findings that men consistently score higher on the UPE and the EPR subscales, are consistent with previous results from studies among French adults (Camilleri et al., 2015), German adults (van Dyck et al., 2016) and US college students (Tylka & Kroon Van Diest, 2013). Therefore, in some respects men are more intuitive eaters than women. They seem to be more likely to allow themselves a variety of foods and the foods they desire, and are less likely to eat foods for the purpose of regulating emotional states. It is intriguing to note that in the present study and also studies conducted in Germany (van Dyck et al., 2016), Canada (Carbonneau et al., 2016), and New Zealand (personal communication, 2018), the UPE and B-FCC subscales were not correlated or comparatively weakly correlated. This contrasts with the finding of Tylka and Kroon Van Diest (2013) who reported these subscales to be inversely related (men: −0.34, women −0.23, both p < .001). Cronbach's alpha's for both UPE and B-FCC subscales were above 0.70 in several studies (van Dyck et al., 2016; Carbonneau et al., 2016; New Zealand, personal communication, 2018), so there must be other reasons for these studies not reproducing the finding of Tylka and Kroon Van Diest (2013). The inverse association observed in the US study (Tylka and Kroon Van Diest (2013) is understood in terms of IE involving a balance between allowing guilt-free enjoyment of those foods one desires and choosing foods that support one's wellbeing (Tribole & Resch, 2012). This balance may mean that in some circumstances (such as when there is a craving or desire for a certain food), an intuitive eater would allow themselves to enjoy this food. In the absence of such a desire, an intuitive eater would choose foods that support their health and are also enjoyable. Clearly further research is needed to explore interrelationships between these aspects of IE. It is noteworthy that two UPE items refer to the specific circumstance of craving or desiring a certain food (“If I am craving a certain

110 INTUITIVE EATING AND FOOD CHOICES

food, I allow myself to have it” and “I allow myself to eat what food I desire at the moment”); however, the other four items and all three B-FCC items refer to eating behavior or attitudes in general (Tylka & Kroon Van Diest, 2013). People may vary in the degree to which they experience such craving/desiring circumstances (which may be linked with intake of highly palatable foods). However, according to the notion of body wisdom, a person may also desire to eat foods that honour their health. An important area for future research is the investigation of the interplay of the different aspects of IE, possibly in combination with other factors. Such an interplay may be important in influencing dietary quality. For example, a person's susceptibility to the food environment may influence the extent to which they experience cravings or the desire for certain foods. The motivation for giving oneself unconditional permission to eat may also be important: an abandonment of dieting and a reduction in self- control, or a conscious decision not to fight occasional food-related desires while most of the time choosing to eat foods that honour health. Strengths of this study include its large random population sample, inclusion of men and women and analysis of all four IE dimensions. Most intuitive eating research has focused only on women (Schaefer & Magnuson, 2014; Van Dyke & Drinkwater, 2014). Furthermore, many studies using the IES were conducted with convenience samples of students (Tylka & Kroon Van Diest, 2013) or community groups (Carbonneau et al., 2016; Ruzanska et al., 2017) and recruited via print advertisment or online recruitment strategies (van Dyck et al., 2016). In contrast, the sample for the present study was based on random selection of men and women from the phone book since this recruitment strategy has the potential for greater representation of the general population. Nevertheless, the sample comprises a higher proportion of older adults and more highly educated persons, which may limit the generalizability of the findings to the entire Swiss population. The use of anonynimized self-administered questionnaires is likely to have limited social desirability bias, however, we cannot exclude the possibility that participants reported in a favourable manner. Moreover, a sFFQ was used to assess food intake, which does not entirely capture individual variations in portion sizes. However, most between-person variability in food intake is explained by frequency of intake rather than portion size (Willett, 2013). It is important to note that the research question in the present study is not dependent on precise estimates of nutrient or energy intake, but rather on food choices related to key components of dietary pattern. Of course, as with other dietary assessment methods, FFQ data can be subject to conscious or unconscious under- and over-reporting of true food intake (Maurer et al., 2006). This may apply particularly to restrained eaters (i.e. those with low unconditional permission scores) where intakes of high fat/sugar foods may be underestimated (Asbeck et al., 2002). The sFFQ used in the present study was based on the Nurse's Health

111 INTUITIVE EATING AND FOOD CHOICES

Study sFFQ (adapted to the Swiss context), the latter of which has been shown to be a valid measure of food intake patterns (Hu et al., 2016; Willett et al., 1985).

Conclusion Unconditional permission to eat was moderately correlated with poorer diet quality scores, whereas the other three aspects of intuitive eating showed only a few small positive correlations with food intake, including small positive associations of diet quality scores with EPR and RHSC in women. The lack of a clear link between B-FCC and food intake may reflect a lack of awareness of subtle body signals (i.e. of feeling well after consistently eating nutritious foods). French and German translations of the UPE and B-FCC aspects of IE may require refinement in order to enhance internal consistency. Further research is needed to determine the impact of IE interventions on food intake, and also to explore how different aspects of IE may interact to influence food intake.

References

Asbeck, I., Mast, M., Bierwag, A., Westenhöfer, J., Acheson, K. J., & Müller, M. J. (2002). Severe underreporting of energy intake in normal weight subjects: Use of an appropriate standard and relation to restrained eating. Public Health Nutrition, 5(5), 683–690. Bochud, M., Chatelan, A., Blanco, J.-M., & Beer-Borst, S. (2017). Anthropometric characteristics and indicators of eating and physical activity behaviors in the Swiss adult population. Retrieved from https://menuch.iumsp.ch/index.php/catalog/4/download/58. Camilleri, G. M., Mejean, C., Bellisle, F., Andreeva, V. A., Kesse-Guyot, E., Hercberg, S.,et al. (2017). Intuitive eating dimensions were differently associated with food intake in the general population-based NutriNet-Sante study. Journal of Nutrition, 147(1), 61–69. https://doi.org/10.3945/jn.116.234088. Camilleri, G. M., Méjean, C., Bellisle, F., Andreeva, V. A., Sautron, V., Hercberg, S., et al. (2015). Cross-cultural validity of the Intuitive Eating Scale-2. Psychometric evaluation in a sample of the general French population. Appetite, 84, 34–42. Carbonneau, E., Begin, C., Lemieux, S., Mongeau, L., Paquette, M. C., Turcotte, M., ... Provencher, V. (2017). A Health at Every Size intervention improves intuitive eating and diet quality in Canadian women. Clinical Nutrition, 36(3), 747–754. https://doi.org/10.1016/j.clnu.2016.06.008. Carbonneau, E., Carbonneau, N., Lamarche, B., Provencher, V., Bégin, C., Bradette-Laplante, M.,., et al. (2016). Validation of a French-Canadian adaptation of the intuitive eating scale-2 for the adult population. Appetite, 105, 37–45. Cohen, Jacob (1988). Statistical power analysis for the behavioral sciences. Hillsdale, NJ:Erlbaum(2nd.). Cole, Renee E., & Horacek, Tanya (2010). Effectiveness of the my body knows when intuitive- eating pilot program. American Journal of Health Behavior, 34(3), 286–297.

112 INTUITIVE EATING AND FOOD CHOICES

van Dyck, Z., Herbert, B. M., Happ, C., Kleveman, G. V., & Vogele, C. (2016). German version of the intuitive eating scale: Psychometric evaluation and application to an eating disordered population. Appetite, 105, 798–807. Federal Statistical Office (2016). Portrait of Switzerland. Results from the population census.Neuchâtel. Hawley, Greer, Horwath, Caroline, Gray, Andrew, Bradshaw, Alison, Katzer, Lisa, Joyce,J., et al. (2008). Sustainability of health and lifestyle improvements following a nondieting randomised trial in overweight women. Preventive Medicine, 47(6), 593–599. Herman, C Peter, & Polivy, Janet (1983). A boundary model for the regulation of eating. Psychiatric Annals, 13(12), 918–927. Hu, F. B., Satija, A., Rimm, E. B., Spiegelman, D., Sampson, L., Rosner, B., ... Willett, W. C. (2016). Diet assessment methods in the Nurses' health studies and contribution to evidence-based nutritional policies and guidelines. American Journal of Public Health, 106(9), 1567–1572. Järvelä-Reijonen, Elina, Karhunen, Leila, Sairanen, Essi, Rantala, Sanni, Laitinen, Jaana,P., et al. (2016). High perceived stress is associated with unfavorable eating behavior in overweight and obese Finns of working age. Appetite, 103, 249–258. Johansson, G., & Westerterp, K. R. (2008). Assessment of the physical activity level with two questions: Validation with doubly labeled water. International Journal of Obesity, 32(6), 1031–1033. Kerin, J. L., Webb, H. J., & Zimmer-Gembeck, M. J. (2018). Intuitive, mindful, emotional, external and regulatory eating behaviours and beliefs: An investigation of the core components. Appetite, 132, 139–146. Kourlaba, G., & Panagiotakos, D. B. (2009). Dietary quality indices and human health: A review. Maturitas, 62(1), 1–8. Lassale, Camille, Gunter, Marc J., Romaguera, Dora, Peelen, Linda M., Van der Schouw, Yvonne T., Beulens, Joline WJ., ... Huybrechts, Inge (2016). Diet quality scores and prediction of all-cause, cardiovascular and cancer mortality in a pan-european cohort study. PLoS One, 11(7), e0159025. Leblanc, V., Provencher, V., Begin, C., Corneau, L., Tremblay, A., & Lemieux, S. (2012). Impact of a health-at-every-size intervention on changes in dietary intakes and eating patterns in premenopausal overweight women: Results of a randomized trial. Clinical Nutrition, 31(4), 481–488. Madden, Clara EL., Leong, Sook Ling, Gray, Andrew, & Horwath, Caroline C. (2012). Eating in response to hunger and satiety signals is related to BMI in a nationwide sample of 1601 mid-age New Zealand women. Public Health Nutrition, 15(12), 2272–2279. Malik, Vasanti S., Willett, Walter C., & Hu, Frank B. (2013). Global obesity: Trends, risk factors and policy implications. Nature Reviews Endocrinology, 9(1), 13. Maurer, Jaclyn, Taren, Douglas, L., Teixeira, Pedro, J., et al. (2006). The psychosocial and behavioral characteristics related to energy misreporting. Nutrition Reviews, 64(2), 53– 66. Mensinger, J. L., Calogero, R. M., Stranges, S., & Tylka, T. L. (2016). A weight-neutral versus weight-loss approach for health promotion in women with high BMI: A randomized- controlled trial. Appetite, 105, 364–374. Mötteli, Sonja, & Dohle, Simone (2017). Egocentric social network correlates of physical activity. Journal of Sport and Health Science (in press).

113 INTUITIVE EATING AND FOOD CHOICES

Oliver, Georgina, Wardle, Jane, & Gibson, E Leigh (2000). Stress and food choice: A laboratory study. Psychosomatic Medicine, 62(6), 853–865. O'Connor, B. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer's MAP test. Behavior Research methods, Instruments, & Concepts, 32(3), 396–402. Ruzanska, Alexandra, Ulrike, & Warschburger, Petra (2017). Psychometric evaluation of the German version of the Intuitive Eating Scale-2 in a community sample. Appetite, 117, 126–134. Schaefer, Julie T., & Magnuson, Amy B. (2014). A review of interventions that promote eating by internal cues. Journal of the Academy of Nutrition and Dietetics, 114(5), 734–760. Smith, TeriSue, & Hawks, Steven R. (2013). Intuitive eating, diet composition, and the meaning of food in healthy weight promotion. American Journal of Health Education, 37(3), 130– 136. van Sonderen, Eric, Sanderman, Robbert, & Coyne, James C. (2013). Ineffectiveness of reverse wording of questionnaire items: let's learn from cows in the rain. PLoS One, 8(7) e68967-e68967. Swiss Federal Statistical Office (2016). Alter, Zivilstand, Staatsangehörigkeit [Age, marital status,nationality.https://www.bfs.admin.ch/bfs/de/home/statistiken/bevoelkerung/stan d-entwicklung/alter-zivilstand-staatsangehoerigkeit.html. Tribole, Evelyn, & Resch, Elyse (2012). Intuitive eating. Macmillan. Tylka, T. L. (2006). Development and psychometric evaluation of a measure of intuitive eating. Journal of Counseling Psychology, 53(2), 226. Tylka, T. L., Calogero, R. M., & Danielsdottir, S. (2015). Is intuitive eating the same as flexible dietary control? Their links to each other and well-being could provide an answer. Appetite, 95, 166–175. Tylka, T. L., & Kroon Van Diest, A. M. (2013). The intuitive eating scale–2: Item refinement and psychometric evaluation with college women and men. Journal of Counseling Psychology, 60(1), 137–153. Tylka, T. L., & Wilcox, J. A. (2006). Are intuitive eating and eating disorder symptomatology opposite poles of the same construct? Journal of Counseling Psychology, 53(4), 474. Van Dyke, Nina, & Drinkwater, Eric J. (2014). Review article relationships between intuitive eating and health indicators: Literature review. Public Health Nutrition, 17(8), 1757– 1766. Waijers, P. M., Feskens, E. J., & Ocke, M. C. (2007). A critical review of predefined diet quality scores. British Journal of Nutrition, 97(2), 219–231. https://doi.org/10.1017/S0007114507250421. Willett, W. (2013). Food frequency methods. Nutritional epidemiology (pp. 82). (4 ed.). New York: Oxford University Press. Willett, W., Sampson, Laura, Stampfer, Meir J., Rosner, Bernard, Bain, Christopher, Witschi, Jelia, ... Speizer, Frank E. (1985). Reproducibility and validity of a semiquantitative food frequency questionnaire. American Journal of Epidemiology, 122(1), 51–65. Wirt, A., & Collins, C. E. (2009). Diet quality–what is it and does it matter? Public Health Nutrition, 12(12), 2473–2492.

114

Chapter 5

Self-control as a moderator of the effect of hedonic hunger on overeating and snacking 5 Self-control as a moderator of the effect of hedonic hunger on overeating and snacking

A study based on the Swiss Food Panel 2.0 Caroline Horwatha, Désirée Hagmannb, Christina Hartmannb a Department of Human Nutrition, University of Otago, Dunedin, New Zealand b ETH Zurich

Manuscript submitted for publication: Horwath, C., Hagmann, D., Hartmann, C. The Power of Food: Self-control moderates the effect of hedonic hunger on overeating and snacking.

HEDONIC HUNGER AND SELF-CONTROL

Abstract

The purpose of the study was to explore whether self-control moderates the associations between hedonic hunger and snacking behavior as well as overeating tendencies. Data from the first wave of the Swiss Food Panel 2.0 study was analyzed (N = 5,238, from the German- and French-speaking part of Switzerland, 51% men). Measures included hedonic hunger assessed with the Power of Food Scale (PFS), dispositional self-control, overeating tendencies and aspects of snacking behavior (i.e. snack frequency, intake of high fat salty snack foods, intake of high sugar foods) assessed with a semiquantitative food frequency questionnaire. Higher scores on the PFS and lower self-control capacities were correlated with higher overeating frequency and higher snack food consumption. Four separate moderation analyses revealed that the effect of power of food on overeating and snacking behavior was significantly attenuated by self-control. Results of the present study indicate that people who are highly sensitive to living in an obesogenic environment but also have high levels of self-control exhibit less overeating and snacking behavior, including less frequent intake of unhealthy snacks, than those low in self-control. Consequently, self-control may prevent overeating and thus may serve as a protective factor that decreases the risk of becoming overweight in individuals who are highly sensitive to the food environment.

116 HEDONIC HUNGER AND SELF-CONTROL

5.1 Introduction

Food consumption in the absence of physiological hunger is common, and the drive or desire to consume food for pleasure in the absence of an actual need for calories has been referred to as hedonic (or pleasure-based) hunger (Lowe & Butryn, 2007). The ready availability of highly palatable energy-dense foods in the modern food environment may contribute to hedonic hunger (Lowe & Butryn, 2007). The Power of Food Scale (PFS, Lowe et al., 2009) was originally developed to assess sensitivity to living in an environment where highly palatable foods are omnipresent, or the preoccupation with such foods. People high in hedonic hunger have increased activation in visual processing regions of the brain in response to both images and words depicting highly palatable foods (Bullins et al., 2013; Rejeski et al., 2012), and are more likely to choose unhealthy snack foods when presented with a menu of options (Van Dillen & Andrade, 2016; Van Dillen, Papies, & Hofmann, 2013). Research suggests that higher PFS scores may predict preoccupation with palatable foods regardless of whether there is a caloric deficit or not (Rejeski et al., 2012). However, a recent review (Espel-Huynh, Muratore, & Lowe, 2018) concluded that this preoccupation with food does not necessarily translate into increased food intake since available studies show that PFS scores are not consistently associated with food consumption or Body Mass Index (BMI). This raises the question of whether hedonic hunger confers any adverse effects. Based on the limited research available, the review’s authors suggest a close link between hedonic hunger and feelings of a loss of control over eating or binge eating (Espel-Huynh et al., 2018). The observation that some individuals maintain a healthy BMI throughout life despite living in an obesogenic environment suggests that individuals are not universally affected by the food environment (Lowe et al., 2009). Evidence also suggests that lower inhibitory control of food intake may increase susceptibility to overeating in the presence of palatable food cues (Appelhans, 2009; Appelhans et al., 2011). Inhibitory control is an executive function needed to overrule impulsive or habitual reactions to food cues (Logan & Cowan, 1984), and has been found to be less effective in obese individuals than those who are lean (Nederkoorn, Guerrieri, Havermans, Roefs, & Jansen, 2009). If PFS scores tap into a predisposition to be preoccupied with highly palatable foods, preliminary research suggests that this predisposition may lead to increased food intake when it occurs in combination with lack of self-control or impulsivity (Espel-Huynh et al., 2018). For example, amongst a large sample of European adolescents, a children’s version of the PFS (C-PFS) was moderately associated with a higher intake of unhealthy snacks (Stok et al., 2015). Although self-regulatory competence moderated the impact of hedonic hunger on unhealthy snacking, the effect size for the interaction was very small. In contrast, in a study of 62 obese and overweight women, higher PFS scores predicted

117 HEDONIC HUNGER AND SELF-CONTROL

greater intake of palatable snack foods, but only for those with low levels of inhibitory control (Appelhans et al., 2011). It is also noteworthy that, in a study of mainly healthy weight university students, Nederkoorn et al. (2010) reported that individuals with a strong implicit preference for unhealthy snack foods together with low inhibitory control showed the greatest weight gain over one year.

Aim of the present study. Since hedonic hunger is a measure of a person’s appetitive responsiveness to rewarding properties of the food environment, it is plausible that hedonic hunger may be linked to frequent episodes of overeating and to snacking behavior. Snacking, or the consumption of food outside main meal occasions, may be more influenced than main meals by food-related cues (Cleobury & Tapper, 2014). Snacking and snack food consumption can contribute up to 30% of total energy intake in western societies (Mattes, 2018). The identification of factors that are linked to overeating and snacking behavior, such as hedonic hunger, and possible mediators of this relationship are of great relevance for the prevention and treatment of obesity. Therefore, the aim of the present study was to examine the relationship between hedonic hunger as measured by the PFS and overeating frequency as well as snacking behavior including the frequency of intake of high sugar foods and high-fat salty snacks. These outcome variables were chosen because of their potential contribution to the development of obesity (Duffey & Popkin, 2011). It was hypothesized that self-control would moderate the relationship between hedonic hunger and each of the food-related variables (Figure 5.1). In other words, self-control was expected to attenuate the effects of being sensitive to the food environment on food intake.

Self-control (Moderator)

Overeating frequency Snacking behaviour Power of Food – High sugar foods intake – High fat salty snack foods intake – Snacking frequency

Figure 5.1 Proposed model explaining overeating frequency, high sugar foods consumption, high fat salty snack consumption and overall snacking frequency by power of food scale (PFS) scores and self-control. A moderating effect of self-control on the relationship between PFS scores and the overeating and snacking variables was hypothesized.

118 HEDONIC HUNGER AND SELF-CONTROL

5.2 Methods

5.2.1 Participants and procedure

This study is based on data from the first wave of the Swiss Food Panel 2.0, a longitudinal study of the eating behavior of the Swiss population (for details on the Swiss Food Panel 2.0 see Hagmann, Siegrist, & Hartmann, 2018; Horwath, Hagmann, & Hartmann, 2018). Commencing in spring 2017, a randomly selected sample of inhabitants from the German- and French-speaking parts of Switzerland have completed an annual paper-and-pencil questionnaire that gathers information about food intake, socio-demographic characteristics, and various psychological constructs related to eating behavior. Only those constructs relevant to the present study are described below. The sample, after exclusion of pregnant women, consisted of 5,238 participants (51 % men) with a mean age of 56.5 years (SD = 17.3; range = 20-100) (Table 5.1).

Table 5.1 Sociodemographic characteristics, hedonic hunger, and self-control in males and females of the study sample (N = 5,238).

Mean (SD) or % Men 51% Age [years] 56.48 (17.25) German-speaking part of Switzerland 73.8% Highest educational level Primary and lower secondary school 5.4% Vocational education, middle school 38.1% Higher vocational education 36.9% University 16.2% No education/missing 3.5% Body Mass Index (BMI) Men 25.9 (4.0) Women 23.7 (4.4) Overweight (BMI ≥ 25-29.9) Men 42.1% Women 19.3% Obesity (BMI ≥ 30) Men 12.8% Women 9.0%

119 HEDONIC HUNGER AND SELF-CONTROL

Compared to the general Swiss population (Swiss Federal Statistical Office, 2016), adults aged 20-39 years were underrepresented (17% versus census 33%). The percentage of individuals with a higher educational level (higher professional school or university degree) was higher (53.9%) compared to the census (30%) (Federal Statistical Office, 2016, p. 25). The study was approved by the Ethics Committee of ETH Zurich (EK 2017-N-19).

5.2.2 Measures

Hedonic hunger. A German translation of the 15-item Power of Food Scale (PFS, Cappelleri et al., 2009; Lowe et al., 2009) was provided on request by Schultes et al. (2010) and translated into French by a professional translation service (Deman translations, Germany). The PFS (Cappelleri et al., 2009) consists of three subscales that measure different aspects of hedonic hunger which differ in the proximity of food: Food available (six items), Food Present (four items), and Food Tasted (five items). Examples of items include: “I find myself thinking about food even when I'm not physically hungry” and “If I see or smell a food I like, I get a powerful urge to have some.” Response options ranged from 1 (‘Do not agree at all’) to 5 (‘Totally agree’). The total score for PFS was calculated from the mean of all items (Cronbach’s a = .87 for both language versions). Table 5.2 shows mean value for Power of Food total scores.

Self-control. Dispositional self-control was assessed using the 13-item Brief Self-Control Scale of Tangney and Boone (2004). The scale consisted of 13 items, and the items included statements such as ‘Pleasure and fun sometimes keep me from getting work done (reversed item)’ or ‘I am able to work effectively toward long-term goals’. Participants were asked to rate items on a 5-point Likert scale from 1 (‘does not apply at all’) to 5 (‘applies completely’). For the German-speaking sample we used a German translation of the scale published by Bertrams & Dickhäuser (2009). For the French sample, the items of the German version were translated by a professional translating service (Deman translations, Germany) as there was no validated French translation available. For the analyses, a sum score was calculated (according to Tangney et al., 2004) but only for those participants who had answered all items of the scale (Table 5.2). Cronbach’s alpha was α =.78.

120

Table 5.2 Mean values (SD) and correlations for Power of Food Scale scores, overeating and snacking variables.

N1 Observed Mean SD Correlations range 1 2 3 4 5 1 Power of Food Scale [total score] 5018 1-5 2.43 0.68 1 2 Self-control 4890 17-65 46.92 7.64 -.43* 1

3 Overeating [frequency per week] 5208 1-5 1.75 0.80 .36* -.39* 1 4 Snacking [frequency per week] 5208 0-6.92 1.32 1.30 .17* -.17* .21* 1 5 High sugar foods [standard portions per 5038 0-154 7.59 8.01 .10* -.14* .16* .32* 1 week] 6 High fat salty snack foods [standard 5192 0-28 0.98 1.67 .11* -.15* .16* .23* .35* portions per week] Note. *p<.001, 1 N varies due to missing values

HEDONIC HUNGER AND SELF-CONTROL

Dietary assessment. Participants completed a semiquantitative food frequency questionnaire (sFFQ) which consisted of a subset of food items and nine response options from the Nurses’ Health Study sFFQ, a valid measure of food intake patterns (Hu et al., 2016; Willett et al., 1985). Usual consumption frequencies in the last year for 47 food items and food groups were assessed. As in the original Nurses’ Health Study questionnaire, standard portions for every food were provided and participants were asked to indicate how many standard portions of the food they usually consumed. The nine response options were recoded to reflect the number of standard portions consumed per week: 4 or more per day (coded as 28 portions per week), 3 per day (coded 21), 2 per day (coded 14), 1 per day (coded 7), 5-6 per week (coded 5.5), 2- 4 per week (coded 3), 1 per week (coded 1), 1-3 per month (coded 0.5) and seldom/never (coded 0). In the present study, only the consumption frequencies of high sugar foods and high-fat salty snack foods were analyzed. Consumption frequencies of the following six items were summed in order to estimate intake of high sugar foods per week: (1) candies/gummy bears, (2) cookies, (3) chocolate, e.g. pralines, bars, (4) chocolate spread, e.g. Nutella, (5) sweet pastries, e.g. cake, pie, (6) milky pudding, mousse au chocolate. Intake of high-fat salty snack foods intake was assessed with one item. People were asked to indicate how often they consume salty snacks such as salted nuts, chips, or crackers.

Overeating frequency. As an estimate of the participants’ perception of their frequency of overeating, they were asked: ‘How often do you eat too much or overeat?’. Five response options were available from ‘rarely/never’ (recoded 1) to ‘daily’ (recoded 5).

Snacking frequency. Respondents were asked how often they usually eat something between meals (snack) 1) in the morning, 2) in the afternoon, and 3) in the evening. The aim here was to investigate the frequency of between meal consumption of any food types. In addition to exploring general snacking at these specific times, respondents were also asked about their 4) frequency of intake of additional sweet and savoury snack foods. Participants answered all four questions on a five-point response scale ranging from daily (coded 360), 4- 6 times per week (coded 260), 1-3 times per week (coded 104), 1-3 times per month (coded 26) to less or never (coded 0). Values were summed and divided by 52 to create a snacking variable; higher values indicated a higher in between meal consumption frequency per week (Hartmann, Siegrist, & van der Horst, 2013).

Weight status. Based on the self-reported body weight (in kg) and body height (in m2), the participants’ Body Mass Index (BMI) was calculated (BMI = body weight in kg divided by body height in m2). Prevalence of overweight and obesity in males and females of the study sample are presented in Table 5.1.

122 HEDONIC HUNGER AND SELF-CONTROL

5.2.3 Statistical analysis

Pearson correlations between PFS, self-control, overeating frequency and snacking behavior were calculated. In order to test the proposed model (Figure 5.1) showing the mediating role of self-control on the relationship between PFS scores and snacking behavior as well as overeating frequency, four moderation analyses were conducted according to the method proposed by Hayes (2013). Snacking behavior was conceptualized as overall snacking frequency, consumption of high sugar foods as well as consumption of high fat salty snack foods. For all models, PFS scores (X) and the moderator variable self-control (M) were mean- centered before analyses by adding ‘center=1’ within the PROCESS command (Hayes, 2013, p. 230). Simple slopes for PFS scores on snacking behavior and overeating for people low (- 1 SD), moderate and high (+1 SD) in self-control were calculated separately.

5.3 Results

5.3.1 Descriptive statistics Mean values, standard deviations (SD) and Pearson correlation coefficients for PFS scores, overeating and snacking variables are displayed in Table 5.2. PFS showed a moderate inverse association with self-control (r=-.43, p<.001) and a moderate positive association with overeating frequency (r=.36, p<.001). The correlations between PFS and snacking behavior as conceptualized in terms of snacking frequency (r=.17, p<.001), high sugar foods (r=.10, p<.001) and high fat salty snack foods (r=.11, p<.001) intake, were statistically significant but much weaker. Self-control showed a moderate inverse association with overeating (r= -.39, p<.001). Overall snacking frequency was positively correlated with intake of high sugar foods (r=.32, p<.001) and high fat salty snack foods (r=.23, p<.001).

5.3.2 Conditional effects of power of food on snacking behavior (Moderation effects) Table 5.3 provides the unstandardized regression coefficients (b), standard errors (SE), t, and p-values as well as the lower (LLCI) and upper (ULCI) bound confidence intervals for b for the four models estimating the moderating effect of self-control on the relationship between PFS and overeating frequency as well as snacking behavior.

123 HEDONIC HUNGER AND SELF-CONTROL

Table 5.3 Results of four moderation analyses with self-control as moderator for the relationship between Power of Food and overeating frequency as well as snacking behavior.

b SE t p LLCI ULCI R2 Model 1 Overeating (N=4751) Constant 1.73 .01 156.28 <.001 1.71 1.75 .21 Self-control1 -.03 <.01 -19.46 <.001 -.03 -.03 Power of food1 .28 .02 16.19 <.001 .24 .31 Power of food x Self-control -.01 .00 -6.96 <.001 -.02 -.01 .008

Model 2 High fat salty snack foods intake (N=4739)

Constant .93 .02 37.99 <.001 .89 .98 .03 Self-control1 -.03 .00 -7.56 <.001 -.03 -.02 Power of food1 .13 .04 3.44 <.001 .06 .20 Power of food x Self-control -.02 <.01 -4.87 <.001 -.03 -.01 .005

Model 3 High sugar foods intake (N=4623) Constant 7.33 .12 58.95 <.001 7.09 7.57 .029 Self-control1 -.12 .02 -7.34 <.001 -.16 -.09 Power of food1 .46 .19 2.38 <.017 .08 .83 Power of food x Self-control -.11 .02 -5.60 <.001 -.15 -.07 .007

Model 4 Snacking frequency (N=4747) Constant 1.32 .02 65.93 <.001 1.27 1.34 .046 Self-control1 -.02 .00 -7.53 <.001 -.03 -.01 Power of food1 .22 .03 7.36 <.001 .16 .28 Power of food x Self-control -.01 .00 -4.00 <.001 -.02 -.01 .003

Note: 1 Variables were mean centered before analyses. Graphs for models 1–4 can be found in Figure 5.2 to Figure 5.5.

The overall Model 1 was significant (F(3, 4747)=416.57, p<.001) and explained 22% of the variance in overeating. All main effects and the interaction PFS x self-control were significant. Inclusion of the interaction term resulted in a significantly better model fit (Fchange(1, 2 4747)=48.40, p<.001, R change=.01). Separate simple slope analyses revealed that PFS was positively related to overeating frequency when self-control was low (one SD below the mean: b=.37, t(4747)=18.36, p<.001). This positive association was attenuated when self-control was

124 HEDONIC HUNGER AND SELF-CONTROL high (1 SD above the mean: b=.18, t(4747)=7.69, p<.001). Figure 5.2 illustrates the relationship between PFS and overeating. The differences in the slopes indicate the moderation effect for different levels of self-control. The results show that the association between PFS scores and overeating frequency is weaker in people with high dispositional self-control, whereas those with low self-control capabilities and high PFS scores reported a higher frequency of overeating. Thus, the hypothesis that the relationship between PFS and overeating frequency is attenuated by high levels of self-control was confirmed. The overall Model 2 was significant (F(3, 4735)=50.56, p<.001) and explained 4% of the variance in high fat salty snack consumption (Table 5.3). All main effects and the interaction term were significant. Inclusion of the interaction term improved the model fit (Fchange(1, 2 4735)=23,76, p<.001, R change=.005). Separate simple slope analyses revealed that PFS was positively related to high fat salty snack consumption when self-control was low (b=.28, t(4735)=6.20, p<.001). This positive association was no longer significant when self-control was high (b=-.02, t(4735)=-0.35, p=.724). The overall Model 3 was significant (F(3, 4619)=45.75, p<.001) and explained 4% of the variance in high sugar foods consumption (Table 5.3). The main effect of self-control and the interaction term was significant, however, the main effect of PFS was not. Inclusion of the 2 interaction term improved the model fit (Fchange(1, 4619)=31,31, p<.001, R change=.007). Separate simple slope analyses revealed that PFS was positively related to high sugar foods consumption frequency when self-control was low (b=1.32, t(4619)=5.80, p<.001). This positive association was no longer significant when self-control was average (b=.46, t(4619)=2.38, p=.017) and high (b=-.41, t(4619)=-1.55, p=.121). The overall Model 4 was significant (F(3, 4743)=76.95, p<.001) and explained 4.6% of the variance in overall snacking frequency (Table 5.3). All main effects and the interaction term were significant. Inclusion of the interaction term improved the model fit (Fchange(1,4743)=16.00, 2 p<.001, R change=.003). Separate simple slope analyses revealed that PFS was positively related to snacking frequency when self-control was low (b=.32, t(4743)=8.92, p<.001). This positive association was attenuated when self-control was high (b=.13, t(4743)=-3.00, p<.01). For model 2, model 3 and model 4 similar slope patterns were observed as in Figure 5.2. For a visual display of these moderation effects see Figure 5.3 to Figure 5.5.

125 HEDONIC HUNGER AND SELF-CONTROL

Overeating frequency per week

2.2

y=1.96+0.37*x

Self-control low (–1SD)

Self-control medium (mean = 0) 2.0 Self-control high (+1SD) y=1.73+0.28*x

1.8

y=1.51+0.18*x 1.6

1.4

1.2 –1SD 0 +1SD Low High Power of food Figure 5.2 Association between Power of Food and overeating frequency for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control (N=4751).

126 HEDONIC HUNGER AND SELF-CONTROL

High-fat salty snack foods intake per week

1.4

Self-control low (–1SD) 2.0 Self-control medium (mean = 0) 1.2 Self-control high (+1SD) y=1.13+0.28*x

1.0

y=0.93+0.13*x

0.8

y=0.74+0.02*x

0.6 –1SD 0 +1SD Low High Power of food Figure 5.3 Association between Power of Food and frequency of high fat salty snack foods intake for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control.

127 HEDONIC HUNGER AND SELF-CONTROL

High sugar foods intake per week

Self-control low (–1SD)

Self-control medium (mean = 0) 2.0 Self-control high (+1SD) 9

y=18.27+1.32*x

y=7.33+0.46*x

y=16.39+0.41*x

6 –1SD 0 +1SD Low High Power of food Figure 5.4 Association between Power of Food and frequency of high sugar foods intake for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control.

128 HEDONIC HUNGER AND SELF-CONTROL

Snacking frequency per week

1.8

Self-control low (–1SD)

Self-control medium (mean = 0) 2.0 y=1.46+0.32*x Self-control high (+1SD) 1.6

y=1.31+0.22*x

1.4

y=1.15+0.13*x

1.2

1.0 –1SD 0 +1SD Low High Power of food Figure 5.5 Association between Power of Food and frequency of snacking for high (+1SD), medium (mean=0) and low (-1SD) dispositional self-control.

129 HEDONIC HUNGER AND SELF-CONTROL

5.4 Discussion

In the modern food environment, highly palatable foods are readily available, enabling many opportunities for consuming food for pleasure in the absence of physical hunger. This study revealed a moderate association between an individual’s susceptibility to the food environment (conceptualized as hedonic hunger) and frequency of overeating, and small correlations between PFS scores and snacking behaviour. These results are consistent with previous research. For instance, Shuz et al (2015) showed that higher PFS scores predicted consumption of a higher number of snacks per day. Moreover, a positive correlation between PFS scores and ‘unhealthy’ snack consumption (assessed using a single item referring to the average daily number of snacks such as candy bars, crisps and fried snacks) was observed in a large study of European adolescents (Stok et al., 2015). In a representative community sample of men and women, PFS scores were associated with self-reported frequency of unhealthy snack intake; however, this relationship was no longer significant after controlling for habit strength (assessed using a self-report measure of the degree to which snacking was habitual or automatic) (Verhoeven, Adriaanse, Evers, & de Ridder, 2012). Based on previous evidence and the results from the present study utilizing a large adult sample, one may conclude that it is likely that hedonic hunger exerts negative effects on aspects of eating behavior. Further results of the present study also support the hypothesis that under conditions of high self-control, the effects of hedonic hunger on aspects of eating behavior (i.e. snacking behavior and overeating frequency) are attenuated. These results are also consistent with a small number of previous studies (Appelhans et al., 2011; Stok et al., 2015). Thus, findings suggest that self-control may serve as a protective factor since participants with high PFS scores in combination with high levels of self-control exhibit less susceptibility to overeating or unhealthy snacking. Present findings are also consistent with previous research on dispositional self-control and it’s association with improvements in eating behavior (Keller, Hartmann, & Siegrist, 2016). Small to medium sized effects were observed in the present study for the correlational associations between PFS, self-control and eating behavior. Considering the tested interaction term PFS x self-control, slightly higher effects were observed in the present study compared to the study by Stok et al. (2015) who tested whether self-regulatory competence moderated the impact of hedonic hunger on unhealthy snack intake (with snack intake assessed using a single item). However, when interpreting the results one should keep in mind that effects of self-control in the multifactorial domain of eating have been found to be lower in comparison with other life domains such as work (De Ridder & Lensvelt-Mulders, 2018).

130 HEDONIC HUNGER AND SELF-CONTROL

Strengths and limitations Key strengths of this study are its large sample of men and women, and use of a randomized recruitment strategy that has the potential for representation of the general population. Many previous PFS studies included primarily female participants (Espel-Huynh et al., 2018). Nevertheless, the proportion of older adults and highly educated persons was higher in the present sample than in the general Swiss population, which may limit the generalizability of findings. The use of anonynimized self-administered questionnaires is likely to have limited social desirability bias, however, we cannot exclude the possibility of underestimation of overeating and snacking frequency (Heitmann & Lissner, 1995). This may contribute to the small correlations between snacking and PFS scores in the present study. There may also be specific limitations in the assessment of snacking frequency given the lack of conceptual clarity about the definition of a snack as opposed to a meal (Gatenby, 1997). Foods are classified by consumers as snacks according to the time of the day they are consumed or the food type (Gatenby, 1997). Food consumed timewise close to the main meal might be interpreted as part of the main meal or as snack depending on the individual’s view. In the present study, people practicing a non-traditional meal-time pattern or a nibbling pattern could not be detected because the maximum number of between meal eating occasions per day that could be assessed with our question format was four. Thus, a ceiling effect cannot be ruled out.

Conclusion and Implications Results of the present study indicate that people who are highly sensitive to living in an obesogenic environment but also have high levels of self-control exhibit less overeating and snacking behavior, including less frequent intake of unhealthy snacks, than those low in self- control. Consequently, self-control may prevent overeating and thus serve as a protective factor that decreases the risk of becoming overweight in individuals who are highly sensitive to the food environment. Implementation of regulatory interventions to address the determinants of obesogenic environments are vital (Swinburn et al., 2011); however, the present results suggest that the development of self-control could be an important strategy to support improvements in eating behavior. Self-control can be developed and this has been shown to result in favorable food intake (Houben & Jansen, 2011; van Koningsbruggen, Veling, Stroebe, & Aarts, 2014). Our results suggest that self-control may lead both directly to less overeating and unhealthy snacking, in addition to attenuating the effects of being sensitive to the food environment on food intake.

131 HEDONIC HUNGER AND SELF-CONTROL

References

Appelhans, B. M. (2009). Neurobehavioral inhibition of reward-driven feeding: implications for dieting and obesity. Obesity, 17(4), 640-647. Appelhans, B. M., Woolf, K., Pagoto, S. L., Schneider, K. L., Whited, M. C., & Liebman, R. (2011). Inhibiting food reward: delay discounting, food reward sensitivity, and palatable food intake in overweight and obese women. Obesity, 19(11), 2175-2182. Bertrams, A., & Dickhäuser, O. (2009). Messung dispositioneller Selbstkontroll-Kapazität: eine deutsche Adaptation der Kurzform der Self-control Scale (SCS-KD) [Measuring dispositional self-control capacity. A German adaptation of the short form of the Self- control Scale (SCS-K-D)]. Diagnostica, 55(1), 2-10. Bullins, J., Laurienti, P. J., Morgan, A. R., Norris, J., Paolini, B., & Rejeski, W. J. (2013). Drive for consumption, craving, and connectivity in the visual cortex during the imagery of desired food. Frontiers in Aging Neuroscience, 5, 77. Cappelleri, J. C., Bushmakin, A. G., Gerber, R. A., Leidy, N. K., Sexton, C. C., Karlsson, J., & Lowe, M. R. (2009). Evaluating the Power of Food Scale in obese subjects and a general sample of individuals: development and measurement properties. International Journal of Obesity, 33(8), 913-922. doi:10.1038/ijo.2009.107 Cleobury, L., & Tapper, K. (2014). Reasons for eating ‘unhealthy’snacks in overweight and obese males and females. Journal of Human Nutrition and Dietetics, 27(4), 333-341. De Ridder, D. T., & Lensvelt-Mulders, G. (2018). Taking stock of self-control: A meta-analysis of how trait self-control relates to a wide range of behaviors. In Self-Regulation and Self-Control (pp. 221-274): Routledge. Duffey, K. J., & Popkin, B. M. (2011). Energy density, portion size, and eating occasions: contributions to increased energy intake in the United States, 1977–2006. PLoS medicine, 8(6), e1001050. Espel-Huynh, H., Muratore, A., & Lowe, M. (2018). A narrative review of the construct of hedonic hunger and its measurement by the Power of Food Scale. Obesity Science & Practice. Federal Statistical Office. (2016). Portrait of Switzerland. Results from the population census. Neuchâtel. Gatenby, S. J. (1997). Eating frequency: methodological and dietary aspects. British Journal of Nutrition, 77, 7-20. Hagmann, D., Siegrist, M., & Hartmann, C. (2018). Taxes, labels, or nudges? Public acceptance of various interventions designed to reduce sugar intake. Food Policy, 79, 156-165. doi:10.1016/j.foodpol.2018.06.008 Hartmann, C., Siegrist, M., & van der Horst, K. (2013). Snack frequency: associations with healthy and unhealthy food choices. Public Health Nutrition, 16(8), 1487-1496. doi:10.1017/S1368980012003771 Hayes, A. (2013). Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach. New York: The Guilford Press. Heitmann, B. L., & Lissner, L. (1995). Dietary underreporting by obese individuals--is it specific or non-specific? BMJ, 311(7011), 986-989. Horwath, C., Hagmann, D., & Hartmann, C. (2018). Intuitive eating and food intake in men and women: Results from the Swiss food panel study. Appetite, 135, 61-71. doi:10.1016/j.appet.2018.12.036

132 HEDONIC HUNGER AND SELF-CONTROL

Houben, K., & Jansen, A. (2011). Training inhibitory control. A recipe for resisting sweet temptations. Appetite, 56(2), 345-349. Hu, F. B., Satija, A., Rimm, E. B., Spiegelman, D., Sampson, L., Rosner, B., . . . Willett, W. C. (2016). Diet Assessment Methods in the Nurses' Health Studies and Contribution to Evidence-Based Nutritional Policies and Guidelines. American Journal of Public Health, 106(9), 1567-1572. doi:10.2105/AJPH.2016.303348 Keller, C., Hartmann, C., & Siegrist, M. (2016). The association between dispositional self- control and longitudinal changes in eating behaviors, diet quality, and BMI. Psychology & Health, 31(11), 1311-1327. doi:10.1080/08870446.2016.1204451 Logan, G. D., & Cowan, W. B. (1984). On the ability to inhibit thought and action: A theory of an act of control. Psychological Review, 91(3), 295. Lowe, M. R., & Butryn, M. L. (2007). Hedonic hunger: a new dimension of appetite? Physiology & Behavior, 91(4), 432-439. Lowe, M. R., Butryn, M. L., Didie, E. R., Annunziato, R. A., Thomas, J. G., Crerand, C. E., . . . Halford, J. (2009). The Power of Food Scale. A new measure of the psychological influence of the food environment. Appetite, 53(1), 114-118. doi:10.1016/j.appet.2009.05.016 Mattes, R. D. (2018). Snacking: A cause for concern. Physiology and Behavior, 193(Pt B), 279-283. doi:10.1016/j.physbeh.2018.02.010 Nederkoorn, C., Guerrieri, R., Havermans, R., Roefs, A., & Jansen, A. (2009). The interactive effect of hunger and impulsivity on food intake and purchase in a virtual supermarket. International Journal of Obesity, 33(8), 905. Nederkoorn, C., Houben, K., Hofmann, W., Roefs, A., & Jansen, A. (2010). Control yourself or just eat what you like? Weight gain over a year is predicted by an interactive effect of response inhibition and implicit preference for snack foods. Health Psychology, 29(4), 389. Rejeski, W. J., Burdette, J., Burns, M., Morgan, A. R., Hayasaka, S., Norris, J., . . . Laurienti, P. J. (2012). Power of food moderates food craving, perceived control, and brain networks following a short-term post-absorptive state in older adults. Appetite, 58(3), 806-813. Schultes, B., Ernst, B., Wilms, B., Thurnheer, M., & Hallschmid, M. (2010). Hedonic hunger is increased in severely obese patients and is reduced after gastric bypass surgery. American Journal of Clinical Nutrition, 92(2), 277-283. doi:10.3945/ajcn.2009.29007 Schuz, B., Schuz, N., & Ferguson, S. G. (2015). It's the power of food: individual differences in food cue responsiveness and snacking in everyday life. The International Journal of Behavioral Nutrition and Physical Activity, 12, 149. doi:10.1186/s12966-015-0312-3 Stok, F. M., De Vet, E., Wardle, J., Chu, M. T., De Wit, J., & De Ridder, D. T. (2015). Navigating the obesogenic environment: how psychological sensitivity to the food environment and self-regulatory competence are associated with adolescent unhealthy snacking. Eat Behav, 17, 19-22. doi:10.1016/j.eatbeh.2014.12.003 Swinburn, B. A., Sacks, G., Hall, K. D., McPherson, K., Finegood, D. T., Moodie, M. L., & Gortmaker, S. L. (2011). The global obesity pandemic: shaped by global drivers and local environments. The Lancet, 378(9793), 804-814. Swiss Federal Statistical Office. (2016). Alter, Zivilstand, Staatsangehörigkeit [Age, marital status, nationality]. https://www.bfs.admin.ch/bfs/de/home/statistiken/bevoelkerung/stand- entwicklung/alter-zivilstand-staatsangehoerigkeit.html

133 HEDONIC HUNGER AND SELF-CONTROL

Tangney, J. P., F., B. R., & Boone, A. L. (2004). High self-control predicts good adjustment, less pathology, better grades, and interpersonal success. Journal of Personality, 72(2), 271-324. Van Dillen, L. F., & Andrade, J. (2016). Derailing the streetcar named desire. Cognitive distractions reduce individual differences in cravings and unhealthy snacking in response to palatable food. Appetite, 96, 102-110. Van Dillen, L. F., Papies, E. K., & Hofmann, W. (2013). Turning a blind eye to temptation: How cognitive load can facilitate self-regulation. Journal of Personality and Social Psychology, 104(3), 427. van Koningsbruggen, G. M., Veling, H., Stroebe, W., & Aarts, H. (2014). Comparing two psychological interventions in reducing impulsive processes of eating behaviour: Effects on self-selected portion size. British Journal of Health Psychology, 19(4), 767- 782. Verhoeven, A. A., Adriaanse, M. A., Evers, C., & de Ridder, D. T. (2012). The power of habits: Unhealthy snacking behaviour is primarily predicted by habit strength. British Journal of Health Psychology, 17(4), 758-770. Willett, W. C., Sampson, L., Stampfer, M. J., Rosner, B., Bain, C., Witschi, J., . . . Speizer, F. E. (1985). Reproducibility and validity of a semiquantitative food frequency questionnaire. American Journal of Epidemiology, 122(1), 51-65.

134

Chapter 6

Public acceptance of interventions aimed at reducing sugar intake 6 Public acceptance of interventions aimed at reducing sugar intake

A study based on the Swiss Food Panel 2.0 Désirée Hagmann, Michael Siegrist, Christina Hartmann ETH Zürich

Published manuscript Hagmann, D., Siegrist, M., Hartmann, C. (2018). Taxes, labels, or nudges? Public acceptance of various interventions designed to reduce sugar intake. Food Policy 79, 156- 165. doi: 10.1016/j.foodpol.2018.06.008

PUBLIC ACCEPTANCE OF INTERVENTIONS

Abstract

This study investigated public acceptance of several specific government interventions to lower sugar intake in the population, using data from the first wave (2017) of a large survey (Swiss Food Panel 2.0) on eating behavior conducted in the German- and French-speaking regions of Switzerland (N=5238; 48.7% men). Acceptance varied considerably among different interventions; the least intrusive (i.e., a front-of-package label on products highlighting the sugar content and public health campaigns) garnered the most support, while more restrictive interventions (i.e., taxation, substitution with artificial sweeteners, and the reduction of portion sizes) generated higher resistance. Sugar consciousness and diet-related health consciousness were the strongest predictors of acceptance. Support was stronger among women, dieters, residents from the French-speaking areas of Switzerland, and people living in urban areas. Certain risk groups including overweight participants and those consuming higher amounts of sugar-sweetened beverages were more strongly opposed to these kinds of interventions. The different levels of acceptance must be taken into account by governments when planning interventions designed to reduce sugar intake.

136 PUBLIC ACCEPTANCE OF INTERVENTIONS

6.1 Introduction

Today, sugar is a firm component of our daily nutrition. It is loved and consumed in various forms: as a sweet snack, for sweetening coffee or tea, as a sugar-sweetened beverage (SSB) in addition to a meal or in between, or often almost unknowingly in the form of hidden sugars in many processed foods. Excessive consumption of dietary sugars, however, is a considerable public health problem. It is associated with poor diet quality (World Health Organization [WHO], 2015b), can contribute to weight gain and the development of obesity (Te Morenga, Mallard, & Mann, 2012), and increases the risk of various non-communicable diseases, including cardiovascular diseases, type 2 diabetes, and dental caries (e.g., Moynihan, 2016; Popkin & Hawkes, 2016). However, there is an ongoing controversy regarding the exact nature and strength of these associations (e.g., Stanhope, 2016). To date, no uniform definition of dietary sugars in foods and beverages is used, and also the recommended maximum sugar intake varies considerably (Wittekind & Walton, 2014). This limits comparability between studies (and nations) and makes it difficult to evaluate adherence to dietary recommendations for sugar (Newens & Walton, 2016). It might also partly explain the controversial findings about health risks associated with excessive sugar consumption (e.g., Te Morenga et al., 2012). The WHO (2015b) recently recommended that intake of free sugars (all mono- and disaccharides added during processing or preparation, as well as sugars naturally present in honey, syrups, fruit juices, and fruit juice concentrates) should make up less than 10% of the daily energy intake, or even better less than 5% (WHO, 2015b). However, most of the recent nationally representative dietary surveys rather reported the intake of added sugars, which only include those dietary sugars added during production, omitting the naturally occurring sugars included in the definition of free sugars (e.g., Erickson & Slavin, 2015). According to different reviews, added sugar intake in adults ranges from 7% to 11% of the daily energy intake in Europe (Azais-Braesco, Sluik, Maillot, Kok, & Moreno, 2017) and from 6.3% to 16.3% worldwide (Newens & Walton, 2016). Thus, a large part of the adult population seems to exceed the recommended intake of dietary sugars (Azais-Braesco et al., 2017).

6.1.1 Previous government interventions in the field of sugar consumption

Public health policymakers in many countries, wanting to change health behavior related to sugar intake, are considering interventions or have already taken action to reduce sugar consumption in the population (Popkin & Hawkes, 2016). The most commonly-implemented

137 PUBLIC ACCEPTANCE OF INTERVENTIONS measures to date are taxation, restrictions on marketing to children, reduction of availability, public awareness campaigns, and front-of-package labels (Popkin & Hawkes, 2016). In the last few years, many countries around the world, including Mexico, France, and Saudi Arabia, as well as several cities in the US have implemented taxes on products containing high proportions of added sugar, mainly on SSBs (Backholer, Blake, & Vandevijvere, 2017; Popkin & Hawkes, 2016). More countries (e.g., Ireland and the UK) are planning to introduce them in the near future (Backholer et al., 2017). Tax rates vary considerably between countries, going from 2% of the product price in the Navajo Nation (US) to 100% in Saudi Arabia (Backholer et al., 2017; Popkin & Hawkes, 2016). Most countries, however, impose a rate lower than the minimum 20% recommended by the WHO (2015a). So far, there is some evidence that the introduction of taxes may generate a decrease in per capita purchases of SSBs (e.g., in Mexico; Colchero, Popkin, Rivera, & Ng, 2016). However, while taxation might work in some societies, its effectiveness in other countries and its long-term effect on obesity rates require further investigation. Aside from these regulatory measures that focus on the individual, some countries have made efforts to restrict the activities of the food and beverage industry. In the UK, Ireland, and South Korea, for example, governments have restricted marketing activities for soft drinks, especially those targeted at children (e.g., Popkin & Hawkes, 2016). To date, no country has entirely forbidden marketing activities for foods and beverages high in sugar, and evidence for the efficacy of such restrictions is still missing (Popkin & Hawkes, 2016). Some governments also try to regulate the maximum amount of sugar allowed in certain processed foods and drinks. For instance, in Switzerland, several food manufacturers have pledged to reformulate yogurts and breakfast cereals to reduce their sugar content by 2018 (Swiss Federal Food Safety and Veterinary Office, 2017). But sugar reduction in processed foods may lead to changes in sensory attributes such as flavor and texture and thus reduce consumer liking and acceptance (e.g., Markey, Lovegrove, & Methven, 2015). Evidence from sensory studies on different food categories suggests that sugar reduction is acceptable for consumers if it happens gradually and does not go below certain food-specific thresholds (Chollet, Gille, Schmid, Walther, & Piccinali, 2013; Oliveira et al., 2016; Pineli et al., 2016). A relatively new approach to modifying health-related behaviors has attracted increasing attention. Nudging (from “nudge”, giving someone a soft push) means that the environmental conditions (or the choice architecture) are changed or optimized so that the performance of certain desireable behaviors becomes more likely (Thaler & Sunstein, 2008) or, in this case, so that healthier behaviors are facilitated. In the context of sugar consumption, one way to nudge consumers is to reduce the availability of unhealthy foods and beverages. For example, governments in the US and UK have made great efforts to reduce availability in schools (Hawkes et al., 2015; Popkin & Hawkes, 2016). There is some evidence that these

138 PUBLIC ACCEPTANCE OF INTERVENTIONS measures can reduce consumption inside the environment where availability is restricted, but not beyond it. The reduction of portion sizes is another example of nudging in this context. As many studies have demonstrated (Steenhuis & Poelman, 2017), people consume more calories when offered a larger portion (the so-called portion size effect). Some studies have shown that this effect also works the other way around, so that energy consumption can be reduced by decreasing portion size (e.g., Freedman & Brochado, 2010). Too large a reduction, however, might promote compensatory behaviors such as eating a second portion, adding other calorie-dense items (Gibson et al., 2017), or just adding additional sugar. Nonetheless, there is evidence that nudges can promote healthier food choices (Arno & Thomas, 2016). Since nudges can be implemented cost-effectively, they constitute a promising way to combat excessive sugar consumption (Arno & Thomas, 2016). Another widely used health-promoting intervention is the dissemination of information to consumers through health campaigns and front-of-package labels. For instance, a variety of labeling systems exists that highlight the level of sugar among other key nutrients (e.g., saturated fat and sodium) and caloric information (e.g., Hawley et al., 2013; Hersey, Wohlgenant, Arsenault, Kosa, & Muth, 2013). Research has indicated that nutritional labels can help consumers make healthier food choices (Hersey et al., 2013) if they are quickly and easy understandable (for a more detailed review see Hawley et al., 2013; Hersey et al., 2013). The traffic-light labeling system has shown itself to be among the most promising formats in terms of comprehensibility and consumer friendliness (Siegrist, Leins-Hess, & Keller, 2015; Temple & Fraser, 2014).

Acceptance of government interventions When governments have to decide on the implementation of an intervention, they mainly consider three aspects: effectiveness, cost, and public acceptance (Diepeveen, Ling, Suhrcke, Roland, & Marteau, 2013). Knowing about public support for policy measures in advance is relevant because a lack of it can evoke a strong response. For instance, in countries with a direct democracy, such as in Switzerland, the public can call for a referendum against policy measures; if this opportunity is not given, then disagreement can also be expressed indirectly by deselecting responsible politicians. Above all, if there is little acceptance, this might negatively impact the effectiveness of those measures (Diepeveen et al., 2013). A recent review by Diepeveen et al. (2013) showed that public acceptance depends on several factors. First, it depends on the nature of the targeted health behavior; with most public support for interventions against smoking compared to interventions trying to effect behavior changes related to diet, physical activity, or alcohol consumption. Second, acceptance depends on the type of intervention employed; less intrusive actions (e.g., health campaigns), those already put into practice (i.e., those that people are already familiar with), and those

139 PUBLIC ACCEPTANCE OF INTERVENTIONS directed at children are the best supported. Third, people are more likely to accept interventions when these are aimed at behaviors that they do not perform themselves (e.g., non-smokers are better acceptors of anti-smoking campaigns). Fourth, acceptance also depends on the gender and age of potential supporters; women and older people seem to be stronger supporters (Diepeveen et al., 2013). To the best of our knowledge, few studies have compared acceptance in a broad range of interventions in the field of sugar reduction (e.g., Petrescu, Hollands, Couturier, Ng, & Marteau, 2016). Also, few studies have investigated the association between dietary behavior and body weight status on the one hand and acceptance of those measures on the other hand, and the results of those studies are mixed (Diepeveen et al., 2013). Thus, our study has the following aims. First, it aims to compare the acceptances of distinct types of interventions, all aimed at reducing sugar intake, in a large Swiss sample. Second, it aims to find relevant variables that predict acceptance of interventions, particularly, to see whether population groups with an increased risk for adverse health effects (i.e., being overweight and reporting a higher intake of sugary foods and drinks) show rather support for or resistance against these policy measures. Third, the study aims to identify groups based on similarities in their acceptance of individual interventions and to compare them in the light of selected characteristics.

6.2 Methods

6.2.1 Participants This study used data from the first wave of the Swiss Food Panel 2.0, a longitudinal study of the dietary behavior of the Swiss population. This study follows the Swiss Food Panel conducted by the group from 2010 to 2014 (e.g., Hartmann, Dohle, & Siegrist, 2014). Data collection for the Swiss Food Panel 2.0 started in spring 2017. Mail surveys were sent to a random sample of residents in the German- and French-speaking parts of Switzerland. Most were randomly selected from the phone book, but some extra addresses of people aged 20– 30 were bought from an address company to increase the percentage of younger people, who are often not registered in the phone book. Questionnaires were returned by 5,781 people (response rate = 25.1%). Participants who did not record their gender or age, and those who completed less than 50% of the questionnaire were excluded from the analysis. Women who were pregnant at the time of the survey (n = 348) were also excluded, because Body Mass Index (BMI) was a factor in the analyses. The final sample consisted of 5,238 participants. Of these, the proportion of males was 48.7%, the mean age was 56.5 years (SD = 17.3, range 20–100), and 73.8% (n = 3,866) were German-speaking. Concerning age, the study sample is

140 PUBLIC ACCEPTANCE OF INTERVENTIONS not fully representative of the general Swiss population (Swiss Federal Statistical Office, 2016), as young adults aged 20–39 years are slightly underrepresented at 17% (census: 33.4%).

6.2.2 Questionnaire and individual measures Swiss Food Panel Questionnaire. The survey consists of a self-reporting paper-and-pencil questionnaire sent annually to the same participants. It includes a food frequency questionnaire (FFQ), questions about socio-demographic and lifestyle factors as well as various psychological variables related to eating and physical activity behaviors. For the present study, not all assessed constructs are relevant.

Dietary behavior assessment. To gain insights into the participants’ dietary habits, a self- reporting FFQ was used. For a large range of foods and beverages (including fruits and vegetables, dairy products, starchy foods, meat and fish, meat replacement products, sweets and savories, and soft and alcoholic drinks), participants were asked to indicate how often they consumed a predefined portion of them (e.g., a handful of fruit, 100–120 g of meat, or a glass / 200 mL of SSB). Portion sizes were defined according to the standard portion definition of the Swiss Society for Nutrition (2017). Frequency of consumption was measured with nine response options ranging from “4 or more per day” to “Rarely/never”. For further analyses, the number of weekly portions was calculated for each item.

Diet quality. The diet quality score was calculated using nine indicator variables: weekly portions of fruit (excluding fruit juice; Mdn = 7.00, IQR = 11.00), vegetables and salad (Mdn = 14.00, IQR = 11.00), sweet and salty snacks (Mdn = 6.75, IQR = 7.50), SSBs (54.6% reported “rarely/never”), meat and processed meat (Mdn = 7.00, IQR = 6.50), whole grain products (Mdn = 4.50 , IQR = 8.00), breaded or fried meat/fish (50.8% reported “rarely/never”), alcohol (Mdn = 3.00, IQR = 6.50), and the frequency of fast food consumption (82.6% reported “rarely/never”). These foods and beverages were selected because their high/low consumption has been shown to have either a positive or a negative impact on health (e.g., Malik, Willett, & Hu, 2013) and/or based on existing recommendations for a healthy, balanced diet (Swiss Society for Nutrition, 2017). A sample-based diet quality score was developed, which has already been used in a similar form in other research projects (e.g., Mötteli, Siegrist, & Keller, 2017). Since the variables were not normally distributed, the median was used as a cut-off value. For fruit, vegetables and salad, and whole-grain products, one point was assigned if consumption was at or above the median. For the other indicator variables, one point was assigned if the consumption frequency was at or below the median. Finally, the nine variables were summed up to a diet quality score. The higher the score, the closer the participant has

141 PUBLIC ACCEPTANCE OF INTERVENTIONS adhered to a healthy, balanced diet according to the recommendations of the Swiss Society for Nutrition (2017). The score ranged from 0 (unhealthy diet) to 9 (healthy diet) and was approximately normally distributed. In dietary studies, the use of diet quality or healthy eating indices based on cut-off values for different food and beverage groups is a common and established method (for an overview, see, e.g., Lassale et al., 2016). Moreover, combined with other risk factors, these indices have been shown to have good predictive abilities for health and mortality risk (Lassale et al., 2016).

Health related behaviors. Four items were used to assess diet-related health consciousness: “I think it is important to eat healthily.”, “My health is dependent on how and what I eat.”, “If one eats healthily, one gets ill less frequently.”, and “I am prepared to leave a lot, to eat as healthily as possible.” (adapted from Schifferstein & Oude Ophuis, 1998; published by Dohle, Hartmann, & Keller, 2014;(Hartmann, Dohle, & Siegrist, 2013). Self-control was measured with the Brief Self-Control Scale (Tangney, Baumeister, & Boone, 2004) which has a good internal consistency (Cronbach’s a = .78). Sugar consciousness, the degree of paying attention to sugar in one’s own diet, was assessed with one item (“How much attention do you pay to the sugar content in your diet?”). Answers could range from 1 “not at all” to 7 “very much”. Dieting status was assessed with the question “Are you or have you been on a diet during the past 12 months (e.g., eating less or avoiding certain foods)?”. The response scale was binary (yes/no).

Weight status / Body Mass Index (BMI). BMI was calculated by dividing the self-reported body weight (in kg) by the square of the body height (in m2). Respondents with a BMI ≥ 25 kg/m2 were classified as overweight. For males, the mean BMI in the sample was 25.9 kg/m2 (SD = 4.00), with a range of 16.1 kg/m2 to 62.5 kg/m2. For females, the mean BMI in the sample was 23.7 kg/m2 (SD = 4.42), with a range of 15.4 kg/m2 to 54.7 kg/m2. Overall, 54.9 % (n = 1,454) of the males and 28.3% (n = 710) of the females were overweight. This differs slightly from the average percentages of overweight adults in the Swiss population, (50% of the men and 32% of the women; Swiss Federal Statistical Office, 2012).

Sociodemographic characteristics. The study collected information about gender, age, and language region (German- or French-speaking part of Switzerland). The educational level was divided into three categories: low (primary and secondary school or no education), medium (vocational school), and high (college or university degree). The net monthly household income was also divided into three categories: low (up to 5,000 Swiss francs), medium (5,001 to 9,000 Swiss francs), and high (more than 9,000 Swiss francs). Participants had to indicate whether they lived in an urban, suburban, or rural area. For the analysis of rural-urban differences,

142 PUBLIC ACCEPTANCE OF INTERVENTIONS urban and suburban residencies were summed, and the total was compared with rural residence.

Acceptance of interventions to reduce sugar intake. Eight items (see Table 6.1) were used to assess participants’ acceptance of the types of intervention potentially available to governments trying to achieve sugar intake reduction in the population.

Table 6.1 Items used to assess acceptance of different interventions to reduce sugar intake (English translation).

Type of Factor Cronbach Item intervention loading ’s alpha 1 The availability of foods containing Reducing .71 .81 high levels of sugar should be availability reduced (e.g., no sales at vending (nudging) machines). 2 Taxation of sugar is an efficient Sugar tax .61 means of fighting against overweight. 3 The state should forbid Advertisement ban .77 advertisements for foods high in sugar. 4 The state should limit portion sizes of Reducing portion .76 foods high in sugar. sizes (nudging) 5 The sugar content should be clearly Sugar label .58 visible on a label on the package of (providing foods high in sugar. information + nudging) 6 The state should prescribe the Sugar reduction in .76 maximum amount of sugar that the products food industry may allow foods to contain. 7 Sugar in breakfast cereals and yogurt Substitution of .38 should be replaced by artificial sugar by artificial sweeteners. sweeteners 8 Population-based information Public health .62 campaigns promoting a reduction of campaigns sugar content in the diet should be conducted.

Note. Following Principal Component Analysis with Varimax rotation, all items loaded on a single component. Factor loadings of the items on this component and the Cronbach’s alpha of the scale are indicated.

143 PUBLIC ACCEPTANCE OF INTERVENTIONS

Various strategies were considered (see Table 6.1), including the introduction of a tax on sugar, nudges, and public health campaigns, etc. Respondents were provided with a short introductory statement: “Given the overweight problem (in our society), a lot of measures have been proposed (in order to tackle it).” They were then asked to indicate their acceptance of different specific measures (“How do you rate the following strategies to reduce sugar consumption in the Swiss population?”). Each item was rated on a 7-point response scale ranging from 1 (“do not agree at all”) to 7 (“fully agree”). The Cronbach’s a for the whole scale was .81. The principal component analysis (PCA) revealed a single component for general acceptance of interventions to reduce sugar consumption (see Table 6.1).

6.2.3 Data analysis The statistical analyses were done using IBM SPSS Statistics 23. A principal component analysis with Varimax rotation was conducted to test the factor structure of the eight acceptance items. Group comparisons for gender, dieting status, language (political) region, overweight status, and other variables were calculated using t tests for independent samples. A one-way analysis of variance (ANOVA) with repeated measures was applied to investigate whether different interventions achieved different levels of public acceptance. Moreover, hierarchical regression analysis was used to predict the general acceptance of interventions by socio-demographic and health/diet-related variables. The variable for general acceptance was built by calculating the mean value of all acceptance items. A hierarchical cluster analysis, using Ward’s method as the clustering algorithm and the squared Euclidean distance as the proximity measure, was conducted to identify groups of participants based on similarities in their acceptance of specific interventions. Initially, solutions of two to five clusters were generated. The final cluster solution was chosen based on the inspection of the agglomeration schedule and content considerations. To describe the characteristics and test differences between clusters, one-way ANOVA and c2 tests were conducted. Gabriel’s post hoc tests were used to make post hoc comparisons of the groups, as this method is appropriate when sample sizes are very different.

144 PUBLIC ACCEPTANCE OF INTERVENTIONS

6.3 Results

6.3.1 Sugar consumption and correlates Self-reported consumption frequencies of sweets (e.g., candies, sweet pastries, chocolate, and milk desserts), SSBs, and sugared breakfast cereals (e.g., Frosties and sugared granola) served as indicators of sugar consumption. Although males and females had similar weekly consumption frequencies of sweets (M = 7.57, SD = 8.55 and M = 7.62, SD = 7.43, respectively), t (5036) = -.19, p = .85, males consumed significantly more SSBs than females (M = 2.13, SD = 4.88 vs. M = 1.18, SD = 3.57), t (5148) = 7.93, p < .001. Males also consumed significantly more sugared breakfast cereals than females (M = 1.11, SD = 3.02 vs. M = 0.80, SD = 1.91), t (5125) = 4.37, p < .001 (see Figure 6.1).

Figure 6.1 Gender differences in the consumption frequencies of foods and beverages.

Age was negatively correlated with the consumption frequency of SSBs (r = -.17, p < .001) and sugared breakfast cereals (r = -.13, p < .001), but not with the consumption of sweets (r = -.02, p = .10). BMI correlated positively with SSB consumption (r = .07, p < .001), but not with the consumption of sweets (r = .01, p = .37) and sugared breakfast cereals (r = .001, p = .97; see also Table 6.2).

145

Table 6.2 Pearson correlations between socio-demographic, dietary, and health-related variables.

Variable 1 2 3 4 5 6 7 8 9 10 1 Age - 2 Body Mass Index .15*** - 3 Sweets (portions/week) -.02 .01 - 4 Sugar-sweetened beverages -.17*** .07*** .19*** - (glasses/week) 5 Sugar-sweetened breakfast -.13*** .001 .22*** .13*** - cereals (portions/week) 6 Diet quality .22*** -.16*** -.22*** -.34*** -.13*** - 7 General acceptance of .06*** -.04** -.06*** -.16*** -.03 .25*** - interventions 8 Sugar consciousness .10*** -.04** -.19*** -.25*** -.09*** .39*** .42*** - 9 Health consciousness .09*** -.12*** -.11*** -.19*** -.03** .36*** .34*** .43*** - 10 Self-control .17*** -.19*** -.14*** -.13*** -.09*** .26*** .02 .18*** .23*** - Note. ***p < .001, **p < .01.

PUBLIC ACCEPTANCE OF INTERVENTIONS

6.3.2 Acceptance of interventions to reduce sugar intake The mean acceptance ratings and 95 % confidence intervals for each intervention type are shown in Figure 6.2. The two most accepted interventions were the labeling of high-sugar products and public health campaigns. The three least accepted interventions were the substitution of sugars by artificial sweeteners, taxation, and reduction of portion sizes. One- way ANOVA with repeated measures showed that the level of acceptance differed significantly between interventions, F(6336,32048.72) = 2238.86, p < .001 (with Huynh-Feldt correction for sphericity is not given). Post hoc tests with Bonferroni correction showed significant differences between most strategies. No significant difference was observed between the acceptance of an advertisement ban (M = 4.23, SD = 2.10) and acceptance of reduction of sugar in products through food industry regulations (M = 4.21, SD = 2.21). The distributions of the acceptance scores for the single interventions are shown in Figure 6.3. Females showed higher support than males for most interventions (p < .001; for taxation, p < .01), except for substitution by artificial sweeteners, which was more strongly supported by men (p < .001). Residents of the French-speaking part of Switzerland were shown to be much more open to most of the suggested interventions than were those in the German-speaking part, except for the substitution of sugar by artificial sweeteners in breakfast cereals and yogurt (see Table 6.3). Findings on overweight were mixed: Overweight participants showed less support for interventions such as public health campaigns, reducing

Figure 6.2 Mean acceptance ratings for different interventions and 95% confidence intervals in the study sample (N=5,059).

147 PUBLIC ACCEPTANCE OF INTERVENTIONS availability, and reduction of sugar content in products (p £ .001) compared to those who were not overweight. By contrast, the overweight population was significantly more open to the substitution of sugar by artificial sweeteners (p < .001). A comparison of dieters and non- dieters showed that dieters seemed to accept all of the suggested interventions significantly more readily (p < .001) than participants who were not dieting at the time of the survey or during the past 12 months preceding the survey.

6.3.3 Predicting general acceptance of interventions

Correlations for general acceptance with age, BMI, diet quality, consumption frequencies of foods and beverages with high sugar content, and health-related variables—sugar consciousness, diet-related health consciousness, and self-control—are shown in Table 6.2. The results of the hierarchical regression analysis predicting general acceptance of sugar consumption interventions by socio-demographic and diet/health-related variables are shown in Table 6.4. The model, including all variables, was significant, F(13,4285) = 138.54, p < .001, and explained 26% of the variance. The strongest predictors of general acceptance were sugar consciousness and health consciousness: Participants with higher sugar consciousness and those with higher health consciousness showed stronger support for interventions. Language/political region also had a significant effect, with higher support among residents in the French-speaking part of Switzerland than in the German-speaking part. Higher self-control and higher SSB consumption frequency were associated with lower acceptance of interventions. Being overweight (BMI ≥ 25 kg/m2) was associated with lower support for interventions. Participants living in rural areas were less accepting than those in the cities and suburbs. The significance of age and gender differences disappeared once diet- and health- related variables were included in the model. Educational level and household income had no significant influence on the acceptance of interventions.

148

Figure 6.3 Distribution of the acceptance scores in the sample (N = 5,162) for each intervention type. The response scale ranged from 1 (“do not agree at all”) to 7 (“fully agree”) with higher values indicating higher acceptance.

Table 6.3 Mean acceptance of different interventions for participants from the German- and French-speaking parts.

Intervention type German French speaking speaking n = 3,823a n = 1,342a M SD M SD t(df) p

Taxation 3.32 2.08 3.60 2.12 -4.25 (5148) <.001 Reducing availability 4.69 2.05 5.25 1.95 -8.78 (5159) <.001 Advertisement ban 4.01 2.08 4.88 2.03 -13.25 (5141) <.001 Reducing portion sizes 3.31 2.05 4.45 2.12 -17.35 (5130) <.001 Sugar label 5.83 1.61 6.21 1.33 -7.68 (5160) <.001 Sugar reduction in products 3.88 2.18 5.19 1.99 -19.35 (5139) <.001 Substitution by artificial sweeteners 2.69 1.89 2.46 1.78 3.78 (5143) <.001 Public health campaigns 5.07 1.88 5.90 1.53 -14.61 (5148) <.001 a n varies due to missing values

PUBLIC ACCEPTANCE OF INTERVENTIONS

Table 6.4 Hierarchical regression analysis predicting general acceptance of interventions aimed at reducing sugar intake in the population (N = 4,216).

Step B SE B β Step 1 (Socio-demographic variables) Constant 3.63 .12 Gender (female = 1, male = 0) .35 .04 .14*** Age .01 .001 .09*** Income (medium = 1 vs. low = 0) .05 .05 .02 Income (high = 1 vs. low = 0) -.04 .05 -.01 Education (medium = 1 vs. low = 0) -.06 .08 -.02 Education (high = 1 vs. low = 0) -.04 .08 -.02 Region (French-speaking = 1, .62 .05 .21*** German-speaking = 0) Residence (rural = 1, -.17 .04 -.07*** urban or suburban = 0) Step 2 (Diet/health-related variables) Constant 2.70 .16 Gender (female = 1, male = 0) .06 .04 .02 Age .002 .001 .02 Income (medium = 1 vs. low = 0) .04 .04 .01 Income (high = 1 vs. low = 0) -.08 .05 -.03 Education (medium = 1 vs. low = 0) -.07 .07 -.03 Education (high = 1 vs. low = 0) -.12 .07 -.05 Region (French-speaking = 1, .52 .04 .17*** German-speaking = 0) Residence (rural = 1, -.13 .04 -.05*** urban or suburban = 0) Consumption of sweets .003 .002 .02 Sugar-sweetened beverage consumption -.02 .004 -.06*** Sugar consciousness .25 .01 .33*** Health consciousness .21 .02 .19*** Self-control -.01 .002 -.08*** Overweight (yes = 1, no = 0) -.10 .04 -.04** Note: R2 = .26. Effect size (Cohens f 2) for the whole model: f 2 = R2 / 1 - R2 = 0.26 / 1 - 0.26 = 0.35. According to (Cohen, 1988) an f 2 ³ 0.35 corresponds to a large effect.

***p < .001, **p < .01.

151 PUBLIC ACCEPTANCE OF INTERVENTIONS

6.3.4 Clusters based on acceptance of individual interventions

In the cluster analysis, four groups were identifiable according to their level of acceptance of specific interventions (see Table 6.5). The clusters were described and compared based on various characteristics summarized in Table 6.6. The General Supporters cluster (n = 1,881) showed the highest acceptance across all interventions compared to the other groups. The group consisted of 53.3% females, and had the highest number of urban residents (28.3%). This group also had the lowest SSB consumption as well as the highest diet quality and health consciousness. The General Opponents cluster (n = 318) was the smallest, with the largest proportion of males (67.9%), residents from the German-speaking part of Switzerland (90.6%), and rural residents (56.2%). It had the highest SSB consumption among all groups, the lowest diet quality and health consciousness, and the highest overweight proportion (50.3%). The Supporters of “Soft” Interventions Only cluster represented the largest group (n = 2,015). This group showed relatively strong support only for the less intrusive (“soft”) interventions—public health campaigns and sugar labeling. Slightly over half of the group (55.6%) were males, and a relatively large proportion of the group (55.5%) had a high educational level. The Opponents of Restrictive3 Interventions Only cluster —rejecting taxation and substitution of sugars by artificial sweeteners—was smaller (n = 845), and 53.5% of the group was female. No significant differences between clusters were found for self-control and sugared breakfast cereal consumption. For overweight status, c2(3) = 12.34, p = .006, and consumption of sweets, F(3,4871) = 5.29, p = .001, the overall difference between clusters was significant.

3 By “restrictive” we mean interventions that are hard to ignore and make alternative behavior difficult rather than how strongly they intervene or physically change the product, its availability, appearance or presence.

152

Table 6.5 Mean acceptance ratings of the four clusters for each intervention strategy.

Supporters of General Opponents of General “Soft” Supporters Restrictive Opponents Interventions n =1,881 Interventions n = 318 Only (37.2%) Only (6.3%) n = 2,015 n = 845 (39.8%) (16.7%) M SD M SD M SD M SD F (3,5055) Reducing availability 3.90a 2.02 6.12b 1.19 5.05c 1.72 2.46d 1.74 788.08*** Taxation 2.56a 1.70 5.22b 1.58 1.85c 0.90 1.93c 1.48 1476.19*** Advertisement ban 2.83a 1.64 5.70b 1.41 5.25c 1.47 1.77d 1.33 1580.22*** Reducing portion sizes 2.21a 1.39 5.15b 1.72 4.30c 1.82 1.36d 0.77 1411.64*** Sugar labeling 5.73a 1.45 6.56b 0.82 6.38b 0.89 2.17c 1.32 1377.14*** Sugar reduction in products 2.85a 1.82 5.57b 1.65 5.46b 1.42 1.46c 0.84 1344.56*** Substitution by artificial sweeteners 2.13a 1.47 3.22b 2.09 2.96b 1.91 1.43c 0.92 181.10*** Public health campaigns 4.87a 1.82 6.00b 1.38 5.73c 1.36 2.31d 1.66 572.08*** Note. One-way ANOVA was used to investigate differences in acceptance between clusters. Mean values within rows with different superscript letters are significantly different (using the Gabriel’s post hoc test, p < .001): a,b,c,d for significant differences between clusters. Higher values indicate higher acceptance. *** p < .001.

Table 6.6 Description of the four clusters by means of different characteristics.

Supporters of General Opponents of General “Soft” supporters Restrictive opponents Interventions Only n = 1,881e (37.2%) Interventions n = 318e n = 2,015e Only (6.3%) (39.8%) n = 845e (16.7%) % or SD % or M SD % or M SD % or M SD c2 (df) or M F(df1,df2) Socio-demographic characteristics Gender (% males) 55.6 46.7 46.5 67.9 c2(3) = 73.74*** Age (years) 55.4a,b 17.2 58.1c 17.1 53.4b 16.7 58.2a,c,d 17.9 F(3, 5055) = 18.90** Region (% from German- 81.8 66.6 67.0 90.6 c2(3) = 186.33*** speaking part) Household income (%) c2(6) = 26.49*** Low 26.5 - 32.5 - 29.0 - 37.0 - Middle 40.8 - 39.5 - 40.3 - 34.0 - High 32.7 - 28.0 - 30.7 - 29.0 - Education (%) c2(6) = 34.84*** Low 5.5 - 9.2 - 6.0 - 10.0 - Middle 39.0 - 37.4 - 38.7 - 45.8 - High 55.5 - 53.5 - 55.2 - 44.2 - Overweight 42.5 - 40.2 - 40.5 - 50.3 - c2(3) = 12.34** (% BMI ≥ 25) Residence c2(6) = 44.58*** Urban 25.8 - 28.3 - 22.0 - 21.6 - Suburban 32.3 - 33.5 - 34.7 - 22.2 - Rural 41.8 - 38.1 - 43.3 - 56.2 -

Dietary behavior Sweets (portions/week) 7.59a,b 7.32 7.21a 7.75 7.87a,b 6.99 9.08b 13.95 F(3, 4871) = 5.29** Sugar-sweetened beverages 1.91a 4.53 1.11b 3.45 1.55a,b 3.80 3.85c 7.34 F(3, 4976) = 39.76*** (glasses/week) Sugared breakfast cereals 1.01 2.40 0.88 2.36 0.91 2.26 1.20 4.35 F(3, 4955) = 1.97 ns Diet quality 4.74a 2.05 5.54b 1.98 5.01a 1.97 4.19c 1.89 F(3, 4861) = 70.82*** Health-related behavior Health consciousness 4.7a 1.1 5.3b 1.1 5.0c 1.1 4.2d 1.3 F(3, 4988) = 144.48*** Self-control 46.8 7.6 47.1 7.8 46.4 7.3 47.2 8.1 F(3, 4753) = 1.87ns Note. One-way ANOVA and c2 tests were used to investigate differences between clusters. Mean values within rows that have no superscript letter in common are significantly different (using Gabriel’s post hoc test, p < .001): a,b,c,d for significant differences between clusters. e n varies due to missing values. **p < .01, ***p < .001, ns = not significant.

PUBLIC ACCEPTANCE OF INTERVENTIONS

6.4 Discussion

Government interventions are important activities to encourage consumers to change their health-related behaviors (e.g., to make better nutritional choices), but maximum effectiveness depends heavily on sufficient acceptance by the public (Diepeveen et al., 2013). One of the aims of this study was to investigate and compare the acceptance of several specific interventions designed to reduce sugar intake. The results of the study are in line with those of others (Diepeveen et al., 2013) showing that the least intrusive interventions—public health campaigns and nutritional information on the packaging—received the highest support. Strongly regulatory interventions—such as taxing products containing high amounts of sugar or complete replacement of sugar by artificial sweeteners— were the least well accepted. It is possible that the idea of a sugar tax was unpopular simply because people would be annoyed at having to pay more for everyday foods and beverages. This is particularly true for consumers with a high consumption of these products, thus explaining the lower support among participants with higher SSB consumption. Another reason to reject the idea of a sugar tax might be that many people are not convinced that this measure is effective. In a study of Petrescu et al. (2016), perceived effectiveness was an important factor associated with the acceptance of policy measures. The observed low acceptance of artificial sweeteners might be due to some consumers’ perception of them as unhealthy or even as a health risk (e.g., Bearth, Cousin, & Siegrist, 2014); these respondents would therefore not consider this strategy to be an appropriate solution to the sugar issue. Reduction of portion sizes was also less acceptable than some other interventions, possibly because people fear proportionally higher prices per unit (i.e., that they would get less product for the same price). They might also be afraid that smaller portions could not satisfy their appetites in the same way. Reducing availability was better accepted than some other restrictions, probably because people assumed that there would be sufficient alternative sources where they could buy these products (e.g., cross-border shopping is easy in a small country like Switzerland, with short distances to neighboring countries). The sugar label was rated the most popular intervention in our study, although this does not necessarily mean that consumers would actively use it when making food decisions. A study conducted in European countries involving a variety of food products found that only around 17% of consumers looked for nutritional information on food products (Grunert, Fernandez-Celemin, Wills, Storcksdieck Genannt Bonsmann, & Nureeva, 2010). However, the proportion of consumers who did consult the labels was highest in the UK (27%), where public awareness of nutritional labeling is much more prominent than in the other countries surveyed. We identified several factors linked to the overall acceptance of interventions aimed at reducing sugar intake. Sugar consciousness and health consciousness were the strongest

156 PUBLIC ACCEPTANCE OF INTERVENTIONS among the investigated predictors, and both were associated with higher acceptance. People with strong manifestations of these characteristics might already be more motivated to take care of their sugar intake and their health in general. They might also be more convinced that the reduction of sugar intake is a good strategy to prevent negative health outcomes, so they would naturally support activities that promote these behaviors. That women expressed stronger support compared to men also fits well with previous findings that women are generally more health conscious (Hartmann, Siegrist, & van der Horst, 2013) and more interested in health topics (Kennedy & Funk, 2015). Overweight individuals and those consuming higher amounts of SSB were less supportive of the interventions. This shows that people tend to refuse any intervention that tries to restrict behaviors they want to engage in and is thus against their self-interest (e.g., Diepeveen et al., 2013). Conversely, support is stronger among groups that do not (or only weakly) engage in the targeted health behavior (Diepeveen et al., 2013), which was confirmed by our data. People tend to better accept policy measures if they affect the behavior of others; this applies across different behavioral domains (e.g., for tobacco and alcohol control interventions; Diepeveen et al., 2013). Apart from their own behavior not being limited through these measures, they might also somehow feel affected by the possible consequences of others’ unhealthy behavior (e.g., health care costs they would have to bear) and thus have a certain interest in supporting preventive interventions. However, the results suggest the importance of taking negative attitudes toward interventions into account, especially in risk groups, since these are exactly the population groups at whom these interventions are aimed. The findings concerning self-control go somewhat in the opposite direction; lower self- control was associated with higher acceptance of interventions. As individuals with higher self- control seem to be better able to successfully control their eating behaviors and weight over time (Keller, Hartmann, & Siegrist, 2016), they might consider government interventions unnecessary. By contrast, individuals lacking self-control might struggle with weight control and therefore perceive interventions as helpful. A cultural effect related to the sugar issue appears to exist in Switzerland, as residents in the French-speaking part showed more support for the suggested interventions than those in the German-speaking part (except for the substitution by artificial sweeteners). This is a classic example of the traditional divergences in (socio)political issues between these two parts of Switzerland (Büchi, 2000; Etter, Herzen, Grossglauser, & Thiran, 2014; Eugster, Lalive, Steinhauer, & Zweimüller, 2011). The French-speaking Swiss place much higher demands upon the government than German-speaking Swiss, expecting the government to take responsibility and play an active role in social issues (Eugster et al., 2011). This stronger belief in paternalism might be one explanation for the stronger support for government interventions.

157 PUBLIC ACCEPTANCE OF INTERVENTIONS

Another explanation for this cultural effect could be an orientation to different neighboring countries, each with its own health policy. For example, SSB taxation was introduced in France in 2012 and is well accepted in the population (Julia, Mejean, Vicari, Peneau, & Hercberg, 2015), while the topic is still under debate in Germany. This might influence the attitudes of residents in the bordering regions of Switzerland. In addition to the cultural effect, an urban-rural contrast seems to exist, as people living in urban or suburban areas considered interventions more acceptable than did the rural population. This might reflect a more conservative attitude in rural communities, where health interventions aimed at changing health behavior are more likely to be rejected than in cities. Among the clusters, the proportion of rural residents in the group of the General Opponents cluster was by far the highest (56.2%). The present study used the food frequency questionnaire method, which does not provide a comprehensive dietary assessment. Thus, it was not possible to control for underreporting of energy-dense foods, a common phenomenon in dietary surveys using self- report methods (Willett, 2013). However, the FFQ is a suitable tool for measuring dietary behavior on a population level (Cade, Thompson, Burley, & Warm, 2002). This study did not consider all sources of sugar (e.g., hidden sugars in processed foods), so the participants’ actual sugar intake is probably higher than estimated. The study found a higher consumption frequency of SSBs and sugared breakfast cereals among men. As we could not compare the relative contributions of sugar to the daily energy intake, this result might also simply reflect that men in general have higher caloric intakes than women do (Azais-Braesco et al., 2017). In our study, we asked the participants’ opinions in a rather general way. Future studies could focus on acceptance using specific scenarios for each policy measure to find out how acceptance varies when certain parameters are consciously changed, such as asking for the acceptance of different tax rates or tax rates expressed in monetary terms. Another possibility is to provide different scenarios of how products would change in terms of taste, calorie content, and effects on health, among others, if artificial sweeteners were used instead of sugar.

6.4.1 Conclusion and policy implications Our study provided insights into the acceptance of various policy measures that try to move the population’s sugar consumption habits toward a healthier direction. It demonstrated that the level of public acceptance varies considerably depending on the type of intervention suggested, with the most support for the least intrusive interventions—package labeling and public health campaigns—and higher resistance to the more restrictive interventions—

158 PUBLIC ACCEPTANCE OF INTERVENTIONS taxation, substitution by artificial sweeteners, and reduction of portion sizes. A high percentage (93.7%) of the study participants were in favor of the less intrusive or “soft” interventions. Previous research has suggested that once an intervention is implemented, acceptance tends to increase (Diepeveen et al., 2013). However, if there is strong opposition to a certain intervention, it might also provoke detrimental behavior (e.g., cross-boarder shopping in the case of taxation), which might have negative economic and social effects without solving the original problem. People with increased health risks, namely overweight individuals and those consuming higher amounts of SSBs, who would benefit the most from the suggested measures most strongly refuse them. This is an important finding that policy makers should be aware of when planning new interventions. Policy makers should also think of ways to better reach these population groups to initiate the intended behavior change. Furthermore, for the conceptualization of interventions, it is necessary to consider also differences in acceptance between gender, age groups, and regions (cultures) because each reflects a different readiness for specific interventions. In Switzerland, for example, new measures such as taxes on SSBs could be introduced in the French-speaking regions first, as people there are more in favor of this kind of intervention. If residents in other regions see that this measure is successful, then their opinion may swing favorably, so subsequent implementation might become easier. In the present study, only 37.2% of the participants said they would support the implementation of a national sugar tax. Future studies should investigate how the level of acceptance varies depending on the tax rate as well as on the purpose for which the revenues will be spent. Real-life examples show that the latter varies; many governments want to spend revenues for health purposes more or less related to sugar consumption, while others place the money into a general fund not necessarily related to health and prevention (Backholer et al., 2017). For instance, in Albany, California, the revenues from the SSB tax go into a general city fund; in Cook County, Illinois, the primary purpose of the tax was to alleviate budget deficits (later, however, the tax was repealed); and in Boulder, Colorado, revenues are primarily planned for supporting health, wellness, and chronic disease programmes (Backholer et al., 2017). As sugar consciousness and health consciousness have been shown to be relatively strong predictors of acceptance, it is particularly important to sensitize people to pay more attention to the amounts and types of sugars they consume. Most previous studies (including the present) and the interventions considered have focused mainly on the consumption of products with obviously high sugar contents, such as SSBs and sweet-tasting foods. Future research should also consider the intake of hidden sugar, as this also contributes to overall

159 PUBLIC ACCEPTANCE OF INTERVENTIONS sugar consumption. Possibly, people may be more willing to accept interventions focusing on reducing hidden sugar, since these aim to protect the population against a risk they do not consciously choose to take. Such interventions could focus on the reduction of hidden sugars in products (regulations), better declaration (labeling) on packages, or awareness campaigns. Further studies should also concentrate on obtaining more insights into the acceptance of interventions among younger adults and adolescents, as sugar consumption is particularly high in these age groups (Newens & Walton, 2016), creating a great need for appropriate interventions at this age. A previous experimental study suggested that some interventions might reduce younger people’s preference for and probability of purchasing SSB (Bollard, Maubach, Walker, & Ni Mhurchu, 2016). The effects of specific interventions aimed at young people in real-life settings are worth investigating further. Moreover, an interesting research question would be whether young adults and adolescents are more sensitive to certain interventions such as taxation. Finally, our study demonstrated that the provision of information about the sugar content using front-of-package labels is the clear favorite among the suggested interventions across all groups. Besides its generally high acceptance among consumers, labels can help people make healthier food choices (Hawley et al., 2013) and might be a motivation for the food and beverage industry to reformulate its products to get better ratings (Kanter, Vanderlee, & Vandevijvere, 2018). Further studies are needed to identify which of the existing labeling formats is the most advantageous in terms of helping consumer groups with different needs and preconditions (e.g., in terms of literacy) make healthier, informed food choices.

References

Arno, A., & Thomas, S. (2016). The efficacy of nudge theory strategies in influencing adult dietary behaviour: a systematic review and meta-analysis. BMC Public Health, 16, 676. doi:10.1186/s12889-016-3272-x Azais-Braesco, V., Sluik, D., Maillot, M., Kok, F., & Moreno, L. A. (2017). A review of total & added sugar intakes and dietary sources in Europe. Nutr J, 16(1), 6. doi:10.1186/s12937-016-0225-2 Backholer, K., Blake, M., & Vandevijvere, S. (2017). Sugar-sweetened beverage taxation: an update on the year that was 2017. Public Health Nutrition, 20(18), 3219-3224. doi:10.1017/S1368980017003329 Bearth, A., Cousin, M.-E., & Siegrist, M. (2014). The consumer’s perception of artificial food additives: Influences on acceptance, risk and benefit perceptions. Food Quality and Preference, 38, 14-23. doi:0.1016/j.foodqual.2014.05.008 Bollard, T., Maubach, N., Walker, N., & Ni Mhurchu, C. (2016). Effects of plain packaging, warning labels, and taxes on young people's predicted sugar-sweetened beverage

160 PUBLIC ACCEPTANCE OF INTERVENTIONS

preferences: an experimental study. The International Journal of Behavioral Nutrition and Physical Activity, 13(1), 95. doi:10.1186/s12966-016-0421-7 Büchi, C. (2000). “Röstigraben”. Das Verhältnis zwischen deutscher und französischer Schweiz. Geschichte und Perspektiven. [The “Rösti ditch”: The relationship between the German and French speaking part of Switzerland. History and perspectives]. Zürich: NZZ Verlag. Cade, J., Thompson, R., Burley, V., & Warm, D. (2002). Development, validation and utilisation of food-frequency questionnaires - a review. Public Health Nutrition, 5(4), 567-587. doi:10.1079/PHN2001318 Chollet, M., Gille, D., Schmid, A., Walther, B., & Piccinali, P. (2013). Acceptance of sugar reduction in flavored yogurt. Journal of Dairy Science, 96(9), 5501-5511. doi:10.3168/jds.2013-6610 Cohen, J. (1988). Statistical Power analysis for the behavioral sciences. New York: Psychology Press. Colchero, M. A., Popkin, B. M., Rivera, J. A., & Ng, S. W. (2016). Beverage purchases from stores in Mexico under the excise tax on sugar sweetened beverages: observational study. BMJ, 352, h6704. doi:10.1136/bmj.h6704 Diepeveen, S., Ling, T., Suhrcke, M., Roland, M., & Marteau, T. M. (2013). Public acceptability of government intervention to change health-related behaviours: a systematic review and narrative synthesis. BMC Public Health, 13. doi:doi:10.1186/1471-2458-13-756 Dohle, S., Hartmann, C., & Keller, C. (2014). Physical activity as a moderator of the association between emotional eating and BMI: evidence from the Swiss Food Panel. Psychology & Health, 29(9), 1062-1080. doi:10.1080/08870446.2014.909042 Erickson, J., & Slavin, J. (2015). Total, added, and free sugars: are restrictive guidelines science-based or achievable? Nutrients, 7(4), 2866-2878. doi:10.3390/nu7042866 Etter, V., Herzen, J., Grossglauser, M., & Thiran, P. (2014). Mining democracy. Paper presented at the ACM Conference on online social networks 2014 (COSN’14), Dublin, Ireland. Eugster, B., Lalive, R., Steinhauer, A., & Zweimüller, J. (2011). The demands for social insurance: Does culture matter? The Economic Journal, 121, F413-F448. doi:10.1111/j.1468-0297.2011.02479.x. Freedman, M. R., & Brochado, C. (2010). Reducing portion size reduces food intake and plate waste. Obesity (Silver Spring), 18(9), 1864-1866. doi:10.1038/oby.2009.480 Gibson, S., Ashwell, M., Arthur, J., Bagley, L., Lennox, A., Rogers, P. J., & Stanner, S. (2017). What can the food and drink industry do to help achieve the 5% free sugars goal? Perspect Public Health, 137(4), 237-247. doi:10.1177/1757913917703419 Grunert, K. G., Fernandez-Celemin, L., Wills, J. M., Storcksdieck Genannt Bonsmann, S., & Nureeva, L. (2010). Use and understanding of nutrition information on food labels in six European countries. Z Gesundh Wiss, 18(3), 261-277. doi:10.1007/s10389-009- 0307-0 Hartmann, C., Dohle, S., & Siegrist, M. (2013). Importance of cooking skills for balanced food choices. Appetite, 65, 125-131. doi:10.1016/j.appet.2013.01.016 Hartmann, C., Dohle, S., & Siegrist, M. (2014). Time for change? Food choices in the transition to cohabitation and parenthood. Public Health Nutrition, 17(12), 2730-2739. doi:10.1017/S1368980013003297

161 PUBLIC ACCEPTANCE OF INTERVENTIONS

Hartmann, C., Siegrist, M., & van der Horst, K. (2013). Snack frequency: associations with healthy and unhealthy food choices. Public Health Nutrition, 16(8), 1487-1496. doi:10.1017/S1368980012003771 Hawkes, C., Smith, T. G., Jewell, J., Wardle, J., Hammond, R. A., Friel, S., . . . Kain, J. (2015). Smart food policies for obesity prevention. The Lancet, 385(9985), 2410-2421. doi:10.1016/s0140-6736(14)61745-1 Hawley, K. L., Roberto, C. A., Bragg, M. A., Liu, P. J., Schwartz, M. B., & Brownell, K. D. (2013). The science on front-of-package food labels. Public Health Nutrition, 16(3), 430-439. doi:10.1017/S1368980012000754 Hersey, J. C., Wohlgenant, K. C., Arsenault, J. E., Kosa, K. M., & Muth, M. K. (2013). Effects of front-of-package and shelf nutrition labeling systems on consumers. Nutrition Reviews, 71(1), 1-14. doi:10.1111/nure.12000 Julia, C., Mejean, C., Vicari, F., Peneau, S., & Hercberg, S. (2015). Public perception and characteristics related to acceptance of the sugar-sweetened beverage taxation launched in France in 2012. Public Health Nutrition, 18(14), 2679-2688. doi:10.1017/S1368980014003231 Kanter, R., Vanderlee, L., & Vandevijvere, S. (2018). Front-of-package nutrition labelling policy: global progress and future directions. Public Health Nutrition, 1-10. doi:10.1017/S1368980018000010 Keller, C., Hartmann, C., & Siegrist, M. (2016). The association between dispositional self- control and longitudinal changes in eating behaviors, diet quality, and BMI. Psychology & Health, 31(11), 1311-1327. doi:10.1080/08870446.2016.1204451 Kennedy, B., & Funk, C. (2015). Public interest in science and health linked to gender, age and personality. Retrieved from Lassale, C., Gunter, M. J., Romaguera, D., Peelen, L. M., Van der Schouw, Y. T., Beulens, J. W., . . . Tzoulaki, I. (2016). Diet Quality Scores and Prediction of All-Cause, Cardiovascular and Cancer Mortality in a Pan-European Cohort Study. PloS One, 11(7), e0159025. doi:10.1371/journal.pone.0159025 Malik, V. S., Willett, W. C., & Hu, F. B. (2013). Global obesity: trends, risk factors and policy implications. Nature Reviews: Endocrinology, 9(1), 13-27. doi:10.1038/nrendo.2012.199 Markey, O., Lovegrove, J. A., & Methven, L. (2015). Sensory profiles and consumer acceptability of a range of sugar-reduced products on the UK market. Food Research International, 72, 133-139. doi:10.1016/j.foodres.2015.03.012 Mötteli, S., Siegrist, M., & Keller, C. (2017). Women's social eating environment and its associations with dietary behavior and weight management. Appetite, 110, 86-93. doi:10.1016/j.appet.2016.12.014 Moynihan, P. (2016). Sugars and Dental Caries: Evidence for Setting a Recommended Threshold for Intake. Advances in Nutrition, 7(1), 149-156. doi:10.3945/an.115.009365 Newens, K. J., & Walton, J. (2016). A review of sugar consumption from nationally representative dietary surveys across the world. Journal of Human Nutrition and Dietetics, 29(2), 225-240. doi:10.1111/jhn.12338 Oliveira, D., Reis, F., Deliza, R., Rosenthal, A., Gimenez, A., & Ares, G. (2016). Difference thresholds for added sugar in chocolate-flavoured milk: Recommendations for gradual sugar reduction. Food Research International (Ottawa, Ont.), 89(Pt 1), 448-453. doi:10.1016/j.foodres.2016.08.019

162 PUBLIC ACCEPTANCE OF INTERVENTIONS

Petrescu, D. C., Hollands, G. J., Couturier, D. L., Ng, Y. L., & Marteau, T. M. (2016). Public Acceptability in the UK and USA of Nudging to Reduce Obesity: The Example of Reducing Sugar-Sweetened Beverages Consumption. PLoS One, 11(6), 18. doi:10.1371/journal.pone.0155995 Pineli, L. L. O., Aguiar, L. A., Fiusa, A., Botelho, R. B. A., Zandonadi, R. P., & Melo, L. (2016). Sensory impact of lowering sugar content in orange nectars to design healthier, low- sugar industrialized beverages. Appetite, 96, 239-244. doi:10.1016/j.appet.2015.09.028 Popkin, B. M., & Hawkes, C. (2016). Sweetening of the global diet, particularly beverages: patterns, trends, and policy responses. The Lancet Diabetes & Endocrinology, 4(2), 174-186. doi:10.1016/s2213-8587(15)00419-2 Schifferstein, H. N. J., & Oude Ophuis, P. A. M. (1998). Healthrelated determinants of organic food consumption in the Netherlands. Food Quality and Preference, 9, 119-133. Siegrist, M., Leins-Hess, R., & Keller, C. (2015). Which front-of-pack nutrition label is the most efficient one? The results of an eye-tracker study. Food Quality and Preference, 39, 183-190. doi:10.1016/j.foodqual.2014.07.010 Stanhope, K. L. (2016). Sugar consumption, metabolic disease and obesity: The state of the controversy. Critical Reviews in Clinical Laboratory Sciences, 53(1), 52-67. doi:10.3109/10408363.2015.1084990 Steenhuis, I., & Poelman, M. (2017). Portion Size: Latest Developments and Interventions. Curr Obes Rep, 6(1), 10-17. doi:10.1007/s13679-017-0239-x Swiss Federal Food Safety and Veterinary Office. (2017). Zuckerreduktion [Sugar reduction]. Retrieved from https://www.blv.admin.ch/blv/de/home/lebensmittel-und- ernaehrung/ernaehrung/massnahmen-ernaehrungsstrategie/zuckerreduktion.html Swiss Federal Statistical Office. (2012). Schweizerische Gesundheitsbefragung (SGB) [Swiss Federal Statistical Office: Swiss Health Survey (SHS)]. Neuchâtel, Switzerland: BFS Bundesamt für Statistik. Swiss Federal Statistical Office. (2016). Alter, Zivilstand, Staatsangehörigkeit [Age, marital status, nationality]. Retrieved from https://www.bfs.admin.ch/bfs/de/home/statistiken/bevoelkerung/stand- entwicklung/alter-zivilstand-staatsangehoerigkeit.html Swiss Society for Nutrition. (2017). Schweizer Lebensmittelpyramide [Swiss food pyramid]. Retrieved from http://www.sge-ssn.ch/ich-und-du/essen-und- trinken/ausgewogen/schweizer-lebensmittelpyramide Tangney, J. P., Baumeister, R. F., & Boone, A. L. (2004). High Self-Control Predicts Good Adjustment, Less Pathology, Better Grades, and Interpersonal Success. Journal of Personality, 72(2), 271-322. Te Morenga, L., Mallard, S., & Mann, J. (2012). Dietary sugars and body weight: systematic review and meta-analyses of randomised controlled trials and cohort studies. BMJ, 346, e7492. doi:10.1136/bmj.e7492 Temple, N. J., & Fraser, J. (2014). Food labels: a critical assessment. Nutrition, 30(3), 257- 260. doi:10.1016/j.nut.2013.06.012 Thaler, R. H., & Sunstein, C. R. (2008). Improving Decisions About Health, Wealth, and Happiness. New Haven: Yale University Press. WHO. (2015a). Fiscal Policies for Diet and Prevention of Noncommunicable Diseases. Technical Meeting Report. Geneva: World Health Organization.

163 PUBLIC ACCEPTANCE OF INTERVENTIONS

WHO. (2015b). Guideline: Sugars intake for adults and children. Geneva: World Health Organization. Willett, W. (2013). Nutritional epidemiology. New York: Oxford University Press. Wittekind, A., & Walton, J. (2014). Worldwide trends in dietary sugars intake. Nutrition Research Reviews, 27(2), 330-345. doi:10.1017/S0954422414000237

164

Chapter 7

Nutrition labels and their effect on healthiness perception of salty snack food 7 Nutrition labels and their effect on healthiness perception of salty snack food

A randomized controlled online experiment Désirée Hagmann, Michael Siegrist ETH Zurich

Manuscript submitted for publication: Hagmann, D., & Siegrist, M. Nutri-Score, multiple traffic light and incomplete nutrition labelling on food packages: Effects on consumers’ accuracy in identifying healthier snack options.

NUTRITION LABELS

Abstract

Front-of-package (FOP) nutrition labels are designed to help consumers more easily evaluate the healthiness of foods and to promote healthier food choices. An online experiment with Swiss consumers (N = 1,313) was conducted to compare the effects of different nutrition label formats on consumers’ evaluation of the healthiness of snack foods. For 15 salty snacks, all of which were real brands available from a large Swiss retailer, participants were asked to select the healthier option in 105 pairwise comparisons. The participants were randomly assigned to one of five conditions: control (only the FOP presented), table (plus nutrition facts table), multiple traffic light (MTL; plus MTL label), Nutri-Score (plus Nutri-Score), or partial (plus Nutri-Score on half the products). To objectively classify the products according to their healthiness, the nutrient profiling system of the British broadcast regulator Ofcom/Food Standards Agency (FSA), which also underlies the Nutri-Score label, was used. The results suggest that consumers’ evaluation of the healthiness of snacks is fairly accurate even without any nutrition information on the food packaging. The Nutri-Score led to the greatest accuracy in identifying the healthier of two snacks but had only a minimal effect on the evaluation when only some of the products were labelled. Both FOP labels were superior to the FOP with and without the nutrition facts. This indicates that for maximum effectiveness, the labelling of all available products is needed. The perceived usefulness and public support of a mandatory implementation were higher for the MTL than for the Nutri-Score label, but for the latter, perceived usefulness, and public acceptance were higher among participants who gained familiarity with the label during the experiment than among those who did not.

166 NUTRITION LABELS

7.1 Introduction

When grocery shopping, consumers confront many different kinds of information on product packaging. In view of the increasingly unhealthy dietary habits in many countries (e.g. increased consumption of energy-dense, highly processed foods and snack products; Jones & Richardson, 2007; Mattes, 2018; WHO, 2003), the provision of unambiguous and comprehensible nutrition information is important. Nutrition labels, particularly those with front- of-package (FOP) positioning, are intended to help consumers evaluate the healthiness of processed foods and thus to enable informed food choices (WHO, 2003). In addition to this, it is hoped that the presence of labels on food packages will create an incentive for the food industry to reformulate their products and offer healthier options (Kanter, Vanderlee, & Vandevijvere, 2018) in order to avoid adverse effects on the marketing of their products and negative evaluations of the products themselves.

7.1.1 Nutrition label formats Various nutrition label formats are currently in use (Kanter et al., 2018), and they differ in several respects: the type of nutrients on which they focus (e.g., highlighting only critical nutrients or considering health-promoting nutrients as well), the kind of presentation / design features they use (e.g., using numbers, colour codes, shapes, letters), and how directive they are (Hodkins et al., 2009). The mandatory nutrition facts table on the back of the package can be considered a nondirective label, because it provides detailed numerical information about the nutritional components of a product without explicitly evaluating the product’s healthiness. Semidirective nutrition labels such as the multiple traffic light (MTL) signpost, use visual cues such as colour codes or symbols to communicate an evaluation of the product’s critical-nutrient content. On the MTL label, each nutrient attribute (amount of fat, saturated fatty acids, sugar, and salt/sodium) is represented by a separate symbol that indicates whether the amount is low (green), medium (amber), or high (red). These labels do not provide a global evaluation of the product’s healthiness. Directive labels, by contrast, provide a summary evaluation of the heatlhiness of a product without any detailed information. These summary labels include simple labels placed only on foods that meet certain healthiness criteria (e.g., Keyhole, Green Tick, Choices label) and graded labels (e.g., Nutri-Score, Health Star Rating) (Julia & Hercberg, 2017). According to epidemiological nutrition research, healthy diets contain plenty of fruit, vegetables, fibre, plant-based sources of fat and protein, and low amounts of fat, saturated fat, total sugar, and salt, among others (Willett & Stampfer, 2013). A relatively new method, nutrient profiling enables an evaluation and ranking of food products according to the

167 NUTRITION LABELS healthiness of their nutritional composition (WHO, 2017). Various nutrient profiling models exist such as the Ofcom/FSA nutrient profiling model (Food Standards Agency, 2011) and the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014). Each of these models includes a different number of health-relevant nutrients, and the models serve as a basis for the classification schemes on nutrition labels and the determination of food-related health taxes (Rayner, 2017). Currently, there is no consensus regarding which model should be considered the gold standard for objectively defining the healthiness of foods (Poon et al., 2018). However, the Ofcom/FSA nutrient profiling model is one of the most well-known and well-validated models (Rayner, 2017), and it is clearly considered the gold standard by a growing number of countries and food producers, which are introducing the Nutri-Score (the label based on this model) to communicate the healthiness of products to consumers in a simple way.

7.1.2 Effect of nutrition labels on consumers’ evaluation of healthiness Numerous studies have evaluated the impact of nutrition labels on consumers’ perception of the healthiness of foods and have sought to determine which of the available formats is the best means of communicating nutrition information (e.g., Borgmeier & Westenhoefer, 2009; Egnell, Talati, Hercberg, Pettigrew, & Julia, 2018; Gorski Findling et al., 2018; Hawley et al., 2013; Hersey, Wohlgenant, Arsenault, Kosa, & Muth, 2013; Hodkins et al., 2009; Jones & Richardson, 2007; Julia & Hercberg, 2017; Roberto et al., 2012; Siegrist, Hartmann, & Lazzarini, 2019; Watson et al., 2014). Studies based on eye-tracking methods have suggested that compared to the standard nutrition facts panel, FOP labels, especially those that use a traffic light system, are better able to catch consumers’ attention and direct their attention to the nutrients most relevant to healthiness assessment (Becker, Bello, Sundar, Peltier, & Bix, 2015; Jones & Richardson, 2007; van Herpen & Trijp, 2011). This may be due to the more prominent placement of such FOP labels and their design features (Becker et al., 2015). Similarly, the results of another eye-tracking study (Siegrist, Leins-Hess, & Keller, 2015) suggest that the visual information processing of the MTL label overall is more efficient than that of the Guideline Daily Amount (GDA) and the nutrition facts table: The MTL label was processed more quickly than the GDA label (but less quickly than the nutrition facts table), and participants focussed on more relevant information when reading the MTL label or the GDA than when reading the nutrition facts table. The nutrition facts table contains only numerical information, which can be difficult to understand, especially for consumers who have limited literacy skills (Campos, Doxey, & Hammond, 2011; Roberto & Khandpur, 2014). Consequently, several studies that compared different label formats have found that the MTL system resulted in more accurate healthiness

168 NUTRITION LABELS evaluations compared to no label and some other label formats, such as the GDA and a simple healthier choice tick (Borgmeier & Westenhoefer, 2009; Roberto et al., 2012), but other studies did not find substantial differences between different types of FOP labels (Hodgkins et al., 2015; Watson et al., 2014). More recent studies have included the new Nutri-Score label developed in France (Julia & Hercberg, 2017). This label provides a graded, colour-coded summary evaluation of a product’s healthiness, ranging from a dark-green A (most healthy) to a dark-red E (most unhealthy) and considers the content of various health-promoting and critical ingredients. Previous findings suggest that this label is even easier for consumers to understand and results in more accurate healthiness evaluations compared to the MTL and other labelling systems (Ducrot et al., 2015; Egnell et al., 2018). However, Gorski Findling et al. (2018) found that the MTL led to greater accuracy in identifying the healthier of two foods compared to a labelling scheme based on 0–3 stars, another type of graded summary label. Previous studies have differed widely in terms of their design, the labels (or versions of labels) compared, the food categories used, and how the stimuli were presented to the consumers. For example, many studies have not used real brands available in supermarkets (Borgmeier & Westenhoefer, 2009; Egnell et al., 2018; Watson et al., 2014) or have presented labels only, without a concrete product (Hieke & Wilczynski, 2012; Jones & Richardson, 2007), both of which make the decision situation less realistic. Furthermore, few studies have included the Nutri-Score label (Julia & Hercberg, 2017), and to the best of our knowledge, no study has investigated the impact of nutrition labels on consumers’ perception of the healthiness of a realistic set of salty snack foods from existing brands. The results of all the studies we are aware of are based on the assumption that all available products carry a label and can therefore be compared by consumers. However, because in many cases the implementation of nutrition labels is not mandatory (Buttriss, 2018; Kanter et al., 2018), it is likely that situations occur in which only some of the available products are labelled. In Switzerland, for example, the French food company Danone recently began to place the Nutri-Score label on all their dairy products (Danone, 2019), whereas other producers of dairy products have not implemented it. This raises the question of whether nutrition labels such as the Nutri-Score are equally effective when they are not present on all available products. To the best of our knowledge, no previous study has investigated the effectiveness of a nutrition label under the condition of incomplete labelling.

169 NUTRITION LABELS

7.1.3 Study aims

The first aim of this study was to compare two interpretive FOP nutrition labels (MTL and Nutri- Score) in terms of their effect on consumers’ healthiness evaluation of salty snacks, as well as to compare these labels to the standard nutrition facts table and the absence of nutrition information. In order to create a relatively realistic shopping-choice situation (high ecological validity), a range of snacks offered by the same Swiss retailer, all of which are real brands available at stores, were used. We focused on salty snacks because this product category seems highly relevant considering that snacking contributes nearly one-third of daily energy intake in European countries (Mattes, 2018). In addition, salty snacks usually contain critical amounts of sodium, fats, and sugar (British Nutrition Foundation, 2016). Nevertheless, this product category offers some variability in terms of healthiness, which makes it appropriate for the purpose of the present study. The second aim of this study was to investigate whether the effectiveness of the Nutri-Score label differs when it appears on only some of the products. Moreover, this study explored, among a representative sample of Swiss consumers, the perceived usefulness of the Nutri-Score and the MTL labels compared to the nutrition facts table and the ingredients list, as well as public support for the mandatory introduction of these two labels.

170 NUTRITION LABELS

7.2 Materials and Methods

7.2.1 Selection of snacks In order to choose a set of salty snacks people might encounter simultaneously in a real-world shopping situation, we used a range of products offered by a large Swiss retailer. Initially, a larger set of salty snacks was considered, and 15 snack products from this set were ultimately selected. The following criteria were considered for the final selection:

• All products should be available from the same retailer/store. • The products should exhibit a certain variability in terms of healthiness (overall and in terms of fat, sugar, and salt content), type, ingredients, and origin (animal or vegetable) to ensure a variability in the labels (e.g., all Nutri-Score categories A–E should be represented). • The products should not be overly similar (e.g., if salted pretzels are included, pretzel sticks should not be chosen).

The product characteristics of the salty snacks used in the experiment are presented in Table 7.1. Information about the nutritional values and ingredients was taken from the product packaging. If a relevant piece of information was missing, the website of the producer or retailer of the product was consulted.

171

Table 7.1 Characteristics of the salty snack products used in the choice task.

Product name Product Brand Energy Fat Saturated Total Salt Dietary Carbo- Protein Nutri- UK HCST description kcal/ g/100 g fat sugar g/ fibre hydrat g/100 g Score Ofcom/FSA tier 100 g (MTL g/100 g g/100 g 100 g g/ es nutrient colour) (MTL (MTL (MTL 100 g g/100 profiling colour) colour) colour) g score Corn chia waffles Corn waffles Coop Karma 400 2.5 0.5 2 <0.01 2 83 9 A -3 1 with chia (green) (green) (green) (green) Tortilla chips Tortilla chips Coop Qualité 474 20 2 1.5 0.90 4 64 7 B 1 2 Nature & Prix (red) (amber) (green) (amber) DAR-VIDA Nature Whole-grain Hug AG 411 11 1 1 1.60 10 63 10 C 7 2 crackers (amber) (green) (green) (red) Paprika chips Pepper chips Zweifel 544 34 2 5 1.20 5 51 6 C 8 3 (red) (amber) (green) (amber) Popcorn Popcorn Coop 471 21 1.5 0.5 1.90 9 58 8 C 9 4 (red) (green) (green) (red) Wasabi-coated Wasabi-coated Coop Fine 538 34 7 13 0.90 6 31 24 C 10 4 peanuts peanuts Food (red) (red) (amber) (amber) Vegetable chips Vegetable Coop Fine 517 35 4 21 1.30 13 40 5 D 11 3 chips Food (red) (amber) (amber) (amber) Graneo Mild Chili Multigrain Zweifel 471 19 1.5 7 2.40 5 65 8 D 12 3 chips (red) (green) (amber) (red) Bretzeli classic Salted pretzels Roland 386 4.1 0.5 4.2 3.80 2.6 74 12 D 12 3 Murten AG (amber) (green) (green) (red) Mini Tuc Crackers Mondelez 486 19 1.7 5.9 1.72 2.5 69 8.8 D 13 2 International (red) (amber) (amber) (red) Jack Link’s Beef Beef jerky Jack Link’s 260 3.5 1.7 12 5.10 0 15 42 D 16 3 Jerky Original (amber) (amber) (amber) (red) Chips de crevettes Shrimp chips SIBEL 530 30.3 3 7.1 1.95 0.9 62 2.2 D 17 4 (red) (amber) (amber) (red) Mini-Twist flûtes Pastry bars Kambly 485 23 17 5 1.80 0 56 12 E 24 4 (red) (red) (green) (red) Goldfish Wheat Kambly 440 14 11 4.6 3.80 0 67 9.9 E 26 4 crackers (amber) (red) (green) (red) Party sticks Salami sticks Malbuner 497 39 15 1 5.10 0 1.5 35 E 26 4 (red) (red) (green) (red)

NUTRITION LABELS

7.2.2 Procedure An online study was conducted that consisted of an experimental part and a short questionnaire distributed subsequently. In the experimental part, participants performed a choice task: For each possible pairwise combination of 15 salty snacks, they were asked to indicate which of the two snacks was healthier (105 comparisons in total). The pairwise comparisons were presented in an optimum order, as suggested by Ross (1934), in order to establish a maximum time period between the presentation of the same product and balanced variability in the position (left or right). By means of a script programmed by the online panel company Respondi, the participants were randomly assigned to one of the following five conditions: • (1) FOP only condition (control condition): In this condition, only a picture of the front of package of each snack product was presented, without any additional information or evaluation of the product’s nutritional content.

• (2) MTL label condition: A German adaptation of the MTL label (Department of Health/Food Standards Agency, 2016) was created (see Figure 7.1) and presented below each snack picture. This nutrient-specific label provides an evaluation of the content of fat, saturated fat, sugar, and salt per 100 g of the product. Colour coding is used to highlight the content of these four nutrients as either low (green), medium (amber), or high (red) according to reference values defined by the Food Standards Agency (Department of Health/Food Standards Agency, 2016). Additionally, the label indicates the energy content (which is not evaluated by a colour) and provides numerical information about the nutrient content of a standard snack portion (i.e., 25 g according to the Swiss Society for Nutrition).

• (3) Nutrition facts table (table) condition: In this condition, the standard back-of-package nutrition facts table was presented below each product (see Figure 7.1). By default, this table contains the nutritional values per 100 g of the product for energy, fat, saturated fat, carbohydrates, sugar, fibre, protein, and salt.

• (4) Nutri-Score condition: In this condition, the Nutri-Score label introduced in France (Julia & Hercberg, 2017) was presented below each snack package (see Figure 7.1). This label is based on the nutrient profiling system of the UK Food Standards Agency (2011), which evaluates the overall healthiness of a food product according to its nutritional composition (Food Standards Agency, 2011). For the healthiness classification, the product’s content of several health-promoting and critical nutrients is evaluated (i.e., content of energy, fruit, vegetables and nuts, fibre, saturated fat, total sugar, sodium, and protein). This results in

173 NUTRITION LABELS

a single nutrient profiling score. On the Nutri-Score label, this score is subdivided into five categories represented by capital letters and colour coding, ranging from A (dark green) to E (dark red), where A represents foods considered the most healthy and E represents those considered the least healthy. Thus, this summary label provides a graded overall evaluation of the healthiness of a food product.

• (5) Partial Nutri-Score condition (partial): In order to simulate a choice situation in which only some products are labelled, approximately half of the products (7 of 15) were randomly selected, and each of these was presented with the Nutri-Score label. By contrast, for the remaining half, only the front of the package was presented, without any further information or labelling.

To determine the relative healthiness of the 15 snacks, the Ofcom/FSA nutrient profiling (NP) score of each product was calculated based on its nutritional composition (Food Standards Agency, 2011; see Appendix 1). The score represents an objective and validated measure for the healthiness of a food and is based on a given food’s nutrient content per 100 g. For its calculation, 0–10 ‘A’ points are assigned for each unhealthy aspect (i.e., for the amount of energy, saturated fatty acids, total sugar, and sodium – representing a maximum of 40 ‘A’ points in total), and 0–5 ‘C’ points are assigned for each healthy aspect (i.e., for the content of fruits, vegetables, and nuts, fibre, and protein4; – representing a maximum of 15 ‘C’ points in total). For the final Ofcom/FSA NP score, ‘C’ points are subtracted from ‘A’ points. The final score lies somewhere between –15 and 40, with higher scores indicating lower healthiness. Foods scoring 4 or above are classified as ‘less healthy’ (for more details, see Food Standards Agency, 2011). In the choice task, responses were counted as correct if the product with the lower nutrient profiling score (= healthier) was selected or if the difference in the nutrient profiling score was between 0 and 1. The Ofcom/FSA nutrient profiling score of each snack product used in the experiment is shown in Table 7.1.

4 Points for protein are not included if the ‘A’ point total is ³ 11 and if fruit, vegetables, and nuts score less than 5 points.

174 NUTRITION LABELS

FOP only Multiple traffic light Nutrition table Nutri-Score (control condition)

Figure 7.1 Product examples for the experimental conditions (in the partial Nutri-Score condition, half of the stimuli from the Nutri-Score condition and half of the stimuli from the control condition were presented).

175 NUTRITION LABELS

Questionnaire. After the choice task, participants completed a short questionnaire. Among other things, they were asked to report how useful they considered the Nutri-Score label, the MTL label, the nutrition facts table, and the ingredients list for the healthiness evaluation (on a 7-point Likert-type scale ranging from 1 ‘not at all useful’ to 7 ‘very useful’). Participants were also asked whether they thought these two labels should be mandatory on products in Switzerland (response options: ‘yes’, ‘no’, or ‘I don’t know’). To enable all participants from all conditions to answer this question, the same short description of the two labels was provided to participants in all conditions. The purchase frequency of prepackaged snacks was assessed using a 7-point Likert-type scale ranging from 1 ‘rarely/never’ to 7 ‘very often’, and the consumption frequency of salty snacks was measured using nine categories ranging from ‘4 times or more a day’ to ‘rarely/never’. Educational level was grouped into three categories: low (compulsory school), medium (vocational or middle school), or high (higher vocational education or university degree). Consumers’ educational background was collected in order to be able to avoid confounding through these variables.

7.2.3 Study participants The participants were recruited in the German-speaking part of Switzerland through the online panel company Respondi. All respondents gave their written consent and received a monetary incentive of CHF 1.14 (USD 1.14) for their participation in the study. Quotas for sex and age groups were defined in order to obtain samples representative of the Swiss population in each condition and to minimise confounding through these variables. The sample size required to detect small effects (Cohen’s f = .10) was calculated. Given an alpha level of 0.05 and a power of 0.80, a minimum sample of 240 participants per condition was needed (Cohen, 1988).5 In total, 1,561 participants completed the online study (see Table 7.2). Sixty-nine participants were excluded because they had an unrealistically short response time (less than half of the median processing time for the online study of the respective condition; Mdncontrol =

900 sec.; Mdnmtl = 1,555 sec.; Mdntable = 1,423 sec.; Mdnnutri-score = 983 sec.; Mdnpartial = 993 sec.). Data on processing time was collected automatically based on the time stamps identifying the beginning and end of the online study. After excluding participants with unrealistically short processing time, the median processing times per condition were as follows: Mdncontrol = 924 sec.; Mdnmtl = 1,655 sec.; Mdntable = 1,473 sec.; Mdnnutri-score = 999 sec.;

Mdnpartial = 1,001 sec. An additional 179 participants were excluded because their responses exhibited low consistency: At the end of the choice task, five randomly selected comparisons of products

5 The required sample size was initially determined for the calculation of ANOVAs.

176 NUTRITION LABELS were repeated (for every of these comparisons, there was just one correct anwer). Participants were considered inconsistent responders if they answered two or more of the five repeated comparisons differently than the same comparisons answered before. After exclusion, N = 1,313 participants remained in the sample. The number of participants per condition is shown in Table 7.2, along with information about the sex, age, and educational level of the participants remaining in the study (N = 1,313), separately for the five conditions and for the total sample. Prior to the experiment, the participants were asked if they had red-green colourblindness or difficulties differentiating between green, amber, and red (three coloured circles were presented). Overall, 64 participants indicated that they had either one or both problems in colour vision. All analyses were run once with and once without these participants, and the results were the same. Moreover, apart from colour-coded information, both of the nutrition labels used in the study contained written healthiness cues (‘high/medium/low’ or ‘A to E’, respectively). The study was approved by the Ethics Committee of ETH Zurich (EK 2018-N-101).

177

Table 7.2 Study participants in each condition: Recruited sample, excluded participants and demographic characteristics.

Control MTL Nutrition Nutri- Partial Total F(df1,df2) or c2 condition table Score Nutri- sample (df) Score Recruited sample [n] 313 307 312 318 311 1561 Unrealistic response time [n] 2 30 21 9 7 69 Inconsistent responses [n] 45 32 35 21 46 179 Final sample [n] 266 245 256 288 258 1313 Sex c2(4) = .38, ns Males (%) 43.6 45.3 44.1 45.8 51.6 46.1 Females (%) 56.4 54.7 55.9 54.2 48.4 53.9 Mean Age (SD) [years] 48.2 49.1 48.1 48.3 49.7 48.7 F(4,1308) = .45, (16.3) (16.5) (16.4) (15.8) (16.6) (16.3) ns Age groups [years] c2(4) = .66, ns 18–39 (%) 35.7 31.4 32.8 33.7 30.2 32.8 40–64 (%) 44.4 46.5 46.1 46.5 43.0 45.3 65+ (%) 19.9 22.0 21.1 19.8 26.7 21.9 Educational level c2(8) = .44, ns Low1 (%) 4.1 4.9 3.9 3.8 5.8 4.5 Medium2 (%) 60.9 55.9 55.9 57.3 49.6 56.0 High3 (%) 35.0 39.2 40.2 38.9 44.6 39.5

Note. 1Low = compulsory school; 2Medium = vocational or middle school; 3High = higher vocational education or university degree. ns = not significant.

NUTRITION LABELS

7.2.4 Data analysis The data analysis was carried out with IBM SPSS Statistics version 25. To compare the five experimental conditions (control, MTL, table, Nutri-Score, partial) in terms of the percentage of correct healthiness evaluations, Welch’s analysis of variance (ANOVA) was used, because the homogeneity of variances assumption for ANOVA was violated and because the dependent variable and its residuals were not normally distributed in most conditions. To test differences between conditions, Games–Howell post hoc tests were used. All analyses were repeated using the nonparametric Kruskal–Wallis test, and the results were largely similar. As in a previous study (Siegrist et al., 2019), the weighted inaccuracy was calculated in addition to the percentage of correct choices. This measure takes the magnitude of the errors in the healthiness evaluations into account, that is, the degree to which the compared products differed regarding healthiness. In the first step, for each pairwise comparison, 0 points were assigned if the answer was correct, and the difference between the Ofcom/FSA scores for the compared snacks was assigned if the answer was not correct. In the second step, all these deviations were summed and finally divided by the number of comparisons, resulting in an average weighted inaccuracy per comparison. To compare public support of the Nutri-Score and the MTL label between participants who encountered the respective label in the experiment and those who did not, Pearson’s c2 tests were used. To check for confounders, one-way ANOVAs were used to compare the conditions in terms of participants’ purchase frequency of prepackaged snacks and salty snack consumption. The perceived usefulness of different labels/types of nutrition information on the product package was analysed using a repeated-measures ANOVA. Exploratory t-tests for independent samples were conducted to analyse differences in perceived usefulness between participants who encountered a label during the experiment and those who did not (for this purpose, the means of the conditions of those who did not encounter the label were pooled).

7.3 Results

7.3.1 Healthiness evaluation

Proportion of correct choices. The median proportion of comparisons in which the healthier snack product (classified according to the Ofcom/FSA model) was correctly identified was significantly higher than the chance probability (i.e., 50%) in all conditions (see Figure 7.2).

179 NUTRITION LABELS

The Welch’s ANOVA revealed that the five conditions significantly differed in the proportion of correct choices, F(4,651.95) = 141.71, p < .001. Games–Howell post hoc tests showed that participants in the Nutri-Score condition made the most correct evaluations (M = 86.1, SD = 11.9) compared to participants in each of the four other conditions (p < .001). In the MTL condition, the proportion of correct evaluations (M = 74.3, SD = 8.1) was significantly higher than in the control condition (M = 66.9, SD = 8.4), the table condition (M = 67.2, SD = 10.7), and the partial condition (M = 72.0, SD = 8.3), p < .001. No difference was observed in the participants’ performance between the control condition and the table condition. The analysis was repeated once with all the participants (N = 1,561), that is, without excluding those with unrealistic response times and inconsistent responses (see Table 7.2).6

Figure 7.2 Boxplots of the proportion of correct choices in the five conditions. The objective healthiness of the snack products was determined on the basis of the UK Ofcom/FSA nutrient profiling model (Food Standards Agency, 2011). The means of conditions with unlike superscript letters (a–d) differ significantly from each other (based on Games–Howell post hoc tests, p < .001).

6 Using the full sample (N = 1561), the results were largely similar (Welch’s ANOVA: F(4,775.09) = 130.44, p < .001) and the Games–Howell post hoc tests revealed the same differences between the conditions (Nutri-Score: M = 84.9, SD = 12.7; MTL: M = 73.8, SD = 8.5; control: M = 66.9, SD = 8.3; table: M = 67.3, SD = 10.4; partial: M = 71.1, SD = 8.7; p £ .001).

180 NUTRITION LABELS

Magnitude of errors in healthiness evaluation. For the average weighted inaccuracy, the same pattern was observed as in the corresponding analysis of the proportions of correct choices (see Figure 7.3). The Welch’s ANOVA was significant (F(4,651.69) = 133.34, p < .001). Games–Howell post hoc comparisons revealed that the Nutri-Score condition exhibited the lowest level of inaccuracy (M = 1.02, SD = 1.21) compared to each of the four other conditions. Participants in the MTL condition (M = 2.04, SD = 0.90) made less inaccurate choices compared to participants in the control condition (M = 3.04, SD = 1.12), the table condition (M = 2.97, SD = 1.32), and the partial condition (M = 2.49, SD = 1.00). The table and control condition did not differ from each other. The analysis was repeated once with all the participants (N = 1,561), that is, without excluding those with unrealistic response times and inconsistent responses (see Table 7.2).7

Figure 7.3 Boxplots of the average weighted inaccuracy per comparison in the five conditions (this measure takes the magnitude of the errors in the healthiness evaluations into account, that is, the degree to which the compared products differed regarding healthiness). The objective healthiness of the snack products was determined on the basis of the UK Ofcom/FSA nutrient profiling model (Food Standards Agency, 2011). The means of conditions with unlike superscript letters (a–d) differ significantly from each other (based on Games–Howell post hoc tests, p < .001). The maximum possible average inaccuracy per comparison (i.e., if all choices had been incorrect) was 1004/105 = 9.56. The maximum difference between the healthiest and the least healthy product was 29 Ofcom/FSA points (see Appendix 1).

7 Using the full sample (N = 1561), the results were largely similar (Welch’s ANOVA: F(4,776.65) = 124.57, p < .001) and the Games–Howell post hoc tests revealed the same differences between the conditions (Nutri-Score: M = 1.14, SD = 1.31; MTL: M = 2.13, SD = 1.00; control: M = 3.04, SD = 1.09; table: M = 2.98, SD = 1.29; partial: M = 2.59, SD = 1.05; p < .001).

181 NUTRITION LABELS

Stability of the results. The analysis of the differences between the experimental conditions was repeated (see Supplementary Material) using a different nutrient profiling model to classify the products according to their healthiness, the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014). The classification of the products with the HCST and the Ofcom/FSA system was discordant in 38.1% of the pairwise comparisons, which is comparable to the discordance rate found by Poon et al. (2018). Spearman’s rank correlation between the two systems for the 15 snack products was high (rS = .65, p < .001). The results of the Welch’s ANOVA and Games–Howell post hoc tests using the HCST criterion (see Supplementary Material) differed in some respects from the results obtained based on the Ofcom/FSA criterion. First, participants in the MTL condition now exhibited more accurate evaluations than the Nutri-Score. Second, participants in the table condition now performed better than participants in the control condition. And third, participants in the partial condition and the control condition no longer differed. What remained the same was (1) both FOP labels were superior to the FOP with and without the nutrition facts and (2) participants in the partial condition always performed worse than those in the Nutri-Score condition.

7.3.2 Label preference: Perceived usefulness and support of mandatory implementation A repeated-measures ANOVA revealed significant differences in the perceived usefulness of the four types of nutrition information (F(3) = 73.56, p < .001). Across all conditions, consumers considered the MTL label (M = 5.44, SD = 1.51) and the ingredients list (M = 5.37, SD = 1.48) more useful for the healthiness evaluation of foods than the nutrition facts table (M = 5.21, SD = 1.49). The Nutri-Score label was perceived as the least useful type of nutrition information in the overall sample (M = 4.77, SD = 1.76). However, the perceived usefulness of the Nutri- Score label was rated significantly higher among participants who gained familiarity with it during the experiment, that is, those in the Nutri-Score condition and the partial condition (see Table 7.3). The Nutri-Score and partial conditions did not differ with respect to the perceived usefulness of the Nutri-Score (t(544) = .92, p = .36). Overall, 73.2% of the study participants agreed that the MTL label should be mandatory on processed/prepackaged foods in Switzerland (10.4% ‘do not know’), and only 49.1% were in favour of the mandatory use of the Nutri-Score label (19.6% ‘do not know’). Similarly, participants who were exposed to the Nutri-Score label in the experimental task exhibited significantly higher support (63.2% of participants in the Nutri-Score condition and 60.1% of the participants in the partial condition) than did participants who were not exposed to the label in the experiment (40.2% supported a mandatory implementation); see Table 7.4.

182

Table 7.3 Exploratory analysis of the perceived usefulness of different types of nutrition information. Means and standard deviations of all participants, presented separately for those participants who were exposed to the corresponding information/label during the experiment and those who were not.

Type of information All Experience No experience during experiment [N=1313] during experiment

M (SD) M (SD) M (SD) t (df) p MTL 5.44 (1.51) 5.52 (1.42)a, [n = 245] 5.42 (1.53)d [n = 1068] ns Nutri-Score 4.77 (1.76) 5.31 (1.58)b [n = 288]* 4.43 (1.81)e [n = 767] t (1053) = 7.26 < .001 5.19 (1.54)c [n = 258]* t (1023) = 6.01 < .001 Nutrition facts table 5.21 (1.49) 5.29 (1.38) [n = 256] 5.19 (1.51)f [n = 1057] ns

Ingredients list 5.37 (1.47) - - -

Note. The scale for assessing perceived usefulness ranged from 1 ‘not at all useful’ to 7 ‘very useful’. t-tests for independent samples were conducted to compare perceived usefulness of types of nutrition information between participants with and without experience during the experiment. a Mean (SD) of MTL condition. b Mean (SD) of Nutri-Score condition. c Mean (SD) of Partial condition. d Pooled mean (SD) of conditions FOP, table, Nutri-Score, and partial Nutri-Score; e Pooled mean (SD) of conditions FOP, MTL, table; f Pooled mean (SD) of conditions FOP, MTL, Nutri-Score, and partial Nutri-Score. *The Nutri-Score and partial Nutri-Score conditions did not differ with respect to the perceived usefulness of the Nutri-Score t(544) = .92, p = .36. ns = not significant.

Table 7.4 Public support of a mandatory implementation of the MTL and Nutri-Score labels. Percentage of participants who would support a mandatory implementation of the label in Switzerland is shown for the whole sample and separately for those who were exposed to the label during the experiment and those who were not.

Type of information Public support All Experience No experience during experiment during experiment

% % % c2(2) p MTL N = 1313 n = 245 n = 1068

yes 73.2 74.3 72.9a 3.49 ns no 16.4 13.1 17.1a don’t know 10.4 12.7 9.9a

Nutri-Score N = 1313 n = 288 n = 767

yes 49.1 Nutri-Score: 63.2 40.2b 46.15 < .001 no 31.2 24.7 35.7b don’t know 19.6 12.2 24.1b

n = 258 n = 767 yes Partial: 60.1 40.2b 31.26 < .001 no 25.2 35.7b

don’t know 14.7 24.1b

Note. aMean (SD) of conditions FOP, table, Nutri-Score, and partial Nutri-Score; bMean (SD) of conditions FOP, MTL, table; Pearson’s c2 tests were conducted to compare public support between participants with and without experience during the experiment. ns = not significant.

NUTRITION LABELS

7.3.3 Purchase and consumption frequency Across all conditions, 5.6% of the participants reported rarely/never buying prepackaged snacks, and 9.6% reported rarely/never consuming them. The five conditions did not differ significantly, either in the consumers’ reported purchase frequency of prepackaged foods (F (4,1308) = 2.38, p = .05) or in their consumption frequency of salty snacks (F (4,1308) = 1.25, p = .29).

7.4 Discussion

The provision of unambiguous and easy-to-understand nutrition information in the form of nutrition labels is considered an important strategy for helping consumers identify healthier food options and, hopefully, for promoting healthier food choices. However, there is still a lack of consensus about which format best communicates the nutrition information. A main objective of this experimental study was to compare the effects of different kinds of labels and types of nutrition information on consumers’ evaluation of the healthiness of snacks. Furthermore, the study investigated whether accuracy in identifying healthier snack options differed when the FOP label, in this case the Nutri-Score, was present on only some of the products. The results indicate that the presence of interpretive FOP labels, in this case the Nutri-Score label and the MTL signpost, led to more accurate evaluations of the healthiness of salty snacks compared to the standard nutrition facts table or the absence of nutrition information. Following the premise that the Ofcom/FSA score is the gold standard for classifying products according to their healthiness, the Nutri-Score label resulted in the most accurate healthiness evaluations and thus may be the most effective way to communicate this standard to consumers. These results are in line with previous findings that the Nutri-Score label recently introduced in France (Julia & Hercberg, 2017) may be more effective than the MTL system in terms of helping consumers accurately evaluate the healthiness of foods. However, our results also suggest that the effectiveness of the Nutri-Score label depends strongly on how pervasively it is used; the label is less effective if only some products carry it. It is very likely that partial use of other FOP labels would similarily have a weaker effect, but this needs to be tested in future studies. Our results suggest that even in the absence of explicit nutrition information on the product package, consumers seem to have a certain intuitive ability to evaluate the relative healthiness of snacks accurately, which is significantly above the chance level or guessing. A possible explanation for this is that consumers make use of heuristics, or simple rules of thumb, when they do not have sufficient information or lack the time to use complex decision-making

185 NUTRITION LABELS strategies (Gigerenzer & Gaissmaier, 2011). The heuristics used for the evaluation of the healthiness of salty snacks presented in our study could be beliefs like ‘a snack that contains fibre or whole grains is healthier’ or ‘a plant-based snack is healthier than an animal-based snack’, which led to correct evaluations in many but not all cases. Because we used well- known snack products, another explanation could be that the consumers in our study were already familiar with these products and their nutritional composition. However, for other food categories, such as breakfast cereals, previous studies have found even higher accuracy in selecting healthier food options in the absence of a label (Siegrist et al., 2019). The provision of the nutrition facts table did not result in a more accurate healthiness evaluation compared to the control group. The results of previous research are inconclusive. Jones and Richardson (2007), for example, showed that consumers often lack sufficient skills to interpret nutrition information presented in the nutrition facts table, which results in less accurate evaluations. Results from Siegrist et al. (2019), by contrast, indicate that accuracy in evaluating the healthiness of breakfast cereals slightly increased when consumers had the nutrition facts table at hand compared to a no-information condition, but they also showed that the accuracy in choosing healthier options increases with the frequency with which the nutrition table is consulted. Because we did not include a measure of how often and how intensively the participants in our experiment consulted the table, it remains unclear whether they actually used this information for their decisions and how accurately they compared the nutritional values of the snack options. Future studies investigating the effectiveness of nutrition labels should more frequently combine choice tasks with other methods, such as eye tracking. This will provide additional insights into consumers’ visual attention to and processing of nutrition labels on food packaging, which has been investigated in previous studies (Reale & Flint, 2016; Siegrist et al., 2015). In line with previous research (Borgmeier & Westenhoefer, 2009; Ducrot et al., 2015; Gorski Findling et al., 2018), our study suggests that interpretive FOP nutrition labels, such as the MTL and the Nutri-Score labels, lead to greater accuracy in choosing healthier food options compared to no nutrition information and the standard nutrition facts table. Participants who had the Nutri-Score label at hand to evaluate the healthiness of the snacks performed better than those who had the MTL. This is in line with most studies that have compared the effectiveness of the Nutri-Score and the MTL (Ducrot et al., 2015; Egnell et al., 2018). A possible explanation for why the Nutri-Score may lead to greater accuracy in choosing healthier foods is that the label provides a relatively clear and directive evaluation of the product’s overall healthiness, whereas interpreting the information on the MTL label may require more mental effort from the consumer. Making a decision based on the MTL might be more complicated because different nutritional aspects have to be considered and weighed

186 NUTRITION LABELS against each other. This process may be especially complex if all possible traffic light colours are present. In a qualitative study with Mexican consumers, De la Cruz-Gongora et al. (2017) observed that most participants were confused when the MTL label contained mixed colours and perceived greater difficulty in choosing healthier products when this was the case. On average, the participants in the Nutri-Score condition in our experiment also needed much less time to complete the decision task than did participants in the MTL condition. This may indicate that in addition to its better understandability, the Nutri-Score label has the advantage of allowing consumers to make choices more quickly. This is highly relevant considering that in real-world food-purchasing situations, people usually do not spend much time on their decisions (Grunert, Wills, & Fernandez-Celemin, 2010). Several experimental studies conducted in France have demonstrated that with the help of the Nutri-Score label, the foods consumers shopped for online as well as in real supermarkets were significantly healthier compared to no label, the MTL, a star-based format, and other label formats (Crosetto, Lacroix, Muller, & Ruffieux, 2017; Julia & Hercberg, 2017). However, further studies are needed to evaluate the relative effectiveness of the MTL and the Nutri-Score. As mentioned above, in all the studies we are aware of, consumers were confronted with an optimal situation in which all products carried a label. In real-world situations, this might not be the case either, because the implementation of the label happens on a voluntary basis, goes slowly, or is applied only to specific food categories (Kanter et al., 2018). Our results suggest that under more realistic conditions (i.e., labels only displayed on some products), the Nutri-Score label has only a minimal effect on the accuracy of consumers’ healthiness evaluations and is therefore not as effective as it could be if all available products were labelled. This finding is relevant for public policy makers considering an implementation of a new nutrition label. In this study, the presence of the Nutri-Score or the MTL label led to a higher accuracy in evaluating the healthiness of snacks compared to the FOP with or without the nutrition facts table of about 1–2 Ofcom/FSA points per comparison. Future studies are needed to determine whether the observed effect on healthiness perception is of clinical relevance, that is, the degree to which it actually impacts consumers’ food choices and long-term health. Consumers generally seem to like the idea of nutrition labels on products and show greater support for such public health measures compared to other types of interventions, such as taxes on unhealthy foods (Hagmann, Siegrist, & Hartmann, 2018). Our study found substantially higher public support of the MTL label than of the Nutri-Score label in the overall sample, but the participants who gained some familiarity with the Nutri-Score label during the decision task reported considerably higher support than those who did not see it. This finding confirms the conclusions of previous research that acceptance of health policy measures

187 NUTRITION LABELS increases as people become more familiar with them (Diepeveen, Ling, Suhrcke, Roland, & Marteau, 2013). Similarly, we showed that consumers perceive the Nutri-Score label as a less useful tool for evaluating products’ healthiness compared to the MTL label, the nutrition facts panel, and the ingredients list. Because the Nutri-Score labelling format is relatively new, consumers may not be familiar with it. This migh help to explain why perceived usefulness was higher among those who gained familiarity with the label during the experimental task than among those who did not (although it cannot be confirmed that participants who did not see the Nutri-Score label during the task had never been exposed to it before). An alternative explanation for the higher perceived usefulness of the MTL label could be that consumers may have a need for transparent information that allows them to draw their own conclusions. Gaining practical experience with the new Nutri-Score label seems to be associated with both higher public support of mandated use and higher perceived usefulness of the label.

Strengths and limitations Although the study was not conducted in a real shopping environment, it used a selection of salty snack products that is representative of the range of products Swiss consumers could encounter in real-world grocery-shopping situations. Moreover, the results of this study are based on a large sample that includes an equal number of males and females and is representative of the general Swiss population in terms of age (Swiss Federal Statistical Office). A possible limitation of this study is that there exists no objective criterion for healthiness that could be used to compare the effectiveness of the Nutri-Score label and the MTL label; that is, because of the nutrients on which the available criteria are based, none of them is unbiased with respect to these label formats. Therefore, based on our results, we cannot definitively conclude which of the two label formats is more effective in helping consumers making more accurate healthiness evaluations. Moreover, it remains an open question whether in real-world shopping situations, healthiness is such an important criterion for consumers when choosing snack foods. The results of previous research suggest that attributes such as taste, price, convenience, and brand are the most important characteristics considered by consumers when making snack food decisions (Forbes, Kahiya, & Balderstone, 2015). Furthermore, in a study of consumers in the United Kingdom, Grunert et al. (2010) found that the percentage of people who use nutrition information when shopping varies depending on the food category and is somewhat lower for ‘unhealthy’ product categories, such as salty snacks (22%) and sweets (16%), than for other categories, like yoghurt (38%) or breakfast cereals (34%). More studies are needed to further evaluate the effects of the Nutri-Score compared to other label formats on label use in different food categories and on food choices in real-world situations.

188 NUTRITION LABELS

Conclusions Interpretive FOP nutrition labels help consumers identify healthier snack options. Both investigated FOP labels were superior to the FOP with and without the nutrition facts table. If the Ofcom/FSA model is considered the gold standard for classifying foods according to their healthiness, the Nutri-Score seems the most effective label for communicating this standard to consumers, resulting in the most accurate healthiness evaluations. If another standard for the classification of healthiness is used (i.e., the HCST tier system), the Nutri-Score is less effective. The preferred gold standard therefore determines which FOP label is most suitable. For the Nutri-Score, the study showed that when only some of the products contain the label, its effect is only minimally different from the control group. However, whether this finding applies to other label formats remains to be tested.

7.5 Supplementary Materials

Analysis based on the HCST nutrient profiling system To test the stability of the results and rule out any bias caused by the used nutrient profiling system, all analyses were repeated using a second objective criterion to determine the healthiness of the snack products, the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014). This nutrient profiling model considers the content of four critical nutrients (total fat, saturated fat, sodium, and total sugar) per food specific reference amount as defined by Health Canada (2016). The HCST system categorises foods into four tiers, with foods in tier 1 considered the most healthful and foods in tier 4 the least healthful, based on their content of the aforementioned critical nutrients. For the evaluation of the choice task, responses were considered correct if the snack product in the lower tier was chosen or if both products belonged to the same tier. The average weighted inaccuracy was calculated in a manner analogous to the calculation of the Ofcom/FSA Score. The HCST tier of each snack product used in the experiment is shown in Table 7.1.

Results Proportion of correct choices. The median proportion of comparisons in which the healthier snack product (classified according to the HCST system) was correctly identified was significantly higher than the chance probability (i.e., 50%) in all conditions (see Figure 7.4). The Welch’s ANOVA showed significant differences in the proportion of correct choices between the five conditions (F(4,649.59) = 178.27, p < .001). Participants in the MTL condition made the most correct choices (M = 83.0, SD = 4.7) compared to participants in each of the four other conditions (p < .001). In the Nutri-Score condition (M = 80.5, SD = 5.8), the proportion

189 NUTRITION LABELS of correct choices was significantly higher than in the table condition (M = 75.3, SD = 7.6), the partial condition (M = 73.8, SD = 5.9), and the control condition (M = 72.6, SD = 5.7), p < .001. Participants in the partial condition did not differ significantly from either the participants in the control condition or those in the table condition.

Figure 7.4 Boxplots of the proportion of correct choices in the five conditions. The objective healthiness of the snack products was determined on the basis of the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014). The means of conditions that have no superscript letter (a–d) in common differ significantly from each other (based on Games–Howell post hoc tests, p < .001).

Magnitude of errors in healthiness evaluation. The Welch’s ANOVA was significant (F(4,647.40) = 184.81, p < .001). Games–Howell post hoc comparisons revealed that participants in the MTL condition exhibited the lowest level of inaccuracy (M = 0.20, SD = 0.07) compared to participants in each of the four other conditions (p < .001). In the Nutri-Score condition (M = 0.24, SD = 0.09), inaccuracy was lower than in the control condition (M = 0.37, SD = 0.10), the table condition (M = 0.32, SD = 0.13), and the partial condition (M = 0.34, SD = 0.09), p < .001. No difference in accuracy was observed between the partial and the table conditions (see Figure 7.5).

190 NUTRITION LABELS

Figure 7.5 Boxplots of the average weighted inaccuracy per comparison in the five conditions (this measure takes the magnitude of the errors in the healthiness evaluations into account, i.e. how much the compared products differed regarding healthiness). The objective healthiness of the snack products was determined on the basis of the Health Canada Surveillance Tool (HCST) tier system (Health Canada, 2014). The means of conditions with unlike superscript letters (a–d) differ significantly from each other (based on Games–Howell post hoc tests, p < .001 – except for partial vs. control, p < .01). The maximum possible average inaccuracy per comparison (i.e., if all choices had been incorrect) was 111/105 = 1.06. The maximum difference between the healthiest and the least healthy product was 3 HCST tiers (see Table 7.1).

References

Becker, M. W., Bello, N. M., Sundar, R. P., Peltier, C., & Bix, L. (2015). Front of Pack Labels Enhance Attention to Nutrition Information in Novel & Commercial Brands. Food Policy, 56, 76-86. doi:10.1016/j.foodpol.2015.08.001 Borgmeier, I., & Westenhoefer, J. (2009). Impact of different food label formats on healthiness evaluation and food choice of consumers: a randomized-controlled study. BMC Public Health, 9, 184. doi:10.1186/1471-2458-9-184 British Nutrition Foundation. (2016). 2019(January 25, 2019), Foods high in saturated fat, salt and sugar. Retrieved from https://www.nutrition.org.uk/attachments/article/718/Foods%20high%20in%20fat,%20 salt%20and%20sugars%20resource.pdf

191 NUTRITION LABELS

Buttriss, J. L. (2018). The role of nutritional labelling and signposting from a European perspective. Proceedings of the Nutrition Society, 77(3), 321-330. doi:10.1017/S0029665117004098 Campos, S., Doxey, J., & Hammond, D. (2011). Nutrition labels on pre-packaged foods: a systematic review. Public Health Nutrition, 14(8), 1496-1506. doi:10.1017/S1368980010003290 Cohen, J. (1988). Statistical Power analysis for the behavioral sciences. New York: Psychology Press. Crosetto, P., Lacroix, A., Muller, L., & Ruffieux, B. (2017). Modification des achats alimentaires en réponse à cinq logos nutritionnels. Cahiers de Nutrition et de Diététique, 52(3), 129- 133. doi:10.1016/j.cnd.2017.04.002 Danone. (2019). Nutri-Score. Retrieved from http://www.danone.ch/unsere- ueberzeugung/nutri-score/ De la Cruz-Gongora, V., Torres, P., Contreras-Manzano, A., Jauregui de la Mota, A., Mundo- Rosas, V., Villalpando, S., & Rodriguez-Oliveros, G. (2017). Understanding and acceptability by Hispanic consumers of four front-of-pack food labels. The International Journal of Behavioral Nutrition and Physical Activity, 14(1), 28. doi:10.1186/s12966- 017-0482-2 Department of Health/Food Standards Agency. (2016). Guide to creating a front of pack (FoP) nutrition label for pre-packed products sold through retail outlets. Retrieved from https://www.gov.uk/government/publications/front-of-pack-nutrition-labelling-guidance Diepeveen, S., Ling, T., Suhrcke, M., Roland, M., & Marteau, T. M. (2013). Public acceptance of government intervention to change health-realated behaviours: a systemativ review and narrative synthesis. BMC Public Health(13). doi:10.1186/1471-2458-13-756 Ducrot, P., Mejean, C., Julia, C., Kesse-Guyot, E., Touvier, M., Fezeu, L., . . . Peneau, S. (2015). Effectiveness of Front-Of-Pack Nutrition Labels in French Adults: Results from the NutriNet-Sante Cohort Study. PloS One, 10(10), e0140898. doi:10.1371/journal.pone.0140898 Egnell, M., Talati, Z., Hercberg, S., Pettigrew, S., & Julia, C. (2018). Objective Understanding of Front-of-Package Nutrition Labels: An International Comparative Experimental Study across 12 Countries. Nutrients, 10(10). doi:10.3390/nu10101542 Food Standards Agency. (2011). Nutrient Profiling Technical Guidance. Retrieved from https://www.gov.uk/government/publications/the-nutrient-profiling-model Forbes, S. L., Kahiya, E., & Balderstone, C. (2015). Analysis of Snack Food Purchasing and Consumption Behavior. Journal of Food Products Marketing, 22(1), 65-88. doi:10.1080/10454446.2014.949992 Gigerenzer, G., & Gaissmaier, W. (2011). Heuristic decision making. Annual Review of Psychology, 62, 451-482. doi:10.1146/annurev-psych-120709-145346 Gorski Findling, M. T., Werth, P. M., Musicus, A. A., Bragg, M. A., Graham, D. J., Elbel, B., & Roberto, C. A. (2018). Comparing five front-of-pack nutrition labels' influence on consumers' perceptions and purchase intentions. Preventive Medicine, 106, 114-121. doi:10.1016/j.ypmed.2017.10.022 Grunert, K. G., Wills, J. M., & Fernandez-Celemin, L. (2010). Nutrition knowledge, and use and understanding of nutrition information on food labels among consumers in the UK. Appetite, 55(2), 177-189. doi:10.1016/j.appet.2010.05.045

192 NUTRITION LABELS

Hagmann, D., Siegrist, M., & Hartmann, C. (2018). Taxes, labels, or nudges? Public acceptance of various interventions designed to reduce sugar intake. Food Policy, 79, 156-165. doi:10.1016/j.foodpol.2018.06.008 Hawley, K. L., Roberto, C. A., Bragg, M. A., Liu, P. J., Schwartz, M. B., & Brownell, K. D. (2013). The science on front-of-package food labels. Public Health Nutrition, 16(3), 430-439. doi:10.1017/S1368980012000754 Health Canada. (2014). The Development and Use of a Surveillance Tool: The Classification of Foods in the Canadian Nutrient File According to Eating Well with Canada’s Food Guide. Retrieved from http://publications.gc.ca/collections/collection_2014/sc- hc/H164-158-2-2014-eng.pdf Health Canada. (2016). Table of Reference Amounts for Food. Retrieved from https://www.canada.ca/en/health-canada/services/technical-documents-labelling- requirements/table-reference-amounts-food.html Hersey, J. C., Wohlgenant, K. C., Arsenault, J. E., Kosa, K. M., & Muth, M. K. (2013). Effects of front-of-package and shelf nutrition labeling systems on consumers. Nutrition Reviews, 71(1), 1-14. doi:10.1111/nure.12000 Hieke, S., & Wilczynski, P. (2012). Colour Me In--an empirical study on consumer responses to the traffic light signposting system in nutrition labelling. Public Health Nutrition, 15(5), 773-782. doi:10.1017/S1368980011002874 Hodgkins, C. E., Raats, M. M., Fife-Schaw, C., Peacock, M., Groppel-Klein, A., Koenigstorfer, J., . . . Grunert, K. G. (2015). Guiding healthier food choice: systematic comparison of four front-of-pack labelling systems and their effect on judgements of product healthiness. British Journal of Nutrition, 113(10), 1652-1663. doi:10.1017/S0007114515000264 Hodkins, C., Raats, M., Barnett, J., Wasowicz-Kirylo, G., Stysko-Kunkowska, M., Gulcan, Y., . . . Storcksdieck, S. (2009). Derivation of a typology for nutritional labelling. Report prepared for the EC. Brussels: EUFIC. Jones, G., & Richardson, M. (2007). An objective examination of consumer perception of nutrition information based on healthiness ratings and eye movements. Public Health Nutrition, 10(3), 238-244. doi:10.1017/S1368980007258513 Julia, C., & Hercberg, S. (2017). Development of a new fron-of-pack nutrition label in France: the five-color Nutri-Score. Public Health Panorama, 3(4), 537-820. Kanter, R., Vanderlee, L., & Vandevijvere, S. (2018). Front-of-package nutrition labelling policy: global progress and future directions. Public Health Nutrition, 21(8), 1399-1408. doi:10.1017/S1368980018000010 Mattes, R. D. (2018). Snacking: A cause for concern. Physiology and Behavior, 193(Pt B), 279-283. doi:10.1016/j.physbeh.2018.02.010 Poon, T., Labonte, M. E., Mulligan, C., Ahmed, M., Dickinson, K. M., & L'Abbe, M. R. (2018). Comparison of nutrient profiling models for assessing the nutritional quality of foods: a validation study. British Journal of Nutrition, 120(5), 567-582. doi:10.1017/S0007114518001575 Rayner, M. (2017). Nutrient profiling for regulatory purposes. Proceedings of the Nutrition Society, 76(3), 230-236. doi:10.1017/S0029665117000362 Reale, S., & Flint, S. W. (2016). The Impact of Menu Label Design on Visual Attention, Food Choice and Recognition: An Eye Tracking Study. Journal of Sensory Studies, 31(4), 328-340. doi:10.1111/joss.12216

193 NUTRITION LABELS

Roberto, C. A., Bragg, M. A., Seamans, M. J., Mechulan, R. L., Novak, N., & Brownell, K. D. (2012). Evaluation of consumer understanding of different front-of-package nutrition labels, 2010-2011. Preventing Chronic Disease, 9, E149. doi:10.5888/pcd9.120015 Roberto, C. A., & Khandpur, N. (2014). Improving the design of nutrition labels to promote healthier food choices and reasonable portion sizes. International Journal of Obesity (2005), 38 Suppl 1, S25-33. doi:10.1038/ijo.2014.86 Ross, R. T. (1934). Optimum orders for the presentation of pairs in the method of paired comparisons. The Journal of Educational Psychology, 25, 375-382. doi:10.1037/h0070754 Siegrist, M., Hartmann, C., & Lazzarini, G. A. (2019). Healthy choice label does not substantially improve consumers' ability to select healthier cereals: results of an online experiment. British Journal of Nutrition, 121(11), 1313-1320. doi:10.1017/S0007114519000448 Siegrist, M., Leins-Hess, R., & Keller, C. (2015). Which front-of-pack nutrition label is the most efficient one? The results of an eye-tracker study. Food Quality and Preference, 39, 183-190. doi:10.1016/j.foodqual.2014.07.010 Swiss Federal Statistical Office. Population. Retrieved from https://www.bfs.admin.ch/bfs/en/home/statistics/population/effectif- change/population.html Swiss Society for Nutrition. Schweizer Lebensmittelpyramide [Swiss food pyramid]. Retrieved from http://www.sge-ssn.ch/ich-und-du/essen-und-trinken/ausgewogen/schweizer- lebensmittelpyramide van Herpen, E., & Trijp, H. C. (2011). Front-of-pack nutrition labels. Their effect on attention and choices when consumers have varying goals and time constraints. Appetite, 57(1), 148-160. doi:10.1016/j.appet.2011.04.011 Watson, W. L., Kelly, B., Hector, D., Hughes, C., King, L., Crawford, J., . . . Chapman, K. (2014). Can front-of-pack labelling schemes guide healthier food choices? Australian shoppers' responses to seven labelling formats. Appetite, 72, 90-97. doi:10.1016/j.appet.2013.09.027 WHO. (2003). Diet, nutrition and the prevention of chronic diseases. Report of the joint WHO/FAO expert consultation. WHO Technical Report Series 916 (TRS 916). Geneva: World Health Organization. WHO. (2017). Nutrient profiling. Retrieved from https://www.who.int/nutrition/topics/profiling/en/ Willett, W. C., & Stampfer, M. J. (2013). Current evidence on healthy eating. Annual Review of Public Health, 34, 77-95. doi:10.1146/annurev-publhealth-031811-124646

194

Chapter 8

General Discussion 8 General Discussion

GENERAL DISCUSSION

8.1 Introduction

There is an urgent need to find effective ways to change unhealthy eating habits in order to reduce the currently high prevalence of overweight and diet-related diseases in many populations. The World Health Organization (2003) classified unhealthy eating behaviours as one of the major risk factors for health that are modifiable. This is in fact a positive message, indicating that overweight and illness are not inevitable risks but that there are many ways to counteract their development. However, even though adherence to dietary guidelines is associated with considerably better health outcomes, compliance with these guidelines is not as it should be in many countries, including Switzerland (Schneid Schuh, Campos Pellanda, Guessous, & Marques-Vidal, 2018). Apart from motivational barriers to following a healthier diet, other individual factors, such as poor knowledge and skills with which to translate these guidelines into daily food choices, as well as environmental conditions may explain these low levels of adherence. There is still a great need to better understand these determinants of healthy eating. The present doctoral thesis therefore focused not only on several individual determinants, such as consumers’ motives and skills, but also on determinants of the social and macro-level environment (nutrition labelling and social influences on the acquisition of cooking skills). The findings presented in this thesis were obtained using both population-based survey methods involving a large sample of Swiss consumers and, in one study, an experimental approach. The following section summarises the main findings of the studies included in this thesis and provides a discussion of their implications for public health promotion. Subsequently, in section 8.3, some methodological limitations of the studies are addressed. The final section of the chapter offers conclusions derived from the present research.

8.2 Central findings and implications for research and practice

8.2.1 Acquisition of cooking skills and their importance for healthy eating Cooking skills are considered a prerequisite for the preparation of healthy home-cooked meals, and they may reduce dependency on industrially produced meals, which are frequently less healthy in terms of nutritional composition and energy content (Kanzler, Manschein, Lammer, & Wagner, 2015; Lin, Guthrie, & Frazão, 1999; van der Horst, Brunner, & Siegrist, 2010). Consequently, poor cooking skills have repeatedly been associated with a less healthy diet (Hartmann, Dohle, & Siegrist, 2013; McGowan et al., 2017). To promote better cooking skills

196 GENERAL DISCUSSION and more frequent home cooking, it is essential to know more about how these skills are acquired and which factors contribute to their development. In Chapter 2, a study was presented using cross-sectional survey data from the second wave of the Swiss Food Panel 2.0. Regarding the acquisition of cooking skills, it has been found that for people in Switzerland, as for people in other countries, the mother is the primary source of learning regarding the cooking and preparation of foods. For males, their partners/spouses are important sources of such learning, and for females, cooking courses and self-study using diverse media are important sources. In contrast to the literature, which suggests that people generally have less time for home cooking today, we found that people from younger generations spent more time cooking together with their parents than did people from older generations. Moreover, our results suggest that a stronger involvement of children in cooking activities may promote the development of cooking skills, because better skills were observed among these individuals in adulthood. In line with previous studies (Hartmann et al., 2013), associations between cooking skills and healthier food consumption have been found. More specifically, better cooking skills were associated with higher overall diet quality and more frequent vegetable consumption in both genders, and females with better cooking skills reported a lower consumption of unhealthy foods such as sugar-sweetened beverages, sweets, salty snacks, and fast food. The presented study contributes to the evidence that cooking skills are associated with healthier dietary intake and that involvement in cooking activities early in life has a positive effect on cooking skills in adulthood. Some practical implications can be derived from the present findings. First, in view of the positive associations between cooking skills and healthy food consumption, the acquisition and improvement of such skills should be encouraged among people of all ages. For children, early familiarisation with cooking activities and more opportunities for observational learning seem to contribute to the development of cooking skills. Increased involvement at home should therefore be promoted, especially in boys, because males still reported less frequent involvement in home cooking during childhood than females. For a variety of reasons, involvement in home cooking may not be warranted in all families, and especially for children in such families, promoting cooking skills at school may be an opportunity to get them excited about cooking (e.g., with a compulsory cooking class or playful and creative prevention programmes at schools). In our study, few people had learnt to prepare food in cooking classes at school; therefore, we conclude that this may be an underutilised source of learning, especially for males. Similarly, it is necessary to encourage cooking skills in adults. To the best of our knowledge, there are currently no intervention programmes in Switzerland that specifically target adults. In a recent comprehensive review of cooking-focused interventions directed

197 GENERAL DISCUSSION towards adults, Reicks, Kocher, and Reeder (2018), concluded that this type of intervention has the potential to improve dietary intake and psychosocial outcomes. However, the majority of these intervention studies had methodological limitations and a high degree of heterogeneity (e.g., regarding content, duration of the intervention, and whether they involved co- interventions aimed at components other than cooking). As a result, the evidence about the effectiveness of interventions remains unclear, and it is not possible to make sound recommendations about the optimal design of cooking interventions directed towards adults and which elements they should contain (Reicks et al., 2018). Our results suggest that a programme to promote cooking skills in parents and their children may be an effective intervention. Such a programme could encourage the development of parents’ and children’s cooking skills at the same time, while promoting more frequent shared cooking and family meals at home. Other interventions could focus on the dissemination of easy and healthy cooking practices through new media (e.g., video tutorials, regular suggestions via app or email) to boost motivation for cooking and healthy food preparation. However, further research in this field, especially intervention studies with high methodological quality, is needed to evaluate the short- and long-term effects of cooking interventions on cooking skills and dietary behaviour.

8.2.2 Motives for meat avoidance and healthy meat consumption levels Motives are strong drivers of behaviour. The results of the study of meat consumption presented in Chapter 3 suggest that various motives are relevant for meat avoidance and reduced meat consumption and that the strength of some motives varies depending on how people identify themselves with a specific diet style. For individuals who state that they want to renounce meat completely (i.e., vegetarians, vegans, and pescatarians), ethical concerns about animal slaughter and welfare, concerns about the environmental friendliness of meat consumption, and taste preferences have been identified as stronger motives for avoiding meat than for omnivores who reported consciously reducing their meat consumption. Health motives in general were identified as similarly important in all groups. The intention to lose weight, a specific aspect of health motivation, was generally a less prevalent motive but appeared to be a stronger motive in female low-meat consumers than in female vegetarians. In line with what previous studies have found in different populations (Ruby, 2012), we observed inconsistencies between what people indicated as their diet style and their self- reported meat consumption. Many participants who described themselves as vegetarian (vegan or pescatarian) reported more or less regular meat consumption, a phenomenon that was more pronounced in males. Moreover, we found that a substantial number of self-identified low-meat consumers exceed the recommended amounts for moderate meat intake (three

198 GENERAL DISCUSSION portions of 100–120 g per week). This may indicate that many meat consumers have false reference standards about what is considered a low (and healthy) level of meat consumption. This is an important finding from a public health point of view. In future studies, people’s reference standards and if they are aware of recommendations for meat intake from official dietary guidelines should be further investigated in order to determine if there is a need for awareness campaigns to promote healthy levels of meat consumption in the public. A further goal of the study was to identify factors predicting meat consumption among all self-perceived meat reducers/avoiders. Concerns about animal welfare and a preference for vegetarian dishes generally predicted lower meat intake, whereas higher perceived difficulty of avoiding meat and weight-loss motives were associated with higher consumption. Based on the present research, several important concerns could be addressed by health promotion programmes aimed at reducing meat consumption. First, many people seem to have wrong perceptions of adequate meat intake, and there seem to be large discrepancies between consumers’ self-perception with respect to diet style and their reported meat consumption. Second, many people seem to find it difficult to avoid meat in their diet despite their intention to do so. Further studies should clearly identify the difficulties faced and provide strategies for dealing with them. Third, promoting familiarity with vegetarian dishes may strengthen preferences for more plant-based foods and help people considering these as alternatives to meat. This could be initiated, for example, through the use of nudging strategies, such as offering meatless menus as default options in canteens and restaurants, a strategy found to be effective in promoting more plant-based menu choices (Campbell-Arvai, Arvai, & Kalof, 2012). Finally, because we observed a substantially higher level of meat consumption in males than in females who identified themselves as vegetarians, a further goal of public health campaigns could be to dismantle the strong association between meat consumption and the concept of masculinity in order to increase social acceptance of plant-based diets among males. However, further research is needed to determine why male vegetarians seem to face more difficulties in following meatless diets than their female counterparts.

8.2.3 Intuitive eating and its associations with healthy food choices and body weight Intuitive eating (IE) has been proposed as a promising alternative to traditional diets, which have shown only limited success in achieving long-term weight loss (Tribole & Resch, 1995). Instead of deliberately restricting food intake, this nondieting approach outlines four principles intended to help individuals reach a healthy body weight: (1) listen to body signals of hunger and satiety – that is, eat when hungry and stop eating when full; (2) eat for physical reasons rather than to deal with negative emotions (i.e., no emotional eating); (3) give oneself

199 GENERAL DISCUSSION unconditional permission to eat (UPE) what one desires, excluding the idea of ‘forbidden’ foods (i.e., no restrained eating); and (4) choose foods according to bodily needs (‘body-food choice congruence’ B-FCC; Tribole & Resch, 2012; Tylka & Kroon Van Diest, 2013). Although there is evidence from cross-sectional studies that individuals with a more intuitive eating style have a lower body mass index (BMI) (Van Dyke & Drinkwater, 2014), the associations of IE and food intake are not yet fully understood (Carbonneau et al., 2017; Mensinger, Calogero, Stranges, & Tylka, 2016). To better understand these associations, it is important to determine which IE principles are good strategies that should be promoted in the public. Chapter 4 presented a study that evaluated whether an IE style is associated with healthier food choices and a healthier body weight. In line with other cross-sectional research, more intuitive eaters were found to have lower BMI (Van Dyke & Drinkwater, 2014) and a lower tendency to overeat. Regarding food intake, the results were mixed. Most aspects of IE showed only a few minor associations with healthier food intake. By contrast, UPE was related to higher consumption of unhealthy foods and lower consumption of healthy foods. To conclude, most principles of IE seem to have a weak association with healthy eating. Moreover, caution is needed when teaching people the principle of ‘eating without any limits’ (i.e., UPE), because it seems to promote unhealthy food choices and should therefore be considered as problematic as its opposite, an overly restrained eating style. Even though IE is cross-sectionally associated with healthier body weight and a lower tendency to overeat, further longitudinal studies are needed to evaluate whether IE is an effective strategy for successful weight loss and maintaining a healthy body weight in the long term. Finally, there are some concerns about the construct of IE as a whole that need to be addressed. First, findings from studies that have evaluated associations of IE with related constructs (including emotional and external eating, dietary restraint, mindful eating, and overeating regulation) raise some doubts about the novelty of (at least parts of) the construct (Kerin, Webb, & Zimmer-Gembeck, 2019). For example, Kerin et al. (2019) found strong negative correlations between the IE subscale ‘eating for physical rather than emotional reasons’ and emotional eating (r = –.80) and between the IE subscale ‘unconditional permission to eat’ and dietary restraint (r = –.68). Thus, it appears that these components of IE are conceptually not new but rather the opposites of already well-known eating styles positively framed as adaptive eating patterns. Another concern is related to the question of whether IE is suitable for everyone. IE requires a certain level of awareness and ability to correctly perceive and interpret bodily signals, also referred to as interoceptive sensitivity (Herbert, Blechert, Hautzinger, Matthias, & Herbert, 2013). If people lack these interoceptive abilities, they may not properly implement IE strategies and therefore may not experience the desired positive effects on eating behavior

200 GENERAL DISCUSSION and body weight. As pointed out by previous studies involving healthy and eating-disordered patients, it is important for IE intervention programmes to carefully take this into account and, where necessary, provide more intensive training of these abilities (Herbert et al., 2013; Richards, Crowton, Berrett, Smith, & Passmore, 2017).

8.2.4 The role of self-control in moderating the effect of high hedonic hunger tendency on overeating and excessive snacking In the modern food environment, highly palatable foods are easily available in many places. Some individuals are more susceptible to these temptations and more preoccupied with food even when not physically hungry, exhibiting higher hedonic hunger or a stronger drive to eat for pleasure in the absence of an energy deficit (Lowe & Butryn, 2007; Witt & Lowe, 2014). However, high levels of hedonic hunger have not consistently been associated with higher intake of palatable foods and higher BMI (Espel-Huynh, Muratore, & Lowe, 2018). Some research has suggested that high hedonic hunger may lead to overconsumption of palatable, unhealthy foods only in individuals who also have lower inhibitory control (or self-control) over their food intake (Appelhans et al., 2011; Espel-Huynh et al., 2018). In Chapter 5, the role of self-control in moderating the effect of high hedonic hunger on overeating frequency and snacking behaviour was investigated. The study revealed a moderate association between hedonic hunger (measured with the Power of Food Scale; Lowe et al., 2009) and overeating tendency, and a minor association between hedonic hunger and intake of unhealthy snacks (i.e., high sugar, fat, and/or salt content). As hypothesised, it was found that in individuals with a high susceptibility to temptations of the food environment, a high level of self-control has a protective effect; that is, high self-control seemed to attenuate the negative effect of high hedonic hunger on overeating and snacking behaviour. Moreover, self-control was also directly associated with a lower tendency to overeat and consume unhealthy snacks. In conclusion, self-control seems to be a protective factor, because it decreases the risk for unhealthy snacking and excessive weight gain. It seems to positively influence eating behaviour through both direct and indirect mechanisms. Results from several studies suggest that self-control or inhibitory control over food-related responses is an ability that can be improved by training, resulting in the mitigation of unhealthy eating behaviours (Allom, Mullan, & Hagger, 2016; Houben & Jansen, 2011). Therefore, encouraging the training of inhibitory control strategies may be an important approach, especially in individuals with a high susceptibility to temptations of their food environment.

201 GENERAL DISCUSSION

8.2.5 Interventions against excessive sugar intake and their public acceptance

Excessive sugar consumption, especially in the form of sugar-sweetened beverages (SSB), is a major problem in many industrialised countries. Various attempts have been made to tackle the problem, ranging from information-oriented strategies, such as mass media campaigns to raise public awareness about added sugars in foods and beverages, the use of FOP labels to warn against high sugar content, up to more restrictive regulatory strategies, such as the implementation of sugar taxes (Boles, Adams, Gredler, & Manhas, 2014; Popkin & Hawkes, 2016). In addition to the available evidence of the effectiveness of these measures, public acceptance is an important factor that must be considered (Diepeveen, Ling, Suhrcke, Roland, & Marteau, 2013; Sekhon, Cartwright, & Francis, 2017). It has been argued that public acceptance is an important prerequisite for the success of interventions and also because in case of a strong opposition, an implementation is less likely (Diepeveen et al., 2013; Stok et al., 2016). In Chapter 6, current public support of various interventions seeking to reduce excessive sugar consumption in the Swiss population was investigated. Moreover, several predictors for the acceptance of such measures were identified. Similarly to what was found in other populations (Diepeveen et al., 2013), information-oriented measures that are minimally intrusive and sensitise the public to the issue (i.e., sugar nutrition label, health campaigns) were the most popular interventions, whereas regulatory measures (i.e., taxation) were the most unpopular. Interventions based on the nudging approach (i.e., reducing availability, reducing portion size) were neither very popular nor extremely unpopular. It is likely that the latter are associated with a certain feeling of paternalism and are therefore viewed more critically. Stronger regulations such as taxes are even harder to bypass and might be negatively perceived by many people as limiting their freedom of choice. General acceptance of such interventions seems to be influenced by several factors. Males and participants belonging to certain risk groups (i.e., participants with high SSB consumption, overweight participants) exhibited greater resistance to interventions. Moreover, there was a generally higher acceptance among participants from the French-speaking compared to those from the German-speaking part of Switzerland, which demonstrates a strong influence of culture on a populations’ readiness for public health interventions. Higher health consciousness, particularly a higher awareness of one’s own sugar intake, was associated with better acceptance. Different clusters of consumers have been identified based on their acceptance of specific prevention measures. Most participants (about 94%) seem to be at least supportive of ‘soft’ interventions (i.e., information-oriented interventions). Only a minority of the participants (about 6%) were against any of the policies for reducing sugar intake referred to in the survey.

202 GENERAL DISCUSSION

Unfortunately, the best-accepted measures are frequently not the most effective. For example, available research has suggested that public information campaigns have only limited effects on changes in dietary behaviours (e.g., effects of campaigns to promote fruit and vegetable intake in the public) and has offered mixed evidence about the efficacy of nutrition labels (Capacci et al., 2012; Mozaffarian, Angell, Lang, & Rivera, 2018). By contrast, some nudging strategies (e.g., reductions in portion and package sizes), as we also showed, gain less support, but may be more effective (Cadario & Chandon, 2019). Regarding the effect of excise taxes on SSB, studies from the last few years have suggested that these taxes lead to lower demand for the taxed beverages. Price elasticity, or the responsiveness of the demand as a consequence of higher prices/taxes, was found to be higher in low- and middle-income countries compared to high-income countries, and higher in lower-income compared to higher- income households (Chaloupka, Powell, & Warner, 2019). Moreover, lessons learnt from the impact of tobacco taxes on the demand of these products suggest that young people are more responsive to price increases (Chaloupka et al., 2019). This might be the case for sugar taxes as well; however, this needs to be investigated further. According to several studies, acceptance seems to increase with higher perceived effectiveness of these policies (Cadario & Chandon, 2019; Petrescu, Hollands, Couturier, Ng, & Marteau, 2016). Thus, identifying and correcting misconceptions about the efficacy of specific policies may increase acceptance for unpopular but effective policies (Cadario & Chandon, 2019). However, beyond that, more studies are needed that systematically and thoroughly evaluate the short- and long-term effects of different policies on changes in sugar consumption and healthy eating in general.

8.2.6 The effect of nutrition labels on consumers’ healthiness perception of snacks The provision of nutrition information on prepackaged foods is used as a strategy to promote well-informed and healthy food choices. Many consumers have difficulties with interpreting the information provided by the nutrition facts table that is mandatorily printed on the back of prepackaged foods in most countries (European Food Information Council, 2018). Therefore, different front-of-package (FOP) labels have been proposed to help consumers more easily and quickly evaluate the healthiness of products at the point of sale. A new label, the Nutri- Score, was recently introduced, and results from preliminary studies with French consumers suggest that this label has positive effects on consumers’ healthiness evaluations and on healthier food purchases in experimental settings (Ducrot et al., 2015; Egnell, Talati, Hercberg, Pettigrew, & Julia, 2018; Julia & Hercberg, 2017). The implementation of this label system in Switzerland was also recently discussed (Public Health Schweiz, 2019).

203 GENERAL DISCUSSION

A randomised online experiment was conducted (see Chapter 7) to compare the effect of the Nutri-Score label on consumers’ accuracy in identifying the healthier of two snack options to the effect of the multiple traffic light (MTL) label, the mandatory nutrition facts table, no nutrition information on the packaging, and to the effect of labelling only some of the products with the Nutri-Score. Generally, both FOP labels led to a somewhat higher accuracy in the healthiness evaluation compared to the nutrition facts table and no label. However, based on the results, it is not possible to clearly determine which FOP label format, the Nutri-Score or the MTL, is more effective. Depending on the criterion that was used to operationalise the objective healthiness of the products, either the Nutri-Score or the MTL resulted in the most accurate evaluations. Labelling only some of the products with the Nutri-Score seemed to have little or no effect on the evaluations. This is a new finding that, to the best of our knowledge, has not been reported in any previous study. One can conclude that consumers’ evaluation of the healthiness of snacks seems to be relatively accurate even in the absence of nutrition information. FOP labels, if applied to all the products, led to higher accuracy; however, the effects were not very substantial. Therefore, the practical implications of these differences for consumers’ healthiness perceptions remain undetermined, especially in terms of how they translate into healthier food choices. A recent review concluded that nutrition label use in general is associated with healthier diets but that studies of the influence of FOP label use on healthy eating are scarce (Anastasiou, Miller, & Dickinson, 2019). More studies are thus needed to evaluate the effect of FOP labels on healthy food choices and as well as to determine the direction of the influence, that is, whether FOP labels increase the likelihood of healthy food choices or, conversely, whether people with already healthier eating styles tend to make more use of these labels. Moreover, future studies should investigate which consumer subgroups benefit the most from specific FOP labels.

8.3 Limitations of the studies

Most of the research questions included in this thesis were investigated with cross-sectional data from the Swiss Food Panel 2.0. This longitudinal study was carefully designed and conducted from a methodological point of view. Great importance was given to the random selection of participants in order to obtain as representative a sample as possible. The study sample included approximately an equal number of male and female participants. Even though it was not perfectly representative of the Swiss population, the panel sample was relatively large (N > 5,000 in the first wave), and participants of all age groups and educational levels were represented. However, there are some methodological limitations that need to be addressed.

204 GENERAL DISCUSSION

8.3.1 Dietary assessment

Habitual food intake was measured using a semiquantitative food frequency questionnaire (sFFQ) adapted from the Nurses’ Health Study dietary questionnaire for the Swiss context (Hu et al., 2016; Willett, 2013). Even though this kind of questionnaire has been shown to be a valid and suitable method for the assessment of dietary behaviour in large population studies (Cade, Thompson, Burley, & Warm, 2002; Hu et al., 2016; Willett, 2013), it has some limitations that must be acknowledged. The first issue is related to the self-report nature of the food frequency questionnaire. It has repeatedly been observed that in self-report measures, people tend to consciously or unconsciously underreport their real energy/food intake, with a higher tendency in certain population subgroups, such as overweight and obese individuals and people with high dietary restraint (Asbeck et al., 2002; Hill & Davies, 2007). This might have led to an underestimation of the real consumption frequencies of high-calorie foods, especially in these subgroups. In addition to the assessment of consumption frequencies using the sFFQ method, information about the consumed portion sizes was collected. For this purpose, for every food item a predefined standard portion was provided in accordance with the standard portions defined by the Swiss Society for Nutrition (2011) – for example 100–120 g of meat, one egg, or a handful or 120 g of fruit. Participants were asked to report how many of these predefined portions they usually consumed. However, many consumers seem to have difficulties in accurately estimating the portion sizes they consume (Bucher et al., 2017; Godwin, Chambers, Cleveland, & Ingwersen, 2006; Guthrie, 1984). Moreover, some participants may have difficulties converting their estimated own consumed volumes into the number of predefined standard portions asked for in the questionnaire. Evaluation studies of such sFFQs have suggested that for foods that come in natural units (e.g., one egg, one slice of bread), a specification of the standard portion can increase clarity, whereas for foods that do not come in natural units, it can result in the ignoring of portion size specifications (Willett, 2013). A further issue that was observed in our panel study is that consumers seem to have difficulties when asked to estimate their consumption of specific types of a food group separately. Calculating the overall consumption in a food category by summing up such items may lead to an overestimation of the true consumption. This was observed, for example, in fruit and vegetable consumption in previous studies (Bogers et al., 2003), and it may be an alternative explanation for the implausibly high overall meat intake and the general high intake of meat observed among the participants in the food panel study. Another limitation is related to the evaluation of diet quality. Based on sFFQ data, two own short diet quality indices were developed and used for the analyses. The first used predefined cutoff values oriented towards the dietary guidelines of the Swiss Society for

205 GENERAL DISCUSSION

Nutrition (2011), and the other one was sample based (analyses showed that these two indices highly correlated with r = .64, p < .001). No existing validated diet quality index was used, because most of the available indices require information about the intake of specific macro- or micronutrients (e.g., sodium, saturated fatty acids), which could not be calculated based on the sFFQ data. Because the indices used have not been validated and did not take all aspects of a healthy diet into account, errors in the estimation of diet quality cannot be fully ruled out. However, these two indices were created according to principles similar to those underlying existing validated indices (Lassale et al., 2016).

8.3.2 Assessment of weight status To determine the weight status of the participants in the Swiss Food Panel 2.0, BMI was calculated using self-reported information about body weight and height. Even though BMI is generally considered a valid measure of weight status and most feasible for the use in large population studies, because self-report data are often inaccurate, BMI might lead to misclassification in some cases (Rothman, 2008). Previous studies have found that weight is frequently underestimated when self-report measures are used, with overweight individuals exhibiting a greater tendency to report lower body weight than they actually have (Spencer, Appleby, Davey, & Key, 2002). Other studies have found more frequent underreporting of weight by females than males (Merril & Richardson, 2009). Similarly, it was found that body height is frequently overestimated, especially in males and older individuals (Merril & Richardson, 2009; Spencer et al., 2002). Both the underreporting of body weight and the overreporting of body height may have led to an underestimation of the prevalence of overweight when the BMI criterion was used. Another issue associated with BMI is that it does not perfectly reflect a person’s body fat percentage (Rothman, 2008). It may, for example, overestimate body fat and thus overweight in persons with high muscle mass, such as in power athletes. Therefore, results involving BMI should be interpreted with some caution.

8.3.3 Possible selection bias Another issue that needs to be addressed is possible selection bias. The Swiss Food Panel 2.0 study population consisted of adults who voluntarily participated in the study, investing time and providing personal information over a longer period of time without a monetary incentive. It is very likely that those who participated were more motivated than those who were unwilling to participate because they were more interested in and aware of nutritional topics and related health issues. Furthermore, because filling in the survey is very time-consuming, this might have discouraged certain people, such as those with high occupational or private workloads and time pressures, from participating. Moreover, to understand the questions and complete

206 GENERAL DISCUSSION the survey, sufficient German or French language skills were needed, which might have discouraged persons with migration backgrounds from participating. Thus, it is possible that the study sample included less of the vulnerable population subgroups, which to a certain extent may limit the generalisability of the findings. Following the first wave (2017), a nonresponse analysis was conducted to evaluate the reasons for nonresponse and differences in sociodemographic characteristics compared to the study participants. A random sample of 265 nonresponders from the German-speaking area were selected (200 of those originally recruited from the phone book and 65 of those recruited from the address company Schober). A maximum of three attempts to call these persons were made at different times of day. Of all the persons contacted, 145 nonresonders (55%) were reached. Of these, 98 nonresponders were willing to provide details about their age, educational background, and the reason(s) for nonparticipation. Nonresponders were comparable to those who participated in the study in terms of the proportion of males (45% vs. 48%) and mean age (56.8 vs. 56.5 years), but the former had a lower educational level (low [compulsory school]: 16.5% vs. 5.4%; high [higher secondary school, college or university]: 30.4% vs. 52.6%). The main reasons for nonparticipation mentioned were lack of time and general unwillingness to participate in surveys.

8.3.4 Social desirability bias The food panel surveys were conducted anonymously, with self-administered questionnaires. However, because the questionnaire contains some sensitive questions, we cannot rule out social desirability bias, that is, possible distortion of self-report data due to the tendency of study participants to respond in a way that would allow them to avoid embarrassment or to present themselves in a more positive light (e.g., Fisher, 1993). Besides the measures mentioned above (i.e., body weight, height, consumption frequencies of unhealthy and healthy foods), social desirability might have influenced responses to questions about other factors, such as overeating tendency, income, educational attainment, or motives for meat avoidance.

207 GENERAL DISCUSSION

8.4 Final conclusions

Despite the fact that healthy diets could substantially reduce overweight and many associated diseases, and despite the many attempts that have been made to promote healthy eating behaviours in the population, adherence to dietary guidelines is surprisingly low in Swiss adults. The main goal of the thesis was to improve the understanding of factors that facilitate or hinder healthy eating and to derive implications for the design of effective prevention measures. In summary, the thesis provides new insights into a variety of individual and environmental determinants of healthy eating. First, on the individual level, it was shown that better cooking skills, certain motives for meat avoidance, and diet-related self-control were associated with aspects of healthier or less unhealthy food consumption, respectively. Therefore, these skills and motives should be encouraged. Most principles of IE had only limited positive effects on food consumption; in fact, giving oneself ‘unconditional permission to eat’ may even promote unhealthier eating patterns. Therefore, it remains questionable whether teaching people such an eating style is a viable alternative to dieting. Second, on the macro-environmental level, it was shown that the Nutri-Score and MTL FOP labels have a positive effect on the accuracy of consumers’ evaluations of the healthiness of snack products. However, the effects were not substantial compared to the provision of nutrition information with the nutrition facts table or no information on the product packaging. Moreover, if FOP labels are to be effective, it seems necessary to apply them completely on all products. Third, public acceptance of prevention measures to reduce sugar consumption varied greatly depending on the type of intervention and several sociodemographic and health-related factors. This has implications for deciding which public health interventions to implement and identifying the types of intervention for which there may be a need to identify and change negative perceptions. This thesis supports the view that healthy eating is influenced by multiple factors on the individual and environmental levels. There is no single approach that solves the problem and is equally appropriate for all individuals. Rather, public health promotion measures need to tackle several levels. In general, one must say although many of the observed effects on healthy eating are small, taken together, they can make a difference.

208 GENERAL DISCUSSION

References

Allom, V., Mullan, B., & Hagger, M. (2016). Does inhibitory control training improve health behaviour? A meta-analysis. Health Psychology Review, 10(2), 168-186. doi:10.1080/17437199.2015.1051078 Anastasiou, K., Miller, M., & Dickinson, K. (2019). The relationship between food label use and dietary intake in adults: A systematic review. Appetite, 138, 280-291. doi:10.1016/j.appet.2019.03.025 Appelhans, B. M., Woolf, K., Pagoto, S. L., Schneider, K. L., Whited, M. C., & Liebman, R. (2011). Inhibiting food reward: delay discounting, food reward sensitivity, and palatable food intake in overweight and obese women. Obesity (Silver Spring), 19(11), 2175- 2182. doi:10.1038/oby.2011.57 Asbeck, I., Mast, M., Bierwag, A., Westenhofer, J., Acheson, K. J., & Muller, M. J. (2002). Severe underreporting of energy intake in normal weight subjects: use of an appropriate standard and relation to restrained eating. Public Health Nutrition, 5(5), 683-690. doi:10.1079/PHN2002337 Bogers, R. P., Dagnelie, P. C., Westerterp, K. R., Kester, A. D. M., van Klaveren, J. D., Bast, A., & van den Brandt, P. A. (2003). Using a Correction Factor to Correct for Overreporting in a Food-Frequency Questionnaire Does Not Improve Biomarker- Assessed Validity of Estimates for Fruit and Vegetable Consumption1. The Journal of Nutrition, 133(4), 1213-1219. doi:10.1093/jn/133.4.1213 Boles, M., Adams, A., Gredler, A., & Manhas, S. (2014). Ability of a mass media campaign to influence knowledge, attitudes, and behaviors about sugary drinks and obesity. Preventive Medicine, 67 Suppl 1, S40-45. doi:10.1016/j.ypmed.2014.07.023 Bucher, T., Weltert, M., Rollo, M. E., Smith, S. P., Jia, W., Collins, C. E., & Sun, M. (2017). The international food unit: a new measurement aid that can improve portion size estimation. The International Journal of Behavioral Nutrition and Physical Activity, 14(1), 124. doi:10.1186/s12966-017-0583-y Cadario, R., & Chandon, P. (2019). Viewpoint: Effectiveness or consumer acceptance? Tradeoffs in selecting healthy eating nudges. Food Policy, 85, 1-6. doi:10.1016/j.foodpol.2019.04.002 Cade, J., Thompson, R., Burley, V., & Warm, D. (2002). Development, validation and utilisation of food-frequency questionnaires - a review. Public Health Nutrition, 5(4), 567-587. doi:10.1079/PHN2001318 Campbell-Arvai, V., Arvai, J., & Kalof, L. (2012). Motivating Sustainable Food Choices. Environment and Behavior, 46(4), 453-475. doi:10.1177/0013916512469099 Capacci, S., Mazzocchi, M., Shankar, B., Macias, J. B., Verbeke, W., Perez-Cueto, F. J., . . . Traill, W. B. (2012). Policies to promote healthy eating in Europe: a structured review of policies and their effectiveness. Nutrition Reviews, 70(3), 188-200. doi:10.1111/j.1753-4887.2011.00442.x Carbonneau, E., Bégin, C., Lemieux, S., Mongeau, L., Paquette, M.-C., Turcotte, M., . . . Provencher, V. (2017). A Health at Every Size intervention improves intuitive eating and diet quality in Canadian women. Clinical Nutrition, 36, 747-754. doi:10.1016/j.clnu.2016.06.008 Chaloupka, F. J., Powell, L. M., & Warner, K. E. (2019). The use of excise taxes to reduce tobacco, alcohol, and sugary beverage consumption. Annual Review of Public Health, 40, 187-201. doi:10.1146/annurev-publhealth-

209 GENERAL DISCUSSION

Diepeveen, S., Ling, T., Suhrcke, M., Roland, M., & Marteau, T. M. (2013). Public acceptance of government intervention to change health-realated behaviours: a systemativ review and narrative synthesis. BMC Public Health(13). doi:10.1186/1471-2458-13-756 Ducrot, P., Mejean, C., Julia, C., Kesse-Guyot, E., Touvier, M., Fezeu, L., . . . Peneau, S. (2015). Effectiveness of Front-Of-Pack Nutrition Labels in French Adults: Results from the NutriNet-Sante Cohort Study. PloS One, 10(10), e0140898. doi:10.1371/journal.pone.0140898 Egnell, M., Talati, Z., Hercberg, S., Pettigrew, S., & Julia, C. (2018). Objective Understanding of Front-of-Package Nutrition Labels: An International Comparative Experimental Study across 12 Countries. Nutrients, 10(10). doi:10.3390/nu10101542 Espel-Huynh, H. M., Muratore, A. F., & Lowe, M. R. (2018). A narrative review of the construct of hedonic hunger and its measurement by the Power of Food Scale. Obes Sci Pract, 4(3), 238-249. doi:10.1002/osp4.161 European Food Information Council. (2018). Global update on nutrition labelling. Retrieved from https://www.eufic.org/images/uploads/healthy-living/Executive-Summary-GUNL- 2018-V2.pdf Fisher, R. J. (1993). Social Desirability Bias and the Validity of Indirect Questioning. Journal of Consumer Research, 20(2), 303-315. Godwin, S., Chambers, E. t., Cleveland, L., & Ingwersen, L. (2006). A new portion size estimation aid for wedge-shaped foods. Journal of the American Dietetic Association, 106(8), 1246-1250. doi:10.1016/j.jada.2006.05.006 Guthrie, H. A. (1984). Selection and quantification of typical food portions by young adults. Journal of the American Dietetic Association, 84, 1440-1444. Hartmann, C., Dohle, S., & Siegrist, M. (2013). Importance of cooking skills for balanced food choices. Appetite, 65, 125-131. doi:10.1016/j.appet.2013.01.016 Herbert, B. M., Blechert, J., Hautzinger, M., Matthias, E., & Herbert, C. (2013). Intuitive eating is associated with interoceptive sensitivity. Effects on body mass index. Appetite, 70, 22-30. doi:10.1016/j.appet.2013.06.082 Hill, R. J., & Davies, P. S. W. (2007). The validity of self-reported energy intake as determined using the doubly labelled water technique. British Journal of Nutrition, 85(4), 415-430. doi:10.1079/bjn2000281 Houben, K., & Jansen, A. (2011). Training inhibitory control. A recipe for resisting sweet temptations. Appetite, 56(2), 345-349. doi:10.1016/j.appet.2010.12.017 Hu, F. B., Satija, A., Rimm, E. B., Spiegelman, D., Sampson, L., Rosner, B., . . . Willett, W. C. (2016). Diet Assessment Methods in the Nurses' Health Studies and Contribution to Evidence-Based Nutritional Policies and Guidelines. American Journal of Public Health, 106(9), 1567-1572. doi:10.2105/AJPH.2016.303348 Julia, C., & Hercberg, S. (2017). Development of a new fron-of-pack nutrition label in France: the five-color Nutri-Score. Public Health Panorama, 3(4), 537-820. Kanzler, S., Manschein, M., Lammer, G., & Wagner, K. H. (2015). The nutrient composition of European ready meals: protein, fat, total carbohydrates and energy. Food Chemistry, 172, 190-196. doi:10.1016/j.foodchem.2014.09.075 Kerin, J. L., Webb, H. J., & Zimmer-Gembeck, M. J. (2019). Intuitive, mindful, emotional, external and regulatory eating behaviours and beliefs: An investigation of the core components. Appetite, 132, 139-146. doi:10.1016/j.appet.2018.10.011 Lassale, C., Gunter, M. J., Romaguera, D., Peelen, L. M., Van der Schouw, Y. T., Beulens, J. W., . . . Tzoulaki, I. (2016). Diet Quality Scores and Prediction of All-Cause,

210 GENERAL DISCUSSION

Cardiovascular and Cancer Mortality in a Pan-European Cohort Study. PloS One, 11(7), e0159025. doi:10.1371/journal.pone.0159025 Lin, B.-H., Guthrie, J., & Frazão, E. (1999). Away-From-Home Foods Increasingly Important to Quality of American Diet. Washington, DC: United States Department of Agriculture. Economic Research Service, Agriculture Information Bulletin No. 749. Lowe, M. R., & Butryn, M. L. (2007). Hedonic hunger: a new dimension of appetite? Physiology and Behavior, 91(4), 432-439. doi:10.1016/j.physbeh.2007.04.006 Lowe, M. R., Butryn, M. L., Didie, E. R., Annunziato, R. A., Thomas, J. G., Crerand, C. E., . . . Halford, J. (2009). The Power of Food Scale. A new measure of the psychological influence of the food environment. Appetite, 53(1), 114-118. doi:10.1016/j.appet.2009.05.016 McGowan, L., Caraher, M., Raats, M., Lavelle, F., Hollywood, L., McDowell, D., . . . Dean, M. (2017). Domestic cooking and food skills: A review. Critical Reviews in Food Science and Nutrition, 57(11), 2412-2431. doi:10.1080/10408398.2015.1072495 Mensinger, J. L., Calogero, R. M., Stranges, S., & Tylka, T. L. (2016). A weight-neutral versus weight-loss approach for health promotion in women with high BMI: A randomized- controlled trial. Appetite, 105, 364-374. doi:10.1016/j.appet.2016.06.006 Merril, R. M., & Richardson, J. S. (2009). Validity of Self-Reported Height, Weight, and Body Mass Index: Findings From the National Health and Nutrition Examination Survey, 2001-2006. Preventing Chronic Disease, 6(4). Mozaffarian, D., Angell, S. Y., Lang, T., & Rivera, J. A. (2018). Role of government policy in nutrition-barriers to and opportunities for healthier eating. BMJ, 361, k2426. doi:10.1136/bmj.k2426 Petrescu, D. C., Hollands, G. J., Couturier, D. L., Ng, Y. L., & Marteau, T. M. (2016). Public Acceptability in the UK and USA of Nudging to Reduce Obesity: The Example of Reducing Sugar-Sweetened Beverages Consumption. PloS One, 11(6), e0155995. doi:10.1371/journal.pone.0155995 Popkin, B. M., & Hawkes, C. (2016). Sweetening of the global diet, particularly beverages: patterns, trends, and policy responses. The Lancet Diabetes & Endocrinology, 4(2), 174-186. doi:10.1016/s2213-8587(15)00419-2 Public Health Schweiz. (2019). Medienmitteilung zum Thema Nutri-Score [Media release on the Nutri-Score]. Retrieved May 24, 2019, from https://www.public-health.ch/de/news- de/medienmitteilung-zum-thema-nutri-score/ Reicks, M., Kocher, M., & Reeder, J. (2018). Impact of Cooking and Home Food Preparation Interventions Among Adults: A Systematic Review (2011-2016). Journal of Nutrition Education and Behavior, 50(2), 148-172 e141. doi:10.1016/j.jneb.2017.08.004 Richards, P. S., Crowton, S., Berrett, M. E., Smith, M. H., & Passmore, K. (2017). Can patients with eating disorders learn to eat intuitively? A 2-year pilot study. Eat Disord, 25(2), 99- 113. doi:10.1080/10640266.2017.1279907 Rothman, K. J. (2008). BMI-related errors in the measurement of obesity. International Journal of Obesity (2005), 32 Suppl 3, S56-59. doi:10.1038/ijo.2008.87 Ruby, M. B. (2012). Vegetarianism. A blossoming field of study. Appetite, 58(1), 141-150. doi:10.1016/j.appet.2011.09.019 Schneid Schuh, D., Campos Pellanda, L., Guessous, I., & Marques-Vidal, P. (2018). Trends and determinants of change in compliance to dietary guidelines in a Swiss community- dwelling sample. Preventive Medicine, 111, 198-203. doi:10.1016/j.ypmed.2018.03.008

211 GENERAL DISCUSSION

Sekhon, M., Cartwright, M., & Francis, J. J. (2017). Acceptability of healthcare interventions: an overview of reviews and development of a theoretical framework. BMC Health Services Research, 17(1), 88. doi:10.1186/s12913-017-2031-8 Spencer, E. A., Appleby, P. N., Davey, G. K., & Key, T. J. (2002). Validity of self-reported height and weight in 4808 EPIC-Oxford participants. Public Health Nutrition, 5(4), 561- 565. doi:10.1079/PHN2001322 Stok, F. M., de Ridder, D. T., de Vet, E., Nureeva, L., Luszczynska, A., Wardle, J., . . . de Wit, J. B. (2016). Hungry for an intervention? Adolescents' ratings of acceptability of eating- related intervention strategies. BMC Public Health, 16, 5. doi:10.1186/s12889-015- 2665-6 Swiss Society for Nutrition. (2011). Schweizer Lebensmittelpyramide [Swiss food pyramid]. Retrieved April 17, 2019, from http://www.sge-ssn.ch/ich-und-du/essen-und- trinken/ausgewogen/schweizer-lebensmittelpyramide Tribole, E., & Resch, E. (1995). Intuitive eating: A recovery book for the chronic dieter. New York: St. Martin’s Press. Tribole, E., & Resch, E. (2012). Intuitive eating. New York: St. Martin’s Press. Tylka, T. L., & Kroon Van Diest, A. M. (2013). The intuitive eating scale–2: Item refinement and psychometric evaluation with college women and men. Journal of Counseling Psychology, 60(1), 137-153. van der Horst, K., Brunner, T. A., & Siegrist, M. (2010). Ready-meal consumption: associations with weight status and cooking skills. Public Health Nutrition, 14(2), 239-245. doi:10.1017/S1368980010002624 Van Dyke, N., & Drinkwater, E. J. (2014). Relationships between intuitive eating and health indicators: literature review. Public Health Nutrition, 17(8), 1757-1766. doi:10.1017/S1368980013002139 WHO. (2003). Diet, nutrition and the prevention of chronic diseases. Report of the joint WHO/FAO expert consultation. WHO Technical Report Series 916 (TRS 916). Geneva: World Health Organization. Willett, W. (2013). Nutritional epidemiology. Oxford: Oxford University Press. Witt, A. A., & Lowe, M. R. (2014). Hedonic hunger and binge eating among women with eating disorders. International Journal of Eating Disorders, 47(3), 273-280. doi:10.1002/eat.22171

212 DANKSAGUNG

DANKSAGUNG

Die vorliegende Doktorarbeit ist das Ergebnis einer spannenden Reise, bei der ich viel Neues gelernt und viele interessante Menschen getroffen habe. An dieser Stelle möchte ich die Gelegenheit nutzen, mich bei einigen Menschen zu bedanken, die mich besonders bei dieser Arbeit unterstützt haben. Zuallererst möchte ich mich bei Prof. Dr. Michael Siegrist für die tolle Chance bedanken, die er mir gegeben hat, dieses Projekt überhaupt in Angriff nehmen zu können, für seine stets wertvollen Ratschläge sowie seine wohlwollende und wertschätzende Art als Chef, die ich während der ganzen Zeit sehr geschätzt habe. Ebenfalls möchte ich mich bei Dr. Christina Hartmann bedanken, die mich auf meiner Reise als Betreuerin sehr gut begleitet, immer wieder motiviert und mir hilfreiche Feedbacks gegeben hat und mit der ich neben den fachlichen Besprechungen auch viele lustige Momente erleben durfte. Bei Prof. Dr. Joachim Westenhöfer möchte ich mich an dieser Stelle ebenfalls bedanken für sein Interesse an meiner Arbeit und seine freundliche Bereitschaft als Ko- Referent an meiner Prüfung mitzuwirken. I also want to thank Dr. Caroline Horwath for the very pleasant collaboration with her on two exciting topics, I appreciated and enjoyed it a lot. Zuletzt, aber für mich am wichtigsten, möchte ich mich bei meiner Mutter bedanken, für ihre Liebe, Inspiration und Unterstützung bei allem, was ich im Leben tue.

213

CURRICULUM VITAE

Désirée Astrid Hagmann Born January 22, 1984, in Basel, Switzerland

Educational background

Since 2016/09 PhD student at ETH Zurich, Department of Health Sciences and Technology, Consumer Behavior 2007/09 – 2009/06 Master of Science in Psychology at University of Basel Focus: Developmental psychology and personality 2003/10 – 2006/06 Bachelor of Science in Psychology at University of Basel 1999/08 – 12/2002 Matura at Gymnasium Münchenstein

Professional experience

Since 2016/09 Research assistant / PhD student ETH Zurich, Department of Health Sciences and Technology, Consumer Behavior Project: The Swiss Food Panel 2.0 2016/03 – 2016/08 Traffic psychologist Practice for traffic psychology and psychodiagnostics, PD Dr. phil. U. Gerhard 2010/10 – 2016/02 Staff recruiting and marketing Personalbearatung das team ag 2010/01 – 2010/02 Internship in neuropsychology Memory Clinic, University hospital Basel 2007/01 – 2008/12 Internship in Traffic psychology (part-time) Practice for traffic psychology and psychodiagnostics, PD Dr. phil. U. Gerhard 2007/05 – 2008/08 Research assistant (part-time) Department Developmental Psychology and Personality, University of Basel 2005/02 – 2005/06 Internship in School psychology Jugendpsychologischer Dienst Rheinfelden AG

214

215