DSpace https://erepo.uef.fi
Rinnakkaistallenteet Terveystieteiden tiedekunta
2017
Easy-going, rational, susceptible and struggling eaters: A segmentation
study based on eating behaviour tendencies
Pentikäinen Saara
Elsevier BV
info:eu-repo/semantics/article
info:eu-repo/semantics/acceptedVersion
© Elsevier Ltd.
CC BY-NC-ND https://creativecommons.org/licenses/by-nc-nd/4.0/
http://dx.doi.org/10.1016/j.appet.2017.09.001
https://erepo.uef.fi/handle/123456789/4454
Downloaded from University of Eastern Finland's eRepository
Accepted Manuscript
Easy-going, rational, susceptible and struggling eaters: A segmentation study based on eating behaviour tendencies
Saara Pentikäinen, Anne Arvola, Leila Karhunen, Kyösti Pennanen
PII: S0195-6663(16)30990-4 DOI: 10.1016/j.appet.2017.09.001 Reference: APPET 3602
To appear in: Appetite
Received Date: 21 December 2016 Revised Date: 1 September 2017 Accepted Date: 3 September 2017
Please cite this article as: Pentikäinen S., Arvola A., Karhunen L. & Pennanen Kyö., Easy-going,
rational, susceptible and struggling eaters: A segmentation study based on eating behaviour tendencies, Appetite (2017), doi: 10.1016/j.appet.2017.09.001.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
M AN US CR IP T
AC CE PT ED
Easy-going, rational, susceptible and struggling eaters: A segmentation study based on eating behaviour tendencies
Saara Pentikäinena*, Anne Arvolaa, Leila Karhunenb, Kyösti Pennanena
a VTT Technical Research Centre of Finland Ltd., Finland (saara.pentikainen@vtt.fi; anne.arvola@vtt.fi, kyosti.pennanen@vtt.fi)
b Department of Clinical Nutrition, Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Finland
(leila.karhunen@uef.fi)
*Corresponding author
M AN US CR IP T
AC CE PT ED
Abstract
Eating behaviour tendencies, emotional eating (EE), uncontrolled eating (UE) and cognitive restraint (CR), are associated with various indicators of physical and mental health. Therefore, it is important to understand these tendencies in order to design interventions to improve health.
Previous research has mostly examined eating behaviour tendencies individually, without considering typical combinations of these tendencies or their manifestation in well-being and food choices. This study aimed to understand the interactive occurrence of EE, UE and CR in two independent populations. Finnish (n=1060) and German (n=1070) samples were segmented on the basis of their responses to a modified Three-Factor Eating Questionnaire (TFEQ-R15). Well-being, coping strategies and food consumption habits of the segments were studied.
Segmentation revealed four segments: “Susceptible”, “Easy-going”, “Rational” and “Struggling”.
These segments were similar in both countries with regard to well-being, coping strategies and food choices.
EE and UE co-occurred, and these tendencies were mainly responsible for differentiating the segments. Members of the “Rational” and “Easy-going” segments, who had low scores for EE and UE, tended to experience vitality and positive emotions in life, and contentment with their eating.
By contrast, the “Susceptible” and “Struggling” segments, with more pronounced tendencies towards EE and UE, experienced lower levels of vitality and less frequently positive emotions, applied less adaptive coping strategies and experienced more discontent with eating.
The results of the current study suggest that it is possible to identify segments, with differing eating habits, coping strategies and well-being on the basis of the eating behaviour tendencies EE, UE and CR. We discuss possible viewpoints for the design of interventions and food products to help people towards psychologically and physiologically healthier eating styles.
Keywords:
Eating behaviour tendency; eating style; segmentation; Three-Factor Eating Questionnaire; coping;
emotions; vitality; emotional eating; uncontrolled eating; cognitive restraint; food choice
M AN US CR IP T
AC CE PT ED
1. Introduction 1
2
The need to solve eating-related problems, such as disordered eating or obesity, has inspired eating 3
behaviour research over the past decades (French, Epstein, Jeffery, Blundell, & Wardle, 2012).
4
However, combining the information provided by standard scales in novel ways may provide new 5
insights into eating behaviour, as was recently reported by Bouhlal et al. (Bouhlal, McBride, 6
Trivedi, Agurs-Collins, & Persky, 2017). Bouhal’s research group combined measures of appetite 7
for palatable foods, binge eating, disinhibition, food neophobia, pickiness and satiety 8
responsiveness in order to identify population segments based on eating behaviour tendency. They 9
identified two multi-trait phenotypes showing associations with Body Mass Index (BMI) and self- 10
efficacy, and concluded that such a holistic approach could assist in tailoring and improving 11
interventions for weight management or other eating-related problems.
12
The Three-Factor Eating Questionnaire (TFEQ) (Stunkard & Messick, 1985) and a shortened 13
version, the TFEQ-R18 (Karlsson, Persson, Sjöström, & Sullivan, 2000), are among the most 14
frequently applied psychometric measurement tools for studying eating behaviour. The TFEQ-R18 15
captures tendencies for cognitive restraint, uncontrolled eating and emotional eating. Previous 16
studies have mostly examined each of these tendencies independently and have often focused on 17
specific population categories, such as obese people or women (Anglé et al., 2009; Elfhag & Linné, 18
2005; Jaakkola, Hakala, Isolauri, Poussa, & Laitinen, 2013; Jeanes et al., 2017; Järvelä-Reijonen et 19
al., 2016; Nevanperä et al., 2012). We argue that going beyond the traditional approach, observing 20
eating behaviour in a more holistic way and in the general population, would enable the 21
development of more effective approaches (such as interventions and products) to overcome eating- 22
related problems.
23
The current study aims to identify eating style-based population groups by analysing the interactive 24
occurrence of emotional eating, uncontrolled eating and cognitive restraint in two general 25
M AN US CR IP T
AC CE PT ED
populations. We describe the identified segments on the basis of their well-being, coping strategies, 26
food consumption and contentment with eating. We expect our results to provide new perspectives 27
on eating behaviour research and building blocks to target interventions and products to the 28
identified eating style-based segments.
29
The next section firstly reviews the literature on the concepts of cognitive restraint, uncontrolled 30
eating and emotional eating. The subsequent paragraphs describe the variables applied for segment 31
description and their relevance in the context of eating behaviour and the rationale for applying 32
them in segment description. On this basis, expected relations of these variables with eating style 33
segments are described.
34
2. Literature review 35
36
Emotional eating is a tendency to respond to negative emotions by eating (Karlsson et al., 2000).
37
Positive correlations have been found between emotional eating and BMI (Elfhag & Linné, 2005).
38
Furthermore, covariation has been identified between emotional eating and snack food intake (De 39
Lauzon et al., 2004), consumption of sweet foods in adults (Konttinen et al., 2010) and larger meal 40
portion sizes in general populations (Spence et al., 2016). In addition to these eating-related factors, 41
other associations with emotional eating include psychological distress and depressive symptoms in 42
adults (Järvelä-Reijonen et al., 2016; Konttinen et al., 2010; Pidgeon et al., 2013) and occupational 43
burnout in women (Nevanperä et al., 2012).
44
Uncontrolled eating is loss of control over eating (Karlsson et al., 2000). Uncontrolled eating has 45
been shown to correlate positively with BMI (Cornelis et al., 2014), consumption of energy dense 46
foods in middle-aged adults (De Lauzon et al., 2004) and energy intake and central adiposity in 47
mothers (Jaakkola et al., 2013). Larger meal portion sizes in general populations (Spence et al., 48
2016), binge eating behaviours in women with polycystic ovary syndrome (Jeanes et al., 2017), as 49
M AN US CR IP T
AC CE PT ED
well as occupational burnout in women (Nevanperä et al., 2012) and reduced cognitive functioning 50
(Calvo et al., 2014) have also been shown to correlate with uncontrolled eating.
51
The third eating behaviour tendency, cognitive restraint, reflects an inclination to regulate food 52
intake through conscious restriction rather than using physiological cues (Karlsson et al., 2000).
53
Unlike in the case of emotional eating and uncontrolled eating, the results regarding cognitive 54
restraint are mixed. A positive correlation has been found between cognitive restraint and 55
consumption of healthy foods, and a negative correlation with energy intake and meal portion size 56
among adults (De Lauzon et al., 2004; Spence et al., 2016). Other studies have found correlations 57
with unhealthy phenomena. For example, positive correlations have been identified between 58
cognitive restraint and weight gain in adolescents (Elfhag & Linné, 2005), young women (Anglé et 59
al., 2009) and in normal weight adolescents and adults (De Lauzon-Guillain et al., 2006). In short, 60
cognitive restraint appears to be a contradictory concept in terms of its implications for health and 61
well-being.
62
Taken together, these studies suggest that uncontrolled eating and emotional eating are associated 63
with negative phenomena: 1) obesity or weight management problems, 2) poorer psychological 64
well-being, and 3) a less healthy diet, whereas cognitive restraint may involve both positive and 65
negative implications. On the basis of these results, we expected respondents with higher tendencies 66
towards uncontrolled and emotional eating to use more unhealthy foods and to indicate poorer well- 67
being, whereas cognitive restraint was expected to have less clear relationships with other variables.
68
As indicators of subjective well-being, the present study applies subjective vitality and perceived 69
prevalence of positive and negative emotions in one’s life (Diener, 2006; Macht, 2008). In addition 70
to their reflection of well-being, emotions are also measured because of their various 71
interconnections with eating behaviour as antecedents as well as consequences of eating (e.g.
72
Macht, 2008), as the above results from TFEQ studies have also shown.
73
M AN US CR IP T
AC CE PT ED
The concept of vitality originates from Self-Determination Theory (Deci & Ryan, 1980; Ryan &
74
Deci, 2000; Ryan & Frederick, 1997), which postulates that self-determination is reflected as 75
vitality in an individual (Nix, Ryan, Manly, & Deci, 1999; Ryan & Deci, 2008). Vitality has been 76
associated with self-control, as well as multiple other positive outcomes related to well-being 77
(Muraven, Gagné, & Rosman, 2008), happiness, and satisfaction with life (Uysal, Satici, Satici, &
78
Akin, 2014). In general, a vital person possesses better life skills and well-being than his or her less 79
vital counterparts (Fini, Kavousian, Beigy, & Emami, 2010; Uysal et al., 2014). Considering these 80
results, population groups with less problematic eating styles should report more vitality than the 81
groups with more adverse eating styles.
82
The concept of coping strategies refers to the various strategies that people use to cope with stress 83
and negative emotions (Duhachek, 2005). In the context of eating behaviour, disordered eating has 84
been seen as a maladaptive coping response to negative affect or stress (Heatherton & Baumeister, 85
1991). Coping studies state that strategies focusing on solving the problem actively and on framing 86
it positively are more adaptive (i.e. related to positive well-being outcomes) than strategies focusing 87
on denying the problem and on shifting attention away from it (Boals, Vandellen, & Banks, 2011;
88
Penley, Tomaka, & Wiebe, 2002). The review study by Ball and Lee (2000) concluded that 89
individuals with disordered eating apply coping strategies that focus on emotions and avoidance 90
rather than on actions (Ball & Lee, 2000). In line with this, emotional eating interrelates with 91
greater reliance on emotion-oriented coping and avoidance in women (Spoor, Bekker, Van Strien, 92
& van Heck, 2007). Hence, we expected that the segments with non-problematic eating styles 93
would tend to apply active coping strategies, whereas the groups with problematic eating styles 94
probably rely more on emotion-oriented and avoidance strategies.
95
M AN US CR IP T
AC CE PT ED
3. Material and methods 96
Participants and procedures 97
Data were collected from participants in Finland and Germany. As eating behaviour tendencies 98
measured by the TFEQ-R18 have shown validity in several cultural and contextual settings 99
(Cornelis et al., 2014; De Lauzon et al., 2004; Järvelä-Reijonen et al., 2016; Spence et al., 2016), 100
similar results were expected from both countries. The data collection was conducted at the end of 101
2015 as an internet survey in Finland (n=1060) and in Germany (n=1070). Participants were 102
recruited from an online panel of a market research company in both countries. The German panel 103
consisted of 250 000 people and the Finnish panel included 78 000 people at the time of data 104
collection. The market research company invited their randomly selected panel members to 105
participate in the study. The email invitation included information on the purpose of the study, the 106
research consortium and the funding body, estimated answering time and contact information of the 107
researchers. The respondents accepted the invitation and indicated their consent by clicking the 108
survey link (National Advisory Board on Research Ethics, 2009). The respondents received panel 109
points to compensate their time and effort. In order to approximately represent the target 110
populations, the samples were stratified by gender, age group and area of residence. Table 1 111
describes the sample and general population characteristics in Finland and Germany (The European 112
Commission, 2017) . 113
Table 1. Sample and population characteristics.
114
Finnish sample (n=1060)
Finnish population
German sample (n=1070)
German population
Gender (%)
Male 52.3 49.1 49.7 49.3
Female 47.7 50.9 50.3 50.7
M AN US CR IP T
AC CE PT ED
Age groups (%)
18-29 years 23.7 17.9 19.4 17.2
30-39 years 15.9 18.4 16.6 17.3
40-49 years 20.8 17.5 20.8 19.8
50-59 years 21.2 19.6 20.4 22.4
60-74 years 18.4 26.6 22.7 23.3
Education (%)
Non-academic 69.3 63.8 74.0 75.3
Academic degree 30.7 36.2 26.0 24.7
Source for population characteristics: (The European Commission, 2017) 115
116
Measures 117
The three-factor eating questionnaire TFEQ-R18 118
The TFEQ-R18 questionnaire was used to measure eating behaviour tendencies (Karlsson et al., 119
2000). TFEQ-R18 measures cognitive restraint (CR) with 6 items (e.g. “I deliberately take small 120
helpings as a means of controlling my weight”, “I do not eat some foods because they make me 121
fat”), uncontrolled eating (UE) with 9 items (e.g. “Sometimes when I start eating, I just can’t seem 122
to stop”, “When I see a real delicacy, I often get so hungry that I have to eat right away”) and 123
emotional eating (EE) with 3 items (e.g. “When I feel blue, I often overeat”). Item 1 in the 124
uncontrolled eating scale was modified from “When I smell a sizzling steak or a juicy piece of 125
meat…” to a more general version “When I smell a delicious food…” in order to avoid those who 126
do not consume meat ignoring the item. According to the original measure (Karlsson et al., 2000), 127
all items except one were measured with 4-point scales. An 8-point scale was used in one item of 128
cognitive restraint, i.e. “On a scale of 1 to 8, where 1 means no restraint in eating and 8 means total 129
restraint, what number would you give yourself?”. Following de Lauzon et al. (2004, p. 2373), the 130
M AN US CR IP T
AC CE PT ED
raw scale scores were transformed to a 0–100 scale [((raw score – lowest possible raw 131
score)/possible raw score range) * 100] (De Lauzon et al., 2004). Higher scores indicate a greater 132
tendency towards the measured subscale.
133
Well-being indicators 134
The frequency of experiencing different emotions in life was measured using Richin’s scale for 18 135
emotions (Richins, 1997). The question read “How often do you experience the following emotions 136
in your life”. Answers were given with a 5-point scale, from 1=never, to 5=very often. The means 137
of the items in each category were calculated to form variables for positive emotions (including 138
items such as excitement, joy, and contentment) and negative emotions (including items such as 139
loneliness, guilt, and anger). Higher scores indicate a greater tendency towards the measured scale.
140
Cronbach’s alphas for the positive emotion scales were 0.849 in Finland and 0.856 in Germany. The 141
alphas for negative emotion scales were 0.842 in Finland and 0.840 in Germany. Values higher than 142
0.70 were deemed acceptable (Nunnally & Bernstein, 1994).
143
Subjective vitality was measured by adopting the vitality index scale from Ryan and Frederick 144
(1997). The scale consists of seven items (such as “I feel alive and vital”, “I look forward to each 145
new day” and “I feel energized”) which are measured on a 7-point scale, from 1=not at all true, to 7 146
=very true (Ryan & Frederick, 1997). Higher scores indicate a higher level of vitality. Cronbach’s 147
alphas for the vitality scales were 0.901 in Finland and 0.854 in Germany.
148
Coping strategies 149
Coping strategies were measured with a multidimensional coping scale adapted from Duhachek 150
(2005). Following Duhachek (2005), the respondents were first asked to imagine a stressful 151
situation in which they felt negative emotions and then to evaluate their tendency towards applying 152
each of the strategies. Strategies typically used by the respondents for coping with negative 153
emotions were measured with a list of 36 items (Duhachek, 2005). Answers were given with a 7- 154
M AN US CR IP T
AC CE PT ED
point scale, from 1=I never use this strategy, to 7=I often use this strategy. These single coping 155
strategies have been shown to form the following eight coping factors(examples in quotes): 1) 156
Action “Try to make a plan of action”, 2) Rational thinking “Analyze the problem before reacting”, 157
3) Positive thinking “Try to look at the bright side of things”, 4) Emotional support “Seek out others 158
for comfort”, 5) Emotional venting ”Let my feelings out somehow”, 6) Instrumental support “Try to 159
get advice from someone about what to do”, 7) Avoidance “Avoid thinking about it”, 8) Denial 160
“Deny that the event happened”. In turn, these eight coping factors form three coping dimensions:
161
active coping (factors 1-3) (alphas: Finland 0.971 and Germany 0.918), expressive support seeking 162
(factors 4-6) (alphas: Finland 0.962 and Germany 0.873) and avoidance (factors 7-8) (alphas:
163
Finland 0.938 and Germany 0.810) as higher order categories (Duhachek, 2005).
164
Food consumption 165
Food consumption frequency was estimated by asking respondents to rate their consumption 166
frequency of 39 food products (including dairy products, meat products, functional foods, 167
convenience foods, snacks, vegetables, fruits and berries, beverages, alcoholic beverages, cereal 168
products, pasta and rice, and superfoods) on a 6-point scale in which 1=never and 6=on a daily 169
basis.
170
M AN US CR IP T
AC CE PT ED
Discontent with eating habits 171
Discontent with eating habits was measured with a scale developed for the purposes of this study.
172
The question read, “How often are you discontent with your own or your family’s eating habits at 173
different meal times?” Respondents were then asked to evaluate their discontent with different meal 174
times (breakfast time, lunch time, afternoon, dinner time, late evening and night) with a 5-point 175
scale (1=never, 2=rarely, 3=sometimes, 4=often, 5=very often). For the analyses, an overall level of 176
discontent was calculated as a mean of the evaluations for different meal times. Cronbach’s alphas 177
for the scale were 0.878 in Finland and 0.833 in Germany.
178
Data analyses 179
Confirmatory factor analysis on TFEQ-R18 180
Confirmatory factor analyses (IBM SPSS AMOS 22) using maximum likelihood estimation were 181
performed for the TFEQ-R18 to study the fit of the three-factor model to the Finnish and German 182
data separately. This analytical step was considered necessary because the factor structure of the 183
TFEQ-R18 has been revealed to be unstable in previous studies (Cappelleri et al., 2009; Chong et 184
al., 2016). Relative chi-square (X2 / df) (because the chi-square test is sensitive to high sample 185
sizes), comparative fit index (CFI), root mean square error of approximation (RMSEA) and 186
standardised root mean square residual (SRMR) were used to determine the model fit (Hooper, 187
Mullen, Hooper, Coughlan, & Mullen, 2008; Iacobucci, 2010). The model fit was deemed 188
acceptable when the relative chi-square was ≤ 5 (some researchers argue that the value should be 189
less than 5, whereas others require that it should be less than 2), CFI ≥.95, RMSEA ≤.06, and 190
SRMR ≤.08 (Hooper et al., 2008; Hu & Bentler, 1999; Iacobucci, 2010). In addition, Cronbach’s 191
alphas were calculated for the factors to evaluate their internal consistency. Values higher than 0.70 192
were deemed acceptable (Nunnally & Bernstein, 1994).
193
M AN US CR IP T
AC CE PT ED
Fit indices for the three-factor model of the TFEQ-R18 suggested refinement for both the Finnish 194
and German models (Table 2). Cronbach’s alphas showed acceptable internal consistency for the 195
factors in both countries, although cognitive restraint (CR) was on the borderline (Finland:
196
CR=0.70, EE=0.89, UE=0.89; Germany: CR=0.71, EE=0.90, UE=0.88). Both Finnish and German 197
data sets revealed three items related to cognitive restraint to be weak. These items were: “How 198
frequently do you avoid `stocking up' on tempting foods?” “How likely are you to consciously eat 199
less than you want?” and “On a scale of 1 to 8, where 1 means no restraint in eating (eating 200
whatever you want, whenever you want it) and 8 means total restraint (constantly limiting food 201
intake and never `giving in'), what number would you give yourself?” The squared multiple 202
correlations for the items were less than 0.20, indicating high levels of errors (Hooper et al., 2008).
203
The removal of these items from the refined models (designated as TFEQ-R15) led to an acceptable 204
fit in both data sets (Table 2). Cronbach’s alphas for the refined models indicated improved internal 205
consistency for the factors (Finland: CR=0.79, EE=0.89, UE=0.89; Germany: CR=0.78, EE=0.90, 206
UE=0.88). Therefore, in further analyses the TFEQ-R15 will be used instead of the original TFEQ- 207
R18.
208
Table 2. Fit indices for the models in the Finnish and German data 209
X2 / df CFI RMSEA SRMR
Finnish data Model
TFEQ-R18 6.2615 .920 .070 .1090
TFEQ-R15 4.4300 .963 .057 .0382
German data Model
TFEQ-R18 6.6810 .912 .073 .1078
TFEQ-R15 4.6804 .959 .059 .0373
X2/df, relative Chi-square; CFI, Comparative fit index, RMSEA, Root mean square error of 210
approximation, SRMR, Standardised root mean square residual, TFEQ=Three-Factor Eating 211
Questionnaire 212
Cut-off points: X2/df ≤5; CFI ≥ .95; RMSEA ≤ .06; SRMR ≤ .08 213
214
M AN US CR IP T
AC CE PT ED
Segmentation 215
Segmentation was accomplished separately, based on the TFEQ-R15, for Finnish and German 216
datasets following the two-stage cluster analysis method. This method integrates hierarchical 217
clustering techniques with iterative partitioning methods (Punj & Stewart, 1983). This procedure 218
defines data segments in a powerful way (Helsen & Green, 1991; Homburg, Workman, & Jensen, 219
2002). At first, the number of clusters in both Finnish and German data was explored by performing 220
Ward’s hierarchical cluster analysis with squared Euclidian distances on three random and 221
overlapping subsamples, in which each subsample represented approximately 33 % of the data 222
(Homburg et al., 2002; McDonald, Leckie, Karg, Zubcevic-Basic, & Lock, 2016; Napoli, 223
Dickinson-Delaporte, & Beverland, 2016). The number of potential segments was decided based on 224
observation of the dendograms of each subsample (McDonald et al., 2016; Napoli et al., 2016). The 225
observation determined a four-cluster solution as the most acceptable (all dendograms indicated 226
four clusters, except for one German dendogram which indicated three clusters). Subsequently, 227
segmentation was performed using k-means cluster analysis in which the four clusters were used as 228
a starting point (Homburg et al., 2002; Napoli et al., 2016); Punj & Stewart, 1983). Finally, one- 229
way analysis of variance (ANOVA) was performed to confirm that the differences between the 230
segments were statistically significant.
231
Factoring of food product categories 232
In order to reduce the number of variables (from individual products to food product categories), 233
and to categorize listed foods in a way that reflected respondents’ thinking in each country, a series 234
of principal axis factor analyses (with varimax rotation) was performed separately for the Finnish 235
and German datasets. Principal axis factoring was chosen because the scales were not normally 236
distributed (Costello & Osborne, 2005).
237
M AN US CR IP T
AC CE PT ED
Analyses were performed iteratively in several phases. In each iteration phase, food items were 238
dropped out if they had low communalities (< .30), high loading on more than one factor, and / or if 239
they did not seem to fit the semantic content of that factor. For example, “Soft drinks enriched with 240
vitamins” loaded high on factors “Superfoods”, “Unhealthy snacks”, and “Low fat dairy” in the 241
Finnish data. Thus, the meaning of that food for consumers remained ambiguous, and the item was 242
omitted from subsequent analyses. Another example of variables dropped out was “Vitamins and 243
other nutritional supplements”, which had low communality (.20), and loaded on two factors:
244
“Superfoods” and “Healthy foods”. Iterative analyses were continued until a statistically acceptable 245
and meaningful solution was reached. The final solution in the Finnish data included 21 products 246
(KMO= 0.77, Bartlett’s test of sphericity, p=.00). Six factors were extracted with eigenvalues above 247
1, which explained the 60.5 % of variance. The first factor, designated “Super and functional foods”
248
included superfood products, superfood powders, herbal teas with functional properties, coconut 249
butter or water, fresh smoothies and juices, and healthy snack bars. The second factor was 250
designated “Unhealthy snacks and convenience foods” and included fast foods, snacks such as 251
potato chips, sugar-sweetened beverages, and convenience foods. The third factor was “Healthy 252
foods”, including fresh chicken meat, fresh fish, fruits and berries, and vegetables. The fourth factor 253
was “Sweets”, which included sweet pastries, sweets and chocolate, and ice cream. Beer and wine 254
constituted the fifth factor, designated “Alcoholic beverages”. Finally, the sixth factor included low 255
fat cheese and low fat yoghurt and was designated “Low fat dairy products”.
256
The German data revealed five factors with eigenvalues above 1, which explained the 60.3 % of 257
variance. Due to low communalities (<.30) and cross loadings, these products were dropped out.
258
The final solution included 22 products (KMO= 0.83, Bartlett’s test of sphericity, p=.00). The first 259
factor, “Superfoods”, included protein bars, superfood powders, superfood products, healthy snack 260
bars, bowel friendly yoghurts, protein enriched quarks, fresh smoothies and juices and herbal teas 261
with functional properties. The second factor was “Unhealthy snacks and convenience foods” and 262
M AN US CR IP T
AC CE PT ED
consisted of fast foods such as pizza and hamburger, sweets and chocolate, sweet pastries, snacks 263
such as potato chips, convenience foods and sugar-sweetened beverages. “Healthy foods” was the 264
third factor, which included vegetables, fruits and berries and dark bread. The fourth factor 265
consisted of fresh chicken meat, fresh red meat and fresh fish and was designated “Fresh meat and 266
fish”. The fifth factor was designated “Alcoholic beverages” and included wine and beer.
267
Statistical tests to compare segments 268
Cross tabulation with chi-square and pairwise comparisons with the z-test were performed to 269
compare the gender, age and education groups among segments. One-way analysis of variance 270
(ANOVA) with Tukey’s test for pairwise comparisons was used to compare the segments’ well- 271
being indicators, coping strategies, food consumption and contentment with eating.
272
4. Results 273
274
Segmentation based on TFEQ-R15 275
Segmentation based on the TFEQ-R15 revealed four similar segments in Finland and Germany.
276
Figure 1 presents the identified segments and the means for each eating behaviour tendency among 277
them. Segment 1 was designated “Susceptible”, since the group had high scores for uncontrolled 278
eating and emotional eating but low scores for cognitive restraint, thus being susceptible to both 279
external and internal cues for eating, and with low attempts to control eating. The share of the 280
respondents belonging to this segment was 20.4 % in Finland and 19.8 % in Germany.
281
The second segment had low scores for all the three factors, indicating a carefree eating style, and 282
was thus designated “Easy-going”. The “Easy-going” segment covered 30.8 % and 29.8 % of the 283
Finnish and German samples, respectively.
284
M AN US CR IP T
AC CE PT ED
Segment 3 showed high attempts to consciously restrict eating, which was manifested in high 285
cognitive restraint scores. However, scores for the other tendencies were low and thus the segment 286
was designated “Rational” (29.5 % and 31.3 % of the Finnish and German samples respectively).
287
The fourth segment had high scores for all three factors, indicating high susceptibility to external 288
and internal triggers for eating, and at the same time strong attempts to control eating. This suggests 289
that individuals in this segment struggle with eating and the segment was thus designated 290
“Struggling”. The share of “Struggling” respondents was 19.3 % and 19.1 % in the Finnish and 291
German samples, respectively.
292
293
294
295
Figure 1 Eating styles in the identified segments in Finland and Germany. Bars represent means of 296
eating behaviour tendencies (scale 0-100 with standard deviations). ANOVA was used to compare 297
M AN US CR IP T
AC CE PT ED
the segments and Tukey’s test for pairwise comparisons. In both the samples the “Rational” and 298
“Struggling” groups had significantly higher scores for cognitive restraint than the “Susceptible”
299
and “Easy-going” groups (p<0.001 in all pairwise comparisons), and the “Susceptible” and 300
“Struggling” groups had higher scores for uncontrolled eating and emotional eating compared to the 301
“Easy-going” and “Rational” groups (p<0.001 in all pairwise comparisons).
302
303
Segment profiles 304
Demographics 305
Table 3 presents the gender and age distributions and percentages of people with academic degrees 306
in the segments. To sum up, the segments “Susceptible” and “Struggling” tended to differ from 307
“Rational” and in some cases from “Easy-going”. The “Susceptible” and “Struggling” segments 308
tended to include more female respondents, more younger age groups, and less older age groups. In 309
the Finnish sample, the “Struggling” respondents were more often educated to academic degree 310
level, but in Germany no statistical differences were observed for education or for gender.
311
M AN US CR IP T
AC CE PT ED
Table 3 Gender and age distributions and percentages of people with academic degrees in the Finnish and German samples and segments.
312
Different superscript letters in a row indicate a statistically significant difference between segments.
313
Finland Germany
Sample (n=1060)
Susceptible (n=216)
Easy-going (n=327)
Rational (n=313)
Struggling (n=204)
sig. Sample (n=1070)
Susceptible (n=212)
Easy-going (n=319)
Rational (n=335)
Struggling (n=204)
sig.
Gender (%)
Male 52.3 39.8a 59.6b 57.5b 45.6a <0.001 49.7 48.6 53.6 51.0 42.6 0.096
Female 47.7 60.2b 40.4a 42.5a 54.4b 50.3 51.4 46.4 49.0 57.4
Age groups (%)
18-29 years 23.7 28.7b 19.9a 15.7a 36.8b <0.001 19.4 27.4b 14.4a 13.4a 28.9b <0.001
30-39 years 15.9 19.4b 18.3b 9.6a 18.1b 16.6 19.8b,c 13.8a,b 12.5a 24.5c
40-49 years 20.8 17.6a 19.9a 23.3a 21.6a 20.8 23.1a 20.1a 19.7a 21.6a
50-59 years 21.2 17.6,a,b 23.9b,c 25.6c 14.2a 20.4 16.5a,b 26.3c 22.1b,c 12.3a
60-74 years 18.4 16.7b 18.0b 25.9c 9.3a 22.7 13.2a 25.4b 32.2b 12.7a
Education (%)
Non-academic 69.3 69.4b 73.7b 71.2b 59.3a <0.001 74.0 75.9 74.6 72.8 73.0 0.847
Academic degree
30.7 30.6a 26.3a 28.8a 40.7b 26.0 24.1 25.4 27.2 27.0
314 315
M AN US CR IP T
AC CE PT ED
Well-being 316
Figure 2 presents the means for negative and positive emotions and vitality in life for each segment 317
in Finland and in Germany. The segments in both countries differed in the frequency with which 318
they experienced negative and positive emotions and vitality in their lives. In the Finnish data, the 319
“Rational” segment experienced positive emotions more often than the “Susceptible” segment, and 320
the “Susceptible” segment experienced more negative emotions compared to all the other segments.
321
The “Struggling” segment experienced negative emotions more frequently compared to the 322
“Rational” and “Easy-going” segments. The “Rational”, Easy-going” and “Struggling” segments 323
experienced more vitality in their lives than the “Susceptible” segment. The “Rational" segment 324
also experienced more vitality compared to the “Easy-going” segment.
325
In the German segments, the “Easy-going” segment experienced positive emotions more frequently 326
than the “Susceptible” segment. The “Struggling” and “Susceptible” segments experienced negative 327
emotions more frequently than the “Easy-going” and “Rational” segments. The “Easy-going”
328
segment experienced more vitality than the “Susceptible” segment.
329
330
331
332
Figure 2 Well-being indicators (negative and positive emotions (scale 1-5) and perceived vitality 333
(scale 1-7)) in the Finnish and German segments. Bars represent means with standard deviations.
334
Different letters indicate a statistically significant difference (p<0.05) between the segments.
335
M AN US CR IP T
AC CE PT ED
Coping strategies 336
The mean tendencies to utilize different coping strategies are presented in Figure 3. In both 337
countries, the segments were different in terms of the coping strategies that were typically applied.
338
In the Finnish sample, the “Rational” segment was more likely to use active coping than the 339
“Susceptible” or “Easy-going” segments. The “Susceptible” segment used active coping the least 340
frequently. The ”Struggling” segment used expressive support seeking coping strategies more often 341
than the other segments. The “Easy-going” segment used these strategies the least frequently. The 342
“Susceptible” and “Struggling” segments used avoidance type coping strategies more often than the 343
“Easy-going” and “Rational” segments.
344
In the German sample, the “Easy-going” and “Rational” segments used active coping more often 345
than the “Susceptible” segment. The “Struggling” segment used expressive support seeking and 346
avoidance more often than the other segments. The “Susceptible” segment used avoidance more 347
often than the “Easy-going” segment.
348
349
350
Figure 3 Coping strategies in the segments on a scale from 1 to 7. Bars represent means with 351
standard deviations. Different letters above bars indicate a statistically significant difference 352
between segments.
353
354
M AN US CR IP T
AC CE PT ED
Food consumption 355
In Finland, there were differences between segments in the consumption of all the studied food 356
categories (Figure 4). The “Rational” segment consumed healthy foods more frequently than the 357
other segments and the “Struggling” segment consumed superfoods more frequently than the other 358
segments. The “Susceptible” and “Struggling” segments consumed more unhealthy snacks, 359
convenience foods and sweets than the “Rational” or “Easy-going” segments. The “Rational” and 360
“Easy-going” segments consumed alcoholic beverages more frequently than the “Susceptible” or 361
“Struggling” segments, and the “Rational” and “Struggling” segments consumed low fat dairy 362
products more frequently than the “Easy-going” or “Susceptible” segments.
363
In Germany, the segments differed with regard to the consumption of superfoods, unhealthy snacks 364
and convenience foods and healthy foods. The “Rational” segment consumed healthy foods more 365
frequently than the other segments. The “Struggling” segment consumed superfoods the most 366
frequently and healthy foods the least frequently. The “Easy-going” segment consumed superfoods 367
the least frequently. The “Susceptible” and “Struggling” segments consumed more unhealthy 368
snacks and convenience foods than the “Rational” or “Easy-going” segments did.
369
M AN US CR IP T
AC CE PT ED
370
371
Figure 4 Food consumption habits of the Finnish and German segments. Vertical bars represent 372
factor points with standard deviations.
373
Discontent with eating habits 374
Figure 5 shows the average discontent with the respondents’ own and their families’ eating habits 375
among each segment. In both the Finnish and German data the average discontent was 2.2 ± 0.8 on 376
a scale from 1 to 5. The segments differed in their level of discontent. The “Easy-going” and 377
“Rational” segments reported the least discontent, whereas the “Susceptible” and “Struggling”
378
segments experienced more discontent in both Finland and Germany.
379
M AN US CR IP T
AC CE PT ED
380
381
Figure 5 Discontent with one’s own or family’s eating habits on a scale from 1 to 5. Bars represent 382
means with standard deviations. Different letters above bars indicate a statistically significant 383
difference between segments.
384
385
Summary of the segment characteristics 386
Approximately one third of the respondents fell into the ”Easy-going” segment, a group of people 387
that had low scores for all three tendencies of eating behaviour, in both the Finnish and German 388
samples. In the Finnish sample, the proportion of men in this group was higher than that of women.
389
The “Easy-going” segment indicated experiencing negative emotions less frequently than the 390
“Susceptible” or “Struggling” segments. In comparison to the other segments, the “Easy going” did 391
not have a clearly characteristic food consumption pattern (Figure 4). However, they consumed less 392
foods belonging to the category of healthy foods than the “Rational” segment. The “Easy-going”
393
segment were still content with their eating habits. They tended to use expressive support seeking or 394
avoidance as coping strategies less frequently than the other segments.
395
The “Rational” group covered approximately one third of the studied samples. They differed from 396
the other segments in their relatively high scores for cognitive restraint and low scores for 397
emotional eating and uncontrolled eating. As in the “Easy-going” group, the proportion of men was 398
M AN US CR IP T
AC CE PT ED
higher than that of women in the Finnish sample. Older people were more prevalent in the 399
“Rational” segment than in the other segments. Compared to the other groups, they experienced 400
more positive emotions and vitality. The food consumption of this group appeared to be healthy, as 401
was indicated by their relatively frequent consumption of foods in the category of healthy foods and 402
less frequent consumption of foods in the categories of unhealthy snacks or convenience foods. The 403
“Rational” group were also content with their eating habits and tended to use active coping 404
strategies instead of avoidance.
405
The “Susceptible” segment represented the opposite of the “Rational” segment on the basis of their 406
eating styles: the group had high scores for emotional eating and uncontrolled eating, whereas the 407
scores for cognitive restraint were low. Approximately one fifth of the respondents fell into this 408
group in both samples. In the Finnish sample, the proportion of women in this group was higher 409
than that of men. The “Susceptible” group tended to experience negative emotions more frequently, 410
and positive emotions and vitality less frequently, compared to the other groups. The food choices 411
of this group appeared to be relatively unhealthy, as indicated by the consumption of foods 412
belonging to the categories of unhealthy snacks, convenience foods and sweets. The “Susceptible”
413
group experienced discontent with their eating habits and tended to use avoidance type coping 414
instead of applying active coping strategies.
415
The “Struggling” group, comprising approximately one fifth of the respondents, had relatively high 416
scores for each of the three tendencies of eating behaviour. In both countries, the group members 417
were more likely to be women than men, and were younger than the members of other segments.
418
The members of this group experienced negative emotions more often than the members of the 419
“Easy-going” or “Rational” groups. They consumed foods in the categories of unhealthy snacks and 420
convenience foods relatively often, but, conversely, they seemed to be the most keen to consume 421
foods categorized as superfoods. They also indicated discontent with their own eating habits. In 422
M AN US CR IP T
AC CE PT ED
terms of coping strategies, this group indicated a tendency to apply all of the addressed strategies;
423
active coping, expressive support seeking and avoidance.
424
M AN US CR IP T
AC CE PT ED
5. Discussion 425
426
The current study identified eating style segments based on eating behaviour tendencies and defined 427
the segments using well-being indicators, coping strategies, food consumption and contentment 428
with eating. The information obtained about the segments and their characteristics can be exploited 429
when targeting interventions and products.
430
Four eating-style segments were revealed: “Easy-going”, “Rational”, “Susceptible” and 431
“Struggling”. Emotional eating and uncontrolled eating were more relevant than cognitive restraint 432
in differentiating the groups in terms of well-being, coping strategies and contentment with eating.
433
Two of the identified segments, “Easy-going” and “Rational”, with low scores for these tendencies, 434
were rather similar in perceiving more positive emotions, having more adaptive coping strategies, 435
and experiencing more contentment with eating. The other two segments, “Struggling” and 436
“Susceptible”, with high scores for emotional eating and uncontrolled eating, were similar in their 437
tendency to experience more negative emotions and discontentment with eating, and to apply 438
avoidance as a coping strategy. These results are consistent with earlier studies on the concept of 439
disinhibition of eating, a concept that overlaps with emotional eating and uncontrolled eating 440
(Karlsson et al., 2000). Disinhibition of eating has been shown to interrelate with low self-esteem 441
and poor psychological health (Bryant, King, & Blundell, 2007).
442
The literature reports contradictory results concerning the links between cognitive restraint and 443
food intake or weight management (Anglé et al., 2009; De Lauzon-Guillain et al., 2006; De Lauzon 444
et al., 2004; Elfhag & Linné, 2005; Jaakkola et al., 2013; Keränen, Strengell, Savolainen, &
445
Laitinen, 2011). The current study found that the “Rational” group, which had high scores for 446
cognitive restraint but low scores for emotional eating and uncontrolled eating, appeared to be 447
doing well regarding food choices and life in general. By contrast, the “Struggling” segment, with 448
high scores not only for cognitive restraint but also for emotional eating and uncontrolled eating, 449
M AN US CR IP T
AC CE PT ED
showed difficulties in their food choices and psycho-behavioural factors. This finding indicates that 450
the interactions of cognitive restraint with uncontrolled eating and emotional eating could explain 451
the previous contradictory results. A previous study by Haynes et al. nicely supports the present 452
finding (Haynes, Lee, & Yeomans, 2003). They found that people who have high restraint but low 453
disinhibition do not change their eating habits in a stressful situation, whereas a group with both 454
high restraint and high disinhibition increase their intake in response to stress. Thus, it appears that 455
high restraint is a beneficial feature when it is not accompanied by high tendencies for emotional 456
eating and uncontrolled eating.
457
The holistic approach to segmentation applied in this study offers new possibilities for healthcare 458
and food companies to profile individuals based on eating styles and to provide them with more 459
targeted interventions or products (cf. Bouhlal et al 2017). For example, the “Struggling” group 460
appears to strive for better food choices, but to struggle with their tendencies to eat in response to 461
negative emotions and thus experience loss of control over eating. Thus, interventions that aim to 462
assist in managing emotion-based eating impulses and cues originating from the external eating 463
environment could be beneficial for easing the struggle. The “Struggling” group also seek support 464
to cope with negative emotions; this indicates that they might be susceptible to both peer and 465
professional support for eating-related problems. As this group consumes unhealthy snacks and 466
convenience foods, and superfoods, but strives for better choices, they might be open to healthier 467
options in these product categories. For example, snacks or convenience foods enriched with 468
healthy ingredients could serve this purpose.
469
Similar to the “Struggling” group, the “Susceptible” group is vulnerable to internal and external 470
eating cues, but they show low cognitive restraint, indicating only modest efforts towards conscious 471
regulation of food intake. However, their level of discontent with eating implies that they are aware 472
of their problems with eating. Therefore, nutrition interventions might benefit this group. In 473
addition to the aforementioned approaches discussed in relation to the “Struggling” group, this 474
M AN US CR IP T
AC CE PT ED
group could benefit from actions that support conscious regulation of food intake. Likewise, 475
development of enriched convenience food products might be one way to support this group 476
towards healthier eating.
477
In contrast to the “Struggling” and “Susceptible” groups, internal and external eating cues do not 478
appear to affect the “Easy-going” group. Moreover, they do not experience negative emotions or 479
discontent with eating frequently. Therefore, it appears that although this group might benefit from 480
diet improvements (as they do not regularly consume healthy foods), they might not feel the need 481
for changes. Thus, interventions that do not require strong initial motivation, but rather offer 482
opportunities and triggers to healthy behaviour, would be suitable for the “Easy-going” group.
483
Interventions, especially those targeted at the automatic system modulating behaviour, which do not 484
require cognitive capacity, such as choice architecture interventions (Bucher et al., 2016), could be 485
a potentially useful way to support healthy food choices among this group.
486
In the “Struggling” and “Susceptible” segments the proportion of young women was higher than in 487
the other two segments. This is in line with the notion that young women are more prone to eating- 488
related problems (Gilmour Flint et al., 2008; Power, 2016). It would be interesting to examine in 489
more detail whether this type of eating style is typical for this specific generation or more of an age- 490
related phenomenon. For example, could accumulating life experience alleviate the struggle?
491
Would members of these groups eventually become “Rational” eaters in cases where they manage 492
to develop their coping skills and become less prone to the eating cues arising from their emotions 493
and from the environment? If so, then understanding the events during their lives which enable such 494
changes in behaviour would be of considerable benefit in the development of interventions. This 495
issue could be studied in a longitudinal setting in which the dynamics of eating behaviour could be 496
observed throughout a lifetime.
497
M AN US CR IP T
AC CE PT ED
Despite the fact that both the original TFEQ (Stunkard & Messick, 1985) and the revised and 498
shortened version, TFEQ-R18 (Karlsson et al., 2000), were developed for obese populations, these 499
questionnaires have later been applied to studying the eating behaviour of populations with varying 500
weight categories. The TFEQ-R18 questionnaire has been concluded to be valid, for example, in 501
general populations (De Lauzon et al., 2004) and in young females (Anglé et al., 2009). However, 502
some previous studies have reported, similar to us, problems in the factor structure of the TFEQ- 503
R18 (Chong et al., 2016; De Lauzon et al., 2004) and the TFEQ-21, (TFEQ-R18 with three 504
additional items for emotional eating) (Cappelleri et al., 2009). In accordance with our results, the 505
reported problems suggest that not all the items for cognitive restraint reliably reflect that 506
dimension. As pointed out by Capelleri et al (2009), this phenomenon might be due to cultural 507
reasons. Furthermore, it was notable in our results that emotional eating and uncontrolled eating 508
appeared together with the discovered eating style segments, suggesting that they are reflecting the 509
same phenomenon. In the original Three-Factor Eating Questionnaire by Stunkard & Messick 510
(1985), these two tendencies were in fact generally included within a single factor called 511
disinhibition. Indeed, strong correlations have been reported between both emotional eating and 512
uncontrolled eating and the original disinhibition scale (Karlsson et al., 2000). Consequently, both 513
the previous and present findings suggest that the emotional and uncontrolled eating scales both 514
reflect the same kind of phenomenon, a dysfunctional eating tendency, whereas cognitive restraint 515
is not necessarily an indication of any problem.
516
The current study has limitations. Firstly, the scales to measure discontent with eating habits and 517
those to measure food consumption frequencies were not validated scales from previous studies.
518
The item measuring discontent with eating habits was ambiguous, including both discontent with 519
both personal and family eating habits. The ambiguity of this item might have distorted the results.
520
Second, the current study was based on respondents’ self-reports. In general, respondents tend to 521
under-report unwanted behaviors, such as consumption of unhealthy foods, experiences of negative 522
M AN US CR IP T
AC CE PT ED
emotions or usage of avoidance- or denial-based coping strategies. Thus, the results should not be 523
interpreted as factual measures of food consumption or emotions, but as individuals’ subjective 524
notions. However, in addition to information e.g. about food consumption, information about 525
individuals’ own interpretations is also valuable. Third, despite the stratified sampling, the Finnish 526
sample was not fully concordant with the population characteristics. In particular, respondents over 527
60 years old were under-represented in the sample. This under-representation may be because daily 528
use of the internet is considerably less common in age groups over 65 years than in other age groups 529
in Finland (Official Statistics of Finland, 2015). This slight bias towards a younger population in the 530
sample should be considered when evaluating the results of the Finnish sample.
531
In conclusion, the results of this study show that eating style segments based on population 532
sampling can be identified using logical interactions with well-being indicators. The findings 533
regarding eating styles, and their reflections in well-being and coping strategies, provide new 534
possibilities for healthcare and society to offer more targeted interventions or products.
535
536
Funding 537
This work was supported by Tekes – the Finnish Funding Agency for Innovation under the project 538
“Vital Selfie” (no. 2726/31/2014).
539
M AN US CR IP T
AC CE PT ED
6. References 540
541
Anglé, S., Engblom, J., Eriksson, T., Kautiainen, S., Saha, M.-T., Lindfors, P., … Rimpelä, A. (2009). Three 542
factor eating questionnaire-R18 as a measure of cognitive restraint, uncontrolled eating and emotional 543
eating in a sample of young Finnish females. The International Journal of Behavioral Nutrition and 544
Physical Activity, 6, 41. https://doi.org/10.1186/1479-5868-6-41 545
Ball, K., & Lee, C. (2000). Relationships between psychological stress, coping and disordered eating: A 546
review. Psychology & Health, 14(November 2014), 1007–1035.
547
https://doi.org/10.1080/08870440008407364 548
Boals, A., Vandellen, M. R., & Banks, J. B. (2011). The relationship between self-control and health: The 549
mediating effect of avoidant coping. Psychology & Health, 26(8), 1049–1062.
550
https://doi.org/10.1080/08870446.2010.529139 551
Bouhlal, S., McBride, C., Trivedi, N., Agurs-Collins, T., & Persky, S. (2017). Identifying eating behavior 552
phenotypes and their correlates: A novel direction toward improving weight management interventions.
553
Appetite, 111, 142–150. https://doi.org/10.1016/j.appet.2016.12.006 554
Bryant, E. J., King, N. A., & Blundell, J. E. (2007). Disinhibition: its effects on appetite and weight 555
regulation. Obesity Reviews, 9(5), 409–419. https://doi.org/10.1111/j.1467-789X.2007.00426.x 556
Bucher, T., Collins, C., Rollo, M. E., McCaffrey, T. A., De Vlieger, N., Van der Bend, D., … Perez-Cueto, 557
F. J. A. (2016). Nudging consumers towards healthier choices: a systematic review of positional 558
influences on food choice. British Journal of Nutrition, (9), 1–12.
559
https://doi.org/10.1017/S0007114516001653 560
Calvo, D., Galioto, R., Gunstad, J., & Spitznagel, M. B. (2014). Uncontrolled eating is associated with 561
reduced executive functioning. Clinical Obesity, 4(3), 172–9. https://doi.org/10.1111/cob.12058 562
Cappelleri, J. C., Bushmakin, A. G., Gerber, R. A., Leidy, N. K., Sexton, C. C., Lowe, M. R., & Karlsson, J.
563
(2009). Psychometric analysis of the Three-Factor Eating Questionnaire-R21: results from a large 564
diverse sample of obese and non-obese participants. International Journal of Obesity, 33(6), 611–620.
565
https://doi.org/10.1038/ijo.2009.74 566
Chong, M. F.-F., Ayob, M. N. M., Chong, K. J., Tai, E.-S., Khoo, C. M., Leow, M. K.-S., … Khoo, E. Y.-H.
567
(2016). Psychometric analysis of an eating behaviour questionnaire for an overweight and obese 568
Chinese population in Singapore. Appetite, 101, 119–124. https://doi.org/10.1016/j.appet.2016.03.005 569
Cornelis, M. C., Rimm, E. B., Curhan, G. C., Kraft, P., Hunter, D. J., Hu, F. B., & Van Dam, R. M. (2014).
570
Obesity susceptibility loci and uncontrolled eating, emotional eating and cognitive restraint behaviors 571
in men and women. Obesity, 22(5), 135–141. https://doi.org/10.1002/oby.20592 572
Costello, A. B. A., & Osborne, J. J. W. (2005). Best practices in exploratory factor analysis: Four 573
recommendations for getting the most from your analysis. Practical Assessment, Research &
574
Evaluation, 10(7). https://doi.org/10.1.1.110.9154 575
De Lauzon-Guillain, B., Basdevant, A., Romon, M., Karlsson, J., Borys, J. M., & Charles, M. A. (2006). Is 576
restrained eating a risk factor for weight gain in a general population? American Journal of Clinical 577
Nutrition, 83(3), 132–138.
578
De Lauzon, B., Romon, M., Deschamps, V. V., Lafay, L., Borys, J.-M., Karlsson, J., … Fleurbaix Laventie 579
Ville Sante (FLVS) Study Group. (2004). The Three-Factor Eating Questionnaire-R18 is able to 580
distinguish among different eating patterns in a general population. Nutritional Epidemiology, 134(9), 581
2372–2380. https://doi.org/134/9/2372 [pii]
582
