Women's social eating environment and its associations with dietary behavior and weight management

Appetite 110 (2017) 86e93

Contents lists available at ScienceDirect

Appetite journal homepage: www.elsevier.com/locate/appet

Women's social eating environment and its associations with dietary behavior and weight management € tteli*, Michael Siegrist, Carmen Keller Sonja Mo Department of Health Sciences and Technology, ETH Zurich, Switzerland

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 July 2016 Received in revised form 7 December 2016 Accepted 12 December 2016 Available online 13 December 2016

As an unhealthy social eating environment is considered a risk factor for obesity, this study aimed to examine women's regular eating networks and the extent to which diet-related variables were associated with those of their regular eating companions. In Study Part I (N ¼ 579), an egocentric network approach was used to investigate women's perceptions of their eating networks. In Study Part II (N ¼ 262), the participants' most important eating companions responded to a similar survey, and the corresponding answers were matched. The results showed that women shared their meals most frequently with spouses and other family members. Women who dined more often with healthy eaters reported on average a higher diet quality and a lower body mass index (BMI), which were also significant after controlling for individual factors. Study Part II expanded these results by showing that different diet-related factors such as diet quality, eating styles and BMI were correlated between women and their most important eating companions (r ¼ 0.16e0.30, p < 0.05). Moreover, an actorepartner interdependence model revealed that a higher diet quality of the eating companions was associated with a lower BMI in women, controlled for their own eating behavior (b ¼ 0.45, p < 0.05). This study showed similarities and interdependence between women's dietary behavior and body weight and those of their regular eating companions. This might indicate that regular eating networks have a shared understanding of what constitutes a normal diet, which might be an important factor to consider in the promotion of healthy eating. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Eating companion Eating behavior Diet quality Body weight Weight management Social environment

Given the global life expectancy of 71 years (WHO, 2015), adults eat about 58,035 main meals comprising breakfast, lunch, and dinner in the course of their lives. The majority of people eat at least one or two of these meals in the company of others, primarily their spouses or other family members, depending on their age, work, and housing situation (Sobal & Nelson, 2003). Due to the frequency of eating with close relations, people's eating behavior is likely similar to that of their regular eating companions. Consequently, an unhealthy social eating environment might be a risk factor for obesity (Higgs & Thomas, 2016). Therefore, this study investigated with whom women in Switzerland most frequently shared their meals and to what extent the characteristics of their regular eating companions were associated with their dietary behavior and weight management to identify social promoters of obesity. Women were the focus of this study because compared to men,

* Corresponding author. Department of Health Sciences and Technology, ETH Zurich, Universitaetstrasse 22, 8092 Zurich, Switzerland. €tteli). E-mail address: [email protected] (S. Mo http://dx.doi.org/10.1016/j.appet.2016.12.014 0195-6663/© 2016 Elsevier Ltd. All rights reserved.

they are generally motivated to adhere to a healthy diet (Leblanc, Begin, Corneau, Dodin, & Lemieux, 2015) and are more involved in controlling their body weight (Rolls, Fedoroff, & Guthrie, 1991; Wardle et al., 2004). Nevertheless, adherence to dietary recommendations is low for both women and men in Switzerland (de Abreu et al., 2013), and about 32% of women are overweight or obese (BFS, 2012). Thus, other factors, such as the social environment, might be associated with women's eating behavior over and above individual factors (Story, Kaphingst, Robinson-O'Brien, & Glanz, 2008). Indeed, previous studies have shown that perceptions about what and how much significant others eat are predictors of people's own eating behaviors (Ball, Jeffery, Abbott, McNaughton, & Crawford, 2010; Pelletier, Graham, & Laska, 2014; Robinson, Blissett, & Higgs, 2013), while associations are stronger for closer relationships (Barclay, Edling, & Rydgren, 2013). Additionally, longitudinal social network analyses have demonstrated that certain eating patterns converge among spouses, other family members, and friends over the years (Pachucki, Jacques, & Christakis, 2011; de

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

la Haye, Robins, Mohr, & Wilson, 2013). Social convergence has been also established for other health behaviors, such as weight gain (Christakis & Fowler, 2007; Trogdon, Nonnemaker, & Pais, 2008), leading to the conclusion that people's health might be interconnected among their social networks (Smith & Christakis, 2008). However, only a few studies have examined aspects of both eating behavior and body weight together although unhealthy eating patterns are known as important predictors of obesity (Malik, Willett, & Hu, 2013; Sundararajan, Campbell, Choi, & Sarma, 2014). From a social network perspective, three different mechanismsdsocial influences, homophily, and shared environmentsdcan lead to similarities in eating behaviors, which act jointly and can only be disentangled by sophisticated longitudinal network analyses (Higgs & Thomas, 2016; Valente, 2010). More precisely, a substantial body of research has indicated that eating behavior is strongly influenced by other people in various ways (Higgs & Thomas, 2016). For instance, experimental research have found robust effects of social modeling, which describes the phenomenon of people using their co-eaters as immediate reference values for their own eating behavior (Cruwys, Bevelander, & Hermans, 2015; Vartanian, Spanos, Herman, & Polivy, 2015). This means that a person is likely to eat more or to select a certain type of food when his or her eating companion is doing so. For instance, people are less likely to choose healthy foods in the presence of an unhealthy eating companion (Robinson & Higgs, 2013). Generally, people tend to follow social norms about what and how much is appropriate or normal to eat, especially when they identify themselves with the norm referent (Higgs & Thomas, 2016). Social norms are kinds of implicit cues set by the behaviors of significant others, with the power to affect people's food choices (Higgs, 2015). In the context of regular eating networks, such as family meals, people aim to negotiate social norms and shared values through which they identify themselves as a unique group (Bove, Sobal, & Rauschenbach, 2003; Giacoman, 2016). Social norms then constitute an important force to keep up eating habits and structures of regularly shared meals, for instance, by defining the place, time, and components of a typical meal (Giacoman, 2016). Besides social influences, two other mechanisms might play a role in explaining the similarities in eating patterns among regular eating companions. First, similar individuals more likely tend to form relationships with one another, which is called homophily (McPherson, SmithLovin, & Cook, 2001). For example, women and their spouses may have had pre-existing similar attitudes and behaviors toward healthy eating before they dined regularly together. Second, a shared environment can lead to similar eating patterns. For instance, people who are exposed to the same food supply (e.g., dine in the same type of restaurant) make similar food choices (Barclay et al., 2013; Cohen-Cole & Fletcher, 2008; Sobal & Hanson, 2014). Based on social norms, pre-existing similarities, and shared food environments, it can be assumed that women and their regular eating companions have similar eating behaviors, which might result in similar body weights. Therefore, this study investigated for the first time the extent to which women's dietary behavior and body weight are associated with those of their regular eating companions. More precisely, these authors investigated different factors of healthy eating such as nutrition knowledge, the goal of a healthy diet, diet quality, eating styles, body mass index and shared eating frequency. Insights about how women's social eating environment is related to their dietary behavior and weight management might contribute to a better understanding of how healthy eating and body weight can be promoted more effectively. Additionally, it might help clarify why some studies found a positive association between regularly shared meals (such as family meals)

87

and healthier eating patterns and a lower body weight, whereas other studies did not (Fulkerson, Larson, Horning, & NeumarkSztainer, 2014; Sobal & Hanson, 2014). This study included two parts. Study Part I investigated women's regular eating networks and whether these were associated with women's dietary behavior and body weight. Study Part 2 examined the interrelationship of dietary behavior and body weight between women and their most important eating companions. 1. Study Part I The first part of the study applied an egocentric network approach (Valente, 2010) to investigate the participants' regular eating networks and their associations with the participants' diet quality and body mass index (BMI). A postal survey was conducted with a random sample of female adults from the general population in the German-speaking part of Switzerland. The participants answered the survey consisting of questions about themselves and the people with whom they most often ate their main meals. The outcome measures were the participants' diet quality and BMI as main determinants of health (Popkin, Kim, Rusev, Du, & Zizza, 2006). The independent variable was a calculated healthy-eating score reflecting whether the participants ate more often with healthy or unhealthy eaters. Additionally, the following control variables were assessed: age, nutrition knowledge, and motivation (goal of achieving a healthy diet and body weight) as prerequisites for behavioral changes (Baranowski, Cullen, Nicklas, Thompson, & Baranowski, 2003) and eating styles (restrained, emotional, and external eating) as influencing factors on food choices and body weight (Keller & Siegrist, 2015a). In line with previous research, it was expected that the participants ate most often with their spouses and other family members (Sobal & Nelson, 2003). It was also hypothesized that the participants who dined more often with healthy eaters would have a higher diet quality and a lower BMI than the participants who had meals more frequently with unhealthy eaters. 1.1. Methods 1.1.1. Participants and procedure A cross-sectional study design was used in which survey forms were mailed to 2000 randomly selected addresses with female or family names in the telephone directory (N ¼ 1860 valid addresses). Each accompanying letter was addressed to a female person over 18 years old whose birthday was coming next, informing the potential participant about the study's aim and requesting her to complete the questionnaire. The participants received no financial compensation, but the enclosed return envelopes were preaddressed and prepaid. After three weeks, each nonresponder received a reminder with an additional questionnaire. In total, 579 questionnaires were returned (N ¼ 31.1%). Twelve cases had to be excluded from further analyses because of double participation (n ¼ 2), missing values over 50% (n ¼ 9), and a lack of all personal data (n ¼ 1). Of the remaining 567 respondents, 65 (11.5%) indicated having no regular eating companions, which was significantly associated with older age (r ¼ 0.21, p < 0.001) but not with other study variables of interest. For the purpose of this study, the participants with no eating companions were excluded. Thus, the final sample consisted of 502 female adults between 19 and 95 years old (M ¼ 56.97, SD ¼ 15.26). Their educational levels ranged from supplementary school (8.8%), vocational or university preparation school (42.7%), and higher vocational school (27.3%) to a university degree (20.7%). The missing values of all study variables were between 0.4 and 4.2%. This study was approved by the ETH Ethics Committee, Zurich, Switzerland.

88

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

1.1.2. Individual measures Diet quality. A diet-quality score was calculated as a gradient of the participants' adherence to a healthy diet based on six validated food frequency questions that have been proven to be indicators of healthy or unhealthy eating behaviors (Hartmann, Siegrist, & van der Horst, 2013; Malik et al., 2013). These items included weekly portions of vegetables/salad (Mdn ¼ 20.0, IQR ¼ 16.0), fruits excluding juice (Mdn ¼ 10.0, IQR ¼ 9.0), sweets (Mdn ¼ 4.0, IQR ¼ 5.0), and sugar-sweetened beverages (68.0% zero portions), and the frequency of consuming fast food (85.8% rarely or never) and prepared food (76.4% rarely or never) per week. One portion was defined as a handful of vegetables/salad or fruits, one glass of beverage, or one piece of pastry. The medians of the six food items were used as cutoff values; 1 point was assigned if the participants' consumption of vegetables/salad and fruits was above the median, and 0 point was given if their consumption was below the median. Regarding unhealthy foods, 1 point was assigned if the participants' consumption of sweets, sugar-sweetened beverages, fast food, and prepared food was below the median; 0 point was recorded if their consumption was above the median. Thus, the diet-quality score ranged from 0 (proxy for unhealthy eating) to 6 (proxy for healthy eating) (M ¼ 3.74, SD ¼ 1.36; range ¼ 0e6). BMI. The participants' BMI was calculated by dividing selfreported weight (kg) by self-reported height2 (m2). The mean BMI was 23.89 kg/m2 (SD ¼ 4.79), ranging from 15.06 to 55.17. Nutrition knowledge. For each participant, a sum score was calculated based on the seven items of the Rasch-based and vali€ tteli, Barbey, Keller, Bucher, & Siegrist, 2016) dated PKB-7 scale (Mo that measured practical knowledge about balanced meals (M ¼ 5.00, SD ¼ 1.35; range ¼ 0e7). A typical item of this scale was as follows: “Does a portion of spinach and ricotta ravioli with basil pesto contain the recommended serving of vegetables per meal?”. Goal of achieving a healthy diet and body weight. The participants' goal of adhering to a healthy and balanced diet and maintaining or achieving a healthy body weight was assessed with the first strategy of the weight management strategies inventory (WMSI). The WMSI consisted of three items (Keller & Siegrist, 2015b), as follows: “It is my goal to eat sufficient fruits and vegetables every day.” “It is my goal to adhere to a balanced diet.” “It is my goal to maintain or achieve a normal and healthy body weight.” The answers were given on a rating scale ranging from 1 ¼ not agree at all to 7 ¼ fully agree (M ¼ 5.92, SD ¼ 1.24; range ¼ 1e7; a ¼ 0.80). Eating styles. The three eating stylesdemotional (response to negative emotions), external (response to environmental cues), and restrained eating (cognitive restriction)dwere assessed with a short German version (Grunert, 1989) of the Dutch Eating Behavior Questionnaire (Van Strien, Frijters, Bergers, & Defares, 1986). The answers were given on rating scales ranging from 1 ¼ never to 7 ¼ very frequently (emotional eating: M ¼ 1.98, SD ¼ 0.91; range ¼ 1e5; a ¼ 0.95; external eating: M ¼ 2.79, SD ¼ 0.66; range ¼ 1e4.8; a ¼ 0.80; and restrained eating: M ¼ 3.10, SD ¼ 0.90; range ¼ 1e5; a ¼ 0.85). 1.1.3. Eating-network measures The characteristics of the eating companions were assessed with an egocentric network approach by applying a “name generator” question (Marquez et al., 2014; Marsden, 1990, 2011). The participants were asked to note the first names or initials of up to three adults with whom they dined most frequently. Subsequently, a set of questions about each eating companion was asked, specifying the following characteristics: (a) the relationship between the participant and her eating companion (partner, another family member [such as the participant's parents, siblings, or grown-up children], friend, co-worker, neighbor, or other roles); (b) the frequency of shared meals (daily, 4e6 times a week, 1e3 times a week,

1e3 times a month, or seldom); (c) her eating companion's gender (male ¼ 0, female ¼ 1); (d) her eating companion's age (years); and (e) her judgment on whether her eating companion adhered to a healthy diet (healthy eater ¼ 1, unhealthy eater ¼ 1). Information about each eating companion was provided from the participant's point of view and therefore described her perception of her eating network. Note that the women's regular eating networks comprised adults only; dining with children and adolescents was not questioned. 1.1.4. Data analysis The eating frequency was recoded as daily ¼ 7, 4e6 times a week ¼ 5, 1e3 times a week ¼ 2, 1e3 times a month ¼ 0.5, and seldom ¼ 0. The eating-network size was calculated by summing up the number of reported eating companion(s), comprising one to three persons since the participants with no eating companions were excluded in advance (M ¼ 2.07, SD ¼ 0.91). To evaluate the extent to which the participants dined with healthy or unhealthy eaters, a healthy-eating score was calculated for each participant's eating network. The value for a healthy-eating (1) or an unhealthyeating companion (1) was multiplied by the corresponding eating frequency with this person (7, 5, 2, 0.5, or 0 times per week). Subsequently, these products were summed up for all reported eating companions of each participant and then divided by the eating network size. Thus, for all participants, the score could range from 7 to 7. Two regression analyses were conducted to predict the participants' diet quality and BMI through the calculated healthy-eating score of the eating networks, controlled for individual factors, such as age, knowledge, motivation, and eating styles. Other associations among variables were determined with Pearson correlation coefficients. Statistical tests were performed, using the IBM SPSS version 22 for Mac (SPSS Inc.). 1.2. Results and discussion 1.2.1. Dietary behavior and weight management Half of the participants (51.8%; n ¼ 257) did not follow the Swiss dietary recommendation of consuming three portions of vegetables and salad per day, while 29.7% (n ¼ 148) consumed one or more portions of sweets per day. Compliance with dietary recommendations was even worse in previous studies (de Abreu et al., 2013). Furthermore, 33.3% (n ¼ 166) of the participants had a calculated BMI  25 kg/m2, corresponding to the average proportion of overweight women (32.0%) in Switzerland (BFS, 2012). 1.2.2. Eating networks Of all participants, 38.4% (n ¼ 193) shared their meals with one eating companion, 16.1% (n ¼ 81) with two, and 45.5% (n ¼ 228) with three. The participant age was negatively associated with the eating-network size (r ¼ 0.21, p < 0.01) and the mean eating frequency (r ¼ 0.15, p < 0.01). This indicated that older people had less regular eating companions and dined less often with them. In line with previous findings (Sobal & Hanson, 2014; Sobal & Nelson, 2003), the participants shared on average five meals per week primarily with their spouses and other family members. Table 1 summarizes the characteristics of the eating companions. It shows that the first-named eating companions were on average the most important ones, namely, the persons with whom the participants dined most often and had the closest relationships. 1.2.3. Association of eating networks with diet quality and BMI The correlation matrix in Table 2 shows that both diet quality and BMI were correlated with the other study variables in the expected directions. Two regression analyses were conducted to examine whether the participants' diet quality and BMI could be

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

89

Table 1 Characteristics of participants' regular eating companions.

Relationship, n (%) Romantic partner/spouse Another family member Friend Neighbor Co-worker Other roles Eating frequency Daily, n (%) 5 times per week, n (%) 2 times per week, n (%) 0.5 times per week, n (%) Seldom Age (years), M (±SD) Age range Gender Male, n (%) Female, n (%) Eating behavior Healthy eater, n (%) Unhealthy eater, n (%)

Eating companion 1 (n ¼ 480)

Eating companion 2 (n ¼ 309)

Eating companion 3 (n ¼ 250)

46.0 28.5 12.5 3.8 5.4 2.9

6.8 46.3 23.0 3.2 17.2 2.6

3.6 38.4 26.0 3.6 22.8 5.2

41.5 16.7 21.5 14.2 5.8 52.39 (16.61) 18e97

9.7 10.0 32.4 34.3 13.3 50.84 (18.55) 18e96

4.8 7.2 27.2 38.4 22.0 50.45 (18.13) 18e91

61.0 37.5

30.7 68.9

38.8 59.2

42.7 56.5

47.2 50.8

46.0 52.0

Note. N ¼ 502 female participants. Each participant named at least one eating companion. Percentages do not add up to 100 due to missing data.

predicted by the calculated healthy-eating scores of their eating networks. After controlling for individual factors, such as age, nutrition knowledge, motivation, and eating styles, the healthyeating score of the eating network showed as a weak but significant predictor in both regression analyses (Table 3). This means that, over and above individual factors, the participants who dined more often with healthy-eating companions had a higher diet quality and a lower BMI than the participants who shared meals more frequently with unhealthy-eating companions. An additional analysis revealed the eating frequency with healthy eaters was an important factor to consider since the percentage of healthy eaters and the eating frequency alone were not significant predictors of the participants' diet quality. Overall, these results confirmed that descriptive eating norms are significant predictors of people's own eating behaviors (Ball et al., 2010; Pelletier et al., 2014; Robinson et al., 2013). The results contribute to the literature by showing that the perceived eating behavior of the eating companions was also a significant predictor of the participants' body weight. As the results of Study Part I were based on an egocentric network design, the characteristics of the eating companions (e.g., whether or not they adhered to a healthy diet) reflected the participants' perceptions or social norms. Hence, the participants conceivably perceived themselves as being more similar to their eating companions than they actually were. Therefore, Study Part II expanded the results of Study Part I by analyzing the associations

between the participants and their first-named eating companions, using two separate data sources.

2. Study Part II Study part II involved two main goals. First, it was examined to what extent eating behavior and body weight were correlated between the participants and their most important eating companions. As those were the persons with whom the participants most frequently ate and had close relationships, associations were predicted for diet quality, nutrition knowledge, motivation, eating styles, and BMI. For these purposes, the answers of the participants and their first-named eating companions in Study Part I were matched. Additionally, it was explored whether similarities in diet quality were stronger for older (versus younger) spouses, who reflected stable eating companions over the years. As longitudinal social network analyses have shown that some eating patterns converge over time among closely related people (Pachucki et al., 2011; de la Haye et al., 2013), an age effect might indicate social influence although causal mechanisms could not be truly established in a cross-sectional design. The second goal was based on the results of Study Part I and the previous literature that an unhealthy social eating environment might be a risk factor for obesity (Higgs & Thomas, 2016; Story et al., 2008). It was tested whether the participants' body weight could be predicted by their eating

Table 2 Pearson correlation coefficients between the diet-quality score, BMI and other study variables.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Diet-quality score BMI Age Nutrition knowledge Goal of achieving a healthy diet and body weight Restrained eating External eating Emotional eating Healthy-eating score of the eating network

1

2

3

4

5

6

7

8

9

e 0.12** 0.23** 0.09* 0.43** 0.19** 0.19** 0.10* 0.13**

e 0.15** 0.01 0.12** 0.05 0.14** 0.28** 0.12**

e 0.31** 0.08* 0.00 0.27** 0.09* 0.05

e 0.22** 0.15** 0.12** 0.13** 0.05

e 0.32** 0.06 0.02 0.11*

e 0.06 0.16** 0.08

e 0.49** 0.00

e 0.02

e

Note. N ¼ 460 participants included in the regression analyses (Table 3). *p < 0.05, **p < 0.01.

90

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

Table 3 Regression analyses predicting participants' diet quality and BMI by the healthy-eating score of their eating networks, controlled for individual variables. Diet quality

Individual variables Age (years) Nutrition knowledge Goal of achieving a healthy diet and body weight Restrained eating External eating Emotional eating Eating-network variable Healthy-eating score of the eating network R2 adjusted n

BMI

B

SE B

b

B

SE B

b

0.02 0.08 0.40 0.12 0.23 0.07

0.00 0.05 0.05 0.07 0.10 0.07

0.19*** 0.08 0.35*** 0.08 0.11* 0.05

0.07 0.06 0.54 0.32 0.28 1.54

0.02 0.17 0.18 0.24 0.37 0.26

0.21*** 0.02 0.14** 0.06 0.04 0.29***

0.03 0.24 463

0.02

0.08*

0.16 0.14 480

0.06

0.12**

Note. *p < 0.05, **p < 0.01, ***p < 0.001. Predictors: Scores in nutrition knowledge ranged from 0 to 6. The goal of achieving a healthy diet and body weight was measured on a scale from 1 to 7. Eating styles were measured on a scale from 1 to 5. Healthy-eating scores ranged from 7 to 7. Outcomes: Scores in diet quality ranged from 0 to 6.

companions' diet quality while controlling for the participants' own eating behavior, using the actorepartner interdependence model (APIM) (Kenny, Kashy, & Cook, 2006). 2.1. Methods 2.1.1. Participants and procedure In the same survey used in Study Part I, potential participants were asked to forward a similar questionnaire to their first-named eating companions. This second questionnaire was enclosed in a separate preaddressed and prepaid envelope, attached to the end of the primary questionnaire. Of the 502 female participants in Study 1, 52.2% (N ¼ 262) successfully forwarded the second questionnaire to their most frequent eating companions. Thus, the answers of each female participant (N ¼ 262) could be matched with those of her corresponding eating companion, according to a survey code and demographic variables, such as gender and age. There were no significant differences (all p's > 0.1) in individual variables (e.g., age, education, BMI, nutrition knowledge, goal of achieving a healthy diet and body weight, and diet-quality score) and eating-network variables (e.g., mean age, percentage of females, and percentage of healthy eaters) between the participants in Study Part I who did and did not forward the second questionnaire. Consequently, Study Part II included both the answers of the 262 female participants (subsample of Study Part I) and the independently assessed answers of their corresponding eating companions. For the sake of clarity, the term “participants” always refers to the female subsample of Study Part I, and the others are called eating companions. Note that despite the matched answers, the whole survey was conducted anonymously. The missing values of all study variables were between 0 and 3.4%. 2.1.2. Measures The BMI, the goal of achieving a healthy diet and body weight, nutrition knowledge, eating styles, diet quality, eating frequency, and relationship were assessed and recoded in the same manner as described in Study Part I for both the participants and their eating companions. Concerning the diet-quality score, gender-specific medians were used for the consumption-frequency questions because the men's mean intake amounts were lower for vegetables/salad (Mdnmale ¼ 16.0, IQR ¼ 15.0; Mdnfemale ¼ 20.0, IQR ¼ 14.0) and fruits (Mdnmale ¼ 5.0, IQR ¼ 8.0; Mdnfemale ¼ 10.0, IQR ¼ 9.0) but higher for sugar-sweetened beverages (Mdnmale ¼ 1.0, IQR ¼ 4.0; Mdnfemale ¼ 0.0, IQR ¼ 2.0). However, one eating companion did not disclose his/her gender and was thus excluded from the analyses

(final N ¼ 261). The dieting status was also determined (a current dieter: yes or no). 2.1.3. Data analysis The associations between variables were examined using correlation coefficients. To examine whether the diet-quality scores of both the participants and their regular eating companions were associated with the participants' BMI, a predictive APIM was used as depicted in Fig. 1 (Cook & Kenny, 2005). The APIM is a model of dyadic relationships that accounts for the interdependence of two socially connected individuals (in this study, the participants and their eating companions) when actor and partner effects are estimated on an outcome variable (Cook & Kenny, 2005). Table 4 shows the non-independence of BMI and diet quality between the participants and their eating companions, thus requiring appropriate statistical methods, such as a multilevel analysis. The APIM was estimated by using a multilevel approach with a two-intercept model (random intercepts and fixed slopes) for distinguishable dyads, and the predictors were mean centered (for more detailed information about the model calculation, see Kenny et al., 2006 [Chapter 7]). More precisely, four different associations between the predictor (diet quality) and the outcome (BMI) were specified (see Fig. 1) and estimated by the restricted maximum likelihood (REML) method in a sequential approach. After controlling for age, gender, and dieting status, the first model included the two actor effects predicting the participants' BMI by the participants' diet quality and the eating companions' BMI by the eating companions' diet quality. The final model included the two partner effects predicting the participants' BMI by the eating companions' diet quality and the eating companions' BMI by the participants' diet quality, simultaneously controlling for the covariates and the actor effects. The relative contribution of adding the partner effects was estimated using the likelihood ratio test by calculating the difference in the 2 log likelihood between the first and the final model, using the maximum likelihood (ML) estimation and degrees of freedom n ¼ 2 (Kenny et al., 2006). Statistical tests were performed by using the IBM SPSS version 22 for Mac (SPSS Inc.). 2.2. Results and discussion 2.2.1. Agreement between perception and self-report To verify whether each participant's perception about her most frequent eating companion was accurate, a correlation coefficient was calculated between the participants' perceptions (whether their eating companions were perceived as healthy or unhealthy

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

91

Fig. 1. Final APIM of the participants and their eating companions, according to Cook and Kenny (2005). (A) ¼ participants' actor effect, (P) ¼ participants' partner effect, (A0 ) ¼ eating companions' actor effect, (P0 ) ¼ eating companions' partner effect, Ep., EEC ¼ residuals (unexplained variance), single-headed arrows: predictive paths, double-headed arrows: correlations.

Table 4 Concordance of diet-related variables between participants and their eating companions. Participants N ¼ 261 women

Diet-quality score BMI (kg/m2) Nutrition knowledge Goal of achieving a healthy diet and body weight Restrained eating Emotional eating External eating

Eating companions N ¼ 261 (59.8% men)

Pearson correlation coefficients

M

SD

M

SD

r

3.82 23.96 5.04 6.00 3.15 1.98 2.82

1.34 4.74 1.42 1.15 0.86 0.89 0.68

3.18 25.12 4.48 5.36 2.87 1.75 2.97

1.59 4.03 1.59 1.38 0.88 0.80 0.67

0.23** 0.16* 0.31** 0.20** 0.16* 0.12 0.30**

Note. *p < 0.05, **p < 0.01. Nutrition-knowledge scores ranged from 0 to 7. The goal of achieving a healthy diet and body weight was measured on a scale from 1 to 7. Diet-quality scores ranged from 0 to 6. Eating styles were measured on a scale from 1 to 5.

eaters) and their eating companions' self-reports (calculated diet quality scores), which was r ¼ 0.35, p < 0.01. Taking into account the binary measure for the participants' perceptions, this result suggests some agreement between the participants' perceptions and their eating companions' self-evaluations. 2.2.2. Concordance in diet-related variables The degree to which the participants' eating behavior and body weight were associated with those of their most important eating companions was assessed by the correlation coefficients of two separately collected data sources. The diet-quality scores, based on six self-reported consumption-frequency questions, were calculated for both the participants and their corresponding eating companions. The concordance of other diet-related variables, such as BMI, nutrition knowledge, the goal of achieving a healthy diet and body weight, and eating styles, was also determined. As presented in Table 4, all diet-related variables, with the exception of emotional eating, were significantly correlated between the participants and their most important eating companions, with weak to modest correlation coefficients. An additional analysis examining these correlations separately for spouses and nonspouses revealed that they were in the same range for all variables except knowledge and motivation. The participants' nutrition knowledge and their goal of achieving a healthy diet and body weight were significantly correlated only with nonspouses (r ¼ 0.45, p < 0.01 and r ¼ 0.37, p < 0.01, respectively), which might be explained by the gender differences related to these variables. Additionally, age was negatively correlated with the absolute difference in the dietquality scores of the participants and their spouses (r ¼ 0.29, p < 0.01), indicating higher similarity levels in the diet quality of older eating companions. These results are in line with and expand previous studies that found similarities in some eating patterns and

body weight among closely related people, which converged over the years (Christakis & Fowler, 2007; Pachucki et al., 2011).

2.2.3. Association between diet quality and BMI After controlling for age, gender, and dieting status, the first APIM model showed significant actor effects. The participants with a higher diet quality had a significantly lower BMI; similarly, eating companions with a higher diet quality had a lower BMI, in line with the expectations based on previous research (Sundararajan et al., 2014). Table 5 presents the estimates of the multilevel analysis for both models. In the final model, the partner effects were added to test whether the participants' BMI could be predicted by the diet quality of their eating companions, simultaneously controlling for the effect of their own eating behavior. The results showed that higher scores in the diet quality of the eating companions were significantly associated with a lower BMI of the participants, and higher scores in the diet quality of the participants were also significantly associated with a lower BMI of their eating companions. This final model revealed a better fit with the data than the first model (difference in 2 log likelihood ¼ 8.55, df ¼ 2, p < 0.05), shown in Step 2 of Table 5 and depicted in Fig. 1. The presence of significant partner effects indicated that the dietary behavior and body weight of the women and their most important eating companions were interdependent (Kenny et al., 2006). However, after estimating the actor and partner effects of diet quality on BMI, a significant correlation of the BMI's residuals remained (r ¼ 0.20, p < 0.01). This means that the diet-quality scores could not fully explain the similarities in BMI between the women and their eating companions, which seems plausible because a variety of other lifestyle factors (e.g., physical activity) could influence BMI.

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

92

Table 5 Multilevel fixed effects predicting BMI by the participants' and the eating companions' diet quality. BMI Step 1

Covariates Gender Age Dieting status Actor effects Participants' diet quality Eating companions' diet quality Partner effects Participants' diet quality Eating companions' diet quality

Step 2

B

SE B

B

SE B

2.21*** 0.04*** 1.88***

0.51 0.01 0.39

2.07*** 0.06*** 1.99***

0.51 0.01 0.40

0.68** 0.48*

0.22 0.19

0.57* 0.38*

0.23 0.19

0.38* 0.45*

0.19 0.19

Note. N ¼ 243 participants and their eating companions. *p < 0.05, **p < 0.01, ***p < 0.001. Gender: (0) male, (1) female; dieting status: (0) dieter, (1) non-dieter. Diet-quality scores ranged from 0 to 6.

3. General discussion As people are embedded in social networks and shared environments, their health behaviors might also be interconnected (Cohen-Cole & Fletcher, 2008; Smith & Christakis, 2008). Regular eating networks are stable groups in which closely connected people share their meals several times a week (Sobal & Nelson, 2003). Thus, similar eating behaviors can be assumed among regular eating companions, which might also result in similar body weights. To our best knowledge, this study is the first to investigate associations in different diet-related aspects and body weight between women and their regular eating companions by applying a social network approach. As assumed and in line with previous findings (Leblanc et al., 2015; Rolls et al., 1991; de Abreu et al., 2013), women did not follow the Swiss dietary recommendations, and one-third of them were overweight although they stated being motivated toward a healthy diet and body weight. Moreover, the results confirmed previous findings that on average, people most often eat with their spouses and other family members (Sobal & Nelson, 2003) and that eating patterns and body weight tend to be similar and to converge over time among these relations (Christakis & Fowler, 2007; Pachucki et al., 2011). The results also contribute to existing literature by providing new insights into the interdependence of dietary behavior and body weight among women's regular eating networks. Study Part I showed that over and above individual factors, such as age, knowledge, and motivation, women's diet quality was higher and their body weight was lower when they dined more frequently with healthy-eating companions than with unhealthyeating companions. These results indicated that an unhealthy social eating environment might be a risk factor for the development of unhealthy eating patterns and obesity, as suggested in previous reviews (Higgs & Thomas, 2016; Story et al., 2008). Accounting for the egocentric network design used in Study Part I, which viewed all variables from the participants' perspective, Study Part II applied a more objective approach that included the separate answers of the women and their most important eating companions. The results of Study Part II confirmed and expanded those of Part I by showing that eating behavior and body weight were both correlated and interdependent between women and their regular eating companions. More precisely, women and their most important eating companions tended to be similar in diet-related factors such as diet quality and eating styles as well as in BMI. Moreover, the

estimation of the APIM revealed that a higher diet quality of the eating companions was significantly associated with a lower BMI in women, controlled for their own diet quality. This finding also occurred for the association between women's diet quality and their eating companions' BMI, indicating that a healthy or an unhealthy lifestyle was clustered among regular eating networks. In summary, the interdependence of eating behavior and body weight among regular eating companions might indicate a shared understanding of what constitutes a normal diet. Previous literature has shown evidence that regular eating companions, such as spouses and other family members, actively aim to negotiate social norms and shared values about what is perceived as appropriate food choices and intake amounts (Bove et al., 2003; Giacoman, 2016; Higgs & Thomas, 2016). Depending on these factors, regularly shared meals, such as family meals, might more or less conform to dietary guidelines although they are often generally perceived as healthy meals (Sobal & Hanson, 2014). For instance, for some eating networks, it might be natural to eat dessert after the main course, whereas other people might eat dessert only on special occasions. Therefore, to promote healthy eating behaviors among women more effectively, it seems important to identify and improve inappropriate social eating norms and values of their regular eating networks (Higgs & Thomas, 2016). Furthermore, women's most important eating companions need to be included in the prevention and treatment of obesity (e.g., nutrition counseling and family meal interventions) as social support is considered one of the most effective strategies to improve health behaviors (Smith & Christakis, 2008). Several limitations of this study have to be addressed. First, due to the cross-sectional study design, no causal conclusion can be inferred regarding the similarity levels of diet-related variables between women and their regular eating companions. Second, the response rate in the primary survey (31%) was not as large as preferable to assure a representative sampling. However, the distributions of age and educational levels were consistent with those of other studies based on survey data in the same population €tteli et al., 2016), and the overweight rate reflected that of the (Mo average female population in Switzerland (BFS, 2012). Despite the randomized survey, a self-selection bias toward people who were more interested in the topic than the general population cannot be ruled out. Third, besides nutrition knowledge and motivation for a healthy diet, other important factors (e.g., physical activity) influencing BMI were ignored. Factors about the preparation and location of regularly shared meals, as well as how long the participants had been dining together with their regular eating companions, were also not assessed, which might be interesting factors to examine in future studies. Additionally, due to the survey usability, the perceived eating behavior of the eating companions was only a binary measure (healthy versus unhealthy eaters), in contrast to the calculated diet quality scores of the participants. Forth, all variables were based on self-reports, which can be biased toward socially desirable answers. Finally, the findings cannot be generalized to the general population since only female participants were investigated. Thus, future studies are needed to examine whether similar results can also be found for men, who are generally less motivated to adopt a healthy diet. In conclusion, this study showed that over and above individual factors, women's eating behavior and body weight were interdependent with those of their regular eating companions, who were mostly their spouses and other family members. The results demonstrated that a healthy or an unhealthy lifestyle was clustered among regular eating networks, indicating that they seem to have a shared understanding of what constitutes a normal diet. Identifying and improving social eating norms and addressing eating networks instead of individuals might therefore be effective strategies to

€tteli et al. / Appetite 110 (2017) 86e93 S. Mo

promote healthy eating and a healthy body weight. Acknowledgements The authors thank Stefan Klenke who helped collecting the data. References de Abreu, D., Guessous, I., Vaucher, J., Preisig, M., Waeber, G., Vollenweider, P., et al. (2013). Low compliance with dietary recommendations for food intake among adults. Clinical Nutrition, 32(5), 783e788. http://dx.doi.org/10.1016/ J.Clnu.2012.11.022. Ball, K., Jeffery, R. W., Abbott, G., McNaughton, S. A., & Crawford, D. (2010). Is healthy behavior contagious: Associations of social norms with physical activity and healthy eating. International Journal of Behavioral Nutrition and Physical Activity, 7(1), 86. http://dx.doi.org/10.1186/1479-5868-7-86. Baranowski, T., Cullen, K. W., Nicklas, T., Thompson, D., & Baranowski, J. (2003). Are current health behavioral change models helpful in guiding prevention of weight gain efforts? Obesity Research, 11, 23Se43S. http://dx.doi.org/10.1038/ Oby.2003.222. Barclay, K. J., Edling, C., & Rydgren, J. (2013). Peer clustering of exercise and eating behaviours among young adults in Sweden: A cross-sectional study of egocentric network data. BMC Public Health, 13(1), 1e13. http://dx.doi.org/ 10.1186/1471-2458-13-784. BFS. (2012). Schweizerische Gesundheitsbefragung (SGB) [Swiss Federal Statistical Office: Swiss Health Survey (SHS)]. Neuch^ atel, Switzerland: BFS. Bove, C. F., Sobal, J., & Rauschenbach, B. S. (2003). Food choices among newly married couples: Convergence, conflict, individualism, and projects. Appetite, 40(1), 25e41. http://dx.doi.org/10.1017/S0195-6663(02)00147-2. Christakis, N. A., & Fowler, J. H. (2007). The spread of obesity in a large social network over 32 years. New England Journal of Medicine, 357(4), 370e379. http://dx.doi.org/10.1056/NEJMsa066082. Cohen-Cole, E., & Fletcher, J. M. (2008). Is obesity contagious? Social networks vs. environmental factors in the obesity epidemic. Journal of Health Economics, 27(5), 1382e1387. http://dx.doi.org/10.1016/j.jhealeco.2008.04.005. Cook, W. L., & Kenny, D. A. (2005). The actor-partner interdependence model: A model of bidirectional effects in developmental studies. International Journal of Behavioral Development, 29(2), 101e109. http://dx.doi.org/10.1080/ 01650250444000405. Cruwys, T., Bevelander, K. E., & Hermans, R. C. (2015). Social modeling of eating: A review of when and why social influence affects food intake and choice. Appetite, 86, 3e18. http://dx.doi.org/10.1016/j.appet.2014.08.035. Fulkerson, J. A., Larson, N., Horning, M., & Neumark-Sztainer, D. (2014). A review of associations between family or shared meal frequency and dietary and weight status outcomes across the lifespan. Journal of Nutrition Education and Behavior, 46(1), 2e19. http://dx.doi.org/10.1016/j.jneb.2013.07.012. Giacoman, C. (2016). The dimensions and role of commensality: A theoretical model drawn from the significance of communal eating among adults in Santiago, Chile. Appetite. http://dx.doi.org/10.1016/j.appet.2016.08.116. Grunert, S. (1989). Ein Inventar zur Erfassung von Selbstaussagen zum Ern€ ahrungsverhalten [An inventory for determination of eating behaviors through self-reporting]. Diagnostica, 35(2), 167e179. Hartmann, C., Siegrist, M., & van der Horst, K. (2013). Snack frequency: Associations with healthy and unhealthy food choices. Public Health Nutrition, 16(08), 1487e1496. http://dx.doi.org/10.1017/S1368980012003771. de la Haye, K., Robins, G., Mohr, P., & Wilson, C. (2013). Adolescents' intake of junk food: Processes and mechanisms driving consumption similarities among friends. Journal of Research on Adolescence, 23(3), 524e536. http://dx.doi.org/ 10.1111/jora.12045. Higgs, S. (2015). Social norms and their influence on eating behaviours. Appetite, 86, 38e44. http://dx.doi.org/10.1016/j.appet.2014.10.021. Higgs, S., & Thomas, J. (2016). Social influences on eating. Current Opinion in Behavioral Sciences, 9, 1e6. http://dx.doi.org/10.1016/j.cobeha.2015.10.005. Keller, C., & Siegrist, M. (2015a). Does personality influence eating styles and food choices? Direct and indirect effects. Appetite, 84, 128e138. http://dx.doi.org/ 10.1016/j.appet.2014.10.003. Keller, C., & Siegrist, M. (2015b). The weight management strategies inventory (WMSI). Development of a new measurement instrument, construct validation, and association with dieting success. Appetite, 92, 322e336. http://dx.doi.org/ 10.1016/j.appet.2015.05.037. Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York: The Guilford Press. Leblanc, V., Begin, C., Corneau, L., Dodin, S., & Lemieux, S. (2015). Gender differences

93

in dietary intakes: What is the contribution of motivational variables? Journal of Human Nutrition and Dietetics, 28(1), 37e46. http://dx.doi.org/10.1111/ jhn.12213. Malik, V. S., Willett, W. C., & Hu, F. B. (2013). Global obesity: Trends, risk factors and policy implications. Nature Reviews Endocrinology, 9(1), 13e27. http:// dx.doi.org/10.1038/Nrendo.2012.199. Marquez, B., Elder, J. P., Arredondo, E. M., Madanat, H., Ji, M., & Ayala, G. X. (2014). Social network characteristics associated with health promoting behaviors among Latinos. Health Psychology, 33, 544e553. http://dx.doi.org/10.1037/ hea0000092. Marsden, P. V. (1990). Network data and measurement. Annual Review of Sociology, 435e463. Marsden, P. V. (2011). Survey methods for network data. The SAGE handbook of social network analysis, 370e388. McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415e444. http://dx.doi.org/ 10.1146/Annurev.Soc.27.1.415. €tteli, S., Barbey, J., Keller, C., Bucher, T., & Siegrist, M. (2016). Measuring practical Mo knowledge about balanced meals: Development and validation of the brief PKB-7 scale. European Journal of Clinical Nutrition, 70(4), 505e510. http:// dx.doi.org/10.1038/ejcn.2015.173. Pachucki, M. A., Jacques, P. F., & Christakis, N. A. (2011). Social network concordance in food choice among spouses, friends, and siblings. American Journal of Public Health, 101(11), 2170e2177. http://dx.doi.org/10.2105/AJPH.2011.300282. Pelletier, J. E., Graham, D. J., & Laska, M. N. (2014). Social norms and dietary behaviors among young adults. American Journal of Health Behavior, 38(1), 144e152. http://dx.doi.org/10.5993/AJHB.38.1.15. Popkin, B. M., Kim, S., Rusev, E. R., Du, S., & Zizza, C. (2006). Measuring the full economic costs of diet, physical activity and obesity-related chronic diseases. Obesity reviews, 7(3), 271e293. http://dx.doi.org/10.1111/j.1467789X.2006.00230.x. Robinson, E., Blissett, J., & Higgs, S. (2013). Social influences on eating: Implications for nutritional interventions. Nutrition research reviews, 26(02), 166e176. http:// dx.doi.org/10.1017/S0954422413000127. Robinson, E., & Higgs, S. (2013). Food choices in the presence of 'healthy' and 'unhealthy' eating partners. British Journal of Nutrition, 109(4), 765e771. http:// dx.doi.org/10.1017/S0007114512002000. Rolls, B. J., Fedoroff, I. C., & Guthrie, J. F. (1991). Gender differences in eating behavior and body-weight regulation. Health Psychology, 10(2), 133e142. http:// dx.doi.org/10.1037/0278-6133.10.2.133. Smith, K. P., & Christakis, N. A. (2008). Social networks and health. Annual Review of Sociology, 34, 405e429. http://dx.doi.org/10.1146/ annurev.soc.34.040507.134601. Sobal, J., & Hanson, K. (2014). Family dinner frequency, settings and sources, and body weight in US adults. Appetite, 78, 81e88. http://dx.doi.org/10.1016/ j.appet.2014.03.016. Sobal, J., & Nelson, M. K. (2003). Commensal eating patterns: A community study. Appetite, 41(2), 181e190. http://dx.doi.org/10.1016/S0195-6663(03)00078-3. 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, 253e272. http://dx.doi.org/10.1146/ annurev.publhealth.29.020907.090926. Sundararajan, K., Campbell, M. K., Choi, Y.-H., & Sarma, S. (2014). The relationship between diet quality and adult obesity: Evidence from Canada. Journal of the American College of Nutrition, 33(1), 1e17. http://dx.doi.org/10.1080/ 07315724.2013.848157. Trogdon, J. G., Nonnemaker, J., & Pais, J. (2008). Peer effects in adolescent overweight. Journal of Health Economics, 27(5), 1388e1399. http://dx.doi.org/ 10.1016/j.jhealeco.2008.05.003. Valente, T. W. (2010). Social networks and health: Models, methods, and applications. Oxford University Press. Van Strien, T., Frijters, J. E., Bergers, G., & Defares, P. B. (1986). The Dutch Eating Behavior Questionnaire (DEBQ) for assessment of restrained, emotional, and external eating behavior. International Journal of Eating Disorders, 5(2), 295e315. http://dx.doi.org/10.1002/1098-108X(198602). Vartanian, L. R., Spanos, S., Herman, C. P., & Polivy, J. (2015). Modeling of food intake: A meta-analytic review. Social Influence, 10(3), 119e136. http:// dx.doi.org/10.1080/15534510.2015.1008037. Wardle, J., Haase, A. M., Steptoe, A., Nillapun, M., Jonwutiwes, K., & Bellisle, F. (2004). Gender differences in food choice: The contribution of health beliefs and dieting. Annals of Behavioral Medicine, 27(2), 107e116. http://dx.doi.org/10.1207/ S15324796abm2702_5. WHO. (2015). Health in 2015 from MDGs to SDGs. Retrieved from http://www.who. int/gho/mortality_burden_disease/life_tables/situation_trends_text/en.