Do emerging adults know what their friends are doing and does it really matter? Methodologic challenges and associations of perceived and actual friend behaviors with emerging adults’ disordered eating and muscle building behaviors

Abstract

Disordered eating and muscle building behaviors are common among emerging adults, and friends may be a particularly salient social influence. Epidemiologic research often includes questions about participants’ perceptions of their friends. A less common approach, with greater logistical challenges, is to ask for friend nominations and then survey friends about their actual behaviors. The comparability of these different approaches is unknown. This study addresses the following research questions: 1) What is the feasibility of collecting data from emerging adults’ friends in epidemiologic research? 2) Do perceptions of friends’ weight- and shape-related behaviors align with friends’ actual behaviors? and 3) Are perceptions or friends’ actual behaviors more strongly and consistently associated with emerging adults’ behaviors? Participants (N = 2383) in the EAT (Eating and Activity over Time)-2018 study in Minnesota, USA, were asked to nominate up to three friends and provide data about those friends’ weight- and shape-related behaviors (i.e. perceptions); nominated friends were invited to complete an abbreviated survey and report on their own same behaviors (i.e. actual). Among the invited friends, 191 responded and were linked to the 152 EAT 2018 participants who nominated them. Descriptive statistics, Spearman’s correlations, and logistic regression were used to address the research questions. The response rate for nominated friends was very low (9.9%), suggesting this approach may have low feasibility for epidemiologic studies of emerging adults. Emerging adults’ perceptions of their nominated friends’ weight and shape-related behaviors generally did not align well with the behaviors reported by those friends. Furthermore, analytic models found different associations between friends’ behavior and EAT 2018 participant behaviors, depending on the measure of friends’ behavior used (perceived or actual). Careful consideration of the pros and cons of each study design is essential to build an evidence base and support interventions regarding emerging adults’ weight- and shape-related health.

1. Introduction

1.1. Weight- and shape-related behaviors in emerging adulthood

Disordered eating and muscle building behaviors are common among emerging adults (ages 18–29) (Neumark-Sztainer et al., 2018; Neumark-Sztainer et al., 2012a; Christoph et al., 2019; Nagata et al., 2018). For example, approximately half of emerging adult women reported using risky behaviors (e.g. fasting, diet pills) to control their weight. (Neumark-Sztainer et al., 2018) Similarly, the use of some muscle building behaviors, such as consuming protein supplements, is reported by more than one quarter of emerging adults. (Neumark-Sztainer et al., 2020) A substantial body of evidence indicates that such behaviors increase risk for weight gain, clinical eating disorders, poor cardiometabolic health, heart disease, certain types of cancer, and lower quality of life (Nagata et al., 2018; Neumark-Sztainer et al., 2012b; Hart et al., 2020; NIDA, 2018). Supporting positive behaviors in this age group is a priority to prevent morbidity and early mortality.

1.2. Friends and health behaviors

Social ecological models posit that factors at multiple levels (e.g. intrapersonal, interpersonal, institutional, societal) affect health and health behaviors (Bronfenbrenner, 1979; Sallis et al., 2008). Interpersonal contacts, such as family members, friends, and other peers can influence an individual’s own behavior through modeling, pressure, emotional or material support, normative expectations, and other means. Emerging adulthood affords the opportunity to re-center and reorganize relationships; in this life stage, peer relationships and influence take precedence over family relationships and influence, and friends are a particularly salient social context (Arnett, 2014; Nelson, 2021). Research demonstrates that friends’ attitudes, behaviors, and interactions may be associated with weight, disordered eating, and muscle building behaviors during these life stages, either through homophily (i.e. the tendency for those with similar attitudes and behaviors to establish friendships) or through influence (i.e. friends’ influencing each other’s attitudes and behaviors to be more similar over time) (Fletcher et al., 2011; Cunningham et al., 2012). Methods used in this body of research vary widely. A common approach is to survey participants about perceptions of their friends’ attitudes and behaviors, such as concern about healthy eating or dieting frequency. Studies using this method with adolescent and emerging adult samples have shown that perceptions of friends’ attitudes and behaviors are associated with one’s own similar attitudes and behaviors (Eisenberg and Neumark-Sztainer, 2010; Keel et al., 2013; Rosenrauch et al., 2017; Miething et al., 2018; Leahey et al., 2011). For example, longitudinal research by our team has previously found that participants’ perceptions of their friends’ dieting behaviors during adolescence were positively associated with participants’ self-report of chronic dieting, disordered eating behaviors, and binge eating in emerging adulthood. (Eisenberg and Neumark-Sztainer, 2010).

A less common alternative method for examining the role of friends’ attitudes and behaviors is to gather data directly from both participants and their friends about their own attitudes and behaviors. This method is more cumbersome, as it requires participants to identify or “nominate” their friends and researchers to follow-up with participants’ friends to query their health attitudes and behaviors directly (described herein as “actual” attitudes and behaviors, in spite of the known limitations of self-reported data). In addition to providing self-reported attitudes and behaviors from both parties in the relationship, this method also enables characteristics of the friends, friend relationships (e.g. reciprocity of friendship nomination) and the friend network (e.g. density) to be analyzed. We and other researchers have found associations between primary participants’ weight, disordered eating, and muscle building behaviors and these same actual behaviors among their friends (Eisenberg et al., 2012; Goldschmidt et al., 2014; Forney et al., 2019; Becker et al., 2018).

1.3. Perceived vs. actual behaviors

Asking young people to report on other people’s attitudes and behaviors is methodologically straightforward and accomplished more easily than reaching out to friends to report on their own behaviors as part of survey research. However, participant perceptions of their peers’ attitudes and behaviors may not align with peers’ actual attitudes and behaviors. Several studies have found that young people typically do a poor job of estimating risk behaviors among their peers in general (e.g., others at their school or college). Research has demonstrated this phenomenon particularly for substance use behaviors (Perkins et al., 1999, 2019; Prinstein and Wang, 2005; Kilmer et al., 2006), but for weight and dietary behaviors as well (Perkins et al., 2010a, 2010b, 2015, 2018; Lally et al., 2011), with a tendency to overestimate risky behaviors and underestimate health promoting behaviors among the peer group at large. We would expect greater accuracy in reporting on the attitudes and behaviors of specific friends than reporting on peers in general, based on conversations and direct observations. However, these perceptions may still be subject to considerable bias towards young people’s own behaviors due to a desire to share values and behaviors with friends. To our knowledge, this issue has not been explored in the research literature on disordered eating and muscle building behaviors.

As studies linking data from nominated friends to a primary participant are methodologically challenging and costly, large investigations of this nature are relatively rare. Approaches include using a “closed network,” in which participants are asked to nominate their friends from an existing list (e.g. all students in their school), or an “open network,” in which participants can nominate any friend. The advantage of such complex studies is the opportunity for friends to report on their own behavior. This method is expected to do a better job of assessing friends’ true behaviors, especially in survey research where anonymity or confidentiality are protected and concerns about social desirability bias are therefore lessened (especially in self-reporting behaviors that might be unpopular or illegal, such as steroid use). However, the value of this trade-off is unclear. This study asks whether it is feasible to collect high quality data from emerging adults’ friends on a large scale, and whether the resulting data on actual behaviors have greater utility than measures of perceived friend behaviors in understanding the influence of friends on weight- and shape-related behaviors.

1.4. The present study

As described above, theoretical frameworks and empirical research suggest the importance of understanding the role of friends in disordered eating and muscle-building behaviors of young people. Emerging adulthood is a unique period with regards to social development, (Arnett, 2014), yet this stage has been the focus of relatively few studies on this topic. The use of friends’ self-reported data is rare in studies of emerging adults’ disordered eating and muscle building behaviors – perhaps due to challenges and logistical considerations of collecting such data – and we are not aware of any existing studies that have examined the alignment of perceived behaviors and friends’ actual behaviors in this population. Investigating these strategies empirically is a critical step in advancing the epidemiology in this area. Building on our earlier work examining adolescents’ and their friends using a closed network design, (Eisenberg et al., 2012; Bruening et al., 2012; Bruening et al., 2014; Sirard et al., 2013) the present study utilizes an opportunity to assess the feasibility of collecting data from emerging adults’ friends using an open network approach, in light of the importance of friends from different domains at this stage of life (e.g. college/university, workplace, childhood). We further compare perceived and actual friend behaviors and their associations with disordered eating and muscle building behaviors. This study therefore addresses the following research questions: 1) What is the feasibility of collecting data from emerging adults’ friends in epidemiologic research? 2) Do perceptions of friends’ disordered eating and muscle building behaviors align with friends’ actual behaviors? and 3) Are perceptions or friends’ actual behaviors more strongly and consistently associated with emerging adults’ behaviors?

Findings are expected to contribute to studies of measurement val-idity for reporting on others’ behaviors, as well as advancement of etiologic models of disordered eating and muscle-building behaviors. Disentangling the relevance of perceptions and actual friend behaviors will also identify appropriate targets for future intervention research building on this body of work. For example, research on associations between individuals’ perceptions of friends’ health behaviors and their own behaviors, as well as the observed clustering of health behaviors among friend groups, provide a basis for peer-focused intervention activities aimed at behavior change within a social group. Such interventions have shown some promise in the areas of substance use and sexual health (Hunter et al., 2019), but have rarely been applied to disordered eating and muscle building behaviors (Buller et al., 1999; Wing and Jeffery, 1999; Gotsis et al., 2013). Further understanding of the role of friends’ actual behavior can illuminate prevention opportunities for these additional behaviors.

2. Methods

2.1. Study design and sample

EAT 2010–2018 (Eating and Activity over Time) is a population-based longitudinal study of dietary intake, physical activity, weight control behaviors, weight status, and factors associated with these outcomes in young people. (Neumark-Sztainer et al., 2020 and Larson et al., 2020) Participants were originally recruited in 2009–2010 as adolescents attending one of 20 public middle and high schools in Minneapolis and St. Paul. (Neumark-Sztainer et al., 2012a) Participants were invited to complete follow-up surveys as emerging adults in 2017–2018, and were further asked to invite their friends to participate as part of a substudy of friends’ influences on weight- and shape-related behaviors. The current cross-sectional analysis focuses on the data reported by participants and their friends as part of EAT 2018. (Due to the overlapping nature of ages and roles in this study, we will use the terms “emerging adult” and “EAT 2018 participant” to refer to the primary participant involved in the ongoing EAT 2018 study, and the term “friend” to refer to those who were nominated to take part in the friends’ sub-study.)

Of the original 2793 adolescent participants in 2009–2010, 410 (14.7%) were lost to follow-up for various reasons, primarily due to providing insufficient contact information at baseline. The remaining 2383 young people were re-contacted, and the EAT 2018 sample includes 908 emerging adult women and 649 emerging adult men, plus 11 who indicated another gender identity (mean age = 22.0 ± 2.0 years; n = 1568, 65.8% response). The EAT 2018 sample was diverse with regards to race/ethnicity: 23.4% were White, 22.1% were Black or African-American, 17.5% were Hispanic or Latino/a, 22.7% were Asian, 14.3% identified as another racial or ethnic group or marked more than one category. There were some minor differences in background variables for those who were lost to follow-up from 2010 (e.g. race, gender) and did not participate in the EAT 2018 survey; however, the analysis described here does not make use of the full EAT 2018 survey sample. The current analysis includes only EAT 2018 survey respondents who had a nominated friend participate (as described below), and results based on analyses with this convenience sample are not expected to generalize to the original school-based population.

2.2. Data collection and friend nomination

Invitations to participate in the online follow-up survey were mailed with a small financial incentive; multiple attempts were made to contact non-responders through mailed reminders, email, phone calls, text messaging, social media, and home visits. All EAT 2018 participants were mailed an additional financial incentive following survey completion.

As part of the emerging adult survey, participants were asked to nominate up to three close friends (ranked by closeness), respond to questions about each friend, and provide contact information for each friend, or check a box “I do not have a close friend.” There was no specification on the survey to indicate that friends should be chosen by the gender of the friend. Friends were then invited via text or email to complete an abbreviated version of the EAT 2018 survey (“Friends Survey”) for a small financial incentive, with up to four reminders sent as needed. When friends were nominated but no contact information was provided, up to three invitation cards were mailed to the EAT 2018 participant with a request to distribute cards by hand to their nominated friends (N = 1466 friends). In select cases (N = 9 friends), individuals who completed the Friends Survey provided contact information (name, initials, address) that did not match the information provided during the friend nomination process. These friend responses were discarded, resulting in a sub-sample of 1126 EAT 2018 participants who nominated 1919 friends. Among the invited friends, 191 responded and were linked to the 152 EAT 2018 participants who nominated them.

Data collection ran from June 2017 to November 2018 and was conducted by the Office of Measurement Services (https://oms.umn.edu/) at the University of Minnesota. The University of Minnesota Institutional Review Board Human Subjects Committee approved all protocols used at each time point.

2.3. Survey development and measures

The EAT 2018 Survey and Friends Survey were based on previous versions of the Project EAT Survey, () with revisions to increase the relevance for this age group. Three focus groups were conducted to pretest an initial draft of the 2018 survey with a separate sample of emerging adults. Discussion questions probed both the survey items and the friend nomination process (e.g. What could we do to make it more likely that you would provide contact information for your friends? What challenges might people of your age have in providing the emails or phone numbers for three friends? If you got an invitation to do this study because a friend nominated you, what would make you want to follow the link and find out more?). Feedback from the 29 focus group participants was used to reword or eliminate problematic survey measures, expand on topic areas of interest, and revise the friend nomination protocol. After the survey was finalized, the test-retest reliability of measures was examined using data from a subgroup of 112 emerging adult participants who completed the 2018 survey twice within a period of three weeks. Psychometric properties given below were examined in the full sample of participants who responded to the EAT 2018 survey.

Emerging adults’ own disordered eating and muscle building behaviors, perceived friend behaviors, and actual friend behaviors were assessed on the EAT 2018 Survey and Friend Survey, as detailed in Table 1 EAT 2018 participants reported their gender and date of birth, which was used with date of survey return to calculate age.

Table 1.

Measurement of weight- and shape-related behaviors.

Construct	EAT 2018 participant behaviors	Perceived friend behaviors	Friends’ actual behaviors
Frequent dieting	How often have you gone on a diet during the last year? By “diet” we mean changing the way you eat so you can lose weight.	For each nominated friend: Does this friend do any of the following to lose weight or keep from gaining weight … Go on a diet? b. Take diet pills, vomit, use laxatives, or use diuretics	How often have you gone on a diet during the last year? By “diet” we mean changing the way you eat so you can lose weight.
	5 response options: “never” to “I am always dieting”; 5+ times contrasted with fewer (test-retest r = 0.66)	Response options: yes, no, I don’t know	5 response options: “never” to “I am always dieting”; 5+ times contrasted with fewer (test-retest r = 0.66)
Extreme weight-control behaviors	4 items: Have you done any of the following things in order to lose weight or keep from gaining weight during the past year? Used diet pills, made myself vomit (throw up), used laxatives, used diuretics (water pills).	For each nominated friend: Does this friend do any of the following to lose weight or keep from gaining weight … Take diet pills, vomit, use laxatives, or use diuretics?	4 items: Have you done any of the following things in order to lose weight or keep from gaining weight during the past year? Used diet pills, made myself vomit (throw up), used laxatives, used diuretics (water pills).
	Response: yes/no for each behavior; use of any contrasted with none (test-retest agreement = 93%)	Response options: yes, no, I don’t know	Response: yes/no for each behavior; use of any contrasted with none (test-retest agreement = 93%)
Use protein powder, shakes, or a pre-workout drink	2 items: Have you done any of the following things in order to increase your muscle size or tone during the past year? Used protein powder or shakes, used a pre-workout drink acids, hydroxyl methylbutyrate [HMB], DHEA, or growth hormone)	For each nominated friend: How often does this friend do the following things to increase muscle size or tone … use protein powder, shakes, or a preworkout drink (such as Jack3D, Cellucor C4 or JYM)?	2 items: Have you done any of the following things in order to increase your muscle size or tone during the past year? Used protein powder or shakes, used a pre-workout drink acids, hydroxyl methylbutyrate [HMB], DHEA, or growth hormone)
	Response: yes/no for each behavior; use of either contrasted with none (test-retest agreement = 88%)	Response options: Never, Rarely, Sometimes, Often, I don’t know	Response: yes/no for each behavior; use of either contrasted with none (test-retest agreement = 88%)
Use steroids or another muscle building substance	2 items: Have you done any of the following things in order to increase your muscle size or tone during the past year? Used steroids, used another muscle-building substance (such as creatine, amino acids, hydroxyl methylbutyrate [HMB], DHEA, or growth hormone)	For each nominated friend: How often does this friend do the following things to increase muscle size or tone … use steroids or another muscle-building substance (such as creatine or growth hormone)	2 items: Have you done any of the following things in order to increase your muscle size or tone during the past year? Used steroids, used another muscle-building substance (such as creatine, amino acids, hydroxyl methylbutyrate [HMB], DHEA, or growth hormone)
	Response: yes/no for each behavior; use of either contrasted with none (test-retest agreement = %)	Response options: Never, Rarely, Sometimes, Often, I don’t know	Response: yes/no for each behavior; use of either contrasted with none (test-retest agreement = %)

EAT 2018 participants also reported their friends’ gender (male, female, different gender identity), and the frequency with which they were in touch with that friend via five different methods (in person, phone, text/email, video calling, social media), with response options ranging from never/rarely (1) to daily (6). For each friend, frequency values were summed for use in analysis (mean = 20.6, SD = 5.2, range = 6–30).

2.4. Data analysis

To address the first research question (what is the feasibility of collecting data from emerging adults’ friends?), we provide descriptive statistics of the subset of participants who nominated friends and had at least one friend respond.

To address the second research question (do perceptions of friends’ disordered eating and muscle building behaviors align with friends’ actual behaviors?), we first provide descriptive statistics regarding whether emerging adults report they don’t know about their friends’ behaviors. After deleting “don’t know” responses, we used Spearman correlations to test the association between emerging adults’ perceptions of their friends’ behaviors and their friends’ actual behavior, using friends as the unit of analysis (in order to include cases in which a participant had more than one friend respond).

To address the third research question (are perceptions or friends’ actual behaviors more strongly and consistently associated with emerging adults’ behaviors?), logistic regression models were used to generate odds ratios of emerging adults engaging in each disordered eating and muscle building behavior based on 1) the emerging adult’s perception of their friends’ same behaviors (deleting “don’t know” responses), and 2) friends’ actual behaviors (modeled separately). Perceived and actual behavior variables were standardized to increase comparability of estimates, given measurement differences shown in Table 1. Emerging adult participants were the unit of analysis, and models controlled for frequency of communication and accounted for clustering (i.e. more than one friend per participant) using SAS proc genmod, in order to obtain robust standard errors and accurate inference. Additionally, average marginal effects were derived using the margins macro (shown in the supplemental table). Given established differences in disordered eating and muscle building behaviors in emerging adulthood (Neumark-Sztainer et al., 2018; Christoph et al., 2019; Nagata et al., 2020), all analyses were stratified by gender of the EAT 2018 participant. Only one EAT 2018 participant identifying with another gender had data from a nominated friend and could not be included in analysis.

3. Results

3.1. Feasibility of collecting friend data

As shown in Table 2, most (71.8%) participants nominated at least one close friend to be contacted for the sub-study; 410 participants reported that they did not have a close friend. The gender distribution and mean age of the subsample who nominated friends is very similar to the full emerging adult sample. There were slight differences in race/ethnicity between the full sample and those nominating friends, with white participants making up 23.4% of the full sample but 28.3% of the sample nominating friends. Among those who nominated any friends, almost half (48.0%) nominated only one. An additional 33.6% nominated two friends, and 18.7% nominated three friends.

Table 2.

Characteristics of the EAT 2018 participants indicating feasibility of obtaining friend nominations and friend responses.

	EAT 2018 participants		EAT 2018 participants who nominated ≥ 1 friend		EAT 2018 participants with ≥1 friend responding
	N	%	N	%	N	%
Total	1568	100	1126	71.8a	152	13.5b
Gender
Men	649	41.4	461	41.1	34	22.4
Women	908	57.9	651	58.0	117	77.0
Different gender	11	0.7	11	1.0	1	0.7
Race/ethnicity
White	366	23.4	318	28.3	41	27.0
Black/Afr. Am.	345	22.1	228	20.3	23	15.1
Hispanic	274	17.5	193	17.2	27	17.8
Asian	355	22.7	240	21.4	47	30.9
Mixed/other races	223	14.3	143	12.8	14	9.2
Number of friends nominated
0	442	28.2	–	–	–	–
1	541	34.5	541	48.0	117	77.0
2	377	24.0	377	33.5	31	20.4
3	208	13.3	208	18.5	4	2.6
Frequency of being in touch with friends (mean; range =5–30)	-		20.0 (SD = 5.2)		20.6 (SD = 5.2)
Age (mean)	22.5 (SE = 2.0)		22.5 (SD = 2.0)		21.6 (SD = 1.87)

^aPercent of EAT 2018 YA sample.

^bPercent of EAT 2018 sample with ≥1 nominated friend.

Despite applying best survey practice contacting methods, only 152 EAT 2018 participants had at least one friend complete the Friend Survey (13.5% of those who nominated any friends, and 9.7% of the full EAT 2018 sample). Over three-quarters of these 152 participants had data from only one friend. The subsample who successfully had at least one nominated friend complete the Friend Survey included a higher percentage of emerging adults who identified as women (77.0%), more Asian (30.9%) and less Black/African American (15.1%) representation, and was slightly younger (mean age = 21.6) than the subsample who did not have their nominated friend(s) compete the survey. A total of 191 friends responded to the Friends Survey, which was 9.9% of the 1924 who were invited. The gender of the responding friends (as reported by the EAT 2018 participants who nominated them) was 79.0% women, 20.5% men, and 0.5% another gender identity.

The prevalences of each disordered eating and muscle building behavior are shown in Table 3 for the sample of 152 EAT 2018 participants with at least one friend responding and for the EAT 2018 participants who did not have any friends provide data. Using protein powders, shakes, or a pre-workout drink was the most common of these behaviors, reported by over one-third of emerging adult men. Dieting, weight control behaviors, and other muscle building behaviors were less common. The prevalence of each behavior among those with at least one friend was similar to or slightly lower than the prevalence in the remainder of the sample, but these differences were only statistically significant for women’s use of steroids or other muscle building substances.

Table 3.

Prevalence of each disordered eating and muscle building behavior (N = 1568 EAT 2018 participants).

	EAT 2018 participants with ≥1 friend responding (N = 152)				EAT 2018 participants with no friends responding (N = 1416)				p-valuea
	Men		Women		Men		Women		Men	Women
	N	%	N	%	N	%	N	%
Frequent dieting (5+ times)	4	11.8	14	12.1	58	9.6	107	13.8	.684	.610
Extreme weight-control behaviors	4	11.8	14	12.2	38	6.4	147	19.6	.223	.057
Protein powder, shakes, or a pre-workout drink	12	35.3	17	14.8	218	36.5	153	20.0	.886	.190
Steroids or another muscle building substance	5	15.2	12	10.5	82	13.8	138	18.1	.825	.046

^aFor comparison of those with and without ≥1 friend responding.

3.2. Alignment of perceived and actual friend behaviors

The prevalence of “I don’t know” responses to the four perception items is shown in Table 4 stratified by gender of the EAT 2018 participant. The proportion of perception item responses that were “I don’t know” ranged from 8.3% to 28.6%, across the behaviors. Correlations between emerging adults’ perceptions of their friends’ behaviors and their friends’ actual behaviors are shown in Table 5, stratified by gender of the EAT 2018 participant. For all behaviors except use of steroids and other muscle building substances, these correlations were statistically significant among emerging adult women. For example, the correlation between emerging adult women’s perception that their friends diet to lose weight or keep from gaining weight and the matching friends’ report of actually dieting was 0.31 (p < .001). Most correlations were low (r < 0.30), indicating a weak relationship. Associations between perceived and actual behaviors were not statistically significant among emerging adult men, with the exception of dieting behavior (r = 0.40, p < .05).

Table 4.

Prevalence of “I don’t know” response for perceptions of friend disordered eating and muscle building behaviors (N = 152 EAT 2018 participants reporting perceptions of 191 nominated friends).

How often does this friend …	% who don’t know
	Men	Women
Diet to lose weight/keep from gaining weight	14.3	18.1
Extreme weight-control behaviors	12.2	12.4
Use protein powder, shakes, or a pre-workout drink	28.6	10.4
Use steroids or another muscle building substance	19.1	8.3

Table 5.

Spearman correlations between emerging adult perceptions of their friends’ behaviors and friends’ actual disordered eating and muscle building behaviors (N = 191 friends).

	Men	Women
Diet to lose weight/keep from gaining weight	.40*	.31***
Extreme weight-control behaviors	–	.28**
Use protein powder, shakes, or a pre-workout drink	−.02	.28**
Use steroids or another muscle building substance	.10	−.03

^*p < .05

^**p < .01

^***p < .001.

– not calculable.

3.3. Associations between friend behaviors and EAT 2018 participant behaviors

Associations between the disordered eating and muscle building behaviors of friends and EAT 2018 participants are shown for both perceived and actual friend report of behaviors in Table 6. For emerging adult women, their reported perception of their friends’ behavior was associated with greater odds of their own use of the same behavior for frequent dieting and use of protein powders, shakes, or pre-workout drinks. For example, a one standard deviation increase on the scale of perceived friend dieting frequency was illustratively associated with twice the odds of dieting oneself (OR = 2.03, CI: 1.14, 3.61) among emerging adult women. For emerging adult men, an association between perceived friend and participant behavior was evident only for using protein powders, shakes, or a pre-workout drink (OR = 6.41, CI: 1.50, 27.32).

Table 6.

Odds ratios (and 95% confidence intervals) of each disordered eating and muscle building behavior in emerging adults, given perceived friend behavior or friends’ actual behavior – standardized variables (N = 152 EAT 2018 participants)^a.

	Men		Women
	Perceived	Actual	Perceived	Actual
Frequent dieting (5+ times)	2.24 (0.65, 7.82)	5.78 (1.68, 19.83)	2.03 (1.14, 3.61)	1.57 (1.01, 2.45)
Extreme weight-control behaviors	–	1.95 (0.81, 4.70)	1.53 (0.95, 2.44)	1.50 (0.99, 2.27)
Use protein powder, shakes, or a pre-workout drink	6.41 (1.50, 27.32)	1.03 (0.56, 1.89)	1.97 (1.26, 3.09)	1.09 (0.65, 1.84)
Use steroids or another muscle building substance	1.33 (0.71, 2.50)	1.73 (0.82, 3.61)	–	1.33 (0.84, 2.10)

Boldface font indicates statistical significance (p < .05).

– not calculable.

^aAdjusted for frequency of communication with friend.

For women, friends’ actual behavior was associated with elevated odds of frequent dieting (OR = 1.57, CI: 1.01, 2.45). Emerging adult men also had significantly greater odds of frequent dieting when their friends reported this same behavior (OR = 5.78, CI: 1.68, 19.83). Other behaviors examined here were not associated with friends’ actual behaviors.

4. Discussion

This study is one of the first to examine the feasibility of collecting data from a community sample of nominated friends of emerging adults and to compare perceptions of friends’ disordered eating and muscle building behaviors to friends’ actual self-report of the same behaviors. In spite of extensive protocol development efforts (including input from the target population), multiple friend recruitment methods, and numerous attempts to contact EAT 2018 participants’ nominated friends, the response rate among friends was low, suggesting that nominating and collecting data from friends may not be feasible for large epidemiologic studies of emerging adults. Among EAT 2018 participants, the majority believed they could accurately report the behaviors of their nominated friends. However, emerging adults’ perceptions of their nominated friends’ weight and shape-related behaviors generally did not align with the behaviors reported by those friends. Furthermore, analytic models found different associations between friends’ behavior and EAT 2018 participant behaviors, depending on whether the perceived or actual measure of friends’ behavior was used.

Findings illustrate the importance of distinguishing between the assessment of emerging adults’ perception of their friends’ behaviors and the assessment of friends’ actual behaviors in epidemiologic research. These differences may stem from the fact that some behaviors are more observable or visible to peers than others. For example, self-induced vomiting or diet pill use may be intentionally concealed or simply not observed, as they typically take place in more private locations (in contrast to dieting, which may be more commonly observed or referenced in social settings). Likewise, stigmatized or illegal behaviors (e.g. steroid use) may not be discussed openly with friends. Differences in the public versus private nature of the behaviors examined here might affect the extent to which emerging adults can accurately report on their friends.

Regardless of the underlying reason, perceived and actual friend data cannot be interchanged, and it is therefore imperative to consider the pros and cons to each method and collect the data that are most relevant to the research question. Data on emerging adults’ perceptions of their friends’ behaviors offer the advantages of being relatively quick and less expensive to collect and having a relatively high response rate (compared to secondary recruitment of friends); this method is commonly used in studies of friend influence. Perceptions of friends’ behaviors are often associated with individuals’ behaviors, as observed here and elsewhere (Eisenberg and Neumark-Sztainer, 2010; Keel et al., 2013; Rosenrauch et al., 2017; Miething et al., 2018; Leahey et al., 2011; Pelletier et al., 2015; Harmon et al., 2016; Rice and Klein, 2019), particularly among women (Eisenberg and Neumark-Sztainer, 2010; Keel et al., 2013; Miething et al., 2018). It is important to note that such associations may be an artifact of people’s desire to share values and behaviors with their friends, which may bias perceptions towards the individual’s own behavior. Perceptions may also be a critical mechanism of friends’ influence (conceptually, an individual’s awareness of a friend’s behavior is usually required in order to be affected by that behavior). However, according to our findings, emerging adults’ perceptions of their friends’ behaviors are generally not an accurate proxy for their friends’ report of their disordered eating and muscle building behaviors. Researchers should be clear in reporting when they utilize measures of perception (including in titles and abstracts), and acknowledge the strengths and limitations of this approach. Cognitive testing with emerging adults regarding what they notice about their friends’ disordered eating and muscle building behaviors and how and why they would respond to survey measures of perception may be a useful step. Such work could inform if and how perception measures of friends’ behaviors can be re-worded, statistically adjusted or re-interpreted to increase consistency with friends’ actual behaviors for studies in which actual behaviors are the intended focus but resources are more limited.

On the other hand, identifying, recruiting, and surveying nominated friends can provide rich and strong data that can generate critical insights into understanding the role of friends’ behavioral influences in emerging adulthood. The advantage of obtaining data on friends’ actual behaviors may offset the greater expense, time commitment, and lower response rate for certain studies. If friends’ actual behavior is the key variable of interest, certain methods building on the approach used in the present study may improve response rates. First, a small financial incentive was provided to responding friends in the present study. An additional incentive for the primary survey respondent for each friend nominated or successfully recruited may improve nomination rates. The 28% indicating they had no close friends is higher than expected, suggesting that EAT 2018 participants may have been reluctant to provide survey data regarding perceptions of their friends (adding to the length of their own survey) or provide contact information for their friends (potentially due to privacy concerns). Second, there are multiple ways in which the process of recruiting nominated friends may be strengthened, including using texts to immediately contact friends (thereby reducing the chance that a nominated friend’s contact information would change in the intervening period) and designing the electronic survey invitation to be provided from the emerging adult as a known contact (to avoid spam folders or deletion of a text or email from an unfamiliar source). Third, recruitment of friends may also be enhanced by limiting nominations to a closed network, such as an existing roster of students at the same school or college, as we and others have done previously (Eisenberg et al., 2012; Goldschmidt et al., 2014; Forney et al., 2019; Becker et al., 2018; Bruening et al., 2012; Bruening et al., 2014; Sirard et al., 2013; de la Haye et al., 2010; de la Haye et al., 2011; de la Haye et al., 2013; Ali et al., 2011; Wouters et al., 2010). Although this approach may miss important friends from outside the proscribed network, it has the advantage of providing reliable contact information or pathways (e.g. through the school or college). These enhanced methods were not used in the present study due to financial and time constraints.

The use of perceived or actual friend behaviors in research also has implications for interventions. Intervention activities focused on friends can only be successful if they address the appropriate target based on the guiding evidence – either changing friends’ actual behaviors or changing people’s perceptions of their friends’ behaviors. If friends’ actual behaviors are more healthful than they are perceived to be, activities to bring perceptions into alignment may be an important step. In contrast, actually changing friends’ behaviors may form the basis of peer education, peer modeling, and social support interventions. Existing research on friend (or other peer) influence interventions has largely focused on addressing substance use and sexual behaviors (Hunter et al., 2019); similar approaches could be used for disordered eating and muscle building behaviors, making it particularly important to design studies to support this advancing field.

4.1. Strengths and limitations

This unique study builds on our prior research on friends’ influence using a closed network design, and is one of the first to capture both perceptions of and friends’ actual weight- and shape-related behaviors, for numerous specific behaviors, in an open friend network. The initial sample size was large; this was a strength, as the relatively small resulting convenience sample of emerging adults with perceived and actual friend behavior data was large enough to permit statistical analyses. The original EAT sample is also population based and originated in urban school districts serving diverse communities. The sample is diverse with regards to post-high school settings, rather than focusing solely on college and university students (a common population for studies of this age group).

However, the generalizability of the findings regarding associations between friends’ behavior and emerging adults’ behavior is limited by the low response rate among friends and small analytic sample. Specifically, several regression models were underpowered, as signaled by relatively large (>2.0) but non-significant odds ratios and models that would not converge. Additionally, the analytic sample was ultimately a convenience sample which may not be generalizable to all emerging adults and their friends. Finally, the primary aim of the parent study did not include tests of friend recruitment or data collection protocols, so the effectiveness of specific methodologic enhancements (e.g. recruitment incentives for EAT 2018 participants) is unknown.

5. Conclusion

The study of friends’ influences on the disordered eating and muscle building behaviors of emerging adults can take many forms; the present study identified challenges inherent in relying on perceptions of friend behaviors as well as challenges of collecting data from nominated friends. Data collected through these two methods are not interchangeable, and researchers are strongly advised to plan, conduct, and report studies with careful consideration of the type of data they need to address their research questions. Additional research is needed to extend our understanding of associations between perceived friend behaviors, actual friend behaviors, and emerging adults’ own disordered eating and muscle building behaviors, in order to inform targets for future friend-focused interventions.