Relationship Status Predicts Lower Restrictive Eating Pathology for Bisexual and Gay Men across 10-year Follow-up

Abstract

Objective

Cross-sectional studies support that bisexual and gay (BG) men are at increased risk for eating pathology, and romantic relationships may buffer against risk; however, no studies have examined this association longitudinally. The current study examined how romantic relationships impact the trajectory of eating pathology in BG versus heterosexual men.

Method

BG (n=51) and heterosexual (n=522) men completed surveys of health and eating behaviors at baseline and 10-year follow-up.

Results

For BG men, being single at baseline prospectively predicted an increase in Drive for Thinness scores over 10-year follow-up. Additionally, for BG men in relationships at baseline, lower relationship satisfaction predicted an increase in Drive for Thinness scores over time. Conversely, these relationship variables did not predict trajectory of eating pathology for heterosexual men.

Discussion

Implications for theoretical models of risk, including objectification theory and sexual minority stress theory, and prevention, including peer-led cognitive dissonance based interventions, are discussed.

Community-based studies have consistently demonstrated that bisexual and gay (BG) men report increased eating pathology compared to heterosexual men.^1-5 One promising sociocultural explanation for BG males' high-risk status stems from Siever,¹ who suggested that BG males experience heightened emphasis on a lean, muscular physique within the gay male community and face increased pressure to obtain an ideal physique in order to attract a male partner.^{1, 6-7} This leads BG men to internalize the lean and muscular body ideal, thus increasing risk for disordered eating, particularly related to restriction,⁸ compared to heterosexual men.¹ Potentially relevant to this conceptualization is objectification theory,⁹ which was originally developed, in part, to explain eating disorder risk among women. Objectification theory states that females are socialized to view their physical appearance from an observer's perspective, as a sexual object. This internalized objectification is proposed to lead to greater body monitoring, which contributes to internalization of the thin ideal.¹⁰ Internalized objectification then leads to increased body shame, which could subsequently lead to restrictive eating attitudes and behaviors in an attempt to conform to the thin ideal. Attracting male partners appears to play an important role in this objectification, as men place greater emphasis on physical appearance when looking for romantic partners than women,^{3, 6-7, 11} resulting in males engaging in more frequent visual inspection or “gazing” of potential partners.

According to objectification theory, habitually monitoring the body to attract the male gaze may be more relevant for those seeking male partners. Given that single BG men and BG men who are dissatisfied with their relationships may be more likely to habitually monitor their body to attract the male gaze, they may be at heightened risk to develop restrictive eating pathology compared to heterosexual men. Consistent with male partners contributing to increased eating pathology, Tylka and Andorka¹² found in a cross-sectional study that for gay men, partner pressures to be lean and muscular were associated with greater internalization of the mesomorphic body ideal. This internalization was then associated with greater body dissatisfaction and disordered eating behaviors. Importantly, while bisexual males may be interested in attracting both male and female partners, applications of objectification theory would imply that any interest in attracting a male romantic partner could increase body monitoring to attract the male gaze. Thus, both bisexual and gay men remain subject to pressures to conform to the lean and muscular physical ideal that men find attractive.

In line with these hypotheses, Brown and Keel⁸ found that single BG men had increased drive for thinness and dieting frequency compared to BG men in relationships and all heterosexual men. This association was specific to restrictive eating pathology and did not extend to bulimic symptoms, consistent with a more direct association between pursuit of a lean and muscular ideal and restrictive eating pathology. Regarding relationship satisfaction, the same study found that increased relationship satisfaction was associated with reduced bulimic symptoms for BG, but not heterosexual, men. These results were specific to bulimic symptoms and did not extend to restrictive eating pathology. The authors suggested that BG men in less satisfying relationships may be less equipped to cope with stress effectively, which may increase vulnerability to engage in maladaptive coping strategies, like binge-eating. Although results from Brown and Keel⁸ provided an important first step in establishing how romantic relationships may moderate associations between sexual orientation and eating disorder risk in men, the cross-sectional study design constrains conclusions that may be drawn regarding temporal associations between relationship variables and eating disorder risk. Given that eating pathology impacts psychosocial functioning,¹³ the presence of more restrictive pathology may impact likelihood of being in a romantic relationship, and greater bulimic pathology may negatively impact satisfaction within romantic relationships. Thus, longitudinal designs are needed to interpret the direction of these associations. An important aspect of objectification theory is that self-objectification leads to changes in internalization of the body ideal.¹⁰ If this occurs during sensitive periods, such as the transition from late adolescence to young adulthood or adulthood to midlife,¹⁴ this may impact ongoing risk for eating pathology, independently of future relationship status or satisfaction.

The purpose of the present study was to examine how relationship status and relationship satisfaction affect the longitudinal trajectory of eating pathology in BG men. We hypothesized that BG men who were single at baseline should demonstrate greater increases in restrictive eating pathology scores over 10-year follow-up compared to BG men in relationships and all heterosexual men. Further, we hypothesized that among men in relationships, low baseline relationship satisfaction should predict increased restrictive eating pathology in BG, but not heterosexual, men and this moderation effect should not extend to bulimic symptoms. Although the latter hypothesis differs from what would be predicted based on cross-sectional results from Brown and Keel,⁸ these predictions are consistent with how relationships would affect internalization of the body ideal over time once the impact of bulimic symptoms on relationship variables is controlled.

Methods

Participants

Participants for the present study were drawn from three cohorts, originally recruited from a prestigious northeastern university in 1982, 1992, and 2002 and followed at 10-year intervals. Sexual orientation identity was first queried in surveys distributed in 2002 (the 3^rd wave of the longitudinal study), representing baseline assessment for the current study, and follow-up data were collected in 2012 (4^th wave) in 573 men. The sexual identity item was phrased, “What is your sexual orientation?” with response options including “heterosexual”, “homosexual”, or “bisexual.” At baseline, the men in the 2002 cohort were M (SD) = 20.02 (1.59) years old (range = 17.00 - 23.00), the men in the 1992 cohort were M (SD) = 30.02 (1.66) years old (range = 27.00 - 34.00), and the men in the 1982 cohort were M (SD) = 40.20 (1.78) years old (range = 37.00 - 45.00). Thus, at follow-up, the men in each cohort were approximately 10 years older (2002 cohort M [SD] = 30.56 [1.54]; 1992 cohort M [SD] = 40.39 [1.56]; 1982 cohort M [SD] = 50.53 [1.75]). Given that men may “come out” at various times over their development, men who reported that they were gay or bisexual at either baseline or follow-up were included in the BG group (n=51; 16 self-identified as bisexual and 35 as gay). The remaining 522 men self-identified as heterosexual at both assessment waves. Bisexual and gay men were analyzed as a combined group, consistent with research on sexual plasticity in men^15-16 and previous research on eating pathology,^{8, 17-19} and to maximize power for analyses of sexual minority groups.

Notably, the proportion of BG men in the current sample (51/573 or 9%) was higher than estimates of sexual orientation identity from population-based samples.²⁰ The proportion of heterosexual and BG men did not differ among cohorts (χ²[2] = 1.74, p = .42). Racial or ethnic background did not differ across BG and heterosexual men (χ²[6] = 7.37, p = .29), but did differ by cohort (χ²[12] = 22.72, p = .03), with younger cohorts having greater ethnic diversity, reflecting a trend towards universities including greater ethnic diversity over time. The ethnic/racial breakdown of the sample included within the present study was: 76% Caucasian, 13% Asian, 5% Hispanic, 4% African American, and 2% other/undisclosed.

Overall, 66% of men contacted for the 2002 survey and 63% of men contacted for the 2012 survey participated at follow-up. Participation in the 2012 survey did not differ between BG and heterosexual men (χ²[1] = .50, p = .48). No significant differences were observed for baseline variables between participants and non-participants in follow-up assessments (all p-values >.05), with one exception: the youngest cohort of men (2002 cohort) was less likely to participate (p<.001).

Procedure

Participants were mailed consent forms and a self-report survey on eating and health behaviors with addressed and stamped envelopes at baseline and 10-year follow-up. Surveys were mailed up to 3 times to maximize participation rates, and participants were given the opportunity to complete surveys on-line at both waves. Participants' who completed the follow-up survey were entered into a raffle for the opportunity to win one of 50 $100 awards. No compensation was provided for completion of the baseline surveys. No significant differences were found between participants opting to complete paper versus on-line surveys. All study procedures were approved by the relevant institutional review boards at the universities where the research was conducted, and all participants provided informed consent prior to the completion of surveys.

Measures

Surveys and measures match those previously reported for cross-sectional associations between relationship variables and eating pathology in BG and heterosexual men.^{8, 19}

Body mass index (BMI)

BMI was determined from self-reported height and weight and calculated as weight in kilograms divided by height in meters squared (kg/m²). Self-reported height and weight have been identified as valid estimates of objective height and weight in epidemiological studies of men in the United States.^21-22

Relationship Status

Relationship status was measured through a single, dichotomous-response item assessing, “Are you currently involved in a steady relationship?” (“No” or “Yes”). Prior studies examining how relationship status affects disordered eating have used comparable single-item measures of relationship status.^{8, 19, 23-24}

Relationship Satisfaction

For individuals who endorsed being involved in a steady relationship, relationship satisfaction was assessed through a single item asking, “How satisfied are you with your current relationship?” on a 7-point scale, from 1 (Not at all) to 7 (Completely satisfied).

Disordered Eating Variables

Items assessing Drive for Thinness and Bulimia subscales from the Eating Disorder Inventory (EDI)²⁵ were used to evaluate drive for thinness and bulimic symptoms, respectively. The internal consistency of the Drive for Thinness subscale was α =.85 for heterosexual men and α = .93 for BG men at baseline, and α =.98 for heterosexual men and α = .99 for BG men at follow-up. Internal consistency for the Bulimia subscale at baseline was α =.69 for heterosexual men and α = .86 for BG men, and was α =.99 at follow-up for both heterosexual and BG men. The somewhat lower internal consistency for heterosexual men at baseline may reflect a restricted range in item responses given that they represent a generally lower risk group in college.

Data Analyses

Outcome variables were positively skewed at baseline and follow-up and were log transformed prior to running analyses. Missing data were imputed using expectation maximization in SPSS 19, to avoid biased estimates of associations, as recommended by Schafer and Graham.²⁶ Separate hierarchical moderated linear regression models were run to determine whether sexual orientation identity at baseline moderated the association between relationship variables at baseline and eating pathology outcomes at follow-up (EDI Drive for Thinness and EDI Bulimia), controlling for baseline values of the outcome variables. All analyses also controlled for age and BMI. Baseline levels of the outcome variable, along with relevant covariates, were included in Step 1, the main effects of sexual orientation identity and relationship status (or relationship satisfaction) were included in Step 2, and the interaction between these two variables was included in Step 3 (see Table 2). Because we controlled for baseline levels of the outcome variables, the models within the present study estimated change in the outcome variables over 10 years. Of note, for models that included relationship satisfaction as a predictor, the sample was restricted to the subsample of men in a “steady relationship” at baseline. Significant interactions were probed to determine the simple effect of relationship variables separately for BG and heterosexual men. Interactions were also probed at high and low levels of relationship status (i.e., “single” or “in a steady relationship”) and relationship satisfaction (i.e., 1 SD above and below the mean). All predictor variables were centered prior to inclusion in the regression models and tolerance values were above acceptable levels (all values > .70), minimizing concerns regarding multicollinearity amongst predictors. Sexual orientation identity was coded as heterosexual = 0 and BG = 1, and relationship status was coded as single = 0 and in a relationship = 1.

Table 2. Hierarchical Moderated Linear Regression Model Predicting EDI Drive for Thinness Scale Scores at 10-Year Follow-Up from Sexual Orientation and Relationship Status or Satisfaction.

	Follow-up EDI Drive for Thinness Scale Scores n=570					Follow-up EDI Drive for Thinness Scale Scores N=379
Predictors (all Men)	ΔR2	β	t	p	Predictors (Men in Relationships)	ΔR2	β	t	p
Step 1	.483***				Step 1	.482***
Baseline Drive for Thinness		.67	19.77	<.001	Baseline Drive for Thinness		.66	15.84	<.001
Age		-.01	-0.38	.70	Age		-.01	-0.29	.78
BMI		.04	1.27	.20	BMI		.09	2.11	.04
Step 2	.005				Step 2	.001
S.O. Identity		.01	0.29	.77	S.O. Identity		-.04	-1.12	.26
Relationship Status		-.06	-1.72	.09	Relationship Satisfaction		-.02	-0.50	.62
Step 3	.004*				Step 3	.006*
S.O. Identity X Relationship Status		-.07	-2.18	.03	S.O. Identity X Relationship Satisfaction		-.08	-2.11	.04
Total R2	.492***				Total R2	.489***

Note. EDI = Eating Disorder Inventory; BMI = body mass index; S.O. Identity = sexual orientation identity

^*p<.05

^***p<.001

Results

Table 1 presents mean comparisons between BG and heterosexual males for all variables. Age and BMI did not differ between BG and heterosexual men at baseline, with mean BMI values at baseline for both BG and heterosexual men falling within the normal weight status range (>18.5 kg/m² - <25 kg/m²). At baseline, BG men were more likely to be single and, among those in relationships, to have lower relationship satisfaction than heterosexual men. As expected, BG men endorsed higher eating pathology than heterosexual men, at both baseline and 10-year follow-up.

Table 1. Mean Differences on Baseline and Follow-Up Variables for Heterosexual and BG Men.

	Heterosexual (n=522)		BG (n=51)
Variable	M	SD	M	SD	F(1,571)/χ2*	p
Baseline
Age	29.41	8.65	27.40	9.09	2.44	.12
BMI	24.43	3.03	23.70	3.55	2.60	.11
Single	161	30.8	29	56.9	14.19	<.001
Relationship Satisfaction	5.94	1.19	5.41	1.40	4.05	<.05
EDI Drive for Thinness	8.79	1.45	12.44	6.62	32.92	<.001
EDI Bulimia	9.33	2.83	10.95	4.64	13.33	.001
10-year Follow-Up
EDI Drive for Thinness	9.71	3.36	12.65	6.03	29.56	<.001
EDI Bulimia	9.34	2.57	11.15	4.50	19.39	<.001

Note. BG = bisexual or gay; BMI = body mass index; EDI = Eating Disorder Inventory. For the “single” variable, descriptives are expressed in terms of the number and percentage of single heterosexual and BG males within the sample.

^*df F range=379-571; df χ²=1.

Drive for Thinness

Relationship Status

The overall regression model significantly predicted increases in Drive for Thinness over 10 years (R² = .492, F(6, 563) = 91.04, p < .001; see Table 2). As predicted, a significant interaction between sexual orientation identity and relationship status was found, such that for BG men, being in a relationship at baseline predicted decreases in Drive for Thinness scores over the 10-year span (β = -.26, p = .008) while there was no significant association for heterosexual men (β = -.04, p = .28). Further, among men who were single at baseline, being BG at baseline predicted increases in Drive for Thinness scores over time compared to heterosexual men (β = .10, p = .02). In contrast, for men in relationships, there were no differences in Drive for Thinness over time (β = -.04, p = .46). Importantly, the sexual orientation identity by relationship status interaction remained significant when controlling for follow-up relationship status (p = .02).

Relationship Satisfaction

For relationship satisfaction analyses, the overall regression model significantly predicted increases in Drive for Thinness scores over time (R² = .489, F(6,372) = 59.35, p < .001; see Table 2). Higher BMI at baseline was associated with increased Drive for Thinness scores over time. As predicted, there was also a significant interaction between sexual orientation identity and relationship satisfaction. For BG men, greater relationship satisfaction at baseline predicted decreases in Drive for Thinness scores over 10 years (β = -.30, p = .04). In contrast, for heterosexual men, no significant association was found (β = .01, p = .82). Relatedly, at high levels of baseline relationship satisfaction, being BG at baseline was associated with decreases in Drive for Thinness over time, at the level of significance (β = -.11, p = .050), whereas sexual orientation identity had no impact on change in Drive for Thinness at low levels of relationship satisfaction (β = .03, p = .57). In analyses controlling for follow-up relationship satisfaction, the interaction between sexual orientation identity and baseline relationship satisfaction was just above level of significance (p = .052).

EDI Bulimia Scores

Relationship Status

Regression analyses demonstrated that the overall model significantly predicted increases in Bulimia scores over time (R² = .540, F(5, 563) = 110.33, p < .001; see Table 3). Having a higher BMI and being younger at baseline were associated with greater maintenance of Bulimia scores over 10 years. Further, there was a significant main effect of sexual orientation identity, with BG males having higher bulimic symptoms compared to heterosexual men over the 10 year follow-up. No other significant predictors were found for Bulimia scores over time in this model.

Table 3. Hierarchical Moderated Linear Regression Model Predicting EDI Bulimia Scale Scores at 10-Year Follow-Up from Sexual Orientation and Relationship Status or Satisfaction.

	Follow-up EDI Bulimia Scale Scores n=570					Follow-up EDI Bulimia Scale Scores N=379
Predictors (all Men)	ΔR2	β	t	p	Predictors (Men in Relationships)	ΔR2	β	t	p
Step 1	.535***				Step 1	.563***
Baseline Bulimia Scores		.68	21.40	<.001	Baseline Bulimia Scores		.68	18.41	<.001
Age		-.07	-2.03	.04	Age		-.10	-2.76	.01
BMI		.10	3.03	<.01	BMI		.12	3.21	.001
Step 2	.005*				Step 2	.005
SexOr		.07	2.23	.03	S.O. Identity		.04	1.16	.25
Relationship Status		-.01	-0.31	.76	Relationship Satisfaction		-.04	-1.24	.21
Step 3	.000				Step 3	.002
S.O. Identity X Relationship Status		-.01	-0.15	.88	S.O. Identity X Relationship Satisfaction		-.04	-1.22	.22
Total R2	.540***				Total R2	.570***

Note. EDI = Eating Disorder Inventory; BMI = body mass index; S.O. Identity = sexual orientation identity

^*p<.05

^***p<.001

Relationship Satisfaction

The overall regression model significantly predicted changes in Bulimia scores over time (R² = .570, F(6,372) = 82.12, p < .001; see Table 3). Lower age and higher BMI at baseline predicted greater maintenance of Bulimia scores over time. There were no other significant predictors of Bulimia score trajectory.

Importantly, given possible differences across results between bisexual and gay men, we also ran analyses excluding bisexual men and compared beta weights for all follow-up analyses for comparisons of heterosexual men to BG men vs. comparisons of heterosexual men with gay men. Across all follow-up analyses, we found no differences in beta weights, further supporting our approach combining BG men. Additionally, given that the time lapse between baseline and follow-up variables was approximately 10 years, men's relationship status or satisfaction could change between assessment periods. Given the likely influence of relationship variables at follow-up on eating pathology (and vice versa) at follow-up, we also conducted analyses controlling for follow-up relationship variables. Importantly, results remained unchanged, increasing confidence in the prospective associations between baseline relationship variables and 10-year trajectory of eating pathology.

Discussion

Relative to their elevated risk status, BG men represent an understudied group within the field of eating disorders. Thus, research informing models of eating pathology within this population is imperative. The present study sought to examine how relationship status and relationship satisfaction affect the trajectory of eating pathology for BG men as a means for examining one model for risk in this group. Our results demonstrate that romantic relationships differentially impact drive for thinness over time for BG and heterosexual men, lending support to applications of objectification theory for understanding risk in BG men. For BG men, being single at baseline prospectively predicted greater increases in restrictive eating pathology, but not bulimic symptoms, 10 years later. Similarly, for BG men in relationships, lower relationship satisfaction prospectively predicted greater increases in drive for thinness at 10-year follow-up, but not bulimic symptoms. In contrast, for heterosexual men, relationship status and satisfaction demonstrated no impact on eating pathology trajectory.

By extending cross-sectional findings⁸ to demonstrate prospective associations between relationship variables and eating pathology over a 10-year follow-up, findings support Siever's¹ model for increased eating pathology in BG men. Consistent with his explanation, romantic relationships appear to have an enduring impact on restrictive eating pathology that may be explained by their influence on internalization of body ideals. Finding that effects for both relationship status and relationship satisfaction were specific to restrictive eating pathology provides stronger support that BG men in satisfied relationships may be buffered against pressures to be ultra-lean and muscular. It is also important to note that BG men were more likely to be single at baseline and, among those in relationships, more likely to report lower relationship satisfaction at baseline than heterosexual men. Thus, not only were BG men more influenced by these variables but they were also less likely to have these buffers in their lives at baseline. Both effects could contribute to increased eating pathology in this at-risk group. Finally, the lack of effect of relationship status and relationship satisfaction for heterosexual men replicates and extends previous findings demonstrating the minimal impact of marital status²³ and relationship satisfaction^{8, 19} on disordered eating in heterosexual men. However, it is notable that other romantic relationships variables, such as partner pressure to be lean, have been associated cross-sectionally with disordered eating in predominately heterosexual men.²⁷ Thus, while BG men may be more likely to be confronted with pressures to be lean and muscular in their relationship, when heterosexual men are confronted with these pressures, they also demonstrate increased eating pathology. Therefore, while relationship status and satisfaction may be more important for BG men, pressures to conform to the lean and muscular ideal within romantic relationships should be directly assessed in all men, regardless of sexual orientation.

Longitudinal relationship satisfaction results differ from previous cross-sectional research demonstrating that for BG men, higher relationship satisfaction was associated with reduced bulimic symptoms, but not restrictive eating pathology.⁸ Longitudinal results within the present study correspond with how romantic relationships may contribute to risk over time in models of eating pathology for BG men.¹² Consistent with possible applications of objectification theory, low relationship satisfaction for BG men may lead to greater body monitoring to attract the male gaze. This may increase pressures to be lean in order to attract a mate, leading to internalization of a lean and muscular ideal that could influence the trajectory of restrictive eating pathology. In contrast, cross-sectional associations between relationship satisfaction and bulimic symptoms may reflect the impact of bulimic symptoms on psychosocial function. ⁸

Internalization of the body ideal occurs at a young age,²⁸ and is reinforced through peers and potential or actual romantic partners.¹² Once ingrained, internalization of the body ideal may have an enduring influence on attitudes and behaviors that are compatible with the ideal (i.e., restrictive eating pathology), even up to 10 years later, when relationship status and/or satisfaction have changed. Future studies should examine whether internalization of the body ideal mediates this relationship by incorporating at least three waves of assessment.

While we suggest that our results may be explained by Siever's theory or applications of objectification theory, it is important to note that there are other possible theoretical frameworks, such as minority stress theory,²⁹ that could explain findings. Research supports that the chronic stress of stigma and discrimination that BG men face as being part of a minority group predict negative mental health outcomes,²⁹ including body image concerns.³⁰ Thus, high-quality, supportive romantic relationships may mitigate these chronic social stressors for BG men, buffering against eating pathology over time. Thus, longitudinal studies incorporating at least three waves of data collection could examine both sexual minority stress and internalization of the lean/muscular ideal as possible mediators of associations found in the present study.

While the present study found significant effects for relationship variables on the trajectory of restrictive eating pathology in BG men, it is important to note that only a small percentage of the overall variance was accounted for by these associations, and thus, other factors not evaluated in this study have an important impact. Further, while the longitudinal design of the current study is a strength, we were restricted to two assessment waves because questions about sexual orientation identity were not asked before 2002. Thus, models of mediation could not be included. Although our study included three age cohorts in which baseline data represented the transition from late adolescence to young adulthood, the transition from young adulthood to adulthood, and the transition from adulthood to midlife, our sample of BG men was too small to evaluate whether cohort moderated effects. Future research should examine whether there are key sensitive periods in which being in a relationship may exert the greatest influence over future risk for eating pathology. Further, analyses within the present study combined bisexual and gay men into one group (BG men) to increase statistical power. While several studies support this approach,^{8, 17-19} this increased the heterogeneity in the sample, which may have affected results. Relatedly, we do not know the gender of the partners for men who identified as bisexual. If bisexual men's steady partners were female, this could also have affected the results. However, it is important to note that we still found significant results in the expected direction combining BG men, and the beta weights in analyses conducted excluding bisexual men did not differ from those employing the combined BG group. Nonetheless, future studies should replicate these results with larger samples examining gay and bisexual men, separately, and across the spectrum of sexual minority men (gay, bisexual, transgender, questioning), to determine if there are any differences among these groups.

Additional limitations include the use of a selective sample of men from a prestigious northeastern university, and absence of other potential variables of interest that are more specific to eating pathology in men (i.e. drive for muscularity), or variables related to sexual minority stigma and discrimination. It is also very important to acknowledge that we did not assess relationship length and level of commitment given the need to employ a brief survey. Additionally, men may have interpreted the meaning of “steady relationship” differently, which may have introduced heterogeneity among responses, and we do not have information on the course of men's relationships between baseline and follow-up. Surveys did not assess whether men were in monogamous or non-monogamous (e.g., polyamorous) relationships. Men in non-monogamous relationships could be continuously interested in attracting partners, which also could have impacted results. Further, sexual orientation identity, relationship status, and relationship satisfaction were assessed using single items, thus increasing susceptibility to measurement error. Importantly, these limitations would all have the effect of increasing error which would diminish effect sizes and reduce the likelihood of finding significant effects. Thus, significant effects found in this first longitudinal study of relationship variables' impact on trajectory of eating pathology may underestimate the true effects of these variables, also contributing to the relatively small amount of variance explained by these variables in current models.

Despite these limitations, the present study benefited from several strengths. Key among these was the use of a longitudinal design in which we were able to establish prospective associations between predictors and outcomes. This is the first study, to our knowledge, to examine the prospective impact of relationship status and satisfaction on eating pathology over time in BG men. In addition, we used psychometrically sound measures of eating pathology that demonstrated good internal consistency, reliability, and longitudinal stability, which increase confidence that we had valid measures. Finally, we observed a high retention rate, particularly given the long duration of follow-up, and we observed minimal evidence of biased attrition and accounted for this in our approach to missing values.

Our results have potential implications for interventions for BG men, regardless of relationship status. Identifying that single BG men and those with low relationship satisfaction may be at higher risk for restrictive eating pathology provides an important opportunity for intervention and prevention. Specifically, BG men may benefit from a cognitive dissonance-based intervention³¹ that addresses and encourages participants to challenge cultural pressures to be lean and muscular in order to attract a male partner. If replicated, results from the present study could be incorporated into the content of dissonance-based interventions to discuss how BG men place pressure on one another to achieve the ideal and their opportunities to work against the objectification of other BG males. Further, building stronger relationships within the context of group-based interventions could also help buffer against sexual minority stress by increasing social support among BG males. Such efforts may have the potential to translate findings stemming from theoretical models of risk into interventions to prevent the development of eating disorders in BG men.