Comorbidity and age of onset of eating disorders in gay men, lesbians and bisexuals

Abstract

Objective

This study examines the prevalence of psychiatric disorders among lesbian, gay, and bisexual (LGB) men with eating disorders.

Method

388 white, black, Latino LGB men and women were sampled from community venues. DSM-IV diagnoses of anorexia, bulimia, and binge eating disorder were assessed using the World Health Organization’s Composite International Diagnostic Interview.

Results

Gay and bisexual men with eating disorders were more likely to have an anxiety or substance abuse disorder compared than gay and bisexual men without eating disorders, while lesbian and bisexual women with eating disorders were more likely to have a mood disorder compared to lesbian and bisexual women without an eating disorder. For individuals diagnosed with an eating and anxiety or major depressive disorder, the onset of the psychiatric disorder was more likely to precede the onset of the eating disorder.

Conclusion

Researchers should study potential explanations of the relationship eating and psychiatric disorders among LGB men and women.

1. Introduction

Research has consistently demonstrated that there is a high prevalence of comorbid psychiatric disorders among individuals with eating disorders. Studies using both clinical and community samples have shown that individuals with eating disorders are more likely to have comorbid mood or anxiety disorders than individuals without eating disorders (Bushnell et al., 1994; Garfinkel et al., 1995; Godart et al., 2003; Halmi et al., 1991; Herpetz-Dahlmann et al., 2001; Hudson et al., 1987a; 1987b; Lilenfeld et al., 1998; Rastam et al., 1995). Studies have also found that individuals with eating disorders have higher rates of specific phobia and generalized anxiety disorder compared to individuals without eating disorders (Garfinkel et al., 1995; Kendler et al., 1991; Lilenfeld et al., 1998; Walters and Kendler, 1995). Recent reviews of this literature have suggested that among individuals with eating disorders, the lifetime prevalence of comorbid anxiety disorders ranges from 23% to 75%, and of mood disorders, 24% to 90% (Godart et al., 2002; Swinbourne and Touyz, 2007). Further, having a psychiatric disorder may place individuals at risk for developing an eating disorder. Among individuals with eating disorders, 17–71% had a mood disorders and 57–94% had anxiety disorders with an onset prior to the eating disorder (Brewerton et al., 1995; Bulk et al., 1997; Deep et al., 1995; Godart et al., 2000, 2003; Kaye et al., 2004; Kendler et al., 1991; Schwalberg et al., 1992; Zaider et al., 2002).

Although anorexia and bulimia nervosa occur primarily in women, 5–20% of people with eating disorders are men (Braun et al., 1999; Carlat & Camargo, 1991; Woodside et al., 2001). Studies further suggest that a disproportionate number of these men are gay and bisexual. In both community (Olivardia et al., 1995) and clinical (Carlat & Camargo, 1997; Herzog et al., 1984) samples of men with eating disorders, 14 to 42%--compared with about 4% of the U.S male population (Mosher et al., 2005)—are gay or bisexual. Compared with heterosexuals, gay and bisexual men seem to have high prevalence of body image dissatisfaction (Morrison et al., 2004), eating disorder symptoms (Austin et al., 2009; Wichstrom, 2006), and eating disorders (Carlat et al., 1997; Feldman & Meyer, 2007a; Herzog et al., 1984).

Although it is a commonly held convention that lesbians and bisexual women are less likely to have an eating disorder than heterosexual women, several studies have found no significant differences between the two groups in terms of eating disorder diagnoses (Feldman & Meyer, 2007a) or eating disorder symptoms (Moore & Keel, 2002; Share & Mintz, 2002). In one study, Wichstrom (2006) found that lesbian and bisexual women had higher levels of disordered eating compared to heterosexual women. In a sample of 13,795 youth aged 12–23 years, Austin et al. (2009) found that participants who self-identified as “mostly heterosexual,” bisexual, and lesbian were more likely than participants who self-identified as heterosexual to report binge eating, but only “mostly heterosexuals” and bisexuals were more likely than heterosexuals to report purging. Further, Feldman and Meyer (2007a) found that there was no significant difference in the prevalence of full syndrome eating disorders between lesbian, gay, and bisexual (LGB) men and women.

Research has suggested that LGB individuals may also be at a particularly high risk for having a psychiatric disorder. Studies of the general population have shown that LGB individuals have more anxiety, mood, and substance use disorders than heterosexual individuals (Cochran et al., 2000; Cochran et al., 2003; Cochran & Mays, 2000a, 2000b, Gilman et al., 2001). A recent meta-analysis of such studies showed that LGB individuals have about twice the odds of mood, anxiety, and substance use disorders as heterosexuals (Meyer, 2003).

Despite the prevalence of both eating and psychiatric disorders in this population, there has been no comorbidity study using a community-based sample. Only one study has assessed eating disorder comorbidity using DSM criteria (Carlat et al., 1997), but this study did not have a comparison group of men without eating disorders with which prevalence rates of psychiatric disorders could be compared. There has only been one study that has examined the relationship between disordered eating and depression in lesbian and bisexual women, however it did not use DSM-IV criteria to define disorders (Strong et al., 2000). Research on eating disorder comorbidity has critical implications for the treatment of eating disorders as it can lead to early identification of risk for eating disorders and design of appropriate prevention interventions. Men with eating disorders have a higher prevalence of mood, anxiety, and substance abuse or dependence disorders compared to men without eating disorders (Mangweth et al., 1997; Olivardia et al., 1995; Striegel-Moore et al., 1999a, 1999b; Woodside et al., 2001). There has also been a considerable amount of research that has documented the comorbidity between eating disorders and psychiatric disorders in women (Godart et al., 2002, 2007; Swinbourne & Touyz, 2007). Most studies have found that there are no significant differences in terms of psychiatric comorbidity between men and women with eating disorders (Braun et al., 1999; Striegel-Moore et al., 1999a; Tanofsky et al., 1997; Woodside et al., 2001). Fewer studies found that men with eating disorders were significantly more likely to have a comorbid psychiatric diagnosis than women with eating disorders (Bramon-Bosch et al., 2000; Tanofsky et al., 1997). None of these studies, however, analyzed eating disorder comorbidity among sexual orientation subgroups.

To address this gap in knowledge about eating disorder comorbidity in LGB populations, we describe eating disorder comorbidity and age of onset in a non-probability community sample of LGB persons. Based on the existing literature in the general population, we hypothesize that LGB persons with a lifetime diagnosis of eating disorders will have a higher prevalence of other psychiatric disorders compared with LGB men and women without eating disorders. We further hypothesize that comorbid disorders will predate eating disorders.

2. Methods

2.1 Study Design

This study is based on data from Project Stride, a research study that investigated associations between stress, identity, and mental health among diverse groups defined by sexual, racial/ethnic, and gender identity in New York City. We report on a subgroup of study participants described below. Two previous papers have been written on eating disorders using these sample based on the entire sample (n= 516; Feldman and Meyer, 2007a) and then a subsample of gay and bisexual men (n=193; Feldman and Meyer, 2007b).

2.2 Sampling

Participants were sampled by direct solicitation in diverse New York City venues (e.g., business establishments, such as bookstores and cafes, social groups, outdoor areas, such as parks), and through snowball referrals. Sampling venues were selected to ensure a wide diversity of cultural, political, ethnic, and sexual representation within the demographics of interest. To reduce bias, venues were excluded from our venue-sampling frame if they were likely to over- or underrepresent people receiving support for mental health problems (e.g., 12-step programs, HIV/AIDS treatment facilities), or significant life events (e.g., organizations that provide services to people who have experienced domestic violence).

Between February 2004 and January 2005, 25 outreach workers visited a total of 274 venues in 32 different New York City zip codes. Participants were eligible if they were between the ages of 18–59 years old, New York City residents for two or more years, and self-identified as: a) heterosexual or lesbian, gay, or bisexual; b) male or female; and c) white, black or Latino (participants may have used other identity terms in referring to these social groups). Eligible participants were then invited to participate in a face-to-face interview that lasted a mean of 3.8 hours (SD = 55 minutes). The cooperation rate (the proportion of complete interviews out of eligible people who were contacted) for the study was 79% and the response rate (the proportion of complete interviews out of all eligible people, whether contacted or not) was 60% (calculated according to standard formulas of the American Association for Public Opinion Research, 2005). Response and cooperation rates did not vary greatly by gender, race, and sexual orientation. Recruitment efforts were successful at reaching individuals who resided in diverse New York City neighborhoods and avoided concentration in particular “gay neighborhoods” that is often characteristic of sampling of LGB populations. Interviewed individuals resided in 128 different New York City zip codes and no more than 3.8% of the sample resided in any one zip code area. The research protocol was reviewed by the Western Institutional Review Board.

2.3 Participants

This analysis is based on the subsample of LGB male and female participants (n = 388). The mean age for the gay and bisexual men in this sample is 33 (SD = 9). Of the men, 85% identify as gay, and 14% as bisexual, and 23% of the gay and bisexual men had a high school diploma or less education. The mean age for the lesbian and bisexual women in the sample is 32 (SD=9.5), and 78% identify as gay, while 21% identify as bisexual. Of the women, 21% had a high school diploma or less. The black and Latino participants had a lower socio-economic status as indicated by lower education and lower income.

2.4 Measures

Eating disorder diagnoses were made using the computer-assisted personal interview version 19 of the WMH-CIDI, a fully structured measure used in the National Comorbidity Study (www.hcp.med.harvard.edu/ncs) (Kessler and Ustun, 2004; Kessler et al., 2005). We assessed the presence of both lifetime and current (12 months) eating disorders, including full syndrome anorexia, bulimia, and binge eating disorder. To classify participants, we used the algorithms from Hudson et al.’s (2007) study of the prevalence of eating disorders in the National Comorbidity Survey Replication (www.hcp.med.harvard.edu/ncs/eating.php) with one exception: We used Hudson et al.’s stricter criterion for self-evaluation and considered a positive response to EA17f (did you feel like your self-esteem and confidence depended on your weight or body shape?) as satisfying criterion D for the DSM-IV bulimia diagnosis (self-evaluation is unduly influenced by body shape and weight), as opposed to using EA1 (was there a time in your life when you had a great deal of concern about or strongly feared being too fat or overweight?), EA17e (around the time you were binge eating, were you very afraid that you would gain weight?), or EA17f to satisfy this criterion. We also used Hudson et al.’s criteria for binge eating disorder, which includes three or more (instead of DSM-IV’s six or more) months of symptoms.

Consistent with others (Cotrufo et al., 2004; Woodside et al., 2001), we defined subclinical anorexia as characterized by (a) having a fear or gaining weight or becoming fat and (b) experiencing disturbance in how one perceives one’s body. Subclinical bulimia was defined using the same criteria as full syndrome bulimia except there was no requirement regarding the frequency of binging and compensatory behavior. We also used Hudson et al.’s (2007) algorithm for subclinical binge eating disorder, which was defined as binge eating episodes that occur at least twice a week for at least three months, and not during the course of anorexia, bulimia, or full syndrome binge eating disorder. All of the subclinical diagnosis categories included full syndrome and subthreshold cases. We used this expanded category in this study because the eating disorder literature has suggested that the full syndrome diagnosis criteria may be too restrictive (Sloan et al., 2005; Striegel-Moore et al., 2005).

Psychiatric diagnoses were also made using the WMH-CIDI (Kessler and Ustun, 2004; Kessler et al., 2005). We assessed DSM-IV lifetime and current (12 months) diagnoses for anxiety disorders (panic disorder, specific phobia, social phobia, generalized anxiety disorder), mood disorders (major depressive disorder and dysthymia; bipolar disorder was not a possible diagnosis with this version of the WMH-CIDI), and substance use disorders (alcohol and drug abuse and dependence).

Retrospective age-of-onset reports were obtained in the WMH-CIDI using the following questions that are designed to emphasize the importance of accurate responses. First, the respondent is asked if they can remember the exact age s/he had the syndrome. Participants who answer “no” were probed for a bound of uncertainty by moving up the age range incrementally (e.g., “Was it before you started school?” “Was it before you became a teenager?”). Age of onset was defined as the age at which all of the symptoms necessary to make a full syndrome diagnosis were present concurrently.

2.5 Statistical analysis

To test the differences in prevalence of lifetime and current psychiatric disorders among LGB men and women with a lifetime full syndrome or subclinical disorders as compared to those without eating disorders (the reference group) we used logistic regression analyses. In these analyses, we present the odds ratios (OR) and 95% confidence intervals (CI). Where risk estimate statistics could not be computed due to empty expected cells, we used Fisher’s exact test (drug dependence in gay and bisexual men, and alcohol abuse and dependence in lesbian and bisexual women). We used logistic regression analyses and Fisher’s exact test to analyze the gender differences in the lifetime and current prevalence of psychiatric disorders in LGB men and women with lifetime eating disorder diagnoses.

To test whether onset of comorbid disorders was more likely to have occurred before or after the onset of eating disorders, we used a one-sample test of proportion (Fleiss, 1981). Only participants with full syndrome eating disorders were included in the age of onset analysis. Data were analyzed using SPSS statistical software (Version 13.0).

3. Results

As displayed in Table 1, gay and bisexual men with a lifetime history of full syndrome or subclinical eating disorder were significantly more likely to have a lifetime diagnosis of specific phobia, generalized anxiety disorder, drug dependence disorder, any anxiety disorder, and any psychiatric disorder than gay or bisexual men without an eating disorder. Gay and bisexual men with a lifetime history of full syndrome or subclinical eating disorder were also significantly more likely to have a current diagnosis of alcohol abuse, drug abuse and dependence disorder, and any psychiatric disorder than gay or bisexual men without a lifetime history of having an eating disorder.

Table 1.

Lifetime and current prevalence of DSM-IV/WMH-CIDI psychiatric disorders in lesbians, gay men, and bisexuals with and without a lifetime diagnosis of eating disorders^* (n=388)

	Gay and bisexual men (n=193)						Lesbian and bisexual women (n=195)
	Comorbidity with lifetime disorders			Comorbidity with 12-month disorders			Comorbidity with lifetime disorders			Comorbidity with 12-month disorders
	ED	No ED		ED	No ED		ED	No ED		ED	No ED
			OR (CI)			OR (CI)			OR (CI)			OR (CI)
	(n=30)	(n=163)		(n=30)	(n=163)		(n=19)	(n=176)		(n=19)	(n-176)
	%	%		%	%		%	%		%	%
Anxiety disorders
Panic disorder	10	4	2.4 (0.6, 10.1)	3	3	1.3 (0.1, 12.7)	5	5	1.1 (0.1, 9.8)	5	3	1.9 (0.2, 17.1)
Specific phobia	37	19	2.4 (1.0, 5.7)	27	14	2.2 (0.8, 5.5)	37	26	1.6 (0.6, 4.4)	32	18	2 (0.7, 5.8)
Social Phobia	37	22	2.4 (0.9, 4.8)	20	12	1.7 (0.6, 5)	21	21	1.0 (0.3, 3.3)	21	14	1.7 (0.5, 5.5)
Generalized anxiety disorder	20	8	2.8 (1, 8.3)	10	6	2 (0.4, 7.4)	5	8	0.6 (0.08, 5.1)	5	3	1.9 (0.2, 17.1)
Any anxiety disorder	63	39	2.6 (1.2, 6)	43	27	2 (0.9, 4.4)	53	45	1.3 (0.5, 3.5)	53	31	2.4 (0.9, 6.3)
Mood disorders
Major Depressive disorder	37	25	1.7 (0.7, 4.0)	17	14	1.2 (0.4, 3.7)	58	32	2.8 (1.1, 7.5)	37	18	2.7 (0.9, 7.5)
Dysthymia	7	6	1.1 (0.2, 5.2)	3	6	0.6 (0.07, 4.8)	11	5.1	2.2 (0.4, 11)	5	4	1.3 (0.1, 11.5)
Any mood disorder	37	25	1.7 (0.7, 4)	17	14	1.3 (0.4, 3.7)	58	32	2.8 (1.1, 7.5)	37	18	2.7 (0.9, 7.5)
Substance abuse disorders
Alcohol abuse	13	14	0.9 (0.3, 3.0)	13	4	4.0 (1.1, 15.2)	5	17	0.2 (0.3, 2.1)	0	3	χ2= 0.66, nsa
Alcohol dependence	10	10	1.0 (0.2, 3.7)	3	3	1.1 (0.1, 9.6)	16	10	1.6 (0.4, 6.2)	0	1	χ2=0.21, nsa
Drug abuse	20	16	1.3 (0.5, 3.5)	17	4	5.2 (1.5, 18.4)	21	15	1.5 (0.5, 5.0)	11	3	3.3 (0.6, 17.8)
Drug dependence	23	5	5.8 (2, 17.8)	13	0	χ2=22.2, p< .001a	21	17	1.3 (0.5, 3.5)	5	2	3.2 (0.3, 32.4)
Any substance abuse disorder	46	35	1.6 (0.7, 3.5)	20	8	2.8 (1.0, 8.3)	47	39	1.4 (0.5, 3.6)	11	6.8	1.6 (0.3, 7.8)
Any comorbid psychiatric disorder	83	62	3.0 (1.1, 8.4)	63	38	2.8 (1.2, 6.3)	89	72	3.1 (0.7, 14.3)	68	45	2.6 (0.9, 7.3)

^*Includes cases who met criteria for a full syndrome eating disorder or a subclinical eating disorder.

^aFisher’s exact test performed because odds ratios are not calculable

Lesbian and bisexual women with a lifetime history of full syndrome or subclinical eating disorder were significantly more likely to have a lifetime diagnosis of a major depressive disorder (MDD) or any depressive disorder compared to lesbian and bisexual women without a lifetime history of an eating disorder. There were no significant differences between lesbian and bisexual women with and without a lifetime history of a full syndrome or subclinical eating disorder in terms of the prevalence of current psychiatric diagnoses. We also did not find any gender differences in psychiatric comorbidity among LGB men and women with lifetime eating disorder diagnoses.

The mean age of onset for a full syndrome eating disorder was 19.2 years (SD= 6.6). In Table 2, we present the proportion of participants with onset of comorbid disorders preceding the onset of eating disorders as compared with those with onset of comorbid disorders following the onset of eating disorder. Of 22 people with comorbid anxiety disorders, 19 (86%) had an anxiety disorder prior to the onset of the eating disorder, and only 1 (4%) had an anxiety disorder following the onset of an eating disorder. Of 14 people with a comorbid MDD, 11 (77%) had MDD prior to the onset of the eating disorder, and only 1 (7%) had MDD following the onset of an eating disorder. There were no significant differences in onset of comorbid substance use disorders relative to the onset of an eating disorder.

Table 2.

Gender differences in the lifetime and current prevalence of DSM-IV/WMH-CIDI psychiatric disorders in lesbians, gay men, and bisexuals with lifetime eating disorder diagnoses^* (n=49)

	Comorbidity with lifetime disorders			Comorbidity with 12-month disorders
	Men	Women		Men	Women
			OR (CI)			OR (CI)
	(n= 30)%	(n= 19)%		(n=30)	(n= 19)
	%	%		%	%
Anxiety disorders
Panic disorder	10	5.3	0.5 (0.05, 5.2)	3.3	5.3	1.6 (0.09, 27.4)
Specific phobia	36.7	36.8	1.0 (0.31, 3.3)	26.7	31.6	1.2 (0.4., 4.5)
Social Phobia	36.7	21.1	0.5 (0.1, 1.7)	20	21.1	1.0 (0.2,. 4.4)
Generalized anxiety disorder	20	5.3	0.2 (0.02, 2)	10	5.3	0.5 (0.04., 5.2)
Any anxiety disorder	63.3	52.6	0.6 (0.2, 2.1)	43.3	52.6	1.4 (0.4, 4.6)
Mood disorders
Major Depressive disorder	36.7	57.9	2.4 (0.7, 7.7)	16.7	36.8	2.9 (0.7, 11.1)
Dysthymia	6.7	10.5	1.7 (0.2, 12.8)	3.3	5.3	1.6 (0.09, 27.4)
Any mood disorder	36.7	57.9	2.4 (0.7, 7.7)	16.7	36.8	2.9 (0.7, 11.1)
Substance abuse disorders
Alcohol abuse	13.3	5.3	0.4 (0.04, 3.5)	13.3	0	χ2= 2.70, nsa
Alcohol dependence	10	15.8	1.7 (0.3, 9.4)	3.3	0	χ2=0.63, nsa
Drug abuse	20	21.1	1.0 (0.2, 4.4)	16.7	10.5	0.5 (0.1, 3.4)
Drug dependence	23.3	21.1	0.8 (0.2, 3.5)	13.3	5.3	0.3 (0.04, 3.5)
Any substance abuse disorder	46.7	47.4	1.0 (0.3, 3.2)	20	10.5	0.5 (0.08, 2.6)
Any comorbid psychiatric disorder	83.3	89.5	1.7 (0.3, 9.8)	63.3	68.4	1.2 (.04, 4.2)

^*Includes cases who met criteria for a full syndrome eating disorder or subclinical eating disorder

^aFisher’s exact test performed because odds ratios are not calculable

4. Discussion

This is the first study to examine the comorbidty and the age of onset of eating disorders using a reliable and valid measure of DSM-IV psychiatric disorder in a non-probability community sample of LGB persons. We found that LGB men and women with eating disorders were more likely than LGB men and women without an eating disorder to have a comorbid psychiatric disorder. We did not find significant gender differences in psychiatric comorbidity among LGB men and women with lifetime eating disorder diagnoses, which is consistent with prior research (Braun et al., 1999; Striegel-Moore et al., 1999a; Tanofsky et al., 1997; Woodside et al., 2001).

Our study has several limitations. First, our sample is based on non-probability sampling, which may bias prevalence estimate but even this should not affect our estimates of comorbidity. Of greatest concern would be a bias that led to over- or underrepresentation of individuals with mental disorders in our sample. Our sampling strategy was designed to minimize such bias and is a great improvement over current studies of eating disorders among LGB populations. These studies have primarily used two types of samples: college students (Lakkis et al., 1999) and clinical samples (Carlat & Camargo, 1997). Most these studies recruited volunteers by describing the study’s focus on disordered eating and body satisfaction. Such samples may overrepresent volunteers whose interest in the topic may have been motivated by having greater difficulties around body image and eating disorders than nonvolunteers. The latter, clinical samples cannot be used for estimating disorder prevalence. Also, because eating disorders are rare, our sample size did not allow us to examine the prevalence of comorbid psychiatric disorders by specific eating disorder subtype. Studies that employ larger samples of people with eating disorders and psychiatric disorders are necessary to continue to add to our knowledge of this topic among LGB populations.

We found that the gay and bisexual men with an eating disorder had a significantly higher prevalence of anxiety and substance abuse disorders than gay and bisexual men without an eating disorder. These results are consistent with studies of predominantly heterosexual men with eating disorders in the general population, who were more likely to have anxiety and substance abuse disorders compared with men without eating disorders (Mangweth et al., 1997; Olivardia et al., 1995; Striegel-Moore et al., 1999b, Woodside et al, 2001).

Lesbians and bisexual women with eating disorders were more likely to have a mood disorder than the lesbians and bisexual women who did not have an eating disorder. This too is consistent with what has been found in studies of eating disorders using samples comprised largely of heterosexual women (Bushnell et al., 1994; Hudson et al., 1987a; 1987b; Garfinkel et al., 1995; Herpetz-Dahlmann et al., 2001; Lilenfeld et al., 1998). Therefore, our results suggest that the patterns for comorbidity among LGB men and women are similar to those found in the general population.

We found that LGB individuals with eating disorders who had comorbid anxiety disorder (men) or MDD (women) were more likely to have had the comorbid disorders prior to—rather than simultaneous to, or after—the onset of the eating disorder. This is consistent with other studies that have reported that comorbid psychiatric disorders (excluding substance use disorders) predate eating disorders (Brewerton et al., 1995; Bulik et al., 1997; Deep et al., 1995; Godart et al., 2000, 2003; Kaye et al., 2004; Kendler et al., 1991; Zaider et al., 2002). Like other researchers, we found that substance use disorders did not predate eating disorders (Johnson et al., 2002, Stice et al., 2004: Zaider et al., 2002).

Our results are consistent with the hypothesis that anxiety and mood disorders could increase risk for eating disorders (Schwalberg et al., 1992). Other studies have also found that depressive symptoms and negative affect predicted the onset of eating disorder symptoms (Leon et al., 1999; Zaider et al., 2002). For example, Johnson et al. (2002) found that youths with depressive disorders and co-occurring psychiatric disorders during early adolescence were five times more likely than youths without any psychiatric disorders to develop eating disorders during middle adolescence or early adulthood.

Our study is one of first to describe patterns of comorbidity in LGB individuals. Future research should investigate psychiatric comorbidities of LGB youth with eating disorders. This is particularly important because of the high risk that LGB person have for both eating disorders (Austin et al., 2004; 2009; French et al., 1996; Williamson & Hartley, 1998) and psychiatric disorders (Fergusson et al., 2005; Kipke et al., 2007). It should be noted that our report of age of onset of eating disorders uses the age at which all criteria for a full DSM eating disorder diagnosis were met. In some cases, however, individuals may exhibit subclinical eating disorder symptoms at younger ages, preceding or independent of formal diagnosis of eating disorders. Because we focus on eating disorders as defined by DSM-IV criteria, we did not include these cases. Le Grange et al. (2006) compared subthreshold and threshold eating disorders and found that bulimia and eating disorder, not otherwise specified-bulimia nervosa (EDNOS-BN) were statistically indistinguishable in terms of subjective binge eating disorder episodes, dietary restraint, shape and weight concerns, and comorbid personality/psychiatric characteristics. Le Grange et al. suggested that the shared features between full syndrome bulimia nervosa and EDNOS-BN appeared more significant than the differences and served to highlight the clinical significance of EDNOS (e.g., the EDNOS-BN group engaged in clinically significant levels of binge eating and purging). Other studies have found that there are significant similarities between subthreshold and threshold eating disorders (Crow et al., 2002; Le Grange et al., 2004). Le Grange et al. suggested that the ED-NOS-BN participants’ eating disorder symptoms were sufficiently severe to warrant treatment, however these individuals have traditionally been excluded from outcome studies and third party reimbursement. The current criteria, therefore, may fail to identify individuals who do not meet criteria for a full syndrome eating disorder, however have clinically significant symptoms that indicate the need for treatment.

Future research may focus on comorbidity with subclinical eating disorder symptoms to provide greater insight into the development of eating disorders and patterns of comorbidity. Our study is a retrospective and correlational, preventing us from examining causal relations. Future studies should investigate the causal role of mood and anxiety disorders in the etiology of eating disorders. It is possible that eating disorders develop as a compensative behavior to cope with stress, anxiety, and depression. Behaviors such as bingeing and purging might be maladaptive ways to cope with negative affect or manage stress (Cattanach & Rodin, 1988; Heatherton & Baumeister, 1991). Studies have found that stress and trauma are associated with disordered eating (Crowther et al., 1999; Harrington, et al., 2006).

A stress and coping model for eating disorders is particularly important to consider in the context of LGB populations. Meyer’s (2003) minority stress model focuses on understanding the prevalence of health and mental health concerns among LGB individuals as a result of the chronic stress related to being part of a stigmatized population. In support of this model, studies have found that internalized homophobia is linked to disordered eating and poor body image among gay and bisexual men (Reilly & Rudd, 2006; Williamson & Hartley, 1998). Further, Feldman and Meyer (2007b) found that there was a significant relationship between having a history of childhood sexual abuse and an eating disorder diagnosis among gay and bisexual men.

The finding that mood and anxiety disorders precede eating disorders in LGB individuals has important implications for prevention. Assessment and identification of disordered eating, a precursor to eating disorder, among individuals with mood and anxiety disorders could help in preventing full-syndrome eating disorder. Clinicians who work with LGB patients should be cognizant of the risk for eating disorders when treating anxiety disorders in gay and bisexual men and depression in lesbian and bisexual women. Conversely, clinicians treating eating disorders in LGBs should be cognizant of the high risk for comorbid mood, anxiety and substance use disorders.

Table 3.

Age of onset of eating disorders^* and comorbid psychiatric disorders among lesbians, gay men, and bisexuals (n=31)

	Psychiatric disorder occurred before eating disorder	Psychiatric disorder developed simultaneous to or after eating disorder	z
aAny anxiety disorder (n=22)	86% (n=19)	4% (n=1)	9.23, p < 0.001
bMajor Depressive Disorder (n=14)	77% (n=11)	7% (n=1)	5.22, p < 0.001
Any substance use disorder (n=14)	57% (n=8)	43% (n=6)	0.54, p = 0.6

^*Includes cases who met criteria for a full syndrome eating disorder or a subclinical eating disorder.

^aAge of onset data was missing for 4 participants who were diagnosed with specific phobia. Two individuals had another anxiety disorder, and 2 participants had no other anxiety disorder, 9% (n=2).

^bMissing, 14% (n=2)