|Home | About | Journals | Submit | Contact Us | Français|
A growing number of studies among adult women have documented disparities in overweight adversely affecting lesbian and bisexual women, but few studies have examined sexual orientation-related patterns in weight status among men or adolescents. We examined sexual orientation group trends in body mass index (BMI; kg/m2), BMI Z-scores, and overweight using 56,990 observations from 13,785 adolescent females and males in the Growing Up Today Study, a large prospective cohort of U.S. youth. Participants provided self-reported information from six waves of questionnaire data collection from 1998 to 2005. Gender-stratified linear regression models were used to estimate BMI and BMI Z-score and modified Poisson regression models to estimate risk ratios (RR) for overweight, controlling for age and race/ethnicity, with heterosexuals as the referent group. Among females, we observed fairly consistently elevated BMI in all sexual orientation minority groups relative to heterosexual peers. In contrast, among males we documented a sexual-orientation-by-age interaction indicating steeper increases in BMI with age from early to late adolescence in heterosexuals relative to sexual orientation minorities. Additional prospective research is needed to understand the determinants of observed sexual orientation disparities and to inform appropriate preventive and treatment interventions. The long-term health consequences of overweight are well-documented and over time are likely to exact a high toll on populations with elevated rates.
Overweight has become one of the most critical public health issues affecting the nation today, and overweight in adolescence is of particular concern because of the high likelihood of continuing into adulthood.(1) Given its myriad health consequences, (2) elevated body mass index (BMI) in some population subgroups is likely to presage disparities in both acute and chronic diseases in the long-term. A growing number of studies among adult women have documented disparities in BMI adversely affecting sexual orientation minority women.(3) Working with population-based data from women ages 15 to 44 years participating in the National Survey of Family Growth, Boehmer et al. found higher rates of overweight (BMI ≥ 25 kg/m2) in lesbian (70.5%) compared to heterosexual (50.3%) women, though they did not find a higher rate in bisexual women (51.5%).(4) Other studies have found elevated BMI in both lesbian and bisexual women. For instance, in the Nurses' Health Study II, a longitudinal cohort study of women, Case and colleagues found a larger proportion of both lesbians (56.0%) and bisexual women (50.3%) ages 32-51 years to be overweight compared to heterosexual women (44.2%).(5)
Few studies have examined sexual orientation-related patterns in BMI among adult men or adolescents. In the National Health Interview Survey, a nationwide probability sample of U.S. civilian, noninstitutionalized adults ages 18 and older, Heck and Jacobson found high rates of overweight (BMI ≥ 25 kg/m2) in adult men living in households with male partners but even higher rates in men in households with women partners (56.6% vs. 71.5%).(6) The Geneva Gay Men's Health Survey, which collected data from males ages 15 and older in Switzerland, found gay men to be half as likely to be overweight compared to matched Swiss population controls.(7) However, one large statewide survey of adolescents in Minnesota found no differences in BMI across orientation groups in males or females.(8)
To further our understanding of sexual orientation differences in weight-status patterns in adolescents of both genders, we undertook a descriptive epidemiologic analysis using data from the Growing Up Today Study (GUTS), a large prospective cohort of U.S. youth. In addition, we sought to describe how sexual orientation-related weight-status patterns may vary by gender and age from early through middle to late adolescence.
Begun in 1996, the GUTS cohort was designed to assess diet, physical activity, and weight change in youth. All participants were ages 9 to 14 years at baseline and are the children of women in the Nurses' Health Study II.(9) With permission of their mothers, the children were enrolled in the GUTS cohort if they returned completed questionnaires at baseline: 9,039 girls and 7,843 boys were enrolled. Eight follow-up questionnaires have been administered in subsequent years, and the study is ongoing.(10) Paper questionnaires were mailed to participants at their home address each data collection wave for them to complete and mail back to study staff. Beginning in 2001, a password-protected online version of the questionnaire was also made available, which could be completed and submitted online to study staff. Participants were assured that their responses would be kept confidential and that no information about their survey responses would be given to their parents, health care providers, or others outside the research team. The Brigham and Women's Hospital institutional review board approved this study.
For the present analysis, adolescent and young adult females and males were included if they were ages 12 years to 23 years when they responded to GUTS questionnaires in one or more of six waves of data collection (1998, 1999, 2000, 2001, 2003, and 2005). Ninety-three percent of the cohort described their race/ethnicity as white.
In 1999, the GUTS questionnaire began including an item assessing sexual orientation, which was adapted from one used on the Minnesota Adolescent Health Survey.(11) The item read: “Which of the following best describes your feelings? (1) completely heterosexual (attracted to persons of the opposite sex), (2) mostly heterosexual, (3) bisexual (equally attracted to men and women), (4) mostly homosexual, (5) completely homosexual (gay/lesbian, attracted to persons of the same sex), (6) not sure.” Adolescents who chose the “mostly homosexual” or “completely homosexual” response options were too few to analyze separately, so responses from these participants were combined to create a lesbian category for females and gay category for males. Sexual orientation was assessed in 1999, 2001, 2003, and 2005, and group assignment was updated each year it was reported. Because orientation was not collected in 1998 and 2000, orientation reported in the subsequent year was assigned (i.e. 1999 used for 1998, 2001 used for 2000). To assess appropriateness of assignment of sexual orientation for 1998 and 2000 wave data, we conducted sensitivity analyses to investigate whether results differed when data from 1998 and 2000 were excluded. Direction of effects for sexual orientation subgroups did not change and magnitudes of associations were very similar; therefore, sexual orientation was assigned to data gathered in 1998 and 2000 as described above and included in all analyses.
Height and weight were self-reported at each survey wave and were used to calculate BMI (kg/m2). Cutoffs for overweight were coded using International Obesity Task Force (IOTF) standards, which has set age- and sex-specific BMI values for ages under 18 years that correspond with a BMI of 25 kg/m2 at age 18 years.(12) IOTF guidelines are designed to facilitate analyses of databases that include both children and adults. For participants 18 years and older, we defined overweight as BMI ≥ 25 kg/m2. For ages 12 to 20 years, BMI was converted to age- and sex-specific BMI Z-scores using Centers for Disease Control and Prevention (CDC) standards.(13)
To examine sexual orientation group differences in trends in BMI (kg/m2), BMI Z-scores, and overweight, we structured the six waves of longitudinal data into a person-period database for repeated measures analyses.(14) We used generalized estimating equations (GEE)(15) to account for correlated data resulting from repeated measures from the same individuals and clustering by siblings within the same family. Linear regression models estimated beta coefficients for BMI (kg/m2) and BMI Z-score and modified Poisson regression models(16) estimated risk ratios (RR) and 95% confidence intervals (CI) for overweight. Risk ratios are preferable to odds ratios in cases where the rare-outcomes assumption is violated (i.e. outcome prevalence greater than 10%).(17) Heterosexuals served as the referent group in all models, and sexual orientation was assessed at the same time that heights and weights were assessed except in the 1998 and 2000 survey waves, as described above. Multivariable models controlled for age and race/ethnicity and were stratified by sex. Age was modeled as a continuous variable and centered at the sex-specific mean for females (17.3 years) and males (16.9 years). For models estimating BMI (kg/m2), an additional age-centered-squared term was included to account for nonlinearity in the relationship between age and BMI.(18) Age-by-sexual-orientation-group interaction terms were tested. Models estimating sexual orientation differences in BMI (kg/m2) and overweight included observations from participants when they were ages 12 through 23 years. Models estimating BMI Z-score included observations from participants aged 12 through 20 years, as per CDC guidelines.(13)
To assess whether gender may modify associations between sexual orientation and attained weight status in young adulthood, we estimated the risk of overweight in the 21-23-year-old age group in models including data from both females and males and controlling for age, race/ethnicity, gender, sexual orientation, and gender-by-sexual-orientation interaction terms. All analyses were conducted using the SAS statistical package version 9.1.(19) SAS PROC GENMOD was used to generate GEE models.(15)
A total of 14,853 participants ages 12 to 23 years provided 63,327 observations in response to the 1998, 1999, 2000, 2001, 2003, and 2005 GUTS questionnaires. Of these observations, 4,534 were excluded for missing information on sexual orientation, and 499 observations were excluded for responding “unsure” to the orientation question, as very few described themselves as “unsure” at the older ages, making their numbers too small for analyses. An additional 1,286 observations were excluded for missing information on height or weight, and finally 18 observations were excluded as implausible heights, weights, or BMI values. To define implausible BMI values, we used CDC criteria for heights and weights (age- and sex-specific Z-score values below −6 or above +6) and BMI (age- and sex-specific Z-score values below −4 and above +5).(13) In addition, heights that shrank two or more inches in consecutive years were considered implausible and excluded. In sum, these exclusions resulted in an analytic sample of 56,990 observations provided by 13,785 participants, representing 81.7% of the GUTS cohort. (See Table 1.) The majority of observations included in analyses were from participants who responded to four or more of the six waves of data collection: 10.1% of participants contributed data at one time point, 11.1% at two points, 12.5% at three points, and 66.4% at four or more time points.
Figure 1 shows age- and sex-specific trends in mean BMI, mean BMI Z-scores, and frequency of overweight for each sexual orientation group. Among females throughout adolescence, sexual orientation minority subgroups fairly consistently reported higher BMI than did heterosexuals, though for lesbians, the P-value was marginal (P=0.06). BMI Z-scores were higher in bisexuals compared to heterosexuals, and “mostly heterosexual” and bisexual females were more likely to be overweight than heterosexual peers. The age-by-sexual orientation interaction terms did not reach statistical significance (Wald P>0.05 on 3 degrees of freedom for each outcome); though the 1 degree of freedom test of an age-by-bisexual orientation was significant in the BMI Z-score model. (See Table 2.)
Among males at the youngest ages, indicators of weight status were generally elevated in sexual orientation minorities compared to heterosexuals; however, sexual minority males had smaller increases in BMI over time through adolescence than did heterosexual males. By late adolescence, gay males in particular reported lower BMI than did heterosexuals. Visual examination of patterns for males in the Figure suggests that a cross-over in sexual orientation trends in BMI occurs around ages 17-18 years. Results of multivariable models with males indicated that age-by-sexual-orientation interactions were statistically significant (on 3 degrees of freedom, Wald P<0.01 in models predicting BMI and BMI Z-score and Wald P=0.02 in the model predicting overweight). (See Table 2.)
Additional analyses indicated that gender modified the relationship between sexual orientation and attained weight status by ages 21-23 years such that lesbian and “mostly heterosexual” females were more likely to be overweight and gay and “mostly heterosexual” males were less likely to be overweight compared to their same-gender heterosexual peers (the gender-by-sexual orientation interaction was statistically significant on 3 degrees of freedom, Wald P<0.01).
Working with six waves of data from a prospective cohort, we examined temporal patterns in weight status as they differentially affected sexual orientation subgroups throughout adolescence. Among females, we observed fairly consistently elevated BMI in sexual orientation minority groups relative to heterosexual peers. In contrast, among males we documented a sexual-orientation-by-age interaction not seen in females in our analyses, and this pattern was most pronounced in gay males. Specifically, in males we observed evidence of steeper increases in BMI with age from early to late adolescence in heterosexuals relative to sexual orientation minorities. We also observed a significant gender-by-sexual orientation interaction in attained weight status by ages 21-23 years. The prevalence of overweight in young adulthood was much higher in lesbians (39%) and heterosexual males (42%) than seen in heterosexual females (26%) and gay males (26%).
Studies with adult women have consistently found lesbian and bisexual women to have BMIs higher than those of heterosexual women.(3) Our research indicates that this disparity among women begins by adolescence. In addition, our study identified adolescent girls who describe themselves as “mostly heterosexual” as a group at elevated risk for overweight. Little attention has been given to understanding weight status among “mostly heterosexuals,” who are the largest sexual orientation minority subgroup (12.0% of females by age 22 years in our cohort compared to 4.3% who are bisexual or lesbian at those ages).
Our finding of higher BMI in heterosexual adolescent males relative to sexual orientation minority males by late adolescence is consistent with two studies in adult men.(6, 7) In contrast, our findings differ from one study conducted with a school-based sample of adolescents that did not find sexual orientation to be associated with BMI;(8) however, data from this Minnesota school-based study were collected more than 20 years ago, so discrepancies with our findings may be due to changes in temporal trends in childhood overweight or other factors.
In interpreting our finding of a steeper increase in BMI from ages 12-14 to 21-23 years in heterosexual compared to sexual minority males, it is important to note that heterosexual males in our cohort reported the lowest mean BMI and lowest prevalence of overweight of the four sexual orientation subgroups at ages 12-14 years and gay males report the highest. This difference at the youngest ages may be in part due to transient subgroup differences in maturation(20) in which boys who are aware of same-sex sexual feelings at ages 12-14 years and describe themselves as gay, bisexual, or “mostly heterosexual” at these ages may be further along in pubertal development than same-age peers who describe themselves as heterosexual. At subsequent ages, the heterosexual males as a group in our cohort make steeper gains in weight than the sexual minority subgroups during the years of pubertal development, perhaps a sort of maturational catch-up. Maturational catch-up, however, could not explain patterns observed by ages 21-23 years in attained BMI and weight status, a period in which gay males have been surpassed by the other subgroups and report the lowest attained BMI and lowest prevalence of overweight.
The multiple factors that are likely to underlie sexual orientation-related weight status disparities are not understood, but attention to recent research on stress and its relationship to weight-related behaviors and obesity may illuminate plausible pathways, particularly to explain disparities observed in females. Common manifestations of the stress response, both behavioral and physiological, have been linked to weight status. Stress-related coping behaviors, including unhealthful dietary and physical activity patterns, television viewing, and disrupted sleep patterns, have been identified as risk factors for weight gain and obesity.(21, 22) A study of college-age women found that higher stress was connected with binge eating.(23) Emotional vulnerability and difficulty regulating negative moods has also been positively associated with binge eating(24) and eating in the absence of hunger.(25) In children and preteens, television viewing is a common coping strategy in response to stress.(26, 27) Stress-related physiological perturbations, such as dysregulation of diurnal cortisol, have also been linked with obesity.(28-32)
Sexual minority adolescents experience unique stressors in environments that stigmatize minority sexual orientations.(33) For instance, lesbian, gay, and bisexual youth experience higher rates of bullying and violence victimization than do heterosexual youth.(34-36) Concealing one's sexual orientation can be a stressor(33), but also disclosure that results in conflict with and rejection by parents can contribute to stress on sexual minority youth.(37, 38) Youth whose parents are aware of their minority sexual orientation report significantly higher rates of verbal victimization related to sexual orientation than those whose parents are not aware.(39) Futhermore, at-school victimization(40) and family rejection(37) have been linked with elevated rates of substance use, suicidality, and sexual risk behaviors in sexual minority youth. Building on this evidence, it is plausible that sexual minority youth may also experience behavioral and physiological stress responses that have been linked with weight-related behaviors and weight status.
The observed sexual-orientation-related disparities in weight status suggest important differences in weight-related behaviors. A small number of studies have begun to examine weight-related behaviors of sexual orientation minority populations, but findings are sparse and contradictory. Most of these studies have been focused on adult women rather than youth or males. The Women's Health Initiative found that lesbians consumed fewer servings of fruits and vegetables per day than did heterosexual women.(41) Aaron and colleagues, in a study comparing a sample of 1,017 lesbians to the general population of women participating in the CDC's Behavioral Risk Factor Surveillance System Survey, found no orientation group difference in the samples in sedentary behavior (approximately one third of each group reported no physical activity)(42), and another study comparing heterosexual, bisexual, and gay men and women found no differences across sexual orientation groups in sports participation.(43) Yet, three studies with adult women have found lesbians to be more physically active than heterosexual women.(5, 42, 44) Higher rates of binge eating have been found in sexual orientation minorities, both female and male, compared to heterosexuals, including in the GUTS cohort.(8, 45, 46)
Importantly, in our study, when examining risk of overweight by young adulthood, we observed that gender significantly modified the relationship between sexual orientation and risk of overweight. Specifically, at ages 21-23 years, lesbian and “mostly heterosexual” females were at higher risk of overweight than heterosexual females; whereas, gay and “mostly heterosexual” males were at lower risk of overweight than were heterosexual males. This finding suggests that a hypothesized pathway as laid out above, from sexual-minority-related stress to behavioral and physiological stress responses to excess weight gain may apply to female more so than to male sexual minorities. It is possible that in response to sexual-minority-related stress, females are more likely than males to adopt specific coping behaviors linked with weight gain. Interestingly, an experimental study by Rutters et al. provides some support for gender differences in the effect of stress on excess caloric intake.(25) In a study of eating in the absence of hunger with adult women and men, Rutters and colleagues found not only that study participants exposed to stress consumed more kilocalories compared to those in the control condition but that the subset of participants who had scored high on a measure of disinhibited eating (defined as a loss of control over eating when palatable food is available or when emotionally upset(47)) were the most vulnerable to increased caloric intake in response to the stress condition. Women in the study had significantly higher disinhibited eating scores than did men.(25) Further research will be essential to examine the many questions that remain as to whether and how gender differences in weight-related aspects of stress responses and specifically disinhibited eating may contribute to the distinct gender-by-sexual-orientation patterns observed in weight status in adolescence and young adulthood.
Our study has several limitations. The GUTS cohort is not a representative sample, and participants, most of whom are of white race/ethnicity, are children of registered nurses and therefore are likely to encompass a more narrow range of the socioeconomic spectrum than exists in the U.S. as a whole. As a result, generalizability may be limited. It is worth noting, though, that our study is not affected by sexual orientation-related enrollment bias, as our sample was not recruited on the basis of sexual orientation. While the GUTS cohort is relatively large and our analyses included a total of 56,990 observations drawn from multiple assessments, small sample sizes particularly affected stability of estimates for lesbians and bisexual males, the two smallest subgroups. For instance, though data presented in the Figure would suggest that lesbians have relatively low BMI in the youngest age group and bisexual males to have relatively high BMI at the oldest age group, these are periods when their numbers are most sparse (see Table 1); therefore, patterns should be interpreted with caution. Alternately, there may be some meaningful group differences that were not found to be statistically significant due to small sample sizes (type II error). A strength of the sexual orientation item used on the GUTS questionnaires is that it is constructed as a scale and allows variability at the heterosexual end of the range, enabling us to identify the large subset youth who experience both-sex attractions and describe their feelings as “mostly heterosexual.” This group would have been overlooked had the survey included a conventional measure of sexual orientation identity, which typically uses the more condensed response options heterosexual, bisexual, and lesbian/gay.
On GUTS questionnaires, height and weight are self-reported, which is less accurate than measured values, though adolescents have been found to be able to provide valid reports of height and weight.(48) It is unknown whether there may be sexual-orientation-related bias in self-reported height and weight. For instance, it is plausible that gay males may underreport their weight more so than heterosexual males, due to greater weight concerns among gay males.(49, 50) Methodological studies will be needed to investigate this potential source of bias in research on weight status.
Our study demonstrates that sexual orientation group disparities in BMI are evident early in adolescence. We found a fairly consistent relationship between minority sexual orientation and elevated BMI in females, particularly at the older ages in the observed period. In contrast, in males we identified a significant age-by-sexual-orientation interaction. Gains in BMI from early through late adolescence in heterosexual males was notably higher than that in sexual orientation minority males. We also found a significant gender-by-sexual-orientation interaction in attained weight status by young adulthood in which sexual minority females but not sexual minority males were at elevated risk of overweight. The long-term health consequences of excess weight are well-documented(2) and over time are likely to exact a high toll on communities with elevated rates. Sexual orientation group disparities found among females in the GUTS cohort are as large as those observed in other national studies comparing BMI across racial/ethnic and socioeconomic groups.(51) Additional prospective research among both genders is needed to understand the determinants of these differences. Mechanisms driving orientation group patterns in weight status are likely to be multifaceted and need to be identified to inform effective preventive and treatment intervention strategies for all youth.
The authors would like to thank Catherine Berkey, PhD, Graham Colditz, MD, DrPH, Sarah Wylie, BA, and the GUTS team of investigators for their contributions to this paper and the thousands of young people across the country participating in the Growing Up Today Study. This study was funded by the Boston Obesity Nutrition Research Center, Robert Wood Johnson Foundation, and grants HD045763, DK46834, DK59570, and HL03533 from the National Institutes of Health. S.B. Austin, H.L. Corliss, and A.E. Field are supported by the Leadership Education in Adolescent Health project, Maternal and Child Health Bureau, HRSA grant 6T71-MC00009-17. J. Haines is supported by grant 200510MFE-154556-10955 from the Canadian Institutes of Health Research.
Disclosure: The authors have no conflicts of interest.