Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Arch Gen Psychiatry. Author manuscript; available in PMC 2010 June 11.
Published in final edited form as:
PMCID: PMC2883912

Prenatal Hormone Exposure and Risk for Eating Disorders: A Comparison of Opposite-Sex and Same-Sex Twins



Although the sex difference in eating disorder prevalence has typically been attributed to psychosocial factors, biological factors may also play a role. Prenatal testosterone exposure is a promising candidate, as it masculinizes behavior in animals and humans via its permanent effects on the central nervous system.


We examined whether in utero testosterone exposure has masculinizing effects on disordered eating (DE) by comparing opposite-sex (OS) and same-sex (SS) twins. Twin type (SS versus OS) is considered a proxy measure of prenatal hormone exposure, as females from OS pairs are exposed to more testosterone in utero than females from SS pairs. A linear trend in mean levels of DE was predicted based on expected prenatal testosterone exposure, with SS female twins exhibiting the highest levels of DE followed by OS female twins, OS male twins, and SS male twins.


Participants included 304 SS female twins, 59 OS female twins, 54 OS male twins, and 165 SS male twins from the Michigan State University Twin Registry (MSUTR).

Main Outcome Measures

Overall levels of disordered eating were assessed with the Minnesota Eating Behavior Survey.


Confirming hypotheses, DE exhibited significant linear trends with SS female twins exhibiting the highest levels of DE followed by OS female twins, OS male twins, and SS male twins. This linear trend could not be accounted for by levels of anxiety or socialization effects. Indeed, OS female twins exhibited lower levels of DE compared to an independent sample of undergraduate women (N = 69) who were raised with one or more brothers.


The masculinization of DE in OS female twins is unlikely to be due to socialization effects alone. Biological factors, such as the masculinization of the central nervous system by prenatal testosterone, may also contribute to sex differences in DE prevalence.

Keywords: eating disorder, disordered eating, sex difference, organizational effects, testosterone, twins

Clear sex differences in anorexia nervosa (AN) and bulimia nervosa (BN) exist in which females outnumber males by about 10:11. Sociocultural differences, such as pressure for thinness in women, have typically been used to explain this sex difference2, 3. However, biological factors may also play a significant role4.

Gonadal hormones (i.e., androgens, estrogens, and progestins) are one set of biological factors that may be particularly important given their influence on sexually differentiated behavior5, 6. Indeed, much of human sexual differentiation is thought to occur from organizational effects of the primary androgen, testosterone7, 8. During prenatal development, testosterone (either directly or through its metabolites dihydrotestoterone and 17 ß-estradiol) masculinizes cells throughout the body and the brain5. Male genital development is dependent upon the presence of testosterone, whereas female genital development occurs in the absence of testosterone7. Likewise, while testosterone masculinizes the developing central nervous system (CNS) in males, it is the absence of testosterone that averts such masculinization in females8, 9.

Animal studies confirm that testosterone's masculinization of the CNS translates into several sexually dimorphic behaviors, including eating behaviors. Animal studies have consistently shown that females rodents adjacent to males in utero later exhibit masculinized physiology (e.g., they have later vaginal opening, display greater sensitivity to testosterone), morphology (e.g., they have longer anogenital distance), and patterns of behavior (e.g., they mate and impregnate later, they are more likely to mount other females and exhibit increased aggression) (see Ryan & Vandenbergh10). The magnitude of the masculinization depends on many factors, with female rodents adjacent to males tending to be intermediate between other female and male rodents on a given trait, rather than completely masculinized10. Male rodents adjacent to other males show similar effects (e.g., they display greater sensitivity to testosterone, longer anogenital distance, and increased aggression), but the effects are less pronounced than those in females. These masculinized patterns of behavior have been attributed to the prenatal transfer of testosterone from the male pup to the sibling through the mother's bloodstream11 or amniotic fluid12, 13, as the masculinizing influence of adjacent male fetuses is blocked when mothers are treated with antiandrogens (see Clemens et al.14). These findings suggest that variations in testosterone exposure during critical periods of development can permanently alter wide-ranging sexually dimorphic traits.

The masculinizing effects of early testosterone exposure extend to food intake in male and female mammals. In gonadally intact mammals, testosterone increases food intake which is considered a “masculinized” effect. For example, female rats neonatally treated with testosterone show increased (i.e., masculinized) food intake15, 16 and body weight17 as adults. Ovariectomized female rats treated with androgens show increased food intake and greater sensitivity to the effects of testosterone on feeding16. Similarly, male rats castrated on the day of birth show permanent decreases in body weight and decreases in responsiveness to the effects of testosterone on food intake16.

Prenatal exposure to testosterone in humans cannot be readily examined or manipulated. Consequently, researchers have studied OS twins as a proxy for the masculinizing effects of testosterone on human behavior. Females from OS twin pairs share a prenatal environment with a male co-twin and are thought to be exposed to increased levels of prenatal testosterone10. Consistent with expectations, these females exhibit more masculine personality characteristics (e.g., increased sensation seeking18, 19 and aggressive behavior19) than females from SS twin pairs. Importantly, these differences have not been systematically related to pubertal stage, menarcheal status, or the activational effects of circulating testosterone19. Postnatal socialization (i.e., being raised with a male co-twin) could conceivably influence these masculinized patterns; however, masculinization has been observed for traits that are unlikely to be socially influenced. For example, females from OS twin pairs exhibit more masculine cerebral lateralization (i.e., left-hemisphere dominance when processing verbal-auditory stimuli20), spatial ability21, otoacoustic emissions22, tooth size23, and dental asymmetry24. Thus, prenatal testosterone exposure likely alters a variety of sexually dimorphic traits in females from OS twin pairs, and importantly, the masculinization of these traits is at least partially independent of psychosocial factors.

To date, no study has used the OS twin study design to examine the masculinizing effects of prenatal testosterone on disordered eating (DE) in humans. As noted above, animal studies consistently support the masculinizing effects of prenatal testosterone exposure on feeding behaviors. Given the central role of food intake in all forms of eating disorders, prenatal testosterone seems likely to also influence DE symptoms.

In addition, a recent study found significant associations between increased prenatal testosterone exposure (i.e., measured using 2D:4D finger-length ratios25) and lower levels of DE in adult women4. These findings were significant in suggesting that prenatal testosterone exposure may masculinize brain circuits that affect disordered eating in women and serve as a protective factor against eating disorder development. Nonetheless, these data cannot speak to whether prenatal testosterone accounts for sex differences in DE. Studies comparing OS to SS twins might clarify whether sex differences in prenatal testosterone explain sex differences in eating disorder prevalence.

The current study is the first to investigate the potential masculinizing effects of prenatal testosterone exposure on sex differences in DE using the OS twin study design. DE was compared across four groups of twins (i.e., females from SS twin pairs, females from OS twin pairs, males from OS twin pairs, males from SS twin pairs) to determine whether levels of DE differ by expected levels of prenatal testosterone exposure. A linear trend for DE was hypothesized, such that females from SS twin pairs would show the highest levels of DE, followed by females from OS twin pairs, males from OS twin pairs, and males from SS twin pairs. Importantly, we also examined potential confounds to these predicted linear relationships. First, we compared levels of anxiety across OS and SS twins. Anxiety is a significant risk factor for DE26 that has been shown to be influenced by the organizational effects of testosterone27. Given these links, we investigated whether our proxy for prenatal testosterone exposure (i.e., twin type) demonstrated effects on DE that were independent of its effects on anxiety. In addition, we sought to rule-out psychosocial explanations (i.e., socialization from growing up with a male co-twin) for masculinization of DE by comparing females from OS twin pairs to non-twin females raised with one or more brothers. Females from OS twin pairs were expected to exhibit more masculinized levels of DE than non-twin females who were raised with a male sibling.


This study was approved by the Michigan State University Institutional Review Board. Written informed consent was obtained after study procedures were explained to all participants.



The sample included 582 twins participating in the community-based Michigan State University Twin Registry (MSUTR28). Sample sizes by twin type were as follows: 304 (172 monozygotic (MZ; identical); 132 dizygotic (DZ; fraternal)) same-sex female twins, 59 DZ opposite-sex female twins, 541 DZ opposite-sex male twins, and 165 (103 MZ; 62 DZ) same-sex male twins. Sample sizes vary slightly in statistical analyses due to missing data (see discussion below).

Since rates of disordered eating vary across ethnic groups29, ethnicity was assessed using a demographic questionnaire to ensure that it does not account for differences in DE across groups. The majority of twins self-identified as Caucasian (84.4%) and were largely in the middle-to-upper levels of socioeconomic status (see Table 1) based on the Hollingshead index of social position30. There were no significant differences in ethnicity (X2 = 14.02, df = 12, p = .30), age (F (3, 578) = 0.86, p = .46), or socioeconomic status (X2 = 15.21, df = 12, p = .23) between twin types.

Table 1
Demographic information for twin and non-twin participants.

The MSUTR used a variety of recruitment methods (see Klump & Burt28 for a detailed description), including newspaper and email listserv advertisements, recruitment flyers, and recruitment mailings sent to twin pairs identified by university registrars’ offices. The majority of the sample (94.2%) was recruited using these methods. Additional twin pairs (5.8%) were recruited later in the study through the use of birth records obtained from the Michigan Department of Community Health (MDCH). Importantly, MSUTR twins appear to be representative of men and women from the general population of Michigan in terms of racial categorization (see Table 2). Indeed, there were no significant differences in participant race between MSUTR twins and individuals from the recruitment region (i.e., within a 2-hour drive of MSU).

Table 2
Racial Categories for MSUTR Twins and Individuals from the Recruitment Region.

Most (86.2%) twin participants completed the assessments in our MSU laboratory. However, if a twin could not travel to our MSU laboratory, a questionnaire packet was mailed to their home (14% of the sample). Twins who completed mailed assessments (M = 7.90; SD = 6.03) had significantly higher levels of DE than those assessed in the laboratory (M = 6.13; SD = 5.79; t(1, 579) = 2.46; p = .01). Thus, assessment method was included as a covariate in DE analyses (see Statistical Analyses below).

Non-Twin Controls

The non-twin control sample included 69 undergraduate women with at least one brother with whom they were raised. These women were recruited in undergraduate classes and a volunteer research pool at a large Midwestern university. Like the twin sample, the majority of non-twin participants were Caucasian (72.5%) and in the middle-to-upper level income class range (see Table 1). Importantly, OS female twins and non-twin participants did not significantly differ in terms of age (t (1, 125) = 0.21, p = .84) or socioeconomic status (X2 = 3.28, df = 4, p = .51). However, the non-twin participants showed a trend towards being significantly more ethnically diverse than the OS twin sample (X2 = 7.61, df = 4, p = .11). Consequently, ethnicity was included as a covariate in analyses comparing OS twin and non-twin women.


Disordered Eating

The Total Score from the Minnesota Eating Behavior Survey2 (MEBS31, 32) was used to assess overall DE including levels of body dissatisfaction (i.e., dissatisfaction with the size and/or shape of one's body), weight preoccupation (i.e., preoccupation with dieting, weight, and the pursuit of thinness), binge eating (i.e., the tendency to engage in episodes of overeating as well as having attitudes conducive to binge eating), and compensatory behavior (i.e., the tendency to use or contemplate using behaviors such as self-induced vomiting and laxatives to control weight). Higher scores on the Total Score indicate higher levels of pathological eating attitudes and behaviors. Importantly, these types of DE symptoms have shown robust sex differences32, with females exhibiting significantly more of these symptoms than males. Previous research has supported the reliability and validity of this measure in both genders32. The MEBS Total Score successfully discriminates between girls with clinical eating disorders (i.e., AN and BN) and age-matched, non-eating disorder controls32. In addition, the MEBS factor structure has been replicated across male and female samples32.

In the current study, the MEBS Total Score internal consistency was excellent for females (α = .90) and males (α = .84). There was also ample variability in MEBS scores, with a significant proportion of SS male twins (1.8%; n =3), OS male twins (5.6%; n =3), OS female twins (12.1%; n = 7), SS female twins (18.6%; n = 57), and non-twin females (18.8%; n = 13) scoring above the MEBS Total score mean (score = 15) for young adult women with AN or BN (Minnesota Twin Family Study, unpublished data, 2007).


The State-Trait Anxiety Inventory (STAI33) was used to assess trait levels of anxiety (i.e., stable or enduring anxiety) in the twin sample only. The trait subscale exhibits expected sex differences, with higher levels of trait anxiety in females27. Previous research has also documented its reliability and validity in both genders33. Internal consistency of this subscale was excellent in the current sample (i.e., male twins α = .88; female twins α = .94).


Body Mass Index (BMI; Weight (in kilograms)/ Height (in meters) squared) was calculated and used as an indication of adiposity. In the twin sample, height was measured with a wall mounted scale, and weight was measured with a digital scale. For the non-twins, height and weight were assessed via self-report. Previous research suggests that self-reported height and weight are valid and reliable measures of actual height and weight34-36. There were no differences in BMI between OS and SS twins of the same gender. However, because males had higher BMIs than females, and BMI was related to the MEBS Total Score in both genders (r's = .12-.68; all p's < .001), BMI was included as a covariate in all DE analyses.

Statistical Analyses

Data Preparation

Missing data were addressed using a prorating system. If 10% or less of the MEBS Total Score or STAI scale items were missing, then scores were prorated. If more than 10% of the items were missing, the scores were coded as missing. In total, MEBS data were coded as missing for one (1.7%) OS female twin, and STAI data were coded as missing for two (3.7%) OS male twins and three (1%) SS female twins. The MEBS score exhibited a positively skewed distribution and was therefore log transformed prior to analyses. A p value of .05 was used for all analyses.

Comparisons of OS and SS Twins

Hierarchal linear models (HLM) were used to examine mean differences in DE and anxiety across OS and SS twins. HLM is an extension of the general linear model (GLM) where factors are assumed to have a linear relationship with the dependent variable. HLM is ideal for examining the hypotheses in this study, as the non-independence of the twin data can be accounted for by nesting a level one variable within a level two unit. In the case of twins, the individual twins are the level 1 unit, while the twin pair (identified by a “family” variable that is shared by co-twins) is the level two unit.

Two categorical predictors were used in the HLM analyses: twin sex (male or female) and co-twin's sex (male or female). Each level of these predictor variables could have a different linear effect on the dependent variable (e.g., disordered eating). A significant effect of prenatal testosterone exposure in these analyses would be suggested if there is a main effect for co-twin's sex (e.g., levels of DE differ significantly as a function of having a male or female co-twin). This statistical model has been referred to previously as the Actor-Partner Interdependence Model37, 38.

Separate HLM models were conducted for DE and anxiety. An additional HLM mediation model, following the Baron & Kenny39 method, was then run with co-twin's sex as the predictor, trait anxiety as the mediator, and twin sex, BMI, and assessment method as covariates. This model examined whether anxiety mediates the association between prenatal testosterone exposure (measured via co-twin's sex) and DE.

Following recommendations37, 38, categorical variables (i.e., twin sex, co-twin's sex, assessment method) were effect coded (i.e., sex: 1 for males, -1 for females; assessment method: 1 for home assessment, -1 for laboratory assessment), and continuous variables (i.e., BMI, anxiety) were centered prior to conducting HLM analyses.

Comparisons of OS Twins and Non-Twins

An analysis of covariance (ANCOVA) (accounting for BMI and ethnicity) was used to investigate mean differences in DE between OS female twins and non-twin females. Notably, because having more than one biological older brother has previously been associated with increased prenatal testosterone exposure40, Pearson correlations were used to examine whether the number of older brothers was significantly associated with DE in the twin and non-twin samples. There were no significant associations between number of biological older brothers and DE (non-twin sample: r = .07, p =.16; twin sample: r = -.11, p = .41).


Comparisons of OS and SS Twins

Table 3 presents results from the HLM analyses. As expected, a significant main effect of twin sex on DE was observed, with females reporting significantly higher levels of DE than males (see Table 3). More importantly, however, findings suggested that the co-twin's sex also influences levels of DE. A significant main effect of co-twin's sex was found for the MEBS Total Score (see Table 3), indicating that in both males and females, decreased (i.e., more masculinized) levels of DE are associated with having a male co-twin. Examination of the means revealed the expected linear trend where SS female twins, who purportedly had the lowest level of prenatal testosterone exposure, showed the highest levels of DE, followed by OS female, OS male, and SS male twins (see Figure 1). Trend tests confirm the significance of this linear trend [F(1, 577) = 92.27; p < .0001]. Importantly, this linear trend cannot be accounted for by increased concordance for high or low levels of DE in MZ female or male twins, respectively. When analyses were re-run including only DZ twins, HLM results remained unchanged (see Table 3). The linear trend also remained significant [F(1, 301) = 49.80; p < .0001], with SS female DZ twins having the highest mean level of DE, followed by OS female DZ twins, OS male DZ twins, and SS male DZ twins.

Figure 1
Mean scores by sex of twin and sex of co-twin: a) Minnesota Eating Behavior Survey (MEBS) Total Score; b) State-Trait Anxiety Inventory (STAI) Trait Anxiety score. The sex of twins are represented along the x-axis, and the sex of co-twins are represented ...
Table 3
HLM results for disordered eating and anxiety symptoms in twins.

HLM results for anxiety paralleled those for DE. Main effects for twin sex and co-twin's sex were observed (see Table 3), as were linear trends in anxiety means across all twins [F(1, 573) = 39.83; p < .0001; see Figure 1] as well as the DZ twin sample only [F(1, 297) = 17.28; p < .0001]. Importantly, however, the mediation model in the full twin sample showed that anxiety only partially mediates associations between prenatal testosterone exposure (i.e., measured by co-twin's sex) and DE (see Table 4). Specifically, although the relationship between co-twin's sex and DE (Step 1) decreased when anxiety was included in the model (Step 4), co-twin's sex remained a significant predictor of DE.

Table 4
HLM examining anxiety as a mediator of the relationship between co-twin's sex and disordered eating in all twins.

Comparisons of Female OS Twins and Non-Twins

ANCOVA results indicated a significant main effect for group in which females from OS twin pairs reported significantly lower levels of DE than non-twin females with at least one brother (see Table 5). Thus, socialization from growing up with a male sibling does not appear to account for the masculinization of DE in OS female twins. Nonetheless, since the magnitude of age differences between non-twin female participants and their brothers may have influenced the amount of time spent together, a post-hoc ANCOVA was run after eliminating non-twin females whose closest-in-age brother was greater than 2 years older or younger. Results remained unchanged, as OS female twins (M = 6.59, SD = 5.63) continued to have lower levels of DE than non-twin females who had at least one brother (M = 9.01, SD = 4.65, N = 22) [F(1, 73) = 5.42; p = .02, d = .45]. In addition, post hoc Pearson correlations indicated no significant association between DE and the magnitude of age difference between non-twin females and their closest-inage brother (r = .05, p = .70).

Table 5
Mean differences in disordered eating between female opposite-sex twins and non-twins.


This was the first study to use OS and SS twin pairs to explore the masculinizing effects of prenatal testosterone exposure on the development of DE. Overall, our data were consistent with a “free martin” effect (i.e., masculinized behavior due to in utero exposure to testosterone41, 42) and provide evidence that increased levels of prenatal testosterone exposure may masculinize DE. Specifically, a linear trend in mean levels of DE was observed, where SS female twins showed the least masculinized (i.e., highest) DE, followed by OS female twins, OS male twins, and SS male twins. Although several factors likely influence sex differences in eating pathology, our findings highlight that biological factors likely play a role in this process. Specifically, masculinization of the CNS by prenatal testosterone may decrease risk for DE in males versus females.

The masculinization of DE does not appear to be completely explained by levels of anxiety or by socialization effects. While anxiety partially mediated the association between prenatal testosterone exposure and disordered eating, supporting some shared transmission43, primary conclusions remained unchanged even when controlling for levels of anxiety. Further, OS female twins had lower levels of DE than non-twin females who were raised with a brother. Admittedly, there may be something unique about having a brother the exact same age that accounts for the significantly lower levels of DE in OS female twins compared to non-twin females. Although plausible, it is unclear how having a same-age brother (rather than a brother who is close in age, but not the same age) could fully account for the lower (i.e., masculinized) levels of DE in OS female twins compared to non-twin females. Moreover, findings remained unchanged even after controlling for age differences between non-twin females and their brothers.

The masculinized pattern of DE observed in our OS female twins parallels previous findings in animals (e.g., intrauterine position effects; see Ryan & Vandenbergh10) and humans (e.g., girls with congenital adrenal hyperplasia44, 45) that support the masculinization of sexually dimorphic traits by prenatal testosterone exposure. If prenatal testosterone exposure does masculinize DE, as it does with several other sexually dimorphic traits, it is important to understand the mechanisms underlying the effects.

There are at least two ways in which prenatal testosterone exposure may influence predispositions to DE: 1) testosterone may organize anatomical and functional differences between the sexes in terms of key DE characteristics (e.g., food intake, satiety) that increase risk for eating disorders, and/or 2) testosterone may have organizational effects on sensitivity to circulating ovarian hormones later in life, which then differentially activates predispositions to DE. While definitive data are not available to support either of these specific mechanisms, speculative hypotheses can be proposed on the basis of preliminary animal and human studies.

Extant animal research indicates that early testosterone exposure organizes food intake and accounts for well-documented sex differences in feeding and body weight (i.e., male rats eat and weigh more than females). These effects in animal research are isolated to feeding behaviors (i.e., increased and decreased food intake) rather than cognitive components (e.g., weight preoccupation, body dissatisfaction) of eating disorders. Nonetheless, biological mechanisms for sex differences in DE may map more directly onto these core feeding behaviors than the cognitive features of eating disorders. If so, the masculinization of DE in males and OS female twins may partially occur through the organization of brain structures or functions that influence food intake. Structural brain differences between males and females are region-specific46, with significant sexual dimorphisms in brain areas (e.g., anterior cingulate, amygdala, orbital frontal cortex, hypothalamus, insula46) involved in food intake and body weight (for review, see Kaye et al.47). These areas are also marked by high levels of sex steroid receptors during critical periods of early brain development , suggesting that their structural differentiation may be influenced by prenatal and perinatal testosterone exposure46, 48, 49.

Sex differences in DE may also be due to sex differences in brain function. Few studies have examined this possibility, although two studies observed sex differences in brain responses to hunger50 and satiety50, 51 in several brain regions (e.g., dorsolateral prefrontal cortex). Clearly, additional studies examining sex differences in brain function are needed, but initial results are promising in suggesting that males and females may differ in food-related brain processes that are important for eating disorders.

Prenatal testosterone exposure may also affect DE by decreasing CNS sensitivity to gonadal hormones in adulthood. Prenatal testosterone exposure in female rodents decreases their sensitivity to ovarian hormones later in life16. These effects may be particularly important for feeding behaviors, since ovarian hormones directly influence food intake and DE in adult women. Specifically, decreased levels of estrogen, and increased levels of progesterone, are associated with increased food intake and binge eating in clinical52 and non-clinical53 samples of adult women. Indeed, changes in ovarian hormones appear to drive menstrual cycle changes in binge eating52, 53. Therefore, lower rates of DE in adult OS female twins may be due to increased prenatal testosterone exposure which decreases sensitivity to the activating effects of ovarian hormones on DE. While all of these hypotheses are speculative, they are worthy of additional investigation as they highlight potential pathways between the masculinizing effects of testosterone and correspondingly lower levels of DE symptoms.

Despite the strengths of this study, several limitations must be noted. First, participants were not clinically diagnosed with an eating disorder. Therefore, it is unclear how well the study's findings generalize to clinical populations. Although DE symptoms lie on a continuum54-57, show prospective associations with EDs58, and are considered precursive58, it will be important for future studies to replicate our results with clinical populations.

Second, the extent to which our participants are representative of the general population is somewhat unclear. Our twin samples appear to be roughly representative of the surrounding population based on racial categories. Non-twin female participants (who were college students) were comparable to OS female twins in terms of age and socioeconomic status, and roughly comparable in ethnicity. Nonetheless, future studies should use larger, population-based samples to replicate our results and ensure generalizability of the findings.

Third, we were unable to directly assess levels of prenatal testosterone exposure. The most empirically supported and accepted view (see Collaer & Hines6) is that prenatal testosterone and its metabolites (i.e., dihydrotestoterone and 17 ß-estradiol) masculinize the brain and sexually dimorphic behavior in males and females6; thus, prenatal testosterone exposure seems very likely to play some role in our observed effects. Evidence for the feminizing effects of prenatal estrogen in males is conflicting, with most studies showing no significant effects6. Nonetheless, we did not directly measure prenatal gonadal hormone exposure, and thus, cannot definitively rule-out possible feminization (or demasculinization) of DE by prenatal estrogen exposure in OS male twins. Future research therefore is needed to confirm that the masculinization effects are due to prenatal testosterone exposure.

Fourth, information on placenta type (i.e., monochorionic or dichorionic) was not available for MSUTR participants, but this information would be important to examine in future twin studies of prenatal effects. Finally, future research should also investigate differences in brain processes related to feeding (e.g., hunger, food intake, satiety) and differential sensitivity to ovarian hormones in OS versus SS twins. These investigations will likely provide new insight into how prenatal hormone exposure alters biological mechanisms, and ultimately, influences predispositions to DE in males and females.

In conclusion, findings from this study contribute to the literature on sex differences in eating disorder prevalence. Sociocultural factors (e.g., pressure for thinness in women) have typically been used to explain the sex difference in eating disorder prevalence. However, our results suggest that the masculinizing effects of prenatal testosterone, characteristic of male development, may also play a significant role.


Funding/Support: This research was funded in part by grant T32-MHO70343 (KMC) from the National Institute of Health; R21-07054201 (KLK) from the National Institute of Mental Health; MH58703 (SMB) from the National Institute of Mental Health.


Parts of this manuscript have been presented at the Eating Disorders Research Society meeting, Toronto, Ontario, September 29 – October 1, 2005 and the Academy for Eating Disorders International Conference on Eating Disorders, Baltimore, Maryland, May 2-5, 2007.

1The sample size for OS male twins is smaller than that of OS female twins because five OS male twins did not complete their assessments.

2The Minnesota Eating Behavior Survey (MEBS; previously known as the Minnesota Eating Disorder Inventory (M-EDI)) was adapted and reproduced by special permission of Psychological Assessment Resources, Inc., 16204 North Florida Avenue, Lutz, Florida 33549, from the Eating Disorder Inventory (collectively, EDI and EDI-2) by Garner, Olmstead, Polivy, Copyright 1983 by Psychological Assessment Resources, Inc. Further reproduction of the MEBS is prohibited without prior permission from Psychological Assessment Resources, Inc.

Conflicts of Interest and Financial Disclosure: None reported.

Data Access and Responsibility: All authors had full access to all the data in this study. Dr. Klump takes responsibility for the integrity of the data and the accuracy of the data analysis.


1. Association AP . Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition - Text Revision (DSM-IV-TR) American Psychiatric Association; Washington, DC: 2000.
2. Miller MN, Pumariega AJ. Culture and eating disorders: a historical and cross-cultural review. Psychiatry. 2001;64:93–110. [PubMed]
3. Rolls BJ, Fedoroff IC, Guthrie JF. Gender differences in eating behavior and body weight regulation. Health Psychol. 1991;10:133–142. [PubMed]
4. Klump KL, Gobrogge KL, Perkins P, Thorne D, Sisk CL, Breedlove SM. Preliminary evidence that gonadal hormones organize and activate disordered eating. Psychol Med. 2006;12:1–8. [PubMed]
5. Becker J, Arnold A, Berkley K, Blaustein J, Eckel L, Hampson E, et al. Strategies and methods for research on sex differences in brain and behavior. Endoncrinology. 2005;146:1650–1673. [PubMed]
6. Collaer M, Hines M. Human behavioral sex differences: a role for gonadal hormones during early development? Psychol Bull. 1995;118:55–107. [PubMed]
7. Wilson JD. Sexual differentiation. Annu Rev Physiol. 1978;40:279–306. [PubMed]
8. Breedlove SM. Sexual differentiation of the human nervous system. Annu Rev Psychol. 1994;45:389–418. [PubMed]
9. Morris J, Jordan C, Breedlove SM. Sexual differentiation of the vertebrate nervous system. Nat Neurosci. 2004;7:1034–1039. [PubMed]
10. Ryan BC, Vandenbergh JG. Intrauterine position effects. Neurosci Biobehav Rev. 2002;26:665–678. [PubMed]
11. Meulenberg P, Hofman J. Maternal testosterone and fetal sex. J Steroid Biochem Mol Biol. 1991;39:51–54. [PubMed]
12. Even M, Dhar M, vom Saal F. Transport of steroids between fetuses via amniotic fluid in relation to the intrauterine position phenomenon in rats. J Reprod Fertil. 1992;96:709–716. [PubMed]
13. Fels E, Bosch LR. Effects of prenatal administration of testosterone on ovarian function in rats. Am J Obstet Gynecol. 1971;111:964–969. [PubMed]
14. Clemens LG, Gladue BA, Coniglio LP. Prenatal endogenous androgenic influences on masculine sexual behavior and genital morphology in male and female rats. Horm Behav. 1978;10:40–53. [PubMed]
15. Madrid JA, Lopez-Bote C, Martin E. Effect of neonatal androgenization on the circadian rhythm of feeding behavior in rats. Physiol Behav. 1993;53:329–335. [PubMed]
16. Wade GN. Gonadal hormones and behavioral regulation of body weight. Physiol Behav. 1972;8:523–534. [PubMed]
17. Donohoe TP, Stevens R. Effects of ovariectomy, estrogen treatment and CI-628 on food intake and body weight in female rats treated neonatally with gonadal hormones. Physiol Behav. 1983;31:325–329. [PubMed]
18. Resnick S, Gottesman I, McGue M. Sensation seeking in opposite-sex twins: An effect of prenatal hormones? Behav Genet. 1993;23:323–329. [PubMed]
19. Cohen-Bendahan CC, Buitelaar JK, van Goozen SH, Orlebeke JF, Cohen-Kettenis PT. Is there an effect of prenatal testosterone on aggression and other behavioral traits? A study comparing same-sex and opposite-sex twin girls. Horm Behav. 2005;47:230–237. [PubMed]
20. Cohen-Bendahan CC, Buitelaar JK, van Goozen SH, Cohen-Kettenis PT. Prenatal exposure to testosterone and functional cerebral lateralization: a study in same-sex and opposite-sex twin girls. Pscyhoneuroendocrinology. 2004;29:911–916. [PubMed]
21. Cole-Harding S, Morstad A, Wilson J. Spatial ability in members of opposite-sex twin pairs. Behav Genet. 1988;18:710.
22. McFadden D. A masculinizing effect on the auditory systems of human females having male co-twins. Proc Natl Acad Sci. 1993;90:11900–11904. [PubMed]
23. Dempsey P, Townsend G, Richards L. Increased tooth crown size in females with twin brothers: evidence for hormonal diffusion between human twins in utero. Am J Human Biol. 1999;11:577–586. [PubMed]
24. Boklage C. Interactions between same-sex dizygotic fetuses and the assumption of Weinberg difference method epidemiology. Am J Human Genet. 1985;37:591–605. [PubMed]
25. Manning JT. Digit Ratio: A Pointer to Fertility, Behavior, and Health. Rutgers University Press; Piscataway, NJ: 2002.
26. Bulik CM, Sullivan PF, Fear JL, Joyce PR. Eating disorders and antecedent anxiety disorders: a controlled study. Acta Psychiatr Scand. 1997;92:101–107. [PubMed]
27. Zehr JL, Culbert KM, Sisk CL, Klump KL. An association of early puberty with disordered eating and anxiety in a population of undergraduate women and men. Horm Behav. in press. [PMC free article] [PubMed]
28. Klump KL, Burt SA. The Michigan State University Twin Registry (MSUTR): genetic, environmental and neurobiological influences on behavior across development. Twin Res Hum Genet. 2006;9:971–977. [PubMed]
29. Striegel-Moore RH, Dohm FA, Kraemer HC, Taylor CB, Daniels S, Crawford PB, Schreiber GB. Eating disorders in white and black women. Am J Psychiatry. 2003;160:1326–1331. [PubMed]
30. Hollingshead A. Four factor index of social status. Yale University; New Haven, CT: 1975.
31. Klump KL, McGue M, Iacono W. Age differences in genetic and environmental infuences on eating attitudes and behaviors in preadolescent and adolescent female twins. J Abnorm Psychol. 2000;109:239–251. [PubMed]
32. von Ranson KM, Klump KL, Iacono WG, McGue M. The Minnesota Eating Behavior Survey: a brief measure of disordered eating attitudes and behaviors. Eating Behaviors. 2005;4:373–392. [PubMed]
33. Spielberger CD, Gorsuch RL, Lushene RE, Vagg RE, Jacobs GA. Manual for the State-Trait Anxiety Inventory. Consulting Psychologists Press; Palo Alto, CA: 1983.
34. Palta M, Prineas RJ, Berman R, Hannan P. Comparison of self-reported and measured height and weight. Am J Epidemiol. 1982;115:223–230. [PubMed]
35. Stunkard AJ, Albaum JM. The accuracy of self-reported weights. Am J Clin Nutr. 1981;34:1593–1599. [PubMed]
36. Wing RR, Epstein LH, Ossip DJ, LaPorte RE. Reliability and validity of self-report and observers’ estimates of relative weight. Addict Behav. 1979;4:133–140. [PubMed]
37. Campbell L, Kashy DA. Estimating actor, partner, and interaction effects for dyadic data using PROC MIXED and HLM: A user-friendly guide. Pers Relatsh. 2002;9:327–342.
38. Kashy DA, Kenny DA. The analysis of data from dyads and groups. In: Reis CMJ HT, editor. Handbook of Research Methods in Social and Personality Psychology. Cambridge Unversity Press; New York: 2000. pp. 451–477.
39. Baron R, Kenny DA. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. J Pers Soc Psychol. 1986;51:1173–1182. [PubMed]
40. Williams ME, Pepitone SE, Christensen BM, Cook AD, Huberman NJ, Breedlove NJ, et al. Finger-length ratios and sexual orientation. Nature. 2000;404:455–456. [PubMed]
41. Lillie FR. Theory of the free martin. Science. 1916;43:611. [PubMed]
42. Lillie FR. The Free-Martin, a study of the action of sex hormones in foetal life of cattle. J Exp Zool. 1917;23:371–452.
43. Keel PK, Klump KL, Miller KB, McGue M, Iacono WG. Shared transmission of eating and anxiety disorders. Int J Eat Disord. 2005;38:99–105. [PubMed]
44. Berenbaum S, Hines M. Early androgens are related to childhood sex-typed toy preferences. Psychol Sci. 1992;3:203–206.
45. Berenbaum S, Resnick S. Early androgen effects on aggression in children and adults with congenital adrenal hyperplasia. Psychoneuroendocrinology. 1997;22:505–515. [PubMed]
46. Goldstein JM, Seidman LJ, Horton NJ, Makris N, Kennedy DN, Caviness VS, et al. Normal sexual dimorphism of the adult human brain assessed by in vivo magnetic resonance imaging. Cereb Cortex. 2001;11:490–497. [PubMed]
47. Kaye WH, Wagner A, Frank G, Bailer UF. Review of brain imaging in anorexia and bulimia nervosa. In: Wonderlich S, Mitchell JE, de Zwaan M, Steiger H, editors. Annual Review of Eating Disorders: Part 2. Radcliffe Publishing; Oxford: 2006. pp. 113–129.
48. McEwen BS. Gonadal steroid influences on brain development and sexual differentiation. In: Greep R, editor. Reproductive Physiology. University Park; Baltimore: 1983. pp. 99–145. [PubMed]
49. Pilgrim C, Hutchison JB. Developmental regulation of sex differences in the brain: can the role of gonadal steroids be redefined. Neuroscience. 1994;60:843–855. [PubMed]
50. Del Parigi A, Chen K, Gautier JF, Salbe A, Pratley R, Ravussin E, et al. Sex differences in the human brain's response to hunger and satiation. Am J Clin Nutr. 2002;75:1017–1022. [PubMed]
51. Smeets P, de Graff D, Stafleu A, van Osch M, Nievelstein R, van der Grond J. Effect of satiety on brain activation during chocolate tasting in men and women. Am J Clin Nutr. 2006;83:1297–1305. [PubMed]
52. Edler C, Lipson SF, Keel PK. Ovarian hormones and binge eating in bulimia nervosa. Psychol Med. 2007;37:131–141. [PubMed]
53. Klump KL, Culbert KM, Edler C, Keel PK. Fluctuations in binge eating and ovarian hormones over the menstrual cycle in a community sample of women. in preparation.
54. Tylka TL, Subich LM. Exploring the construct validity of the eating disorder continuum. J Couns Psychol. 1999;46:268–276.
55. Scarano GM, Kalodner-Martin CR. A description of the continuum of eating disorders: implications for intervention and research. J Couns Dev. 1994;72:356–361.
56. Mintz LB, O'Halloran S, Mulholland AM, Schneider PA. Questionnaire for eating disorder diagnoses: reliability and validity of operationalizing DSM-IV criteria into a self-report format. J Couns Psychol. 1997;44:63–79.
57. Drewnowski A, Yee DK, Kurth CL, Krahn DD. Eating pathology and DSM-IH-R bulimia nervosa: a continuum of behavior. Am J Psychiatry. 1994;151:1217–1219. [PubMed]
58. Jacobi C, Hayward C, de Zwaan M, Kraemer HC, Agras WS. Coming to terms with risk factors for eating disorders: application of risk terminology and suggestions for a general taxonomy. Psychol Bull. 2004;130:19–65. [PubMed]