Introduction
Animal and human data suggest the contribution of genetics, epigenetics, and prenatal hormonal exposure, particularly from androgens Reference Berenbaum and Beltz1–Reference Fisher, Ristori, Morelli and Maggi4 in the development of male- and female-typical behaviors and, in humans, sexual behavior. Sexual orientation is typically considered to have three constructs: sexual identity, sexual attraction, and sexual behavior Reference Epstein, McKinney, Fox and Garcia5 that exist on a continuum. 6 Sexual behavior is highly correlated with sexual identity, that is, how an individual perceives their sexuality.
Rare conditions characterized by excessive or very low androgen exposure provide a model for assessing hormonal influences on sexual orientation. Females with classical congenital adrenal hyperplasia (CAH), which involves prenatal androgen excess, appear more likely to be attracted to females, being less likely to be exclusively or almost exclusively heterosexual compared with women without the condition. Reference Hines2 These findings also suggest a trend with degree of androgen excess. Reference Hines2 Studies of complete androgen insensitivity syndrome (an intersex condition that affects the cellular response to androgens) demonstrate a tendency toward an androphilic (i.e., attraction to males) sexual orientation in individuals with female external genitalia. Reference Hines, Ahmed and Hughes7,Reference Wisniewski, Migeon and Meyer-Bahlburg8 These human models demonstrate the role of endogenous androgens in shaping sexual attraction in phenotypical females.
Animal experiments suggest the importance of androgens in sexual phenotype, with in utero exposure to high testosterone concentrations normally leading to a male-typical neural and behavioral phenotype, and lower concentrations resulting in a female-typical phenotype. Reference Hines2 In humans, the role of early endogenously derived androgen exposure (as opposed to pharmacologically influenced concentrations) during critical periods in the development of sexual orientation and gender identity—an individual’s sense of self as female, male, neither or other—as well as other influences are unclear. Reference Berenbaum and Beltz1,Reference Hines9
Diethylstilbestrol (DES) exposure during pregnancy provides an exogenous and widespread example of prenatal hormonal exposure in humans to assess its role in sexual behavior and sexual identity. From the 1940s through the early 1970s, DES, a prototypical endocrine disruptor, was used to treat several million pregnant women in the United States and Europe Reference Noller and Fish10 causing adverse reproductive and other health conditions in the offspring. Reference Hoover, Hyer and Pfeiffer11 While a potent synthetic estrogen, rodent studies demonstrate increased progesterone and testosterone as well as estrogen production in the prenatally DES-exposed. Reference Haney, Newbold and McLachlan12 We have reported that females prenatally exposed to DES (i.e., the second generation) were significantly less likely to report being gay/lesbian, particularly those with exposure to a low cumulative dose of DES, compared to unexposed women. Reference Troisi, Palmer and Hatch13
Findings from animal studies demonstrate that the effects of prenatal DES exposure can be transmitted to subsequent generations through direct or epigenetic mechanisms. Reference Gray Nelson, Sakai, Eitzman, Steed and McLachlan14–Reference Baccarelli and Bollati17 While there are examples of intergenerational inheritance of epigenetic effects in humans, there has been less evidence for transgenerational inheritance which appear in the fourth generation when passed through the maternal line during pregnancy. Reference Fitz-James and Cavalli18,Reference Ghai and Kader19 Assuming prenatal effects of DES on the maternal ova persist until conception thereby escaping global demethylation, they could be incorporated in the embryo at fertilization. In this case, offspring ova were directly exposed to DES while their mother was in utero.
We assessed whether the observed associations of DES with sexual behavior and identity in women exposed to DES in utero exist in the third generation by collecting similar information, as well as querying participants regarding their gender identity and whether it was concordant with the sex to which they were assigned at birth.
Methods
National Cancer Institute (NCI) combined DES cohorts follow-up study
In the early 1990s, the NCI established the Combined DES Cohorts Follow-up Study to assess health outcomes in women exposed to DES during pregnancy, and in their prenatally exposed sons and daughters. Included were new and previously followed cohorts at Dartmouth College, Boston University, New England Medical Center, the University of Chicago, and Texas Methodist Health Center. The NCI’s DES study has been described in detail elsewhere. Reference Hatch, Palmer and Titus-Ernstoff20 Follow-up of the prenatally exposed (second generation) of the combined cohorts began in 1994 with a mailed questionnaire, and subsequent questionnaires were mailed approximately every five years in 1997, 2001, 2006, 2011, and 2016.
Third-generation cohort baseline
In 2001, we assembled for the first time a third-generation cohort comprised of the adult offspring who were assigned female at birth of prenatally DES-exposed and unexposed second-generation women participating in the NCI’s DES Follow-up Study. Reference Titus-Ernstoff, Troisi and Hatch21 Through a review of the second-generation women’s parity records, we identified 1781 (967 exposed and 814 unexposed) age-eligible (≥ 18 years) third-generation females. Exposed mothers gave permission to contact 516 (53%), and unexposed mothers 382 (47%) of their daughters (Figure 1). Of these, similar proportions of exposed 464 (90%), and unexposed 329 (86%) third-generation females completed a baseline questionnaire.
Figure 1. DES Third Generation Study eligibility and enrollment.
In 2009, we identified 1195 (857 exposed, 338 unexposed) newly eligible (became 18 years old since the last questionnaire) third-generation females; mothers gave permission to contact 416 (49%) exposed and 161 (48%) unexposed. Of these, 334 (80%) exposed and 138 (86%) unexposed third-generation females completed the baseline questionnaire. Third-generation participants from the Texas cohort were included through 2009 only.
Third-generation cohort – follow-up
In 2009, we sent follow-up questionnaires to 789 (462 exposed and 327 unexposed) participants who were previously enrolled in the 2001 baseline study (4 participants could not be re-approached due to Institutional Review Board regulations at one of the study centers). Of these, 381 (83%) exposed and 280 (73%) unexposed individuals completed questionnaires by mail, telephone, or web.
In 2019, a second follow-up questionnaire was sent to participants who responded to either the 2001 or 2009 baseline questionnaire and did not refuse or were not deceased in 2009. Of 1071 who were eligible (638 exposed and 433 unexposed), a total of 763 (71%) including 458 (72%) exposed and 305 (67%) unexposed participated in the 2019 follow-up.
In summary, mothers gave permission to contact about 50% of eligible female offspring. Of those offspring who were sent a questionnaire, 80–90% completed the baseline questionnaire, and 67–83% completed a follow-up questionnaire.
Questionnaire information
The 2009 follow-up questionnaire (all questionnaires are in eAppendix) included a question on sexual behavior: “What best describes your sexual partners during your adult life?” Response options were “Only men; mostly men; mostly women; only women; no sexual contact; or prefer not to answer.” The 2019 follow-up questionnaire repeated the sexual behavior question, and additionally queried sexual identity (“Which of the following best represents how you think of yourself? Gay or lesbian; Bisexual; Straight/heterosexual (not gay, lesbian, or bisexual); Asexual (no interest in sex); Other, please specify; and prefer not to answer),” and gender identity (“Which of the following best represents how you currently think of yourself? Woman; man; other than a man or a woman, please specify; prefer not to answer”).
When participants left the question blank, they were categorized with those who indicated they preferred not to respond in a nonresponse category A variable for non-heterosexual behavior was defined by combining responses indicating mostly men; mostly women; and only women as sexual partners. A variable for non-heterosexual identity was defined by combining responses for gay/lesbian; bisexual; other; and asexual, for the sexual identity variable. Attained education (highest grade completed) was collected on all questionnaires. Exposure to DES occurred early in fetal development, so most covariates could be considered possible mediators of any influence of exposure on the outcomes studied.
Eligibility
Eligibility for the current analysis required completion of either the 2009 or 2019 follow-up questionnaire which queried participants on sexual partners; sexual identity and gender identity were asked only the 2019 questionnaire. Of the participants who responded to either of the follow-up questionnaires, 982 (n= 592 and n = 390 of the exposed and unexposed, respectively) answered any of the relevant questions, with 661 answering the 2009 questions, 763 answering the 2019 questions, and 442 answering questions on both the 2009 and 2019 questionnaires.
Maternal DES exposure
The participants’ mothers’ exposure to DES during pregnancy or the absence of exposure was determined from their medical record. For all combined cohort participants, the presence or absence of vaginal epithelial changes (VECs) in the mother was available for one of the original subcohorts from Boston University, and from the University of Chicago. VEC is correlated with earlier gestational administration and higher DES dose in utero, Reference Herbst, Poskanzer, Robby, Friedlander and Scully22 and was used as a proxy for these, and possible biological susceptibility to DES, in the analyses.
Statistical analysis
We used logistic regression to compute odds ratios (OR) and 95% confidence intervals (CIs) for the association between DES exposure and sexual behavior (any non-heterosexual behavior vs. only heterosexual; no sexual contact; prefer not to answer/missing), and the association between DES exposure and sexual identity (any non-heterosexual identity vs. heterosexual; prefer not to answer/missing) using multinomial logistic regression models. In primary models, we adjusted for birth year. In subsequent models, we additionally controlled for education. Associations were evaluated by presence or absence of VEC in exposed mothers (vs. unexposed as the reference group). Estimates were reported for the individual categories of non-heterosexuality only when the number was >5. Responses to the question regarding gender identity were presented by count only, due to the small number of participants reporting a gender identity different from their assigned sex at birth.
Human subjects protection
The current study was approved by the institutional review boards at each participating center (Dartmouth College, Boston University, New England Medical Center, the University of Chicago, and Texas Methodist Health Center), the data coordinating center (Westat, Inc.), and the NCI. Informed consent was based on participant’s return of a completed questionnaire.
Results
Compared with unexposed participants, a greater proportion of third-generation DES-exposed women were born in later calendar years and were therefore slightly younger at the end of follow-up (mean age 37.5 and 39.2 years in the exposed and unexposed, respectively). Attained education was high overall, and greater for the DES-exposed than for the unexposed participants (Table 1). Self-reported race/ethnicity was predominantly white (98 and 97% of exposed and unexposed, respectively).
Table 1. Distribution (%) of demographic factors and original cohort in participants whose mothers were prenatally DES-exposed and -unexposed
a At the most recent questionnaire.
Sexual behavior
Five percent or less of the exposed and unexposed participants selected “prefer not to respond” or left the response blank for sexual orientation (Table 2). Most of the exposed (85%) and unexposed (83%) reported having male sexual partners only, with an additional 6.4% and 9.2%, respectively, reporting having mostly male sexual partners.
Table 2. Odds ratio (OR) and 95% confidence intervals (CIs) for the association of maternal prenatal DES exposure and sexual behavior and sexual identity
a All models adjust for birth year. ORs are not reported for categories with n<5. Percentages may not add to 100 because of rounding.
b ORs for DES are from a multinomial logistic regression model with non-heterosexual behavior (defined as mostly men, mostly women, only women), no sexual contact, and prefer not to respond/missing in separate categories versus heterosexual (only men) as the outcome variable.
c The ORs for specific categories of non-heterosexual behavior are from a multinomial model with mostly men, mostly women, only women, no sexual contact, and prefer not to respond/missing versus heterosexual as the reference group.
d Sexual and gender identity were only ascertained on the 2019 questionnaire.
e ORs for DES are from a multinomial logistic regression model with non-heterosexual identity (defined as gay/lesbian, bisexual, asexual, and other), and prefer not to respond/missing versus straight/heterosexual as the reference group.
f The ORs for specific category of non-heterosexual identity are from a multinomial model with gay/lesbian, bisexual, asexual, other, and prefer not to respond/missing versus straight/heterosexual as the reference group.
In models that included DES exposure, birth year, and education, participants born in later calendar years were more likely to report non-heterosexual behavior (OR per year 1.1; 95% CI 1.0–1.1; data not in Table). Participants with less than a college education were more likely to report non-heterosexual behavior compared with college graduates (OR 2.4; 95% CI 1.4–3.9; data not in table), while the OR for more than a college education compared with college graduate was 0.95 (95% CI 0.58–1.6; data not in table).
The OR for the association of maternal prenatal DES exposure and sexual behavior (non-heterosexual vs. heterosexual) adjusted for birth year was 0.71 (95% CI 0.46–1.1) (Table 2) and was similar with adjustment for education (0.76; 95% CI 0.49–1.2). This association appeared to be explained by the inclusion of “mostly men” in the non-heterosexual group (vs. only men in the heterosexual group). The OR for DES and having mostly men for sexual partners compared with only men was 0.63 (95% CI 0.39–1.0; Table 2). In other words, individuals whose mothers were prenatally exposed to DES were less likely to report non-heterosexual behavior as represented by having only male partners compared with having mostly male partners. The OR for no sexual contact was 1.2 (95% CI 0.44–3.3) and for those preferring not to respond or missing responses was 1.0 (0.52–1.9).
Concordance of sexual identity and sexual behavior
We assessed the concordance of sexual identity with sexual behavior among participants with information on both (n= 715). Among those who reported heterosexual identity (n= 665), 630 (99%) also reported having opposite sex partners only, while 35 (1%) reported having some same-sex partners. Among those who reported non-heterosexual identity (n= 50), 39 (78%) reported having some same-sex partners, while 11 (22%) reported having only opposite sex partners. Among those who reported heterosexual behavior, (n= 641), 630 (98%) reported heterosexual identity, while ∼ 2.0% reported bisexual (n= 4), other (n= 4), asexual (n= 3). Among those who reported non-heterosexual behavior (mostly men, mostly women, only women partners, n = 74), 35 (47%) reported their identity as heterosexual, 20 (27%) bisexual, 13 (17%) gay/lesbian identity, 4 (5.3%) other, and 2 (2.7%) asexual.
Sexual identity
Sexual identity was mostly comprised of those reporting a straight/heterosexual identity (88% exposed and 89% unexposed; Table 2). Close to 1% of the exposed and the unexposed reported “other” for sexual identity; this response prompted a request to specify, and a review of the write-ins indicated that all were consistent with non-heterosexual identity and were classified as such. Like the association of age and sexual behavior, birth year (in models that included education and DES) was positively associated with non-heterosexual identity (OR per year 1.0; 95% CI 1.0–1.1). The OR for the association of non-heterosexual identity with education was OR 1.2 (95% CI 0.53–2.8 for less than a college education compared with college graduates, and the OR 0.83 (95% CI 0.45–1.6) for more than a college education compared with college graduate.
The OR for the association of DES and non-heterosexual identity (vs. straight/heterosexual identity) adjusted for birth year was 0.99 (95% CI 0.55–1.8), compared with the unexposed (Table 2). Results were similar with further adjustment for education (OR 1.0 (95% CI 0.57–1.8). The OR for DES and identifying as bisexual compared with heterosexual/straight was 0.61 (0.27–1.4).
Gender identity
Only three participants, two DES-exposed and the other unexposed, reported gender identity that was different from their biological sex (Table 2).
We assessed whether exposed mothers’ VEC status (as a proxy for DES higher dose and earlier gestational timing of exposure) was related to sexual behavior (n= 622) and identity among (n= 475) participants for whom VEC status was available. Roughly similar proportions of DES-exposed individuals had a positive or a negative VEC status. ORs for non-heterosexual behavior were 0.90 (95% CI 0.49–1.7) in DES-exposed participants whose mothers had a history of VEC and 0.48 (95% CI 0.24–0.97) in participants whose mother did not have a VEC history compared to participants unexposed to DES. ORs for non-heterosexual identity were 1.1 (95% CI 0.47–2.4) in DES-exposed participants whose mothers had a history of VEC and 0.87 (95% CI 0.35–2.1) in participants whose mothers did not have VEC compared to the unexposed (Table 3).
Table 3. Odds ratio (OR) and 95% confidence intervals (CIs) for maternal prenatal DES exposure status and presence or absence of vaginal epithelial changes (VECs) with sexual behavior and sexual identity
a Sexual behavior was missing for 12 DES-exposed with VEC, 16 DES-exposed without VEC, and 6 DES-unexposed; sexual identity was only ascertained on the 2019 questionnaire and was missing for 7, 10 and 2 participants, respectively.
b OR are from binary logistic regression models with non-heterosexual (defined as mostly men, mostly women, only women) versus heterosexual (only men) behavior as the reference group, and non-heterosexual (defined as gay/lesbian, bisexual, asexual, and other) versus heterosexual (straight/heterosexual) identity as the outcome variable, DES and VEC status as the exposure variable, and birth year as a covariate.
c Analysis of VEC includes participants from only one of the original subcohort from Boston University, and the University of Chicago.
Discussion
The medical travesty of DES exposure during pregnancy impacted health in two generations of women and may—through direct exposure to the ova and/or epigenetic mechanisms—impact the health of a third generation of individuals as well. Though banned in the United States since 1971, continued study of DES exposure offers a unique model to assess the influence of a potent synthetic estrogen on human development, including important aspects of sexuality and identity, across generations. These data were used to gain insight into the role of natural variation in in utero hormone exposure on sexual behavior and identity. In our analysis, using information from the largest cohort of individuals with documented DES exposure status, we observed no consistent evidence that sexual behavior and sexual identity in third-generation daughters was influenced by exposure of their mothers to DES in utero.
Results of previous investigations in participants who were exposed to DES in utero (second generation), Reference Troisi, Palmer and Hatch13,Reference Titus-Ernstoff, Perez and Hatch23 demonstrated that DES-exposed women were 10% more likely to have ever married, 30% less likely to report sexual partners of the same sex, Reference Titus-Ernstoff, Perez and Hatch23 and about 40% less likely to report being non-heterosexual Reference Troisi, Palmer and Hatch13 than unexposed women. Our current findings are not consistent with an influence of DES on sexual behavior or identity in the third generation. While the OR was below one for the association of DES and non-heterosexual behavior, this was mainly explained by including having “mostly men” but not “only men” for sexual partners. behavior was in the form of having some same-sex partners but mostly opposite sex partners. ORs for non-heterosexual behavior were lower in DES-exposed participants whose mothers did not have a history of VEC but not in those whose mothers did have VEC, compared to the unexposed. The presence of VEC is associated with greater DES dose and earlier gestational exposure, therefore it is possible that DES effects on sexual behavior may only occur at lower doses and with later gestational exposure. Reference Herbst, Poskanzer, Robby, Friedlander and Scully22 There is evidence of effects of lower doses in the second-generation females prenatally exposed to DES among whom the association of being less likely to report being gay/lesbian was stronger among those with exposure to a low cumulative dose of DES, compared to unexposed women. Reference Troisi, Palmer and Hatch13 The lower odds of non-heterosexual behavior in the third generation among females whose mothers did not have VEC may also be due to chance. To the best of our knowledge, there are no other published studies on the effects of DES exposure on sexual behavior or identity among offspring of mothers exposed prenatally.
The statistical power of our analysis was limited by the small numbers of individuals who reported sexual behavior or sexual identity other than heterosexual, with even fewer reporting transgender identity. Our results may have been affected by underreporting of non-heterosexuality stemming from social stigmatization, or if mothers excluded transgender males from their reported daughters during recruitment. However, our study found similar results for the proportion of participants reporting non-heterosexuality compared with a large Gallop poll (Jones, 2023). Reference Jones J.M.24 In that poll, 10.5% of Americans born in 1981-1996 reported being non-heterosexual compared with 8.6 and 11% for sexual behavior, and in our study, 7.0% and 6.6% for sexual identity, respectively, for DES-exposed and -unexposed. Moreover, the OR for the participants preferring not to answer or whose responses were missing was not elevated for non-heterosexual behavior and identity. We would not expect misclassification of sexual behavior or sexual identity to be differential with respect to DES exposure, as hypotheses about these possible associations are not widely known in the general population, and thus any bias introduced would be expected to be toward the null. Participants with less than a college education were over two times more likely to report non-heterosexual behavior compared participants with greater attained education. While mother’s DES exposure status was associated with greater education, adjusting for education did not change the results. It is possible that second-generation mothers could have been exposed to other estrogenic endocrine disruptors ubiquitous in the environment, but unless this exposure varied systematically by mothers’ prenatal DES exposure status, it would not bias the estimates that we present for DES. Also, DES is a particularly potent estrogen.
Despite these limitations, these are unique data on a generation whose mothers were exposed prenatally to high doses of synthetic estrogen during a particularly important development period. In summary, while previous findings suggested an association between prenatal DES exposure and sexual identity, Reference Troisi, Palmer and Hatch13 our current findings in the offspring of those exposed in utero provide little evidence of an effect of maternal prenatal DES exposure on the next generation’s sexual behavior or sexual identity. We are limited in our conclusions regarding the influence of maternal prenatal DES exposure on the next generation’s gender identity owing to very few individuals reporting gender identity different from their biological sex in either the exposed or unexposed group. The small number of transgender individuals did not allow evaluation of whether DES could be involved in the formation of gender identity. These data do not support intergenerational effects of DES in the third generation.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2040174424000175.
Acknowledgments
We thank the participants in the Continuation of Follow-up of DES-exposed cohorts for their long-standing commitment to the study.
Financial support
This research has been supported with Federal funds from the intramural research program of the Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health.
Competing interests
None.
Open access
