Key Points
Question
Is exposure to diethylstilbestrol during pregnancy associated with adverse multigenerational neurodevelopmental outcomes?
Findings
A cohort study of 47 450 women in the Nurses’ Health Study II found significantly elevated odds for attention-deficit/hyperactivity disorder in the grandchildren (third generation) of users of diethylstilbestrol, a potent endocrine disruptor.
Meaning
Exposure to endocrine disruptors during pregnancy may be associated with multigenerational neurodevelopmental deficits.
Abstract
Importance
Animal evidence suggests that endocrine disruptors affect germline cells and neurodevelopment. However, to date, the third-generation neurodevelopmental outcomes in humans have not been examined.
Objective
To explore the potential consequences of exposure to diethylstilbestrol or DES across generations—specifically, third-generation neurodevelopment.
Design, Setting, and Participants
This cohort study uses self-reported health information, such as exposure to diethylstilbestrol during pregnancy and attention-deficit/hyperactivity disorder (ADHD) diagnosis, from 47 540 participants enrolled in the ongoing Nurses’ Health Study II. The 3 generations analyzed in this study were the participants (F1 generation), their mothers (F0 generation), and their live-born children (F2 generation).
Main Outcomes and Measures
Participant- and mother-reported exposure to diethylstilbestrol during pregnancy and physician-diagnosed child ADHD.
Results
The total number of women included in this study was 47 540. Of the 47 540 F0 mothers, 861 (1.8%) used diethylstilbestrol and 46 679 (98.2%) did not while pregnant with the F1 participants. Use of diethylstylbestrol by F0 mothers was associated with an increased risk of ADHD among the F2 generation: 7.7% vs 5.2%, adjusted odds ratio (OR), 1.36 (95% CI, 1.10-1.67) and an OR of 1.63 (95% CI, 1.18-2.25) if diethylstilbestrol was taken during the first trimester of pregnancy. No effect modification was observed by the F2 children’s sex.
Conclusions and Relevance
This study provides evidence that diethylstilbestrol exposure is associated with multigenerational neurodevelopmental deficits. The doses and potency level of environmental endocrine disruptors to which humans are exposed are lower than those of diethylstilbestrol, but the prevalence of such exposure and the possibility of cumulative action are potentially high and thus warrant consideration.
This cohort study uses data from the Nurses Health Study II to assess the risk of attention-deficit/hyperactivity disorder among grandchildren of women exposed to diethylstilbestrol during pregnancy.
Introduction
The prevalence of neurodevelopmental disorders, like attention-deficit/hyperactivity disorder (ADHD), has been increasing.1,2 Children with ADHD have compromised quality of life3 and substantial educational problems.4 Most studies, but not all,5 suggest the negative symptoms of childhood ADHD persist into adulthood6,7,8 and considerably worsen the quality of life for those adults,9,10 increasing their risk for obesity,11,12 many other morbidity and mortality health indicators,13,14,15 and even premature death.16 Therefore, identifying the modifiable factors that contribute to these generational and multigenerational risks is of immense importance.
Endocrine-disrupting chemicals (EDCs) comprise potentially modifiable factors that have been linked to neurodevelopmental disorders,17,18,19,20,21 including ADHD.17 Interest in the potential multigenerational and transgenerational consequences of EDCs has increased given the hypothesized biological mechanism involving epigenetic reprogramming of the germline.22,23,24,25,26 For example, exposure to di(2-ethylhexyl) phthalate, a known EDC, has been shown to alter third-generation behavior and stress responses, observed corticosterone levels, and pituitary gene expression and behavior in mice.27 Wolstenholme et al28 also reported changes in third- to fifth-generation social interactions following exposure to bisphenol A in mice. Despite compelling evidence from toxicological studies, epidemiological evidence in humans of EDC-associated multigenerational neurodevelopmental outcomes is currently unavailable, likely because of the lack of large cohorts with relevant information across multiple generations.
In this study, we used data from the Nurses’ Health Study II (NHSII) to examine the potential third-generation outcomes (ADHD) of diethylstilbestrol, a potent perinatal EDC. Between 1938 and 1971, diethylstilbestrol was given to pregnant women to prevent pregnancy complications. The exact number of women worldwide who used diethylstilbestrol is unknown, but between 5 and 10 million women in the United States were estimated to have used it.29 Prescription of the drug started to phase out after a 1953 study found diethylstilbestrol had no actual treatment value,30 and it was banned in 1971 after another study linked it to vaginal adenocarcinomas in female children of the women who used diethylstilbestrol.31 Later, it was found to be associated with multiple other reproductive adverse outcomes in female offspring.29 Only a few studies have considered EDCs and their outcomes for the third generation. For example, exposure to diethylstilbestrol during pregnancy was linked to hypospadias in the grandsons32,33,34 and to delayed menstrual regularization in the granddaughters35 of the exposed women. To our knowledge, no epidemiological study to date has assessed the multigenerational neurodevelopmental consequences of diethylstilbestrol or any other EDCs.
Methods
Data Collection
Study Population
In 1989, a total of 116 686 female registered nurses between the ages of 25 and 42 years enrolled in the NHSII prospective cohort by completing a mailed questionnaire. At baseline, participants resided in 14 US states (California, Connecticut, Indiana, Iowa, Kentucky, Massachusetts, Michigan, Mississippi, New York, North Carolina, Ohio, Pennsylvania, South Carolina, and Texas), but they have since moved to all 50 states. The NHSII has a retention rate of more than 90%. Follow-up questionnaires are mailed to participants every 2 years to gather updated information on lifestyle and risk factors, medication use, and major illness occurrence. Our study was approved by the institutional review board of Brigham and Women’s Hospital, Boston, Massachusetts, and the human subjects committee of Harvard T. H. Chan School of Public Health, Boston. Participants provided implied informed consent by completing and returning the questionnaire.
All NHSII participants (F1 generation) were born between 1946 and 1964, when diethylstilbestrol was still being prescribed. In 1993, the mailed questionnaire included questions about the use of diethylstilbestrol by the participant’s mother (F0 generation) during her pregnancy with the F1 participant. We excluded from our analyses F1 participants who did not return the 1993, 2005, or 2013 questionnaires (the relevant surveys that collected outcome and exposure data) and who did not report any live-born children. Because children (F2 generation) with ADHD were identified only by their year of birth, we also excluded F1 participants with multiple pregnancies (eg, twins, triplets) and same-year births (from different pregnancies), as matching the response to the appropriate child born in the same year would not be possible. The final data set for the main analysis included 47 540 F0 mothers as well as 106 198 live-born F2 children.
Exposure Assessment
In 1993, a supplementary questionnaire was mailed to 2742 F1 participants who reported that their F0 mothers had been exposed to diethylstilbestrol while pregnant with the F1 participants. Among the 2317 participants (84.5%) who responded, 2032 (87.7%) reported being certain or somewhat certain of their mothers’ exposure, 123 (5.3%) were not certain, and 162 (7.0%) reported no such exposure. Thus, we considered only those who were certain or somewhat certain of their mothers’ exposure, excluding those who were not certain or had originally said yes but later reported no exposure. The 1993 questionnaire also included questions on trimester-specific diethylstilbestrol use—that is, any exposure during the first, second, or third trimester of pregnancy.
In 2001, a total of 29 070 F0 mothers received a questionnaire pertaining to their pregnancy with the F1 participants; it included a question about diethylstilbestrol use. We observed very good agreement between the mothers’ responses to the 2001 survey and prior reports by the F1 in 1993 (κ = 0.74)36,37 that did not vary by ADHD status.
Outcome Assessment
The outcome of interest was an F1 participant’s report of whether any of her children ever received a doctor’s diagnosis of ADHD. This question was first asked in the 2005 questionnaire, although at that time the date of birth of the child or children with ADHD was not asked. This question was repeated in the 2013 questionnaire, further requesting information on the birth years of the children with an ADHD physician diagnosis.
Maternal reports of ADHD have been found to be reliable.38 Moreover, in a validation study involving 92 F1 participants who reported in the 2005 questionnaire of having a child diagnosed with ADHD, we found that the F2 children scored high on the ADHD Rating Scale-IV.39 All girls scored above 90%, 81.1% of boys scored above 80%, and 63.8% of boys scored above 90%.40
For our main analysis, we included information on F1 participants who provided concordant responses (92.6% concordance) in the 2005 and 2013 questionnaires to the question of whether any of their children had been diagnosed with ADHD by a physician, using the child-specific responses from the 2013 questionnaire to minimize any potential outcome misclassification. We also conducted analyses of all participants who responded to the 2013 questionnaire to include their children, who may have received a diagnosis after 2005.
Covariates
Potential confounders of the association between F0 mothers’ diethylstilbestrol exposure and F2 children’s ADHD could only be variables that preceded diethylstilbestrol exposure41 (ie, F0 mother–level variables). Therefore, we did not consider participant or child variables as covariates in the health models except for self-identified participant race/ethnicity, which was assumed to reflect the mother’s race/ethnicity. In the 1999 questionnaire, participants were asked whether their mothers smoked during pregnancy. In the 2005 questionnaire, participants were asked about their family socioeconomic status at their birth as well as their mother’s lifestyle, educational level, and occupation. We adjusted for the participant’s race (white, African American, Asian, or other), ethnicity (Hispanic—yes or no), and year of birth (and squared year of birth to account for nonlinear confounding); mother’s smoking during pregnancy (yes or no); mother’s home ownership at the time of pregnancy (yes or no); mother’s education (<9 years, 1-3 years of high school, 4 years of high school, 1-3 years of college, or ≥4 years of college); father’s educational level (same categories); mother’s occupation (professional, executive/manager, sales/clerical work, mechanic/skilled, machine operator/driver, service worker, laborer/unskilled work, farming, military, or housekeeping); and father’s occupation (same categories).
Because of previous reports of an association between F0 mother’s diethylstilbestrol exposure and F1 participant’s depression37 as well as between depression and ADHD,42 in a sensitivity analysis we adjusted for the mother’s depression history, which was reported by the participants in the 2005 survey. However, this sensitivity analysis was only a secondary analysis, as information was not available on the timing of the diagnosis of the mother’s depression relative to her index pregnancy. In a secondary analysis, we adjusted for the mother’s age when the participant was born because age at delivery was unknown for 1928 F0 mothers (4.1%) in our study. We included a missing indicator for other covariates with missing responses, which arose mostly when the participant did not know the response for her mother.
Statistical Analysis
Data Analysis
In utero exposure to diethylstilbestrol is known to affect women’s reproductive ability, which could affect the number of grandchildren within the F0 generation and the likelihood that any F2 child has ADHD—that is, the distribution of ADHD given diethylstilbestrol potentially may depend on the number of children within the F1 generation—for which our data provide evidence (results not shown). This structure is defined as informative clustering.43,44,45 Standard generalized estimating equation approaches are not appropriate for this analysis and would result in invalid estimates and inferences.43,44,45 We used cluster-weighted generalized estimating equations with a logit link to account for multiple children within the F1 generation, adjusting for potential confounders with the inverse of the cluster size (ie, number of children per participant) as the weights to account for informative clustering.45 We also assessed potential effect modification by F2 children’s sex by including in the model an interaction term between F0 mothers’ diethylstilbestrol use and F2 children’s sex.
We considered any diethylstilbestrol exposure during pregnancy and trimester-specific exposures. These exposures were not mutually exclusive and were modeled as 3 binary variables along with another variable for those who did not know about trimester-specific timing; all were included simultaneously in 1 model. Results were similar when we repeated this analysis using separate models for each trimester and excluding the 302 F0 mothers (with 625 F2 children) who did not know the trimester of exposure.
Our results are reported as odds ratios (ORs) and 95% CIs. Statistical significance was assessed at α = .05; all P values presented are for 2-tailed tests. All analyses were conducted using SAS statistical software, version 9.4 (SAS Institute Inc).
Sensitivity Analyses
As stated, we repeated the analyses using the ADHD-related responses to only the 2013 questionnaire. In addition, we conducted sensitivity analyses to assess the outcome of potential unmeasured confounding by calculating the E value—that is, the minimal strength at which a confounder would need to be associated with both the outcome and the exposure of interest to fully explain the observed association.46
Results
In the main analysis, we included information on 47 540 F0 mothers, 861 (1.8%) of whom were exposed to diethylstilbestrol while pregnant with the F1 participants and 46 679 (98.2%) who were not. Table 1 presents descriptive characteristics of the mothers and participants by diethylstilbestrol use. The F1 participants had a median of 2 children (range, 1-12) if their F0 mothers did not use diethylstilbestrol while pregnant and a median of 2 children (range, 1-5) if their mothers used diethylstilbestrol. Table 2 presents descriptive characteristics of the F1 participants and their F0 mothers by the ADHD status of the F2 children. The total number of children was 106 198, and 5587 (5.3%) of them were diagnosed with ADHD. The median F2 year of birth was 1983 (interquartile range, 1978-1988). The grandmothers of 137 F2 children with ADHD (2.5%) were exposed to diethylstilbestrol while pregnant with the F1 participants.
Table 1. Characteristics of 47 540 F1 Participants and Their F0 Mothers, by F0 Diethylstilbestrol Use During Pregnancy.
| Characteristic | Diethylstilbestrol Use During F0 Pregnancy, No. (%) | |
|---|---|---|
| Yes (n = 861) | No (n = 46 679) | |
| F1 race | ||
| White | 822 (95.5) | 44 338 (95.0) |
| African American | 6 (0.7) | 509 (1.1) |
| Asian | 10 (1.2) | 633 (1.4) |
| Other | 23 (2.7) | 1199 (2.6) |
| F1 ethnicity | ||
| Hispanic | 6 (0.7) | 582 (1.2) |
| F0 educational level, y | ||
| <9 | 28 (3.3) | 3228 (6.9) |
| 1-3 High school | 90 (10.5) | 5332 (11.4) |
| 4 High school | 394 (45.8) | 22 921 (49.1) |
| 1-3 College | 244 (28.3) | 10 263 (22.0) |
| ≥4 College | 95 (11.0) | 4199 (9.0) |
| F1 Did not know | 10 (1.2) | 736 (1.6) |
| Missing | 5 (0.6) | 244 (0.5) |
| F1 father’s educational level, y | ||
| <9 | 62 (7.2) | 5001 (10.7) |
| 1-3 High school | 90 (10.5) | 6008 (12.9) |
| 4 High school | 283 (32.9) | 17 511 (37.5) |
| 1-3 College | 185 (21.5) | 7568 (16.2) |
| ≥4 College | 227 (26.4) | 9198 (19.7) |
| F1 did not know | 14 (1.6) | 1393 (3.0) |
| Missing | 3 (0.3) | 261 (0.6) |
| F0 occupation | ||
| Professional | 123 (14.3) | 5094 (10.9) |
| Executive, manager | 12 (1.4) | 284 (0.6) |
| Sales/clerical work | 73 (8.5) | 4502 (9.6) |
| Mechanic/skilled | 4 (0.5) | 355 (0.8) |
| Machine operator/inspector/driver | 15 (1.7) | 594 (1.3) |
| Service worker | 13 (1.5) | 893 (1.9) |
| Laborer/unskilled worker | 18 (2.1) | 1542 (3.3) |
| Farming | 7 (0.8) | 728 (1.6) |
| Military | 0 (0) | 12 (0) |
| No work outside the home | 594 (69.0) | 32 060 (68.7) |
| F1 did not know | 2 (0.2) | 615 (1.3) |
| F1 father’s occupationa | ||
| Professional | 184 (21.4) | 8186 (17.5) |
| Executive, manager | 138 (16.0) | 5011 (10.7) |
| Sales/clerical work | 87 (10.1) | 4706 (10.1) |
| Mechanic/skilled | 188 (21.8) | 10 891 (23.3) |
| Machine operator/inspector/driver | 65 (7.5) | 4273 (9.2) |
| Service worker | 35 (4.1) | 1829 (3.9) |
| Laborer/unskilled worker | 74 (8.6) | 5573 (11.9) |
| Farming | 52 (6.0) | 3983 (8.5) |
| Military | 50 (5.8) | 2310 (4.9) |
| No work outside the home | 1 (0.1) | 186 (0.4) |
| F1 did not know | 5 (0.6) | 831 (1.8) |
| F0 smoking during pregnancy | ||
| Yes | 282 (32.9) | 10 857 (23.3) |
| No | 478 (55.5) | 29 440 (63.1) |
| F1 did not know | 76 (8.8) | 4205 (9.0) |
| Missing | 25 (2.9) | 2177 (4.7) |
| F0 Lifetime history of depression | 142 (16.5) | 4985 (10.7) |
| F1 Total No. of children, mean (SD) | 2.07 (0.8) | 2.24 (0.9) |
Abbreviations: F0, mothers of participants in the Nurses’ Health Study II; F1, participants in the Nurses’ Health Study II.
F1 father’s occupation categories are not mutually exclusive.
Table 2. Characteristics of the 47 540 F1 Participants and Their F0 Mothers, by 106 198 F2 Children’s ADHD Status .
| Characteristic | F2 ADHD Status, No. (%) | |
|---|---|---|
| Yes (n = 5587) | No (n = 100 611) | |
| F1 race | ||
| White | 5376 (96.2) | 95 777 (95.2) |
| African American | 25 (0.4) | 1022 (1.0) |
| Asian | 26 (0.5) | 1303 (1.3) |
| Other | 160 (2.9) | 2509 (2.5) |
| F1 ethnicity | ||
| Hispanic | 48 (0.9) | 1221 (1.2) |
| F0 educational level, y | ||
| <9 | 304 (5.4) | 6899 (6.9) |
| 1-3 High school | 655 (11.7) | 11 133 (11.1) |
| 4 High school | 2601 (46.6) | 49 718 (49.4) |
| 1-3 College | 1365 (24.4) | 22 236 (22.1) |
| ≥4 College | 586 (10.5) | 9103 (9.0) |
| F1 did not know | 76 (1.4) | 1522 (1.5) |
| Missing | 31 (0.6) | 508 (0.5) |
| F1 father’s educational level, y | ||
| <9 | 498 (8.9) | 10 645 (10.6) |
| 1-3 High school | 649 (11.6) | 12 788 (12.7) |
| 4 High school | 1969 (35.2) | 37 810 (37.6) |
| 1-3 College | 1034 (18.5) | 16 415 (16.3) |
| ≥4 College | 1281 (22.9) | 20 084 (20.0) |
| F1 did not know | 156 (2.8) | 2869 (2.9) |
| Missing | 26 (0.5) | 565 (0.6) |
| F0 occupation | ||
| Professional | 655 (11.7) | 10 995 (10.9) |
| Executive, manager | 33 (0.6) | 603 (0.6) |
| Sales/clerical work | 577 (10.3) | 9346 (9.3) |
| Mechanic/skilled | 34 (0.6) | 748 (0.7) |
| Machine operator/inspector/driver | 59 (1.1) | 1218 (1.2) |
| Service worker | 100 (1.8) | 1826 (1.8) |
| Laborer/unskilled worker | 141 (2.5) | 3208 (3.2) |
| Farming | 77 (1.4) | 1608 (1.6) |
| Military | 2 (0) | 31 (0) |
| No work outside the home | 3832 (68.6) | 69 862 (69.4) |
| F1 did not know | 82 (1.5) | 1263 (1.3) |
| F1 father’s occupationa | ||
| Professional | 1127 (20.2) | 17 652 (17.5) |
| Executive, manager | 672 (21.0) | 10 941 (10.9) |
| Sales/clerical work | 622 (11.1) | 10 115 (10.1) |
| Mechanic/skilled | 1288 (23.1) | 23 257 (23.1) |
| Machine operator/inspector/driver | 452 (8.1) | 9116 (9.1) |
| Service worker | 193 (3.5) | 3996 (4.0) |
| Laborer/unskilled worker | 533 (9.5) | 12 012 (11.9) |
| Farming | 353 (6.3) | 9045 (9.0) |
| Military | 336 (6.0) | 4836 (4.8) |
| No work outside the home | 23 (0.4) | 410 (0.4) |
| F1 did not know | 107 (1.9) | 1700 (1.7) |
| F0 smoking during pregnancy | ||
| Yes | 1527 (27.3) | 23 104 (23.0) |
| No | 3254 (58.2) | 63 939 (63.6) |
| F1 did not know | 528 (9.5) | 8864 (8.8) |
| Missing | 278 (5.0) | 4704 (4.7) |
| F0 lifetime history of depression | 839 (16.5) | 10 316 (10.7) |
| F1 total No. of children, mean (SD) | 2.51 (0.9) | 2.60 (1.0) |
Abbreviations: ADHD, attention-deficit/hyperactivity disorder; F0, mothers of participants in the Nurses’ Health Study II; F1, participants in the Nurses’ Health Study II; F2, live-born children of the F1 participants.
F1 father’s occupation categories are not mutually exclusive.
We observed harmful associations between the F0 mothers’ diethylstilbestrol use while pregnant and the F2 child’s ADHD diagnosis (Table 3). Use of diethylstilbestrol by F0 mothers was associated with increased odds of ADHD among the F2 generation (adjusted OR, 1.36; 95% CI, 1.10-1.67). Further adjusting for F0 mothers’ age at delivery (excluding 1928 F0 mothers and 4368 F2 children with unknown age) yielded the same results (OR, 1.35; 95% CI, 1.09-1.66; P = .005). We observed no difference in the effect estimate when we included information on F0 mothers’ depression history in the main model (OR, 1.33; 95% CI, 1.08-1.63; P = .007).
Table 3. Odds Ratios and 95% CIs for F2 Children’s ADHD, by F0 Diethylstilbestrol Use During Pregnancy.
| Exposure | Children, No. | ADHD Cases, No. (%) | OR (95% CI) | P Value |
|---|---|---|---|---|
| Any diethylstilbestrol | ||||
| Unexposed | 104 414 | 5450 (5.2) | 1 [Reference] | |
| Exposed (crude)a | 1784 | 137 (7.7) | 1.43 (1.16-1.76) | .001 |
| Exposed (adjusted)b | 1.36 (1.10-1.67) | .004 | ||
| By trimesterb | ||||
| Unexposed | 104 414 | 5450 (5.2) | 1 [Reference] | |
| First | 950 | 82 (8.6) | 1.63 (1.18-2.25) | .003 |
| Second | 519 | 33 (6.4) | 0.68 (0.35-1.34) | .27 |
| Third | 338 | 27 (8.0) | 1.41 (0.72-2.82) | .32 |
| F1 did not know | 625 | 42 (6.7) | 1.15 (0.80-1.65) | .46 |
Abbreviations: ADHD, attention-deficit/hyperactivity disorder; F0, mothers of participants in the Nurses’ Health Study II; F1, participants in the Nurses’ Health Study II; F2, live-born children of the F1 participants; OR, odds ratio.
Crude model: cluster-weighted generalized estimating equations without including any covariates. A crude model ignoring any F2-generation clustering within the F1 generation yielded an OR of 1.51 (95% CI, 1.27-1.80).
Adjusted model: cluster-weighted generalized estimating equations for F1 race, F1 ethnicity, F1 year of birth and squared year of birth, F0 smoking during pregnancy with F1, whether F0 owned her home at the time of pregnancy, F0 education, F1 father’s education, F0 occupation, and F1 father’s occupation.
Information on sex was available for only 100 786 F2 children, 48 831 (48.5%) of whom were female. Of the 5338 F2 children with ADHD who had available sex information, 3977 (74.5%) were male. We found no effect modification by children’s sex (P = .62 for the interaction).
In the trimester-specific analysis, we observed that the association was limited to diethylstilbestrol use during the first trimester (OR, 1.63; 95% CI, 1.18-2.25). The ORs for the other trimesters were not substantially different from 1 (Table 3).
In sensitivity analyses, among the 51 561 F0 mothers and 115 806 F2 children (8699 [7.5%] of whom had ADHD) identified when only the ADHD reported on the 2013 questionnaire was considered, we found an OR of 1.28 (95% CI, 1.08-1.52; P = .005) for any diethylstilbestrol use during pregnancy. Our sensitivity analyses of potential unmeasured confounding are presented in the eMethods and eFigure in the Supplement. Briefly, we do not believe that the estimated associations can be fully attributed to unmeasured confounding.
Discussion
We conducted a large-scale cohort analysis to assess the association between use of diethylstilbestrol during pregnancy and third-generation ADHD. The observed associations were robust to covariate adjustment and sensitivity analyses. Despite animal evidence of adverse multigenerational consequences—including neurodevelopmental disorders—of EDC exposure,27,28,47 to date only a few studies have explored the potential multigenerational implications of EDC exposure in humans. These studies have only considered diethylstilbestrol exposure, and none has studied neurodevelopmental outcomes. Some studies have reported increased risk of hypospadias in grandsons of women exposed to diethylstilbestrol during pregnancy.32,33,34 Titus-Ernstoff et al35 found delayed menstrual regularization, higher odds of irregular menstrual periods, and fewer live births among women whose grandmothers used diethylstilbestrol during pregnancy. Birth defects have also been found in grandchildren of women who used diethylstilbestrol when pregnant.48
Oocytes that will develop into the F2 child grow while the F1 mother is still in utero. Thus, third-generation implications of diethylstilbestrol (and other EDC) exposures could relate to direct exposure of F1’s oocytes developing during the F0 pregnancy. This does not imply transgenerational inheritance of epigenetic changes (ie, effects that do not derive from direct exposures, which would require F3 assessment), but animal studies do suggest that epigenetics play a role. Evidence from animal studies suggests that the biological pathway for multigenerational and transgenerational consequences of EDCs involves molecular alterations to the germline, mediated through epigenetic mechanisms, to promote outcomes to subsequent generations22,23,24,25,26; this process is known as epigenetic transgenerational inheritance.22,49 As described earlier, F0 exposure to EDCs—specifically to di(2-ethylhexyl) phthalate and bisphenol A—has been linked to behavioral changes and altered social interactions among mice in later generations.27,28 Direct epidemiological evidence of multigenerational and transgenerational neurodevelopmental outcomes of EDCs is lacking, but 2 studies have reported increased risk for ADHD and autism spectrum disorders among boys with hypospadias,50,51 which in F2 has also been linked to F0 diethylstilbestrol exposure during pregnancy.34 Note that diethylstilbestrol is similar structurally and functionally to bisphenol A, albeit more potent.52 However, Newbold et al47 have observed multigenerational consequences of diethylstilbestrol even at very low diethylstilbestrol estrogenic doses comparable to environmental EDC exposures.
Our trimester-specific analyses indicated that first-trimester F0-generation diethylstilbestrol use is associated with significantly increased risk for F2-generation ADHD. The associations with diethylstilbestrol use during the second and third trimesters were slightly weaker and not significant. The attenuated associations with the other trimesters could result from the smaller number of exposed cases (33 for the second trimester and 27 for the third trimester vs 82 for the first trimester). Although additional studies are warranted, these findings could suggest that the first trimester is a critical exposure window. The early phase of gestation is thought to be especially sensitive to maternal influences, resulting in embryonic and germ cell reprogramming, because during this period a wave of genome demethylation followed by de novo remethylation occurs together with the establishment of imprints and determination of sex.53,54
Limitations
We were in the unique position to access information across 3 generations in a well-defined nationwide cohort comprising a large number of F2-generation ADHD cases and extensive follow-up. Our findings, nevertheless, should be interpreted in light of our limitations.
Outcome misclassification is possible. The ADHD assessment was based on maternal reports of a doctor’s diagnosis. Maternal reports have been found to be highly reliable,38 and we found good agreement in the internal validation study. In addition, in a sensitivity analysis including all 2013 questionnaire responses, we observed only a slight attenuation, suggesting that any bias from misclassification was likely minimal.
Some exposure misclassification is likely, as exposure assessment depended on the F1 participants’ report on whether their F0 mothers used diethylstilbestrol. To minimize potential exposure misclassification, we excluded from our analyses any women whose exposure was uncertain. Furthermore, the small validation study in our data comparing F1 responses with those directly obtained from F0 mothers showed very good agreement. Any exposure or outcome misclassification is likely nondifferential, as information on exposure was collected 12 to 20 years prior to the question about the outcome and in 1993 (when the question on diethylstilbestrol use was asked) there was no toxicological or epidemiological evidence of EDC multigenerational outcomes. Indeed, in our validation study, we found that agreement between F0-generation and F1-generation diethylstilbestrol reporting did not vary by ADHD status. Any potential resulting bias, thus, would likely be toward the null.
Some residual confounding is possible given that we used F1-provided information on F0 factors. However, we included information about several socioeconomic status factors, including smoking during pregnancy, and conducted sensitivity analyses; therefore, our findings are not likely to be completely attributable to residual confounding. Finally, although confounding by indication is possible, the observed different consequences across trimester-specific diethylstilbestrol use suggest that this is unlikely.55,56
Conclusions
We found evidence in humans of increased ADHD risk in the third generation following diethylstilbestrol exposure during F0 pregnancy. The levels of a high number of EDCs to which humans are daily exposed are lower than that of the prescribed diethylstilbestrol; thus, extending this work to lower levels of exposure is warranted. Because EDCs are ubiquitous, the high prevalence of exposure and the possibility of cumulative consequences must be considered.
eMethods. Impact of Unmeasured Confounding
eFigure. Estimated E Value for OR = 1.36 and Outcome Prevalence Less Than 15%
References
- 1.American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. http://dsm.psychiatryonline.org. Accessed March 5, 2014.
- 2.Grandjean P, Landrigan PJ. Neurobehavioural effects of developmental toxicity. Lancet Neurol. 2014;13(3):330-338. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Danckaerts M, Sonuga-Barke EJ, Banaschewski T, et al. . The quality of life of children with attention deficit/hyperactivity disorder: a systematic review. Eur Child Adolesc Psychiatry. 2010;19(2):83-105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Loe IM, Feldman HM. Academic and educational outcomes of children with ADHD. J Pediatr Psychol. 2007;32(6):643-654. [DOI] [PubMed] [Google Scholar]
- 5.Moffitt TE, Houts R, Asherson P, et al. . Is adult ADHD a childhood-onset neurodevelopmental disorder? evidence from a four-decade longitudinal cohort study. Am J Psychiatry. 2015;172(10):967-977. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Brod M, Schmitt E, Goodwin M, Hodgkins P, Niebler G. ADHD burden of illness in older adults: a life course perspective. Qual Life Res. 2012;21(5):795-799. [DOI] [PubMed] [Google Scholar]
- 7.Das D, Cherbuin N, Butterworth P, Anstey KJ, Easteal S. A population-based study of attention deficit/hyperactivity disorder symptoms and associated impairment in middle-aged adults. PLoS One. 2012;7(2):e31500. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lensing MB, Zeiner P, Sandvik L, Opjordsmoen S. Quality of life in adults aged 50+ with ADHD. J Atten Disord. 2015;19(5):405-413. [DOI] [PubMed] [Google Scholar]
- 9.Agarwal R, Goldenberg M, Perry R, Ishak WW. The quality of life of adults with attention deficit hyperactivity disorder: a systematic review. Innov Clin Neurosci. 2012;9(5-6):10-21. [PMC free article] [PubMed] [Google Scholar]
- 10.Bernardi S, Faraone SV, Cortese S, et al. . The lifetime impact of attention deficit hyperactivity disorder: results from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). Psychol Med. 2012;42(4):875-887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Cortese S, Moreira-Maia CR, St Fleur D, Morcillo-Peñalver C, Rohde LA, Faraone SV. Association between ADHD and obesity: a systematic review and meta-analysis. Am J Psychiatry. 2016;173(1):34-43. [DOI] [PubMed] [Google Scholar]
- 12.Cook BG, Li D, Heinrich KM. Obesity, physical activity, and sedentary behavior of youth with learning disabilities and ADHD. J Learn Disabil. 2015;48(6):563-576. [DOI] [PubMed] [Google Scholar]
- 13.Spencer TJ, Faraone SV, Tarko L, McDermott K, Biederman J. Attention-deficit/hyperactivity disorder and adverse health outcomes in adults. J Nerv Ment Dis. 2014;202(10):725-731. [DOI] [PubMed] [Google Scholar]
- 14.Sorge GB, Flora DB, van der Maas M, et al. . Using factor analytic models to examine the association between attention-deficit/hyperactivity disorder symptoms and health-related outcomes in a representative population survey. Atten Defic Hyperact Disord. 2015;7(3):225-235. [DOI] [PubMed] [Google Scholar]
- 15.Nigg JT. Attention-deficit/hyperactivity disorder and adverse health outcomes. Clin Psychol Rev. 2013;33(2):215-228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dalsgaard S, Østergaard SD, Leckman JF, Mortensen PB, Pedersen MG. Mortality in children, adolescents, and adults with attention deficit hyperactivity disorder: a nationwide cohort study. Lancet. 2015;385(9983):2190-2196. [DOI] [PubMed] [Google Scholar]
- 17.Colborn T. Neurodevelopment and endocrine disruption. Environ Health Perspect. 2004;112(9):944-949. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Damstra T. Potential effects of certain persistent organic pollutants and endocrine disrupting chemicals on the health of children. J Toxicol Clin Toxicol. 2002;40(4):457-465. [DOI] [PubMed] [Google Scholar]
- 19.de Cock M, Maas YG, van de Bor M. Does perinatal exposure to endocrine disruptors induce autism spectrum and attention deficit hyperactivity disorders? review. Acta Paediatr. 2012;101(8):811-818. [DOI] [PubMed] [Google Scholar]
- 20.Herbstman JB, Sjödin A, Kurzon M, et al. . Prenatal exposure to PBDEs and neurodevelopment. Environ Health Perspect. 2010;118(5):712-719. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kim B-N, Cho S-C, Kim Y, et al. . Phthalates exposure and attention-deficit/hyperactivity disorder in school-age children. Biol Psychiatry. 2009;66(10):958-963. [DOI] [PubMed] [Google Scholar]
- 22.Skinner MK. Endocrine disruptors in 2015: epigenetic transgenerational inheritance. Nat Rev Endocrinol. 2016;12(2):68-70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Anway MD, Skinner MK. Epigenetic transgenerational actions of endocrine disruptors. Endocrinology. 2006;147(6)(suppl):S43-S49. [DOI] [PubMed] [Google Scholar]
- 24.Anway MD, Skinner MK. Epigenetic programming of the germ line: effects of endocrine disruptors on the development of transgenerational disease. Reprod Biomed Online. 2008;16(1):23-25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Skinner MK, Manikkam M, Guerrero-Bosagna C. Epigenetic transgenerational actions of endocrine disruptors. Reprod Toxicol. 2011;31(3):337-343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Walker DM, Gore AC. Transgenerational neuroendocrine disruption of reproduction. Nat Rev Endocrinol. 2011;7(4):197-207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Quinnies KM, Doyle TJ, Kim KH, Rissman EF. Transgenerational effects of di-(2-ethylhexyl) phthalate (DEHP) on stress hormones and behavior. Endocrinology. 2015;156(9):3077-3083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wolstenholme JT, Edwards M, Shetty SR, et al. . Gestational exposure to bisphenol A produces transgenerational changes in behaviors and gene expression. Endocrinology. 2012;153(8):3828-3838. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Giusti RM, Iwamoto K, Hatch EE. Diethylstilbestrol revisited: a review of the long-term health effects. Ann Intern Med. 1995;122(10):778-788. [DOI] [PubMed] [Google Scholar]
- 30.Dieckmann WJ, Davis ME, Rynkiewicz LM, Pottinger RE. Does the administration of diethylstilbestrol during pregnancy have therapeutic value? Am J Obstet Gynecol. 1953;66(5):1062-1081. [DOI] [PubMed] [Google Scholar]
- 31.Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of the vagina: association of maternal stilbestrol therapy with tumor appearance in young women. N Engl J Med. 1971;284(15):878-881. [DOI] [PubMed] [Google Scholar]
- 32.Kalfa N, Paris F, Soyer-Gobillard M-O, Daures J-P, Sultan C. Prevalence of hypospadias in grandsons of women exposed to diethylstilbestrol during pregnancy: a multigenerational national cohort study. Fertil Steril. 2011;95(8):2574-2577. [DOI] [PubMed] [Google Scholar]
- 33.Brouwers MM, Feitz WF, Roelofs LA, Kiemeney LA, de Gier RP, Roeleveld N. Hypospadias: a transgenerational effect of diethylstilbestrol? Hum Reprod. 2006;21(3):666-669. [DOI] [PubMed] [Google Scholar]
- 34.Klip H, Verloop J, van Gool JD, Koster ME, Burger CW, van Leeuwen FE; OMEGA Project Group . Hypospadias in sons of women exposed to diethylstilbestrol in utero: a cohort study. Lancet. 2002;359(9312):1102-1107. [DOI] [PubMed] [Google Scholar]
- 35.Titus-Ernstoff L, Troisi R, Hatch EE, et al. . Menstrual and reproductive characteristics of women whose mothers were exposed in utero to diethylstilbestrol (DES). Int J Epidemiol. 2006;35(4):862-868. [DOI] [PubMed] [Google Scholar]
- 36.Mahalingaiah S, Hart JE, Wise LA, Terry KL, Boynton-Jarrett R, Missmer SA. Prenatal diethylstilbestrol exposure and risk of uterine leiomyomata in the Nurses’ Health Study II. Am J Epidemiol. 2014;179(2):186-191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.O’Reilly EJ, Mirzaei F, Forman MR, Ascherio A. Diethylstilbestrol exposure in utero and depression in women. Am J Epidemiol. 2010;171(8):876-882. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Faraone SV, Biederman J, Milberger S. How reliable are maternal reports of their children’s psychopathology? one-year recall of psychiatric diagnoses of ADHD children. J Am Acad Child Adolesc Psychiatry. 1995;34(8):1001-1008. [DOI] [PubMed] [Google Scholar]
- 39.DuPaul GJ, Power TJ, Reid R, Anastopoulos AD. ADHD Rating Scale—IV: Checklists, Norms, and Clinical Interpretation. New York, NY: Guilford Press; 1998. [Google Scholar]
- 40.Gao X, Lyall K, Palacios N, Walters AS, Ascherio A. RLS in middle aged women and attention deficit/hyperactivity disorder in their offspring. Sleep Med. 2011;12(1):89-91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Hernan MA, Robins JM. Causal Inference. Boca Raton, FL: Chapman Hall/CRC Press; 2018. [Google Scholar]
- 42.Glover V. Maternal depression, anxiety and stress during pregnancy and child outcome; what needs to be done. Best Pract Res Clin Obstet Gynaecol. 2014;28(1):25-35. [DOI] [PubMed] [Google Scholar]
- 43.Seaman S, Pavlou M, Copas A. Review of methods for handling confounding by cluster and informative cluster size in clustered data. Stat Med. 2014;33(30):5371-5387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Williamson JM, Datta S, Satten GA. Marginal analyses of clustered data when cluster size is informative. Biometrics. 2003;59(1):36-42. [DOI] [PubMed] [Google Scholar]
- 45.Seaman SR, Pavlou M, Copas AJ. Methods for observed-cluster inference when cluster size is informative: a review and clarifications. Biometrics. 2014;70(2):449-456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the e-value. Ann Intern Med. 2017;167(4):268-274. [DOI] [PubMed] [Google Scholar]
- 47.Newbold RR, Padilla-Banks E, Jefferson WN. Adverse effects of the model environmental estrogen diethylstilbestrol are transmitted to subsequent generations. Endocrinology. 2006;147(6 suppl):S11-S17. [DOI] [PubMed] [Google Scholar]
- 48.Titus-Ernstoff L, Troisi R, Hatch EE, et al. . Birth defects in the sons and daughters of women who were exposed in utero to diethylstilbestrol (DES). Int J Androl. 2010;33(2):377-384. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Skinner MK. Endocrine disruptor induction of epigenetic transgenerational inheritance of disease. Mol Cell Endocrinol. 2014;398(1-2):4-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Butwicka A, Lichtenstein P, Landén M, et al. . Hypospadias and increased risk for neurodevelopmental disorders. J Child Psychol Psychiatry. 2015;56(2):155-161. [DOI] [PubMed] [Google Scholar]
- 51.Rotem RS, Chodick G, Davidovitch M, Hauser R, Coull BA, Weisskopf MG. Congenital abnormalities of the male reproductive system and risk of autism spectrum disorders [published online February 14, 2018]. Am J Epidemiol. doi: 10.1093/aje/kwx367 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Dodds EC, Lawson W. Synthetic, oestrogenic agents without the phenanthrene nucleus. Nature. 1936;137:996. [Google Scholar]
- 53.Bale TL. Epigenetic and transgenerational reprogramming of brain development. Nat Rev Neurosci. 2015;16(6):332-344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Weaver JR, Susiarjo M, Bartolomei MS. Imprinting and epigenetic changes in the early embryo. Mamm Genome. 2009;20(9-10):532-543. [DOI] [PubMed] [Google Scholar]
- 55.Lipsitch M, Tchetgen Tchetgen E, Cohen T. Negative controls: a tool for detecting confounding and bias in observational studies. Epidemiology. 2010;21(3):383-388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Weisskopf MG, Tchetgen Tchetgen EJ, Raz R. Commentary: on the use of imperfect negative control exposures in epidemiological studies. Epidemiology. 2016;27(3):365-367. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Impact of Unmeasured Confounding
eFigure. Estimated E Value for OR = 1.36 and Outcome Prevalence Less Than 15%