Abstract
Background
Diethylstilbestrol (DES) is an endocrine-disrupting pharmaceutical prescribed to pregnant women to prevent pregnancy complications between the 1940s and 1970s. Although DES has been shown in animal studies to have multigenerational effects, only two studies have investigated potential multigenerational effects in humans on preterm birth (PTB), and none on low birthweight (LBW)—major determinants of later life health.
Methods
Nurses’ Health Study (NHS) II participants (G1; born 1946–64) reported their mothers’ (G0) use of DES while pregnant with them. We used cluster-weighted generalized estimating equations to estimate odds ratios (OR) and 95% confidence intervals (CI) for risk of LBW and PTB among the grandchildren by grandmother use of DES. G1 birthweight and gestational age were considered to explore confounding by indication.
Results
Among 54 334 G0-G1/grandmother-mother pairs, 973 (1.8%) G0 used DES during pregnancy with G1. Of the 128 275 G2 children, 4369 (3.4%) were LBW and 7976 (6.2%) premature. Grandmother (G0) use of DES during pregnancy was associated with an increased risk of G2 LBW [adjusted OR (aOR) = 3.09; 95% CI: 2.57, 3.72], that was reduced when restricted to term births (aOR = 1.59; 95% CI: 1.08, 2.36). The aOR for PTB was 2.88 (95% CI: 2.46, 3.37). Results were essentially unchanged when G1 birthweight and gestational age were included in the model, as well as after adjusting for other potential intermediate variables, such as G2 pregnancy-related factors.
Conclusions
Grandmother use of DES during pregnancy is associated with an increased risk of LBW, predominantly through an increased risk of PTB. Results when considering G1 birth outcomes suggest this does not result from confounding by indication.
Keywords: Diethylstilbestrol, endocrine disrupting chemical, birth outcomes, low birthweight, preterm birth, preterm delivery, prenatal exposures, multigenerational effects, germline cells, confounding by indication
Key Messages.
Grandmothers’ use of diethylstilbestrol during pregnancy is associated with higher risk of premature birth and low birthweight among the grandchildren after adjustment for many potential confounding factors.
The findings from the current study argue against indication bias and mediation of the observed associations by maternal factors.
Effects on the fetal germline could underlie the observed association; the germline effects should be considered in regulatory risk assessment.
Introduction
Diethylstilbestrol (DES) was first developed in 1938 as a synthetic estrogen and is estimated to be five times more potent than the naturally occurring estrogen, estradiol.1 DES was prescribed to approximately 10 million pregnant US women worldwide between the 1940s and 1970s,2 when it was banned by the U.S. Food and Drug Administration (FDA).3 The prescription of DES was strongly promoted since the early 1940s to prevent premature deliveries and miscarriages.4 DES was mainly administered to women with a history of or threatened pregnancy loss, but it became the standard of care for high-risk pregnancies among obstetricians during the later years of use. This drug was even administered to women with otherwise healthy pregnancies, being advertised as ‘routine prophylaxis in all pregnancies’.5 By the 1940s the effectiveness of DES was already being questioned,6 data even suggesting it promoted premature delivery rather than prevented it,7 and later research found multiple adverse outcomes in children exposed in utero.8,9 Although the use of DES was discontinued, humans are currently exposed to a large number of other similarly acting estrogenic endocrine-disrupting chemicals, such as bisphenol A,10 albeit ones typically with lower potency than DES.
More recently, a few studies have linked DES exposure during pregnancy to adverse health outcomes in generations beyond the one exposed in utero.2 Furthermore, animal research suggests such multigenerational effects of endocrine-disrupting chemical exposures.11 Preterm birth (PTB) and low birthweight (LBW) are key birth outcomes linked with early life morbidity and mortality,12 as well as adulthood diseases, such as hypertension, type 2 diabetes and cardiovascular diseases.13 No studies in humans have considered whether pregnancy DES exposure is associated with later generations’ LBW, and only two have considered PTB. Pregnancy DES use was found to be associated with greater risk of PTB in the third generation (grandchildren) in the National Cancer Institute’s Collaborative DES Follow-Up Study,14 and in a combination of the National Cooperative Diethylstilbestrol Adenosis (DESAD) Project and follow-up of a University of Chicago clinical trial.15 Recently, exposure to DES was also associated with suggestively elevated PTB in the fourth generation (great grandchildren; prevalence ratio = 1.31; 95% CI: 0.81, 2.10).16 However, these analyses included limited potential confounders (date of birth, cohort and number of births) and were potentially susceptible to confounding by indication, which occurs when a reason for a drug to be prescribed is associated with the outcome of interest.17 Therefore, the current study sought to examine the question of DES use in association with LBW and PTB in the third generation, while attempting to better address potential confounding especially by indication.
Methods
Informed consent was implied through the return of the baseline questionnaire of participating women. The institutional review boards of Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health approved the study protocol (IRB protocol number: P16731 2008P001152).
Study population
The Nurses’ Health Study II (NHS II) is a prospective cohort study of 116 430 female registered nurses who were born between 1946 and 1964 in the USA, a time when DES was prescribed to pregnant women.18 Since recruitment in 1989, these nurses have returned follow-up questionnaires biennially with information on lifestyle, medication use and the occurrence of major illnesses.19 In the present study, we consider the NHS II participants (hereafter, referred to as nurses) as generation 1 (G1), their mothers as generation 0 (G0) and their children as generation 2 (G2).
The Nurses’ Mothers’ Cohort Study (NMCS) was launched in 2001 and obtained information regarding the prenatal and childhood environment of the NHS II participants directly from their mothers. A total of 39 904 G1 nurses’ mothers who were alive and free of cancer in 2000 participated in the NMCS.20
Our main analytical population derived from the 54 672 nurses (G1) who were not themselves adopted, returned the NHS II questionnaires with information on G0 use of DES and G2 PTB and LBW, reported at least one singleton live birth and were sure of their in utero DES exposure. We excluded any children (G2) with missing information on gestational weeks (n = 536), reported gestational age of less than 20 weeks (n = 109) or missing information on birthweight (n = 700), leaving 54 334 G0/G1 dyads and 128 275 G2. In a secondary analysis, we repeated our main analysis using NMCS data among 22 422 G0/G1 dyads and 53 103 G2 after applying the same exclusion criteria. Our study was approved by the institutional review board of Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health, Boston, MA.
Use of DES assessment
In 1993, NHS II participants (G1) were asked whether their mothers (G0) used DES during the pregnancy with the nurse, and a supplementary questionnaire was then mailed to the 2742 women who said they had been exposed to DES.10 Among the 2317 G1 respondents to the supplementary questionnaire, 2032 (87.7%) reported that they were certain or somewhat certain of their exposure, 123 (5.3%) were not certain and 162 (7.0%) said they had had no such exposure.10 These latter two groups were excluded from our analyses. The supplementary questionnaire also asked about exposure specifically during the first, second or third trimester of pregnancy, or nurses (G1) could report that they didn’t know. In the NMCS direct G0 report of DES use during pregnancy was a binary response (yes/no). G0- and G1-reported DES exposure showed good agreement (κ = 0.74).21 This kappa coefficient did not differ by G2 LBW (κ = 0.69; 95% CI: 0.67, 0.72 for healthy birthweight; κ = 0.73; 95% CI: 0.65, 0.8 for LBW) or PTB (κ = 0.69; 95% CI: 0.66, 0.71 for term births; κ = 0.73; 95% CI: 0.68, 0.78 for PTB).
LBW and PTB assessment
On the 2009 questionnaire, NHS II participants reported their pregnancy histories, including the length of each of their pregnancies in the following categories: 12 to <20 weeks, 20 to <28 weeks, 28 to <32 weeks, 32 to <37 weeks, 37 to <40 weeks, 40 to <43 weeks, and ≥43 weeks. Participants were also asked the birthweight for each of their pregnancies (<5 pounds [lbs] [2.3 kg], 5–5.4 lbs [2.3–2.4 kg], 5.5–6.9 lbs [2.5–3.1 kg], 7.0–8.4 lbs [3.2–3.8 kg], 8.5–9.9 lbs [3.9–4.5 kg], ≥10 lbs [4.5 kg]). We defined LBW as less than 2.5 kg (5.5 lbs)22 and PTB as gestational age less than 37 weeks.23 Our study participants could not be classified according to small for gestational age (SGA) because both birthweight and gestational age were reported as categorical variables. Additionally, the number of G1 stillbirth(s) was counted (0, 1, 2, 3 and 4+).
Covariates
Potential confounders included measures of: grandparents’ (G0) socioeconomic status: grandmother and grandfather education (high school or less/college or more); grandfather (G0) occupation (blue and lower white-collar workers, labourers, farmers, upper white-collar workers or none); grandparental home ownership in G1 infancy (yes/no); grandmother (G0) smoking during pregnancy (yes, no, don’t know); grandmother (G0) age at G1 birth; and G1 race (White or non-White). Mothers’ (G1) year of birth was further included to account for secular trends in the DES use.1 G1 answer to a question ‘How do you feel about your standing in US society using a ladder?’ in 2001 was added as a G1 socioeconomics status variable. G0 alcohol use during pregnancy (yes/no) and whether grandmother ever had a miscarriage (yes/no) were also considered, but only available from the NMCS participants.
Other variables were considered in secondary analyses because they might be mediating factors or address possible confounding by indication, i.e. G1 gestational age (born ≥2 weeks early, ±2 weeks of due date, or >2 weeks late),24 G1 birthweight (<5.5 lbs [2.5 kg], 5.5–9.9 lbs [2.5–4.5 kg], or 10+ lbs [4.5+ kg]),24,25 G1 history of miscarriages/stillbirths (none, 1, 2, 3 or 4+),1 G1 pre-pregnancy body mass index [underweight (< 18.5 kg/m2), healthy/normal (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2) and obese (≥ 30 kg/m2)],26,27 G1 smoking or alcohol use during pregnancy with G2 (yes/no)28,29 and G1 pregnancy complications (yes/no: any of pre-eclampsia/toxaemia, pregnancy-related high blood pressure or gestational diabetes).30 G1 gestational weeks information was only available from the NMCS participants. We additionally considered G2 premature birth (<37 weeks) as a possible mediator for G2 LBW.31
Statistical analysis
Prenatal exposure to DES is known to affect reproductive function of offspring and so could affect the number of grandchildren (G2) of each G0.32 Since the distribution of LBW and PTB could also possibly depend on the number of grandchildren, our analysis is potentially subject to informative clustering.33,34 To address the correlation among multiple children from the same G1 and possible informative clustering simultaneously, we used cluster-weighted generalized estimating equations with a logit link and weighted by the number of G2 per G1, to assess the association between grandmother (G0) use of DES and grandchild (G2) risk of LBW and PTB. Models were adjusted for grandparental (G0) socioeconomic status (grandmother and grandfather education, grandfather occupation and grandparents’ home ownership at G1 birth), grandmother smoking during pregnancy with G1, grandmother age at G1 birth, mother (G1) race, mother year of birth and mother socioeconomic status in a ladder scale. We considered any DES exposure during pregnancy and trimester-specific exposures. Trimester-specific exposures were not mutually exclusive and were modelled as three binary variables plus one for those who did not know about trimester-specific timing, all of which were included simultaneously in the model. We also estimated the relative risk of G1 stillbirth according to G0 use of DES with the incidence rate ratio (IRR), with Poisson distribution and a log link.
In secondary analyses, stratified analyses were performed to examine whether the main association differed by G2 sex. Additionally, we considered mother (G1) birthweight, gestational weeks, smoking during pregnancy with G2, alcohol use during pregnancy with G2, pre-pregnancy BMI, pregnancy complications (pre-eclampsia/toxaemia, pregnancy-related high blood pressure or gestational diabetes) during pregnancy with G2 as possible mediating variables. We first checked for interaction between the G0 DES use and each mediating factor examined by including a multiplicative interaction term in models, but none was detected. Therefore, we used the traditional epidemiological approach for mediation analysis of including the intermediate factor in a logistic regression model of exposure on outcome to assess mediation.35 With this approach, if an association between DES and G2 outcomes is reduced or eliminated after adding the possible mediating factor, the interpretation would be that that factor mediates the association, assuming there is no confounding of the mediator-outcome association with opposite direction to a DES-outcome association. We specifically examined the association of G0 DES use with G1 birthweight and gestational age, and the effect of adjusting for these G1 factors as well as G1 history of miscarriages/stillbirths on results of G0 DES-G2 birth outcomes, to investigate possible confounding by indication. Since some of G2 could have been additionally exposed to G1 use of DES, we performed an additional sensitivity analysis excluding the G2 who were born in 1970 or before. Given the generally low missingness in our data (<10%), our main analyses included missing indicators for any missing data. All analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC).
Results
Among the 54 334 grandmothers, 973 (1.8%) were exposed to DES while pregnant with the G1 nurse and 53 361 (98.2%) were not (Tables 1 and 2). G0 with reported exposure to DES during pregnancy with G1 were more likely to have attended college and have husbands (the grandfather) who had professional occupations. G0 who used DES during pregnancy with G1 were also more likely to have smoked while pregnant with the G1 nurse. G1 participants who were prenatally exposed to DES had a median of two children (range, 1–7), and their unexposed counterparts had a median of three children (range, 1–12). Similar patterns were seen when considering the 22 422 G0 in the NMCS (Supplementary Tables S1 and S2, available as Supplementary data at IJE online).
Table 1.
Characteristics of the G0 (n = 54 334) generation by grandmother (G0) use of DES during pregnancy, reported by G1, in the Nurses’ Health Study II
| Characteristic | Grandmother (G0) use of DES during pregnancy, no. (%) |
|
|---|---|---|
| No (n = 53 361) | Yes (n = 973) | |
| Grandmother’s education, n (%) | ||
| High school or less | 32 360 (60.64) | 532 (54.68) |
| College or more | 15 137 (28.37) | 357 (36.69) |
| Missing | 5864 (11.0) | 84 (8.6) |
| Grandfather’s education, n (%) | ||
| High school or less | 29 224 (54.77) | 467 (48.00) |
| College or more | 17 603 (32.99) | 415 (42.65) |
| Missing | 6534 (12.2) | 91 (9.4) |
| Grandfather’s occupation, n (%) | ||
| Blue-collara | 24 547 (46.0) | 455 (46.8) |
| Labourer | 5497 (10.3) | 71 (7.3) |
| Farmer | 3355 (6.3) | 44 (4.5) |
| Professionalb | 13 523 (25.3) | 318 (32.7) |
| None | 107 (0.2) | 0 (0) |
| Missing | 6332 (11.9) | 85 (8.7) |
| Grandmother home ownership at G1 birth, n (%) | ||
| Yes | 25 805 (48.4) | 490 (50.4) |
| No | 21 803 (40.9) | 407 (41.8) |
| Missing | 5753 (10.8) | 76 (7.8) |
| Grandmother smoking during pregnancy, n (%) | ||
| No | 32 417 (60.8) | 536 (55.1) |
| Yes | 12 239 (22.9) | 297 (30.5) |
| Missing | 8705 (16.3) | 140 (14.4) |
| Grandmother birth year, median (range) | 1927 (1897–1951) | 1926 (1908–1950) |
| Missing, n (%) | 7079 (13.3) | 104 (10.7) |
| Grandmother age (years) at G1 birth, mean (SD) | 27.3 (5.7) | 28.2 (5.3) |
| Missing, n (%) | 7136 (13.4) | 106 (10.9) |
G0, generation 0; G1, generation 1; DES, diethylstilbestrol; SD, standard deviation.
Sales/clerical work, service worker, mechanic/electric/skilled work, machine operator/inspector/driver or military work.
Professional, executive or manager.
Table 2.
Characteristics of the G1 (n = 54 334) and G2 (n = 128 275) generations by grandmother (G0) use of DES during pregnancy, reported by G1, in the Nurses’ Health Study II
| Characteristic | Grandmother (G0) use of DES during pregnancy, no. (%) |
||
|---|---|---|---|
| No (n = 53 361) | Yes (n = 973) | ||
| G1 generation | |||
| Mother’s race, n (%) | |||
| White | 49 989 (93.7) | 936 (96.2) | |
| Non-White | 2572 (4.8) | 27 (2.8) | |
| Missing | 800 (1.5) | 10 (1.0) | |
| Mother’s birthweight, n (%) | |||
| <2.5 kg (5.5 lbs) | 3270 (6.1) | 134 (13.8) | |
| 2.5–4.5 kg (5.5–9.9 lbs) | 42 609 (79.9) | 744 (76.5) | |
| ≥4.5 kg (10 lbs) | 535 (1.0) | 9 (0.9) | |
| Missing | 6947 (13.0) | 86 (8.8) | |
| Mother’s subjective social status in the USA using a ladder scale, n (%) | |||
| Top | 19 058 (35.72) | 368 (37.82) | |
| Middle | 23 824 (44.65) | 457 (46.97) | |
| Bottom | 4916 (9.21) | 71 (7.30) | |
| Missing | 5563 (10.43) | 77 (7.91) | |
| Mother’s birth year, median (range) | 1954 (1946–1964) | 1954 (1946-1964) | |
| G2 generation | Grandchildren | ||
| n = 126 097 | n = 2178 | ||
| Median no. of G2a (range) | 3 (1–12) | 2 (1–7) | |
| Maternal (G1) smoking during pregnancy with G2, n (%) | |||
| No | 81 269 (64.5) | 1464 (67.2) | |
| Yes | 9193 (7.3) | 181 (8.3) | |
| Missing | 35 635 (28.3) | 533 (24.5) | |
| Maternal (G1) alcohol use during pregnancy with G2, n (%) | |||
| No | 78 429 (62.2) | 1395 (64.1) | |
| Yes | 11 989 (9.5) | 250 (11.5) | |
| Missing | 35 679 (28.3) | 532 (24.4) | |
| Prenatal exposure to G1 pre-pregnancy BMI, n (%) | |||
| <18.5 kg/m2 | 1887 (1.50) | 40 (1.84) | |
| 18.5–24.9 kg/m2 | 99 519 (78.92) | 1700 (78.05) | |
| 25.0–29.9 kg/m2 | 18 265 (14.48) | 326 (14.97) | |
| ≥30 kg/m2 | 6426 (5.10) | 112 (5.14) | |
| Maternal (G1) pregnancy-related complications during pregnancy with G2,bn (%) | 11 628 (9.2) | 230 (10.6) | |
| Maternal (G1) age (years) at G2 birth, mean (SD) | 29.1 (5.0) | 29.2 (5.0) | |
| Grandchild birth year, median (range) | 1984 (1956–2010) | 1983 (1964–2007) | |
G1: generation 1; G2: generation 2; DES: diethylstilbestrol; SD: standard deviation; BMI, body mass index.
As of 2009.
Pre-eclampsia/toxaemia, pregnancy-related high blood pressure, or gestational diabetes.
Grandmother (G0) use of DES was strongly associated with increased odds of LBW in the G2 generation (Table 3), which was robust to adjustment for many potentially confounding factors [adjusted OR (aOR) = 3.09; 95% CI: 2.57, 3.72]. In analyses restricted to G2 children who were term births, the effect estimate was substantially decreased (adjusted OR = 1.59; 95% CI: 1.08, 2.36). G0 use of DES was strongly associated with G2 PTB (aOR = 2.88; 95% CI: 2.46, 3.37) (Table 4). In analysis of trimester-specific exposures, the associations were slightly stronger in later trimesters for both G2 LBW and G2 PTB (Tables 3 and 4). Results for both G2 LBW and G2 PTB were similar in analyses of NMCS participants using direct G0 report of her DES use during pregnancy with G1 (Table 5). G0 use of DES had a greater rate of G1 stillbirth than G0 non-users, with adjusted IRR of 1.55 (95% CI: 1.43, 1.67).
Table 3.
Odds ratios (OR) and 95% confidence intervals (CI) for low birthweight (LBW) in the third generation (G2) by grandmother’s (G0) use of DES in the Nurses’ Health Study II
| Grandmother (G0) use of DES during pregnancy | Grandchildren (G2) | Low birthweight,an (%) | OR (95% CI) |
|||
|---|---|---|---|---|---|---|
| Unadjusted | P | Adjustedb | P | |||
| Any DES (n = 54 334 G0/G1, 128 275 G2) | ||||||
| Unexposed | 126 097 | 4183 (3.3) | 1 (Reference) | 1 (Reference) | ||
| Exposed | 2178 | 186 (8.5) | 2.98 (2.48, 3.59) | <0.01 | 3.06 (2.54, 3.68) | <0.01 |
| By trimesterc (n = 54 334 G0/G1, 128 275 G2) | ||||||
| Unexposed | 126 097 | 4183 (3.3) | 1 (Reference) | 1 (Reference) | ||
| First | 1205 | 108 (9.0) | 3.12 (2.45, 3.96) | <0.01 | 3.20 (2.52, 4.08) | <0.01 |
| Second | 648 | 65 (10.0) | 3.53 (2.58, 4.82) | <0.01 | 3.62 (2.64, 4.97) | <0.01 |
| Third | 394 | 39 (9.9) | 3.47 (2.31, 5.21) | <0.01 | 3.66 (2.44, 5.51) | <0.01 |
| G1 did not know | 715 | 63 (8.8) | 3.01 (2.20, 4.12) | <0.01 | 3.06 (2.23, 4.19) | <0.01 |
| Any DES, among term births only (n = 52 676 G0/G1, 120 299 G2) | ||||||
| Unexposed | 118 432 | 1561 (1.3) | 1 (Reference) | 1 (Reference) | ||
| Exposed | 1867 | 34 (1.8) | 1.50 (1.02, 2.22) | 0.04 | 1.54 (1.04, 2.28) | 0.03 |
G0, generation 0; G1, generation 1; G2, generation 2; DES, diethylstilbestrol; P, P-value.
Birthweight ≤2.5 kg (5.5 lbs).
Adjusted for grandmother and grandfather (G0) education, grandfather (G0) occupation, grandparents (G0) home ownership at G1 birth, grandmother (G0) smoking during pregnancy with G1, grandmother (G0) age at G1 birth, mother (G1) race, mother (G1) year of birth and mother (G1) socioeconomic status in a ladder scale.
Trimester-specific exposure is not mutually exclusive.
Table 4.
Odds ratios (OR) and 95% confidence intervals (CI) for preterm birth (PTB) in the third generation (G2) by grandmother’s (G0) use of DES, among the 54 334 nurses (G1) in the Nurses’ Health Study II
| Grandmother (G0) use of DES during pregnancy | Grand-children (G2) | Preterm birtha, n (%) | OR (95% CI) |
|||
|---|---|---|---|---|---|---|
| Unadjusted | P | Adjustedb | P | |||
| Any DES | ||||||
| Unexposed | 126 097 | 7665 (6.1) | 1 (Reference) | 1 (Reference) | ||
| Exposed | 2178 | 311 (14.3) | 2.78 (2.37, 3.25) | <0.01 | 2.88 (2.46, 3.37) | <0.01 |
| By trimesterc | ||||||
| Unexposed | 126 097 | 7665 (6.1) | 1 (Reference) | 1 (Reference) | ||
| First | 1205 | 184 (15.3) | 3.03 (2.47, 3.72) | <0.01 | 3.13 (2.55, 3.85) | <0.01 |
| Second | 648 | 107 (16.5) | 3.32 (2.53, 4.35) | <0.01 | 3.41 (2.59, 4.48) | <0.01 |
| Third | 394 | 74 (18.8) | 3.66 (2.62, 5.11) | <0.01 | 3.83 (2.74, 5.35) | <0.01 |
| G1 did not know | 715 | 90 (12.6) | 2.30 (1.74, 3.05) | <0.01 | 2.38 (1.80, 3.15) | <0.01 |
G0, generation 0; G1, generation 1; G2, generation 2; DES, diethylstilbestrol; P, P-value.
Gestational age ≤37 weeks.
Adjusted for grandmother and grandfather (G0) education, grandfather (G0) occupation, grandparents (G0) home ownership at G1 birth, grandmother (G0) smoking during pregnancy with G1, grandmother (G0) age at G1 birth, mother (G1) race, mother (G1) year of birth and mother (G1) socioeconomic status in a ladder scale.
Trimester-specific exposure is not mutually exclusive.
Table 5.
Odds ratios (OR) and 95% confidence intervals (CI) for birth outcomes in the third generation (G2) by grandmother’s (G0) use of DES in the Nurses’ Mothers’ Cohort Study, among the 22 422 nurses (G1)
| Grandmother (G0) use of DES during pregnancy | Grand-children (G2) | Outcome, n (%) | OR (95% CI) |
|||
|---|---|---|---|---|---|---|
| Unadjusted | P | Adjustedc | P | |||
| For grandchild (G2) low birthweighta | ||||||
| No | 47 247 | 1463 (3.1) | 1 (Reference) | 1 (Reference) | ||
| Yes | 1227 | 94 (7.7) | 2.81 (2.17, 3.65) | <0.01 | 2.72 (2.09, 3.54) | <0.01 |
| Missing | 4629 | 182 (3.9) | 1.35 (1.11, 1.63) | <0.01 | 1.32 (1.09, 1.60) | <0.01 |
| For grandchild (G2) preterm birthb | ||||||
| No | 47 247 | 2719 (5.8) | 1 (Reference) | 1 (Reference) | ||
| Yes | 1227 | 157 (12.8) | 2.62 (2.09, 3.27) | <0.01 | 2.65 (2.11, 3.32) | <0.01 |
| Missing | 4629 | 276 (6.0) | 1.06 (0.90, 1.25) | 0.47 | 1.06 (0.90, 1.25) | 0.5 |
G0, generation 0; G2, generation 2; DES, diethylstilbestrol; P, P-value.
Birthweight ≤2.5 kg (5.5 lbs).
Gestational age ≤37 weeks.
Adjusted for grandmother (G0) race, grandparents’ (G0) education, grandfather (G0) occupation, grandparents’ (G0) home ownership at G1 birth, grandmother (G0) smoking during pregnancy, grandmother (G0) alcohol use during pregnancy, grandmother (G0) age at G1 birth, whether grandmother (G0) ever had a miscarriage (yes/no), mother (G1) year of birth and mother (G1) socioeconomic status in a ladder scale.
When we included the potential mediating factors of G1 pregnancy-related complications and G1 smoking and alcohol use during pregnancy with G2, the results were effectively unchanged. In stratified analyses, the effect of G0 DES use was stronger among boys compared with girls, but they were not materially different (Supplementary Tables S3 and S4, available as Supplementary data at IJE online). G0 DES use was associated with G1 LBW (aOR = 2.34; 95% CI: 1.94, 2.83) and being born >2 weeks early (aOR = 1.93; 95% CI: 1.61, 2.31) in analyses adjusted for the same factors considered for G2 outcomes. However, when G1 LBW, being born >2 weeks early and G1 history of miscarriages/stillbirths were included in models of G2 LBW and PTB, the associations with G0 DES use were effectively unchanged from results without including the G1 birth outcomes (Supplementary Tables S5 and S6, available as Supplementary data at IJE online). The results remained consistent when G2 who were born in 1970 or before were excluded from the analysis (Supplementary Table S7, available as Supplementary data at IJE online).
Discussion
We found grandmother use of DES during pregnancy to be associated with increased risk of LBW and PTB in her grandchildren. Our results were robust to adjustment for many G0- and G1-related potential confounders. The association with LBW was substantially reduced after restricting to full-term births, suggesting an important contribution of PTB to the association with LBW. However, even among full-term births the association of G0 DES with G2 LBW remained elevated.
An important concern, given that DES was a prescribed medication, is that of confounding by indication.17 DES was likely often prescribed because the woman (G0) was thought to be at higher risk for adverse pregnancy outcomes like miscarriage and other pregnancy complications that could lead to LBW and PTB.36 If this heightened risk is also passed on to her child (G1), then a non-causal association could exist between G0 DES use and both LBW and PTB in the G2 (and possibly later) generations (Figure 1). Indeed, G0 DES use was associated with increased odds of LBW and earlier gestational age of the nurse herself (G1), although this association was weaker than the association of G0 DES use with G2 LBW and PTB. Importantly, adjusting for G1 LBW, G1 early gestational age and G1 history of miscarriages did not materially change the associations between G0 DES use and G2 LBW and PTB. If the association of G0 DES use with G2 outcomes was the result of confounding by indication, then: (i) it would likely be weaker than the associations with the G1 birth outcomes (although the G1 associations could appear weaker if the G1-level variables were captured with more error than the G2 outcomes) and (ii) adjustment for G1 factors should have substantially reduced the G0-G2 associations, which was not the case. This is because adjusting for a variable (in this case G1 LBW, G1 early gestation age and G1 history of miscarriage) partially adjusts for factors that causally affect it (in this case, possibly indications for DES use in G0; Figure 1).37,38 Even if G1 factors were measured with some error, the fact that robust G0 DES-G1 factor associations are seen means that adjusting for them should reduce the G0 DES-G2 outcome associations, if they are caused by confounding by indication. Thus, our findings add important evidence against confounding by indication, which was not addressed in the two previous studies,14,15 and instead argue for a direct effect of DES exposure on the risk of LBW and PTB in the second generation.
Figure 1.
Structure of confounding by indication. G0, generation 0; G1, generation 1; G2, generation 2; DES, diethylstilbestrol; LBW, low birthweight, PTB: preterm birth)
Our results were robust to further adjustment for potential mediating factors, such as maternal (G1) LBW, early gestational age, smoking and alcohol use during pregnancy with G2. To the extent our variables captured these G1 factors well, this suggests that the mechanism(s) underlying the G0 DES-G2 outcome findings are not related to G0 DES exposure effects on these maternal (G1) factors. One possible biological mechanism for such multigenerational associations is via direct germline cell exposure and epigenetic alterations. In utero exposures can act both on the fetus (G1) and on the primordial germ cells (PGCs) in the developing fetus.39 PGCs are undifferentiated stem cells that will eventually produce eggs and sperms that will become the G2 generation. During pregnancy, PGCs are formed, actively migrate, receive cues from their somatic environment and undergo cell division.40 Whereas PGC DNA could potentially be directly mutated, it has also been proposed that PGCs lack protection from demethylation and thus that germline epigenetic vulnerability is at its highest during pregnancy.41 Exposure to DES could induce epigenetic modifications involving permanent gene expression changes without changing DNA sequences,42 for example via DNA methylation.43 Epigenetically acquired phenotypes may be heritable,44 and evidence from animal models has shown multigenerational influences of endocrine disruptors.45,46 Of particular relevance for our study, Odum et al.47 observed changes in organ weight in F2 rats when the F0 dam was exposed to DES. Whether DES-induced germline perturbations are maintained through global DNA demethylation periods in fetal development, although plausible, remains to be examined.
Several limitations of our study should be noted. We lacked information on the dose of DES exposure, and some exposure misclassification is likely because DES exposure was determined by nurse participants’ (G1) self-report. However, we excluded G1 who indicated uncertain prenatal DES exposure, to focus on more certain exposure, and results using either G1 or G0 reporting of DES exposure were similar. Furthermore, for bias away from the null, the DES misclassification would have to differ by LBW or PTB, but the kappa coefficients for G0 and G1 DES reporting did not differ by G2 LBW or PTB. Some outcome misclassification is also likely given that we relied on maternal (G1) self-report. However, mother’s recall of children’s birthweight and gestational age has been found to be reliable.48 Further, information on the exposure was collected in 1993 (by G1) or 2001 (by G0), before the outcome questions in 2009, and before the only studies to suggest DES exposure might be related to PTB in later generations.14–16 Thus, any exposure or outcome misclassification is likely to be non-differential. The results from mediation analysis should be interpreted with caution due to the possible effect of measurement errors in the mediating factors.49 However, the approach for collecting those data was standard and, in particular for G1 reporting of data from her own birth, self-report has been found to be valid.50 Moreover, as a study of nurses, we assume that the health information obtained from these health care professionals is likely more reliable than it would be from a general population sample. Last, as in any observational studies, we cannot exclude the possibility of residual confounding. However, we accounted for a large list of potential risk factors, including proxies of G0 socioeconomic status and health risk behaviours during pregnancy (e.g. smoking), which was a limitation of the publications from the previous studies.
Our findings add to a growing literature on multigenerational effects of DES exposure in humans, and provides important data suggesting confounding by indication is not accounting for the findings. Although DES is no longer used in pregnancy, this study adds to data suggesting that its legacy can continue to adversely affect those exposed in utero (second generation) and their children (third generation),14 and many chemicals in use today can have similar actions. Even though those actions may be weaker than DES, there are many more such compounds to which people are exposed today. The possible combined effect of many such exposures is not yet clear.
Four out of five pregnant women in high-income countries are prescribed one or more medications during pregnancy, with a lack of information from middle- or low-income countries.51 Given the potential long-range consequences of medication use in pregnancy, as well as in utero exposure to endocrine disruptors and environmental toxicants in general, fetal germline risk should be included as one of the risk assessment categories by regulatory authorities.52 In addition, the observed multigenerational associations further suggest that the biological dimension of inheritance of effects of environmental exposures should be considered as one of the next public health research priorities.
Data availability
The data underlying this article cannot be shared publicly due to the privacy of individuals that participated in the study. The data will be shared on reasonable request to the Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital.
Supplementary data
Supplementary data are available at IJE online.
Funding
This work was supported by NIH grants P30 ES000002 (National Institute of Environmental Health Sciences), P30 ES009089 (National Institute of Environmental Health Sciences), R01HD094725 (National Institute of Child Health and Human Development), and the Escher Fund for Autism. The Nurses’ Health Study II is supported by infrastructure grant U01 176726 from the National Cancer Institute. The funding sources had no role in the design or conduct of the study: collection, management, analysis and interpretation of data; or preparation, review or approval of the manuscript; or decision to submit the manuscript for publication.
Supplementary Material
Acknowledgements
We would like to thank the participants and staff of the Nurses’ Health Study II, Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital, for their substantial contributions.
Author contributions
Concept and design: M.G.W., G.Y. Acquisition, analysis, or interpretation of data: all authors. Drafting of the manuscript: M.G.W., G.Y. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: M.G.W., G.Y. Obtained funding: M.G.W., G.Y. Administrative, technical or material support: M.G.W., G.Y. Supervision: M.G.W., G.Y., A.R., M-A.K., D.W.
Conflict of interest
None declared.
References
- 1. Reed CE, Fenton SE.. Exposure to diethylstilbestrol during sensitive life stages: a legacy of heritable health effects. Birth Defects Res C Embryo Today 2013;99:134–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.IARC Working Group on the Evaluation of Carcinogenic Risk to Humans. Pharmaceuticals. Lyon, France: International Agency for Research on Cancer, 2012.
- 3. McLachlan JA. Commentary: prenatal exposure to diethylstilbestrol (DES): a continuing story. Int J Epidemiol 2006;35:868–70. [DOI] [PubMed] [Google Scholar]
- 4. Smith OW, Smith GV, Hurwitz D.. Increased excretion of pregnanediol in pregnancy from diethylstilbestrol with special reference to the prevention of late pregnancy accidents. Am J Obstet Gynecol 1946;51:411–15. [DOI] [PubMed] [Google Scholar]
- 5. Troisi R, Hatch EE, Titus L.. The diethylstilbestrol legacy: a powerful case against intervention in uncomplicated pregnancy. Pediatrics 2016;138:S42–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Davis ME, Fugo NW.. Does administration of diethylstilbestrol to pregnant women result in increased output of urinary pregnanediol? Proc Soc Exp Biol Med 1948;69:436–38. [DOI] [PubMed] [Google Scholar]
- 7. Dieckmann WJ, Davis ME, Rynkiewicz LM, Pottinger RE.. Does the administration of diethylstilbestrol during pregnancy have therapeutic value? Am J Obstet Gynecol 1953;66:1062–81. [DOI] [PubMed] [Google Scholar]
- 8. Perez KM, Titus-Ernstoff L, Hatch EE. et al. ; National Cancer Institute's DES Follow-up Study Group. Reproductive outcomes in men with prenatal exposure to diethylstilbestrol. Fertil Steril 2005;84:1649–56. [DOI] [PubMed] [Google Scholar]
- 9. Laronda MM, Unno K, Butler LM, Kurita T.. The development of cervical and vaginal adenosis as a result of diethylstilbestrol exposure in utero. Differentiation 2012;84:252–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. 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;179:186–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Skinner MK. Endocrine disruptors in 2015: epigenetic transgenerational inheritance. Nat Rev Endocrinol 2016;12:68–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. McCormick MC. The contribution of low birthweight to infant mortality and childhood morbidity. N Engl J Med 1985;312:82–90. [DOI] [PubMed] [Google Scholar]
- 13. Barker DJ. In utero programming of chronic disease. Clin Sci (Lond) 1998;95:115–28. [PubMed] [Google Scholar]
- 14. Hoover RN, Hyer M, Pfeiffer RM. et al. Adverse health outcomes in women exposed in utero to diethylstilbestrol. N Engl J Med 2011;365:1304–14. [DOI] [PubMed] [Google Scholar]
- 15. Hatch EE, Troisi R, Wise LA. et al. Preterm birth, fetal growth, and age at menarche among women exposed prenatally to diethylstilbestrol (DES). Reprod Toxicol 2011;31:151–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Titus L, Hatch EE, Drake KM. et al. Reproductive and hormone-related outcomes in women whose mothers were exposed in utero to diethylstilbestrol (DES): a report from the US National Cancer Institute DES Third Generation Study. Reprod Toxicol 2019;84:32–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Walker AM. Confounding by indication. Epidemiology 1996;7:335–36. [PubMed] [Google Scholar]
- 18. Bao Y, Bertoia ML, Lenart EB. et al. Origin, methods, and evolution of the three nurses' health studies. Am J Public Health 2016;106:1573–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Roberts AL, Liew Z, Lyall K, Ascherio A, Weisskopf MG.. Association of maternal exposure to childhood abuse with elevated risk for attention deficit hyperactivity disorder in offspring. Am J Epidemiol 2018;187:1896–906. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Boynton-Jarrett R, Rich-Edwards J, Fredman L. et al. Gestational weight gain and daughter's age at menarche. J Womens Health (Larchmt) 2011;20:1193–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. O'Reilly EJ, Mirzaei F, Forman MR, Ascherio A.. Diethylstilbestrol exposure in utero and depression in women. Am J Epidemiol 2010;171:876–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Cutland CL, Lackritz EM, Mallett-Moore T. et al. ; Brighton Collaboration Low Birthweight Working Group. Low birthweight: case definition & guidelines for data collection, analysis, and presentation of maternal immunization safety data. Vaccine 2017;35:6492–500. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Quinn JA, Munoz FM, Gonik B. et al. ; Brighton Collaboration Preterm Birth Working Group. Preterm birth: case definition & guidelines for data collection, analysis, and presentation of immunisation safety data. Vaccine 2016;34:6047–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Magnus P, Bakketeig LS, Skjaerven R.. Correlations of birthweight and gestational age across generations. Ann Hum Biol 1993;20:231–38. [DOI] [PubMed] [Google Scholar]
- 25. Han Z, Mulla S, Beyene J, Liao G, McDonald SD; Knowledge Synthesis Group. Maternal underweight and the risk of preterm birth and low birthweight: a systematic review and meta-analyses. Int J Epidemiol 2011;40:65–101. [DOI] [PubMed] [Google Scholar]
- 26. Shafei AE, Nabih ES, Shehata KA. et al. Prenatal exposure to endocrine disruptors and reprogramming of adipogenesis: an early-life risk factor for childhood obesity. Child Obes 2018;14:18–25. [DOI] [PubMed] [Google Scholar]
- 27. McDonald SD, Han Z, Mulla S, Beyene J; Knowledge Synthesis Group. Overweight and obesity in mothers and risk of preterm birth and low birthweight infants: systematic review and meta-analyses. BMJ 2010;341:c3428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Shah NR, Bracken MB.. A systematic review and meta-analysis of prospective studies on the association between maternal cigarette smoking and preterm delivery. Am J Obstet Gynecol 2000;182:465–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Patra J, Bakker R, Irving H, Jaddoe VW, Malini S, Rehm J.. Dose-response relationship between alcohol consumption before and during pregnancy and the risks of low birthweight, preterm birth and small for gestational age (SGA) - a systematic review and meta-analyses. BJOG 2011;118:1411–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Sibai BM. Preeclampsia as a cause of preterm and late preterm (near-term) births. Semin Perinatol 2006;30:16–19. [DOI] [PubMed] [Google Scholar]
- 31. Kramer MS. Determinants of low birthweight: methodological assessment and meta-analysis. Bull World Health Organ 1987;65:663–737. [PMC free article] [PubMed] [Google Scholar]
- 32. 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;35:862–68. [DOI] [PubMed] [Google Scholar]
- 33. Seaman S, Pavlou M, Copas A.. Review of methods for handling confounding by cluster and informative cluster size in clustered data. Stat Med 2014;33:5371–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. McGee G, Weisskopf MG, Kioumourtzoglou MA, Coull BA, Haneuse S.. Informatively empty clusters with application to multigenerational studies. Biostatistics 2020;21:775–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Vanderweele TJ, Vansteelandt S.. Odds ratios for mediation analysis for a dichotomous outcome. Am J Epidemiol 2010;172:1339–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Jivraj S, Anstie B, Cheong YC, Fairlie FM, Laird SM, Li TC.. Obstetric and neonatal outcome in women with a history of recurrent miscarriage: a cohort study. Hum Reprod 2001;16:102–06. [DOI] [PubMed] [Google Scholar]
- 37. Greenland S. The effect of misclassification in the presence of covariates. Am J Epidemiol 1980;112:564–69. [DOI] [PubMed] [Google Scholar]
- 38. Ogburn EL, Vanderweele TJ.. Bias attenuation results for nondifferentially mismeasured ordinal and coarsened confounders. Biometrika 2013;100:241–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Skinner MK, Manikkam M, Guerrero-Bosagna C.. Epigenetic transgenerational actions of endocrine disruptors. Reprod Toxicol 2011;31:337–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Nikolic A, Volarevic V, Armstrong L, Lako M, Stojkovic M.. Primordial germ cells: current knowledge and perspectives. Stem Cells Int 2016;2016:1741072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Gkountela S, Zhang KX, Shafiq TA. et al. DNA demethylation dynamics in the human prenatal germline. Cell 2015;161:1425–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Skinner MK. Environmental epigenetics. Environ Epigenet 2015;1:dvv002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Alworth LC, Howdeshell KL, Ruhlen RL. et al. Uterine responsiveness to estradiol and DNA methylation are altered by fetal exposure to diethylstilbestrol and methoxychlor in CD-1 mice: effects of low versus high doses. Toxicol Appl Pharmacol 2002;183:10–22. [DOI] [PubMed] [Google Scholar]
- 44. McLachlan JA. Environmental signaling: what embryos and evolution teach us about endocrine disrupting chemicals. Endocr Rev 2001;22:319–41. [DOI] [PubMed] [Google Scholar]
- 45. Manikkam M, Haque MM, Guerrero-Bosagna C, Nilsson EE, Skinner MK.. Pesticide methoxychlor promotes the epigenetic transgenerational inheritance of adult-onset disease through the female germline. PLoS One 2014;9:e102091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Wolstenholme JT, Edwards M, Shetty SR. et al. Gestational exposure to bisphenol a produces transgenerational changes in behaviors and gene expression. Endocrinology 2012;153:3828–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Odum J, Lefevre PA, Tinwell H. et al. Comparison of the developmental and reproductive toxicity of diethylstilbestrol administered to rats in utero, lactationally, preweaning or postweaning. Toxicol Sci 2002;68:147–63. [DOI] [PubMed] [Google Scholar]
- 48. Adegboye AR, Heitmann B.. Accuracy and correlates of maternal recall of birthweight and gestational age. BJOG 2008;115:886–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. VanderWeele TJ, Valeri L, Ogburn EL.. The role of measurement error and misclassification in mediation analysis: mediation and measurement error. Epidemiology 2012;23:561–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Troy LM, Michels KB, Hunter DJ. et al. Self-reported birthweight and history of having been breastfed among younger women: an assessment of validity. Int J Epidemiol 1996;25:122–27. [DOI] [PubMed] [Google Scholar]
- 51. Stock SJ, Norman JE.. Medicines in pregnancy. F1000Res 2019;8:911. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Escher J, Robotti S.. Pregnancy drugs, fetal germline epigenome, and risks for next-generation pathology: a call to action. Environ Mol Mutagen 2019;60:445–54. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article cannot be shared publicly due to the privacy of individuals that participated in the study. The data will be shared on reasonable request to the Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital.