Diethylstilbestrol (DES) is a powerful synthetic estrogen that acts as a transplacental carcinogen in humans. Beginning around 1940 and continuing through the 1960s, DES was given to as many as 2 million pregnant women in the United States under the mistaken belief that it would reduce risk of miscarriage and complications of pregnancy.1 The now well-established adverse effects of in utero DES exposure in women include infertility,2–4 tissue and structural anomalies of the reproductive tract,2,4,5 pregnancy loss and premature delivery,5–8 and a rare vaginal and cervical clear-cell adenocarcinoma,9,10 There may also be increased risks of breast cancer11 and squamous neoplasia of the cervix.12 In prenatally exposed men, outcomes associated with in utero DES exposure include urogenital anomalies13 and possibly increased risk of infertility14 and testicular cancer.15
The mouse model, which replicates many outcomes associated with women's prenatal DES exposure,16–18 has been useful for investigating the mechanisms through which DES exerts carcinogenic and teratogenic effects on the female reproductive tract. Studies in mice suggest the involvement of persistent epigenetic changes (altered gene expression) caused by excessive hormonal stimulation during a critical window in prenatal development.19–21 The epigenetic alterations appear to involve several families of genes including the hormone-responsive genes lactoferrin20,22 and epidermal growth factor,20,23,24 the Hox25,26 and WNT genes27 involved with embryonic reproductive tract development, and c-fos24,28 and c-jun,28 protooncogenes involved in hormone regulation.
Epigenetic changes caused by prenatal exposures may be transmitted to descendant generations of mice.19,29,30 Experiments have shown an elevated occurrence of reproductive tract tumors, particularly uterine cancer31–33 and benign32 and malignant31 ovarian tumors in the female offspring of parents exposed in utero to DES. The possibility that epigenetic alterations, like genetic mutations, may be heritable in humans has profound implications for an evolutionary impact of in utero chemical exposures, including endocrine disruptors.34,35
We report findings from the only study of offspring whose mothers' prenatal DES exposure status is known. Our data arise from 2 sources: (1) mothers' reports of cancer occurring in their sons and daughters and (2) confirmed reports of cancer and benign pathology arising in a subset of adult daughters participating in a study of the third-generation.
Second Generation Cohort (Mothers)
In the early 1990s, the National Cancer Institute (NCI) established the DES Combined Cohort Follow-up Study to assess health effects of DES exposure. Through the NCI effort, previously followed cohorts were reassembled, and new cohorts were assembled for the first time. The combined cohorts included second-generation women; ie, those who had been exposed to DES in utero, and a comparable group of unexposed women. DES exposure, or the lack thereof, was verified by the medical record. The combined cohort study and the second-generation women have been described in detail elsewhere.36 The study was approved by the institutional review boards at each participating center.
In 1994, the first combined cohort study questionnaires were mailed to 6551 second-generation women, including 4459 exposed to DES in utero and 2092 unexposed. Completed questionnaires were returned by 5707 women (88% of the exposed and 84% of the unexposed). Cohort retention was excellent in the 2 successive follow-up mailings in 1997 (91%) and 2001 (93%). The study participants had 8216 offspring, including 4254 sons (2780 exposed; 1474 unexposed), and 3962 daughters (2539 exposed; 1423 unexposed). Each phase of data collection queried second-generation women for reproductive, hormonal, and medical information, and cancers occurring in family members, including their sons and daughters (ie, the third generation).
Third-Generation Cohort (Daughters)
In 2001, the NCI assembled the third-generation cohort, consisting of the adult daughters (≥18 years of age) of DES-exposed and unexposed second-generation women.37 A review of parity records at all 5 study centers identified 763 exposed and 577 unexposed mothers of 966 exposed and 815 unexposed age-eligible daughters. About half of the mothers, 414 (54%) of the exposed and 297 (52%) of the unexposed, gave permission to contact 515 (53%) exposed and 383 (47%) unexposed daughters. The questionnaires, which queried daughters for hormonal and reproductive information, screening histories, and medical events (including gynecologic biopsies, breast biopsies, and cancer diagnoses), were returned by 793 (88%) of 898 daughters, including 463 (90%) exposed and 330 (86%) unexposed. Pathology reports were obtained to verify self-reported diagnoses and biopsies. A study-related review of histology slides confirmed 1 of 2 reported cases of borderline ovarian cancer. Slides were unavailable for the second case, which involved metastatic disease.
Table 1 shows the number of third-generation women who self-reported a biopsy and provided consent to release their medical records, as well as those for whom records were obtained, biopsies were confirmed, and pathology outcomes were other than normal. The confirmation of cancers occurring in the subset of adult daughters participating in the third-generation study was excellent. Of the 8 self-reported cancers, 7 (5 exposed, 2 unexposed) were confirmed by pathology; consent was not obtained to confirm a melanoma reported by an unexposed woman. For other conditions, confirmation of benign biopsies was reasonably good, generally exceeding 60%.
The analyses are based on 2 sources of information: (1) mothers' reports of cancer affecting their sons and daughters (third-generation offspring) and (2) benign and malignant diagnoses self-reported by adult daughters participating in the third-generation study. Cancer diagnoses reported by the mothers were unverified, and due to the low accuracy with which cervical cancers and precursor lesions are reported,12 analyses involving the mothers' data did not include this tumor site. The analyses of third-generation study participants were based on diagnoses that were verified by the medical record. Nonmelanoma skin cancers were not included in any analyses.
Agreement between the mothers' and daughters' cancer reports was assessed using a Kappa coefficient, with 95% confidence interval (CI). One participating daughter was omitted from the analysis because her mother had never responded to a questionnaire mailing. Consequently, the analysis involved 792 third-generation study participants and their mothers. For comparability with the mothers' unverified reports, agreement was based on the daughters' self-reported cancers (whether or not they were verified). All diagnoses of cancer reported by the daughter had occurred before the mothers' most recent questionnaire completion. Nonmelanoma skin cancers and cervical cancers were omitted. The mothers reported 8 daughters with cancer among the third-generation study participants, and the daughters themselves reported 8 cancers. The Kappa coefficient was 0.87 (95% CI = 0.70–1.00) for the agreement between mothers' and daughters' reports. Overall, 7 of 8 (88%) diagnoses reported by the daughter were also reported by the mother, and 7 of 8 (88%) diagnoses reported by the mother were also reported by the daughter. Agreement was perfect among the unexposed. In the exposed, 1 mother did not report a borderline ovarian tumor self-reported by the daughter, and 1 daughter did not report a thyroid cancer that had been reported by her mother.
When case numbers were sufficient, we used Cox proportional hazards models38 with age as the underlying time scale to compute relative risks (RR) and 95% CIs for the association between DES and outcomes of interest. In the analyses of cancers, follow-up began at the offspring's birth and continued until cancer diagnosis or the most recent survey response, whichever occurred first. When possible, RRs were adjusted for the potential confounders of study center (4 levels) and birth year (tertiles: <1974; 1974–1978; >1978). In analyses where adjustment for study center and birth year was possible, the adjusted RR was similar to the unadjusted RR and to the RR adjusted for birth year alone. Clinical, mammographic, and gynecologic screening histories were similar for the exposed and unexposed women participating in the third-generation study.37 We tested for departure from the proportional hazards assumption using the likelihood ratio test to compare models with and without interaction terms between exposure and age (<20 vs. ≥20).
External comparisons were conducted using the standardized incidence ratio (SIR) to compare observed and expected numbers of cancer cases overall and at specific sites. Expected case numbers were calculated by applying age, calendar year, and (when appropriate) sex-specific cancer rates in whites from the Surveillance, Epidemiology, and End Results Program39 to the person-year distribution of the study population.
Analyses Based on Mothers' Reports
The age distribution was comparable for the third-generation sons and daughters, and the combined frequencies are shown in Table 2. Comparing the exposed to the unexposed, the RR for cancer in the sons and daughters combined (adjusted for birth year and study center) was 1.5 (95% CI = 0.68–3.2). The RR (adjusted for birth year) for cancer in the sons (11 exposed and 2 unexposed cases) was 4.2 (0.92–19). The RR (adjusted for birth year) for cancer in the daughters (10 exposed, 10 unexposed) was 0.70 (0.29–1.7). External comparisons with the general population offered little evidence of increased cancer risk in the exposed sons and daughters combined, in the exposed daughters, or in the exposed sons (Table 3). The external comparisons also suggested a deficit of cancer diagnoses in the unexposed sons, explaining the nonsignificant elevated RR arising from the internal comparison.
Two prenatally DES-exposed women reported a daughter affected by ovarian cancer; the corresponding SIR was 5.3 (95% CI = 1.3–21) based on 0.38 cases expected. One of the reported ovarian cancers was a juvenile granulosa tumor (diagnosis at age 7 years) occurring in a daughter who was age-ineligible for the third-generation study. The other tumor (diagnosis at age 20 years) affected a daughter participating in the third-generation study, who also reported the cancer. None of the mothers, regardless of exposure status, reported a breast or vaginal/vulvar cancer affecting their daughters. There was no indication of an excess of testicular cancer in the sons of exposed mothers (data not shown). The unexposed mothers' reports suggested an excess of leukemia in daughters; the SIR was 4.8 (2.0–11) based on 5 observed and 1 expected case.
Analyses Based on Daughters' Confirmed Reports
The age distribution of third-generation study participants is shown in Table 2. Of the 8 cancers self-reported by the daughters, 7 (5 exposed, 2 unexposed) were confirmed by the medical record. The RR (adjusted for birth year) for the relationship between the mothers' prenatal DES exposure and confirmed cancer in the daughter was 2.0 (0.39–11). After omitting 2 confirmed ovarian cancers occurring in the exposed, the RR was 1.2 (0.20–7.2). Using external comparisons, the SIR for confirmed cancer of any type in the exposed daughters was 2.1 (0.89–5.1) based on 5 observed and 2.35 expected cases (Table 4). In the unexposed, the SIR was 0.98 (0.24–3.9) based on 2 observed and 2.05 expected cases.
The 2 confirmed ovarian diagnoses included a borderline mucinous tumor (diagnosis at age 22) and a borderline serous ovarian tumor with lymph node metastases (diagnosis at age 20) (this diagnosis was also reported by the mother). The SIR for confirmed ovarian cancer reported by the daughters of exposed women was 15 (3.7–59) based on 0.14 expected cases. None of the mothers who denied consent to contact their adult daughters for inclusion in the third-generation study had ever reported a daughter affected by ovarian cancer in follow-up questionnaires asking about cancers in family members. When the SIR was conservatively calculated based on all adult exposed daughters, whether or not they participated in the third-generation study (0.30 expected cases), the excess of ovarian cancer remained, with a SIR of 6.6 (1.7–26).
Both of the confirmed cancers reported by the unexposed daughters were acute lymphoblastic leukemias (ages 2 and 16 at diagnoses). In the unexposed, the SIR for leukemia was 7.3 (1.8–29) based on 0.27 expected cases.
Adjusted for birth year and study center, the RR was 1.5 (0.46–4.9) for the association between the mothers' DES exposure and a benign diagnosis of the uterus/endometrium, ovary, or fallopian tube combined (Table 5). The RR was 5.0 (0.48–52) for benign pathology of the endometrium/uterus specifically, and 0.68 (0.15–3.1) for benign pathology of the ovary. Two exposed daughters and none of the unexposed had a diagnosis of hydatid cysts of the fallopian tube.
Adjusted for birth year and study center, the RR was 1.5 (0.69–3.1) for the mothers' prenatal DES exposure in relation to any cervical dysplasia (mild/moderate/severe) in the daughter, and 0.93 (0.29–2.9) for moderate/severe dysplasia (Table 5). One vaginal biopsy in an exposed daughter produced normal findings. One exposed daughter had a diagnosis of squamous dysplasia of the vulva, and a second exposed daughter had a dysplastic mole of the vulva. After adjustment for birth year and study center, the RR was 1.6 (0.78–3.4) for the combined cervical and vulvar diagnoses (24 exposed, 16 unexposed) (Table 5).
All benign breast diagnoses were fibroadenomas; the RR (adjusted for birth year) for DES in relation to this outcome was 0.57 (0.09–3.6) (Table 5).
Our findings, based on the only study in offspring whose mothers' prenatal DES exposure status is documented by the medical record, do not indicate an overall excess of cancers in the sons or daughters of women who were exposed to DES in utero. However, due to the small number of cases that have occurred in this young population, statistical power is inadequate to rule out adverse outcomes with certainty.
The number of ovarian cancer cases affecting the daughters of prenatally exposed women was greater than expected, based on a total of 3 cases. These included an unverified case of juvenile granulosa ovarian tumor reported by the mother of a daughter who was age-ineligible for the third-generation study, and 2 borderline tumors reported by third-generation study participants and confirmed by the medical record. An excess of ovarian tumors is seen in the descendants of prenatally DES-exposed mice,31,32 but previous evidence in humans is limited to a single case report of a small-cell ovarian cancer occurring in an adolescent girl whose mother was exposed in utero to DES.40
Our finding may be due to chance or possibly to bias, and should be considered preliminary. Although 1 daughter's ovarian cancer was never reported by the mother, the 2 reported diagnoses occurred prior to the initiation of the NCI Combined Cohort Study and may have influenced the mothers' decision to participate. Enrollment was excellent (88%) among exposed mothers, however, which would limit the impact of participation bias on findings arising from their reports. Because only half of adult daughters were approachable for study participation, results based on their reports would be susceptible to bias. Even so, none of the mothers ever reported an ovarian cancer in an adult daughter who did not participate in the third-generation study. Also, the SIR for confirmed ovarian cancer arising from the daughters' reports remained elevated when the expected number of cases was based on all exposed adult daughters, whether or not they participated in the third-generation study. We cannot, of course, exclude the possibility that the general population is an inappropriate comparison group for our study population. Also, it is possible our findings reflect more accurate reporting of ovarian cancers by exposed than unexposed mothers (or daughters), although this seems unlikely. Although prenatal exposure to DES is associated with infertility in women, none of the mothers of the 3 affected daughters reported using fertility drugs. Surveillance or detection bias also seems an improbable explanation for our finding. The daughters of prenatally exposed or unexposed mothers were similar with regard to health screening, including the frequency of recent general physical examinations, gynecologic examinations, colposcopy, and Pap tests.37 The frequency of reproductive tract biopsies was also similar for the exposed and unexposed third-generation study participants. Excluding the ovarian cancer diagnoses, fewer exposed (0.6%) than unexposed (1.2%) women specifically underwent a biopsy of the ovary.
The number of cases of leukemia in the unexposed daughters (but not in the sons) was also greater than expected. Lower birth weight may be associated with decreased leukemia risk,41 and in our study, the SIRs based on reports from the mothers (SIR = 0.58; CI = 0.08–4.1) and daughters (0; 0–5.2) suggested a deficit of leukemia in the exposed daughters, who tend to be of lower birth weight. This would not, however, explain the elevated SIR in the unexposed daughters. Four of the 5 leukemia diagnoses reported by unexposed mothers occurred before the initiation of the combined cohort study, and may have influenced their decision to participate. A high proportion (84%) of unexposed mothers enrolled in the combined cohort study, which would reduce the influence of participation bias. As is also true of the results for ovarian cancer, the general population may not be an appropriate comparison group for our study, or the finding may be due to chance.
We did not find in the sons of exposed women an excess of testicular cancer, a tumor that may be more common in prenatally DES-exposed men.15 Similarly, none of the daughters was affected by vaginal/cervical adenocarcinoma, a rare tumor which occurs in striking excess in women exposed to DES in utero.9,10 A clinical study of third generation daughters did not find gynecologic anomalies, but the participants were few in number (n = 28)42; also, irregularities of the uterus or fallopian tube, which affect the prenatally exposed women, would not be evident upon physical examination. None of the third-generation women had a diagnosis of breast cancer, a tumor which may occur more often in DES-exposed than unexposed women.11 At present, however, the study population is too young for a meaningful assessment of these outcomes.
In this study, we found little evidence of an increased risk of cervical dysplasia, which may affect the DES-exposed second generation women.12 A possible elevation of benign uterine conditions was based on small numbers and inconclusive. Our data did not support an increased risk of benign ovarian pathology, and findings have been inconsistent in studies of second-generation women.43,44 There was no association between the mothers' prenatal DES exposure and the daughters' risk of benign breast disease.
In summary, although statistical power was limited, the current data do not support an excess of overall cancer in the sons or daughters of women who were prenatally exposed to DES, or an excess of benign diagnoses in the exposed daughters. We observed a total of 3 cases of ovarian cancer in the daughters of exposed women, which was more than expected. This finding is preliminary and may be due to chance or possibly bias. Even so, when considered in conjunction with the results of animal studies, our observation reinforces the need for continued follow-up of the third-generation women.
We are grateful to the second and third generation women whose participation in the NCI DES Follow-up Study made this research possible. We thank the coordinators and programmers at each study center for their work on this study, and thank Jorge Gonzales at Dartmouth-Hitchcock Medical Center for his review of the ovarian cancer pathology.
1. Noller KL. In utero exposure to diethylstilbestrol. In: Jones HW III, Wentz AC, Burnett LS, eds. Novak's Textbook of Gynecology
. 11th ed. Baltimore: Williams and Wilkins; 1988:623–642.
2. Senekjian EK, Potkul RK, Frey K, et al. Infertility among daughters either exposed or not exposed to diethylstilbestrol Am J Obstet Gynecol
3. Palmer JR, Hatch EE, Rao Sowmya R, et al. Infertility among women exposed prenatally to diethylstilbestrol. Am J Epidemiol
4. Bibbo M, Gill WB, Azizi F, et al. Follow-up of male and female offspring of DES-exposed mothers. Obstet Gynecol
5. Kaufman RH, Adam E, Binder GL, et al. Upper genital tract changes and pregnancy outcome in offspring exposed in utero to diethylstilbestrol. Am J Obstet Gynecol
6. Barnes AB, Colton T, Gundersen J, et al. Fertility and outcome of pregnancy in women exposed in utero to diethylstilbestrol. N Engl J Med
7. Kaufman RH, Adam E, Hatch EE, et al. Continued follow-up of pregnancy outcomes in diethylstilbestrol-exposed offspring. Obstet Gynecol
8. Goldberg J, Falcone T. Effect of diethylstilbestrol on reproductive function. Fertil Steril
9. 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
10. Hill EC. Clear cell carcinoma of the cervix and vagina in young women. Am J Obst Gynecol
11. Palmer JR, Wise LA, Hatch EE, et al. Prenatal diethylstilbestrol exposure and risk of breast cancer. Cancer Epidemiol Bio Prev
12. Hatch EE, Herbst AL, Hoover RN, et al. Incidence of squamous neoplasia of the cervix and vagina in DES-exposed daughters. Ann Epidemiol
13. Wilcox AJ, Baird DD, Weinberg CR, et al. Fertility in men exposed prenatally to diethylstilbestrol. N Engl J Med
14. Perez K, Titus-Ernstoff L, Hatch EE, et al. Reproductive outcomes in men with prenatal exposure to DES. Fertil Steril
15. Strohsnitter WC, Noller KL, Hoover RN, et al. Cancer risk in men exposed in utero to diethylstilbestrol. J Natl Cancer Inst
16. McLachlan JA, Newbold RR, Bullock BC. Long-term effects on the female mouse genital tract associated with prenatal exposure to diethylstilbestrol. Cancer Res
17. McLachlan JA, Newbold RR, Shah HC, et al. Reduced fertility in female mice exposed transplacentally to diethylstilbestrol (DES). Fertil Steril
18. Newbold RR, McLachlan JA. Vaginal adenosis and adenocarcinoma in mice exposed prenatally or neonatally to diethylstilbestrol. Cancer Res
19. Newbold RR, Padilla-Banks, Jefferson WN. Adverse effects of the model environmental estrogen diethylstilbestrol are transmitted to subsequent generations. Endocrinol
20. Gray Nelson KG, Sakai Y, Eitzman B, et al. Exposure to diethylstilbestrol during a critical development period of the mouse reproductive tract leads to persistent induction of two estrogen-related genes. Cell Growth Differentiation
21. McLachlan JA, Burow M, Chiang T-C, et al. Gene imprinting in developmental toxicology: a possible interface between physiology and pathology. Toxicol Ltrs
22. Li JJ, Hansman R, Newbold R, et al. Developmental exposure to diethylstilbestrol elicits demethylation of estrogen-responsive lactoferrin gene in mouse uterus. Cancer Res
23. Okada A, Sato T, Ohta Y, et al. Effect of diethylstilbestrol on cell proliferation and expression of epidermal growth factor in the developing female rat reproductive tract. J Endocrinol
24. Falck L, Forsberg J-G. Immunohistochemical studies on the expression and estrogen dependency of EGF and its receptor and C-fos proto-oncogene in the uterus and vagina of normal and neonatally estrogen-treated mice. Anatom Rec
25. Ma L, Benson GV, Lim H, et al. Abdominal B (AbdB) Hoxa genes: regulation in adult uterus by estrogen and progesterone and repression in mullerian duct by the synthetic estrogen diethylstilbestrol (DES). Devel Biol
26. Block K, Kardana A, Igarashi P, et al. In utero diethylstilbestrol (DES) exposure alters Hox gene expression in the developing mullerian system. FASEB J
27. Mericskay M, Carta L, Sassoon D. Diethylstilbestrol exposure in utero: a paradigm for mechanisms leading to adult disease. Birth Defects Res
28. Yamashita S, Takayanagi A, Shimizu N. Effects of neonatal diethylstilbestrol exposure on c-fos and c-jun protooncogene expression in the mouse uterus. Histol Histopathol
29. Holliday R. The possibility of epigenetic transmission of defects induced by teratogens. Muta Res
30. Holliday R. The inheritance of epigenetic effects. Science
31. Walker BE. Tumors in female offspring of mice exposed prenatally to diethylstilbestrol. J Nat Cancer Inst
32. Turosov VS, Trukhanova LS, Parfenov YD, et al. Occurrence of tumors in descendants of CBA male mice prenatally treated with diethylstilbestrol. Int J Cancer
33. Newbold RR, Hanson RB, Jefferson WN, et al. Increased tumors but uncompromised fertility in the female descendants of mice exposed developmentally to diethylstilbestrol. Carcinogenesis
34. Guerrero-Bosagno C, Sabat Om, Valladares L. Environmental signaling and evolutionary change: can exposure of pregnant mammals to environmental estrogens lead to epigenetically induced evolutionary changes in embryos? Evol Devel
35. Jablonka E, Lamb M. Epigenetic Inheritance and Evolution
. Oxford: Oxford University Press; 1995.
36. Hatch EE, Palmer JR, Titus-Ernstoff L, et al. Cancer risk after long term follow-up of women exposed to diethylstilbestrol (DES) in utero. J Am Med Assoc
37. Titus-Ernstoff L, Troisi R, Hatch EE, et al. Menstrual and reproductive characteristics of women whose mothers were exposed to DES in utero. Int J Epidemiol
38. Cox DR, Oakes D. Analysis of Survival Data
. London, UK: Chapman Hall; 1984.
39. Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov
) SEER*Stat Database: Incidence - SEER 9 Regs Public-Use, Nov 2003 Sub (1973–2001), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2004., based on the November 2003 submission.
40. Blatt J, Van Le L, Weiner T, et al. Ovarian carcinoma in an adolescent with transgenerational exposure to diethylstilbestrol. J Ped Hem/Oncol
41. Hjalgrim LL, Westergaard T, Rostgaard K, et al. Birth weight as a risk factor for childhood leukemia: a meta-analysis of 18 epidemiologic studies. Am J Epidemiol
42. Kaufman RH, Adam E. Findings in female offspring of women exposed in utero to diethylstilbestrol. Obstet Gynecol
43. Haney AF, Newbold RR, Fetter BF, et al. Paraovarian cysts associated with prenatal diethylstilbestrol exposure. Am J Pathol
44. Wise LA, Palmer JR, Rowlings K, et al. Risk of benign gynecological tumors in relation to prenatal diethylstilbestrol exposure. Obstet Gynecol