Prenatal Diethylstilbestrol Exposure and Cancer Risk in Males

This article is featured in Highlights of This Issue, p. 1767

Diethylstilbestrol (DES) is a non-steroidal estrogen analog that was prescribed to pregnant women predominantly between 1940 and the 1960s for numerous pregnancy complications. In 1971, Herbst and colleagues (1) reported an increased risk of clear cell adenocarcinoma of the cervix and vagina among women prenatally DES-exposed. The cancer experience of men prenatally DES-exposed, however, was not characterized until 2001, when an increase in testicular cancer among men who were prenatally DES-exposed was reported (2). At that time, however, the cohort members (mean age = 41.3 years) had not reached an age at which most cancers normally develop. Consequently, the influence of prenatal DES exposure on cancer risk among middle-aged men has not been well characterized.

The influence of prenatal DES exposure on prostate cancer risk is of particular interest. In an early study of autopsied male perinatal deaths, there was an increased percentage of structural abnormalities in prostate tissue among those who were prenatally DES exposed relative to those who were unexposed (3). Age, race, and positive family prostate cancer history notwithstanding, few adult prostate cancer risk factors have been consistently identified with regard to prostate cancer incidence (4). Body mass index (BMI), adult height, current or recent smoking history and dietary factors have been associated with increased risk of advanced or fatal prostate tumor development (5).

Possibly, study of early life exposures may shed insight into prostate cancer etiology that is currently lacking. Early life experiences contribute to attained adult height, which has been observed to have a positive association with prostate cancer risk (6, 7). Other perinatal factors such as birth weight, preeclampsia during the index pregnancy, premature birth, and possibly perinatal jaundice have also been associated with adult prostate cancer risk (8, 9), all of which may be surrogates for pregnancy estrogen exposure (10). The associations between birth weight and other early risk factors with prostate cancer risk have been inconsistent but could be attributed to small study sizes (10, 11). Approximately 20 years after our initial report, we present the results of an investigation that continues follow-up of a male cohort with prenatal DES exposure and its effect on overall cancer risk and site-specific cancers.

The National Cancer Institute (NCI) DES Combined Cohorts Follow-up Study includes 4,101 men who were (n = 2001) or were not (n = 2,100) prenatally exposed to DES (previously detailed; ref. 2). Briefly, in 1994, the NCI combined cohorts of men available from four independent sources: (i) The Dieckmann cohort included sons of women who were enrolled in a clinical trial of the efficacy of DES on preventing miscarriages (12); (ii) The Mayo Clinic cohort consisted of sons of women who were or were not prescribed DES at the Mayo Clinic between 1940 and 1960 and were identified via medical record review (13); (iii) The Horne cohort included sons of women who were attended by a physician who prescribed DES for infertility; and (iv) The Women's Health Study cohort included sons of women who did or did not take DES during pregnancy and were enrolled in a study of DES' effects on breast cancer risk (14).

Approvals for the study were obtained from the Institutional Review Boards at the respective study sites and the NCI. Participants indicated their informed consent by completion of a questionnaire or telephone interview and by written consent for medical record retrieval.

Follow-up information was available for the Dieckmann, Mayo, and Horne cohorts previous to the combining of the cohorts by NCI. Systematic follow-up of the combined cohorts began in 1994 with a mailed questionnaire to all cohort members. Subsequent questionnaires were administered in 1997, 2001, 2006, 2011, and 2016. Among the identified men, there were 3,422 (1,694 DES-exposed and 1,728 unexposed) who provided some health information before or after 1994 (13, 15). There were 2,924 men (1,448 DES-exposed and 1,476 unexposed) who provided follow-up in 1994 and after. Furthermore, there were 1,978 (990 DES-exposed and 998 unexposed) who provided follow-up though the 2016 follow-up (Fig. 1 and Table 1). The attrition and reason for study withdrawal between the exposed and unexposed groups was similar (Table 1).

A total of 393 (187 exposed and 206 unexposed) cancer cases were identified during the follow-up. Among these, there were 32 (15 exposed and 17 unexposed) primary tumors that occurred in men who were previously diagnosed with another primary tumor. These multiple primary tumors comprised 8.1% of the total tumors identified and 8.0% and 8.3% of the tumors identified among the exposed and unexposed, respectively.

Cancer cases were primarily identified by questions regarding cancer diagnosis, including primary site and year of diagnosis. We sought consent to obtain pathology reports and medical records to confirm cancers for all anatomical sites that were self-reported on the 1994–2011 questionnaires. Pathology and medical records were obtained for 85% of self-reported cases, and confirmation of cancer diagnoses self-reported between 1994 and 2011 by review of these records was high (95%). Similarly, medical records were provided for 74 (77%) of the 96 reported prostate cancer cases. Grade and stage information was abstracted from these reports. Among the 74 reported prostate cancer cases reported in this period for whom records were obtained, there was only one that disconfirmed a participant's self-report and this case was not included in the current analysis. Consequently, self-reports of cancer diagnoses after 2011 were not verified with medical record review but were included in the analysis. Also, stage and grade information were unavailable for prostate cancer cases identified during the 2016 follow-up.

Cancer cases were also identified via passive follow-up. For cohorts originating in Massachusetts, Minnesota, and New Hampshire, cancer or tumor registries (for approximately 88% of participants) were periodically searched from 2008 to 2016 to obtain additional information on reported cases (including tumor stage, when available; tumor grade was consistently unavailable) and to identify new cases. Cases were also identified from death certificates and ICD coding of underlying and contributing causes of death from the National Death Index (NDI) Plus. The NDI was routinely searched for participants who were lost to follow-up or had an unknown cause of death.

For 72 of the 92 prostate cancer cases reported between 1994 and 2011, prostate cancer stage and grade information was available from the medical records obtained. In addition, cancer stage was available for 4 cases identified via cancer registry searches. The TNM staging system as defined by the 8th edition of the American Joint Committee on Cancer (AJCC) was used to classify prostate cancer cases (16). Cases were also defined as cT1c (early-stage) per AJCC when they were not clinically palpable and were detected by elevated PSA levels and subsequent positive needle biopsy.

Prognostic groups were categorized as defined by the AJCC system. The group with the most favorable prognosis, Group I (N = 39), consisted of cases whose tumor was confined to the prostatic capsule (cT1c, or pT2), had a Gleason Score of six or less, and a serum prostate-specific antigen (PSA) level less than 10 ng/mL at diagnosis (Supplementary Table S1). Cases that were in the second prognostic group or higher, Group II+ (N = 34), presented with a tumor that was not confined to the prostatic capsule, had a Gleason Score of seven or higher, or had a PSA level of 10 ng/mL or higher at diagnosis (Supplementary Table S1). Three cases identified through cancer registries lacked a Gleason score or diagnostic PSA level and could not be assigned a prognosis group (Supplementary Table S1). Cases having extracapsular involvement, stageT3 or higher or a PSA level of 20 ng/mL or greater at diagnosis, were classified as having a poor prognosis, group III+ (N = 15). Men with a prostate tumor that either invaded tissue outside of the prostate region or had a Gleason score of 8 or higher were observed to have a poor survival probability (17). There were 5 (1 exposed, 4 unexposed) such cases in this study. All prognostic grouping was conducted by one of the investigators (W.C. Strohsnitter) upon review of the available documentation without consideration of DES exposure status.

For all combined cohort participants, prenatal exposure to DES was documented by the medical record or a physician's note. Medical record data for prescribed dose were available for the Mayo Clinic cohort. Because, the date when DES administration was stopped was inconsistently recorded, total dose over the course of the pregnancy was unavailable for most of this cohort. Consequently, we imputed dose by fitting a linear model using maximum prescribed dose, and week of initial dose among the 181 men for whom complete dose information was available in the medical record. This model indicated that maximum DES dose and the gestational week that DES was started were reasonable predictors of total dose (R² = 0.63), and maximum DES dosage was used as a surrogate for total dose in the DES-exposed men grouped into quartiles. For the Dieckmann cohort, DES dose was prescribed according to the Smith and Smith regimen where cumulative doses were as high as 12 g administered through the length of the pregnancy (12).

Smoking status, alcohol use, body weight, height, family history of cancer, and routine medical screening (including PSA screening) were updated on subsequent questionnaires (in 2006 and 2016 for smoking, in 2016 for alcohol, and in 2006, 2011, and 2016 for body weight). BMI [weight (kg)/height (m)²] was calculated. Participants were categorized as being obese if their BMI was equal to or exceeded 30.

Person-time accrual began on January 1, 1978, when previous health data were consistently available for men in the Dieckmann, the Horne, and Mayo Clinic cohorts. Men whose mothers participated in the Women's Health Study began person-time accrual on January 1, 1994, the approximate date when that cohort was assembled. The age-adjusted relative effect of prenatal DES exposure on overall and site-specific cancer incidence was estimated by comparing exposure-specific rates with the national age-specific rates for white men obtained from the Surveillance, Epidemiology, and End Results (SEER) program (18). Ninety-five percent confidence intervals (CI) for these estimates were calculated using the Wald method (19). Hazard ratios (HR) and their associated 95% CIs derived from Cox proportional hazard modeling (20) compared cancer rates in the exposed and unexposed adjusting for age, birth year, and study cohort. Follow-up time accrual ended at the date of cancer diagnosis, the date of the participant's completion of their last questionnaire, or date of death as determined by review of the NDI record or death certificate, whichever was earliest. For the determination of DES' relative effect on site-specific cancers, follow-up time accrued after the first cancer diagnosis to allow for subsequent development of another primary cancer. A total of 92,588 person years (47,332 DES-exposed, 45,256 unexposed) accrued among the participants included in these analyses. An HR was estimated considering only the initial primary tumor and person-time accrual ended at the initial tumor diagnosis.

The relative rate (RR) of prenatal DES exposure was estimated for overall prostate cancer as well as for stage-specific disease using Poisson regression modeling (21). There was a total of 91,251 person years (46,911 DES-exposed, 44,330 unexposed) accrued among the participants included in this analysis. Because age, race, and family prostate cancer history are the only established risk factors for prostate cancer (4), all prostate cancer risk models only adjust for age. This is because the cohorts are predominantly Caucasian and family prostate cancer history is equally distributed between the exposed and the unexposed (Table 2). DES dose response on prostate cancer was estimated among participants from two of the subcohorts; the Mayo Clinic and the Dieckmann trial.

Five separate analyses were conducted to estimate DES' effect on stage-specific prostate cancer. They include those cases with the most favorable prognosis (Group I), those with a less favorable prognosis (Group II or higher), those with a poor prognosis (Group III or higher), those that were only evident by screen-detection (cT1c only), and those that were observed by Hurwitz and colleagues (17) to be concordant with less than a 10-year survival.

Some animal studies suggest that prenatal estrogen exposure predisposes them to increased prostate cancer due to later estrogenic insults, such as those associated with age and obesity (22–26). Consequently, the influence of obesity on the association between prenatal DES exposure and prostate cancer rate was explored. This was evaluated using height and weight information requested throughout the follow-up. Participants who did not provide complete height and weight information for the duration of their follow-up were excluded from this analysis. Obesity status was treated as time-dependent. Before and up to their first weight and height report, participants accrued obesity-specific person-time according to their first provided height and weight information. Subsequently, they accrued obesity-specific person-time according to their next reported weight and height information. The relative excess rate due to interaction (RERI) was used to evaluate the departure from additivity of the independent effect estimates (19, 27). CIs were constructed around the RERI using the standard delta methods (28).

The percentages of exposed and unexposed men did not differ by family prostate cancer history, vasectomy, and smoking or alcohol use history. DES-exposed men appeared to undergo PSA screening slightly more frequently then unexposed men with 42.2% of DES-exposed men reporting screening every 2 years or more in 2006 compared with 39.3% of unexposed men (Table 2). The mean age of any participant who returned a questionnaire in 1994 or after, when their person-time accrual ended, was 58.9. Among all men who completed a questionnaire in 2016, their mean age at questionnaire completion was 62.0 (Table 2).

Among DES-exposed men, 187 total cancers were identified, including 65 prostate cancer diagnoses. Among unexposed men, 206 total cancers were identified, including 72 prostate cancer diagnoses. Overall cancer rates were slightly higher for both the exposed and unexposed men in the cohort compared with the general population. The standardized incidence ratios (SIR) were 1.08 (95% CI, 0.93–1.25) and 1.19 (95% CI, 1.03–1.37) for the exposed and unexposed, respectively (Table 3). The overall cancer incidence among the DES-exposed men was not increased relative to those unexposed (age-adjusted HR, 0.94; 95% CI, 0.77–1.15; Table 4). The estimate of DES' effect on overall cancer risk was virtually unchanged when excluding subsequent primary tumors (HR, 0.94; 95% CI, 0.76–1.16).

Compared with nationwide prostate cancer rates, the SIRs for prostate cancer were elevated among both the exposed 1.28 (95% CI, 0.99–1.63) and unexposed 1.41 (95% CI, 1.1–1.77; Table 3). Using the internal comparison group, the age-adjusted prostate cancer rates were similar between the exposed and unexposed groups (HR, 0.95; 95% CI, 0.68–1.33; Table 4). We also observed a decrease in urinary system (kidney and bladder) cancer among the DES-exposed (n = 11) relative to the unexposed (n = 23; age-adjusted HR, 0.48; 95% CI, 0.23–1.00; Table 4). The estimate of DES' effect on testicular cancer rates was imprecise because there were only six exposed cases (adjusted HR, 1.74; 95% CI, 0.43–7.07; Table 4).

The age-adjusted RR for prostate cancer in the men from the Mayo Clinic cohort in the lowest DES dose quartile compared with the unexposed men was 1.42 (95% CI, 0.76–2.66; Table 5). There was also no increase in prostate cancer with DES exposure among men in the Dieckmann cohort (age-adjusted RR, 0.77; 95% CI, 0.32–1.87), where high doses were administered during the index pregnancy (12).

There was no evidence of an association between prenatal DES exposure and prostate cancer prognosis. The age-adjusted RRs for DES and prostate cancer risk among cases with the most favorable prognosis (Group I) was 1.05 (95% CI, 0.56–2.0), and 1.0 (95% CI, 0.5–2.0) among cases with a less favorable prognosis (Group II and higher), whereas the RRs for advanced prostate cancer (Group III or higher) and screen-detected prostate cancer were 0.67 (95% CI, 0.24–1.9) and 1.58 (95% CI, 0.83–3.0), respectively (Supplementary Table S2). The age-adjusted estimate for the relative effect of DES on aggressive prostate cancer as observed by Hurwitz and colleagues (17) was 0.25 (95%CI, 0.03–2.23). This estimate, however, is based on 5 cases and is imprecise.

The age-adjusted RR for DES and prostate cancer incidence among those with a BMI of at least 30 was 1.15 (95% CI, 0.59–2.24) and among those with a BMI less than 30, 0.79 (95% CI, 0.51–1.22). The RERI indicating a departure from additivity was 0.34 (95% CI, −0.36–1.03; Supplementary Table S3). Similar to the estimates for DES and prostate cancer, overall there was no apparent independent effect of obesity on prostate cancer rates. Among the DES-unexposed, the association between obesity and prostate cancer was RR, 0.86 (95% CI, 0.49–1.51).

There was no increase in overall cancer risk among the prenatally DES-exposed middle-aged men in this cohort relative to those unexposed. Although previous animal and observational human studies (6–9, 22, 29) lend evidence to the possibility that prostate cancer etiology has an early life component, no increase in overall prostate cancer among the DES-exposed men relative to those unexposed was observed. There was also no observed appreciable increase in cancer at any other anatomical site among DES-exposed men compared with the unexposed. We did, however, observe a decrease in urinary system (bladder and kidney) tumors in exposed men in comparison with those unexposed.

Laboratory studies have suggested that prostate cancer etiology has a hormonal component (30) that in turn may be related to obesity and aging (24, 25). Obesity and aging are associated with increased P450 aromatase conversion of testosterone to estradiol in adipose cells (22, 25, 26). It is speculated from the animal model that the serum ratio of estradiol to testosterone increases with age might also play a role in prostate cancer etiology (30).

Several human studies, however, have shown inverse associations between postnatal estradiol (E₂) and testosterone (T) ratio in archived blood and subsequent prostate cancer risk. In a case–control study of circulating sex hormone levels and prostate cancer risk (31) investigators observed a lower incidence of aggressive prostate cancer among those with increased molar ratios of estradiol to testosterone in blood drawn on average 2.9 years before diagnosis. A similar association was observed when the maximum time between blood draw and cancer diagnosis was 10 (32) and 41 years (33). Platz and colleagues (34), however, observed a positive association between circulating blood levels of estradiol to testosterone ratio and aggressive prostate cancer risk where blood samples drawn a maximum of five years before prostate cancer diagnosis.

Increased total estrogen and estradiol levels were observed in prenatally DES-exposed postmenopausal women relative to those unexposed (35). There have been, however, no known studies investigating sex hormone levels in prenatally DES-exposed adult men. Consequently, it is difficult to speculate whether prenatal DES exposure had an influence on adult male hormone levels. Furthermore, there were too few cases of men with aggressive prostate cancer identified during this follow-up to estimate with reasonable precision what effect, if any, DES exposure had on the risk of prostate cancer with poor prognosis. As a result, the current study does not lend evidence supporting findings of other human studies observing a decreased aggressive prostate cancer risk associated with elevated adult estrogen levels.

Animal studies also have suggested that prenatal xenoestrogen exposure coupled with further estrogenic insults later in life may influence prostate tumorigenesis. Early introduction of estradiol, DES or other estrogen-like compounds were observed in studies of laboratory animals to disrupt and impede epithelial cell differentiation and branching (36). These estrogenic insults early in life followed by estradiol administration to adult rats resulted in an increased incidence of prostate lesions (22). Obesity is associated with an increase in serum estradiol to testosterone ratio, a speculated prostate cancer risk factor (25, 26). Possibly any effect of prenatal DES exposure on prostate cancer rates then require another later exposure such as obesity. There was little evidence, however, of an effect of prenatal DES exposure on prostate cancer rates in those with a high BMI and the RERI did not indicate an appreciable departure from additivity.

At least one animal study offered the possibility that prenatal exposure to low, but not high, DES doses factor into prostate cancer development. Mice prenatally exposed to low DES doses had higher adult prostate weights than those who were exposed to higher DES doses (37). We, however, did not observe an appreciable increase in prostate cancer rates among the men who were prenatally exposed to low or high DES doses.

In 2001, we reported a possible increase in testicular cancer among prenatally DES-exposed men (2). Only one more testicular cancer case (unexposed) developed since that report and the estimate reported in the current study was attenuated compared with that reported in 2001. Nonetheless, a 3-fold increase in the testicular cancer rate among prenatally DES-exposed men compared with men not so exposed was observed in a recent meta-analysis of the association between prenatal DES exposure and testicular cancer (38).

The inverse association observed between prenatal DES exposure and urinary system (combined bladder and kidney) cancers in our male cohort was unexpected. Bladder cancer etiology may, however, have a prenatal component. An increase in bladder cancer was observed in a Chilean population where fetal arsenic exposure occurred via maternal ingestion of drinking water with elevated arsenic levels (39). Also, the frequency of bladder hyperplasia but not tumors was increased among male mice who were prenatally administered arsenic and DES shortly after birth relative to those who were only prenatally administered arsenic (40). Postnatally, estrogen inhibits the conversion of aromatic amines to bladder carcinogens in the liver. Decreased bladder tumor progression subsequent to estrogen administration was also observed in laboratory animals. These two findings may possibly explain the decreased bladder cancer risk observed among females relative to males (41). Furthermore, women with increased parity were observed to have lower bladder cancer incidence (42), lending further supporting evidence of estrogen playing a protective role against the development of bladder cancer. Increasing parity was also observed to be positively associated with renal cancer risk (43). Nonetheless, these observations do not directly associate prenatal DES exposure with decreased urinary system (bladder and kidney) cancer risk in males. Consequently, the inverse association between DES exposure and urinary systems tumors in this male cohort, is, therefore, currently inexplicable.

Increased screening among the prenatally DES exposed men could result in an inflated risk of prostate cancer in this group, but prevalence of screening was similar between the two groups and the risk of early-stage prostate cancer was not different between the exposed and unexposed. Although uncontrolled confounding could influence DES'-estimated effect on prostate cancer incidence, other than age, race, and positive family history of prostate cancer, there are few established risk factors for prostate cancer. The cohort was predominantly Caucasian and the percentage of men with a positive family prostate cancer history was similar in the two groups. Identified risk factors for development of aggressive or fatal prostate cancer include adult height, BMI, current or recent smoking history, and dietary factors (5). No dietary data were available, so it was not possible to adjust the effect estimate for DES on aggressive prostate cancer risk for this factor; confounding could have resulted if diet differed by DES exposure status. The distribution of positive smoking history, and BMI is similar however, among the exposed and unexposed groups. Loss to follow-up could have resulted in underacertainment of cases and could have biased the estimates of DES' effect on overall and site-specific cancer incidence. This is, however, unlikely because the SIRs comparing the cancer incidence among the unexposed with the general population were at or near unity, and the percentages of those who completed the questionnaires in the two groups was similar. Also, although the NDI and Cancer Registry did not cover the entire cohorts, searches were conducted without consideration of exposure status and consequently, differential case ascertainment would be, in this instance, unlikely as well.

The study benefited from prospective follow-up of a large number of men for whom their DES exposure information was relatively reliable. In addition, the information on prostate cancer covariates and modifiers was available for a substantial percentage of the cohort. There did not appear to be an effect of DES on overall prostate cancer risk. This estimate, however, represents mostly those tumors that do not pose an appreciable risk of further development. The estimate of DES' effect on aggressive prostate cancer was imprecise. The paucity of prostate cancer cases also limited the estimation of the interaction between DES exposure and obesity on prostate cancer risk with any reasonable degree of precision. The study was also disadvantaged by the small number of other site-specific cases. Consequently, the influence of DES on site-specific tumors is still uncertain. Nevertheless, the study results lend reassurance that prenatal DES exposure does not appear to pose an appreciable risk to overall cancer development among men.

W.C. Strohsnitter reports other support from NIH during the conduct of the study. M. Hyer reports other support from Information Management Services during the conduct of the study. J.R. Palmer reports grants from National Institutes of Health during the conduct of the study. E.E. Hatch reports serving as a co-author on the DES chapter for UpToDate, and got reimbursed approximately $800–1,000 per year (it varies depending on use of the chapter on line). L. Titus reports other support from Geisel School of Medicine at Dartmouth during the conduct of the study. No disclosures were reported by the other authors.

W.C. Strohsnitter: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft. M. Hyer: Formal analysis, methodology, writing–original draft. K.A. Bertrand: Conceptualization, formal analysis, investigation, methodology, writing–original draft. A.L. Cheville: Data curation, investigation, writing–review and editing. J.R. Palmer: Conceptualization, data curation, supervision, funding acquisition, methodology, writing–review and editing. E.E. Hatch: Methodology, writing–review and editing. K.M. Aagaard: Writing–review and editing. L. Titus: Conceptualization, supervision, investigation, methodology, writing–review and editing. I.L. Romero: Supervision, methodology, writing–review and editing. D. Huo: Investigation, writing–review and editing. R.N. Hoover: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing. R. Troisi: Data curation, formal analysis, supervision, investigation, methodology, writing–original draft, project administration.

This research was supported by contracts with the NCI (HHSN261201500029C to W.C. Strohsnitter and A.L. Cheville; HHSN261201500027C to L. Titus; HHSN261201500028C to I.L. Romero and D. Huo; HHSN261201500026C to J.R. Palmer, K.A. Bertrand, E.E. Hatch, and K.M. Aagaard). The authors are grateful for the diligent efforts of study coordinators Ann Urbanovitch, Melissa Rathbun, Janell Keehn, Hannah Lord, Minji Kang, and Suzanne Lenz. We also offer our sincere thanks to the Sons' cohort members for their invaluable participation in this extended follow-up. This work was supported by contracts awarded by the National Cancer Institute, U.S. National Institutes of Health.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.