Prenatal Diethylstilbestrol Exposure and High-Grade Squamous Cell Neoplasia of the Lower Genital Tract

Abstract

Background

Prenatal diethylstilbestrol (DES) exposure is associated with an excess risk of clear cell adenocarcinoma of the vagina and cervix, and of high-grade squamous neoplasia.

Objective

We explored whether neoplasia risk remains elevated among DES-exposed women as they age.

Study Design

4,062 DES-exposed and 1,837 unexposed daughters were followed for approximately 30 years (1982–2013) for pathology-confirmed diagnoses of cervical intraepithelial neoplasia grade 2+ (CIN2+) of the lower genital tract (n=178). Hazard ratios (HR) and 95% confidence intervals (CI) were estimated adjusting for birth year and individual study cohort.

Results

The cumulative incidence of CIN2+ in the DES exposed women was 5.3% (CI 4.1%–6.5%) and in the unexposed was 2.6% (CI 1.5%–3.7%). The HR for DES and CIN2+ was 1.98 (CI 1.33–2.94), and was similar with further adjustment for frequency of cervical cancer screening (HR 1.97; CI 1.33–2.93). The HR was 2.10 (CI 1.41–3.13) with additional adjustment for other potential confounders. The HR for DES exposure was elevated through age 44 (age <45 HR 2.47; CI 1.55–3.94), but not in women age 45+ HR 0.91; CI 0.39–2.10. In exposed women, HRs for DES were 1.74 (CI 1.09–2.79) among those who had earlier evidence of vaginal epithelial changes (VEC), presumably reflecting glandular epithelium undergoing transformation to normal, adult type squamous epithelium, and 1.24 (CI 0.75–2.06) among those without VEC, compared with unexposed women. The HRs for DES and CIN2+ were higher among women with earlier intrauterine exposure (HR 2.64; CI 1.64–4.25 for <8 weeks gestation and HR 1.41; 0.88–2.25 for 8+), and lowest (HR 1.14; CI 0.59 – 2.20) when exposure began after the 15th week.

Conclusion

CIN2+ incidence was higher among the DES exposed, particularly those with early gestational exposure and VEC. The HR for DES and CIN2+ remained positive and significant until the mid-40s, confirming that the recommendation of annual cytological screening among these women is appropriate. Whether those 45 years of age and older continue to require increased screening is unclear, and would require a careful weighing of possible risks and benefits.

Introduction

The association between prenatal exposure to diethylstilbestrol (DES), a nonsteroidal synthetic estrogen given in pregnancy, and vaginal and cervical clear cell adenocarcinoma in young women was first described nearly forty-five years ago.¹ Subsequently, benign pathological findings of the genital tract were associated with DES, including an increased prevalence of vaginal adenosis, ectropion, and a wider cervical transformation zone.^2,3

The National Cooperative Diethylstilbestrol and Adenosis (DESAD) study sought to determine the prevalence and incidence of cervical and vaginal neoplasia in a large cohort of DES-exposed and unexposed women. Baseline screening examination data from this study did not provide evidence of an increased prevalence of squamous neoplasia;⁴ however, a later follow-up study of incident cases showed a nearly 2-fold increased risk.⁵ In the NCI Combined DES Cohorts Follow-Up Study, Hatch et al.⁶ also reported a doubling of risk of high grade dysplasia in primarily younger women, with adjustment for history of routine cervical cancer screening.

The recognition of the causal association of human papilloma virus (HPV) with cervical neoplasia and the increasingly widespread use of the HPV vaccine has prompted a reevaluation of the guidelines for screening Pap smears. The American College of Obstetricians and Gynecologists (ACOG) issued a revised practice bulletin in November, 2012⁷ that reduced screening cytology frequency to every three years among women aged 21–65 years without risk factors or previous cervical disease. Women with prenatal DES exposure were disqualified from reduced screening frequency because of their increased risk of cervical neoplasia.

The objective of the current study was to determine whether the elevated risk of high-grade squamous neoplasia associated with in utero DES exposure has continued with age.

Materials and Methods

Cohorts

The NCI DES Combined Cohort Follow-up consists of the following: 1) women who participated in the National Cooperative Diethylstilbestrol Adenosis Project ([DESAD] cohort),⁸ 2) women whose mothers participated in a clinical trial of DES in 1951–52 (Dieckmann Cohort),⁹ 3) women whose mothers were treated in a large private infertility practice in Massachusetts, USA (Horne Cohort) and 4) women from Massachusetts, New Hampshire, Maine and the Mayo Clinic whose mothers participated in the Women’s Health Study ([WHS] Cohort), a study of the subsequent health effects of DES in women who were administered DES during their pregnancy.¹⁰

Mailed questionnaires and physical examinations

The follow-up of the four combined cohorts by NCI began in 1994 with a mailed questionnaire, with subsequent questionnaires mailed in 1997, 2001, 2006, and 2011. Prior to the Combined Follow-up questionnaire study (1984–1989), DESAD cohort members were mailed annual questionnaires, and medical records and pathology reports were collected for cancers and gynecologic neoplasia. In 1982 and 1983, many members of the DESAD cohort still participated in annual screening examinations as part of the study protocol. The last routine follow-up of the Dieckmann cohort consisted of a mailed questionnaire in 1990. The Horne cohort was reassembled for follow-up along with unexposed siblings in the mid-1970s, and mailed annual questionnaires through the 1980s. Daughters from the WHS cohort had not been followed before the combined follow-up in 1994.

The Dieckmann and DESAD studies incorporated a comprehensive gynecologic examination around the time of recruitment in the mid-1970s into their original cohorts that systematically identified vaginal epithelial changes (VEC) by means of colposcopy or iodine staining. Identical screening protocols were used for the exposed and unexposed women. VEC is glycogen poor squamous epithelium found in the vagina or exocervix that presumably reflects glandular epithelium undergoing transformation over time to glycogenated, normal adult type squamous epithelium. These changes were more frequent in women prenatally exposed to DES early in pregnancy who also had large cumulative doses of DES by the end of pregnancy.¹¹ No attempt was made to systematically physically examine members of the Horne or WHS cohorts. Institutional Review Board approvals were obtained at the field centers and the NCI. Participants indicated their informed consent by completion of a questionnaire or telephone interview, and/or with written consent for pathology and slide reports.

DES exposure and covariate ascertainment

For all combined cohort participants, prenatal exposure to DES, or the lack thereof, was documented by the medical record or a physician’s note. Gestational week of first DES use was available for 75% of all exposed women, and for 80.2% of the DESAD and Dieckmann exposed. Because data for total cumulative DES dose were available for only 38% of the women, we classified the individual cohorts as high- or low-dose based on differences in prescribing practices by U.S. region (unknown for a subgroup of the WHS). Agreement between the dose categories and individual doses was excellent among those with complete data.¹² Information on highest level of education, smoking status, and frequency of routine medical examinations, including Pap smears, in the last 5 years was collected on the 1994 questionnaire. Smoking status was updated on the 2006 questionnaire, and menopausal status and frequency of Pap smears were ascertained on all five questionnaires and treated as time-dependent.

CIN2+ confirmation

Reports of CIN2+ were available from two sources, records from the original cohorts (1982–1988), and the NCI Combined Cohort Study questionnaires (1989–2011); the methods for confirmation of cases were similar. The Combined Cohort Study questionnaires ascertained new diagnoses of neoplasia, and biopsies of the cervix, vagina or vulva that indicated a precancerous condition (dysplasia or carcinoma in-situ, but not abnormal pap smears only). Pathology records were obtained for reported biopsy-confirmed, genital-tract neoplasia of any grade (including HPV infection). Slides also were requested for pathology-confirmed diagnoses of cervical intraepithelial neoplasia grade 2 and above (CIN2+), and were reviewed by one pathologist (SJR), blinded to DES-exposure status.

Among participants included in the analysis, 5,237 (89% of exposed and 88% of unexposed) completed a questionnaire during the 1994 follow-up or after, and 1,247 women (23.8%) reported having had a biopsy of the lower genital tract. Pathology reports were obtained for 986 (79%) of those reporting a biopsy, and of these, 206 (21%) indicated CIN2+. Pathology reports for the remaining self-reports indicated CIN1 (n=280) or other benign diagnoses (n=500, e.g. squamous metaplasia, inflammation, HPV only). Representative slides were reviewed for 169 (82%) CIN2+ cases confirmed by pathology reports, and 30 cases were downgraded to <CIN2. Eight of the CIN2 cases were glandular cell type, and not included in the main analysis. All 8 glandular lesions were cervical; 7 were DES-exposed (2 were invasive, 1 was adenocarcinoma in situ, 3 were CIN3 and 1 was CIN2), and 1 was unexposed (CIN3). An additional 10 cases were not reported as biopsies, but identified from pathology reports obtained for other reasons. Of the 178 pathology confirmed diagnoses of squamous CIN2+ included in the analysis, 163 were cervix (including n=11 cases of invasive carcinoma of the cervix), 9 were vagina and 6 were vulva (n=2 were invasive). The National Death Index was routinely searched and one additional invasive case was identified, resulting in a total of 3 deaths in the exposed from invasive lower genital tract cancer among CIN2+ cases.

Exclusions and Follow-up Information

Information from patient history and medical record review was used to ascertain prior history of diagnoses and treatments. Table 1 shows exclusions and follow-up information. Of 7,232 women in the study, 460 who were lost to follow-up or died before 1982 were excluded. To limit the analysis to incident disease, 200 cases were excluded due to pathology-confirmed high-grade neoplasia diagnosed before the start of follow-up and an additional 107 were excluded who reported cervical dysplasia before the start of follow-up. Because prior treatment of the cervix may lower the subsequent incidence of CIN 2+, 416 women who received treatment before 1982 were excluded at baseline from the main analyses. In addition, 150 women (0 cases) who had a hysterectomy prior to 1982 were excluded. Out of 5,899 women in the analysis, 5237 (89% of exposed and 89% of unexposed) responded to any of the NCI follow-up questionnaires.

Table 1.

Follow-up Information on Diethylstilbestrol Exposed and Unexposed Daughters

Follow-up Information	Exposed	Unexposed	TOTAL
NCI Combined Cohort	5012	2220	7232
Excluded from analysis:	950	383	1333
Lost to follow-up or deceased before entrya	291	169	460
Diagnosis of CIN2+ before entryb	167	33	200
Biopsy reported before entry, pathology not obtaineda	84	23	107
Hysterectomy before entry	77	73	150
Treatment before 1/1/1982c	331	85	416
Total Participants in Analysisa	4062	1837	5899
Cohort:
DESAD	3267	869	4136
Dieckmann	309	285	594
Horne	277	204	481
WHS	209	479	688
CIN2+ case (squamous cervix, vagina, vulva)b	145	33	178
Biopsy reported, pathology not obtained (censored at reported diagnosis date)	184	71	255
NCI Questionnaire Responsed
Responded to any NCI questionnaire (1994–2011)	3612 (89%)	1625 (89%)	5237
Responded to 1994 questionnaire	3429 (84%)	1596 (87%)	5025
Responded to 1997 questionnaire	3377 (83%)	1526 (83%)	4903
Responded to 2001 questionnaire	3272 (81%)	1451 (79%)	4723
Responded to 2006 questionnaire	2939 (72%)	1368 (75%)	4307
Responded to 2011 questionnaire	2734 (67%)	1243 (68%)	3977
Responded to all questionnaires	2467 (61%)	1160 (63%)	3627

Abbreviations: Diethylstilbestrol, DES.

^aEntry into analysis is 1/1/82 except for WHS, which is date of first NCI questionnaire in 1994 or 1995.

^bDiagnoses (dx) were confirmed by pathology report.

^cTreatments included conization, hot and cold cautery, and multiple biopsies documented by medical records that were considered treatment in the original DESAD study; treatment information was available for the DESAD cohort only.

^dNumber and percentages based on participants included in the analysis.

Statistical Analysis

The analyses focused on the first occurrence of pathology confirmed CIN2+. Person-years at risk for each woman were computed from 1/1/82 (except for the WHS, for which follow-up started in 1994–1995) until the date of first documented diagnosis of CIN2+, date of last known follow-up, or date of last questionnaire response. The start of follow-up was chosen to correspond to the end of follow-up in the original cohort study of incident dysplasia among the DESAD cohort,⁵ which comprises 70% of the current study population. Women were censored at the reported date of hysterectomy, except when CIN2+ was diagnosed at surgery. Participants who reported dysplasia on the questionnaire but for whom we were not able to obtain pathology records were censored at their reported diagnosis year. Among the exposed, follow-up ranged from <1 year to 32.0 years with a median of 25.4 (25th percentile 13.6, 75% percentile 30.0) and among the unexposed was <1 year to 32.0 years with a median of 18.2 years (25th percentile 11.9, 75th percentile 29.9).

The association of DES and CIN2+ was estimated with hazard ratios (HR) and 95% confidence interval (CI) from Cox regression models,¹³ using SAS statistical software (version 9.2).¹⁴ The models included original study cohort (except for models evaluating dose because the Dieckmann cohort had a uniform dose) and birth year, and used age as the time-metric. Additional covariates included education, age at first intercourse, number of sexual partners, smoking status, and menopausal status. Pap smear screening represented the number of Pap smears in the five years before the most recent questionnaire. Categories for each of the covariates are listed in Table 2. Missing values were categorized separately and included in the models; a complete case approach yielded the same results for DES and CIN2+ (not shown). Associations were evaluated by DES dose and timing, and in the Dieckmann and DESAD cohorts, by presence or absence of VEC.

Table 2.

Distribution of High-grade Squamous Neoplasia for Selected Characteristics by Diethylstilbestrol Exposure Status

Characteristics	DES Exposed			DES Unexposed
	# person- years	%	# cases	# person- years	%	# cases
ALL	88098		145	35934		33
Cohort
DESAD	72915	82.8	98	19251	53.6	17
Dieckmann	6408	7.3	16	5695	15.9	8
Horne	5731	6.5	26	4326	12.0	3
WHS	3044	3.5	5	6663	18.5	5
Age (y; time-dependent)
<30	14548	16.5	41	4418	12.3	6
30–34	15110	17.2	41	5152	14.3	9
35–39	15710	17.8	26	5728	15.9	4
40–44	14083	16.0	21	5725	15.9	3
45–49	12146	13.8	6	5537	15.4	5
50+	16505	18.7	10	9377	26.1	6
Birth year
<1953	34732	39.4	48	18095	50.4	15
1953+	53366	60.6	97	17839	49.6	18
Calendar year (time-dependent)
<1995	44544	50.6	98	15373	42.8	17
1995+	43558	49.4	47	20565	57.2	16
Education (1994)
<4-year college	27170	30.8	48	13696	38.1	20
4-year college	30541	34.7	56	11487	32.0	5
Graduate school	23440	26.6	35	8849	24.6	7
Missing	6948	7.9	6	1902	5.3	1
Smoking status (at exit or 2006)
Never	45916	52.1	71	17087	47.5	6
Ever	38094	43.2	69	17589	49.0	25
Missing	4088	4.7	5	1261	3.5	2
Menopausal status (time-dependent)
Premenopausal	73103	83.0	138	28383	79.0	29
Postmenopausal	14503	16.5	6	7342	20.4	4
Missing	492	0.56	1	209	0.58	0
Age at first intercourse
<=18	36511	41.4	72	15642	43.5	16
19+	45167	51.3	67	17502	48.7	14
Missing	6420	7.3	6	2790	7.8	3
# of male partners (1994)
<5	42742	48.5	35	16779	46.7	8
5–9	17168	19.5	46	7948	22.1	9
10+	26102	29.6	59	9330	26.0	13
Missing	2086	2.4	5	1876	5.2	3
# of Pap smearsa
<3	18903	21.5	14	8286	23.1	6
3+	59436	67.5	101	23266	64.8	21
Missing	9760	11.1	30	4382	12.2	6

Abbreviations: High-grade Squamous Neoplasia, CIN2+; Diethylstilbestrol, DES.

^aReported for the 5 years before the most recent questionnaire.

Results

The average age at entry into the analysis was 16.0 and 16.5 years for exposed and unexposed women in the Horne cohort, compared with, respectively, 27.6 and 28.4 in DESAD, 30.1 and 30.0 in Dieckmann, and 42.7 and 43.2 in the WHS. DES-exposed women were slightly younger (median age at last follow-up was 51.3 for exposed and 53.7 for unexposed), less likely to smoke and had higher levels of education than the unexposed participants (Table 2).

In general, we found expected associations between several covariates and the incidence of CIN2+. For example, the incidence of CIN2+ declined with age: compared with women <30, the HRs were 1.0, 0.56, 0.49 and 0.25 for women ages 30–34, 35–39, 40–44 and 45+, respectively (data not shown). CIN2+ incidence adjusted for birth year was higher in women with more male partners and among women who had ever smoked (data not shown). CIN2+ was also elevated among women with a greater number of prior pap smears, although most (67.5% of exposed and 64.8% of unexposed) reported having 3+ pap smears in the last 5 years. Education and age at first intercourse were not materially associated with CIN2+ (data not shown).

The cumulative incidence of CIN2+ adjusted for birth year and cohort for DES exposed (5.3%; CI 4.1%–6.5%) and unexposed women (2.6%; CI 1.5%–3.7%) by age is presented in Figure 1. Overall, the HR for DES and CIN2+ with adjustment for age and cohort was 1.98 (CI 1.33–2.94), and was similar with further adjustment for frequency of cervical cancer screening (HR 1.97; CI 1.33–2.93) (Table 3). The HR was 2.10 (CI 1.41–3.13) with additional education, smoking, age at first sexual intercourse, and number of male sexual partners. Restricting to the 134 cases of CIN2+ confirmed by slide review, the HR was 2.15 (CI 1.35, 3.41), and including cases that were treated prior to 1982, the HR was 1.98 (CI 1.35–2.90). Results were similar including the 8 cases with glandular lesions with the other cases (fully adjusted HR 2.15 (95% CI 1.45–3.17). We did not have adequate statistical power to separately assess glandular lesions. For the exposed and unexposed, respectively, there were 56 and 11 CIN2 cases, 80 and 18 CIN3 cases and 9 and 4 invasive cases. The HR adjusted for birth year and cohort for DES and CIN2 was 2.27 (CI 1.16–4.45) and for CIN3 was 1.98 (CI 1.15–3.38). The HR for DES and invasive squamous cell carcinoma of the cervix (n=7 exposed and n=4 unexposed) was 0.86 (CI 0.23–3.25), with the two additional cases of vulvar cancer (n=9 exposed and n=4 unexposed) the HR was 1.13 (CI 0.32–4.03), and with two additional cases of cervical adenocarcinoma (n=11 exposed and 4 unexposed) the HR was 1.38 (CI 0.41–4.73).

Figure 1. Cumulative incidence of cervical intraepithelial neoplasia 2+ by diethylstilbestrola exposure status and age, adjusted for birth year and cohort.

^aDES=diethylstilbestrol.

The cumulative incidence of CIN2+ rose more steeply with age among the DES-exposed through approximately age 45, when it appeared to level off.

Table 3.

Hazard Ratios and 95% Confidence Intervals for Prenatal Diethylstilbestrol Exposure and High-grade Squamous Neoplasia

	# Cases	HR	95% CI
DES
Adjusted for birth year & cohort	178	1.98	1.33	2.94
+ # of Pap smearsa	178	1.97	1.33	2.93
+ covariatesb	178	2.10	1.41	3.13
Slide confirmed cases onlyc	134	2.15	1.35	3.41
by Cohortd
DESAD	115	1.38	0.82	2.32
Dieckmann	24	1.82	0.78	4.25
Horne	29	6.88	2.06	23.0
WHS	10	2.08	0.60	7.23
by Agee
<30	47	2.66	1.11	6.37
30–34	50	1.97	0.94	4.15
35–39	30	2.87	0.98	8.42
40–44	24	3.11	0.88	11.0
45+	27	0.91	0.39	2.10

Abbreviations: Hazard Ratios, HR; 95% Confidence Intervals, CI; Diethylstilbestrol, DES; High-grade Squamous Neoplasia, CIN2+.

^aAdjusted for birth year, cohort and number of Pap smears over the last 5 years prior to the questionnaire before diagnosis or censoring.

^bIn addition to birth year, cohort and # of Pap smears, covariates include education, smoking, age at first sexual intercourse, and # of male sexual partners.

^cAdjusted for birth year and cohort.

^dAdjusted for birth year.

^eAdjusted for cohort.

The HR for DES and CIN2+ was 1.38 in the DESAD cohort, 1.82 in the Dieckmann cohort, 2.08 in the WHS and 6.88 in the Horne cohort. Twenty-six CIN2+ cases, all but one confirmed with slide review, occurred in exposed women in the Horne cohort, and three slide confirmed cases occurred in the unexposed. The HR for the Horne cohort was 4.36 (CI 1.14–16.7) when follow-up began with the 1994 baseline questionnaire. The median for gestational week at first DES exposure was 5 weeks in the Horne cohort and 10 weeks in both DESAD and Dieckmann (not available for WHS). Eliminating the Horne cohort, the covariate adjusted HR for DES and CIN2+ was 1.68 (CI 1.10, 2.58).

The HR for DES and CIN2+ was elevated among women through 44 years of age and then decreased among those 45+ (Table 3). The cohort-adjusted HR among those <45 years of age was 2.47 (CI 1.55–3.94) and among those 45+ was 0.91 (CI 0.39–2.10; p-value for interaction = 0.03). The DES – CIN2+ association appeared to be stronger among nonsmokers (HR 4.79; CI 2.05–11.2 vs. 1.38; CI 0.85–2.24 among ever smokers, p-value for interaction = 0.01). The HRs were 1.52 (CI 0.54–4.30) and 1.71 (CI 1.04–2.81) among women with <3 and 3+ pap smears over the past 5 years, respectively (p-value for interaction = 0.25). The HR for DES and CIN2+ could not be evaluated by menopausal status because most cases were premenopausal (167/178).

Incidence of cervical dysplasia was greater with earlier vs. later gestational age at first exposure to DES and with higher DES dose (Table 4). Compared with the unexposed, the adjusted HR was 2.64 (CI 1.64–4.25) for women exposed within 8 weeks of the last menstrual period, compared to 1.41 (CI 0.88–2.25) in women exposed at 8 weeks or later. The reduction in HR for DES exposure at each subsequent category of gestational weeks based on categories in Table 4 was 19% (including the unexposed women HR 0.81; CI 0.72, 0.91, p for trend=0.0002 and excluding the unexposed women HR 0.76; CI 0.63–0.91, p for trend=0.003). The association between DES and CIN2+ was slightly stronger among those exposed to high doses (HR 1.73) vs. low doses (HR 1.35), and among the exposed women with VEC (HR 1.74) than in those without VEC (HR 1.24).

Table 4.

Hazard Ratios and 95% Confidence Intervals for High-grade Squamous Neoplasia among Women Prenatally Exposed to DES According to Timing of Exposure during Gestation, Dose, and Vaginal Epithelial Changes at Baseline

	#person- years	# Casesa	HRb	95% CI
Gestational age at initial DES exposurec,d
<=7 weeks	21740	53	2.62	1.63	4.22
8–10 weeks	18021	26	1.55	0.90	2.66
11–14 weeks	12401	16	1.48	0.79	2.76
15+ weeks	15638	14	1.14	0.59	2.20
Unexposed	29270	28	1.00
DES dosee
Low	31674	42	1.35	0.84	2.15
High	54517	100	1.73	1.15	2.60
Unexposed	32340	30	1.00
VECf
Exposed, VEC	39087	67	1.74	1.09	2.79
Exposed, No VEC	39607	46	1.24	0.75	2.06
Unexposed	24944	25	1.00

Abbreviations: Hazard Ratios, HR; 95% Confidence Intervals, CI; Diethylstilbestrol, DES; High-grade Squamous Neoplasia, CIN2+; vaginal epithelial changes, VEC.

^aCases were confirmed by pathology report.

^bAdjusted for birth year, and cohort (except for dose, which is not adjusted for cohort because the Dieckmann cohort had a uniform dose).

^cCalculated as the number of weeks since last menstrual period.

^dExcludes the WHS cohort in which data were unavailable for gestational age at first use.

^eDose is based on cohort, and excludes participants from New Hampshire among whom dose was unavailable (High dose includes Dieckmann, DESAD (Boston), Horne, WHS, and California; low dose includes DESAD (Minnesota, Wisconsin, Texas).

^fThis analysis was performed only in the DESAD and Dieckmann cohorts, which had information on VEC.

Comment

With 15 additional years of follow-up since our last report⁶ the data continue to show that women who were exposed to DES in utero have twice the risk of high-grade squamous neoplasia compared with unexposed women. The association with DES was highest among women who were exposed very early in gestation. These results are consistent with an earlier follow-up of the DESAD cohort⁵ during the annual screening program that found a two-fold increase in cervical dysplasia incidence among the exposed after the first screening examination in the mid-1970s through the early 1980s, and a more recent study using the combined cohorts follow-up data.⁶ In our data, hazard estimates for DES and CIN2+ were similar after adjustment for covariates, including cervical cancer screening history. The increased incidence of CIN2+ among the DES exposed continued through the mid-40s; in exposed women 45+ years, incidence was no longer elevated.

When the National Cancer Institute first commissioned the DESAD project (1974–75), a primary goal of the study was to identify the natural history of DES-related genital tract anomalies. During the DESAD annual examinations, we found that there were frequent discrepancies between atypical forms of squamous metaplasia versus bona fide precancerous squamous cell lesions (dysplasia).¹⁵ Therefore, in the current study we limited our analysis to cases of high grade squamous neoplasia (CIN2+). To further avoid potential misclassification, one pathologist, blinded to DES exposure status but knowledgeable about typical benign changes associated with DES, reviewed pathology confirmed cases. Slides were reviewed for 82% of cases; there was over 80% agreement between pathology report diagnoses of high-grade disease and the slide review.

The unexposed comparison group differed slightly in established risk factors for cervical disease. For example, they were slightly older and were slightly more likely to have ever smoked but adjustment for these factors had a minimal effect on the association of DES and CIN2+. The association between prenatal DES exposure and CIN2+ was higher in the Horne cohort than the other cohorts, but this could be explained partly by their younger age given the DES-CIN2+ age interaction, and the earlier gestational week at first use in the Horne cohort. In addition, the Horne cohort may have included some prevalent cases as they did not receive an examination at baseline as was performed in most of the other women (i.e. DESAD and Dieckmann cohorts). The DES-CIN2+ association was observed in the Dieckmann cohort in which mothers were not prescribed DES based on clinical indication but were participants in a clinical trial, which suggests that the association is not due to an inherently higher genetic risk of women who were prescribed DES. Follow-up was approximately equal between exposed and unexposed women.

DES exposed and unexposed women had identical cervical cancer screening protocols during the DESAD screening program, but more frequent and intense screening among the exposed after the program ended could explain our results. However, self-reported screening patterns were similar and relatively frequent (with most women reporting three or more Pap smears over the previous five years) in exposed and unexposed women, and adjusting for screening behaviors did not materially affect the hazard estimates.

Speculation regarding the biological mechanisms explaining the higher risk of cervical disease in DES exposed women has focused on differences in mechanical and histological attributes of cervical/vaginal tissue, and alterations in immune function. In unexposed women, the transformation zone is tiny. In DES exposed women, the much wider transformation zone is strongly associated with earlier prenatal exposure,² and could explain their greater CIN2+ risk. This area of metaplastic transformation (VEC), reflecting the physiologic and reparative changeover from the embryonic residual and relatively sensitive glandular epithelium to a much tougher squamous epithelium, also may be more susceptible to HPV infection, causative in the genesis of squamous cell neoplasia.¹⁶ Prenatal DES exposure is associated with VEC, and in the DESAD cohort, the intraepithelial neoplasia risk was correlated with the extent of these changes.⁵ Both the prevalence and the extent of VEC markedly declined with age. Thus, the declining incidence of CIN2+ among women after 44 years of age may reflect the shrinking size of the transformation zone as women age.¹⁷

In view of the doubling in risk of high grade cervical squamous neoplasia among women <45 years who were DES exposed in utero, the ACOG 2012⁷ recommendation exempting DES-exposed women from reduced frequency of screening cytology seems prudent. Whether those 45 years of age and older continue to require increased screening is unclear. While the overall estimate for this group was not elevated, the confidence interval was wide indicating that the lack of an effect could be due to chance. Therefore, the ACOG should assess the potential risks and benefits of any change in screening recommendations for these women.