Anderson GL, Judd HL, Kaunitz AM, Barad DH, Beresford SAA, Pettinger M, Liu J, McNeeley SG, Lopez AM, for the Women's Health Initiative Investigators . Effects of Estrogen Plus Progestin on Gynecologic Cancers and Associated Diagnostic ProceduresThe Women's Health Initiative Randomized Trial. JAMA. 2003;290(13):1739-1748. doi:10.1001/jama.290.13.1739
Author Affiliations: Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Wash (Drs Anderson and Beresford and Ms Pettinger); Department of Epidemiology, University of Washington, Seattle (Dr Beresford); Department of Obstetrics and Gynecology, University of California, Los Angeles (Dr Judd); Department of Obstetrics and Gynecology, University of Florida Health Science Center, Jacksonville (Dr Kaunitz); Department of Obstetrics and Gynecology and Women's Health, Albert Einstein College of Medicine, Bronx, NY (Dr Barad); Department of Reproductive Biology, Case Western Reserve University, Cleveland, Ohio (Dr Liu); Department of Obstetrics and Gynecology, Wayne State University, Detroit, Mich (Dr McNeeley); and Department of Internal Medicine, University of Arizona Cancer Center, Tucson (Dr Lopez).
Context The effects of continuous combined hormone therapy on gynecologic cancers
have not been investigated previously in a randomized trial setting.
Objective To determine the possible associations of estrogen plus progestin on
gynecologic cancers and related diagnostic procedures.
Design, Setting, and Participants Randomized, double-blind, placebo-controlled trial of 16 608 postmenopausal
women, who had not had a hysterectomy at baseline and who had been recruited
from 40 US clinical centers between September 1993 and October 1998 (average
follow-up, 5.6 years).
Intervention One tablet per day containing 0.625 mg of conjugated equine estrogens
plus 2.5 mg of medroxyprogesterone acetate (n = 8506) or placebo (n = 8102).
Main Outcome Measure Incident invasive cancer of the ovary and endometrium.
Results In 5.6 years of follow-up, there were 32 cases of invasive ovarian cancer,
58 cases of endometrial cancer, 1 case of nonendometrial uterine cancer, 13
cases of cervical cancer, and 7 cases of other gynecologic cancers. The hazard
ratio (HR) for invasive ovarian cancer in women assigned to estrogen plus
progestin compared with placebo was 1.58 (95% confidence interval [CI], 0.77-3.24).
The HR for endometrial cancer was 0.81 (95% CI, 0.48-1.36). No appreciable
differences were found in the distributions of tumor histology, stage, or
grade for either cancer site. The incidence of other gynecologic cancers was
low and did not differ by randomization assignment. More women taking estrogen
plus progestin required endometrial biopsies (33% vs 6%; P<.001).
Conclusions This randomized trial suggests that continuous combined estrogen plus
progestin therapy may increase the risk of ovarian cancer while producing
endometrial cancer rates similar to placebo. The increased burden of endometrial
biopsies required to assess vaginal bleeding further limits the acceptability
of this regimen. These data provide additional support for caution in the
use of continuous combined hormones.
For years, there has been concern about possible associations of gynecologic
malignancies with postmenopausal hormone therapy. The development of endometrial
hyperplasia and endometrial cancer with unopposed estrogen is well recognized.
To reduce or avoid this complication, progestin has been added,1- 3 although
results from randomized trials are extremely limited. These concerns have
created a need for reasonable monitoring guidelines to follow-up women who
experience vaginal bleeding while taking estrogen plus progestin.
The Women's Health Initiative (WHI) trial of estrogen plus progestin
provides the first opportunity to examine possible associations of gynecologic
malignancies with continuous combined postmenopausal hormone therapy in a
large, randomized, double-blind, placebo-controlled setting. The trial was
stopped early at the recommendation of the independent data and safety monitoring
board on the basis of an increased risk of breast cancer supported by a summary
measure of effects indicating risks exceeded benefits over an average of 5.2
years of follow-up. The initial report provided interim, locally adjudicated
outcomes for endometrial cancer reported through April 30, 2002. This article
summarizes centrally coded gynecologic cancer outcomes and related diagnostic
procedures occurring prior to the announcement of the trial closure and participant
unblinding (July 8, 2002).
Women were recruited at 40 clinical centers in the United States between
September 1993 and October 1998, largely through direct mail. Eligibility
required women to be between age 50 and 79 years, to be postmenopausal, and
to provide written informed consent. Women were excluded if they had preexisting
conditions that contraindicated use of hormones, had health conditions that
suggested a predicted survival of less than 3 years, or were considered likely
to be poor adherers to the study protocol. Only women who had not had a hysterectomy
were considered for this trial. Eligible women were randomized in equal proportions
using a stratified permuted block algorithm to either placebo or to 0.625
mg/d of conjugated equine estrogens plus 2.5 mg/d of medroxyprogesterone acetate,
which was administered in a single tablet (Prempro, Wyeth, St Davids, Pa).
Women who had a prior hysterectomy were randomized to a parallel trial of
estrogen alone and are not included in these results. Study design details
have been published.4,5
All WHI participants provided demographic, medical, reproductive, and
family history information using self-administered questionnaires at baseline.
Prior postmenopausal hormone use was ascertained through a structured interview
asking women to describe the strength, schedule, and duration of each hormone
Endometrial Evaluation. For safety, endometrial
biopsies were performed on every woman interested in the estrogen plus progestin
trial prior to randomization. A 5% cohort of women was randomly selected at
enrollment to undergo routine biopsies at follow-up years 3, 6, and 9. This
cohort was intended to provide a valid comparison of the rates of endometrial
The WHI procedures called for endometrial biopsies to be performed by
WHI trained and certified staff, who were licensed physicians, nurse practitioners,
or physician assistants, using plastic endometrial suction curettes.6,7 Readings were obtained from local pathologists
blinded to randomization assignment.
During follow-up, women with persistent or heavy bleeding were evaluated
by the clinic gynecologist. If a nonroutine biopsy was indicated, study guidelines
permitted the clinic gynecologist to then be unblinded to assist in further
safety evaluation. Women selected for routine biopsies who had a biopsy in
the last 12 months to evaluate bleeding problems were not required to repeat
the procedure. When biopsies could not be accomplished, vaginal ultrasounds
of the endometrium were performed.
At baseline, biopsy evidence of endometrial cancer, complex or adenomatous
hyperplasia, or atypia disqualified women from participating, whereas findings
of simple hyperplasia resulted in temporary exclusion pending resolution.
During study follow-up, evidence of endometrial cancer, complex or adenomatous
hyperplasia, or atypia required permanent discontinuation of study medicines.
Simple hyperplasia identified during follow-up led to unblinding of the consulting
gynecologist. Those participants assigned to placebo had study medications
discontinued and were referred to their health care clinicians for further
management. Those assigned to active therapy continued study medications supplemented
with 20 mg/d of medroxyprogesterone acetate. In such women, the biopsy was
repeated in 3 to 6 months.
An endometrial wall thickness of greater than 0.5 cm on vaginal ultrasound
was considered evidence of an endometrial pathological finding8 and
resulted in exclusion (at baseline) or discontinuation of study pills unless
further evaluation ruled out malignancy.
Cervical Cytology. In all participants, Papanicolaou
tests were performed at 3-year intervals in conjunction with routine pelvic
examinations. Absent an a priori hypothesis of association between estrogen
plus progestin and cervical cancer, these tests were performed by WHI staff
as a courtesy, in conjunction with a required pelvic examination. Pathological
analysis was obtained locally. During follow-up, the protocol was modified
to accept Papanicolaou test results from a participant's health care clinician.
Abnormal test results were referred to the participant's health care
clinician for further diagnostic evaluation and treatment, except when a change
in use of study medicines was required.
Outcome ascertainment procedures have been described.5 Briefly,
semiannual self-reports of new diagnoses were recorded and all associated
medical records were obtained from local health care clinicians and classified
by blinded physician adjudicators at each clinical center. The documents from
all gynecologic cancer cases were forwarded to the WHI clinical coordinating
center for centralized review by cancer coding specialists who were also blinded
to randomization assignment and reported symptoms. Histological codes were
based on the International Classification of Diseases for
Oncology, 2nd Edition.9 Histological
analysis was available from pathology reports for all ovarian and endometrial
cancers. Stage and grade were coded using Surveillance, Epidemiology, and
End Results guidelines.10 In this system, local-stage
ovarian cancer refers to a tumor confined to the ovary with no tumor on the
ovarian surface. A regional stage ovarian cancer represents a tumor that is
present on the ovarian surface, or has evidence of spread to ascites, peritoneal
washings, or other locations within the pelvis. Tumors that spread beyond
the pelvis are considered distant stage. For endometrial cancers, local stage
refers to cancers confined to the endometrium or myometrium/serosa. Regional-stage
tumors have spread only to the pelvis, vagina, and/or the wall of the rectum
or bladder. Information could not be obtained for ovarian cancer stage in
1 woman and ovarian cancer grade in 9 women. Because the number of primary
peritoneal and fallopian tube cancers was small, these are reported together
as other gynecologic cancers.
Tests for differences between groups in the distribution of pathological
features of diseases, and rates and results of endometrial biopsies and Papanicolaou
tests are based on Fisher exact tests. When a diagnostic procedure was performed
multiple times for a woman, her most severe result is reported.
Cancer incidence rate comparisons are presented as hazard ratios (HRs)
and 95% confidence intervals (CIs) from Cox proportional hazards models, stratified
by age and randomization to the WHI dietary trial. Kaplan-Meier estimates
of cumulative hazards are shown as 1 minus the disease-specific failure time
estimates. For the primary results, adjusted 95% CIs are corrected for the
7 outcomes that contributed to the global index, which is consistent with
our initial report.5 This adjustment may not
completely account for the variability in these estimates, given the multiple
outcomes that will be examined in this trial. No adjustments were made for
multiple tests over time because these outcomes make only a minor contribution
to early stopping considerations under WHI monitoring guidelines.11
The primary analyses are based on the intent-to-treat principle. In
the original design, women who had not had a hysterectomy were randomized
to either unopposed estrogen, estrogen plus progestin, or placebo. After information
from the Postmenopausal Estrogen/Progestin Interventions (PEPI) trial12 indicated that it was not feasible to include an
unopposed estrogen group in a long-term prevention study in women who had
not had a hysterectomy, the estrogen group was closed and the 331 women previously
randomized to unopposed estrogen were unblinded and switched to combined hormone
therapy. Mean (SD) duration of exposure to unopposed estrogen in these women
was 5.4 (3.1) months. When appropriate, analyses were performed separating
out the women randomized during this phase to account for the exposure to
unopposed estrogen. Several other sensitivity analyses were conducted. In
per protocol analyses, events and follow-up time occurring more than 6 months
after the participant became nonadherent to study medicines or initiated use
of nonstudy hormones were censored. This analysis preserves the randomization
assignment and attempts to limit the dilution of effects that nonadherence
may entail. As treated analyses (attributing events to the woman's use of
hormones, either study pills or those provided by her health care clinician,
6 months prior to the event) were also performed. Analyses censoring women
at the time of surgical removal of the organs of interest (total hysterectomy
for endometrial cancer and bilateral oophorectomy for ovarian cancer) were
Potential effect modification with known gynecologic cancer risk factors
was assessed in expanded proportional hazards models that included the designated
risk factor and randomization assignment as main effects and the interaction
between these. Participants with missing values were excluded only from the
analyses using the relevant variables.
Throughout this article, we report unadjusted 2-sided P values to indicate the relative strength of evidence in these secondary
analyses. Seventeen tests for interactions with selected baseline characteristics
were examined and accordingly 1 test would be expected to be significant at
the P = .05 level by chance alone. All analyses were
performed using SAS statistical software (version 8.2, SAS Institute Inc,
Many subject characteristics have been presented.5 Use
of unopposed estrogen prior to enrollment was limited (11%), was generally
of short duration (<5 years), and had not taken place recently (Table 1). Prior use of combined hormones
was more common (18%) and also more recent.
The average follow-up time for this report is 5.6 years. At the time
of our interim report, 42% of women randomized to estrogen plus progestin
and 38% of women randomized to placebo had stopped taking their study medications.5 A total of 485 deaths (3%) occurred before July 8,
2002 (Figure 1).
The observed annual incidence rate of ovarian cancer was 34 per 100 000
person-years, somewhat lower than the population-based rate of 45 per 100 000
person-years reported by the Surveillance, Epidemiology, and End Results for
women of this age distribution.13 The rate
in the estrogen plus progestin group was elevated (20 vs 12; HR, 1.58; 95%
CI, 0.77-3.24 [adjusted 95% CI, 0.59-4.23]), but not statistically significant.
Limiting the analyses to invasive epithelial cancers did not change the results
substantially (HR, 1.64; 95% CI, 0.78-3.45). The possibility of an increasing
effect over time is suggested by the Kaplan-Meier estimates of cumulative
hazards (Figure 2A) but likewise
did not reach statistical significance. Controlling for family history of
colorectal cancer, the only baseline risk factor exhibiting a noticeable degree
of imbalance, produced a modest increase in the estimated effect (HR, 2.11;
95% CI, 0.96-4.60). No substantial changes were found in analyses conducted
per protocol (HR, 1.51; 95% CI, 0.64-3.55), as treated (HR, 1.76; 95% CI,
0.87-3.55), or censoring at the time of bilateral oophorectomy (HR, 1.59;
95% CI, 0.78-3.25).
There was no evidence of a difference between treatment groups in the
distribution of histological classes, morphological grade, or stage of disease
at diagnosis (Table 2). No significant
interactions were found with age, race/ethnicity, body mass index, family
history of breast or ovarian cancer, family history of colorectal cancer,
prior use of oral contraceptives, prior exposure to unopposed estrogen, or
prior use of combined hormones. However, power for these tests was limited
because the number of events was small. None of the women reporting a history
of ovarian cancer at baseline experienced a new diagnosis of ovarian cancer.
Ovarian cancer was the reported cause of death in 9 women taking estrogen
plus progestin and 3 women taking placebo (HR, 2.70; 95% CI, 0.73-10.0).
The observed incidence rate of endometrial cancer was 62 per 100 000
person-years, which was also lower than the Surveillance, Epidemiology, and
End Results rate of 83 per 100 000 person-years.13 A
small, nonsignificant reduction in endometrial cancer risk was observed with
estrogen plus progestin use (27 vs 31; HR, 0.81; 95% CI, 0.48-1.36 [adjusted
95% CI, 0.40-1.64]). A similar reduction was observed in cancers arising from
the epithelium (all reported histological classes except stromal sarcoma and
mixed mullerian) (23 vs 30; HR, 0.71; 95% CI, 0.41-1.22). Kaplan Meier estimates
of the cumulative hazards reveal no differences in rates throughout follow-up
beyond what could be readily explained by chance alone (Figure 2B). Three (0.9%) endometrial cancers were diagnosed among
the 331 women originally randomized to estrogen alone compared with 3 (0.5%)
cases among the 573 women randomized to combined hormones and 2 (0.4%) cases
among the 522 women randomized to placebo during this same period. Removing
these cases had a limited impact on the results (21 vs 29; HR, 0.72; 95% CI,
0.41-1.26). The per protocol and as treated analyses for endometrial cancer
also did not yield important differences in overall findings (HR, 0.83; 95%
CI, 0.42-1.64 and HR, 0.80; 95% CI, 0.45-1.43, respectively). Censoring at
the time of hysterectomy provided a similar result (HR, 0.81; 95% CI, 0.49-1.36).
There was no evidence of a difference in the distributions of histological
class, morphological grade, or stage at diagnosis of endometrial cancer by
randomization assignment. No significant interactions were found with age,
race/ethnicity, body mass index, hypertension, smoking status, pack-years
of smoking, prior use of unopposed estrogen, or prior use of estrogen plus
progestin. The possibility of an interaction with diabetes could not be tested
because of the sparseness of these data. One death in the placebo group was
attributed to endometrial cancer.
The numbers of borderline ovarian tumors (1 vs 3) and other gynecologic
cancers (6 vs 1) were too small to provide meaningful comparisons. Using a
combined outcome of invasive ovarian, primary peritoneal, and fallopian tube
cancers resulted in an estimated HR that was slightly higher than invasive
ovarian cancer alone (26 vs 13; HR, 1.92; 95% CI, 0.99-3.74). We note that
one primary peritoneal cancer in the estrogen plus progestin group was detected
by an ultrasound performed to evaluate the endometrium. One leiomyosarcoma
of the uterus was also reported. No difference was detected in cervical cancer
incidence (8 vs 5; HR, 1.44; 95% CI, 0.47-4.42).
Estrogen plus progestin reduced the percentage of unsuccessful biopsies
relative to placebo in the cohort randomly selected for routine surveillance
(21% vs 36%, P<.001), but no difference was observed
in the distribution of findings among women with results available (P = .28) (Table 3).
This comparison may be confounded by a higher proportion of women having multiple
biopsies in the estrogen plus progestin group (47% vs 27%; P<.001) because it is impossible to accurately attribute biopsies
in these women to those conducted in response to reports of bleeding vs those
Among women selected for usual care, the fraction of women taking combined
hormones and requiring diagnostic biopsies increased more than 5-fold over
placebo (33% vs 6%; P<.001) and twice as many
women required multiple biopsies (38% vs 17%; P<.001).
Among women having successful biopsies, a higher proportion taking estrogen
plus progestin had normal findings (85% vs 68%), reflecting increased bleeding
not arising from malignant or premalignant lesions. While the proportion of
biopsies yielding abnormal findings in women taking estrogen plus progestin
was low, the proportion of women having these abnormalities in the estrogen
plus progestin group was not reduced relative to placebo. More simple and
adenomatous hyperplasias and atypias were found in the estrogen plus progestin
group, although this increase may be an artifact of the higher biopsy rate.
More women in the estrogen plus progestin group were examined with ultrasound;
the fraction with repeat ultrasounds was also elevated (Table 4). No significant differences were found in the endometrial
findings of these examinations, but some small differences were noted in the
proportion with other pelvic abnormalities.
Follow-up Papanicolaou test results were available for 94% of trial
participants (Table 5). The distribution
of findings varied significantly (P<.001) with
the estrogen plus progestin group yielding slightly more mild dysplasia, low
grade squamous intraepithelial lesions, or atypia than the placebo group (7.8%
vs 5.5%) and fewer normal results (92% vs 94%).
During 5.6 years of follow-up, 111 women were diagnosed as having invasive
gynecologic cancers (invasive ovarian, 32; endometrial, 58; nonendometrial
uterine, 1; cervical, 13; and other gynecologic, 7). In women randomized to
estrogen plus progestin, a nonsignificant 1.58-fold increase in ovarian cancer
and a nonsignificant 19% reduction in endometrial cancer were observed, relative
to placebo. Data for other gynecologic malignancies were too sparse to provide
meaningful comparisons but were included for completeness. Women randomized
to continuous combined hormones were subjected to more endometrial biopsies
and vaginal ultrasounds and were more frequently found to have mild abnormalities
in routine Papanicolaou tests.
In this trial, women taking estrogen plus progestin were diagnosed as
having invasive ovarian cancer at a rate of 42 per 100 000 person-years,
15 per 100 000 person-years more than the placebo group rate. The possibility
of an increase in ovarian cancer mortality was also noted. Many, but not all,
observational studies have found a modest increased risk of ovarian cancer
or ovarian cancer mortality associated with postmenopausal estrogen use,14- 19 but
few studies have reported results specifically on combined hormones. In a
recently reported US cohort study,20 an association
was found with estrogen alone (odds ratio [OR], 1.6; 95% CI, 1.2-2.0) but
not with combined hormones (OR, 1.1; 95% CI, 0.64-1.7), except possibly in
women previously exposed to estrogen alone (OR, 1.5; 95% CI, 0.91-2.4). That
study did not address the schedule of progestin use. A recent Swedish case-control
study21 (the only currently available study
known to distinguish between continuous combined hormone therapy and sequential
progestin treatment) found an increased risk associated with use of unopposed
estrogen (OR, 1.43; 95% CI, 1.02-2.00) and sequential preparations (OR, 1.54;
95% CI, 1.15-2.05), but not with use of continuous progestin regimens (OR,
1.02; 95% CI, 0.73-1.43). However, the majority of women in this latter study
used a progestin derived from 19-nortestosterone rather than the 17-hydroxyprogesterone
derivative medroxyprogesterone acetate used in the current trial.
While the etiologies of ovarian cancer are poorly understood, a role
for estrogen and progestin is biologically plausible.22 For
example, the gonadotropin hypothesis asserts that the many reproductive history
risk factors having a modest association with ovarian cancer risk act indirectly
by increasing exposure of the ovarian epithelium to estrogen23,24 and
consequently to proliferation and malignant transformation.25 Other
hypotheses regarding inflammation of the ovarian epithelium26 and
retrograde bleeding27 have been introduced
and allow for an increased risk with hormones through an indirect pathway.
The WHI data do not relate directly to the incessant ovulation hypothesis28 because all participants were menopausal at entry.
Progestins have been hypothesized to have a favorable effect on ovarian
cancer incidence22 based on generally consistent
findings of lower risk associated with increasing parity and the use of oral
contraceptives,29 and on animal data describing
a role for progestins in promoting apoptosis.30 The
current WHI trial cannot address this question directly, but the eventual
comparison with the parallel trial of estrogen alone in women with prior hysterectomy
may provide some insight. The WHI data suggest that the continuous combined
estrogen and progestin preparation examined in the trial will have no role
in ovarian cancer prevention.
In women taking estrogen plus progestin, the incidence of endometrial
cancer during the 5.6 years of follow-up was 56 per 100 000 person-years
or 13 fewer cases per 100 000 person-years than observed in women taking
placebo. This difference cannot be distinguished from chance.
This is the first randomized, double-blind, placebo-controlled trial
to demonstrate that endometrial cancer rates for women taking continuous combined
hormones are similar to placebo group rates, indicating that progestin protects
against the increased risk of endometrial cancer associated with unopposed
estrogen. Several reasonably sized, randomized, placebo-controlled trials
of combined hormones have reported a substantial reduction in the formation
of endometrial hyperplasia, the presumed precursor to many endometrial carcinomas
compared with women given estrogen-only therapy.12,31,32 The
results of the present study are also consistent with one smaller randomized
trial33 and 3 observational studies reporting
no increase in endometrial cancer risk associated with continuous combined
estrogen plus progestin therapy of at least 3 years' duration.1- 3
Uterine bleeding was a frequent adverse effect of this regimen, leading
to much more frequent biopsies and ultrasounds in women taking combined hormones
than taking placebo. This increase in bleeding and the gynecologic procedures
needed to diagnose and resolve the ensuing concerns (including hysterectomy)
continue to be a major drawback of combined hormone therapy, including continuous
The WHI data on cervical cancer are too limited to suggest there is
any association between incidence and estrogen plus progestin therapy. The
statistically significant increase in mild abnormalities detected on Papanicolaou
tests is interesting and warrants further investigation.
The WHI estrogen plus progestin trial is the largest, randomized, double-blind,
placebo-controlled trial of continuous combined hormones that has been conducted,
yet some limitations must be acknowledged. The number of gynecologic cancers
observed was small, yielding wide CIs for the overall effects and limited
power to examine possible differential effects in disease subtypes or in subgroups
of women. Use of community-based diagnoses rather than central reading of
pathological specimens constrains our ability to provide detailed subclassification
of disease. As gynecologic cancers were secondary end points of the trial,
issues of multiple testing apply. Adjusted CIs are provided to suggest the
additional conservatism needed in interpreting these results. The trial was
stopped early, which limited the precision of these results and precluded
the examination of longer-term exposure. Finally, this trial tested one means
of administering a single hormone regimen. We do not know the extent to which
these results apply to other postmenopausal hormones. The effects of estrogen
alone will be examined in the parallel trial that is ongoing.
In assessing the overall merit of estrogen plus progestin therapy, the
low rates of gynecologic cancers in the population and the limited precision
in the estimated effects from this trial suggest that these results should
not have an appreciable influence on most women's decision making when seeking
relief for moderate to severe vasomotor symptoms, nor can they resolve questions
of etiology. The possibility of an increased risk of ovarian cancer incidence
and mortality remains worrisome, however, and needs confirmation. The increased
need for diagnostic procedures in response to bleeding is an added burden
and could reasonably affect a woman's decision to use these medicines. These
data provide further support for the recently revised guidelines for the use
of continuous combined estrogen plus progestin therapy.34,35