Endometrial cancer and menopausal hormone therapy in the National Institutes of Health-AARP Diet and Health Study cohort§

Authors

  • James V. Lacey Jr. MD, PhD,

    Corresponding author
    1. Hormonal and Reproductive Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    • Hormonal and Reproductive Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, 6120 Executive Boulevard, MSC 7234, Rockville, MD 20852-7234
    Search for more papers by this author
    • Fax: (301) 402-0916.

  • Michael F. Leitzmann MD, DrPH,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    Search for more papers by this author
  • Shih-Chen Chang PhD,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    Search for more papers by this author
  • Traci Mouw MPH,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    Search for more papers by this author
  • Albert R. Hollenbeck PhD,

    1. AARP, Washington, DC
    Search for more papers by this author
  • Arthur Schatzkin MD, DrPH,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    Search for more papers by this author
  • Louise A. Brinton PhD

    1. Hormonal and Reproductive Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Rockville, Maryland
    Search for more papers by this author

  • This article is a US Government work and, as such, is in the public domain in the United States of America.

  • The Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University Cancer collected incidence data from the Atlanta metropolitan area. The California Department of Health Services, Cancer Surveillance Section collected cancer incidence data from California. The Michigan Cancer Surveillance Program, Community Health Administration, State of Michigan collected cancer incidence data from the Detroit metropolitan area. The Florida Cancer Data System, under contract to the Department of Health (DOH), collected the Florida cancer incidence data used in this report

  • §

    The views expressed herein are solely those of the authors and do not necessarily reflect those of the contractor or the US Department of Health

  • The Louisiana Tumor Registry, Louisiana State University Medical Center in New Orleans collected cancer incidence data from Louisiana. The New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey State Department of Health and Senior Services Cancer collected incidence data from New Jersey. The North Carolina Central Cancer Registry Cancer collected incidence data from North Carolina. The Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg collected cancer incidence data from Pennsylvania. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions

Abstract

BACKGROUND.

Menopausal hormone therapy formulations for women without hysterectomy have included estrogen plus progestin for years, but endometrial cancer risks associated with the use of sequential and continuous estrogen-plus-progestin regimens remain unclear.

METHODS.

The National Institutes of Health-AARP Diet and Health Study included 73,211 women who were ages 50 years to 71 years at baseline and who completed 2 questionnaires (1995–1996 and 1996–1997). Linkage to state cancer registries and mortality indices identified 433 incident endometrial cancers through 2000. Using proportional hazards regression, the authors estimated relative risks (RRs) and 95% confidence intervals (95% CIs) relative to never-use of hormone therapy.

RESULTS.

In 51,312 women who never used hormones or only used estrogen-plus-progestin regimens at doses consistent with current practice, neither sequential estrogen plus progestin (daily estrogen plus progestin for 10–14 days per cycle: RR, 0.74; 95% CI, 0.39–1.40) nor continuous estrogen plus progestin (daily estrogen plus progestin for ≥20 days per cycle: RR, 0.80; 95% CI, 0.55–1.15) had any statistically significant association with endometrial cancer. Long durations (≥5 years) of sequential regimen use (RR, 0.79; 95% CI, 0.38–1.66) and of continuous regimen use (RR, 0.85; 95% CI, 0.53–1.36) were not associated with endometrial cancer.

CONCLUSIONS.

Confirmation that these estrogen-plus-progestin regimens neither increase nor decrease the risk of endometrial cancer could influence menopausal symptom management for women who are considering estrogen-plus-progestin therapy. Cancer 2007. Published 2007 by The American Cancer Society.

Increased risks of endometrial cancer among women who take unopposed estrogen menopausal hormone therapy1 led to the addition of progestin to menopausal hormone therapy formulations that are prescribed to women with intact uteri.2, 3 Progestin was added to counteract unopposed estrogen's proliferative effects on postmenopausal endometrium.4, 5 Estrogen-plus-progestin regimens that included progestin for 7 to 9 days per 28-day cycle did not appear to produce endometrial cancer risks that were lower than the risks in women who had taken unopposed estrogen.6 Estrogen-plus-progestin regimens that included ≥10 days of progestin per 28-day cycle appeared to generate no increases in endometrial cancer risk.7 These sequential or cyclic regimens became commonplace until continuous combined estrogen-plus-progestin regimens, which included both estrogen and progestin taken each day of the cycle, gained widespread acceptance.8

Compared with no use of hormone therapy, the use of continuous regimens has been associated with no increase6, 9 or a decreased10–12 risk of endometrial cancer. Three recent epidemiologic studies reported statistically significant increased endometrial cancer risks13–15 in women who used continuous estrogen plus progestin. Slightly different exposure definitions, low statistical power for specific regimens, or chance may explain the discrepancies across studies. To better understand whether endometrial cancer risk is associated with estrogen-plus-progestin regimens that contain doses of progestin that have been used widely by women in the United States, we analyzed data from a large, prospective study: the National Institutes of Health-AARP (NIH-AARP) Diet and Health Study.

MATERIALS AND METHODS

Study Population

The NIH-AARP Diet and Health Study16 began in 1995 and 1996, when a baseline questionnaire eliciting information on demographic characteristics, dietary intake, and numerous health-related behaviors was mailed to 3.5 million AARP members. Recipients were between ages 50 years and 71 years and resided in 1 of 6 states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) or 2 metropolitan areas (Atlanta, Georgia and Detroit, Michigan). The National Cancer Institute (NCI) Special Studies Institutional Review Board approved this study. All participants provided informed consent.

In total, 617,119 individuals (17.6%) returned the questionnaire, and 566,407 individuals (16.2%) satisfactorily completed the questionnaire. A second mailed questionnaire (1996–1997) collected information on family history of cancer, physical activity, and menopausal hormone therapy. In total, 337,076 individuals (59.5%) completed that questionnaire. After excluding participants who died (N = 1619) or moved out of the study area (N = 547) before their completed second questionnaires were scanned, proxy respondents to the baseline questionnaire (N = 6959) or to the second questionnaire (N = 3424), and 188,118 men, the study population included 136,409 potentially eligible women.

Exposure Ascertainment

The baseline questionnaire asked whether women currently were taking replacement hormones and, if so, for how many years. Participants also were asked about demographics, hysterectomy, oophorectomy, reproductive history, oral contraceptive use, menopausal status, family history of cancer, height, weight, and smoking.

The second questionnaire asked women who used estrogen or progestin pills to report their dates of first and last use, total duration of use, regimen, usual dose, and the name of the pill that they took for the longest time. This questionnaire asked about estrogen and progestin use separately but did not ask about the continuous combined estrogen-plus-progestin pill, which was marketed first in 1995.8

We considered reports of both estrogen and progestin pill use to be estrogen-plus-progestin-only use if the reported dates of first use were within 90 days of each other or if the reported durations of use were identical. Use of estrogen plus progestin after use of unopposed estrogen (N = 3308) or after use of unopposed progestin (N = 263), use of estrogen plus progestin before use of unopposed estrogen (N = 276) or use of unopposed progestin (N = 187), use of unopposed estrogen before use of unopposed progestin (N = 234), and use of unopposed progestin before use of unopposed estrogen (N = 168) constituted separate exposure categories.

Sequential regimens included estrogen plus progestin used for ≤15 days per cycle, which we stratified further into <10 days per cycle (N = 2167) or 10 to 14 days per cycle (N = 5333). Continuous estrogen-plus-progestin regimens included estrogen plus progestin used for 20 to 25 days per cycle (N = 2049) or every day of the cycle (N = 9762). Only 398 women reported using progestins for 15 to 19 days per cycle.

Cohort Follow-up

Two annual linkages provided follow-up information. For changes of address, the cohort was matched to the National Change of Address database (maintained by the U.S. Postal Service) and also was updated based on undeliverable mail, other address change update services, and participants' notifications. For vital status, the Social Security Administration Death Master File identified cohort members who were presumed deceased. A follow-up search of the National Death Index (NDI) Plus provided verification.

Incident Cancers

A probabilistic linkage to 8 state cancer registries used participant-provided name(s), address history, sex, date of birth, and Social Security numbers to identify incident cancers and associated dates of diagnosis with an estimated 90% completeness.17 Dates and causes of death from the NDI linkage also identified fatal cancers.

Analytic Population

To use the detailed hormone therapy data, our analysis included the 136,409 women who completed the second questionnaire. We excluded 9039 women who reported a personal history of cancer other than nonmelanoma skin cancer on either questionnaire, 51,505 women who reported a prior hysterectomy at baseline, and 1497 women whose hysterectomy status was unknown at baseline. We also excluded women who reported that their menstrual periods stopped because of chemotherapy or radiation (N = 65) or surgery (N = 1056) and 36 women who developed nonepithelial endometrial cancer during follow-up. Therefore, the current analysis included 73,211 women.

Through December 31, 2000, 433 women developed endometrial cancer. Most cancers (N = 394) were endometrioid adenocarcinomas, but 28 cancers were serous tumors, and 11 tumors had unknown or unspecified histologic type.

Statistical Analysis

We estimated relative risks (RRs) by using Cox proportional hazards regression (using SAS software, version 8.2; SAS Institute, Inc., Cary NC), with age as the time scale and ties handled by complete enumeration.18 We observed no departures from proportional hazards assumptions. Follow-up began when a completed second questionnaire was scanned and ended on December 31, 2000, or the earliest of the following dates: participant moved out of the registry catchment areas, withdrew from the study, was diagnosed with endometrial cancer, or died from any cause.

All regression models used women who reported no hormone therapy use as the referent group and used mutually exclusive formulation or regimen categories. For combined exposures within formulations (eg, recency and duration), we created single variables based on cross-tabulations of original variables. Estrogen-plus-progestin analyses first were limited to women who never used hormones (N = 37,870) or who only used estrogen plus progestin (N = 20,837). Then, to analyze specific regimens, we restricted analyses further to women who used sequential regimens (progestin for 10–14 days per cycle) with progestin doses of 5.0 mg per day or 10.0 mg per day (N = 4204) or continuous regimens (progestin for ≥20 days per cycle) with progestin doses of 2.5 mg per day or 5.0 mg per day (N = 9238). These analyses excluded estrogen-plus-progestin-only users whose regimen was unknown (N = 1128) or differed from historic prescribing patterns in the United States3: 1) sequential estrogen plus progestin with a progestin dose <1.0 mg per day (N = 254) or 2.5 mg per day (N = 1629), unknown (N = 1349), or other (N = 64); 2) continuous estrogen plus progestin with a progestin dose <1.0 mg per day (N = 413), 10.0 mg per day (N = 159), unknown (N = 1861), or other (N = 140); or 3) progestins used 15 to 19 days per month (N = 398). All regimen-specific RRs were obtained from models that included terms for both regimens.

Compared with no hormone therapy use, estrogen-plus-progestin use was associated positively with younger age, premenopausal status, white race/ethnicity, college education, oral contraceptive use, former smoking, body mass index (BMI) (based on self-reported height and weight at baseline) <25 kg/m2, and self-reported excellent or very good health at baseline. After considering all available confounders, we chose a parsimonious combination of variables that had an association with both exposure and outcome and changed the hormone therapy parameter estimates compared with estimates from models that were adjusted only for age at entry. Our statistical models adjusted for continuous age at entry (in years), race/ethnicity (white vs nonwhite or unknown), oral contraceptive use (ever vs never), and menopausal status at baseline (for values, see Table 1) and were adjusted for BMI (<25 kg/m2, 25–29 kg/m2, or ≥30 kg/m2) through stratification. Additional adjustment for other factors, including calendar time, did not change the results.

Table 1. Risk Factors for Endometrial Cancer Among 73,211 Women
Risk factorNo. of cancersPerson-yearsRR*95% CI
  • RR indicates relative risk; 95% CI, 95% confidence interval; Ref, reference group; BMI, body mass index.

  • *

    RRs were adjusted for age, oral contraceptive use, BMI, race, and menopausal status.

Age at study entry, y
 <575959,941.9  
 57–607655,872.2  
 61–649663,381.8  
 65–6812573,188.3  
 ≥697737,065.3  
Race/ethnicity
 Caucasian403266,345.51.00Ref
 Other/unknown3023,103.90.790.54–1.14
Menopausal status at baseline
 Premenopausal1617,583.70.930.52–1.65
 Natural menopause, before age 45 y3529,743.00.750.52–1.07
 Natural menopause, ages 45–49 y9571,900.10.890.70–1.14
 Natural menopause, ages 50–54 y196130,357.21.00Ref
 Natural menopause, ages ≥55 y6929,479.51.441.10–1.90
BMI at baseline, kg/m2
 <25133134,749.31.00Ref
 25–2911488,208.31.291.00–1.65
 ≥3017458,701.53.012.40–3.78
Oophorectomy at baseline
 No399272,208.91.00Ref
 Yes61116.63.541.58–7.95
 Other ovarian surgery2615,164.11.180.80–1.76
Oral contraceptive use
 Never300170,533.51.00Ref
 Ever130117,131.80.740.59–0.91
Smoking history
 Never227129,565.71.00Ref
 Former164118,175.60.810.66–0.99
 Current3739,075.90.660.46–0.94
Parity
 Nulliparous9451,511.61.00Ref
 14930,672.40.900.64–1.28
 211475,803.90.850.65–1.12
 ≥3173130,536.90.650.50–0.84
Age at menarche, y
 ≤12223137,181.31.00Ref
 13–14179124,767.70.940.77–1.15
 ≥153026,658.20.750.51–1.09
Family history of breast cancer
 No323208,373.61.00Ref
 Yes4535,257.80.810.59–1.10

RESULTS

The 73,211 women accrued 289,449 person-years. The mean durations of follow-up in women who developed endometrial cancer and women who did not develop endometrial cancer were 2.12 years (range, from 11 days to 4.08 years) and 3.95 years (range, from 1 day to 4.17 years), respectively. Five women developed endometrial cancer, and 40 women were censored within 1 month of entry. The mean ± standard deviation ages at entry and exit were 62.4 ± 5.5 years and 66.4 ± 5.5 years, respectively.

Most women were white, postmenopausal, and in their 60s when follow-up began. Women who were overweight (BMI 25–29 kg/m2) or obese (BMI ≥30 kg/m2) at baseline contributed >50% of the total person-time. Endometrial cancer was associated positively with BMI and later age at natural menopause; it was associated inversely with duration of oral contraceptive use, increasing parity, smoking, later age at menarche, and non-Caucasian race/ethnicity; and it was not associated with a family history of breast cancer (Table 1). The standardized incidence ratio (SIR) for endometrial cancer compared with the NCI's Surveillance, Epidemiology, and End Results (SEER) rates (ages 50–79 years) was 1.42 (95% confidence interval [95% CI], 1.29–1.57); the higher rates in our analysis are consistent with the increase in rates that has been reported when endometrial cancer rates in SEER are corrected for hysterectomy status.19

Unopposed Estrogen

Almost 7% of women (N = 4990) reported having used only unopposed estrogen, and 35 of those women developed endometrial cancer (RR, 1.27; 95% CI, 0.89–1.82) (Table 2). Long-duration use (≥5 years) of unopposed estrogen was uncommon but had a statistically significant association with endometrial cancer. Current long-duration use of unopposed estrogen generated a similar positive association.

Table 2. Associations Between Unopposed Estrogen and Endometrial Cancer in 73,211 Women Without Hysterectomy at Baseline
HTNo. of cancersPerson-yearsRR*95% CIRR95% CI
  • HT indicates hormone therapy; RR, relative risk; 95% CI, 95% confidence interval; ET, estrogen therapy; EPT, estrogen-plus-progestin therapy; Ref, reference group.

  • *

    RRs were adjusted for age.

  • RRs were also adjusted for race, body mass index, menopausal status, and oral contraceptive use.

  • Twenty cancers developed in women who reported using both formulations but did not report enough information to determine which formulation was used first. Six cancers developed in women who reported use of other combinations of EPT, ET, or unopposed progestin.

No HT use225149,593.71.00Ref1.00Ref
ET only3519,618.51.160.81–1.661.270.89–1.82
ET then EPT5113,086.02.561.89–3.483.022.21–4.12
Recency of ET-only use
 Former2312,871.31.130.74–1.741.230.80–1.89
 Current126237.71.330.75–2.381.500.84–2.68
Duration of ET-only use, y
 <51615,802.80.670.40–1.110.720.43–1.20
 6–961,259.73.021.34–6.793.391.51–7.65
 ≥10122209.13.331.86–5.974.072.27–7.31
Recency of ET-only use
 Former2312,871.31.130.74–1.741.230.80–1.89
 Current, y
  <544363.10.660.25–1.770.730.27–1.96
  ≥581852.72.741.35–5.553.221.59–6.54

Fewer women (N = 3308; 4.5%) reported having used unopposed estrogen before using estrogen plus progestin (RR, 3.02; 95% CI, 2.21–4.12). The mean ± standard deviation durations of unopposed estrogen use and estrogen plus progestin use were 6.9 ± 7.1 years and 6.9 ± 5.7 years, respectively.

Estrogen Plus Progestin

Eighty-one cancers developed among the women who reported using only estrogen plus progestin (RR, 0.93; 95% CI, 0.85–1.02) (Table 3). Use for ≤1 year had a significant inverse association with endometrial cancer. Other durations of use were not associated with endometrial cancer except for a weak and statistically nonsignificant, positive association with ≥10 years of use. Most women who used estrogen plus progestin and developed endometrial cancer were current users at baseline. Current use was not associated with endometrial cancer, but former use had a statistically significant inverse association with endometrial cancer. The RR for current, long-duration use was 1.43 (95% CI, 0.97–2.12; N = 32 cancers and 18,912 person-years). None of the reported progestin doses had a significant association with endometrial cancer (data not shown).

Table 3. Associations Between the Use of Only Estrogen Plus Progestin and Endometrial Cancer in Women Without Hysterectomy at Baseline
HTAll EPT-only Use, N = 58,707
No. of cancersPerson-yearsRR*95% CIRR95% CI
  • HT indicates hormone therapy; RR, relative risk; 95% CI, 95% confidence interval; EPT, estrogen-plus-progestin therapy; Ref, reference group.

  • *

    RRs were adjusted for age.

  • RRs were adjusted for age, race, body mass index, menopausal status, and oral contraceptive use.

No HT use225149,593.71.00Ref1.00Ref
EPT-only use8182,648.60.880.81–0.960.930.85–1.02
Duration of EPT-only use, y
 ≤1413,916.30.210.08–0.550.220.08–0.60
 2–42322,665.40.750.48–1.160.850.54–1.32
 5–92225,551.10.600.38–0.930.710.45–1.12
 ≥103220,374.61.010.69–1.461.310.89–1.94
Recency of EPT-only use
 Former714,4520.330.16–0.710.360.17–0.77
 Current7467,596.20.770.59–1.010.920.70–1.23
Recency and duration, y
 Former
  <5510,447.50.330.14–0.810.360.15–0.88
  ≥523912.60.340.09–1.370.380.09–1.54
 Current
  <52225,816.70.640.41–1.010.730.47–1.15
  ≥55241,779.50.840.62–1.141.050.76–1.45
Days on progestin
 <10138593.41.050.60–1.841.270.72–2.25
 10–142021,205.30.670.42–1.070.800.49–1.30
 15–1911582.20.430.06–3.100.500.07–3.59
 20–2588123.30.660.33–1.330.790.39–1.60
 Daily3128,665.40.560.33–0.820.650.44–0.96

Neither progestin use for <10 days per cycle (RR, 1.27; 95% CI, 0.72–2.25) nor progestin use for 10 to 14 days per cycle (RR, 0.80; 95% CI, 0.49–1.30) was associated significantly with endometrial cancer. Ever-use of daily progestin had a statistically significant inverse association with endometrial cancer (RR, 0.65; 95% CI, 0.44–0.96), but increasing duration of estrogen plus daily progestin use did not (<5 years: RR, 0.54; 95% CI, 0.31–0.95; 6–9 years: 0.79, 95% CI, 0.43–1.47; ≥10 years: RR, 0.73; 95% CI, 0.34–1.56).

Ever-use of estrogen plus progestin at standard regimen-dose combinations (Table 4) was not associated with endometrial cancer (RR, 0.92; 95% CI, 0.83–1.03). The RRs for both sequential and continuous regimens were below the null and were not statistically significant. No endometrial cancers occurred in former sequential regimen users or former continuous regimen users. Endometrial cancer was not associated with current use of sequential regimens (RR, 0.80; 95% CI, 0.42–1.52) or continuous regimens (RR, 0.87; 95% CI, 0.60–1.26). For both regimens, duration of use was not associated with endometrial cancer. The associations with ever-use of sequential regimens that contained medroxyprogesterone acetate (MPA) at doses of 5 mg per day (RR, 0.81; 95% CI, 0.32–2.03) and 10 mg per day (RR, 0.67; 95% CI, 0.29–1.57) were similar. These analyses excluded women who used progestin for <10 days per cycle at doses of 5 mg per day or 10 mg per day (N = 7 cancers and 3190 person-years; RR, 1.90; 95% CI, 0.90–4.22).

Table 4. Associations Between the Use of Only Estrogen Plus Progestin in Usual Regimen/Dose Combinations and Endometrial Cancer in Women Without Hysterectomy at Baseline
HTEPT Only in Usual Regimen/Dose Combinations, N = 50,506*
No. of cancersPerson-yearsRR95% CIRR95% CI
  • HT indicates hormone therapy; RR, relative risk; 95% CI, 95% confidence interval; EPT, estrogen-plus-progestin therapy; Ref, reference group; NA, not available.

  • *

    Sequential regimens with 5 mg per day or 10 mg per day medroxyprogesterone acetate or continuous regimens with 2.5 mg per day or 5 mg per day medroxyprogesterone acetate.

  • RRs were adjusted for age.

  • RRs were adjusted for age, race, BMI, menopausal status, and oral contraceptive use.

  • §

    Sequential regimens include progestin taken for 10–14 days per month. Continuous regimens include progestin taken for ≥20 days per month.

No HT use225149,593.71.00Ref1.00Ref
EPT-only use4650,145.40.860.77–0.960.920.83–1.03
Regimen§
 Sequential1113,532.60.580.31–1.060.740.39–1.40
 Continuous3536,612.90.660.46–0.950.800.55–1.15
Regimen and duration, y
 Sequential
  ≤10974.6NANANANA
  2–433491.10.670.21–2.110.850.27–2.70
  ≥589066.90.600.30–1.220.790.38–1.66
 Continuous
  ≤115727.20.120.02–0.880.140.02–1.01
  2–41411,054.60.920.53–1.591.070.61–1.85
  ≥52019,822.60.670.43–1.070.850.53–1.36
Selected regimens and doses, mg
 Sequential with 557188.90.600.25–1.450.810.33–2.03
 Sequential with 1069533.50.540.24–1.230.670.29–1.57
 Daily progestin with 2.51816,146.70.770.48–1.260.950.58–1.55

Women who reported taking daily conjugated equine estrogens at a dose of 0.625 mg per day plus daily MPA at a dose of 2.5 mg per day—equivalent to the single-pill, continuous-combined estrogen-plus-progestin regimen—accrued 16,147 person-years. Eighteen of those women developed endometrial cancer (RR, 0.95; 95% CI, 0.58–1.55). Ever-use of progestin every day of the cycle at a dose of 2.5 mg per day or 5 mg per day was not associated significantly with endometrial cancer (RR, 0.77; 95% CI, 0.52–1.14; N = 29 endometrial cancers). The associations with increasing duration of use were also null (<5 years: RR. 0.71; 95% CI, 0.40–1.26; 6–9 years: RR. 0.78; 95% CI, 0.40–1.54; ≥10 years: RR, 0.89; 95% CI, 0.42–1.89).

The average reported age at first use of estrogen plus progestin only was younger both for women who developed endometrial cancer (49.6 years) and for censored women (51.5 years) who used sequential regimens compared with the average age of women who developed endometrial cancer (54.4 years) and censored women (53.9 years) who used continuous regimens.

The crude cumulative incidence of endometrial cancer was 15.0 per 10,000 person-years. Incidence was identical in women who reported no hormone therapy use (15.0 per 10,000 person-years) and women who reported any use of estrogen or progestin (14.9 per 10,000 person-years). Current, long-duration, unopposed estrogen users had the highest incidence (59.5 per 10,000 person-years). Among women who reported using estrogen plus progestin at usual regimen-dose combinations, the lowest incidence occurred among short-duration sequential regimen users (5.6 per 10,000 person-years), the highest incidence occurred among long-duration sequential regimen users (13.2 per 10,000 person-years), and incidence for other exposures was between 8 and 12 per 10,000 person-years.

Among women who never used menopausal hormone therapy, the incidence of endometrial cancer was 6 times higher in obese women (BMI ≥30 kg/m2) than in normal-weight women (BMI <25 kg/m2; 33.7 vs 5.7 per 10,000 person-years, respectively) (Fig. 1). Estrogen-plus-progestin use, regardless of the regimen, did not have a statistically significant association with endometrial cancer in normal-weight women or overweight women (BMI 25–29 kg/m2; data not shown). Most endometrial cancers among estrogen-plus-progestin users occurred in women who were not obese (ie, BMI <30 kg/m2), but most endometrial cancers among nonusers occurred in obese women. Obese women comprised the only subgroup in which estrogen-plus-progestin use was associated inversely with endometrial cancer (RR, 0.50; 95% CI, 0.24–1.02).

Figure 1.

Cumulative incidence of endometrial cancer by hormone therapy use within body mass index (BMI) groups (in kg/m2). CEPT indicates continuous estrogen plus progestin; SEPT, sequential estrogen plus progesterin.

DISCUSSION

In this large, prospective study, estrogen-plus-progestin use was not associated consistently with the risk of developing endometrial cancer. These data do not support the previously reported increased or decreased endometrial cancer risks among continuous estrogen-plus-progestin users.

Certain estrogen-plus-progestin exposures generated statistically significantly reduced risks in our initial analyses. For some exposures, those risks lost their statistical significance after adjustment for confounders. For others (eg, former use and short-duration use), associations were based on few women who developed endometrial cancer and lacked the expected patterns of stronger (inverse) associations among women with more substantial exposures (eg, current or long-duration use). For others (eg, estrogen plus daily progestin), further analyses that were restricted to standard regimen-dose combinations yielded RRS that were only slightly below the null and were not statistically significant.

These findings highlight our study's large sample size and its ability to investigate specific contemporary formulations, regimens, and doses. We limited our estrogen-plus-progestin analyses to women who had only used this formulation. We considered most known potential confounders, although we cannot rule out possible residual positive or negative confounding.

To interpret our findings, the limitations of this study also must be considered. Focusing on the detailed exposure data increased the internal and overall validity but reduced our effective sample size. Our questionnaire queried the pill names and doses that participants used most often; therefore, we could not document less frequently used hormones. We lacked information on hormone therapy use after the second questionnaire, but the short follow-up minimized potential exposure misclassification after study entry. We could not identify which women had undergone a hysterectomy after baseline. Overall hysterectomy prevalence in the United States increased during the study period, but the decline in hysterectomy after age 55 years20 reduces the expected number of NIH-AARP study participants who had hysterectomy after baseline. The response to the first mailed questionnaire was low, but 62% of baseline respondents completed the second questionnaire. Compared with nonrespondents, respondents to the second questionnaire were more likely to be older and non-Hispanic white and were more likely to report their overall health as excellent or very good, a BMI <25 kg/m2, and hormone therapy use. Response was not associated with hysterectomy status, parity, or menopause; therefore, we expect that our results will be acceptably generalizable to women in the United States aged >50 years.

The Million Women Study (MWS), which is a prospective study of >700,000 postmenopausal women in the United Kingdom with intact uteri and no history of cancer, recently reported that use of cyclic (sequential) regimens was not associated significantly with endometrial cancer.11 Its summary of then-published literature showed a statistically significant increased risk (RR, 1.14; 95% CI, 1.01–1.28) for ever-use of sequential regimens compared with never-use.11 Of the 6 studies that were included in the summary estimate, the only study10 that reported a significantly increased risk allowed for previous unopposed estrogen use, which is a strong endometrial cancer risk factor. Our recent cohort study also reported significantly increased risks associated with sequential regimens,14 which were defined as progestin taken for <15 days per month but could have included progestin used for <10 days per month. Adding the NIH-AARP findings to the summary estimate reported in the MWS may decrease the magnitude of the association between endometrial cancer and ever-use of sequential regimens that contain ≥10 days of progestin per cycle.

The MWS also reported a significantly decreased risk among ever-users of continuous combined estrogen plus progestin (RR, 0.71; 95% CI, 0.65–0.90).11 A summary estimate that included 5 other studies was slightly attenuated and was not statistically significant (RR, 0.88; 95% CI, 0.75–1.03).11 Our statistically nonsignificant risk estimate for use of continuous combined regimens (ie, progestins used each day of the cycle) at usual doses (RR, 0.77; 95% CI, 0.52–1.14) fell between those 2 risks. Another United States study with more exposed women (N = 94) reported a nonsignificant, positive association with endometrial cancer7; whereas, in our recent cohort study, we observed a significantly increased endometrial cancer risk among women who used continuous estrogen plus progestin (progestin for ≥15 days per month) based on 15 women who developed endometrial cancer.14 Apart from statistical chance, obvious explanations for these reported inconsistencies have not emerged.

Other cohort14 and case-control7, 9, 10, 12, 13, 15, 21–23 studies have employed variable exposure definitions for estrogen plus progestin. The studies collectively captured different snapshots of the hormone therapy usage patterns that have changed substantially since the middle 1970s. The design similarities between the MWS and our NIH-AARP Diet and Health Study cohort, therefore, offer a rare comparison between similar study cohorts (eg, women in their 50s and 60s, data collection in the late 1990s, and similar duration of follow-up). Formulations used in Europe and the United States include different constituent estrogens and progestins. The cyclic (sequential) regimens that MWS participants reported using were based on 19-nortestosterone derivative progestins, such as northisterone, whereas almost all progestin use in our study was based on 17-hydroxyprogesterone derivatives, such as MPA. The MWS risk estimate in analyses that were restricted to MPA-based, continuous, combined regimens (RR, 0.63; 95% CI, 0.63–0.43–0.93; N = 27 cancers)11 was lower than the equivalent estimate in our analysis (RR, 0.80; 95% CI, 0.55–1.15; N = 29 cancers). Like the MWS data, our data showed that obese women were the only subgroup in which endometrial cancer incidence was noticeably lower among estrogen-plus-progestin users than among nonusers. However, the risks among obese estrogen-plus-progestin users still were higher than the risks among nonobese women, regardless of hormone therapy use.

In our study, the prevalence of estrogen-plus-progestin use (especially sequential regimens) sharply declined as BMI increased. If the lower risks among obese women who used estrogen plus progestin arise because progestins inhibit the increased endometrial proliferation that accompanies higher circulating estrogen levels in heavier women,11 then estrogen-plus-progestin use also should reduce the risk in overweight women (and not just in obese women). However, no consistent pattern appeared for overweight women in our data or in the MWS data. Replication of these findings would provide useful additional information on the influence of exogenous hormones and BMI on endometrial cancer risk.

The lower risks in former users compared with current users and in short-duration users compared with long-duration users cast doubt on the causal nature of the potential decreased risks among continuous regimen users in our study. We did not update exposure information after 1996 to 1997, but other data from the United States during those years24 suggest that most users would have been expected to continue using hormone therapy during follow-up. Cessation of use has been associated positively with side effects, such as vaginal bleeding, which can occur shortly after the initiation of hormone therapy.25 If such bleeding reflects underlying endometrial hyperplasia,6, 26 which is a marker of increased endometrial cancer risk,27 then former use may be associated positively with endometrial cancer. Alternatively, if vaginal bleeding prompts clinical examinations that detect endometrial lesions before invasion or leads to successful hormone treatment, then former use may be associated inversely with endometrial cancer. These dynamics14 open doors for detection bias or confounding by indication to influence the observed associations. Future studies that systematically collect data on reasons for initiation and cessation of hormone therapy (plus subsequent gynecologic conditions) could address these issues.

We did not ask directly about the continuous combined estrogen-plus-progestin pill that was tested in the Women's Health Initiative estrogen-plus-progestin trial,26 in which women who were randomized to receive estrogen plus progestin were nonsignificantly less likely than women who were randomized to receive placebo to develop endometrial cancer (hazard ratio, 0.81; 95% CI, 0.49–1.36; N = 27 vs N = 31 endometrial cancers, respectively). An equivalent dose-regimen combination was not associated with endometrial cancer in our study. Detection, treatment, and censoring of candidate endometrial cancer precursor lesions in randomized clinical trials of menopausal hormone therapy make direct comparisons between observational studies and clinical trials difficult for menopausal hormone therapy and endometrial cancer.14 It seems unlikely that future randomized clinical trials will be designed specifically to evaluate endometrial cancer risks associated with long-duration use of today's most common estrogen-based and progestin-based menopausal hormone therapies. Therefore, continued follow-up of observational studies will provide useful data for risk assessment.

The increased risks among long-duration, unopposed estrogen users in our analysis are consistent with extensive previous data.1 Data on endometrial cancer risk in women who switched from unopposed estrogen to other formulations are more limited. Because endometrial cancer risk remains elevated in former unopposed estrogen users for years after last use,14 we hypothesize that the significantly increased risks in women who used estrogen plus progestin after using unopposed estrogen in our study is because of the prior use of unopposed estrogen rather than the use of estrogen plus progestin.

In conclusion, endometrial cancer risk among women who used continuous estrogen-plus-progestin regimens was neither decreased nor increased significantly. Although accumulating data indicate that the use of estrogen plus progestin does not influence endometrial cancer risk, continued attention to the unresolved epidemiologic and clinical issues associated with endometrial cancer and potential precursors among hormone therapy users could improve the understanding of endometrial carcinogenesis.

Acknowledgements

We thank Leslie Carroll, Matthew Bucher, and Dave Campbell at IMS, Inc., Silver Spring, Maryland for data support.

Ancillary