Hormone replacement therapy and incidence of hormone-dependent cancers in the Norwegian Women and Cancer study

Authors


  • The sponsors had no role in the study design, data analysis, interpretation of the results, writing of the report or the decision to submit the report for publication. No conflicts of interest declared.

Abstract

Increasing use of HRT over the last 2 decades could have contributed to the increasing incidence of cancer in women. Our aim was to investigate the relation between use of HRT and risk of hormone-dependent cancers in a Norwegian cohort of women. The Norwegian Women and Cancer (NOWAC) study is a representative, national, population-based cohort study. This report includes 35,456 postmenopausal women aged 45–64 years who answered a postal questionnaire in 1996–1998 providing information on reproduction, lifestyle and use of HRT. The women were followed up for cancer incidence. The main analyses were restricted to 31,451 postmenopausal women with complete information. Ever use of HRT was reported by 43.5% and current use, by 35% of the women. Current users had an increased risk of breast cancer (adjusted RR = 2.1, 95% CI 1.5–2.5). The risk increased with increasing duration of use (ptrend < 0.0001). Using a regimen of continuous estrogen–progestagen implied an increased risk. Adjusted RRs associated with <5 and ≥5 years' duration of use were 2.6 (95% CI 1.9–3.7) and 3.2 (95% CI 2.2–4.6), respectively. The population-attributable risk of breast cancer due to current use of HRT was 27%. We found no significant increase in risk of ovarian cancer. Neither did we find users of estrogen–progestagen preparations to have any increase in risk of endometrial cancer. Our results suggest that HRT could be considered a major determinant for the increasing incidence of breast cancer in Norway. © 2004 Wiley-Liss, Inc.

HRT had a breakthrough in 1966 with the publication of Feminine Forever.1 According to the author, the main objective of estrogen use was “to slow down the aging rate and to assure the general systemic health of postmenopausal women, the distinctly sexual benefits of estrogen treatment should not be underrated”. Although most women start using HRT to ameliorate climacteric complaints,2 promising results from observational studies during the last 2 decades3, 4, 5, 6, 7, 8, 9 probably have encouraged more women to start and the duration of use to be extended. With the number of postmenopausal women worldwide expected to mount to 1.2 billion by 2030,10 the potential user group for HRT is enormous.

Breast cancer is the most common cancer among women worldwide. Research has shown risk to increase by diagnosed breast cancer in first-degree relatives as well as early menarche, parity, old age at first delivery, late menopause and high postmenopausal BMI. Use of HRT is associated with an increase in breast cancer risk.11 In 1997 the Collaborative Group on Hormonal Factors in Breast Cancer reanalyzed data from 51 epidemiologic studies and found use of HRT to increase the RR of having breast cancer diagnosed by 2.6% each year. With duration of use ≥5 years, the RR of having a breast cancer diagnosis was 1.34.12 Later published results from randomized clinical studies are consistent with these estimates.13 Recent published results from the MWS showed a significant increase in the RR of breast cancer with increasing total duration of use in current users of HRT. The effect was greater for combinations of estrogen–progestagen (RR = 2.00) than for estrogen-only (RR = 1.30).14

In Norway, the range of HRT preparations differs from what we find in the United States and most European countries. Conjugated estrogens have never been available on the Norwegian market, where estradiol dominates, mostly as combinations of estrogen–progestagen with norethisterone as the preferred progestagen. The androgen tibolone was introduced in 1999. Sales of HRT tremendously increased during the 1990s. The prevalence of use in Norway increased in the same period, and by the end of the decade every second woman 55–59 years old had ever used HRT.15 Sales peaked in 2001, only to culminate in 2002 after publication of the results from the WHI trial (Fig. 1).16

Figure 1.

Sales of products with estrogen in Norway 1985–2002 (except OCs).

Our aim was to investigate the relation between use of HRT and risk of hormone-dependent cancers in a Norwegian population-based cohort of postmenopausal women aged 45–64 years.

Abbreviations

BMI, body mass index; CI, confidence interval; FTP, full-term pregnancy; HRT, hormone replacement therapy; MWS, Million Women Study; NOWAC, Norwegian Women and Cancer study; OC, oral contraceptive; PAR%, population-attributable risk percent; RR, relative risk; WHI; Women's Health Initiative.

MATERIAL AND METHODS

The NOWAC study was set up in 1991 as a representative, national, population-based cohort study. The study design, material and procedures as well as aspects of external validity have been described and discussed elsewhere.17 The study population is made up of 2 subsamples: one of 46,978 women who participated in 1996–1997 and one of 44,852 women who first participated in 1991–1992 and then filled in a second questionnaire with identical questions in 1998. After exclusions and age restriction (45–64 years), the study population consisted of 67,336 participants (Fig. 2). Follow-up information was based on linkage to the Cancer Registry of Norway using the unique national identification number given all Norwegians.

Figure 2.

Enrollment of women in the study.

The questionnaires had a common core of questions covering general information (height, weight, income, family, health and lifestyle habits) and special information about menstruation and childbearing as well as use of OCs and information on HRT use. We asked for specific information on ever/current/never use, age at first use, brand name and duration of use for each continuous period of use of the same brand and total duration of use. A pamphlet with photos of almost all the different HRT brands available in Norway since 1953 was enclosed with the questionnaire to help the women recall use of these drugs.

Menopausal status

Menopausal status at recruitment was defined for most women by their answers to whether they still had regular periods, if the periods were irregular or had stopped or whether the woman did not know (due to medicine use, exercise, diseases, etc.). If the periods had stopped, women were asked to give the reason why (natural stop, bilateral oophorectomy, hysterectomy or other reasons) as well as their age when the periods stopped. A woman was considered postmenopausal when stating that her periods had stopped and/or giving the reason for and/or the age when stopping.

However, hysterectomized women and those using HRT before menopause often have a masked menopausal status. To classify these women, we used the same conventions as in the MWS:14 women 53 years or older were classified as postmenopausal (as were 92% of the study population who had not had a hysterectomy). At age 45–52, such women were classified as having an unknown menopausal status. Time since menopause in these women was assumed to be no different from that in women with natural menopause, i.e., <5 years in women aged 53–55, 5–9 years in women aged 55–59 and ≥10 years in women 60 years or older. Women with unknown or irregular status aged 53 or more were also considered postmenopausal. Women with regular periods and no stop information were considered premenopausal. Women with irregular periods or more than one alternative answer were considered perimenopausal if not hysterectomized before age 53.

Women who were younger than 53 and hysterectomized or had started HRT before menopause and women not classified elsewhere were considered as having unknown status.

Current and former users of HRT

Women indicating use of HRT at enrollment were classified as current users. If total duration of use added to the woman's age when starting HRT was equal to her age at enrollment, we also considered her a current user. Former users were ever-users not indicating current use and ever-users with missing values on recency of use.

Ethics

Approval was granted from the Data Inspectorate and the Regional Committee for Medical Research Ethics in Northern Norway. Women gave informed consent for linkage to the Cancer Registry of Norway and the register of death certificates in Statistics Norway.

Data analysis

The Cox proportional hazards model was applied to perform statistical analysis, using the SAS (Cary, NC) software package, version 8.02. RRs and 95% CIs were estimated. Age was used as the primary time variable. In the multivariate models, we included time since start of menopause (0–4, 5–9, ≥10 years), age at menarche (≤12, 13–14, ≥14 year), ever use of OCs (yes/no), BMI (<20, 20–24.99, 25–29.99, ≥30 kg/m2), history of breast cancer in mother (yes/no), regions with a screening program (yes/no), age at first birth (≤20, 21–24, ≥25 years) and parity (0, 1, 2, 3+) as possible confounders. Subjects with missing values for the covariates were excluded from analysis. Models were run with and without alcohol consumption as a covariate. When including this variable, more than 5,000 subjects were excluded due to missing values (one of the questionnaires did not include questions about alcohol consumption). Since inclusion of alcohol consumption did not change the risk estimates for HRT use, the variable was omitted from the models presented here. Never-users of HRT were used as the reference group in the different models. Tests for trend were calculated by introduction of ordinal variables obtained by assigning consecutive integers to values of the categorized variable.

RESULTS

The results are based on follow-up of 67,336 women aged 45–64 years and include 624 incident breast, 129 ovary and 127 endometrial cancers (Table I).

Table I. Study Population According to Menopausal Status, Ever Use of HRT and Numbers of Incident Cancers: the Nowac Study
Menopausal statusNumber of womenEver use of HRTIncident cancers
%MissingBreastOvaryEndometrium
Premenopausal16,6527.78121432916
Perimenopausal6,05414.919747618
Postmenopausal35,45642.91,0793428685
 Time since menopause (years)      
 < 513,51043.52191103124
 5–99,56154.0207972528
 ≥ 1012,12433.26391343033
 Unknown26150.214100
Unknown menopause9,17478.11149288
 Information missing1,47740.4751111
 Age 45–52 years and hysterectomy before menopause2,03942.5391640
 HRT before menopause5,658100.06537
Total67,33636.72,202624129127

To minimize the problem of misclassification of menopausal status in the main analysis, the study sample was restricted to postmenopausal women. Overall, 97% of the postmenopausal women gave information on HRT use. The proportion of current users was 35%. The majority of women used oral preparations. Transdermal preparations were used by <8% of women. Among current users, 56% had used only one HRT preparation and 78% had used the last preparation for more than half their total duration of HRT use. Mean age of the study population was 53.0 years and mean BMI, 25.0 kg/m.2

Overall use and total duration of use in relation to breast cancer

Among the 31,451 postmenopausal women with complete information on ever/never use of HRT, we found the RR of breast cancer to be significantly higher in ever-users than in never-users (RR = 1.9, 95% CI 1.5–2.5). The RR was increased in current, but not in former, users (test for heterogeneity p = 0.002). Increasing the total duration of use significantly increased the risk of breast cancer in current and ever-users, while a corresponding trend could not be found among former users (Table II).

Table II. Estimated RR1 With 95% CI OF Breast Cancer in Relation to Recency and Duration of HRT Use: the Nowac Study
HRT use and total durationPopulation/cases2 (postmenopausal)RR (95% CI)ptrend
  • 1

    Adjusted for age, BMI, age at menarche, ever use of OCs, time since menopause, family history of breast cancer, mammography, parity and age at first delivery.

  • 2

    Numbers may not add up due to missing values for one or more variables.

Use of HRT   
 Never17,756/1301.0 
 Ever13,695/1761.9 (1.5–2.5) 
 Current10,916/1512.1 (1.7–2.7) 
 Former2,618/191.0 (0.6–1.6) 
Total duration of use (years)   
 Ever-users   
  0–13,789/351.4 (1.0–2.1) 
  2–44,899/682.1 (1.6–2.9) 
  5–93,233/442.2 (1.5–3.1) 
  10+1,214/202.2 (1.4–3.6)<0.0001
 Current users   
  0–12,245/231.7 (1.1–2.6) 
  2–44,228/642.4 (1.8–3.3) 
  5–93,016/432.3 (1.6–3.3) 
  10+1,131/161.9 (1.1–3.2)<0.0001

The estimated RR of breast cancer among current users of HRT decreased from 2.7 (95% CI 1.5–4.6) in the first year to 1.9 (95% CI 1.0–3.5) in the fourth year of follow-up.

Type, regimen and duration of use

Focusing on the last period of use, we found the RR of breast cancer to be 2.5 (95% CI 1.9–3.2) in current users of combinations of estrogen–progestagen compared to 1.8 (95% CI 1.1–2.9) in users of estrogen only (test for homogeneity p = 0.17) (Table III).

Table III. Estimated RR1 With 95% CI of Breast Cancer in Current Users in Relation to Type of HRT Used at Baseline: the Nowac Study
Type of HRT usedPopulation/casesRR (95% CI)
  • 1

    Adjusted for age, BMI, age at menarche, ever use of OCs, time since menopause, family history of breast cancer, mammography, parity and age at first delivery.

Estrogen only1,542/181.8 (1.1–2.9)
Estrogen–progestagen7,714/1162.5 (1.9–3.2)
Unknown1,068/121.6 (0.9–3.0)
Estriol592/51.0 (0.4–2.5)

Users of estrogen–progestagen constituted 70% of all users. The RR of breast cancer increased with increasing duration of use in users of estrogen–progestagen. The RR in users for <5 years was 2.3 (95% CI 1.7–3.2) compared to 2.8 (95% CI 2.0–4.0) in users for ≥5 years (Table IV).

Table IV. Estimated RR1 With 95% CI of Breast Cancer in Relation to Duration of Use by Type and Regimen of HRT Used at Baseline: the Nowac Study
Total duration of HRT use by type and regimenPopulation/cases2RR (95% CI)ptrend
  • 1

    Adjusted for age, BMI, age at menarche, ever use of OCs, time since menopause, family history of breast cancer, mammography, parity and age at first delivery.

  • 2

    Numbers may not add up due to missing values for one or more variables.

Estrogen only   
 < 5 years805/132.5 (1.4–4.5) 
 ≥ 5 years712/51.0 (0.4–2.5)0.2
Estrogen–progestagen   
 < 5 years4,669/632.3 (1.7–3.2) 
 ≥ 5 years2,920/512.8 (2.0–4.0)<0.0001
 Sequential regimen   
  < 5 years2,022/191.7 (1.0–2.8) 
  ≥ 5 years1,048/142.2 (1.3–3.8)0.004
 Continuous regimen   
  < 5 years2,616/442.6 (1.9–3.7) 
  ≥ 5 years1,846/373.2 (2.2–4.6)<0.0001

Using a continuous estrogen–progestagen regimen implied significantly higher risk than sequential use (RR = 2.8, 95% CI 2.1–3.8, and RR = 1.8, 95% CI 1.2–2.8, respectively; test for heterogeneity p = 0.04). However, total duration of use of ≥5 years significantly increased the risk of breast cancer in both categories compared to duration <5 years.

PAR%

With 151 incident breast cancers among 10,916 current users and 130 among 17,756 never-users and a population prevalence of 35% (current use), we estimated 27% of the total risk of breast cancer among Norwegian women 45–64 years to be due to current use of HRT. That equals approximately 300 cases of breast cancer per year.

Endometrial and ovarian cancers

The total number of incident endometrial cancers in 27,621 postmenopausal women with an intact uterus and information on HRT use was 75. No more than 30 cancers occurred in the user group. The risk of endometrial cancer was increased in users of estrogen only (RR = 3.2, 95% CI 1.2–8.0) and in users of estriol (RR = 3.1, 95% CI 1.2–7.9).

The total number of incident ovarian cancers in 30,115 postmenopausal women with intact ovaries and information on HRT use was 74. Fewer than half of them (34) occurred in the user group. There was a nonsignificant increased risk of 1.4 (95% CI 0.9–2.3) in current users of HRT (Table V).

Table V. Estimated RR With 95% CI of Endometrial1 and Ovarian2 Cancers in Relation to Type of HRT and Recency of Use: the Nowac Study
Recency and type of HRT useEndometrial cancerOvarian cancer
Population/casesRR (95% CI)Population/casesRR (95% CI)
  • 1

    Included in the analyses are only women with a uterus intact. Adjusted for age, BMI, smoking, ever use of OCs, time since menopause, parity and age at last birth.

  • 2

    Included in the analyses are only women with the ovaries intact. Adjusted for age, ever use of OCs, parity, hysterectomy and age at first delivery.

Recency    
 Never use16,035/451.017,439/401.0
 Ever use11,586/301.2 (0.7–1.9)12,676/341.3 (0.8–2.0)
 Current use9,292/221.1 (0.7–1.9)10,147/301.4 (0.9–2.3)
 Former use2,192/81.4 (0.7–3.0)2,399/40.8 (0.3–2.1)
Type    
 Estrogen only632/53.2 (1.2–8.0)1,048/20.9 (0.2–3.9)
 Estrogen–progestagen7,268/110.7 (0.4–1.4)7,549/231.5 (0.9–2.6)
 Unknown901/10.4 (0.1–2.9)1,016/41.9 (0.7–5.3)
 Estriol491/53.1 (1.2–7.9)534/10.9 (0.1–6.4)

DISCUSSION

Overall use and total duration of use in relation to breast cancer

We found a higher RR of breast cancer in both ever- and current users compared to never-users of HRT. Most ever-users (80%) stated current use. The risk estimate in current users was higher than those in the WHI trial13 and the MWS.14 However, the WHI women used conjugated estrogens (0.625 mg) plus medroxyprogesterone (2.5 mg), both components different from those used in the NOWAC study. The MWS, with more than 800,000 postmenopausal women, found the RR of breast cancer to be 1.66 (95% CI 1.58–1.75) among current compared to never-users of HRT. Different types of estrogen as well as combinations with progestagen were included in the estimate. However, while 40% of the women in the MWS used an estrogen-only preparation, the corresponding proportion in the NOWAC study was 15%. Consequently, the average Norwegian user of HRT will be an estrogen–progestagen user and the cancer impact of both constituents will be reflected in the risk estimate.

An updated analysis from the WHI trial gave a hazard ratio of 1.24 (95% CI 1.02–1.50) in current users of conjugated estrogens–medroxyprogesterone compared to placebo.18 Use of different HRT preparations in the MWS gave different risk estimates, the lowest being among estrogen-only users (RR = 1.3, 95% CI 1.21–1.40) and the highest being among estrogen–progestagen users (RR = 2.0, 95% CI 1.88–2.12).14 We found a corresponding difference in the risk estimates for users of estrogen-only and estrogen–progestagen; however, our risk estimates are higher. Users of estrogen-only preparations in Norway are few, and two-thirds of them are hysterectomized, oophorectomized or both, in contrast to other users of HRT. Sample variation and the possibility of these users being different and having a different baseline risk of cancer from other users could explain the higher estimate in users of estrogen only.

The underlying mechanisms of the effects of progestagens on the breast are complex and not fully understood. There is evidence from experimental animal studies indicating that progestagens can influence breast physiology as well as tumor development,11 and progestagens are known to increase mammographic breast density, an often-used surrogate marker for breast cancer.19 Most studies indicate that addition of progestagens enhances the risk of breast cancer associated with use of estrogen alone.20, 21, 22 Our results are consistent with this finding.

Six of 10 Norwegian estrogen–progestagen users favored a continuous regimen. If continuous progestagen implies increased carcinogenesis, this could explain the higher risk estimate in our study compared to the MWS, with only 36% using a continuous regimen.

We found a significant trend of increasing risk of breast cancer with increasing total duration of use of estrogen–progestagen combinations. The risk increased from year 1, as found in the MWS. The WHI trial did not find a corresponding risk increase for the first 2–3 years of use.13 It is difficult to explain this, but it may be speculated that the high mean BMI of the WHI women had an impact. The risk-generating effect of exogenous estrogen supply is of less importance in overweight postmenopausal women.11 This could suggest that a cumulative effect of years of exposure to HRT is needed to further increase the risk.

The MWS found a positive trend of increased risk in users of estrogen only. The most likely explanation for why we did not is the small sample size and sampling variation. In users of estrogen–progestagen, we found the risks associated with <5 and ≥5 years of use to be somewhat higher than in the MWS.

Type, regimen and duration of use

In the NOWAC study, 60% of current users of estrogen–progestagen preparations used a continuous regimen. We found the risk of breast cancer to be significantly higher among those using a continuous compared to a sequential regimen. Increasing the duration of use was significantly associated with increased risk in users of sequential and continuous regimens.

There are few studies and inconsistent results on the effects of progestagens on breast cancer.23 Various preparations available and a scarcity of data for continuous regimens make it impossible to draw any conclusions. To date, most studies have demonstrated an increased risk for sequential regimens.19, 21, 22 The MWS suggested no large variations between the effects of different progestagens or different regimens.14 There are, however, plausible biologic explanations for both possibilities.11

The main difference between the sequential and continuous regimens used in our study is the dosage of progestagen (10 vs. 30 mg of norethisterone/28 days). The dosage of estradiol is practically identical in both regimens, and we found no large difference in mean duration of use between the 2 groups. A Swedish study published in 1999 found a continuous regimen to confer increased risk. They found use of estrogen with a testosterone-derived progestagen like norethisterone to confer an excess risk of breast cancer per year of use (RR = 1.08, 95% CI 1.03–1.13) compared to use of estrogen with a progesterone-derived progestagen (RR = 0.95, 95% CI 0.80–1.14). The positive association was especially pronounced with a continuously combined regimen.24 The possible differences between compounds, regimens and dosages of progestagens are of particular interest to Norwegian users and need to be further investigated and understood.

Endometrial and ovarian cancers

Our risk estimates do not indicate an increased risk of endometrial cancer among Norwegian users of estrogen–progestagen. The risk estimates for users of estrogen only and oral estriol were significant, though based on few cases. Data from studies on estriol and endometrial cancer are scarce, but others have found similar results.25 There is a convincing body of evidence that addition of a progestagen counteracts the carcinogenic effect of estrogens on the endometrium.11 Limited data exist on estrogen–progestagen use in relation to ovarian cancer, and so far they are inconsistent. While OC use consistently shows an inverse relationship with ovarian cancer, the role of HRT is unclear. A meta-analysis from 1998 showed a modest risk increase (RR = 1.15) with ever use.26 Our samples are too small to draw firm conclusions about associations between HRT and ovarian cancer. However, no significant increase in risk has been observed.

Estimates, misclassification and confounding

Our study has several strengths: it is population-based and fairly representative, missing and recall bias are low and only few HRT preparations were available on the Norwegian market at the time of the study, most of them containing the same estrogen and progestagen.

Although we did not have the opportunity to check prescription data, we believe the reliability of self-reported use of HRT to be high. More than 90% of the women indicating current use also specified a brand name, suggesting high concordance. Moreover, our estimated prevalence numbers on HRT use are compatible with estimated user prevalence obtained from sales data. The prevalence size is decisive for the PAR%. However, even a 15% smaller prevalence would imply a considerable risk percent, accounting for about 250 incident cases of breast cancer. A Swedish study found that information on climacteric estrogen intake could be reliably obtained through a questionnaire.27

Confounding by known risk factors is accounted for by adjusting the estimates and by restricting the study population to postmenopausal women.

The estimated risk of breast cancer among current users of HRT decreased from the first to the fourth year of follow-up, most likely indicating increasing misclassification in the current user group as more and more current users probably discontinue treatment and become former users. This finding supports the earlier conclusion from the collaborative reanalysis of data from 51 epidemiologic studies published in 1997. Although they found the risk of breast cancer diagnosis to increase 2.6% for each year of HRT use, there was no evidence of increased risk of breast cancer ≥5 years after stopping HRT.12

In conclusion, current use of estrogen–progestagen preparations doubles the risk of breast cancer in women 45–64 years old. Our results suggest that HRT could be considered a major determinant for the increasing breast cancer incidence in postmenopausal women in Norway over the last decade. This underlines the importance of emphasizing that HRT should be used only for short periods of time and solely in women with serious peri- or postmenopausal symptoms.

Ancillary