Use of hormone therapy and risk of breast cancer detected at screening and between mammographic screens

Authors


Abstract

Postmenopausal hormone therapy (HT) is associated with increased risk of breast cancer, but in women undergoing breast cancer screening it is not clear whether use of HT is associated with increased risk of breast cancer detected at screening or between screens (interval cancer). Further, it is unclear whether the use of the HTs that have been common in Scandinavia is associated with higher risk of breast cancer than the HTs used in other countries. Our study was based on data from 296,651 women aged 50–69 years, who participated in the Norwegian Breast Cancer Screening Program during 1995–2004. After a mean enrollment time of 3.8 years, 1,512 women were diagnosed with invasive screen detected breast cancer, and 814 with invasive interval breast cancer. Cox regression models were used to estimate hazard ratios (HRs) of breast cancer associated with HT use, after adjusting for confounders. Ever users of HT had a 58% increased risk of breast cancer, compared to never users. The HRs associated with HT use were 1.45 (95% confidence interval (CI) = 1.29–1.63) for screen detected and 1.89 (95% CI = 1.61–2.23) for interval cancer. The difference between screen detected and interval cancer was statistically significant (p = 0.011). The HR of breast cancer increased with duration of HT use, but significantly more so for interval than for screen detected cancer (use of HT for 5 or more years compared to never use; HR = 2.91, 95% CI = 2.10–4.04 and HR = 1.94, 95% CI = 1.51–2.50, respectively; p = 0.002). The population attributable fraction of breast cancer due to HT use was 19.8% overall. Ever users of HT tended to develop a cancer of lower grade. No other differences in histological tumor characteristics were observed between ever and never users of HT among screen detected or interval cancers. The estimated risks of either breast cancer overall with HT use are higher in Norway than reported in similar studies from the U.S. HT-use is a stronger risk factor for interval cancer than for screen detected cancer. The increased risk of interval cancer, which may partly be due to decreased sensitivity of mammograms in HT users, remains a challenge in breast cancer screening programs. © 2006 Wiley-Liss, Inc.

Hormone therapy (HT) has been found to be an independent risk factor for breast cancer1, 2, 3, 4, 5, 6, 7, 8, 9 and HT containing estrogen–progestin combinations have been associated with a higher risk than preparations containing estrogen alone.4, 5, 7, 8, 9 European3, 7, 8, 9 and certain Scandinavian studies3, 7, 9 have reported stronger breast cancer risk estimates among HT users compared to U.S. studies. The possible reasons for the difference are not understood. Two of the previous studies from Scandinavia were small cohort studies, including a few women reporting use of HT for more than 5 years.7, 9 Further and larger studies are thus warranted. Although HT use has declined in Norway after the publication of the Women Health Initiative trial in 2002, HT was widely used in the past decades,9, 10, 11 and is still in use by a large number of women.12

Screening for breast cancer by mammography is recommended by the World Health Organization13 and is offered in most countries in the western part of the world. Currently, most breast cancers in women aged 50–69 are diagnosed by screening.14, 15 Use of HT among women undergoing regular mammography screening and the subsequent effects of HT on the sensitivity of the screening program have frequently been discussed.16, 17, 18, 19, 20 It has been argued that HT users are more likely to attend screening21, 22 and to be diagnosed with smaller, screen detected cancers. On the other hand, HT has been shown to increase mammographic density,23, 24 which could mask the tumor25 and result in interval cancer.17 Interval cancers have less favorable prognostic and predictive tumor characteristics than do screen detected cancers;26, 27 thus a high interval cancer rate is a disadvantage in a screening program. A better understanding of the risk of breast cancer among HT users is obviously important for the overall success of breast cancer screening programs.

As a part of the evaluation of the Norwegian Breast Cancer Screening Program (NBCSP), we wanted to determine the overall relative risk of breast cancer associated with HT use in a large prospective study, and to determine the risk of screen detected versus interval breast cancer associated with HT use.

Material and methods

The NBCSP is a governmentally funded breast cancer screening program administered by the Cancer Registry of Norway. The program started in 1995/96 as a 4-year pilot project in 4 of the country's 19 counties, and has gradually expanded to become nationwide from February 2004.14 The results in our study are based on data from the first 8 years of the program (November 15, 1995 – March 31, 2004). The frequency and selected characteristics of the interval cancers diagnosed during the first 2 years of the pilot project have previously been described.28

Collection of data

All Norwegian women residents aged 50–69 years in the study period, living in the counties that had started the NBCSP, were invited to receive bilateral 2-view screening mammography every second year. A personal letter with specific place and time for examination invited the women 2–4 weeks before their scheduled appointment. Together with the invitation, the women also received a leaflet about the screening program and a 1-page questionnaire. The invitation letter specified that the completed questionnaire should be turned in at the first screening examination. The attendance rate was 76%. A total of 94% of the women participating in the first screening round completed and turned in the questionnaire. All participants signed an informed consent that specified that data related to their screening visit could be used for evaluation and scientific purposes. The current study was part of the evaluation and scientific aspects of the screening program, thus covered by the general ethical approval of the screening program as part of the Cancer Registry of Norway.29

Cancer ascertainment

The Cancer Registry of Norway is one of the oldest in the world, and cancer has since 1951 been a reportable disease by law. As a consequence, the Cancer Registry is essentially complete, i.e. 99.6% for solid tumors.30, 31

Results from all radiological examinations conducted in the NBCSP were reported online to the Cancer Registry, while results of all invasive procedures (fine-needle aspirations, core needle biopsies and open biopsies) were reported on paper-based forms. A total of 98.9% of the cancers in our study were histologically verified. The remaining 1.1% was diagnosed by fine-needle aspiration cytology. Only invasive breast cancers were included in these analyses.

All breast cancers diagnosed as a result of a diagnostic work-up following a positive screening test were classified as screen detected cancers. An interval cancer was defined as a breast cancer diagnosed between the date of a negative screen result and the date of the next scheduled screening examination.

Women who had attended one or more screening examinations in the governmental program were included in the study. All breast cancers diagnosed at private clinics were defined as interval cancers, as we assumed that the majority of these cancers were diagnosed as a result of symptoms or signs noted by the woman or her physician. However, because of the 2-year interval in the governmental program, it is possible that some women will have undergone additional planned screening mammography at private clinics in between the NBCSP screenings. To what extent this occurred is not known, but to address this possible problem, we stratified women by access to mammography at private clinics (private clinics were predominantly found in certain parts of the country during this time). Areas of the country with high access included Akershus, Agder, Hedmark, Hordaland, Oppland and Oslo, while areas with low access included the rest of the country. We also stratified the analyses on whether or not women reported having had mammography before participating in the NBCSP, and whether that examination was performed at a private clinic.

The questionnaire

The questionnaire covered age at menarche and menopause, ages at first and last birth, ever use and duration of use of oral contraceptives (OC) and hormone therapy (HT), family history of breast cancer, education and life style factors such as smoking habits and alcohol consumption.

Age at answering the questionnaire was categorized into 4 groups: 50–54, 55–59, 60–64 and 65+ years old. The questionnaire did not distinguish between HT containing estrogen alone and HT containing estrogen and progestin combined, and did not distinguish between current and former use of HT. Duration of HT use was reported in 4 categories: less than 1 year, 1–4 years, 5–9 years, and 10 years or more. Age at start of HT use was estimated as age at answering the questionnaire minus duration of HT use. Menstruating women who reported never having used HT were defined as premenopausal. Age at menopause was defined as reported age at cessation of menstruation, or the estimated age at start of HT use. The other variables we considered were educational level (primary school, intermediate school or college/university), age at menarche (before or after the age of 13 years), number of births (none, 1–2 or 3+), ever use of OCs (yes/no), alcohol consumption (never, seldom, weekly or more), family history of breast cancer (none or if mother and/or sister was diagnosed before the age of 50 years) and mammography history.

A barcode on the questionnaire was linked to a file with personal information about a unique 11-digit personal identification number carried by all inhabitants in Norway. This number was further linked to the screening database.

Women available for analyses

A total of 298,548 of 341,639 women (87.4%) completed the questionnaire on all covariates used in our study. Of these, 1,897 were diagnosed with invasive screen detected cancer in their prevalent screening in the NBCSP, thus not included as breast cancers in the analyses. Among the remaining 296,651 women, 1,512 women were diagnosed with an invasive screen detected breast cancer in their subsequent screening (screened once or more in the NBCSP before). All the 814 interval cancer cases were included in the analyses.

The enrollment time in the study was calculated as follows: Each woman was included in the study from the date she attended screening and turned in a completed questionnaire. She was followed to whichever date came first: the cut off date of the enrolling period (March 31, 2004), the date of an invasive breast cancer diagnosis or the date of censoring. Reasons for censoring were death (n = 5,514), emigration (n = 1,392), diagnosis of ductal carcinoma in situ (n = 722) and upper age limit (n = 16,658). The upper age limit for follow-up was 2 years after the last invitation in order to identify any interval cancers. Twenty-three women with malignancies other than breast cancer in the breast area (malignant melanoma, lymphoma) were censored at the time of diagnosis. Further, 262 women diagnosed with breast cancer more than 2 years after their last participation were considered noncompliant, thus censored 2 years after their last examination in the NBCSP.

Statistical analyses

Cox regression was used to estimate the hazard ratio (HR) as a measure of the relative risk of screen detected, interval and all breast cancers associated with HT use. We estimated HRs and 95% confidence intervals (CI) and their corresponding trends while adjusting for age at answering the questionnaire, calendar year of enrolling the study, educational level, age at menarche, menopausal status/age at menopause, number of childbirths, use of OCs, alcohol consumption and family history of breast cancer using the categories described earlier. To determine whether these variables as well as the effects of HT use and HT duration differed between screen detected and interval cancers, we conducted a case–case analysis using logistic regression where interval cancers were considered “cases” and screen detected cancers as “controls”. The outcome thus estimated the relative risk of an interval cancer versus a screen detected cancer in ever versus never users of HT.

We tested for interactions between HT use and other variables such as access to private clinics by including interaction variables in the model.

The population attributable fraction (AFp) was calculated to estimate the proportion of breast cancers in this program that could have been avoided if the exposure was eliminated. AFp was calculated as Pe (RR − 1)/(1+ Pe (RR − 1)), where RR is the relative risk for breast cancer (here, HR) in ever users of HT compared to never users and Pe is the proportion of the population exposed, which was estimated from the questionnaire data.32

A p value less than 0.05 was considered statistically significant. All p values provided are double-sided. All analyses were conducted using SPSS (SPSS version 12.0.1 for Windows, SPSS, Chicago, IL).

Results

The women were 50–69 years old on entering the NBCSP. Mean age at answering the questionnaire was 57.6 years for women without breast cancer diagnosed during the study period, 57.7 years for screen detected cancer cases and 57.0 years for interval cancer cases.

A total of 42.6% of all women reported ever use of HT. The proportion of HT users was 42.5% among women without breast cancer, 50.1% among the screen detected cancer cases and 57.4% among the interval cancer cases.

Table I shows the HRs of breast cancer associated with selected risk factors. The relative risk of interval, but not of a screen detected cancer, was associated with educational level (p from case–case analyses comparing screen detected and interval cancers = 0.014). Multi-parity decreased the risk of both screen detected and interval cancer, while ever use of OCs decreased the risk of screen detected cancer and increased the risk of interval one (p from case–case analyses = 0.005). Having a mother or sister with breast cancer diagnosed before age 50 increased the risk of both screen detected and interval cancer.

Table I. Hazard Ratio (HR)1 of Screen Detected and Interval Breast Cancer Associated with Risk Factors in Women Participating in the Norwegian Breast Cancer Screening Program
Risk factors2Screen and interval cancers (N = 2,326)Screen detected cancers (N = 1,512)Interval cancers (N = 814)
NHRNHRNHR
  • 1

    Adjusted for age at answering the questionnaire, educational level, age at menarche, menopausal status/age, number of childbirth, use of oral contraceptive, use of hormone treatment, alcohol consumption, calendar years and familial history of breast cancer required.

  • 2

    n, number of women available for analyses.

  • 3

    Values in parentheses are 95% CIs.

  • 4

    p value from case–case analyses (logistic regression) comparing screen detected with interval cancers.

  • 5

    Age at menopause corrected for duration of HT use.

  • 6

    Breast cancer diagnosed in mother or sister before the age of 50 years.

Age at answering the questionnaire
 50–54 years (n = 134,788)1,024163013941
 55–59 years (n = 55,439)5051.21 (1.08–1.36)33381.28 (1.12–1.48)31671.09 (0.90–1.32)3
 60–64 years (n = 54,386)5241.30 (1.16–1.46)3781.44 (1.25–1.66)1461.04 (0.85–1.28)
 65–69 years (n = 52,038)2730.76 (0.66–0.88)1660.70 (0.58–0.84)1070.90 (0.71–1.14)
  Test for trend p = 0.215 p = 0.350 p = 0.421
  Test of interval vs screen detected cancer4     p = 0.865
Educational level
 Primary school (n = 107,391)716149012261
 Comprehensive (n = 124,709)1,0351.10 (1.00–1.21)6781.07 (0.95–1.21)3571.15 (0.97–1.37)
 Graduate (n = 64,551)5751.16 (1.03–1.30)3441.05 (0.90–1.21)2311.38 (1.14–1.68)
  Test for trend p = 0.011 p = 0.482 p = 0.001
  Test of interval vs screen detected cancer4     p = 0.014
Age at menarche
 ≤13 years (n = 146,592)1,176177614001
 ≥14 years (n = 150,059)1,1500.93 (0.86–1.01)7360.89 (0.79–0.99)4141.02 (0.89–1.17)
  Test of interval vs screen detected cancer4     p = 0.122
Childbirth
 Null-parious (n = 25,853)24911601891
 1–2 childbirth (n = 148, 273)1,2600.96 (0.84–1.11)8000.97 (0.82–1.16)4600.95 (0.75–1.19)
 ≥3 childbirth (n = 122,525)8170.76 (0.66–0.88)5520.80 (0.67–0.95)2650.69 (0.54–0.88)
  Test for trend p = 0.001 p < 0.001 p < 0.001
  Test of interval vs screen detected cancer4     p = 0.203
Ever used OC
 No (n = 175,905)1,426197714491
 Yes (n = 120,746)9000.97 (0.89–1.07)5350.88 (0.79–0.99)3651.17 (1.00–1.36)
  Test of interval vs screen detected cancer4     p = 0.005
Menopausal status/age at menopause5
 Regular periods, no use of HT (n = 31,298)22211391831
 ≥50 years (n = 109,447)8990.78 (0.66–0.92)6040.82 (0.67–1.01)2950.70 (0.53–0.92)
 40–49 years (n = 121,588)9540.74 (0.63–0.88)6140.76 (0.62–0.94)3400.69 (0.52–0.91)
 <40 years (n = 34,318)2510.65 (0.53–0.80)1550.66 (0.51–0.86)960.61 (0.43–0.86)
  Test for trend p = 0.001 p = 0.006 p = 0.033
  Test of interval vs screen detected cancer4     p = 0.982
Currently partake of alcohol
 Never (n = 46,366)350124511051
 Seldom (n = 168,226)1,2060.95 (0.84–1.08)7800.92 (0.79–1.06)4261.04 (0.83–1.29)
 Weekly or often (n = 82,059)7701.11 (0.97–1.27)4871.09 (0.92–1.28)2831.17 (0.92–1.48)
  Test for trend p = 0.009 p = 0.061 p = 0.061
  Test of interval vs screen detected cancer4     p = 0.532
Family history of breast cancer6
No (n = 288,798)2,23011,45217781
Yes (n = 7,853)961.59 (1.30–1.95)601.52 (1.17–1.96)361.72 (1.23–2.40)
Test of interval vs screen detected cancer4     p = 0.572

Risk of breast cancer associated with HT use

Overall, as compared to never users of HT, ever users were at 58% higher risk of breast cancer (Table II). Use of HT increased the relative risk for both screen detected (HR = 1.45, 95% CI = 1.29–1.63) and interval cancer (HR = 1.89, 95% CI = 1.61–2.23). Use of HT for 5 or more years increased the HR (statistically significant) of both screen detected, interval cancer and for breast cancer in total (Table II, all p for trends < 0.001), but the trend for interval cancers was stronger (statistical significantly) than that for screen detected cancer (p from case–case analyses = 0.003). Use of HT for 5 or more years was associated with an almost 3-fold elevated risk of developing interval cancer. In a case–case analysis, there was a statistically higher risk of an interval than a screen detected cancer in HT users as compared to never users (RR = 1.31, 95% CI = 1.06–1.61, p = 0.011).

Table II. Hazard Ratio (HR)1 of Screen Detected and Interval Breast Cancer Associated with use of Hormone Therapy (HT) in Women Participating in the Norwegian Breast Cancer Screening Program
Duration of HT use2Screen and interval cancer (N = 2,326)Screen detected cancer (N = 1,512)Interval cancer (N = 814)
NHRNHRNHR
  • 1

    Adjusted for age at answering the questionnaire, calendar year, age at menarche, educational level, menopausal status/age, number of childbirth, use of oral contraceptive, alcohol consumption and family history.

  • 2

    n, number of women available for analyses.

  • 3

    Values in parentheses are 95% CIs.

  • 4

    p value from case–case analyses (logistic regression) comparing screen detected to interval cancer.

Never use of HT (n = 170,258)1,101175413471
<1 year (n = 22,093)1651.12 (0.95–1.33)31081.08 (0.88–1.33)3571.22 (0.92–1.64)3
1–4 years (n = 52,135)5501.55 (1.39–1.73)3471.45 (1.26–1.66)2031.79 (1.48–2.17)
5–9 years (n = 36,494)3622.03 (1.76–2.35)2131.72 (1.43–2.07)1492.75 (2.16–3.50)
≥10 years (n = 15,671)1482.23 (1.83–2.72)901.94 (1.51–2.50)582.91 (2.10–4.04)
 Test for trend p < 0.001 p < 0.001 p < 0.001
 Test of interval vs screen detected cancer4     p = 0.003
Ever use of HT (n = 126,393)1,2251.58 (1.44––1.74)7581.45 (1.29–1.63)4671.89 (1.61–2.23)
 Test of interval vs screen detected cancer4     p = 0.011

With the HRs given in Table II and a HT prevalence of 42.6%, we estimated the population attributable fraction (AFp) of breast cancer to be 19.8% in total (18.4% for screen detected cancers and 33.8% for interval cancers) (results not shown in table).

We next stratified our results on access to private screening. The relative risk of having a screen detected breast cancer associated with ever use of HT was 1.52 (95% CI = 1.25–1.85) in women residing in areas with high access to mammography at private clinics, and 1.39 (95% CI = 1.21–1.60) in women residing in areas with low access (interaction p = 0.63). The relative risk of interval cancer in areas with high access was 1.89 (95% CI = 1.54–2.31), and essentially identical—1.89 (95% CI = 1.43–2.51)—in areas with low access to private screening (interaction p = 0.89) (results not shown in table).

We further stratified the results on prior mammography and prior mammography at a private clinic. In women reporting mammography prior to participating in the NBCSP, the adjusted HR of screen detected breast cancer associated with ever use of HT was 1.37 (95% CI = 1.17–1.59), while it was 1.56 (95% CI = 1.29–1.88) among those without prior mammography (interaction p = 0.57). For interval cancer the HR was 1.93 (95% CI = 1.58–2.38) for those with prior mammography and 1.64 (95% CI = 1.23–2.19) for those without (interaction p = 0.58). For women with prior mammography at a private clinic, adjusted HR of screen detected breast cancer associated with HT was 1.44 (95% CI = 1.17–1.78) and 1.45 (95% CI = 1.27–1.67) among those without prior mammography (interaction p = 0.73). For interval cancer, the HR was 2.03 (95% CI = 1.50–2.75) for those with prior mammography and 1.82 (95% CI = 1.50–2.22) for those without (interaction p = 0.26) (results not shown in table).

Tumor characteristics in ever versus never HT users

As expected, tumor size was larger, more women were lymph node positive, and grade was worse among interval cancer than screen detected cancers (Table III). These differences were statistically significant both among HT users and never users. Ductal carcinoma was the most common cancer type diagnosed in all women combined, as well as in the following subgroups of cases: screen detected, interval cancers, ever and never users of HT. However, there were more lobular cancers among the interval cancers. Among screen detected cancers, there were no differences between ever and never users of HT, except for grade, which was worse for never users of HT. The results were similar for interval cancers.

Table III. Prognostic and Predictive Tumor Characteristics of Screen Detected and Interval Breast Cancers by use of Hormone Therapy (HT) in Women Participating in the Norwegian Breast Cancer Screening Program
Tumor characteristicsScreen detected cancers1Interval cancers
Never used HT2 (N = 754)Ever used HT (N = 758)Never used HT3 (N = 347)Ever used HT (N = 467)
  • 1

    All p values for comparing screen detected cancers and interval cancers (among never or ever users of HT) were statistically significant (p < 0.01) except for histological morphology among ever users of HT (p = 0.17).

  • 2

    Among screen detected cancers, none of the differences in tumor characteristics were statistically significant in never as compared to ever users of HT (all p values > 0.05), except for Grade (p = 0.021).

  • 3

    Among interval cancers, none of the differences in tumor characteristics were statistically significant in never as compared to ever users of HT (all p values > 0.05), except for Grade (p < 0.001).

Histological morphology
 Invasive ductal carcinoma (%, 95% CI)82.5 (79.6–85.1)80.2 (77.2–83.0)74.1 (69.2–78.7)76.2 (72.1–80.0)
 Invasive lobular carcinoma (%, 95% CI)9.7 (7.7–12.0)12.4 (10.1–15.0)15.2 (11.6–19.4)13.1 (10.1–16.5)
 Other invasive (%, 95% CI)7.8 (6.0–10.0)7.4 (5.6– 9.5)10.6 (7.6–14.4)10.7 (8.1–13.9)
 Missing (n)0000
Tumor size
 Mean (mm, 95% CI)14.0 (13.4–14.7)13.9 (13.3–14.6)21.7 (20.3–23.0)20.3 (18.9–21.7)
 Median (mm)12122018
 Missing (n)39404868
Lymph node statement
 Lymph node positive (%, 95% CI)23.3 (20.1–26.8)25.8 (22.5–29.3)49.5 (43.6–55.3)39.2 (34.4–44.1)
 Missing (n)103965360
Grade
 I (%, 95% CI)34.3 (30.8–38.0)40.6 (36.9–44.4)14.7 (11.0–19.2)25.8 (21.7–30.2)
 II (%, 95% CI)48.3 (44.5–52.1)45.4 (41.7–49.3)49.4 (43.7–55.1)47.3 (42.5–52.2)
 III (%, 95% CI)17.4 (14.7–20.5)14.0 (11.5–16.8)35.9 (30.6–41.5)26.9 (22.8–31.4)
 Missing (n)68743536
Estrogen receptor status
 Positive (%, 95% CI)85.9 (82.6–88.9)84.9 (81.5–87.9)67.9 (61.7–73.6)73.6 (68.4–78.3)
 Missing (n)24924999149
Progesterone receptor status
 Positive (%, 95% CI)63.8 (59.4–68.1)61.2 (56.8–65.5)46.1 (37.9–52.6)48.3 (42.6–53.9)
 Missing (n)265260105152

The results from further analyses of the association between hormone use and tumor size are shown in Table IV. The HR of a small (<15 mm) screen detected cancer associated with ever use of HT was lower (HR = 1.39, 95% CI = 1.20–1.60) than for a large (≥15 mm) screen detected cancer (HR = 1.58, 95% CI = 1.27–1.97) (Table IV), but not statistically significant (p from case–case analysis = 0.370). However, for interval cancers the HRs associated with ever use were significantly higher (HR = 2.37, 95% CI = 1.77–3.18) for a small than for a large interval cancer (HR = 1.62, 95% CI = 1.30–2.02) (p from case–case analysis = 0.114).

Table IV. Hazard Ratio (HR)1 of Screen Detected and Interval Breast Cancer <15 mm and ≥15 mm Associated with use of Hormone Therapy (HT) in Women Participating in the Norwegian Breast Cancer Screening Program
Duration of HT useHR of screen detected cancerHR of Interval cancer
<15 mm (N = 999)≥15 mm (N = 434)<15 mm (n = 267)≥15 mm (n = 431)
  • 1

    Adjusted for age at answering the questionnaire, calendar year, age at menarche, educational level, menopausal status/age, number of childbirth, use of oral contraceptive, alcohol consumption and family history of breast cancer.

  • 2

    Values in parentheses are 95% CIs.

  • 3

    p value from case–case analyses using logistic regression (screen detected cancers <15 mm to those ≥15 mm and interval cancers <15 mm to those ≥15 mm).

Never use of HT1111
<1 year use1.23 (0.96–1.56)20.87 (0.56–1.34)21.34 (0.84–2.38)21.21 (0.83–1.78)2
1–4 years use1.35 (1.14–1.60)1.57 (1.21–2.03)2.48 (1.78–3.45)1.46 (1.11–1.90)
5–9 years1.59 (1.27–1.99)2.27 (1.62–3.18)3.12 (2.05–4.77)2.37 (1.69–3.33)
≥10 years1.69 (1.24–2.31)2.69 (1.69–4.29)3.22 (1.82–5.67)2.61 (1.64–4.15)
 Test for trendp < 0. 0001p < 0.0001p < 0.0001p < 0.0001
 Test for whether HR differ by tumor size3 p = 0.061 p = 0.293
Ever use of HT1.39 (1.20–1.60)1.58 (1.27–1.97)2.37 (1.77–3.18)1.62 (1.30–2.02)
 Test for whether HR differ by tumor size3 p = 0.370 p = 0.114

Discussion

In our study from the Norwegian Breast Cancer Screening Program, NBCSP, ever use of HT increased the relative risk of breast cancer by 58%. The estimated risk increase was higher than that reported in similar studies from the U.S.1, 2, 4, 5, 6 HT-use was a substantially stronger risk factor for interval (HR = 1.89) than for screen-detected cancer (HR = 1.45).

The estimated relative risk of breast cancer associated with HT in our study was higher than that reported in studies from the U.S.1, 4, 5 Our study is the largest cohort study of HT use and breast cancer risk conducted in Scandinavia, and in general—our study supports the results from these previous Scandinavian studies.3, 7, 9 In Norway as in other European countries, estradiol is the predominant estrogen found in HT preparations, while the progestin component is predominantly norethisterone. Some commonly used HT preparations in Norway during the late 1990s contained relatively high amount of both estrogen (2 mg estradiol) and progestin (1mg norethisteroneacetate), while recent accessible preparations contain only half the dose of these hormones. We can speculate that part of the high relative risk of breast cancer associated with HT use observed in our and other Scandinavian studies3, 7, 9 is due to use of relatively high dose regimens of HT preparations. The population attributable fraction of 19.8% suggests that HT was responsible for ∼460 of the 2,326 cancer cases in our study. The estimate indicates that even with declining HT use over time, HT remains a concern for breast cancer screening programs.

The higher risk of interval cancer in HT users as compared with screen detected cancer could be due to several reasons: masking of tumors due to increased mammographic density, increased cell proliferation and rapid growing tumors in HT users, or increased surveillance in HT users.

HT has been shown to increase mammographic breast density.23, 24 Mammographically dense breasts may obscure tumors, which increases the difficulty of detecting breast cancer and thereby decreases the sensitivity of breast screening programs.19, 20, 21, 22 Information on breast density was not available in our study, but several studies,22, 28, 33 including 1 from the NBCSP's pilot project,28 have shown a convincing link between HT and breast density. We found a nonstatistically significant higher risk of a large (≥15 mm) than a small (<15 mm) screen detected cancer in HT users. The results might indicate a masking of small tumors, which later are diagnosed as interval cancer.

HTs containing estrogen and progestin have been found to have direct cell proliferative effects,16, 34 and the increased epithelial growth may subsequently increase the risk of an epithelial tumor. Thus, use of HT may have caused some cells or preexisting clinically latent cancers to grow faster, and this may contribute to the increased risk of interval cancers we observed.

The third reason for the increased risk of interval cancer among HT users could have been the increased medical surveillance in HT users,21, 22i.e. high extent of mammography at private clinics in between 2 scheduled NBCSP's screening examinations. All breast cancers diagnosed at private clinics in between 2 scheduled examinations were defined as interval cancer in our study. However, if the increased risk of interval cancer in HT users was predominantly a consequence of more frequent mammography (at private clinics) in HT users, then we would expect this effect to be limited to areas with high access to private mammography. This was not so. We also observed the same risk for breast cancer associated with HT in women reporting to have had mammography prior to participation in the NBCSP as among those without.

The strengths of our study include the population-based cohort study design and the large sample size. This study was performed in a screening setting similar to that of the Million Women Study (MWS).8 The women completed the questionnaire just before they were screened, but while the MWS excluded breast cancers diagnosed before recruitment, we also excluded breast cancer cases diagnosed in the prevalent screening round. This was done to avoid any bias caused by HT using women suspecting a tumor being more likely to participate. Nevertheless, we were able to show substantial association between HT and breast cancer risk.

Weaknesses of the study were lack of information about preparations' composition, estrogen only versus estrogen and progestin combined, and the fact that the questionnaire did not distinguish between former and current HT use. In addition, we lacked information about mammographic breast density. However, data from another study from Norway show that about 70% of HT users in Norway reported use of estrogen–progesterone therapy, and 14% use of estrogen only.35 The same study suggested that 80% of ever HT users were current users.35 The consumption of HT grew in the 1990s in Norway, peaked in 2001, and as in other countries, fell after the 2002 publication of results from the Women's Health Initiative Trial. In 2001, every second woman 55–59 years old had ever used HT.35 Tibolone was introduced in Norway in 1999,35 but represents only a small fraction of total sales.12

In conclusion, we found a high relative risk of breast cancer associated with HT use, higher than what has been found in studies from the U.S., but similar to that in previous Scandinavian studies. We found an increased HR of both screen detected and interval breast cancer associated with HT use. The relative risk of interval cancer was higher than that of screen detected cancer, with long-term HT use being associated with almost a tripling of risk compared with never users. Despite declined use of HT, our results suggest that the association between breast cancer and HT remains a challenge in breast cancer screening programs. Further research is needed to understand the mechanisms of breast tumor progression in HT users.

Ancillary