Age-specific differences in the relationship between oral contraceptive use and breast cancer
Post on 07-Jun-2016
Age-Specific Differences in the Relationship between Oral Contraceptive Use and Breast Cancer Phyllis A. Wingo, M.S.,* Nancy C. Lee, M.D.,* Howard W. Ory, M.D.,t Valerie Beral, M.B.B.S.,S Herbert B. Peterson, M.D.,* and Philip Rhodes, M.S.5
Background. Nearly all studies have suggested that the use of oral contraceptives (OC) is not associated with the aggregate risk of breast cancer diagnosed in women aged 20-54 years. Because of age-specific differences in the breast cancer-parity relationship and because of age- specific differences in other breast cancer risk factors, the Centers for Disease Control reexamined data from the Cancer and Steroid Hormone Study (CASH) to assess whether OC use has different effects on the risk of breast cancer at different ages of diagnosis.
Methods. This population-based case-control study was designed to examine the relationship between the use of OC and the risk of breast, ovarian, and endome- trial cancer. CASH was conducted in eight geographic areas in the United States during 1980-1982. All partici- pants were interviewed at home with a pretested stan- dardized questionnaire including a calendar of life events and a photograph book of all pills marketed in the United States.
We found that the relationship between the risk of breast cancer and OC use appeared to vary by the age at diagnosis. Among women aged 20-34 years at
Presented at the National Conference on Gynecologic Cancers, Orlando, vorida, April 2-4, 1992.
From the *Division of Reproductive Health, Center for Chronic Disease Prevention and Health Promotion; the tInformation Re- sources Management Office; and the Division of Injury Control, Center for Environmental Health and Injury Control, Centers for Disease Control, Department of Health and Human Services, At- lanta, Georgia, and the $Imperial Cancer Research Fund, Cancer Epi- demiology Unit, Radcliffe Infirmary, Oxford, England.
Supported by interagency agreement 3-Y01-HD-8-1037 be- tween the Centers for Disease Control (Atlanta, Georgia) and the National Institute of Child Health and Human Development, with additional support from the National Cancer Institute (Bethesda, Maryland).
The authors thank James J. Schlesselman, David C. G. Skegg, George L. Rubin, and Peter M. Layde for comments and suggestions on the manuscript; Karen S. Colberg for her computer assistance; and Beverly H. Malone for her word-processing help. The authors ac- knowledge the contributors to the Cancer and Steroid Hormone Study: study design and coordination, The Division of Reproductive Health, Center for Chronic Disease Prevention and Health Promo- tion, Centers for Disease Control; principal investigator, George L.
diagnosis or interview, those who had ever used OC had a slightly increased risk of breast cancer (odds ratio [OR], 1.4; 95% confidence interval [CI], 1.0-2.1) compared with women of the same ages who had never used OC. Among these women, there were no trends of increasing or de- creasing risk with any measure of OC use. Among women aged 35-44 years, there was no association be- tween OC use and breast cancer. Among women aged 45- 54 years, those who used OC had a slightly decreased risk of breast cancer (OR, 0.9; 95% CI, 0.8-1.0). Among these women, risk estimates decreased significantly with in- creasing time since first and last use.
Conclusions. Although the slightly increased risk estimates for the youngest women were compatible with findings by other investigators, the decreased risk esti- mates for the oldest women have not been described in as many studies. Available data provide no reasons to change prescribing practices or the use of OC that are related to the breast cancer risk. Cancer 1993; 71:
Key words: oral contraceptives, breast cancer, case- control studies, epidemiology.
Rubin,M.B.,F.R.A.C.P.;projectdirector,PhyllisA. Wingo,M.S.;proj- ect associates, Nancy C. Lee, M.D., Michele G. Mandel, B.A., Herbert B Peterson, M.D.; data collection centers and principal investigators, Atlanta: Raymond Greenberg, M.D.; Connecticut: J. Wister Meigs, M.D., and W. Douglas Thompson, Ph.D.; Detroit, Michigan: G. Mane Swanson, Ph.D.; Iowa: Elaine Smith, Ph.D.; New Mexico: Charles Key, M.D., and Dorothy Pathak, Ph.D.; San Francisco: Donald Aus- tin, M.D.; Seattle: David Thomas, M.D.; Utah: Joseph Lyon, M.D., and Dee West, Ph.D.; pathology review principal investigators, Fred Gorstein, M.D., Robert McDivitt, M.D., and Stanley J . Robboy, M.D.; project consultants, Lonnie Bumett, M.D., Robert Hoover, M.D., Peter M. Layde, M.D., M.Sc., Howard W. Ory, M.D., M.Sc., James J . Schlesselman, Ph.D., David Schottenfeld, M.D., Bruce Stadel, M.D., Linda A. Webster, M.S.P.H., and Colin White, M.B.B.S.; and pathol- ogy consultants, Walter Bauer, M.D., William Christopherson, M.D., Deborah Gersell, M.D., Robert Kurman, M.D., Allen Paris, M.D., and Frank Vellios, M.D.
Address for correspondence: Phyllis A. Wingo, M.S., Centers for Disease Control, 1600 Clifton Road NE, Mail-stop K 34, Atlanta, GA 30333.
Accepted for publication September 2, 1992.
Oral Contraceptives and Breast Cancer/Wingo et al . 1507
Nearly all epidemiologic reports on the subject confirm that the risk of breast cancer diagnosed in women at ages 20-54 years is the same for those who have ever used oral contraceptives (OC) as for women who have never used them.'-3 In 1983, two reports suggested that using OC before the first term pregnancy or using "high progestogen" OC before age 25 years increased the risk of breast cancer diagnosed at young age^.^,^ More recent publications suggest possible increased risks of breast cancer associated with OC use among specific groups of users, including premenopausal women, nulliparous women, and women with long durations of use, espe- cially before the first term pregnancy.'p6-" Analyses of premenopausal women in the Cancer and Steroid Hor- mone Study data suggested a possible increased risk of breast cancer among nulliparous women aged 20-44 years and a decreased risk among nulliparous women aged 45-54 years.' Although the subgroups of users with increased risks varied across studies, most reports with positive findings involved breast cancer diagnosed at younger ages, usually younger than 35 years.
Several risk factors for breast cancer, including marital status, age at menarche, age at first birth, race, and birthplace alter the breast cancer risk in opposite directions, depending on the age at diagnosi~. '~-'~ The breast cancer-parity relationship varies by age.I3-l7 Women with high parity who are younger than 35 years of age have an increased risk of breast cancer compared with women with low parity of the same age. After a woman reaches approximately 35-40 years of age, parity confers the well-described protection against breast cancer. However, because the incidence of breast cancer before age 35 years is low and because most reports only describe the parity effect for all ages combined, high parity generally is considered to be pro- tective against breast cancer.
One author suggested that the breast cancer-OC relationship may parallel the breast cancer-parity rela- tionship and may vary by age, with an increased risk for women diagnosed at young ages and no risk or a de- creased risk for women diagnosed at older To assess possible age-specific differences in the breast cancer-OC relationship and to pursue the possible anal- ogy with the breast cancer-parity relationship, we reex- amined data from the Cancer and Steroid Hormone Study. 19,'0
Materials and Methods
The methods for the Cancer and Steroid Hormone Study have been described in detail previously.20,21 The study was a population-based case-control study con- ducted in eight geographic locations in the United States. The cases were 471 1 women aged 20-54 years
who were living in the study areas and had histologi- cally confirmed breast cancer, initially diagnosed be- tween December 1, 1980 and December 31, 1982. The control subjects were 4676 women aged 20-54 years who were identified by Waksberg's method of tele- phoning randomly selected residential telephone num- bers in the same locations where the women with breast cancer resided." Control subjects were matched to the expected 5-year geographic-specific age distribution of the patients with breast cancer. All study participants were interviewed in person at home with a pretested standardized questionnaire. The interview lasted ap- proximately 50 minutes and focused on reproductive and contraceptive histories, family histories of cancers, and personal characteristics and behaviors.
We used unconditional logistic-regression methods to adjust the estimates of relative risk within three age groups (ages 20-34, 35-44, and 45-54 years) for vari- ables considered a priori as potential confounder^.'^ The risk factors that were considered as potential con- founders included age at diagnosis or interview, parity, age at menarche, age at first term pregnancy, Quetelet's index of adiposity, menopausal status, family history of breast cancer, history of biopsy for benign breast dis- ease before use of OC, regularity of menses during ado- lescence, frequency of breast examinations, breast feeding, a history of spontaneous or induced abortions, and geographic location. The age at diagnosis or inter- view was assessed in different logistic-regression mod- els both as a continuous variable and as a matched vari- able. Because the results from both approaches were similar, we included age in the final models as a continu- ous variable. We also included those risk factors that altered the crude relative risk estimates by more than 5%. Models that included all confounding factors from any of the three age-specific models yielded compara- ble results.
We used likelihood ratio tests to assess multiplica- tive interactions between OC use and the risk factors of intere~t.'~ When we tested for the statistical significance of trends for the various measures of OC use, we ex- cluded women who had never used OC and included months of OC use in the model as a continuous vari- able.
To assess whether the relationship between OC use and the risk of breast cancer varied by age, we considered several logistic-regression models, Regardless of the age or OC groupings, the interaction between age and OC use was statistically significant (P c 0.05, by likelihood ratio tests).23 Because age seemed to modify the effect of OC use on the risk of breast cancer, we presented all
1508 CANCER Supplement February 25, 2993, Volume 71, No. 4
findings stratified by age at diagnosis or interview: 20- 34 years, 35-44 years, and 45-54 years.
Among women aged 20-34 years, those with breast cancer were more likely than women who were control subjects to be black, to have had their first men- strual period before age 12 years, to have higher panty,
and to be obese. By contrast, among women aged 45-54 years, women with breast cancer were more likely to be premenopausal and nulliparous than women of the same ages who were control subjects (Table 1).
When we examined patterns of OC use between cases and controls by age and the characteristics listed
Table 1. Age-Specific Percent Distributions of Characteristics for Breast Cancer Cases and Controls: Cancer and Steroid Hormone Study, 1980-1982
Age at diagnosis or interview (vr)
20-34 35-44 45-54
Cases Controls Cases Controls Cases Controls
White Black Other/unknown
< High school High school College Unknown
Income ($) < 10,000 10,000-19,999 20,000-29,999 30,000-39,999
< 12 yr 12-14 yr 14+ yr Other/unknown
Menopausal status Premenopausal Perimenopausal Natural menopause Surgical menopause Other/unknown
Parity 0 1-2
3-4 4+ Unknown
Age first menstrual period
Age first term pregnancy Nulliparous women c 20 yr 20-24 yr 25+ yr Unknown
29.2 14.3 9.7
24.2 69.3 6.1
91.0 - -
53.6 17.4 1.1
26.7 18.3 7.7
18.5 71.3 10.1
93.2 - -
29.8 59.9 -
8.2 16.6 24.7
25.9 65.0 8.9
19.4 37.3 27.0
58.9 1 .o
14.5 13.2 4.2
62.6 11.5 0.5
60.6 13.5 0.7 19.2
11.2% 39.4 38.0
11.2 23.9 40.9
45.8 53.3 0.9
11.0 12.9 7.6
65.5 11.8 0.3
21.7 25.0 23.0 28.7
14.9% 33.3 35.9
14.9 16.8 39.8
42.9 56.4 0.7
11.4 12.0 5.9
67.7 11.7 0.4
12.2 25.4 26.0
8.7% 29.3 40.4
8.7 22.1 42.9
Oral Contraceptives and Breast Cancer/ Wingo et al. 1509
Table 1. (Continued)
Age at diagnosis or interview (yr)
20-34 35-44 45-54
Characteristic Cases Controls Cases Controls Cases Controls
Spontaneous and induced abortions None 72.7% 72.5% 70.8% 70.2% 71.3% 69.0% 1 17.7 19.0 19.0 18.5 18.4 19.5
1+ 9.2 7.8 9.4 10.5 8.8 10.3
Unknown 0.4 0.7 0.8 0.8 1.5 1.2 Family history of breast cancer
None 57.6 74.3 57.2 66.0 53.7 61.8 First-degree 9.7 2.6 11.4 5.2 11.7 7.0 Second-degreen 21.4 16.3 22.2 20.1 20.5 18.5 Unknown 11.3 6.8 9.2 8.7 14.1 12.7
Biopsy for benign breast disease before oral contraceptive use
Yes 3.4 1.8 4.9 2.5 13.3 8.6 No 90.5 92.8 86.1 91.2 77.0 82.8 Unknown 6.1 5.4 9.0 6.3 9.7 8.6
in Table 1, we found that cases and controls of the same ages reported similar percentages of use (Table 2). How- ever, across age groups, the patterns of use demon- strated birth cohort effects. Older women were approxi- mately half as likely to have ever used OC as younger women. Among women aged 45-54 years, fewer than 0.1 % first used OC before age 20 years compared with more than 35% of women younger than 35 years of age at diagnosis or interview. The mean ages at first use of OC were 20 years for the women who were youngest at diagnosis or interview, 24 years for women aged 35-44 years, and 32 years for women who were 45-54 years. Among the oldest OC users, approximately 54% of the cases and controls had ever used OC containing more than 50 pg of estrogen compared with 42% of cases and controls who used OC and were younger than 35 years of age at diagnosis or interview. Among parous women, 3% of the oldest cases and controls used OC before the first term pregnancy compared with approximately 5 1 YO of youngest cases and controls.
Among the women who were youngest at diagno- sis or interview, OC users had an increased risk of b...