American Journal of Epidemiology Copyright C 1992 by The Johns Hopkins Urtverslty School of Hygiene and Public Health All rights reserved
Vol. 136, No. 12 Printed in U.S.A.
ORIGINAL CONTRIBUTIONS
A Case-Control Study of the Risk of Breast Cancer in Relation to Oral Contraceptive Use
Lynn Rosenberg,1 Julie R. Palmer,1 E. Aileen Clarke,2 and Samuel Shapiro1
The relation of oral contraceptive use to the risk of breast cancer was evaluated in a case-control study of women under the age of 70 years, conducted in Toronto, Canada, from 1982 through 1986. A total of 607 breast cancer cases identified in a cancer hospital were compared with 1,214 controls matched to the cases on neighborhood and decade of age. Information on oral contraceptive use and risk factors for breast cancer was collected in home interviews. Conditional logistic regression was used to control multiple confounding factors. For women aged 40-69 years (527 cases, 1,054 controls), trie results suggest that oral contraceptive use does not increase the risk of breast cancer. Multivariate relative risk estimates were close to or below 1.0 for long durations of use overall and in various categories of parity status and other factors. For women under the age of 40 years, the data were sparse (80 cases, 160 controls). Although there were some elevated relative risk estimates, most were not statistically significant, and there were no consistent patterns across duration of use. The present data add to the body of evidence that indicates that oral contraceptive use does not adversely affect the risk of breast cancer in older women; the data are inadequate to clarify the effect in younger women. Am J Epidemiol 1992;136:1437-44 breast neoplasms; contraceptives, oral
Many studies have assessed whether oral contraceptive use influences the risk of breast cancer (1-20). In most studies conducted before the 1980s, results were reassuring (1, 2). Because oral contraceptives were first marketed in the early 1960s, however, the women in those studies had little
opportunity to use them for long periods. In addition, few had used oral contraceptives early in their reproductive lives. Over time, there has been a trend toward earlier commencement of use. In some of the more recent studies, oral contraceptive use has been associated with
Received for publication February 6, 1992, and in final form June 1, 1992. Abbreviation: Cl, confidence interval. 1 Stone Epidemiology Unit, School of Public Health, Boston University School of Medicine, Brookline, MA. 2 Ontario Cancer Treatment and Research Foundation, Toronto, Ontario, Canada. Reprint requests to Dr. Lynn Rosenberg, Stone Epidemiology Unit, 1371 Beacon Street, Brookline, MA 02146. This work was supported by the Alcoholic Beverage Medical Research Foundation, Baltimore, MD; cooperative agreements U01 FD01222 and FD-U-000082 from the US Food and Drug Administration; Hoffmann-LaRoche, Inc., Nutley, NJ; Ciba-Geigy Corporation, Summit, NJ, Hoechst AG, Frankfurt, West Germany; McNeil Pharmaceutical, Spring House, PA; Merrell Dow Pharmaceuticals, Inc.,
Cincinnati, OH; Ortho Pharmaceutical Corporation, Raritan, NJ; and National Coffee Association of U.S.A., Inc., New York, NY. The authors are grateful for the collaboration of the attending physicians at Princess Margaret Hospital, Sunnybrook Medical Centre, Toronto General Hospital, and Mt. Sinai Hospital. The authors are also grateful to the interviewers who collected the data, Anne Allen, Rosemary Chepa, Suzanne Gale, Kathleen Gaiespie, Lindsay Hat, Linda Harrel, Virginia Hunter, Beverly Linden, Muriel Relton, Shirley Rothery, Irene Servos, Ann Skene-Metvin, and Linda West; to Verna Cundari and Phyllis Disenhouse who coordinated the study; to Marguerite Angeloni who managed the data; and to Glenn Street and Leonard Gaetano for data analysis.
1437
1438
Rosenberg et al.
increased risks (3-12). Positive associations have been observed for long durations of oral contraceptive use overall, before the birth of the first child, or before age 25; for use at an early age; and for recent use. Although the positive associations have varied across studies, they have been most consistently observed among younger women. Other studies have found no association of oral contraceptive use with increased risk (13-17). Conflicting reports have been published on the largest case-control study, the Cancer and Steroid Hormone Study. At first, the study was interpreted as null (18, 19). Upon reanalysis (20), there was an association of oral contraceptive use with increased risk of breast cancer in women under the age of 35 years, but with no trend for the risk to increase with increasing duration of use; among older women there was a slightly decreased risk of breast cancer among oral contraceptive users, with a significant trend for the risk to decrease with increasing time since first and last use. In this report, we assess the relation of use of oral contraceptives to the risk of breast cancer in women under the age of 70, with data from a case-control study conducted in Toronto, Ontario, Canada. Particular attention is given to the hypotheses that have been raised in previous studies. MATERIALS AND METHODS Data collection
The data were collected from 1982 through 1986 from women under the age of 70 years who lived in metropolitan Toronto, in a study conducted to assess the relation of alcohol consumption to the risk of breast cancer. Participants were interviewed at home after they had given informed consent. Trained interviewers administered standard questionnaires to obtain information on demographic factors, risk factors for breast cancer, tobacco and alcohol use, and use of oral contraceptives, noncontraceptive estrogens, and other drugs. For each episode of drug use, the timing, duration, and preparation were recorded. Data on oral contraceptive use were elicited by questions about
drugs used for contraception, regulation of periods, menstrual problems, breast problems, endometriosis, infertility, and sexual difficulties. Photographs of oral contraceptive pills and packets marketed in Canada and the United States were used to aid recall. Cases. The cases were women with histologically confirmed first occurrences of breast cancer (excluding in situ cancer) who were identified through the records of the Princess Margaret Hospital, a tertiary referral hospital. Over the course of the study, there were 800 eligible patients, of whom 73 (9 percent) were not enrolled because of physician refusal and 97 (12 percent) because of their own refusal. After exclusion of 23 cases for whom controls could not be found, the final case series consisted of 607 women. Estrogen and progesterone receptor status were determined within 6 weeks of the diagnosis for 529 of the 607 cases (21). Controls. Controls were identified from the tax assessment rolls of all Ontario residents. Provincial law mandates the maintenance of these records; they are updated partially each year and completely every 3 years. For each case, two women who were in the same decade of age and lived closest to the case (except for next-door neighbors) were selected. Of 1,874 eligible controls, 1,214 (65 percent) participated. Data analysis
Conditional logistic regression (22) was used to estimate relative risks for oral contraceptive use relative to never use with control for the matching factors, age and neighborhood. As reported previously (23), in the present data increased risk of breast cancer was associated with late menopause, family history of breast cancer, history of fibrocystic breast disease, Jewish religion, and low parity. To control for these and other factors, terms were included in the conditional logistic regression for age at menarche, age at first birth, parity, age at menopause, body mass index (kg/m2), history of breast cancer in a mother or sister, duration of noncontraceptive estrogen use, Jewish religion, cigarette smoking, alcohol consumption, years of
Breast Cancer and Oral Contraceptive Use 1439
no trend of increasing relative risk with increasing duration of use. The estimate for >10 years of use was 0.9 (95 percent CI 0.61.3). Among women below the age of 40 years (80 cases, 160 controls), again there was no tendency for the relative risk estimates to increase as the duration of use increased. The point estimates were elevated for durations of use less than 5 years and approximated 1.0 for longer durations. All the estimates were compatible with 1.0. Data on duration of use according to whether the women were premenopausal or postmenopausal are shown in table 3. Among premenopausal women, the estimate for > 10 years of use was 1.1 (95 percent RESULTS CI 0.7-1.8) and, in postmenopausal women, Ever use of oral contraceptives was re- it was 0.7 (95 percent CI 0.4-1.2). Among women who reported a history of ported by 43 percent of cases and 45 percent breast cancer in a mother or sister (76 cases, of controls; the corresponding percentages for 10 or more years of use were 10 percent 81 controls), the relative risk estimate for and 10 percent (table 1). There was no tenthose who had used oral contraceptives for dency for the relative risk to increase as the >10 years was 1.1 (95 percent CI 0.3-4.0), duration of use increased; i.e., the multivarbased on five case and six control users. iate relative risk estimates for 10-14 and To assess whether risk was related to the >15 years of use were both 0.9. The largest recency of use, we considered oral contraestimate, 1.4, was for use that lasted less ceptive use according to the interval since than 1 year (95 percent confidence interval last use. For women who had ever used oral (CI) 0.9-2.0). For all duration categories, the contraceptives, the relative risk estimate was matched relative risk estimates were similar 0.8 (95 percent CI 0.4-1.5) for use that to the multivariate estimates. Only multicontinued into the year before interview (14 variate estimates are given in further analcases, 40 controls) and 1.2 (95 percent CI yses. 0.8-1.8) for use that continued into the 5In table 2, the duration of use is given year period before interview (53 cases, 96 according to age. Among women aged 40 or controls). For women who had used oral older (527 cases, 1,054 controls), there was contraceptives for 5 or more years, the cor-
education, and use of medical care in the previous year. History of fibrocystic breast disease was not included because oral contraceptive use influences the incidence of this illness, and the timing of the onset in relation to oral contraceptive use was uncertain. After control for 5-year age groups rather than decade of age, the estimates were similar. In addition, the estimates with only age and neighborhood controlled (matched estimates) were generally similar to those from analyses that controlled all potential confounding factors (multivariate estimates).
TABLE 1. Duration of oral contraceptive use in 607 cases of breast cancer and 1,214 neighbor controls: Toronto, Canada, 1082-1986 Duration of contraceptive use (years) Never O CO C\l Q T^ T^
d44J
dodo •r^ r
O
O
O
qinnom w ID 't T- 01
en co co eg
4
dodo i-;
Vol. 136, No. 12 Printed in U.S.A.
ORIGINAL CONTRIBUTIONS
A Case-Control Study of the Risk of Breast Cancer in Relation to Oral Contraceptive Use
Lynn Rosenberg,1 Julie R. Palmer,1 E. Aileen Clarke,2 and Samuel Shapiro1
The relation of oral contraceptive use to the risk of breast cancer was evaluated in a case-control study of women under the age of 70 years, conducted in Toronto, Canada, from 1982 through 1986. A total of 607 breast cancer cases identified in a cancer hospital were compared with 1,214 controls matched to the cases on neighborhood and decade of age. Information on oral contraceptive use and risk factors for breast cancer was collected in home interviews. Conditional logistic regression was used to control multiple confounding factors. For women aged 40-69 years (527 cases, 1,054 controls), trie results suggest that oral contraceptive use does not increase the risk of breast cancer. Multivariate relative risk estimates were close to or below 1.0 for long durations of use overall and in various categories of parity status and other factors. For women under the age of 40 years, the data were sparse (80 cases, 160 controls). Although there were some elevated relative risk estimates, most were not statistically significant, and there were no consistent patterns across duration of use. The present data add to the body of evidence that indicates that oral contraceptive use does not adversely affect the risk of breast cancer in older women; the data are inadequate to clarify the effect in younger women. Am J Epidemiol 1992;136:1437-44 breast neoplasms; contraceptives, oral
Many studies have assessed whether oral contraceptive use influences the risk of breast cancer (1-20). In most studies conducted before the 1980s, results were reassuring (1, 2). Because oral contraceptives were first marketed in the early 1960s, however, the women in those studies had little
opportunity to use them for long periods. In addition, few had used oral contraceptives early in their reproductive lives. Over time, there has been a trend toward earlier commencement of use. In some of the more recent studies, oral contraceptive use has been associated with
Received for publication February 6, 1992, and in final form June 1, 1992. Abbreviation: Cl, confidence interval. 1 Stone Epidemiology Unit, School of Public Health, Boston University School of Medicine, Brookline, MA. 2 Ontario Cancer Treatment and Research Foundation, Toronto, Ontario, Canada. Reprint requests to Dr. Lynn Rosenberg, Stone Epidemiology Unit, 1371 Beacon Street, Brookline, MA 02146. This work was supported by the Alcoholic Beverage Medical Research Foundation, Baltimore, MD; cooperative agreements U01 FD01222 and FD-U-000082 from the US Food and Drug Administration; Hoffmann-LaRoche, Inc., Nutley, NJ; Ciba-Geigy Corporation, Summit, NJ, Hoechst AG, Frankfurt, West Germany; McNeil Pharmaceutical, Spring House, PA; Merrell Dow Pharmaceuticals, Inc.,
Cincinnati, OH; Ortho Pharmaceutical Corporation, Raritan, NJ; and National Coffee Association of U.S.A., Inc., New York, NY. The authors are grateful for the collaboration of the attending physicians at Princess Margaret Hospital, Sunnybrook Medical Centre, Toronto General Hospital, and Mt. Sinai Hospital. The authors are also grateful to the interviewers who collected the data, Anne Allen, Rosemary Chepa, Suzanne Gale, Kathleen Gaiespie, Lindsay Hat, Linda Harrel, Virginia Hunter, Beverly Linden, Muriel Relton, Shirley Rothery, Irene Servos, Ann Skene-Metvin, and Linda West; to Verna Cundari and Phyllis Disenhouse who coordinated the study; to Marguerite Angeloni who managed the data; and to Glenn Street and Leonard Gaetano for data analysis.
1437
1438
Rosenberg et al.
increased risks (3-12). Positive associations have been observed for long durations of oral contraceptive use overall, before the birth of the first child, or before age 25; for use at an early age; and for recent use. Although the positive associations have varied across studies, they have been most consistently observed among younger women. Other studies have found no association of oral contraceptive use with increased risk (13-17). Conflicting reports have been published on the largest case-control study, the Cancer and Steroid Hormone Study. At first, the study was interpreted as null (18, 19). Upon reanalysis (20), there was an association of oral contraceptive use with increased risk of breast cancer in women under the age of 35 years, but with no trend for the risk to increase with increasing duration of use; among older women there was a slightly decreased risk of breast cancer among oral contraceptive users, with a significant trend for the risk to decrease with increasing time since first and last use. In this report, we assess the relation of use of oral contraceptives to the risk of breast cancer in women under the age of 70, with data from a case-control study conducted in Toronto, Ontario, Canada. Particular attention is given to the hypotheses that have been raised in previous studies. MATERIALS AND METHODS Data collection
The data were collected from 1982 through 1986 from women under the age of 70 years who lived in metropolitan Toronto, in a study conducted to assess the relation of alcohol consumption to the risk of breast cancer. Participants were interviewed at home after they had given informed consent. Trained interviewers administered standard questionnaires to obtain information on demographic factors, risk factors for breast cancer, tobacco and alcohol use, and use of oral contraceptives, noncontraceptive estrogens, and other drugs. For each episode of drug use, the timing, duration, and preparation were recorded. Data on oral contraceptive use were elicited by questions about
drugs used for contraception, regulation of periods, menstrual problems, breast problems, endometriosis, infertility, and sexual difficulties. Photographs of oral contraceptive pills and packets marketed in Canada and the United States were used to aid recall. Cases. The cases were women with histologically confirmed first occurrences of breast cancer (excluding in situ cancer) who were identified through the records of the Princess Margaret Hospital, a tertiary referral hospital. Over the course of the study, there were 800 eligible patients, of whom 73 (9 percent) were not enrolled because of physician refusal and 97 (12 percent) because of their own refusal. After exclusion of 23 cases for whom controls could not be found, the final case series consisted of 607 women. Estrogen and progesterone receptor status were determined within 6 weeks of the diagnosis for 529 of the 607 cases (21). Controls. Controls were identified from the tax assessment rolls of all Ontario residents. Provincial law mandates the maintenance of these records; they are updated partially each year and completely every 3 years. For each case, two women who were in the same decade of age and lived closest to the case (except for next-door neighbors) were selected. Of 1,874 eligible controls, 1,214 (65 percent) participated. Data analysis
Conditional logistic regression (22) was used to estimate relative risks for oral contraceptive use relative to never use with control for the matching factors, age and neighborhood. As reported previously (23), in the present data increased risk of breast cancer was associated with late menopause, family history of breast cancer, history of fibrocystic breast disease, Jewish religion, and low parity. To control for these and other factors, terms were included in the conditional logistic regression for age at menarche, age at first birth, parity, age at menopause, body mass index (kg/m2), history of breast cancer in a mother or sister, duration of noncontraceptive estrogen use, Jewish religion, cigarette smoking, alcohol consumption, years of
Breast Cancer and Oral Contraceptive Use 1439
no trend of increasing relative risk with increasing duration of use. The estimate for >10 years of use was 0.9 (95 percent CI 0.61.3). Among women below the age of 40 years (80 cases, 160 controls), again there was no tendency for the relative risk estimates to increase as the duration of use increased. The point estimates were elevated for durations of use less than 5 years and approximated 1.0 for longer durations. All the estimates were compatible with 1.0. Data on duration of use according to whether the women were premenopausal or postmenopausal are shown in table 3. Among premenopausal women, the estimate for > 10 years of use was 1.1 (95 percent RESULTS CI 0.7-1.8) and, in postmenopausal women, Ever use of oral contraceptives was re- it was 0.7 (95 percent CI 0.4-1.2). Among women who reported a history of ported by 43 percent of cases and 45 percent breast cancer in a mother or sister (76 cases, of controls; the corresponding percentages for 10 or more years of use were 10 percent 81 controls), the relative risk estimate for and 10 percent (table 1). There was no tenthose who had used oral contraceptives for dency for the relative risk to increase as the >10 years was 1.1 (95 percent CI 0.3-4.0), duration of use increased; i.e., the multivarbased on five case and six control users. iate relative risk estimates for 10-14 and To assess whether risk was related to the >15 years of use were both 0.9. The largest recency of use, we considered oral contraestimate, 1.4, was for use that lasted less ceptive use according to the interval since than 1 year (95 percent confidence interval last use. For women who had ever used oral (CI) 0.9-2.0). For all duration categories, the contraceptives, the relative risk estimate was matched relative risk estimates were similar 0.8 (95 percent CI 0.4-1.5) for use that to the multivariate estimates. Only multicontinued into the year before interview (14 variate estimates are given in further analcases, 40 controls) and 1.2 (95 percent CI yses. 0.8-1.8) for use that continued into the 5In table 2, the duration of use is given year period before interview (53 cases, 96 according to age. Among women aged 40 or controls). For women who had used oral older (527 cases, 1,054 controls), there was contraceptives for 5 or more years, the cor-
education, and use of medical care in the previous year. History of fibrocystic breast disease was not included because oral contraceptive use influences the incidence of this illness, and the timing of the onset in relation to oral contraceptive use was uncertain. After control for 5-year age groups rather than decade of age, the estimates were similar. In addition, the estimates with only age and neighborhood controlled (matched estimates) were generally similar to those from analyses that controlled all potential confounding factors (multivariate estimates).
TABLE 1. Duration of oral contraceptive use in 607 cases of breast cancer and 1,214 neighbor controls: Toronto, Canada, 1082-1986 Duration of contraceptive use (years) Never O CO C\l Q T^ T^
d44J
dodo •r^ r
O
O
O
qinnom w ID 't T- 01
en co co eg
4
dodo i-;
