Breast Cancer Research and Treatment 18: $23-$26, 1991. 9 1991 Kluwer Academic Publishers. Printed in the Netherlands.
Endogenous hormones and breast cancer: A prospective cohort study Paolo G. Toniolo 1'2, Bernard S. Pasternack 1,2, Roy E. Shore 1,2, Elizabeth Sonnenschein 1, Karen L. Koenig 1, Carl Rosenberg 1, Philip Strax 3 and Selig Strax 3 1Institute of Environmental Medicine, New York University Medical Center, New York, NY, USA; 2 Kaplan Cancer Center, New York University Medical Center, New York, NY, USA; 3 The Guttman Breast Diagnostic Institute, New York, NY, USA
Key words: breast neoplasms, estrogens, prolactin, prospective studies
Summary A cohort study is under way in New York City to evaluate how levels of endogenous reproductive hormones influence the risk of breast cancer. The study, in which approximately 15,000 women are being recruited, utilizes a prospective design in which volunteers are asked to provide repeated specimens of serum during the period 1985-1992. A case-control study nested within the cohort is planned by which specimens from all cases arising in the population and from a randomly selected sample of time-matched controls will be analyzed and compared. As of December 31, 1989, 13,609 volunteers had donated blood specimens, about 50% of whom had already donated more than once. Of the 187 incident breast cancer cases who are expected to arise in the cohort before the end of 1992, 77 have been detected thus far.
Introduction Epidemiologic research on the role of endogenous hormone levels in breast cancer has relied almost exclusively on case-control studies [1]. A major drawback of this approach is that the exposure is measured at, or after, the time of diagnosis of disease, rather than before. Since the temporal sequence of the cause-effect relationship is violated, it is impossible to conclude whether an observed association, or a lack of association, has any relevance to the etiology of the disease. The cohort (follow-up) design, in which hormone levels are determined months or years before cancer is diagnosed, eliminates or at least reduces the temporal defect of case-control studies. Large numbers of subjects are required, however, and
laboratory analyses on all cohort members would be prohibitively expensive. An alternative is to sample from the original cohort [2]. The most widely used sampling scheme, known as 'nested casecontrol' sampling, involves selection of controls from those at risk at the failure time of each case [3]. Instead of analyzing specimens from the entire cohort, those of all cases arising in the population and of a randomly selected sample of time-matched controls are used, thereby reducing labor and costs with only a marginal loss in statistical power. In this paper we describe a cohort study which we are conducting in New York City which we plan to analyze using this approach. Our main research purpose is to determine how levels of endogenous reproductive hormones - initially estrogens and prolactin - influence the risk of breast cancer, uti-
Address for offprints: P. Toniolo, NYU Medical Center, 341 East 25th Street, New York, N.Y. 10010, USA
$24
P G Toniolo et al.
lizing a prospective design with repeated endocrine measurements.
Materials and methods
Women aged 34-65 attending the Guttman Breast Diagnostic Institute (GBDI) in New York City for yearly screenings that include a mammogram and breast examination are asked to participate. Women who have taken hormone medications or have been pregnant in the previous six months are excluded. During the 8 years of data collection (198592), we plan to assemble a cohort of approximately 15,000 women. Thirty ml of venous blood are drawn at entry. Serum is fractionated into 12 or more 1-ml aliquots and frozen at - 80~C for storage. Additional blood donations are solicited at each annual visit to the clinic. Cohort members complete a self-administered questionnaire to report basic demographic information and to answer questions on family history of breast cancer, reproductive history, surgery to reproductive organs, benign breast conditions, physical activity, and recent use of medications. They also complete a self-administered semi-quantitative food-frequency dietary questionnaire adapted from one developed at the National Cancer Institute [4]. Follow-up is based on yearly questionnaires answered by cohort members who return to the GBDI for screenings. Those who fail to return are contacted by mail or telephone. Considerable ef-
fort is expended in locating women who have moved or whose records are inconsistent. Breast cancer cases are confirmed by obtaining a copy of pertinent clinical and pathological information from hospitals. Cases are being identified also through record linkages with the statewide tumor registries of New York, New Jersey, and Connecticut. In the nested case-control study cases are cohort members who develop a carcinoma of the breast at any time after entering the cohort. Individuallymatched controls are selected at random from the appropriate risk set for each case. The risk set for a case consists of all cohort members alive and free of breast cancer at the time of diagnosis of the case who match the case on menopausal status, age at entry, and number and dates of blood donations. For premenopausal cases, risk set definition also includes day of menstrual cycle at the time of first blood sampling, calculated back from the onset of the menses immediately following blood donation. Two controls are selected for each postmenopausal case and four for each premenopausal case. In the event there are not enough controls satisfying the matching requirements in a given risk set, the matching criteria are relaxed one at a time in a prescribed order and amount until a sufficient number of controls are available or until a preset limit of relaxation is reached. Cases and controls are interviewed by telephone by a trained interviewer who is unaware of their
T a b l e 2. Identification of breast cancer cases, by year of diagnosis (Follow-up in progress, data incomplete) T a b l e 1. Distribution of blood donations by n u m b e r of donations and year of entry into the study (as of December 31, 1989)
Year of diagnosis 1985
Year of donation
Blood donation 1st
1985 1986 1987 1988 1989
2nd
5,610 2 5,136 2,465 1,763 2,627 581 868 519 708 13,609 6,670
1986
1987
1988
1989
1985-89 Total
12 2
24 7
24 2
17 3
77 14
-
-
53
Total 3rd
4th
1,082 1,105 937 3,124
425 651 1,124
5th 1 197 198
Incident: * Invasive 5,612 7,601 5,472 2,980 3,012 24,677
In situ
Prevalent: * * Invasive In situ
19 .
25 .
9 .
.
.
* First diagnosed > 180 days after enrollment ** First diagnosed -< 180 days after enrollment
0
Endogenous hormones and breast cancer case or control status. The interview focuses on indicators of hormone exposure as well as potential confounders or risk modifiers, such as reproductive and medical history, lifetime use of alcohol and cigarettes, and history of use of contraceptive and replacement hormones. Surrogate interviews are conducted with close relatives for cases or controls who have died or are incapacitated.
Study progress Table 1 summarizes our progress in subject recruitment and serum specimen collection through December 31, 1989. The cumulated number of blood donations was 24,677, or more than 300,000 1-ml aliquots in storage. Of the cohort members entering in 1985, more than 60% of those age 45 and older and about 50% of those aged less than 45 have contributed two or more blood samples. At enrollment, 42% of cohort members were premenopausal, and 45% were younger than age 45. Table 2 summarizes our progress in identifying breast cancer cases. Although follow-up coverage is above 90% for the years 1985 through 1987, it is obviously less complete for more recent years. The age-distribution of cases, separately identified as prevalent at enrollment or incident thereafter, and as invasive or in situ, is provided in Table 3. In Table 4 we show a comparison between the number of invasive incident breast cancer cases identified through 1988 and the number expected on the basis of the 1978-82 incidence data for the
Table 3. Follow-up. Distribution of breast cancer cases by age at diagnosis (Follow-up in progress, data incomplete) Age
< 40 40-49 50-59 60-69 70 + All ages
Prevalent
3 18 15 17 53
Incident Invasive
In situ
2 20 26 29 77
4 8 2 14
$25
state of New York. Despite our partially complete follow-up, the two distributions look similar.
Discussion Despite extensive epidemiological research, the evidence relating endogenous hormones and breast cancer is unconvincing and often conflicting [1, 5]. Large-scale prospective studies involving longterm storage of biological specimens are needed to address this issue. We are creating a large bank of stored biological specimens, which includes aliquots of serum, red cell membranes, and blood clots, and which is available for investigations in different areas, as well as for international comparisons. State-of-the-art assays could be added to the study as they become available, for many years into the future. Furthermore, in a few years we will have identified sufficient cases of other frequent cancers (lung, ovary, endometrium, colon, and rectum), to conduct studies similar in design to the one just described.
Acknowledgements This work is supported by Grant CA-34588 (to B.S.P.) from the National Cancer Institute, Center Grants CA-16087 and CA-13343 from the National Cancer Institute, and Center Grant ES-00260 from Table 4. Comparison between n u m b e r s of invasive incident breast cancer cases identified in 1985-88 and n u m b e r s expected, by age at diagnosis (Follow-up in progress, data incomplete) Age
Expected*
Observed
< 40 40-44 45-49 50-54 55-59 60-64 65-69 70 + All ages
1.6 5.2 8.6 10.2 14.8 18.0 7.2 0.7 66.3
1 4 11 11 12 15 6 0 60
* Calculated on the basis of the (1978-82) incidence data for New York State.
$26
P G Toniolo et al.
the National Institute of Environmental Health Sciences. We thank F. Mastrota, A. Bassiri, S. Porco, L. Quinones, L. Sen, and B. Whiton for technical assistance. We acknowledge the collaboration of Drs. P. Bruning, M. Levitz, R. Bulbrook, J. Moore, and D. Wang. We also acknowledge the assistance provided by the Guttman Breast Diagnostic Institute in every phase of this project.
References 1. Zumoff B: Hormonal profiles in women with breast cancer (Review). Anticancer Research 8: 627-636, 1988 2. Mantel N: Synthetic retrospective studies and related topics. Biometrics 29: 479-486, 1973 3. Liddel FDK, McDonald JC, Thomas DC: Methods for cohort analysis: appraisal by application to asbestos mining (with discussion). Royal Stat Soc 140: 469-491, 1977 4. Block G, Hartman A M , Dresser CM, et al.: A data-based approach to diet questionnaire design and testing. Am J Epidemiol 124: 453-469, 1986 5. Shore RE, Pasternack BS, Bulbrook RD, et al.: Endocrine and environmental factors in breast cancer: the case for prospective studies. Commentaries on Research in Breast Cancer 3: 1-31, 1983
Endogenous hormones and breast cancer: A prospective cohort study Paolo G. Toniolo 1'2, Bernard S. Pasternack 1,2, Roy E. Shore 1,2, Elizabeth Sonnenschein 1, Karen L. Koenig 1, Carl Rosenberg 1, Philip Strax 3 and Selig Strax 3 1Institute of Environmental Medicine, New York University Medical Center, New York, NY, USA; 2 Kaplan Cancer Center, New York University Medical Center, New York, NY, USA; 3 The Guttman Breast Diagnostic Institute, New York, NY, USA
Key words: breast neoplasms, estrogens, prolactin, prospective studies
Summary A cohort study is under way in New York City to evaluate how levels of endogenous reproductive hormones influence the risk of breast cancer. The study, in which approximately 15,000 women are being recruited, utilizes a prospective design in which volunteers are asked to provide repeated specimens of serum during the period 1985-1992. A case-control study nested within the cohort is planned by which specimens from all cases arising in the population and from a randomly selected sample of time-matched controls will be analyzed and compared. As of December 31, 1989, 13,609 volunteers had donated blood specimens, about 50% of whom had already donated more than once. Of the 187 incident breast cancer cases who are expected to arise in the cohort before the end of 1992, 77 have been detected thus far.
Introduction Epidemiologic research on the role of endogenous hormone levels in breast cancer has relied almost exclusively on case-control studies [1]. A major drawback of this approach is that the exposure is measured at, or after, the time of diagnosis of disease, rather than before. Since the temporal sequence of the cause-effect relationship is violated, it is impossible to conclude whether an observed association, or a lack of association, has any relevance to the etiology of the disease. The cohort (follow-up) design, in which hormone levels are determined months or years before cancer is diagnosed, eliminates or at least reduces the temporal defect of case-control studies. Large numbers of subjects are required, however, and
laboratory analyses on all cohort members would be prohibitively expensive. An alternative is to sample from the original cohort [2]. The most widely used sampling scheme, known as 'nested casecontrol' sampling, involves selection of controls from those at risk at the failure time of each case [3]. Instead of analyzing specimens from the entire cohort, those of all cases arising in the population and of a randomly selected sample of time-matched controls are used, thereby reducing labor and costs with only a marginal loss in statistical power. In this paper we describe a cohort study which we are conducting in New York City which we plan to analyze using this approach. Our main research purpose is to determine how levels of endogenous reproductive hormones - initially estrogens and prolactin - influence the risk of breast cancer, uti-
Address for offprints: P. Toniolo, NYU Medical Center, 341 East 25th Street, New York, N.Y. 10010, USA
$24
P G Toniolo et al.
lizing a prospective design with repeated endocrine measurements.
Materials and methods
Women aged 34-65 attending the Guttman Breast Diagnostic Institute (GBDI) in New York City for yearly screenings that include a mammogram and breast examination are asked to participate. Women who have taken hormone medications or have been pregnant in the previous six months are excluded. During the 8 years of data collection (198592), we plan to assemble a cohort of approximately 15,000 women. Thirty ml of venous blood are drawn at entry. Serum is fractionated into 12 or more 1-ml aliquots and frozen at - 80~C for storage. Additional blood donations are solicited at each annual visit to the clinic. Cohort members complete a self-administered questionnaire to report basic demographic information and to answer questions on family history of breast cancer, reproductive history, surgery to reproductive organs, benign breast conditions, physical activity, and recent use of medications. They also complete a self-administered semi-quantitative food-frequency dietary questionnaire adapted from one developed at the National Cancer Institute [4]. Follow-up is based on yearly questionnaires answered by cohort members who return to the GBDI for screenings. Those who fail to return are contacted by mail or telephone. Considerable ef-
fort is expended in locating women who have moved or whose records are inconsistent. Breast cancer cases are confirmed by obtaining a copy of pertinent clinical and pathological information from hospitals. Cases are being identified also through record linkages with the statewide tumor registries of New York, New Jersey, and Connecticut. In the nested case-control study cases are cohort members who develop a carcinoma of the breast at any time after entering the cohort. Individuallymatched controls are selected at random from the appropriate risk set for each case. The risk set for a case consists of all cohort members alive and free of breast cancer at the time of diagnosis of the case who match the case on menopausal status, age at entry, and number and dates of blood donations. For premenopausal cases, risk set definition also includes day of menstrual cycle at the time of first blood sampling, calculated back from the onset of the menses immediately following blood donation. Two controls are selected for each postmenopausal case and four for each premenopausal case. In the event there are not enough controls satisfying the matching requirements in a given risk set, the matching criteria are relaxed one at a time in a prescribed order and amount until a sufficient number of controls are available or until a preset limit of relaxation is reached. Cases and controls are interviewed by telephone by a trained interviewer who is unaware of their
T a b l e 2. Identification of breast cancer cases, by year of diagnosis (Follow-up in progress, data incomplete) T a b l e 1. Distribution of blood donations by n u m b e r of donations and year of entry into the study (as of December 31, 1989)
Year of diagnosis 1985
Year of donation
Blood donation 1st
1985 1986 1987 1988 1989
2nd
5,610 2 5,136 2,465 1,763 2,627 581 868 519 708 13,609 6,670
1986
1987
1988
1989
1985-89 Total
12 2
24 7
24 2
17 3
77 14
-
-
53
Total 3rd
4th
1,082 1,105 937 3,124
425 651 1,124
5th 1 197 198
Incident: * Invasive 5,612 7,601 5,472 2,980 3,012 24,677
In situ
Prevalent: * * Invasive In situ
19 .
25 .
9 .
.
.
* First diagnosed > 180 days after enrollment ** First diagnosed -< 180 days after enrollment
0
Endogenous hormones and breast cancer case or control status. The interview focuses on indicators of hormone exposure as well as potential confounders or risk modifiers, such as reproductive and medical history, lifetime use of alcohol and cigarettes, and history of use of contraceptive and replacement hormones. Surrogate interviews are conducted with close relatives for cases or controls who have died or are incapacitated.
Study progress Table 1 summarizes our progress in subject recruitment and serum specimen collection through December 31, 1989. The cumulated number of blood donations was 24,677, or more than 300,000 1-ml aliquots in storage. Of the cohort members entering in 1985, more than 60% of those age 45 and older and about 50% of those aged less than 45 have contributed two or more blood samples. At enrollment, 42% of cohort members were premenopausal, and 45% were younger than age 45. Table 2 summarizes our progress in identifying breast cancer cases. Although follow-up coverage is above 90% for the years 1985 through 1987, it is obviously less complete for more recent years. The age-distribution of cases, separately identified as prevalent at enrollment or incident thereafter, and as invasive or in situ, is provided in Table 3. In Table 4 we show a comparison between the number of invasive incident breast cancer cases identified through 1988 and the number expected on the basis of the 1978-82 incidence data for the
Table 3. Follow-up. Distribution of breast cancer cases by age at diagnosis (Follow-up in progress, data incomplete) Age
< 40 40-49 50-59 60-69 70 + All ages
Prevalent
3 18 15 17 53
Incident Invasive
In situ
2 20 26 29 77
4 8 2 14
$25
state of New York. Despite our partially complete follow-up, the two distributions look similar.
Discussion Despite extensive epidemiological research, the evidence relating endogenous hormones and breast cancer is unconvincing and often conflicting [1, 5]. Large-scale prospective studies involving longterm storage of biological specimens are needed to address this issue. We are creating a large bank of stored biological specimens, which includes aliquots of serum, red cell membranes, and blood clots, and which is available for investigations in different areas, as well as for international comparisons. State-of-the-art assays could be added to the study as they become available, for many years into the future. Furthermore, in a few years we will have identified sufficient cases of other frequent cancers (lung, ovary, endometrium, colon, and rectum), to conduct studies similar in design to the one just described.
Acknowledgements This work is supported by Grant CA-34588 (to B.S.P.) from the National Cancer Institute, Center Grants CA-16087 and CA-13343 from the National Cancer Institute, and Center Grant ES-00260 from Table 4. Comparison between n u m b e r s of invasive incident breast cancer cases identified in 1985-88 and n u m b e r s expected, by age at diagnosis (Follow-up in progress, data incomplete) Age
Expected*
Observed
< 40 40-44 45-49 50-54 55-59 60-64 65-69 70 + All ages
1.6 5.2 8.6 10.2 14.8 18.0 7.2 0.7 66.3
1 4 11 11 12 15 6 0 60
* Calculated on the basis of the (1978-82) incidence data for New York State.
$26
P G Toniolo et al.
the National Institute of Environmental Health Sciences. We thank F. Mastrota, A. Bassiri, S. Porco, L. Quinones, L. Sen, and B. Whiton for technical assistance. We acknowledge the collaboration of Drs. P. Bruning, M. Levitz, R. Bulbrook, J. Moore, and D. Wang. We also acknowledge the assistance provided by the Guttman Breast Diagnostic Institute in every phase of this project.
References 1. Zumoff B: Hormonal profiles in women with breast cancer (Review). Anticancer Research 8: 627-636, 1988 2. Mantel N: Synthetic retrospective studies and related topics. Biometrics 29: 479-486, 1973 3. Liddel FDK, McDonald JC, Thomas DC: Methods for cohort analysis: appraisal by application to asbestos mining (with discussion). Royal Stat Soc 140: 469-491, 1977 4. Block G, Hartman A M , Dresser CM, et al.: A data-based approach to diet questionnaire design and testing. Am J Epidemiol 124: 453-469, 1986 5. Shore RE, Pasternack BS, Bulbrook RD, et al.: Endocrine and environmental factors in breast cancer: the case for prospective studies. Commentaries on Research in Breast Cancer 3: 1-31, 1983
