|
|
| |||||||||||||||||||||||||||||||
Background Fibroadenomas are benign breast tumors that are commonly diagnosed in young women and are associated with a slight increase in the risk of breast cancer. These lesions vary considerably in their histologic characteristics. We assessed the correlation between the histologic features of fibroadenomas and the risk of subsequent breast cancer.
Methods We conducted a retrospective cohort study of a consecutive series of patients with fibroadenoma diagnosed between 1950 and 1968. Follow-up data were obtained for 1835 patients (90 percent of those eligible). Fibroadenomas with cysts, sclerosing adenosis, epithelial calcifications, or papillary apocrine changes were classified as complex. The rate of subsequent breast cancer among the patients was compared with the rates in two control groups, women listed in the Connecticut Tumor Registry and women chosen from among the patients' sisters-in-law.
Results The risk of invasive breast cancer was 2.17 times higher among the patients with fibroadenoma than among the controls (95 percent confidence interval, 1.5 to 3.2). The relative risk increased to 3.10 among patients with complex fibroadenomas (95 percent confidence interval, 1.9 to 5.1) and remained elevated for decades after diagnosis. Patients with benign proliferative disease in the parenchyma adjacent to the fibroadenoma had a relative risk of 3.88 (95 percent confidence interval, 2.1 to 7.3). Patients with a family history of breast cancer in whom complex fibroadenoma was diagnosed had a relative risk of 3.72, as compared with controls with a family history (95 percent confidence interval, 1.4 to 10). Two thirds of the patients had noncomplex fibroadenomas and no family history of breast cancer and did not have an increased risk.
Conclusions Fibroadenoma is a long-term risk factor for breast cancer. The risk is increased in women with complex fibroadenomas, proliferative disease, or a family history of breast cancer.
Fibroadenomas exhibit a wide range of cytologic and histologic patterns; the epithelial component can vary from an absence of hyperplastic activity to carcinoma in situ. It would thus be valuable if a subgroup of patients with fibroadenoma could be identified who are at a particularly high risk for breast cancer. To investigate this possibility, we conducted a long-term, retrospective cohort study of patients who presented with fibroadenoma at three Nashville hospitals between 1950 and 1968.
Methods
Criteria for Enrollment
The hospitals in this study were Vanderbilt University Hospital, St. Thomas Hospital, and Baptist Hospital. Recruitment for the study was carried out from 1952 to 1968 for Vanderbilt University Hospital and St. Thomas Hospital and from 1950 to 1968 for Baptist Hospital. Women were considered for this study if a biopsy revealed a fibroadenoma at a study hospital during the recruitment period. A patient's entry biopsy was the first biopsy revealing fibroadenoma during the recruitment period. Patients were eligible if they were residents of Tennessee or Kentucky at the time of the entry biopsy, both breasts were intact within six months after the entry biopsy, and no breast cancer developed before or within six months after the entry biopsy. Patients were excluded if their entry-biopsy slides or hospital charts were lost. At Baptist Hospital, the consent of the patient's surgeon was required for enrollment. Follow-up was begun six months after the entry biopsy was performed.
Control Groups
There were two control groups in this study. The first consisted of the patients' sisters-in-law. The patients (or their next of kin) were asked to list the names and birth years of all sisters-in-law (whether alive or dead). The two sisters-in-law nearest in age to each patient were then selected as controls. Patients who had one sister-in-law or none were assigned that one control or none. Follow-up of each control was begun when she was six months older than the patient had been when her fibroadenoma was diagnosed. To serve as a control, a sister-in-law had to be at risk for her first breast cancer at the start of follow-up. We chose to use sisters-in-law as controls to obtain a population-based control group that would be roughly matched to the patients for socioeconomic factors but that would allow retrospective follow-up without laborious effort.
The second control group consisted of women from the state of Connecticut who were at risk for breast cancer during the recruitment and follow-up periods of this study. Data on the incidence of breast cancer according to both year of diagnosis and five-year age interval were obtained from the Connecticut Tumor Registry8 and used to adjust our risk estimates for temporal trends in the incidence of breast cancer as well as for age at biopsy. A total of 62,848 newly diagnosed cases of breast cancer in women were recorded in this registry between 1950 and 1991 (data collected after 1979 from this registry were obtained from P.D. Sullivan [personal communication]). The denominator population from which the incidence rates in the registry were derived was the total of female residents of the state of Connecticut.
Histologic Examination
The histologic slides of all patients were reviewed by two pathologists working independently. Lesions had to be at least 4 mm in diameter and meet Fechner's criteria1 to be classified as fibroadenomas. The epithelial component of the fibroadenomas was assessed according to criteria similar to those used in our previous studies9,10. Fibroadenomas were classified as complex if they contained cysts greater than 3 mm in diameter, sclerosing adenosis, epithelial calcifications, or papillary apocrine changes9. When sufficient parenchyma was available adjacent to the fibroadenoma, it was classified as being free of proliferative disease, as having proliferative disease without atypia, or as having atypical hyperplasia according to Page's criteria9,10,11. At least 0.5 cm2 of parenchyma or five lobular units were required for such a diagnosis (a lobular unit is a cluster of acini that drain into a common terminal duct). The study pathologists resolved discrepant diagnoses with a double-viewing microscope.
Follow-up
Approximately 12,000 breast biopsies that revealed only benign breast tissue were performed at the study hospitals during the recruitment periods. Of these, 2458 biopsies demonstrated fibroadenomas. Two hundred thirty-two of these biopsies were performed in women who were ineligible for follow-up: 12 biopsies were performed in women who had a history of breast cancer or who had breast cancer within six months after the biopsy revealing a fibroadenoma; 13 in women who underwent bilateral mastectomy within six months after the biopsy; 38 in women who were not residents of Tennessee or Kentucky at the time of biopsy; 41 in women treated by surgeons at Baptist Hospital who did not give permission for follow-up data to be collected; 124 in women whose hospital charts were lost; and 4 in women whose histologic slides were lost. The exact number of women in whom these biopsies were carried out is unknown, because multiple biopsies were performed in some patients. Follow-up was obtained for 2010 of the 2226 biopsies (90.3 percent) performed in 1835 eligible women.
There were 463 patients who had no sisters-in-law, 444 who had one sister-in-law, 808 who had two or more, and 120 who had an unknown number. Thus, 444 patients were matched with a single control and 808 with two controls, for a total of 2060 potential controls. Follow-up was successfully completed in 1651 controls (80.1 percent). Of these, 11 did not meet the criteria for enrollment, for a total of 1640 controls.
Both the patients and their controls (or their next of kin) provided follow-up information about the outcome of breast cancer and about other epidemiologic risk factors for breast cancer through a telephone interview with the same questionnaire. Follow-up was terminated when any of the following events occurred in a subject (patient or control): both breasts were removed for reasons other than breast cancer (49 women), invasive breast cancer was diagnosed (130 women), the subject was interviewed (3095 women), or the subject died of causes other than breast cancer (151 women). There were an additional 50 women whom we were unable to interview but who were known to be free of breast cancer at some point after their entry biopsy. Follow-up of these women was terminated when they were last known to be free of breast cancer. All reported cases of breast cancer were confirmed -- 78 percent by a review of histologic slides, and the remainder by a review of medical records.
Statistical Analysis
The risk of breast cancer among the patients relative to the risk among the women listed in the Connecticut Tumor Registry was derived from standard morbidity ratios with the approach of Breslow et al.12. We used incidence data on breast cancer broken down according to both year of diagnosis and five-year age interval to adjust our relative-risk estimates for age at diagnosis, year of diagnosis, and length of follow-up.
The risk of breast cancer among the patients relative to that among their sisters-in-law was evaluated by proportional-hazards regression analysis13. Model selection was guided by examination of model-deviance statistics14 and the size of parameter estimates for potential confounding variables. All relative-risk estimates were adjusted for age at the biopsy revealing fibroadenoma, parity, age at the birth of the first child, and age at menarche by including appropriate variables and parameters in the models. These covariates appreciably reduced the model deviance. Age at menopause was considered a potential confounding variable but was excluded from our models because its effect on the model deviance was trivial. Interaction terms were included in the models whenever multiple risk factors were assessed simultaneously. Additional parameters were included to adjust for missing values as needed. Different models were used depending on the number of risk factors being simultaneously evaluated. The potential effects of multivariate outliers on our parameter estimates were examined with the delta-beta method15. These analyses showed that most of our relative-risk estimates were not noticeably affected by the exclusion of any given patient. Some patients, however, did have an appreciable influence on the relative-risk estimates that were associated with wide confidence intervals. The assumption of proportional hazards was tested by fitting models with time-dependent-hazard step functions. These step functions had a trivial effect on the model deviance, indicating that the proportional-hazards model remained reasonable more than 20 years after the entry biopsy. These time-dependent models were also used to assess the risk of breast cancer during the first 10 years after biopsy, 11 to 20 years after biopsy, and more than 20 years after biopsy. Confidence intervals of the relative-risk estimates were derived from the asymptotic covariance matrix of the parameter estimates.
Cumulative morbidity curves were derived with the Kaplan-Meier method13 and compared by the log-rank test13. Kaplan-Meier life tables were also used to estimate the median age at menopause in the patients and their sisters-in-law. All P values were calculated with respect to two-sided alternative hypotheses. Hazard regression analyses were performed with the BMDP 2L program,16 and delta-beta residual analyses with the Pecan program17. Standard morbidity ratios were derived with an SAS program18 in combination with a program we wrote in Fortran. All other results were derived with SAS programs18.
Results
Table 1 summarizes the findings of 2458 consecutive biopsy specimens containing fibroadenomas. In 654 of the biopsies (26.6 percent), insufficient adjacent parenchyma had been removed to permit the parenchyma to be evaluated. The proportion of fibroadenomas classified as complex was similar among the biopsy specimens with external parenchyma and those without it. Table 2 shows the distribution of proliferative disease within and adjacent to fibroadenomas in the 1804 consecutive specimens containing external parenchyma that could be evaluated. Proliferative disease was more common adjacent to the complex fibroadenomas than adjacent to the noncomplex lesions (P = 0.002). The study pathologists independently reached the same diagnosis of complex fibroadenoma in more than 90 percent of cases.
|
|
|
|
|
|
|
|
Our findings are consistent with other studies2,3,4,5,6 that found a slight increase in the risk of breast cancer among women with fibroadenoma, but our results also show that the histologic features of the fibroadenoma influence the risk of breast cancer. Fibroadenomas display considerable morphologic variability1. Women whose tumors contained cysts, sclerosing adenosis, epithelial calcifications, or papillary apocrine changes were found to be at increased, and roughly comparable, risk of invasive breast cancer. It thus makes sense on epidemiologic grounds to lump these lesions together under the rubric "complex fibroadenoma." However, it is not clear that these lesions are necessarily related by a common underlying cause.
Of particular importance is the evidence that the increased risk of breast cancer persists for more than 20 years after the diagnosis of fibroadenoma. Since fibroadenomas are commonly diagnosed before the age of 30 years, these lesions provide a means of identifying young women who have an increased risk of breast cancer decades before the onset of invasive disease. Breast cancer develops and evolves over many years; events such as age at first birth or age at menarche affect the risk of breast cancer decades after their occurrence. Thus, there is great interest in identifying the somatic changes that occur early in this process.
Proliferative disease in breast epithelium is an established risk factor for breast cancer9,19,20,21. In our study, proliferative disease in the parenchyma adjacent to a fibroadenoma had an appreciable effect on the risk of breast cancer. The physician should take into account the presence or absence of proliferative disease and the histologic features of the fibroadenoma when discussing the risk with the patient.
A history of breast cancer in a first-degree relative is also an important risk factor among patients with fibroadenomas. Among patients with a family history and either complex fibroadenomas or proliferative disease, the incidence of breast cancer during the first 25 years after diagnosis of the fibroadenoma was 20 percent. It is known22 that abnormalities in the BRCA1 gene on chromosome 17q21.1 increase the lifetime probability of breast cancer to more than 80 percent. Therefore, routinely testing patients with fibroadenomas for mutations in this gene after it has been found and cloned may prove worthwhile.
The clinical implications of these results are most important for women with a family history of breast cancer. For these women, the presence of either adjacent proliferative disease or a complex fibroadenoma adds to the magnitude of their risk of breast cancer. In our view, a diagnosis of a complex fibroadenoma should give a patient with a family history additional incentive to undergo regular mammographic surveillance starting at the age of 35 or 40. Although it may occasionally be technically difficult, the inclusion of some adjacent parenchyma when a fibroadenoma is removed also seems appropriate. Women with noncomplex fibroadenomas who have neither adjacent proliferative disease in the breast nor a family history of breast cancer are not at an elevated risk for the disease. It should be emphasized that two thirds of patients with fibroadenoma have neither a complex lesion nor a family history and may be reassured by the knowledge that their risk of breast cancer is not appreciably influenced by their tumor.
Supported by grants (CA-31698 and CA-50468) from the National Cancer Institute.
We are indebted to Dr. Marie R. Griffin for her helpful suggestions; to Paul D. Sullivan for providing unpublished incidence data on cancer in the state of Connecticut; to Marcia G. Freudenthal, Cheryl D. Sharpe, Rosalie A. Duffy, Julia F. Smith, and Kay B. Covington for their skilled and dedicated pursuit of follow-up information; to over 60 Nashville surgeons and the staffs of the tumor registries and records departments of three Nashville hospitals for their cooperation and aid in patient follow-up; and to R. Janelle Steele for assistance in preparing the manuscript.
Source Information
From the Departments of Preventive Medicine (W.D.D., W.D.P., M.S.R., P.A.S.) and Pathology (D.L.P., F.F.P., C.L.V.-J.), Vanderbilt University School of Medicine, Nashville.
Address reprint requests to Dr. Dupont at Vanderbilt University School of Medicine, Department of Preventive Medicine, A-1124 Medical Center N., Nashville, TN 37232-2637.
References
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
