Relationship Between Mammographic and Histological Risk Factors for Breast Cancer N. F. Boyd,* H. M. Jensen, G. Cooke, H. Lee Han

Information about future risk of breast cancer can be obtained from histological features of the breast epithelium detected in biopsy specimens (7,2) and from radiological features of the breast tissue detected in mammograms (3,4). The 1170

Received January 21, 1992; revised March 19, 1992; accepted May 12, 1992. Supported by a grant from the National Cancer Institute of Canada. N. F. Boyd is the recipient of a National Health Scientist Award, Health and Welfare Canada. In collaboration with the Reference Pathologists of the Canadian National Breast Screening Study—F. Alexander (Tom Baker Cancer Centre, Calgary, Alberta), Y. Boivin, W. Schurch (Departement Pathologie, Hopital Hotel Dieu, Montreal, Quebec), N. Cooter (Department of Pathology, Mount Sinai Hospital, Toronto, Ontario), J. Danyluk (Pathology Department, Misericordia Hospital, Edmonton, Alberta), D. Dawson (Red Deer Regional Hospital, Red Deer, Alberta), T. D'Souza (Department of Pathology, Henderson General Hospital, Hamilton, Ontario), M. Jabi (Canadian Reference Centre, Ottawa Civic Hospital, Ottawa, Ontario), S. Jacob (Anatomo-Pathologie, Hopital du St-Sacrement, Quebec), J. Safneck (Department of Pathology, Health Sciences Centre, University of Manitoba, Winnipeg, Manitoba), D. I. Turnbull (Department of Pathology, Victoria Hospital, London, Ontario), R. Vauclair, Departement Pathologie, Hopital Notre Dame, Montreal, Quebec), A. Worth (Anatomical Pathology and Clinical Pathology, Cancer Control Agency of British Columbia, Vancouver, British Columbia), H. J. Yazdi (Department of Laboratory Medicine, Ottawa Civic Hospital, Ottawa), I. Zayid (Anatomical Pathology Division, Victoria General Hospital, Halifax, Nova Scotia), and A. B. Miller (Canadian National Breast Cancer Screening Program. University of Toronto, Ontario). N. F. Boyd, H. L. Han, Division of Epidemiology and Statistics, Ontario Cancer Institute, Canada. H. M. Jensen, Department of Pathology, University of California, Davis. G. Cooke, Department of Radiology, St. Michael's Hospital, Toronto, Ontario, Canada. *Correspondence to: N. F. Boyd, M.D., FRCPC, Division of Epidemiology and Statistics, Ontario Cancer Institute, 500 Sherbourne St., Toronto. ON M4T 1K9, Canada.

Journal of the National Cancer Institute

Downloaded from http://jnci.oxfordjournals.org/ at Russian Archive on December 9, 2013

Background: Information on breast cancer risk can be obtained both from the histological appearance of the breast epithelium in biopsy specimens and from the pattern of parenchymal densities in the breast revealed by mammography. It is not understood, however, how parenchymal densities influence breast cancer risk or whether these densities are associated with histological risk factors. Purpose: We have estimated, in a large cohort of women, the relative risk of detecting carcinoma in situ, atypical hyperplasia, hyperplasia without atypia, or nonproliferative disease in biopsy specimens from women with different extents of mammographic density. We also examined the association between these histological classifications and radiological features present specifically at the biopsy site. Methods: The source of study material was a population of women aged 40-49 years who were enrolled in the Canadian National Breast Screening Study (NBSS). Mammograms from women who had undergone a biopsy (n = 441) and from a comparison group of women (n = 501) randomly selected from the mammography arm of the NBSS were classified according to the extent of mammographic density. The corresponding histological slides were independently classified by a review pathologist. Results: Compared with women showing no mammographic densities, women with the most extensive densities (i.e., occupying >75% of the breast volume) had a 9.7 times greater risk of developing carcinoma in situ or atypical hyperplasia (95% confidence interval [CI] = 1.75-53.97), a 12.2 times greater risk of developing hyperplasia without atypia (95% CI = 2.97-50.14), and a 3.1 times greater risk of developing nonproliferative disease (95% CI = 1.20-8.11). The gradients in risk were not monotonic across the five classifications of mammographic density. The associations could not be explained by the presence of mammographic densities at the biopsy site, but calcification at the biopsy site was strongly associated with high-risk histological changes (relative risk = 24; 95% CI = 5.0-156.0). Conclusions: These results suggest that the radiological patterns referred to as mammographic dysplasia may influence breast cancer risk by virtue of their association with high-risk histological changes in the breast epithelium. Implications: Identification of the factors responsible for high-risk histological changes may offer new insights into the etiology of breast cancer and potentially lead to the development of methods for its prevention. [J Natl Cancer Inst 84:1170-1179, 1992]

histological risk factors are well defined. Women diagnosed with carcinoma in situ, atypical hyperplasia, or hyperplasia without atypia have, respectively, an approximately 9, 4, or 1.5 times greater risk of developing breast cancer than the general population of women the same age. By contrast, there is no universally accepted method for classifying the mammographic risk factors. The most widely used classification scheme, introduced by Wolfe (5,6), has four categories: Nl, PI, P2, and DY. The lowest risk of breast cancer is associated with the Nl category, in which the breast tissue is made up mainly of fat, and the highest risk is associated with the DY category, in which the breast tissue shows extensive nodular or plaque-like tissue densities referred to as "mammographic dysplasia." Intermediate degrees of risk are associated with the PI and P2 categories, which are characterized by linear densities referred to as "ductal prominence." Quantitative estimates of the proportion of the breast volume occupied by these radiological densities, particularly the nodular or plaque-like densities (the DY pattern), can be used to estimate breast cancer risk. In previous work (7), we have found that quantitative estimation of these densities discriminated more strongly between cases and controls than did Wolfe's original, nonquantitative classification. All other casecontrol studies that have used both Wolfe's system and a quantitative classification of density have found higher estimates of risk with the quantitative classification (8-12), and in Wolfe's

described by Dupont and Page (7). We selected subjects for the present study according to the results of this classification. All subjects with invasive cancer were excluded. All subjects with a histological diagnosis of carcinoma in situ or atypical hyperplasia were selected for inclusion. From the much larger number of subjects diagnosed with hyperplasia without atypia or with nonproliferative disease, we sampled randomly to provide approximately 100 and 200 subjects, respectively, in each of these histological categories. Selection of comparison group. The comparison group was selected by random sampling from the entire NBSS population who had been assigned to the mammography arm and who were aged 40-49 years at entry. Individuals were selected for inclusion in this group regardless of whether they subsequently had a biopsy or were included in the biopsy group described above. The purpose of this comparison group was to estimate the prevalence of each classification of mammographic density in the starting population from which the biopsy subjects were selected. We could then compare these estimates with those seen in the biopsy group. Assembly of material. After selection of biopsy and comparison subjects, letters requesting mammograms and histological slides were sent to NBSS centers and to hospital pathology departments. For biopsy subjects, we requested the most recent mammogram taken before the date of biopsy. For comparison subjects, we requested the mammogram taken at the visit corresponding in time to the biopsy subjects after random selection. For biopsy subjects, we asked for one slide from each tissue block taken from the biopsy specimen, stained with hemMethods atoxylin and eosin. Of the 899 mammograms requested, we received 852 (95%). General Method There were 441 biopsy subjects whose slides had been classiThe source of material for this study was a population of fied by the NBSS pathologists and were requested from the 109 women enrolled in the Canadian National Breast Screening hospitals in which they had been taken. Five hospitals, Study (NBSS) (21), a randomized controlled trial of screening accounting for 47 biopsies (11% of the total), refused to with mammography. Participants in the trial were randomly provide slides, and a further 29 sets of slides could not be assigned either to a group who were screened annually with located. The remaining 104 hospitals provided 365 (83%) of mammography and physical examination or to a control group the 441 slides we requested for 463 subjects. Mammograms who were taught breast self-examination at the first visit but were unavailable for 18 of the 441 sets of slides classified by who received no follow-up radiological or physical examina- the NBSS pathologists, leaving 423 sets of slides for analysis. tions. In the present study, we restricted our attention to Mammograms were unavailable for 12 of the 365 biopsy specimens available for review, leaving 353 sets of slides for incluwomen aged 40-49 years at entry to the NBSS. We assembled mammograms and histological slides from sion in the analysis. Upon receipt of the mammograms and subjects who had been assigned to the mammography arm of slides, we assigned them a study number and distributed them the NBSS and in whom a breast biopsy had been performed to the study pathologist and radiologist, who classified them (biopsy subjects). We also obtained mammograms from a using the methods described below. group of comparison subjects; these women were randomly selected members of the mammography arm of the NBSS who Measurements were similar in age to the biopsy subjects. Mammograms from Radiology. All mammograms were classified by G. Cooke, both groups and histological slides from the biopsy subjects without knowledge of the histological findings, into one of six were then independently classified using the methods described categories according to the percentage (estimated by visual below. inspection) of the breast volume occupied by ductal prominence or mammographic dysplasia. Linear shadows radiating Sampling Methods from the nipple were classified as ductal prominence. Densities Selection of biopsy subjects. Subjects were eligible for that were nodular or plaque-like, or otherwise nonlinear, were inclusion if they had been assigned to the mammography arm classified as mammographic dysplasia and are the subject of of the NBSS and were aged 40-49 years at entry. Biopsy sub- the analysis in this article. If both types of density were present jects were identified from the computerized lists of the NBSS. in the same mammogram, the extent of each type was recorded Histological slides from approximately 2000 biopsy specimens separately. Sample mammograms illustrating the six categories in this group of subjects had been classified by the NBSS refer- of density used in the present analysis are shown in Fig. 1. ence pathologists using a nomenclature similar to that When localization procedures had been performed at the time Vol. 84, No. 15, August 5, 1992

ARTICLES 1171

Downloaded from http://jnci.oxfordjournals.org/ at Russian Archive on December 9, 2013

most recent publications (11,12), quantitative methods of classification have been used. The mechanism(s) by which these radiological changes influence breast cancer risk are not yet clear, and it has not been Firmly established whether a consistent association exists between the radiological and histological risk factors. Some studies have provided evidence for such an association {13-17), but others have not (18-20). In a previous study of women less than 50 years old (16), we found that biopsy specimens from subjects with mammographic dysplasia were significantly more likely to show atypical hyperplasia than specimens from subjects without these radiological changes. No association was found between histological changes and linear radiological signs (ductal prominence). This study population, however, had only a small number of subjects with atypical hyperplasia, a histological change strongly associated with cancer risk, and there was no information about the radiological features present specifically at the biopsy site. In the present study, we report on a large cohort of women taking part in a randomized trial of screening with mammography. Our objective was to estimate the relative risk of detecting carcinoma in situ, atypical hyperplasia, hyperplasia without atypia, and nonproliferative disease in subjects with different extents of mammographic density. We also examined the association between the histological findings on biopsy and the radiological features present at the biopsy site.

Downloaded from http://jnci.oxfordjournals.org/ at Russian Archive on December 9, 2013

Fig. 1. Six categories of density: A) density = to 0% to

Relationship between mammographic and histological risk factors for breast cancer.

Information on breast cancer risk can be obtained both from the histological appearance of the breast epithelium in biopsy specimens and from the patt...
1MB Sizes 0 Downloads 0 Views