High-risk mammographic parenchymal patterns, hormone replacement therapy and other risk factors: a case-control study

Evis Salaa, Ruth Warrenb, Jenny McCanna, Stephen Duffyc, Robert Lubena and Nicholas Daya

a Department of Community Medicine, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 8RN, UK.
b Cambridge and Huntingdon Breast Screening Service, Rosie Maternity Hospital, Robinson Way, Cambridge CB2 2SW, UK.
c MRC-Biostatistics Unit, Institute of Public Health, Robinson Way, Cambridge CB2 2SR, UK.

Reprint requests to: Dr E Sala, Department of Community Medicine, University of Cambridge, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 8RN, UK. E-mail: evis.sala{at}srl.cam.ac.uk


    Abstract
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
Background Mammographic parenchymal patterns are of particular interest because the denser patterns reduce screening sensitivity as well as increasing breast cancer risk, and because they have been shown to be affected by exogenous oestrogens.

Methods We designed a case-control study comprising 200 cases with high-risk (P2 and DY) pattern and 200 controls with low-risk (N1 and P1) pattern. Mammograms were evaluated according to the Wolfe classification.

Results Parity, body mass index (BMI) and current smoking were inversely and independently associated, whereas late age of menarche and history of benign breast disease were positively associated with high-risk mammographic patterns. Current-users of hormone replacement therapy (HRT) were more than twice as likely to have a high-risk pattern than never-users (OR = 2.48, 95% CI : 1.32–4.61). Women who used HRT for more than 5 years were almost three times more likely to have a high-risk pattern than never-users (OR = 2.77, 95% CI : 1.11–6.91). Relative to never-users, women who started HRT before the menopause were more than twice as likely to have a high-risk pattern (OR = 2.53, 95% CI : 1.31–4.87).

Conclusions Careful clinical and mammographical follow-up might be appropriate in women undergoing HRT. The HRT-induced mammographic pattern might be regarded as a new baseline and changes with respect to this new pattern could then be monitored over time.

Keywords Breast parenchymal patterns, mammography, hormone replacement therapy, menopausal status

Accepted 28 January 2000


    Introduction
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
Hormone replacement therapy (HRT) is increasingly prescribed to postmenopausal women for relief of menopausal symptoms as well as for the prevention of heart disease and osteoporosis. However, the general enthusiasm for its beneficial effects has declined because of the possibility of such therapy increasing the risk of development of breast cancer. Epidemiological studies have produced conflicting evidence on this issue.1–7 A recent collaborative re-analysis of data from 51 epidemiological studies has shown that the risk of breast cancer increases with long-term HRT use.8

Mammographic parenchymal patterns, a function of the density and structure of the breast tissue as imaged on a mammogram, have been positively associated with breast cancer risk.9–13 These patterns were classified by Wolfe into four different categories: N1, P1, P2 and DY.9 The high-risk patterns P2 and DY are characterized by greater mammographic density (which show more white areas on the X-ray film). They are associated with increased breast cancer risk, suggesting that the parenchymal patterns, which can be viewed radiographically, may reflect premalignant or at least high-risk pathological changes in breast tissue. Mammographic parenchymal patterns of the breast reflect the development of ducts and lobules which itself depends on endogenous endocrine factors. Oestrogens are responsible for the growth and development of ducts, while progestogens stimulate the growth and development of lobules. During the menopause, the levels of endogenous hormones decrease leading to regression of the ductal and stromal elements and to their fatty replacement. The use of HRT may reverse this process resulting in an increase in mammographic density. The number of studies in this area is limited and their results remain contradictory.14–23

Mammographic parenchymal patterns are of particular interest in this epoch of routine mammographic screening and widespread use of HRT, firstly because the denser patterns reduce screening sensitivity as well as increasing risk,24–26 and secondly because they have been shown to be affected by ingestion of exogenous oestrogens.16,19–23 Thus mammographic patterns may have relevance both to fine-tuning of the screening programme and to breast cancer aetiology and possibly even primary prevention.

Mammographic parenchymal patterns may simply reflect the underlying endocrine mechanisms through which the use of HRT increases the risk of breast cancer. We believe that investigating the association between mammographic parenchymal patterns and HRT will improve our understanding of the aetiology of breast cancer and especially of the complex relationship between female sex hormones and breast cancer.

In this paper we report a case-control study nested within the European Prospective Investigation on Cancer cohort in Norfolk (EPIC-Norfolk).27 Our study assesses the effect of HRT use, duration and timing on mammographic parenchymal patterns assessed according to the Wolfe classification (Table 1Go).


View this table:
[in this window]
[in a new window]
 
Table 1 Wolfe classification of mammographic parenchymal patterns
 

    Material and Methods
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
Study population
Study subjects were members of a cohort of women born between February 1921 and December 1946 who are taking part in the EPIC-Norfolk cohort,27 who attended the prevalence screening round at the Norwich Breast Screening Programme between November 1989 and December 1997 and who were free of breast cancer prior to or at the time of their prevalent screen. A case-control study was designed within the above cohort.

The case-control study
A total of 9484 women were identified by linking databases from EPIC-Norfolk and the National Health Service (NHS) Regional Breast Screening Programme for Norwich, the latter being part of an ongoing evaluation of screening programme effectiveness.28 We aimed to recruit 200 cases with high-risk (P2 and DY) patterns and 200 controls with low-risk (N1 and P1) patterns, matched for age and date of screening. Of these 9484 women identified by linking the databases, 8001 had completed food diaries and 445 of these diaries had been already entered into the EPIC-Norfolk database. A women was excluded from the total study population if: (1) she was diagnosed with a histologically verified breast cancer prior to or at the prevalent screen; (2) she did not respond to the screening invitations; (c) after an extensive search, her screening mammogram or screening records were not located; (d) she had breast implants.

We excluded 45 women on the above criteria. Of these, 17 women were excluded because they developed breast cancer prior to or at the prevalent screen. We were unable to locate screening records for 13 women, 14 women did not respond to screening invitation, and one woman was excluded because she had breast implants which makes the reading of the mammographic pattern difficult.

Cases were defined as women from EPIC-Norfolk cohort with a P2/DY Wolfe's mammographic parenchymal pattern on the prevalence screen mammogram who had been diagnosed as normal at that screen. In order for a case to be eligible, a mammogram had to be classified as P2/DY for both sides and both views by the two readers. Similarly, a control mammogram had to be classified as N1/P1 on both sides and both views by the two readers in order to be eligible. There was inter-reader disagreement for 17 women so these were excluded as potential cases. This left 383 women who satisfied the study criteria: i.e. were classified as either NI/P1 or P2/DY patterns. From these, a total of 203 women with P2/DY mammographic patterns according to Wolfe's classification were identified as cases.

For each case, we wished to select one control with an N1/P1 Wolfe's mammographic parenchymal pattern at the prevalence screen mammogram who had been diagnosed as normal at that screen, matched to the case by date of birth (within one year) and date of prevalence screen (within 3 months). The readers disagreed for 13 women who were excluded as potential controls. A total of 167 women with N1/P1 Wolfe's mammographic patterns were identified as potential controls. Of these, only 141 could be individually matched for birth and prevalent screening date with the cases. The remaining 62 controls were identified among 8001 women with completed food diaries not yet entered on the database using the same criteria. We randomly selected 248 women that satisfied the inclusion criteria and were matched for date of birth and date of screening with the remaining 62 cases. We then read the mammograms to determine the parenchymal pattern and selected as controls the first 62 women with N1/P1 mammographic pattern on both sides and both views that were also individually matched to 62 remained cases. Their diaries were entered afterwards. As a result, 203 cases and 203 individually matched controls remained in the study.

Risk factor data
In addition to dietary data, considerable information was available from the EPIC Health and Lifestyle Questionnaire. With the exception of anthropometric measures such as weight, height, hip and waist that were measured by a trained EPIC nurse, the rest of the variables were obtained from the self-completed EPIC Health and Lifestyle Questionnaire. Details of this instrument are reported elsewhere.27 In this study we examine the subjects' personal and family history for benign breast disease and cancer; menstrual factors and menstrual history; reproductive history; oral contraception and hormone replacement therapy; physical activity; smoking; and anthropometric information such as measures of weight and height. The dietary results will be the subject of another paper.

Women were classified as postmenopausal if they answered ‘No’ to the question ‘Are you still menstruating?’ They were also classified as postmenopausal if they had hysterectomy and bilateral ovarectomy. Women were classified as premenopausal if they answered ‘Yes’ to the question ‘Are you still menstruating?’ and also if they had more than 10 menstrual cycles in the last 12 months. Women were classified as perimenopausal if they answered ‘Yes’ to the question ‘Are you still menstruating?’ and also if they had between 1 and 9 menstrual cycles in the last 12 months. The pre- and perimenopausal groups were combined for purposes of analyses and are referred to as premenopausal below, for the sake of brevity. If women stated that they were still menstruating but they did not report the number of periods in the last 12 months, they were still included in the premenopausal group. Of the women, 74% in the pre-perimenopausal category were premenopausal. In addition, 25 women with unknown menstrual status were included in the analysis of the total study population. Women with unknown age at menopause were excluded from the analysis with respect to age at menopause, but if they reported being postmenopausal they were included in the postmenopausal category.

Statistical methods
Statistical analysis was by conditional logistic regression, which takes into account the matching of controls to cases and produces odds ratio (OR) estimates of relative risk and their 95% CI.29 Descriptive tables complemented the results of these analyses.


    Results
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
Table 2Go shows the OR for the association between Wolfe's high-risk mammographic parenchymal patterns and menstrual and reproductive factors. Relative to premenopausal women, the OR of having a high-risk pattern in women with established menopause was 0.43. The protective effect of the menopause was lost after adjustment for the number of children, HRT use, history of benign breast disease, and body mass index (BMI). The probability of having a high-risk pattern in women with more than four children was one-quarter that in nulliparous women (OR = 0.25, 95% CI : 0.08–0.68). A significant trend of decreasing risk with increasing parity was observed (P = 0.007). Women whose age at menarche was >15 years were more than twice as likely to have a high-risk Wolfe pattern (OR = 2.24, 95% CI : 1.00–4.99) than those whose age at menarche was <11 years. The trend in risk with age at menarche was significant (P = 0.02).


View this table:
[in this window]
[in a new window]
 
Table 2 Odds ratio (OR) estimates for high-risk mammographic patterns according to menstrual and reproductive factors
 
The OR estimates of high-risk patterns according to BMI, oral contraceptive (OC) use, family history, history of benign breast disease, smoking and physical activity are displayed in Table 3Go. The BMI was inversely associated with high-risk patterns. Relative to the lower tertile, women in the upper tertile of the BMI had a significantly lower probability of having a high-risk pattern (OR = 0.30, 95% CI : 0.14–0.61). There was a significant trend across the three categories of BMI (P-value = 0.001). History of benign breast disease was significantly and positively related to a high-risk pattern in the univariate analysis (OR = 3.00, 95% CI : 1.09–8.25). This significance did not persist after adjustment for other variables. Relative to never smokers, current smokers were less than half as likely to have a high-risk pattern (OR = 0.40, 95% CI : 0.16–0.96).


View this table:
[in this window]
[in a new window]
 
Table 3 Odds ratio (OR) estimates for high-risk mammographic patterns according to other factors
 
Table 4Go shows the OR estimates of high-risk patterns according to different patterns of HRT use. Current-users of HRT were more than twice as likely to have a high-risk pattern compared to never-users (OR = 2.48, 95% CI : 1.32–4.61). The duration of HRT use was strongly and positively associated with high-risk pattern. Women who used HRT for longer than 5 years were almost three times more likely to have a high-risk pattern than never-users (OR = 2.77, 95% CI : 1.11–6.91). The trend across the four categories of duration of HRT use was significant (P-value = 0.007). Time since last use of HRT did not have any effect on mammographic parenchymal patterns.


View this table:
[in this window]
[in a new window]
 
Table 4 Odds ratio (OR) estimates for high-risk mammographic patterns according to hormone replacement therapy (HRT) use
 
When we performed the analysis in postmenopausal women separately, HRT user status and the duration of use were no longer significantly associated with high-risk patterns (Table 5Go). We extended our investigation to evaluating the association between high-risk mammographic patterns and time at which HRT was started with respect to the menopause (Tables 4 and 5GoGo). Relative to never-users, women who started taking HRT before they reached menopause were more than twice as likely to have a high-risk pattern (OR = 2.53, 95% CI : 1.31–4.87). Women who started taking HRT after the menopause did not appear to be different from never-users with respect to high-risk patterns. The above findings persisted when the analysis was limited to postmenopausal women (OR = 3.03, 95% CI : 0.93–9.78) but statistical significance was not reached, possibly due to the small numbers in each category.


View this table:
[in this window]
[in a new window]
 
Table 5 Odds ratio (OR) estimates for high-risk mammographic patterns according to hormone replacement therapy (HRT) use, in postmenopausal women only
 
Table 6Go shows the OR estimates for high-risk patterns according to timing of start of HRT use, with respect to menopause within different categories of HRT duration. Women who started HRT pre- or perimenopausally and used it for >5 years were almost ten times more likely to have a high-risk pattern than those who used HRT for >5 years but started it after the menopause (OR = 9.82, 95% CI : 1.84–52.39, P-value for trend = 0.008).


View this table:
[in this window]
[in a new window]
 
Table 6 Odds ratio (OR) estimates for high-risk mammographic patterns according to hormone replacement therapy (HRT) use starting time, within different HRT duration categories
 

    Discussion
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
Our study confirms that parity is inversely and independently associated with high-risk mammographic patterns, consistent with its effect on breast cancer risk. Several epidemiological studies have reported similar results.11,30–32 We also found that late age of menarche is positively associated with high-risk mammographic patterns in contrast to its association with a decreased risk of breast cancer.33 Only two other studies11,32 support our finding but these were limited to premenopausal women and the results may be due to chance. The association of high-risk pattern with a late age at menarche may also be a feature of confounding with nutritional status in early life. Relatively low nutritional status is likely to be associated with both late menarche and less fatty replacement in breast tissue, and hence a higher risk parenchymal pattern. In addition, there was in our study a non-significant but suggestive interaction between the effects of age at menarche and parity (P = 0.1). The OR associated with late (>=13 years) menarche in women with 0–2 children was 2.33 (95% CI : 1.18–4.59), whereas in women with >=3 children, the OR associated with late menarche was 0.78 (95% CI : 0.28–2.09). Thus, the association of late menarche with high-risk patterns is confined to women with <=2 children. The interaction of these two variables in terms of effects on breast cancer risk is not well-researched and should be addressed in future work.

Our study strongly suggests that obese women or those with normal BMI (25–30 kg/m2) are less likely to have a high-risk pattern compared to women with low BMI. This association persisted even after adjusting for other possible confounding factors. This finding is supported by other published studies34,35—in premenopausal women only—and it is intuitively reasonable, since one might expect more fatty replacement and hence less dense pattern in obese women. On the other hand, increased body weight and BMI has been consistently associated with elevated breast cancer risk especially in postmenopausal women.36 This suggests that studies of mammographic parenchymal pattern may underestimate breast cancer risk if body size is not taken into account.

A higher BMI means more adipose tissue generally and potentially more fatty replacement in the breast. The fact that Wolfe classification depends on percentages of the breast with dense parenchyma implies that, in this system, an association with breast size is inevitable. An interesting point is whether it is possible to determine a measure of parenchymal density that is independent of body habitus (anthropometric measures) and breast size.

In our study, current and extended use of HRT was associated with increased probability of having a high-risk pattern in the total study population and this is clearly due to currently pre- and perimenopausal women who are taking HRT. There is a limited number of studies in this area and the results remain inconclusive14,15,17,18 although the most recent studies support our findings.16,19–23 We failed to show a relationship between time since last use of HRT and mammographic parenchymal patterns.

Unique to this study is the demonstration of an association between mammographic parenchymal patterns and when HRT was started with respect to menopause. We found that women who started using HRT while still menstruating were more likely to have a high-risk pattern compared to women not exposed to HRT or to those who started HRT after the menopause. There was a significant interaction between duration and the time of starting use (P = 0.04). We also found that women who started HRT pre- or perimenopausally and used it for >5 years were almost ten times more likely to have a high-risk pattern than those who used HRT for >5 years but started it after the menopause. We are not aware of any other study which has investigated the effect of timing of start of HRT use on the relation between HRT use and Wolfe's mammographic parenchymal pattern, and it would be very interesting to see if other studies show the same phenomenon.

Establishing whether a woman is pre- or postmenopausal is always a difficult issue, especially when she is taking HRT that includes a progestogen. Some women could be reporting regular menstrual cycles because of HRT use but their true menopausal status may be unknown. One limitation of our study may be that we did not have enough information on type of HRT used. In addition, in women who are hysterectomized it is difficult to establish the true time of the menopause. In our study 39 (12%) out of 313 postmenopausal women reported the same age for menopause and simple hysterectomy. Since the timing of the decline in ovarian function obviously affects breast density and the indication for HRT use may be related to a particular endocrine characteristic that is associated with the onset of menopause, there may be a possibility of confounding on account of menopausal age that in turn might affect the validity of our findings.

We found that quite a high proportion of women (25% of the cases and 11% of the controls) started using HRT preor perimenopausally. Among these women 57% were still menstruating at the time they completed the questionnaire. Of these, 87% were premenopausal and 13% perimenopausal. Moreover, among those women using HRT who were still menstruating, 93% of the cases and 67% of controls started using HRT when they were premenopausal, while 7% of the cases and 33% of controls started using HRT when they were perimenopausal. We could not assess the true menopausal status of those women who started HRT premenopausally and were postmenopausal at the time they filled the questionnaire.

There is a possibility of recall bias in reporting health and lifestyle information including data on HRT use. This would be likely to cause non-differential misclassification, which might obscure the real association, since subjects were unaware of their own case-control status. Healthy participant bias might also have affected our study since our study population was composed of women participating in the EPIC study and it is believed that those who participate in health-related studies are more health conscious. In addition, we did not have enough statistical power for some subgroup analyses.

Our results suggest that the combination of starting to use HRT pre- or perimenopausally and extended duration (>5 years) of use increase the probability of having a high-risk pattern, although clearly it would be desirable to see these results replicated in other studies before drawing absolute conclusions. Oestrogens are known promoters of breast cancer.37 The long-term use of HRT, especially before menopause, exposes breast tissue to high levels of oestrogens, whereas the natural menopause decreases endogenous oestrogen level and thus should be protective against breast cancer. From our results it would appear that mammographic parenchymal patterns reflect the above process.

The association of long-term and pre- or perimenopausal use of HRT with an increase in high-risk patterns is consistent with the overview evidence on HRT and breast cancer risk. This is not an argument against HRT, but it suggests that research is required to address the following issues: (1) the effects of different preparations of HRT on mammographic parenchymal patterns; (2) whether long-term use of HRT before the menopause is particularly associated with breast cancer risk, and (3) whether the association between density and Grade 3 cancers and ductal carcinoma in situ (DCIS) holds good for HRT-induced density.

These findings have some practical considerations in care of women: (1) The North American practice of regular screening of women on HRT is supported, and the issue of UK advice against such additional screening should be revisited; (2) It can be expected that the sensitivity of mammography will be impaired by long HRT use, and so two views, two reports and a frequent screening schedule should be considered; (3) Consideration should be given to adding either clinical examination or routine ultrasound where HRT use has resulted in dense patterns.

In conclusion, the findings of our study indicate that careful clinical and mammographical follow-up might be appropriate in women undergoing HRT. The HRT-induced mammographic pattern could be regarded as a new baseline and changes with respect to this new pattern should be monitored over time.


    Acknowledgments
 
We thank Anglia and Oxford Health Authority, R & D Programme for funding this study. We are most grateful to Dr Graham Hurst, director of Norwich Breast Screening Unit. We also thank all the staff of Norwich Breast Screening Unit for their invaluable help during data collection. We also thank the staff of EPIC for their contribution to the study.


    References
 Top
 Abstract
 Introduction
 Material and Methods
 Results
 Discussion
 References
 
1 Brinton LA, Hoover RM, Szklo M, Fraumeni JF. Menopausal estrogen use and risk of breast cancer. Cancer 1981;47:2517–22.[ISI][Medline]

2 Ewertz M. Influence of non-contraceptive exogenous and endogenous sex hormones on breast cancer risk in Denmark. Int J Cancer 1988;42:832–38.[ISI][Medline]

3 Schairer C, Byrne C, Keyl PM, Brinton LA, Sturgeon SR, Hoover RN. Menopausal estrogen and estrogen-progestin replacement therapy and risk of breast cancer (United States). Cancer Causes Control 1994; 5:491–500.[ISI][Medline]

4 Colditz GA, Hankinson SE, Hunter DJ et al. The use of estrogens and progestins and the risk of breast cancer in postmenopausal women. N Engl J Med 1995;332:1589–93.[Abstract/Free Full Text]

5 Schuurman AG, van den Brandt PA, Goldbohm RA. Exogenous hormones and the risk of postmenopausal breast cancer: results from the Netherlands cohort study. Cancer Causes Control 1995;6:416–24.[ISI][Medline]

6 La Vechia C, Negri E, Franceschi S et al. Hormone replacement treatment and breast cancer risk: a cooperative Italian study. Br J Cancer 1995;72:244–48.[ISI][Medline]

7 Newcomb PA, Longnecker MP, Storer BE et al. Long-term hormone replacement therapy and risk of breast cancer in postmenopausal women. Am J Epidemiol 1995;142:788–95.[Abstract]

8 Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormone replacement therapy: collaborative reanalysis of data from 51 epidemiological studies of 52 705 women with breast cancer and 108 411 women without breast cancer. Lancet 1997;350: 1047–59.[ISI][Medline]

9 Wolfe JN. Breast patterns as an index of risk for developing breast cancer. Am J Roentgenol 1976;126:1130–39.[ISI]

10 Wolfe JN. Risk for breast cancer development determined by mammographic parenchymal patterns. Cancer 1976;37:2486–92.[ISI][Medline]

11 Saftlas AF, Szklo M. Mammographic parenchymal patterns and breast cancer risk. Epidemiol Rev 1987;9:146–74.[ISI][Medline]

12 Oza AM, Boyd NF. Mammographic parenchymal patterns: a marker of breast cancer risk. Epidemiol Rev 1993;15:196–208.[ISI][Medline]

13 Sala E, Warren R, McCann J, Duffy S, Day N, Luben R. Mammographic parenchymal patterns and mode of detection: implications for the breast screening programme. J Med Screen 1998;5:180–85.

14 Bland KI, Buchanan JB, Weisberg BF, Hagan TA, Gray LA. The effects of exogenous estrogen replacement therapy of the breast: breast cancer risk and mammographic parenchymal pattern. Cancer 1980; 45:3027–33.[ISI][Medline]

15 Buchanan JB, Weisburg BF, Sandoz JP et al. Selected prognostic variables for mammographic parenchymal patterns. Cancer 1981; 47:2135–37.[ISI][Medline]

16 Berkowitz JE, Gatewood OMB, Goldblum LE, Gayler BW. Hormonal replacement therapy: mammographic manifestations. Radiology 1990;174:199–201.[Abstract]

17 Kaufman Z, Garstin WI, Hayes R et al. The mammographic parenchymal patterns of women on hormonal replacement therapy. Clin Radiol 1991;43:389–92.[ISI][Medline]

18 Berkvist L, Tabar L, Adami H et al. Mammographic parenchymal pattern in women receiving non-contraceptive estrogen treatment. Am J Epidemiol 1989;130:503–10.[Abstract]

19 Stomper PC, Van Voorhis BJ, Ravnikar VA et al. Mammographic changes associated with postmenopausal hormone replacement therapy: a longitudinal study. Radiology 1990;174:487–90.[Abstract]

20 McNicholas MMJ, Heneghan JP, Milner MH et al. Pain and increased mammographic density in women receiving hormone replacement therapy: a prospective study. Am J Roentgenol 1994;163:311–15.[Abstract]

21 Laya MB, Gallagher JC, Schreiman JS, Larson EB, Watson P, Weinstein L. Effect of postmenopausal hormonal therapy on mammographic density and parenchymal pattern. Radiology 1995; 196:433–37.[Abstract]

22 Persson I, Thurfjell E, Holmberg L. Effect of estrogen and estrogen-progestin replacement regimens on mammographic breast parenchymal density. J Clin Oncol 1997;15:3201–07.[Abstract]

23 Porfiri LM, Constanza L, De Felice C, David V, Zichella L. A mammographic evaluation of the morphostructural variations of the breast during hormone-replacement therapy in the menopause. Radiol Med (Torino) 1998;95:573–76.[Medline]

24 Ma L, Fishell E, Wright B, Hanna W, Allan S, Boyd NF. Case-control study of factors associated with failure to detect breast cancer by mammography. J Natl Cancer Inst 1992;84:781–85.[Abstract]

25 Bird R, Wallace T, Yankaskas B. Analysis of cancers missed at screening mammography. Radiology 1992;184:613–17.[Abstract]

26 Laya MB, Larson EB, Taplin SH, White E. Effect of estrogen replacement therapy on the specificity and sensitivity of screening mammography. J Natl Cancer Inst 1996;88:643–49.[Abstract/Free Full Text]

27 Day N, Oakes S, Luben R et al. EPIC in Norfolk: study design and characteristics of the cohort. Br J Cancer 1999;80(S1):95–103.[ISI][Medline]

28 McCann J, Stockton D, Day N. Breast cancer in East Anglia: the impact of the breast screening programme on stage at diagnosis. J Med Screen 1998;5:42–48.[ISI][Medline]

29 Breslow NE, Day NE. Statistical Methods in Cancer Research. Vol 1. The Analysis of Case-Control Studies. Lyon, France: IARC Scientific Publications, 1980.

30 De Waard F, Rombach JJ, Collette HJA, Slatboom S. Breast cancer risk associated with reproductive factors and breast parenchymal patterns. J Natl Cancer Inst 1984;72:1277–82.[ISI][Medline]

31 Berkvist L, Bergstrom R, Tabar L, Adami H. Epidemiologic determinants of mammographic parenchymal pattern. A population based study within a mammographic screening programme. Am J Epidemiol 1987;126:1075–81.[Abstract]

32 Gram IT, Funkhouser E, Tabar L. Reproductive and menstrual factors in relation to mammographic parenchymal patterns among perimenopausal women. Br J Cancer 1995;71:647–50.[ISI][Medline]

33 Rockhill B, Moorman PG, Newman B. Age at menarche, time to regular cycling, and breast cancer (North Carolina, United States). Cancer Causes Control 1998;9:447–53.[ISI][Medline]

34 Brisson J, Morrison AS, Kopans DB et al. Height and weight, mammographic features of breast tissue, and breast cancer risk. Am J Epidemiol 1984;119:371–81.[Abstract]

35 Boyd NF, Lockwood GA, Byng JW, Little LE, Yaffe MJ, Tritchler DL. The relationship between anthropometric measures to radiological features of the breast in premenopausal women. Br J Cancer 1998; 78:1233–38.[ISI][Medline]

36 Van den Brandt P, Dirx MJM, Ronckers CM, van den Hoogen P, Goldbohm AR. Height, weight, weight change, and post-menopausal breast cancer risk: the Netherlands Cohort Study. Cancer Causes Control 1997;8:39–47.[ISI][Medline]

37 Colditz GA. Relationship between estrogen levels, use of hormone replacement therapy, and breast cancer. J Natl Cancer Inst 1998;90: 814–23.[Abstract/Free Full Text]