1172 EDITORIALS Journal of the National Cancer Institute, Vol. 98, No. 17, September 6, 2006

This issue of the Journal contains two articles refi ning and advancing the breast cancer risk assessment model known as the Gail Model ( 1 ). This model has been shown to be well calibrated with respect to predicting the number of cancers likely to develop within cohorts of white American women with specifi c risk fac-tors ( 2 – 4 ) . As needs of the chemoprevention trials (for determin-ing eligibility) and clinicians (in counseling patients) have grown, the Gail Model has been modifi ed to account for history of atypia and race or ethnicity, but until now it has included only non-modifi able risk factors (i.e., age, reproductive history, and biopsy history). It is readily available to practitioners as a user-friendly program ( http://bcra.nci.nih.gov/brc/start.htm ).

Chen et al. ( 5 ) modifi ed the Gail Model to estimate the abso-lute risk of invasive breast cancer on the basis of breast density, weight, age at birth of fi rst live child, and the number of fi rst- degree relatives with breast cancer. The Breast Cancer Detection and Demonstration Project (BCDDP) served as the dataset for the model of Chen et al. and the original Gail Model ( 1 ) . For the current model, women who developed breast cancer, had a biopsy, or were recommended for biopsy and a subset of women remaining normal during the screening phase were followed through December 31, 1995. Breast density measurements for 7200 women were obtained from baseline mammograms from the screening phase of the BCDDP (1973 – 1979).

The importance of breast density as an indicator of breast cancer risk is emphasized in the more ethnically diverse multi-institutional study by Barlow et al. ( 6 ) . They used prospective data from the Breast Cancer Surveillance Consortium (BCSC) to evaluate risk in a cohort of more than 1 million women undergoing screening mammography. They constructed separate risk models for pre- and postmenopausal women. Risk factors for premenopausal women were age, breast density, family history of breast cancer, and a prior breast procedure. Risk factors for postmenopausal women were age, breast density, race or ethnicity, family history of breast cancer, a prior breast procedure, body mass index, natural meno-pause, hormone therapy, and a prior false-positive mammogram. This model is more complex than the conventional Gail Model, but key fi ndings are related to the inclusion of risk factors that are ame-nable to modifi cation (body mass index, hormone therapy, and possibly breast density) and reinforcement of the message of Chen et al. ( 5 ) that breast density is a strong predictor of breast cancer risk. The rich BCSC dataset would serve as an excellent source of validation for the revised model described by Chen et al. ( 5 ).

Suggested modifi cations of the Gail Model are in no way a criticism of its aims or rejection of its approach. Rather, they demonstrate ongoing collaborative efforts of oncologists, epide-miologists, and biostatisticians to introduce chemoprevention into clinical practice as a viable and safe option for high-risk women. Gail et al. embarked upon their biostatistical journey with great care, thoughtfulness, and respect for the women who did or did not have representation in the BCDDP study, and for these reasons Gail et al. have been quite specifi c about the pri-mary applicability of their model to white women undergoing screening mammography. Studies of the Gail Model’s accuracy in multiethnic populations thus far have been preliminary and

Affi liations of authors: Department of Epidemiology, The University of Texas M.D. Anderson Cancer Center, Houston, TX (MLB); Breast Care Center, De-partment of Surgery, Division of Surgical Oncology, University of Michigan Comprehensive Cancer Center, Ann Arbor, MI (LAN).

Correspondence to: Melissa L. Bondy, PhD, Department of Epidemiology, The University of Texas M.D. Anderson Cancer Center, PO Box 301439, Hous-ton, TX 77230-1439 (e-mail: mbondy@mdanderson.org ) or Lisa A. Newman, MD, MPH, Breast Care Center, Department of Surgery, Division of Surgical Oncology, University of Michigan Comprehensive Cancer Center, Ann Arbor, MI (e-mail: lanewman@med.umich.edu ).

DOI: 10.1093/jnci/djj365 © The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org.

Assessing Breast Cancer Risk: Evolution of the Gail Model Melissa L. Bondy , Lisa A. Newman

speculative in nature ( 7 – 9 ) . Current efforts to incorporate breast density into the Gail Model offer tripartite benefi ts: 1) breast den-sity appears to represent a truly individualized risk factor that in-creases the discriminatory accuracy of prediction models to more than 0.6 ( 5 , 6 ) , 2) breast density may be more uniformly predictive of breast cancer risk in multiethnic populations ( 10 , 11 ) , and 3) breast density may also represent a modifi able risk factor that can monitor responsiveness to lifestyle and/or medical prevention strategies. Interestingly, in both studies, breast density was more predictive of risk than family history of breast cancer, despite the fact that two different measures of breast density were used.

In its present form, the calibration strength of the Gail Model substantiates the fulfi llment of its goal to identify cohorts of high-risk patients who are appropriate participants for chemopreven-tion trials. However, the discriminatory accuracy of the Gail Model at 0.5 – 0.6 suggests that it is only slightly better than chance in generating a higher risk estimate for a woman with versus without a diagnosis of breast cancer ( 12 ) . Furthermore, the medi-cal community has defaulted to generalizing the Gail Model, the best available risk prediction model for white women, to women of all racial or ethnic backgrounds. Although the model is quite valuable for defi ning chemoprevention trial eligibilty, it is some-what less able to assist in counseling the individual patient who visits a clinician’s offi ce and desires advice regarding whether she should face the morbidity of a chemoprevention medication. Risk assessment that uses an anatomic-based or tissue-based biomarker would be a more rational tool, but the ability to retrieve evaluable tissue from the general population is an obvious obstacle. Assess-ment of an individual woman’s breast density may be the next-best option, as shown by Chen et al. and by Barlow et al.

Despite the potential for breast density to improve the pre-cision of the Gail Model, it is incumbent upon the medical community to consider the feasibility of its widespread use by looking at contemporary mammography-reporting practices. Mammography facilities in the United States are accredited through the Food and Drug Administration through their adher-ence to the Mammography Quality Standards Act (MQSA) of 1994. MQSA inspections monitor mammogram reports to ensure presentation of clear, fi nal impressions as “ negative, ” “ benign, ” “ probably benign, ” “ suspicious, ” or “ highly suggestive of malig-nancy. ” MQSA descriptors correlate with the numbered catego-ries as defi ned by the American College of Radiology (ACR)

Journal of the National Cancer Institute, Vol. 98, No. 17, September 6, 2006 EDITORIALS 1173

in the Breast Imaging Reporting and Data System (BI-RADS). The ACR has also developed the following set of fi broglandular density descriptors that may be used within the text of a mam-mogram report: “ almost entirely fat ” (<25% density), “ scattered fi broglandular densities ” (25% – 50%), “ heterogeneously dense ” (51% – 75%), and “ extremely dense ” (>75%). The prospective clinical trial of digital versus fi lm-screen mammograms ( 13 ) found that breast density is a useful indicator of women that ben-efi t from digital mammography, and this result has motivated many mammographers to routinely incorporate the BI-RADS density descriptors into their reports. However, it must be noted that mammography facilities currently have no MQSA mandate to include breast density into their reporting mechanism. Further-more, one meta-analysis of breast density and breast cancer risk ( 10 ) found that the correlation is strongest when density is re-ported as percentage density as opposed to qualitative categories, such as the BI-RADS descriptors. Barlow et al. used the ACR descriptors to stratify breast density; however, Chen et al. used a fi ve-category stratifi cation of percentage density measurements. Standardization of breast density reporting must be achieved before breast density can be adapted into a clinically useful risk assessment model.

Inclusion of breast density, and perhaps other modifi able risk factors, is indeed exciting in the ongoing evolution of breast cancer prediction tools and our quest for accurate, individualized estimates. However, we must continue to exercise caution with these adjustments; a perfectly predictive model will be of mini-mal value if its component factors are unavailable, misunder-stood, or inappropriately assigned.

REFERENCES

(1) Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989 ; 81 : 1879 – 86.

(2) Bondy ML, Lustbader ED, Halabi S, Ross E, Vogel VG. Validation of a breast cancer risk assessment model in women with a positive family history. J Natl Cancer Inst 1994 ; 86 : 620 – 5.

(3) Spiegelman D, Colditz GA, Hunter D, Hertzmark E. Validation of the Gail et al. model for predicting individual breast cancer risk. J Natl Cancer Inst 1994 ; 86 : 600 – 7.

(4) Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM, et al. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst 1998 ; 90 : 1371 – 88.

(5) Chen J, Pee D, Ayyagari R, Graubard B, Schairer C, Byrne C, et al. Projecting absolute invasive breast cancer risk in white women with a model that includes mammographic density. J Natl Cancer Inst 2006 ; 98 : 1215 – 26.

(6) Barlow WE, White E, Ballard-Barbash R, Vacek PM, Titus-Ernstoff L, Carney PA, et al. Prospective breast cancer risk prediction model for women undergoing screening mammography. J Natl Cancer Inst 2006 ; 98 : 1204 – 14.

(7) Bondy ML, Newman LA. Breast cancer risk assessment models: applicability to African-American women. Cancer 2003 ; 97 : 230 – 5.

(8) Newman L, Rockhill B, Bondy M, Bernstein L. Validation of the Gail et al. breast cancer risk assessmnet model in African American women based on a multi-center case-control study of 3,283 African American and 5,974 white American women [abstract]. Proc Annu Meet Am Soc Clin Oncol 2002 ; 976 .

(9) Newman LA, Gail MH, Selvan M, Bondy M, Rockhill B, Colditz GA, et al. Proposed revision of the Gail breast cancer risk assessment model for African American women. American Society of Clinical Oncology, 39th Annual Meeting. Vol 22. Chicago (IL): American Society of Clinical Oncology; 2003 .

(10) McCormack VA, dos Santos Silva I. Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis. Cancer Epidemiol Biomarkers Prev 2006 ; 15 : 1159 – 69.

(11) Ursin G, Ma H, Wu AH, Bernstein L, Salane M, Parisky YR, et al. Mammographic density and breast cancer in three ethnic groups. Cancer Epidemiol Biomarkers Prev 2003 ; 12 : 332 – 8.

(12) Rockhill B, Spiegelman D, Byrne C, Hunter DJ, Colditz GA. Validation of the Gail et al. model of breast cancer risk prediction and implications for chemoprevention. J Natl Cancer Inst 2001 ; 93 : 358 – 66.

(13) Pisano ED, Gatsonis C, Hendrick E, Yaffe M, Baum JK, Acharyya S, et al. Diagnostic performance of digital versus fi lm mammography for breast-cancer screening. N Engl J Med 2005 ; 353 : 1773 – 83.