SAN ANTONIO — The widely used Gail model proved no better than a coin toss in predicting breast cancer risk in individual women with atypical hyperplasia on a benign breast biopsy, a case-control study has shown.
“The Gail model couldn't separate who was going to go on to breast cancer. We found that the accuracy of the model for individual women with atypia was essentially a flip of the coin, with a concordance statistic of 0.50 for predicted versus observed outcomes,” Dr. Lynn C. Hartmann reported at the San Antonio Breast Cancer Symposium.
The findings of this case-control study underscore the need for physicians to exercise great caution when using the Gail model to counsel individual women with atypia. The Gail model was designed to assess risk in populations of women, yet it's increasingly being applied in an effort to identify the risk in individuals. And as this study shows, when those women already have atypia on a benign breast biopsy, the Gail model greatly underestimates their breast cancer risk, according to Dr. Hartmann, a professor of oncology at the Mayo Clinic, Rochester, Minn.
Among 9,376 women prospectively followed in the ongoing Mayo Benign Breast Disease Cohort, 3.5% had atypia. During a mean 13.7 years of follow-up after biopsy, 58 of these 331 women (17.5%) were diagnosed with invasive breast cancer. Yet application of the Gail model predicted there would be only 34.9 breast cancers during the same period. In other words, 66% more breast cancers occurred in women with atypia than predicted by the Gail model.
“I think the take-home message here is if you have access to the tissue and you see the phenotype of atypia, the tissue has already integrated the risks featured in the Gail model, both the endogenous risks like family history and the exogenous exposures. Adding them back in is not going to help you further,” the oncologist said.
Dr. Hartmann said the Gail model and most other risk models work best in the setting of hereditary breast cancer. Better screening tools are needed for the broad population of 150 million American women over age 40 who should be getting screened for breast cancer; physicians would then be able to identify within that vast pool of healthy women the 1-2 million who are truly at high risk.
Dr. Hartmann and her Mayo colleagues are developing a tissue-based risk stratification system applicable to the 1 million women per year who undergo breast biopsy showing benign disease.
“If you look at where we do best in predicting cancer risk, it's where we can actually examine the tissue at risk: cervix, colon, esophagus, bladder. We're trying to apply this principle to breast cancer,” Dr. Hartmann explained.
Atypia and other standard histologic findings are incorporated in the evolving Mayo model because they have demonstrated utility in stratifying level of risk. Molecular markers are also of value in this regard, she said. And recent studies conducted in the Mayo Benign Breast Disease Cohort have identified a novel indicator of breast cancer risk: the extent of lobular involution in a biopsy specimen.
The Mayo Clinic's prospective studies of benign breast disease are funded by the Department of Defense.
'The Gail model couldn't separate who was going to go on to breast cancer.' DR. HARTMANN