News

ASCO: Model predicts risk for breast cancer from atypical hyperplasia


 

AT THE ASCO BREAST CANCER SYMPOSIUM

References

SAN FRANCISCO – A woman’s age at biopsy and the number of atypical hyperplasia foci appear to be good predictors of risk for subsequent breast cancer, investigators say.

A review of pathology records and medical history on more than 13,000 women with benign breast disease showed that a predictive model including age and atypia effectively identified those women with atypical hyperplasia at highest risk for developing breast cancer, reported Dr. Amy C. Degnim from the Mayo Clinic in Rochester, Minnesota.

Dr. Amy Degnim

Dr. Amy Degnim

“In our country, approximately 10% of all benign breast biopsies are showing atypical hyperplasia, and with about a million women who undergo biopsy every year, this about 100,000 women every year who are diagnosed with atypical hyperplasia. These women are known to have an increased risk of breast cancer,” Dr. Degnim said at a breast symposium here jointly sponsored by the American Society of Clinical Oncology, American Society For Radiation Oncology, and Society of Surgical Oncology.

Current risk prediction models such as the Breast Cancer Risk Assessment Tool and IBIS Breast Cancer Risk Evaluation Tool tend to underestimate or overestimate risk of breast cancer in women with atypical hyperplasia, prompting the investigators to explore developing a reliable prediction tool, she said

The investigators identified a cohort of 13,538 women diagnosed with benign breast disease at the Mayo Clinic from 1967 through 2001, and they found data on 699 with atypical hyperplasia confirmed by pathology review blinded to outcomes. They collected data on clinical and histologic features and identified breast cancer event through review of medical records and questionnaires.

In addition to the Mayo patients used to develop the model, the authors tested it in a validation sample of 461 women with atypical hyperplasia treated at Vanderibilt University in Nashville, Tennessee.

They rounded up data on potential contributors to a risk-prediction model using Lasso-identified variables that they then plugged into a Cox regression model.

They found that of all possible co-variates, including body-mass index at biopsy, age at menarche, indication for biopsy, number of live births, breastfeeding, family features, and histologic features such as involution or calcifications, only the age at biopsy and number of foci of atypical hyperplasia remained as robust predictors for breast cancer risk.

They then tested the model on data from the 699 women in the development set, who had had a total of 142 breast cancer events over a median follow-up of 8.1 years, and in the external validation set of 461 women who had a total of 114 breast cancer events over a median follow-up of 11.4 years.

Dr. A. Marilyn Leitch

Dr. A. Marilyn Leitch

They found that the concordance between the prediction model and the actual outcomes at 5, 10, and 30 years was 0.607. 0.633, and 0.607, respectively, for the development set.

The model performed a little less well for the validation set, with 5-, 10-, and 30-year concordance of 0.557, 0.584, and 0.557.

Based on their findings, they developed a risk-prediction table showing relative risk for women by age and number of foci (1, 2, or 3 or more). For example, the table shows a 5-year absolute risk of 9.69% for a woman 70-74 years in age at the time of biopsy with 3 or more foci of atypia, compared with just 4.5% for a woman of the same age with only 1 focus. There are risk prediction tables for 5-, 10-, and 30-year absolute risk.

In her commentary on the study, Dr. A. Marilyn Leitch from the University of Texas Southwestern Medical Center in Dallas, noted that the risk-prediction model relies on specific detail in the pathology report for the description of the number of foci of atypical hyperplasia.

“This model can provide estimates that are more informative to the patient than ‘4.5 times risk of breast cancer.’ However, while a high score might motivate a patient for intervention, most patients in the less than 20% risk category, and so we may not have as much persuasion from this model,” she said.

Next Article: