Predictive Model for Breast Cancer in Women With Atypical Hyperplasia


Key Points

  • The risk model included age at biopsy, age at biopsy squared, and number of foci of atypical hyperplasia.
  • The model showed good discrimination (C-statistic of 0.59–0.63) and calibration (0.87–0.91) at 10 years.

As reported in the Journal of Clinical Oncology, Degnim et al have developed a model for predicting breast cancer risk among women with atypical hyperplasia on breast biopsy.

Study Details

The risk model (AH-BC) was developed using retrospective cohorts of women aged 18 to 85 years with pathologically confirmed benign atypical hyperplasia. The model-building cohort (from Mayo Clinic, Rochester) included 699 women, of whom 142 developed breast cancer during a median follow-up of 8.1 years. The validation cohort (Nashville Breast Cohort) included 461 women, of whom 114 developed breast cancer during a median follow-up of 11.4 years.

Clinical risk factors and histologic features of tissue biopsy were selected for model building using Cox proportional hazards regression. Identified features were included in a regression model to estimate breast cancer risk, with death as a competing risk.

Model Performance

The final AH-BC model included three covariates: age at biopsy (by age groups), age at biopsy squared (age group risk score squared), and number of atypical hyperplasia foci (1, 2, and ≥ 3). Younger (aged < 40 years) and older women (aged ≥ 75 years) and women with one atypical hyperplasia focus had lower risk than those diagnosed with atypical hyperplasia at age 45 to 70 years and those with ≥ 2 atypical foci.

Combinations of these variables defined 5-year risk of breast cancer as lower (1% to 3%), intermediate (4% to 7%), and higher (8% to 14%). For the Mayo and Nashville cohorts, 15.5% and 12.8% of patients were at lower risk, 53.1% and 48.6% at intermediate risk, and 31.5% and 38.6% at higher risk.

In the Mayo cohort, the AH-BC model had C-statistics of 0.650 at 10 years and 0.636 at 20 years and ratios of predicted-to-observed cancers of 0.96, 0.87, and 0.92 at 5, 10, and 20 years. In the Nashville cohort, the model had C-statistics of 0.591 at 10 years and 0.606 at 20 years and ratios of predicted-to-observed cancers of 1.46, 0.91, and 1.02 at 5, 10, and 20 years.

The investigators concluded, “We have created a new model with which to refine [breast cancer] risk prediction for women with [atypical hyperplasia]. The AH-BC model demonstrates good discrimination and calibration, and it validates in an external data set.”

The study was supported by grants from the National Cancer Institute and Susan B. Komen.

Amy C. Degnim, MD, of Mayo Clinic, Rochester, is the corresponding author for the Journal of Clinical Oncology article.

The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.