Using Multiple Biomarkers Improves Risk Prediction for Breast Cancer

The addition of three biological markers to conventional breast cancer risk models significantly improved the ability to distinguish women at high and low risk.

Xuehong Zhang, MD, ScD

The addition of three biological markers to conventional breast cancer risk models significantly improved the ability to distinguish women at high and low risk, according to a report presented at the AACR 2016 Annual Conference.

Risk discrimination (high versus low) improved by 10% to 20% with the addition of a genetic risk score, breast tissue density, and postmenopausal hormone levels, as compared with either of two widely used risk models for breast cancer. Risk prediction improved to a greater degree for estrogen receptor (ER)-positive breast cancer.

“While each of the genetic markers, percent mammographic density, and hormones independently improved risk prediction for invasive breast cancer and hormone receptor-positive disease, incorporating all of these factors simultaneously improved models the most, especially among postmenopausal women not using hormone therapy,” said Xuehong Zhang, MD, ScD, an assistant professor of medicine at Harvard Medical School and Associate Epidemiologist Brigham and Women’s Hospital in Boston.

“If validated in independent populations, our findings could help identify women at a higher risk, who would most benefit from chemoprevention, other risk-reducing regimens, or screening,” he said.

Conventional risk-prediction models for breast cancer have generally included only traditional risk factors for breast cancer, such as age, family history, reproductive factors, body mass index, and alcohol intake. The National Cancer Institute’s Gail model and the Rosner-Colditz typify traditional risk-prediction models and have been widely used for years.

The extent to which biological markers of breast cancer risk might improve risk prediction has remained unclear. To address the issue, Zhang and colleagues performed analyses involving three recognized biological markers, each shown to be associated with breast cancer risk in multiple studies.

The genetic risk score is based on 67 breast cancer-associated single nucleotide-polymorphisms identified in a recent meta-analysis of nine genome-wide association studies, mammographic density (defined as percent dense area on a mammogram, divided by total breast area), and levels of the endogenous testosterone, estrange sulfate, and prolactin.

Investigators used the markers, as well as the Gail and Rosner-Colditz models, to evaluate breast cancer risk among participants in the Nurses’ Health Study I and II. The two studies combined included 10,052 women who developed invasive breast cancer and 12,575 women who did not. The two cohorts had about 20 years of follow-up, and stored pre-diagnostic blood samples provided a means to assess hormone levels in the women.

Zhang and colleagues used age-adjusted area under the curve (AUC) as the measure of discrimination for 5-yeare risk of invasive breast cancer. They performed a subgroup analysis limited to risk prediction for ER-positive breast cancer.

The biological markers demonstrated “excellent” discrimination between high and low risk, said Zhang. Both the genetic risk score and mammographic density were associated with about a 2.5-fold increased risk from the lowest- to highest-risk quartiles (P <.01). The range for endogenous hormone levels was 1.5 to 2-fold for each of the three hormones measured (P <.01).

Risk assessment by means of the Gail criteria resulted in an AUC of 55.2, and the Rosner-Colditz model yielded an AUC of 60.2.

The individual biological markers yielded AUC values of 58.2 to 62.1. When all three markers were combined with the conventional risk-assessment tools, the AUC increased to 66, representing an absolute increase of 10.8 versus the Gail model (P <.001) and 6.0 versus the Rosner-Colditz model (P <.001).

Investigators repeated the analyses for prediction of estrogen receptor-positive breast cancer. The Gail and Rosner-Colditz models were associated with AUC values of 55.5 and 60.5, respectively. When the three biomarkers were added to the conventional models, the AUC value increased to 67.2 with the Gail model and 69.9 with the Rosner-Colditz model (P <.001 for both comparisons).

Zhang said the results require validation in other populations. Additionally, the NHS analysis was limited to women of European ancestry, so risk estimates will be revised as need for non-Caucasian populations.

Zang X, Rice M, Tworoger SS, et al. Zhang, Xuehong, Breast Cancer Risk Prediction Models Improved by Adding Multiple Biological Markers of Risk. Presented at: AACR 2016 Annual Meeting; New Orleans, Louisiana, April 16-20, 2016. Abstract 2600.


View more from the 2016 AACR Annual Meeting

Related Videos
View All
Related Content