New Polygenic Risk Score for Breast Cancer Accurate for All Ancestries

A new polygenic risk score (PRS) offers improved breast cancer risk estimates for women of all ancestries in the United States, according to data presented at the virtual 2021 American Society of Clinical Oncology meeting

“Combined with clinical and biological risk factors, this approach may offer improved risk stratification for all women, regardless of ancestry,” says Holly Pederson, MD, Director of Medical Breast Services at Cleveland Clinic.

Teaming up with Myriad Genetics, researchers from Cleveland Clinic and other institutions collaborated to study a PRS comprising 149 single-nucleotide polymorphisms (SNPs), 93 of them well-established for breast cancer and 56 that are ancestry specific. The composite PRS was validated and calibrated for women in the United States of all ancestries and demonstrates significant and meaningful discrimination for subcohorts defined by self-reported ancestry.

Validating PRS for women of non-European ancestry

Currently available PRS scores are powerful breast cancer risk prediction tools for women of European ancestry, with high-risk SNPs identified from large-scale genome-wide association studies (GWAS). However, their use has not been widely adopted in clinical practice in large part because they have not been validated for women of non-European ancestry.

“Clinically, the PRS is in its infancy. There are companies that have commercialized PRSs, but national guidelines have cautioned against widespread use because of the lack of transportability for women of non-European descent,” Dr. Pederson explains.

In September 2020, the revised National Comprehensive Cancer Network clinical practice guidelines advised against the use of current PRS outside the context of a clinical trial, because of “significant limitations in interpretation.”

Both SNP frequencies and linkage between SNPs differ by ancestry. Current PRSs are only available for women of European ancestry, because they would overestimate breast cancer risk in women of non-European ancestries and are less effective at distinguishing between high-risk and low-risk women.

Genetic ancestral data from three major continents

Principle component analysis and clustering analysis show that the three major source ancestries in the United States are European, African and East Asian.

A total of 56 ancestral SNPs were weighted to quantify the contribution of those three ancestries in a given subject. The 93 breast cancer SNPs were the strongest for predicting breast cancer across disparate ancestries. The PRS created for an individual woman is the sum of ancestry-specific PRSs weighted according to her ancestral genetic composition, creating a PRS for all ancestries.

Weighting was based on data from 31,126 self-reported Black women referred for hereditary cancer genetic testing, published data from the Asian Breast Cancer Consortium and data from 24,259 self-reported white/non-Hispanic hereditary cancer testing patients as well as a meta-analysis of the GWAS literature. The weight of one protective SNP was estimated from 9,322 self-reported Hispanic patients.

The PRS for all ancestries was developed in 189,230 women and validated independently in 89,126 additional women. In the validation cohort, about 23% of women had breast cancer in both groups, and 30.5% in each group had a first-degree relative with breast cancer. In both groups about two-thirds were self-reported white, 10 to 12% Black and 2% Asian.

Interestingly, European-derived breast cancer SNPs were common to all ancestries. At least 95% of breast cancer SNPs had a 1% or higher frequency of risk alleles within each of the self-reported patient populations. “There is marked genetic similarity between different ancestries,” Dr. Pederson observes.

Significant discrimination within subcohorts

The new PRS provided significant discrimination within each ancestry subcohort. For those in the top decile, the risk for breast cancer was approximately doubled compared with typical women, while among self-reported Black women the risk was increased by about 44%.

“The discriminatory power in women of African descent was significantly improved and we hope to only improve on the findings further over time. Discrimination for Black women is expected to further improve as additional data become available,” says Dr. Pederson.

Indeed, she believes one of the most important outcomes is that this new framework for developing a PRS for women of all ancestries can be adapted with new data. “The calibration is centered for all populations so that no longer are the estimates inaccurate for other races. It’s really exciting.”