By: Stephanie L. Graff, MD, FACP
In oncology we increasingly turn to precision tools, whether genomic profiles or targeted medications, and so it only makes sense that we start to turn to the same precision instruments to assess cancer risk. In a study in JCO Precision Oncology, Minnier et al. utilize the Million Veteran Program data set to assess clinical breast cancer risk models alone or in combination with multiple single-nucleotide polymorphisms (SNPs) folded into a polygenic risk score (PRS).
We have long recognized that a family history of breast cancer results in approximately a two-fold increase in a risk of developing breast cancer, which is incompletely corrected for by the known high-risk genetic mutations like BRCA 1 or BRCA 2 (accounting for <20% of that increase), or the intermediate genetic risk mutations like CHEK2 or ATM (which may account for another 5% of the increase).
There are wide, common, low-risk variants known as SNPs which likely account for the subsequent increased risk within families. Attempts to combine these common, low-risk variants into a predictable model resulted in PRS313, work which these authors further explored against the diverse Million Veteran Program dataset. Worth noting is that the original body of work that led to the selection of the 313 SNPs for PRS313 was done in largely white volunteers of European ancestry.
The clinical risk models utilized in this study included (1) Individualized Coherent Absolute Risk Estimator (iCARE) tool, and two models developed out of iCARE including (2) Breast and Prostate Cancer Cohort Consortium analysis (iCARE-BPC3), and another based on literature review (iCare-Lit).
These complex, computer software models run data from multiple sources and incorporate many of the data endpoints we are all familiar with as breast cancer risk factors like hormone-replacement therapy or oral contraceptive use, menarche, tobacco or alcohol use, family history, and age at first live birth, among others.
While there may be an imagined future wherein these models are paired with electronic health records or mammography software, they are unrealistic to run in a one-on-one clinical encounter as a means of assessing risk due to their complexity.
A total of 35,130 female participants were included in the million women’s program across 63 Veterans Affairs medical facilities, and included 31.7% African ancestry (AA), and 8.3% Hispanic. Unsurprisingly, age, higher body mass index, and higher use of oral contraceptives and hormone-replacement therapy all increased the likelihood of incident breast cancer diagnosis.
In looking across risk models, BCRAT and iCARE-Lit performed similarly for non-AA ancestries, and BCRAT was slightly superior in AA women. BPC3 underperformed across ancestries. PRS313 alone also only had a modest predictive model, best in non-AA women. When combining PRS313 with clinical models, the prediction accuracy improved, but again, that improvement was largely limited to non-AA persons.
While combining genotype data with clinical risk does improve risk modeling for women of European descent, these complex models and accessibility of profiling remain barriers to implementation. Furthermore, combining precision tools and clinical risk models developed in a Eurocentric model of medicine does not advance cancer risk prediction for AA women.
We need to be working with the Black community to develop tools unique to their lived experiences, risk factors, and diverse genotypes to understand and predict cancer in this population that, while less likely to develop breast cancer, remain the population at the highest risk to die of the disease.
JC Precision Oncology
Barbara Jacoby is an award winning blogger that has contributed her writings to multiple online publications that have touched readers worldwide.