New Publication Demonstrates Value of Combining Clinical and Genetic Risk Models for More Accurate Breast Cancer Risk Prediction
A new, peer-reviewed paper published in JCO Precision Oncology shows the clear value of incorporating genetic data with existing clinical models to better predict a woman’s risk of developing breast cancer. This effort also points the way toward improved risk prediction for individuals of any ethnicity, which is what we think is the most exciting aspect of the study.
Historically, predicting risk for cancer susceptibility is most accurate among people of European descent because the earliest genetic databases used to determine risk were built with data from those countries — Iceland, for example, created one of the earliest and largest genetic databases in the world by generating data from its population. That means anyone who shares common ancestors with people in those databases is more likely to get conclusive results from genetic testing than someone whose ancestors were not represented. Scientists around the world are working hard to address this problem by adding other populations to these databases so that genetic testing can be highly accurate for everyone regardless of ethnicity.
In this new study, Myriad Genetics scientists teamed up with researchers from nine other healthcare institutions and leading cancer centers to test another approach that we believe will significantly improve our community’s ability to generate reliable results for all individuals.
Here’s the concept: while existing clinical models for predicting someone’s risk of developing breast cancer are good, they’re not great. We posited that adding a limited amount of targeted genetic data would hone those models to provide more accurate results, particularly among women with a pronounced family history of breast cancer but who do not have the most well-known genetic variants associated with high risk. If that theory is correct, it would mean that we could improve results for any population simply by gathering similar genetic data among that group of people and adding it to the clinical models already in use, with further modifications as needed.
To test our hypothesis in the population that currently has the most data, we analyzed more than 100,000 women of European ancestry, assessed our results, and then evaluated the same approach in independent validation groups together that had more than 2,000 women. Our study focused on women between 18 and 84 years of age who did not have known pathogenic variants in about a dozen well-studied genes associated with breast cancer risk.
We combined 86 genetic variants that have been associated with breast cancer risk into an aggregated polygenic risk score and plugged it into the frequently used Tyrer-Cuzick cancer clinical risk model. Then, we compared predictions for women’s risk using either just the Tyrer-Cuzick model or the model plus the polygenic risk score. Finally, we validated our results in groups of patients not included in the first round of analysis.
The study demonstrated that polygenic risk scores were highly associated with family history, helping to account for some of the breast cancer risk missed by standard models. Our results showed that adding the genetic data “significantly improved discrimination relative to the Tyrer-Cuzick model for predicting risk of breast cancer [with] excellent calibration across age groups,” as lead author Elisha Hughes and collaborators report in the publication.
Importantly, the adjusted risk scores for many of these individuals would affect the cancer screening recommended for them. For the best chance of detecting cancer early, the American Cancer Society recommends regular screening for women who have at least a 20% lifetime risk of breast cancer. In our study, more than 2,600 women — that’s 8% of the group we analyzed — would have been classified as having less than 20% lifetime risk with the standard Tyrer-Cuzick model but were reclassified as having more than 20% risk with the additional genetic markers. These women should all qualify for enhanced surveillance, but would be missed with standard clinical models alone.
As the authors describe it, “Combined clinical and genetic risk models improve breast cancer risk prediction and may result in better allocation of cancer risk-reduction resources, such as chemoprevention and enhanced imaging techniques, to women with the highest combined risk.”
In addition to the compelling results showing that we can better target cancer screening tests by incorporating genetic data with clinical risk models, this study also proved our theory that adding a small amount of population-specific genetic data can make a huge difference in accuracy. Now that we have validated this approach, the next step will be to gather that kind of data for many different ethnic groups and see how well it improves risk prediction for people of non-European ancestry so that this technology tool can be available for all women.