The Hopkins team developed a model for gauging absolute risk of breast cancer for white women in the US that draws on the influence of nearly a hundred SNPs uncovered through genome-wide association studies, as well as non-modifiable factors like family history and age at menarche, and modifiable factors like BMI and smoking status. The model does not include breast cancer risk genes like BRCA1 or BRCA2.
As the researchers reported in JAMA Oncology today, they found that women in the top and bottom deciles had an average absolute risk of disease of 23.5 percent and 4.4 percent, respectively. However, they also found that women in the top risk decile due to non-modifiable factors who had low BMIs, didn't drink or smoke, and weren't on menopausal hormone therapy had a disease risk similar to that of an average women in the general population.
"People think that their genetic risk for developing cancer is set in stone," senior author Nilanjan Chatterjee from the Bloomberg School of Public Health at Hopkins said in a statement. "While you can't change your genes, this study tells us even people who are at high genetic risk can change their health outlook by making better lifestyle choices such as eating right, exercising and quitting smoking."
To develop their risk prediction model, Chatterjee and his colleagues drew upon genetic data and questionnaire information from 17,171 cases and 19,862 controls from the Breast and Prostate Cancer Cohort Consortium (BPC3), a collection of eight prospective cohorts from Australia, Europe, and the US.
The BPC3 study participants were genotyped at 24 SNPs, and the researchers used this data to develop a polygenic risk score. They then simulated and folded in data on 68 additional SNPs that weren't genotyped in the BPC3 participants. That simulated data was based on previous estimates of associations between those SNPs, case-control status, and family history. That then yielded a polygenic risk score based on 92 SNPs. Each of these SNPs individually has a small effect, but together explain a large degree of disease risk, the researchers said.
The final model of absolute risk Chatterjee and his colleagues presented also incorporated the influence of modifiable risk factors, which they garnered from a number of nationwide surveys such as the National Health Interview Survey and National Health and Nutrition Examination Survey, and age-specific breast cancer rates from the National Cancer Institute-Surveillance, Epidemiology, and End Results Program, and more.
From this, they estimated that the average absolute risk for a 30-year-old white woman in the US of developing breast cancer by the age of 80 is 11.3 percent. They further calculated that if all white women in the US were at the lowest risk from the four modifiable risk factors in their model - BMI, hormone therapy, alcohol use, and smoking - up to nearly 30 percent of breast cancers could be prevented.
In addition, they found that if women with the highest risk of disease due to non-modifiable risk factors fell in the lowest category of modifiable risk, then their overall risk of developing breast cancer would be about the same as an average woman from the general population.
Chatterjee and his colleagues suggested that a model like theirs could be used to stratify risk and motivate people to make lifestyle changes to avoid disease. "Everyone should be doing the right things to stay healthy, but motivating people is often hard," Chatterjee said. "These findings may be able to help people better understand the benefits of a healthy lifestyle at a more individualized level."
A related editorial in JAMA Oncology, however, questioned the Hopkins team's methods and highlighted the study's drawbacks, including how the researchers developed their genotype-based model.
Using simulated data, Vanderbilt University's Jeffrey Smith and his colleagues wrote, is "controversial." They noted using an imputation-based model could be problematic when making predictions.
"[T]he model may best be used to predict and to understand population trends that could be the subject of large public health interventions, as opposed to using it as a risk score calculator to inform a clinical decision for a specific patient," he and his colleagues added.
Smith and his co-authors further cautioned that the Hopkins team did not internally or externally validate their model - a drawback Chatterjee and his colleagues acknowledged - and didn't follow the established criteria for reporting a multivariable prediction model for individual prognosis or diagnosis. In addition, they noted that the study's calibration analysis relied on the same population as the one used to generate the model.
A more rigorous calibration analysis, they wrote, would have increased the confidence in the model's predictive ability, especially as it relies on simulated data. As such, Smith and his colleagues added that the model is not quite ready to guide clinical decision-making, though they wrote "that the overarching goal of the authors and research community is laudable."
Return to News Home