Expanded Risk Model Identifies Women Who Need Additional Breast Cancer Screening

Creating a risk model that includes mammographic features, lifestyle factors, and genetics-based risk prediction scores can help providers identify which women at a high risk for breast cancer are most likely to develop the disease within two years of a negative screening. Providers can use those findings to direct these women for necessary supplemental screening.

With approximately 25 percent of all cancers being interval cancers – those detected between regular screenings – having this type of tool can improve patient outcomes, said a team led by Mikael Eriksson from the department of medical epidemiology and biostatistics at Karolinska Institutet in Sweden. They published the details of their model, as well as its efficacy, on Sept. 8 in Radiology.

“With this risk model, we aimed to create a protocol that has the potential to guide radiologists in identifying women with a negative screening result who are potential candidates for supplemental screening, more frequent screening, or risk-reducing medication,” the team said.

Related Content: Earlier Breast Cancer Screening Reduces Risk of Death

High-risk women are more likely to be diagnosed with a Stage 2 cancer with tumors larger than 20 mm, the team said. They are also less likely to be diagnosed with Stage 1 cancer or have estrogen receptor-positive tumors. With this knowledge, Eriksson’s team augmented existing risk models in an effort to identify more of these interval cancers.

They added details about breast density, masses, and microcalcifications, differences between left and right breasts, and age. In addition, they incorporated menopausal status, family breast cancer history, body mass index, hormone replacement therapy history, and tobacco and alcohol consumption. They also expanded the genetic components of the model by including 313 single nucleotide polymorphisms. Alongside this information, the team also pulled in tumor characteristics, immunohistochemistry markers, and clinical pathological variables from the Swedish national breast cancer registry.

To test the efficacy of their risk model, the team examined screening records from 10,350 women with an average age of 54 who were enrolled in the Karolinska Mammography (KARMA) Project for Risk Prediction of Breast Cancer. These women were screened at 18-to-24-month intervals. Of the group, 974 had a breast cancer diagnosis, and 9,376 were healthy. The average follow-up time was 4.9 years, and they averaged 1.2 years from their most recent breast cancer screening.

Overall, based on their analysis, the team found their full risk model had an area under the curve (AUC) of 0.77. In comparison, a model that just included imaging had an AUC of 0.73, and a model that also included family history and lifestyle factors – but not genetic risk factors – had an AUC of 0.74.

“We have shown that lifestyle factors and polygenic risk scores add to a short-term risk model built on in-depth analysis of three mammographic features and their differences,” they said.

From their results, they determined that women at high risk – who account for 8 percent of all women – experienced an eight-times higher relative risk of being diagnosed with breast cancer. That corresponds to a 30-fold gradient when compared to women of average risk, they said.

“It is recommended that women in this risk category undergo increased surveillance and risk-reduction interventions,” they advised.

Specifically, they added, the KARMA model identifies more women with Stage 2 cancers and tumors of 20 mm or more. And, many of these cancers are only detectable within three years prior to diagnosis.

“This suggests that an effective clinical risk model used for individualizing screening should be a short-term model,” Eriksson’s team said. “Our model identifies women in whom there is a high likelihood that they will be diagnosed with a cancer that was missed or fast-growing.”

Overall, the team concluded, their risk model can play a role in improved clinical care and better patient outcomes.

“With a high discriminatory performance, the model has the potential to support the decision regarding which women should be recalled for supplemental screening or should undergo more frequent screening,” they said, “or in whom risk-reducing medication should be recommended, thereby potentially improving overall prognosis of breast cancer and decreasing breast cancer incidence.”