Breast cancer screening is a vital tool against the deadly disease, yet it faces its share of challenges. Although it reduces breast cancer-related deaths, it also has the potential to detect non-harmful tumors (overdiagnosis), leading to unnecessary treatments. This not only adversely affects some women but also drives up healthcare costs unnecessarily. 'Risk-based screening' is a strategy aimed at customizing screening approaches based on an individual's risk profile, aiming to maximize benefits and minimize drawbacks. Tailoring screening programs based on individual risks was recently identified as a way to refine screening strategies. Presently, most risk-based breast screening models estimate a woman's risk of being diagnosed with breast cancer. However, not all breast cancers are fatal, and the risk of diagnosis doesn't always align with the risk of death post-diagnosis. Now, researchers have devised a new model that accurately predicts a woman's likelihood of both developing and then succumbing to breast cancer within a decade.

The new model developed by a team of researchers at University of Oxford (Oxford, UK) predicts a woman's 10-year combined risk of breast cancer development and subsequent mortality. The aim is to identify women at the highest risk of deadly cancers in order to enhance the effectiveness of screening programs. Such high-risk individuals might be encouraged to initiate screening earlier, receive more frequent screenings, or undergo different types of imaging. This personalized strategy not only has the potential to reduce breast cancer fatalities but also avoid unnecessary screening for women with lower risk. Women with an elevated risk of deadly cancer could also be considered for preventive treatments against the development of breast cancer.

The research team explored four distinct modeling techniques to predict breast cancer mortality risk. Two followed conventional statistical methodologies, while the other two harnessed machine learning, a branch of artificial intelligence. All models incorporated identical data types, including age, weight, smoking history, family history of breast cancer, and hormone therapy (HRT) usage. The models underwent evaluation for their overall predictive accuracy, spanning various women's groups with diverse characteristics such as different age brackets and ethnic backgrounds. An approach called 'internal-external cross-validation' was employed. This method involves dividing the dataset into structurally distinct segments, based on factors like region and time frame, to assess the model's adaptability across different scenarios. The outcomes revealed that a statistical model constructed using 'competing risks regression' outperformed the rest. This model demonstrated the highest accuracy in predicting which women would develop and face breast cancer mortality within a 10-year span. The machine learning models displayed comparatively lower accuracy, particularly for diverse ethnic women's groups.

“This is an important new study which potentially offers a new approach to screening. Risk-based strategies could offer a better balance of benefits and harms in breast cancer screening, enabling more personalized information for women to help improve decision making,” said University of Oxford Professor Julia Hippisley-Cox. “Risk based approaches can also help make more efficient use of health service resources by targeting interventions to those most likely to benefit.” 

http://www.gzjiayumed.com/en/index.asp