London: A group of British scientists has created a novel model that effectively forecasts the probability of a woman developing and succumbing to breast cancer within a decade. The findings, featured in The Lancet Digital Health, were established by analyzing de-identified data from 11.6 million women aged 20 to 90 during the period from 2000 to 2020. This group had no previous history of breast cancer or the precursor condition known as ‘ductal carcinoma in situ’ (DCIS).
The ability to identify individuals at a higher risk of lethal breast cancers could significantly enhance screening procedures. Such high-risk women might be encouraged to initiate screening earlier, undergo more frequent screenings, or be assessed with alternative imaging methods. This personalized approach could potentially reduce breast cancer mortality while preventing needless screening for women with lower risk. Moreover, women with elevated susceptibility to developing life-threatening cancers could also be candidates for interventions aimed at preventing breast cancer, according to experts from the University of Oxford.
Julia Hippisley-Cox, an Epidemiology Professor at the university, remarked, “This is a significant new study that offers a potential novel screening approach. Risk-based strategies could provide a more balanced assessment of the pros and cons of breast cancer screening, providing women with more personalized information to assist in decision-making. Risk-based methods can also optimize the use of healthcare resources by targeting interventions toward those who are more likely to benefit.”
The researchers examined four distinct modelling techniques for predicting the risk of breast cancer mortality. Two methods were conventional statistical models, while the other two involved machine learning, a form of artificial intelligence. All models incorporated identical data variables, such as age, weight, smoking history, family history of breast cancer, and hormone therapy (HRT) use. The models’ precision in overall risk prediction and within diverse groups of women, including those from various ethnic backgrounds and age brackets, was assessed.
The study employed a technique called ‘internal-external cross-validation,’ which divides the dataset into dissimilar parts, based on regions and time frames, to determine how effectively the model can adapt to various circumstances. Outcomes indicated that a statistical model established using ‘competing risks regression’ demonstrated the highest overall accuracy. This model most precisely predicted which women would develop and succumb to breast cancer within a decade.
The machine learning models were comparatively less precise, particularly when applied to different ethnic groups. The researchers noted that if further investigations verify the reliability of this fresh model, it could be deployed to identify women with a heightened risk of fatal breast cancers, who could then benefit from enhanced screening and preventative measures.
