When Less is More — Short Term vs Lifetime Breast Cancer RiskFeb 24, 2021
Two well-known and widely used lifetime breast cancer risk assessment tools include the Tyrer-Cuzick model version 8, and the Gail Model.
However, the more recent short-term risk models that have emerged that use artificial intelligence to provide actionable information at point-of-care may be poised to finally move the needle in precision breast cancer risk assessment.
Data inputs to breast cancer risk models must be:
- Reliable and reproducible
- Derived from standardized measures
- Complete, with no missing values
- Accessible on-demand for risk-based stratification
Most traditional lifetime breast cancer risk models require extensive and time-consuming collection of self-reported clinical patient history and other risk factor data that can be unreliable and subject to recall bias.
The emergence of electronic systems for collection of risk factor data does nothing to overcome these challenges as the integrity of the source data remains unaffected. Furthermore, such systems are deployed in a limited number of mammography practices and are not well-integrated with image archiving systems and automated A.I. solutions that provide image feature and clinical information that is associated with breast cancer risk.
A review of the commonly used risk assessment models for breast cancer suggests they can be complex, confusing, and cumbersome to use. Due to the time-consuming nature of data collection and lack of data integrity associated with patient responses to questionnaires for traditional lifetime risk models, they are impractical to deploy at a population level.
Automated short-term risk models provide actionable information in contrast to their lifetime risk counterparts which rely on long-term and less accurate outputs that do not inform immediate action inpatient care.
A case where less is more.
What is a Short-Term Breast Cancer Risk Model?
Simply put, a short-term risk assessment identifies the level of risk that a patient has for developing breast cancer in the next few years.
Traditional risk models provide 10-year and/or lifetime risk assessments (whereas short-term risk models predict the risk of developing breast cancer in the next 1-5 years) with their accuracy and generalizability varying based on the characteristics of the population they were trained on.
Short-term risk models can be categorized into comprehensive, survey-based short-term risk assessments like the 5-year Gail Model and the Breast Cancer Surveillance Consortium Risk Calculator or automated/rapid short-term risk assessments.
Compared to their more traditional comprehensive survey-based counterparts, A.I. powered short-term risk models excel in rapid generation of standardized risk scores at point-of-care. A recent analysis shows that a simplified short-term risk model using a small set of image-derived factors performs comparably to more complex traditional risk models.
This new breed of A.I. algorithms provides a scalable solution for population-wide deployment of risk-stratification tools.
What Risk Factors do Short-Term Breast Cancer Risk Models Use?
Breast cancer risk models vary in the risk factors they include.
Some of the comprehensive short-term breast cancer risk models include risk factors including, but not limited to, patient age, menarche, biopsy and/or family history, and breast density.
However, studies have shown that breast cancer risk models for use in screening mammography can use a smaller set of readily available risk factors derived solely from mammographic image and pixel data that are reliably reproducible for more efficient and simpler clinical use.
Breast density is now considered to be a necessary risk factor to include in risk models.
Breast density is recognized by national, professional, and regulatory bodies as a key factor in determining breast cancer risk, where it both reduces the effectiveness of mammography due to its masking effect, and is an independent risk factor for breast cancer. Breast cancer risk models that incorporate breast density perform significantly better than those that do not, and the more reproducible and precise the breast density measure is, the more reproducible and precise the risk model results. This is important because it enables health system-wide application of tailored risk-based screening protocols.
New A.I. based breast cancer risk models use the data universally available in digital mammograms that contain both risk factor and pixel data, simplifying and standardizing data collection and improving data integrity.
Increasingly, new A.I. driven automated short-term risk models use mammographic features to produce risk estimates. While they vary from model to model, many include the same core set of risk factors that traditional lifetime risk models default to when other risk factor data is unavailable including but not limited to patient age, breast density, and body mass index.
Why are Short-Term Rapid Risk Assessments More Practical for Clinical Use?
Simply put, they scale effectively for population-wide deployment to support standardized risk-stratification of the population with no additional burden to the health system or patients.
Additionally, studies show that lifetime risk models perform increasingly poorly the further out in time that the predicted risk is computed, especially for higher risk patients. Even when using lifetime risk models, a woman’s lifetime risk changes over time and should be reassessed regularly.
The use of automated risk models enables risk assessment for every mammogram taken at every screening exam within a health system, making risk-based stratification in the clinical setting practicable and pragmatic without the need to alter existing IT infrastructure.
Recent research suggests that risk models using image-derived risk factors can generate standardized and reproducible results, mitigating the risks of poor reliability and reproducibility associated with self-reported risk factors used in traditional risk models.
How to Improve Patient Care with Short-Term Risk Assessments for Personalized Medicine
“One size fits all” is an oxymoron. Why would that be any different for breast cancer screening?
A one-size fits all model for patient care, where all patients receive the same level of resources, can be clinically ineffective and very costly. A health economics headroom analysis considering a health system in the USA with 100,000 women who start annual screening at 50 years old and regularly participate in a breast cancer screening program for the whole of their lives has shown the potential for $300 Million to be saved over the course of their lived by better stratification of high-density patients.
Patients who have normal mammogram results, and are not at a high risk of breast cancer may benefit from longer screening intervals and less frequent mammograms. Just as importantly as identifying women at high risk, it is important to identify women at low risk to avoid unnecessary adjunctive imaging – such women can receive more tailored models of care to minimize the high level of intensive support associated with standard clinical practice.
By risk-stratifying screen eligible patients, the high-intensity resources and adjunctive imaging can be reserved for high-risk patients.
Risk-stratification enables radiologists to identify the appropriate level of care and follow-up services for patients. Rapid short-term risk assessments that provide information at point-of-care are poised to move the needle in precision breast cancer risk assessment by providing risk scores to improve clinical confidence and support clinical decision-making through risk-stratification for tailored patient care.
At the individual level, risk-stratification is a critical step for developing a personalized care plan for each patient. A recent study suggests that risk-stratification may improve breast cancer mortality rates, and could reduce unnecessary harms (physical and psychological) from breast cancer screening. At the health system level, risk-stratification can help with optimizing resources, improving clinical outcomes, and improving overall operational efficiencies.
How to Optimize Operational Efficiency Amidst Radiologist Shortage with Short-Term Breast Cancer Risk Models
Many mammography practices face resource challenges that have been exacerbated by the COVID-19 pandemic. In addition to the existing global shortage in radiologists, there is increasingly a lack of radiology residents opting to pursue mammography as a subspeciality. In fact, studies have shown that the majority of radiology residents have expressed that they would not want to spend more than 25% of their time reading mammograms.
An important driver of this tendency for residents to pursue radiology subspecialties other than mammography includes the stress of interpretation which is linked to the high prevalence of radiologist burnout. Studies have shown that most mammograms are read by radiology generalists, with only approximately 30% of mammograms being read by breast imaging specialists.
Mammography practices can optimize clinical efficiencies by using actionable insights from short-term breast cancer risk models to flag the most challenging high-risk cases to subspecialty breast radiologists instead of general radiologists.
As such, in addition to enabling population-wide risk-based stratification for the development of actionable breast screening follow-up protocols that are patient-specific, short-term breast cancer risk models can mitigate the pressures associated with understaffing by optimizing staff resources for greater clinical practice efficiencies.
The bottom line is — simplified risk models make it possible to cost-effectively scale risk-stratification across a whole population for greater operational efficiencies.
Densitas offers a short-term risk model that uses image-derived factors and provides automated results at point-of-care. Learn more about how adopting a short-term risk model can improve clinical confidence and optimize operational efficiencies in your mammography facility.
Let’s Stay Connected, Subscribe for Updates
Join our email list to stay up to date on the latest advancements in breast health technology.