“Seeing how CanRisk is used makes all our efforts worthwhile” – Antonis Antoniou wins Don Listwin Award

You and your team have been developing breast and ovarian cancer risk prediction models that look at several risk factors. How do you envision their application in clinical practice?

Yes, we have been working on the BOADICEA models — short for the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm.

The models combine a wide range of data from rare genetic variants in high- or moderate-risk cancer genes (such as BRCA1) to common genetic variants and family history of cancer. They also include lifestyle, reproductive, hormonal and anthropometric risk factors, and in the case of breast cancer, mammographic density, to help predict someone’s future risk of developing breast and ovarian cancer.

The models can be very useful for cancer risk stratification. They help us identify who is at higher risk and who could benefit most from prevention and early detection strategies.

For example, they can help determine whether someone should start mammographic screening earlier, undergo more frequent screening, or be offered alternative screening modalities such as MRI if they are at very high risk. The models can also guide decisions about risk-reducing medications or surgeries and identify those who might benefit most from making lifestyle changes.

How did you develop the model, and what sets it apart from other risk prediction tools?

We have been working on these models for over 20 years. The research underpinning BOADICEA involved developing novel statistical methods to model susceptibility to breast and ovarian cancer and numerous studies based on large international consortia to identify and characterise genetic susceptibility variants for these cancers.

In addition, we developed ways to integrate information on cancer risk factors from diverse data resources into comprehensive, multifactorial models. These models have been independently validated across different countries and clinical settings.

One of the key strengths of BOADICEA is its comprehensive approach. It incorporates most of the known risk factors for breast and ovarian cancer, including both genetic and epidemiological data. As such it represents a significant step forward in assessing multifactorial risk, maximising the informativeness of the predictions.

The CanRisk tool uses the BOADICEA algorithm to easily calculate breast or ovarian cancer risk. Tell us how you developed the tool and got it into clinical practice

CanRisk represents a collaborative effort from a large, multidisciplinary team, including statisticians, data scientists, software developers, applied health scientists, clinicians, and patients.

Importantly, we co-developed it with end-users, the healthcare professionals and patients, ensuring it meets the needs of those using it.

The process involved exploring the needs of end users, technical development, prototyping and testing with these users. We went through an iterative process of refining and testing the tool based on feedback, and importantly gaining regulatory approval for its use as a medical device.

The entire process from developing the BOADICEA model to creating CanRisk came with several challenges. Issues like integrating data across different sources, developing complex computational algorithms, and refining the user interface for clinicians and patients were certainly complicated. And of course, then there were the issues around securing regulatory approval as a medical device.

These challenges pushed us all beyond our comfort zones. However, it was through overcoming these challenges that we learned the importance and value of multidisciplinary teamwork – this has been a great example of a collaborative effort.

CanRisk is recommended for determining eligibility for high-risk cancer screening, and informing cancer risk management. Do you think people are receptive to learning about their cancer risk?

Yes, CanRisk is now endorsed in clinical management guidelines worldwide. Just this year, it was also included in the new NICE guidance on managing ovarian cancer risk. CanRisk is already widely used in many clinical settings, particularly in clinical genetics and screening.

Since its release in 2020, CanRisk has been used to perform more than three million breast or ovarian cancer risk assessments, helping to guide decisions on screening and prevention options. CanRisk is also used in several ongoing trials and clinical implementation studies focused on risk-stratified screening and prevention.

For me personally – and I am sure the entire CanRisk team feels the same – seeing this endorsement and how CanRisk is used makes all our efforts worthwhile. It is the best form of recognition we could ask for.

If we take breast cancer for example; since it’s one of the most studied cancers, multiple studies have shown that breast cancer risk assessment itself is widely acceptable to the public. However, more research is needed to understand how well this approach is accepted in primary care, where it could have the biggest impact.

One of the goals of our CRUK CanRisk programme, is to assess the acceptability and feasibility of proactive, multifactorial risk assessment in primary care on a larger scale.

There are exciting developments in screening technologies, such as multicancer detection tests (including liquid biopsies) and advances in imaging, which offer new opportunities for early cancer detection.

However, like most screening methods, they come with challenges, such as the risk of overdiagnosis and overtreatment. For example, multicancer detection tests might require further invasive or expensive investigations, which can have their own risks. Also, managing negative diagnostic tests following a positive screening result and dealing with false reassurance can be complex. Therefore, minimising unnecessary testing among asymptomatic populations is critical.

This is where risk stratification comes in – it allows for more tailored screening, by identifying individuals who are most likely to benefit from the screening tests. This way we can optimise the timing and frequency of screening, while reducing the potential harms from overdiagnosis.

Whilst early detection research has yielded some incredible advances, there is still a long way to go to rule out late diagnosis of cancer… what areas of research give you the most encouragement that we will get there?

As a data scientist, I’m particularly excited about the increasing availability of population-scale multimodal health data, and large population-based cohorts like the UK Biobank and Our Future Health.

Advances in cancer genomics, multiomics, imaging technologies, and machine learning also offer great potential. Together, these innovations will significantly improve our understanding of who is most at risk of developing cancer, and hence advance our ability to prevent, detect and diagnose cancers early.

The field of early detection encompasses a huge variety of approaches and research expertise – how important is it to get the right collaborations in place for any given project?

Collaboration is essential for impactful research.

Our experience with CanRisk and related work, has shown that engaging experts from different disciplines – nationally and internationally – results in more innovative, creative and well-rounded solutions. Each member brings a unique perspective and expertise, that helps ensure the project not only achieves its scientific aims but also addresses the real-world challenges of implementation. I am sure the same principles apply to the entire field of early detection.

I would say that getting the right team in place is essential for success!