Ian Rowe and Jack Halligan showcase how artificial intelligence could be used to improve cancer diagnosis and treatment
Though it may not feel like it, there is cause for optimism in healthcare. Advancing technologies, the increased availability of better therapies and personalised treatment are a few examples of things that are driving this optimism.
Artificial intelligence (AI) and machine learning have huge potential on a broader scale across a range of healthcare applications. AI thrives where there is an abundance of data; Genomics England’s 100,000 Genomes Project is one high-profile example in which AI has been deployed.
We are also seeing how the technology we already have is changing the way we live and our understanding of our own health – something perhaps accelerated by the global COVID-19 pandemic. The necessity for remote medical consultations during the past two years may pave the way for a permanent change in the way we engage with GPs and other primary care services.
Smartphone apps are already being developed to give patients with conditions such as diabetes better support and easier access to healthcare providers. Perhaps patients with a broad range of chronic conditions will soon be better placed to manage their own health.
The way we access our own health records is also changing. Later this year, people in the UK may be able to access new entries in their medical records via the NHS app. In time, it is conceivable that we may have access to our complete medical histories from our mobile phones.
AI-enabled cancer screening
One of the key objectives of emerging AI solutions in healthcare is achieving better outcomes for cancer patients by detecting the disease at an earlier stage.
As shown in Figure 1, more than 50% of cancers detected via screening programmes are stage 1 cancers (in other words, early-stage cancer), which are more likely to be treatable or curable. The use of AI to improve detection rates on screening imaging is one of its most advanced applications in healthcare. This is helped by the fact that imaging in healthcare has been standardised for decades, providing a source of well-coded longitudinal data, which is essential for ‘training’ AI models.
“Even experienced radiologists face a challenge when interpreting mammogram images during cancer breast screening”
Imperial College London’s Institute of Global Health Innovation recently worked with Google Health and the NHS on a trial to evaluate an AI system for breast cancer screening, which was ‘trained’ on 29,000 mammography images. Encouraging results were reported, with the AI system outperforming human readers and demonstrating a small reduction in false positives.
Even experienced radiologists face a challenge when interpreting mammogram images during cancer breast screening, and incorrect interpretations do occur. In the UK, two radiologists must review each mammogram; AI could potentially be used to improve detection at this stage, and/or replace the need for a ‘second reader’. However, the widespread adoption of systems to enable this may still be five to 10 years away.
Protection insurance
Could AI be a game changer for protection insurance? Realistically, not in the short term – perhaps for future generations. Introducing new technologies to healthcare can be a long and difficult process, with one study putting the average period at 17 years. Clinicians and academics also describe an ‘AI chasm’ between the development of algorithms to the demonstration of value through scalable deployment – largely because of the requirements for clinical trials, the restructuring of clinical processes and staff training.
For cancer screening, we may be close to seeing the introduction of AI in a clinical setting, with the goal of refining and potentially increasing accuracy for existing screening processes. In terms of new technology providing benefit to cancer patients, this is an exciting development – perhaps paving the way for future advances. However, in the context of morbidity in the world of protection insurance, it is unlikely to represent material change. Currently, only around 6% of all cancers are detected via screening, and screening is only viable for three types of cancer: breast, cervical and bowel. While the prospect of deploying AI to enhance screening protocols is real, this is a small step on the way to achieving faster and earlier cancer detection on a large scale. The take-up for cancer screening itself remains a challenge, with UK figures currently showing that 74% of women attend breast cancer screening within six months of an invitation. The figure for cervical cancer is lower, at 70%–73%. For bowel cancer, numbers dip again to just 50%–58%.
In terms of morbidity or mortality within the insured population, it is likely that an increase in the uptake of cancer screening will have an impact before technological innovation does, at least in the foreseeable future. This scenario is not beyond the realms of possibility when you consider the ‘Jade Goody effect’ of the late 2000s, where a diagnosis of cervical cancer in an individual with a high media profile resulted in a notable spike in attendances for cervical cancer screenings.
Immunotherapy innovation
Immunotherapies are a range of therapies that use the body’s own immune system to identify and fight against cancer. They are a relatively new arrival in cancer care – most immunotherapy treatments currently available have been approved by the US Food and Drug Administration within the past 10 years.
One type of immunotherapy, monoclonal antibody treatments, was considered a potential game-changer when it first came onto the market. It is considered by many oncologists to be the fourth pillar of cancer treatment, alongside surgery, chemotherapy and radiotherapy. While there is no doubt that immunotherapy has had a positive impact in the field of cancer therapy, it is far from a ‘silver bullet’. For example, lung cancer is the most common cancer death in the UK, with non-small cell lung cancer (NSCLC) being the most common form. However, only approximately 30% of patients with an NSCLC diagnosis are considered strong candidates for immunotherapy. A US study has showed that within this group of patients, five-year survival may be increased by 5% to 25%. While this represents a significant advance in outcomes, life in many cases will be prolonged by less than one year.
“For the time being, our best defence against cancer is prevention through healthier lifestyles”
In recent years, there has been talk of ‘cancer vaccines’ being used to incite an immune response against cancer cells. Unlike many common vaccines (such as the coronavirus vaccines), these would treat disease, rather than preventing it. Many are mRNA vaccines and are being trialled across the globe for almost every type of cancer, although only one cancer vaccine is currently on the market – in the US, for prostate cancer.
These vaccines could treat a broad range of cancers. They are versatile, relatively fast and cheap to manufacture, and may be personalised for individuals’ biomarkers. Although this increases time and cost, early trials are encouraging.
Immunotherapy in the protection claims environment
Immunotherapy drugs will be reasonably familiar to anyone who has recently managed claims for terminal illness. It is often clear when a patient is considered a good candidate for immunotherapy. Being a good candidate for immunotherapy might, for the most fortunate patients, result in materially improved life expectancy. However, this is certainly not the norm, and assessors are generally aware of this.
For chemotherapy, there is good statistical data relating to outcomes over time and survival statistics; immunotherapy, on the other hand, has not been around long enough for this. While this type of data is not usually the sole basis for assessing an individual terminal illness claim, it is important to understand the potential benefits of available treatment.
Prevention still the best defence
Any breakthrough in the detection and treatment of cancer typically attracts headlines, but the fact remains that there are no cure-all solutions on the horizon.
The development of better healthcare through evolving technology is an intricate and painstaking process and, for the time being, our best defence against cancer is prevention through healthier lifestyles.
In the meantime, other innovations and environmental impacts continue to gradually influence and change public health and healthcare.
For protection underwriters and claims assessors, the healthcare environment is already changing, and there are sure to be new challenges ahead. With greater awareness and more opportunity to manage their own healthcare, customers may retain knowledge about their own health that is not in their health records – which, at this point, remains a foundation for managing risk.
On the other hand, there are also likely to be opportunities. With greater empowerment for individuals to effectively understand and manage their own health, it is entirely possible that more favourable underwriting terms and greater access to protection products can be achieved.
Ian Rowe is a claims director at Pacific Life Re
Dr Jack Halligan is a policy fellow in health innovation at Imperial College London and co-founder of Prova Health
Image credit | Getty
