Oncology and marketing – AI has a critical role

Dania Alarcon, Chief Medical Officer of WPP’s VML, says there are four key assumptions that should guide the use of AI in oncology marketing

Marketing has a critical role to play in informing individuals about their health needs. When it comes to oncology, we are seeing the use of AI in oncology move from tentative to the mainstream.

According to the American Society of Clinical Oncology (ASCO), AI is now firmly part of oncology lexicon. It is there across the abstracts of ASCO presentations – not just as a key word but as meaningful evidence of the direction science is taking. Since 2020, each year more than 100 abstracts have mentioned AI in relation to oncology – and this is just in the summaries of presentations at one annual flagship event. In 2023, this number was around 150, and the jury is still out on 2024. 

The overarching theme of these presentations is that AI’s potential can only be met by using robust data and varied sources of data. They also emphasise the need to be intentional about addressing the potential bias that sits within data sets. These presentations run the gamut from using AI to aid the drug development process, to tailoring and customising therapy for specific patients, to predicting what the response to therapy might be based on models and algorithms being developed based on the data available. 

And a new discipline requires a new outlet for critical thinking. The ‘New England Journal of Medicine’ is launching a journal focused specifically on AI in 2024. This will help increase the exploration of AI applications to healthcare, and to help identify the applications that will lead to a better path – and outcome – for patients. 

But first we need the AI bias to be countered with intentional inclusivity, especially in situations where trial data – and even recruitment and outreach – does not include populations that are reflective of the overall patient landscape. We need to figure out how we can address that deficiency at the trial stage to ensure that treatment and recommendations are not shrouded in bias. 

It is true to say that pharma has been slow to adapt to the AI-driven ‘new normal’ we are experiencing in so many other industries. This is natural for a sector that is rightly highly regulated and rigorously tested. But AI is now being surfaced as an important tool in cancer detection, treatment and aftercare. The inflection point has been reached. 

Four principles to guide us 

There are four key assumptions that should guide the use of AI in oncology marketing. 

First, the human element will always be critical when actioning AI findings in marketing. AI alone will never be the solution, but it will be a means of analysing robust datasets, finding patterns and defining actions to address what is learned, and doing so objectively. AI could also be a means of generating extensive content, but it will still need to be sorted and analysed by humans who understand the bigger picture. 

Second, AI can be trained to overcome analytical bias – bias that is endemic in oncology marketing. Traditional methods of research, and a focus on look-alike audiences, result in a lack of objectivity which introduces bias. This will likely result in underserved communities continually being overlooked and underrepresented. AI tools must take this into account if outcomes are to be meaningful to all populations, including the most marginalised.

Third, the potential for AI in oncology will rely on robust, unbiased data sets. Identification of the first AI predictive biomarker for high-risk prostate cancer patients highlighted the potential for building on such robust data sets. After all, AI will only be as good as the data that the AI tool is presented with (and current data is based predominantly on white males from the western world). There is a long way to go before the data can be anything like universal. Collaboration from industry partners will be paramount to aggregating robust data.  

Finally, AI bias must be countered with intentional inclusivity. Finding the optimal balance of AI learnings and human insights will be key to driving better outcomes for all cancer patients, so we must ensure that we make a clear effort towards robust, inclusive data collection, analysis and application. 

These four key assumptions should guide the use of AI in oncology marketing. They are important, not least because the use of AI tools in this area is increasing rapidly but, from what we are seeing in research circles, this is just the start. So, let’s make sure our principles are secure while we pursue this very exciting and important advance for oncology itself and the communications that go with it. 

Healthcare meets marketing 

The strength of AI is its ability to uncover patterns – quickly – through mountains of data. This enables humans to dovetail with AI tools to make predictions and propose the best actions. 

This goes beyond the science of oncology to the communication of the science behind disease to populations, clinicians, governments, regulators and other stakeholders. AI is already making a difference in healthcare outreach and education. For example, WPP’s VML and its Creative Data Group collaborated on a smallpox simulator for Meridian, the maker of smallpox medication, to help governments plan for future infectious disease outbreaks.  

A key learning from Covid is that preparedness is critical to containment. VML’s tool maps how potential smallpox outbreaks might occur throughout the world. This data visualisation tool is a powerful reminder for the appropriate authorities to act and reach the right communities, using the right data, prepare the right messaging and do so at the right time. These are the foundations of any content-driven marketing campaign. 

For oncology, if the data used for prevention, treatment and care is to be representative, there must be diverse representation in clinical trials. This requires impactful outreach materials and recruitment strategies that will help with intentional inclusivity in trials and screening. Commitment from industry partners is already being required by regulatory bodies in planning for diversity in clinical trial recruitment; however, the need to ensure that diverse audiences and voices are integrated into outreach efforts underscores the need for human voices and, in particular diverse human voices, to be part of the conversation.  

AI will also help with the personalisation of communication and outreach in a way that connects in a more meaningful way. It will enable patient care and treatment management choices become more intuitive, and more patient friendly, based on the information that AI provides. But human insight will always be critical when it comes to actioning AI responses.  

And then we can access the opportunities afforded by generative AI to create a vast array of different communication materials to reach at-risk groups and do so in an informed way, with AI guiding the process. There will always be the need for a human lens, optimisation and input that will help us define what parts of this generative content will resonate best and achieve the best outcomes for all.