AI could personalise clinical trial consent, but privacy concerns remain

Blanka Hezelova’s keynote presentation outlined the opportunities and concerns surrounding AI applications in clinical trial conduct. Credit: Athanasios Psimadis/ Arena International Events.

Artificial intelligence (AI) is revolutionising clinical trials, offering ways to improve the consent process, but concerns regarding the complete reliability of the technology for sensitive tasks persist.

The limitations of AI to enhance trial participant engagement, and how it may overcome these, were outlined by Blanka Hezelova, associate director at GSK on 11 June. Hezelova spoke on the impact of AI during a keynote address on the second day of the 12^th Annual Arena International Outsourcing in Clinical Trials UK & IR conference held in London.

Hezelova said the clinical trial industry is in a “transition period”, such that much of the data processing work is being passed from humans to AI to increase capacity and restrict human error. She noted a number of AI applications already entrenched throughout the industry: protocol creation, document review, and simulations are already done using AI to mitigate risk and reduce future operational burdens.

Also highlighted were what Hezelvoa identified as “transformational, long-term changes”. Alongside anticipated AI applications for budget automation, generation of innovative new endpoints, and enhanced biosample operations, she pointed to more direct patient engagement as a focal priority of AI development, typified by personalised consent procedures.

Trial consent forms have often been noted by prospective trial participants as complex, impersonal, and unclear regarding data privacy, presenting a potential barrier to study recruitment. According to Hezelova, AI could offer more data-driven, personalised documents to overcome this barrier to successful trials.

However, there is a range of concerns challenging AI adoption in this role. Several practical risks, namely hallucinations, inadvertent bias, and inaccurate generalisations undermine confidence in the ability of AI to faithfully summarise data, said Hezelova. Additionally, she said, “AI, unfortunately, does have a very serious risk of becoming coercive and manipulative,” raising worries around the integrity of informed consent gained through AI tools.

Data security is another principal concern, said Hezelova, as the technology is at risk of sharing data inappropriately and putting patient privacy at risk. “Regulatory bodies are catching up slowly,” she said, noting a lag as governments struggle to keep pace with technological development.

In response to questions from the audience, Hezelova noted the enduring role of interpersonal interactions to maintain the trust of trial participants, particularly those from underrepresented communities; “You need to work with the GPs [general practitioners], and you need to work in the communities, and you need to work with people they completely trust.”