Can AI Transform Clinical Research?

Artificial intelligence is fundamentally transforming medical research. Digital twins, synthetic data arms… the science is advancing swiftly but raising major issues. Is a digitally simulated control group effective? How can the ethics of algorithms be guaranteed? Faced with international competition, will France manage to combine innovation, scientific rigour and technological sovereignty?

Whether as a generator of images, text or code, AI is experiencing meteoric adoption amongst the general public. But in healthcare, progress is more nuanced. “Barely 10% of practitioners say they’re comfortable with AI,” notes Loïck Menvielle, EDHEC professor and director of the Management in Innovative Health Chair. Yet the integration of AI could profoundly improve medicine, offering automated data analysis, the creation of protocols, anticipation of clinical trajectories or even the intelligent selection of patients in clinical trials.

These tools, based on powerful algorithms, mark a major shift towards more predictive and personalised medicine. But as these technologies become more established, they raise major ethical, methodological and regulatory challenges that healthcare systems must rise to meet.

AI-Assisted Clinical Trials? Yes, but…

“AI is everywhere: on our television screens, computers, phones… We’re constantly talking about it,” observes Caroline Beaufour, innovation lead in clinical development at Servier. “In the reality of our pharmaceutical industry practices, however, it remains a subject for exploration. We’re still working out how and why to deploy it.” For practitioners and clinical researchers, that horizon is still a long way off, as confirmed by Antoine Iannessi, vice president of medical affairs at Median Technologies, an organisation that supports medical sector players integrating AI into their practices: “The technology is mature, but the framework (financial, regulatory or usage-wise) isn’t there yet.”

Andy Karabajakian, director of medical oncology at Owkin, is rather more enthusiastic. “In two clinical care domains in particular, AI is already a reality.” This is the case in radiology, where it offers diagnostic assistance via image analysis (detecting fractures and pulmonary nodules, notably), but AI also has a place in supporting the development of digital pathology. This Franco-American startup has, for instance, developed a deep learning tool, RlapsRisk, capable of predicting cancer relapse risks. “It’s currently being tested at Hôpital Bicêtre [a hospital in the north of Paris].” What remains to be done is getting the tool certified by health authorities — a mandatory step that should also reassure future patients.

A Matter of Defiance

Trust is at the very heart of the problem. “EDHEC conducted a study on the subject,” explains Menvielle. “Of women surveyed, 40% declared they wouldn’t trust this type of solution.” This was according to the 2024 EDHEC IPSOS Bristol Myers Squibb barometer. 

But for Pierre Loulergue, an infectious diseases specialist, immunologist and member of the Ethik-IA think tank, this concern could become an opportunity. “We’re at a historic moment in healthcare. We still have the time and ability to properly guide this technological revolution.” And in so doing, they will also be able to properly reassure patients and healthcare professionals. Karabajakian emphasises the importance of maintaining the “explainability and traceability of these tools” to stem collective mistrust of artificial intelligence.“But lifting the veil on AI and promoting transparency of the process for users isn’t straightforward,” Iannessi cautions. “It requires thinking through all our models with this end in mind from the outset.”

When applied to clinical trials, AI explainability remains equally challenging, as Caroline Beaufour reminds us: “We currently struggle to detect the AI and non-AI components in medical treatment evolution.” Her hope, however, rests on the technology’s capacity to identify patients most likely to respond favourably to treatment. But to do this, like many AI applications across all domains, the tool needs “fuel”: data. And in healthcare, data is scarce and hotly contested.

Programme manager of Digital Ségur at the Ministry of Labour, Health, Solidarity and Family Policy, Olivier Clatz dispels these concerns. The programme he leads aims to generalise and streamline health data use, and France appears destined for more favourable prospects than its European neighbours. “Since late 2023, over 250 million health documents have been exchanged via Mon espace santé [the digital health records tied to France’s social security system]. The country has one of the world’s most advanced digital care pathways.” This volume of data is now set to be enriched by digital imaging data and, soon, genetic analyses. 

Through these testimonies, one conviction has emerged: AI will only revolutionise clinical research if it operates within a robust, transparent and ethical framework. Neither a technological miracle nor a simple tool, it calls for a profound overhaul of our practices, regulations and collaborations. More than just a challenge for innovation, this is a collective project aimed at delivering more precise, more humane and sustainably enhanced healthcare.