ICR discuss clinical trial recruitment

Earlier diagnosis, smarter recruitment — what struck me from Day 2 of the ICR Conference

By Dr. Graham Wylie, ICR Board Member & Executive Chairman, Medical Research Network (MRN)

Day 2 of the ICR Conference put innovative clinical trials in the spotlight, and four speakers across two sessions made the case — separately, but converging — that the future of recruitment looks very different to the present.

A few things stayed with me:

  • First, we are finally seeing artificial intelligence applied to recruitment in ways that move the needle, not just the slide deck.

  • Second, the organizations getting traction are those that have invested in trust as well as technology.

  • Third, the line between diagnosis and recruitment is starting to blur, which has real consequences for how we design and run trials.

Session 5 — Innovative clinical trials

Chaired by Stephanie Jones, Associate Director, Clinical Trial Delivery Unit, Biogen.

Ben Fell, Akrivia Health: AI & Electronic Medical Records

Akrivia is using a large language model (LLM) to read electronic medical records, looking not only for the diagnostic criteria a trial requires but also for the broader signs and symptoms that suggest the disease. Current focus is Alzheimer’s Disease, where they are demonstrably increasing the rate of early diagnosis.

The point that landed for me is specificity. When you bring inclusion and exclusion criteria into the screening logic, the cohort that arrives at the site is materially more likely to qualify. That means less wasted screening, faster recruitment and — crucially — happier site staff who are not running visits that go nowhere.

Sharon Allin, Biogen: Validated Digital Screening Tools

Sharon talked about a validated voice tool on a mobile device that predicts amyloid aggregation with around 80 percent accuracy. It does two jobs at once: it pulls forward possible diagnosis, and it strips screening activity out of the site visit. Sites love it, which is the strongest practical signal you can get for an innovation in this space.

This is also a neat example of what a diagnostic tool to facilitate recruitment should look like — light to deploy, validated, and obviously useful to the people doing the work.

Identifying and characterizing precision cohorts

Cosima Gretton, Our Future Health: Scaling Trust and Diversity

Our Future Health links phenotype and genotype data with electronic medical records to support recruitment screening. The scale is striking. They have 2.6 million participants enrolled, with 1.9 million currently analyzable, and around 770,000 fully genotyped. They are one of three providers to the National Health Service linking primary and secondary healthcare data to a genome database, with a substantial lifestyle questionnaire on top.

What I found more interesting than the numbers was how they got there. They tried recruiting through general practitioners and it did not work. So they went to the high street — physically, with teams talking to people, alongside paper letters. The result is a database with strong diversity across gender, age, ethnicity, socio-economic class and urban or rural location. Their conversion rate sits at 26 percent.

Cosima was very clear about what drives that conversion. It is trust. Participants believe the organization is acting in their interest, and that is what gets them to engage and to stay engaged. We should not lose sight of that — every recruitment system in the world will lose to a trusted relationship.

Practical takeaway: It is possible to join their network. Worth a look for anyone in the ICR community working in patient identification.

Peter Fish, Mendelian: Dissolving the Boundary Between Diagnosis and Enrollment

Mendelian is going after early diagnosis in rare disease in the United Kingdom. They have access to large reference databases (some from outside the United Kingdom) and to electronic medical records via the National Health Service, and they use International Classification of Diseases search criteria combined with free-text mining. They can also look for constellations of signs and symptoms that correlate with a diagnosis even when no diagnosis is recorded.

Current focus is inherited retinal disease — where they are tracking around 30 gene therapy trials coming through the pipeline — and monogenic epilepsy.

What this tells me, and what ties all four speakers together, is that the boundary between diagnostic work and trial recruitment work is dissolving. The same data, the same algorithms and the same patient relationship can move someone from undiagnosed to diagnosed to enrolled. If you are designing a study in a rare disease or hard-to-find population, this is where the action is.

Three takeaways (and a prompt for the DCT community)

  • Artificial intelligence applied to electronic medical records is no longer experimental — Akriva and Mendelian have made it operational.

  • Validated, patient-facing digital tools (voice, mobile) reduce site burden and accelerate diagnosis at the same time.

  • Big databases plus genuine trust beat sophisticated databases on their own — Our Future Health is the proof point.

  • These tools need other support to maximise their benefits — turning identified patients into randomized patients remains a challenge based on their distance from sites and the demand of trial participation on their lives.

For those of us working on decentralized approaches, this is a useful prompt. The patient identification end of the trial is now where the most interesting innovation is happening. If you would like to continue this debate, let us know. The DCT SIG could pick this up.

Ready to deepen your understanding of clinical research industry trends clinical trials? Join the ICR’s membership community, subscribe to our newsletter, or attend one of our upcoming webinars. Stay at the forefront of innovation in clinical research with the Institute of Clinical Research (ICR).

Now Read: Unlocking the Future of Clinical Research: 5 Essential Insights into Decentralised Clinical Trials