Why AI-powered trial matching alone will not fix oncology trial recruitment

^{Author: Emily Wilbrand, MSc, Chief of Staff Reesi & Christoph Hillen, MD, Co-Founder Reesi
The content of this article was provided by Reesi and reflects the views of the authors. ESMO does not endorse the content of this publication and does not take any responsibility for the accuracy, completeness or reliability of the information contained in such article.}

Matching the right cancer patient to the right trial has long been a challenge in oncology and has become increasingly difficult for physicians and research teams over time (1). Contemporary oncology trials frequently contain highly detailed eligibility criteria that require extensive manual protocol review, making eligibility screening both time-consuming and costly for clinical teams (2,3).

AI-powered clinical trial matching platforms have emerged to address this challenge. By aggregating trial information from public registries such as ClinicalTrials.gov and the EU Clinical Trials Register and applying natural language processing, these systems transform unstructured eligibility criteria into structured, searchable variables. In parallel, relevant clinical patient data can be structured and compared against these criteria to identify potentially eligible trials (4). This enables clinicians to identify potentially relevant studies far more efficiently than manual protocol review. Evidence suggests that such tools can reduce screening workload and may improve recruitment outcomes. In hepatocellular carcinoma, for example, an AI-based system reduced screening time from ~150 hours to two hours (5). Similarly, integration of an AI-based trial matching platform into tumor board preparation at a breast cancer center in Germany doubled trial inclusion rates from 16% to 33% (6).

However, better matching alone does not automatically translate into higher enrollment. In practice, recruitment often fails across layers that algorithms alone cannot address. First, trial discovery remains limited when systems match only against trials available at a local site. Second, public trial registries do not always reflect the current operational recruitment status at site level. While updates are required when recruitment status changes, publication delays of several weeks are common, and registries often lack detailed information on cohort availability or site-specific enrollment (7). Third, even when a relevant trial is identified, referral pathways between physicians and study sites are often fragmented or operationally complex (8).

In our view, addressing these barriers requires more than improving algorithms; it requires infrastructure that connects the missing layers of the recruitment process. Broader trial discovery can be enabled through sponsor-agnostic trial search that makes studies visible beyond the local site portfolio (8). Site-level recruitment information requires continuously updated trial data, combining sponsor updates with feedback from study sites and the HCP community to reflect real recruitment status. Finally, referral routing must be simplified through trusted pathways that allow physicians to refer patients to recruiting sites with minimal operational friction (9). Platforms that integrate these elements into a shared infrastructure layer connecting sponsors, study sites, healthcare professionals, and patients are most likely to translate improved matching into actual enrollment. In that setting, AI moves from theoretical matching to real-world access, helping ensure that patients reach the right trial at the right site and time.

References

1. Spira AI et al, Modernizing clinical trial eligibility criteria: recommendations of the ASCO–Friends of Cancer Research laboratory reference ranges and testing intervals work group, Clin Cancer Res (2021) https://pmc.ncbi.nlm.nih.gov/articles/PMC8102342/
2. Garcia S et al, Thoracic oncology clinical trial eligibility criteria and requirements continue to increase in number and complexity, J Thorac Oncol (2017) https://pmc.ncbi.nlm.nih.gov/articles/PMC5610621/
3. Penberthy LT et al, Effort required in eligibility screening for clinical trials, J Oncol Pract (2012) https://pmc.ncbi.nlm.nih.gov/articles/PMC3500483/
4. Gueguen L et al, A prospective pragmatic evaluation of automatic trial matching tools in a molecular tumor board, npj Precision Oncology (2025) https://www.nature.com/articles/s41698-025-00806-y
5. Wang K et al, Evaluation of an artificial intelligence-based clinical trial matching system in Chinese patients with hepatocellular carcinoma, BMC Cancer (2024) https://doi.org/10.1186/s12885-024-11959-7
6. Puppe J et al, Impact of an AI-driven web-based study platform on patient recruitment in breast cancer clinical trials, manuscript in submission (2025)
7. Jones CW et al, Discrepancies between ClinicalTrials.gov recruitment status and actual trial status, BMJ Open (2017) https://doi.org/10.1136/bmjopen-2017-017719
8. Durden K et al, Provider motivations and barriers to cancer clinical trial screening, referral, and operations, Cancer (2023) https://doi.org/10.1002/cncr.35044
9. Afrin LB et al, Improving clinical trial accrual by streamlining the referral process, Int J Med Inform (2015) https://doi.org/10.1016/j.ijmedinf.2014.09.001