Clinical Trial Matching: Finding the Relevant Trial, Not Just a Trial
Trial search rarely fails by returning nothing. It fails by returning a long undifferentiated list in which relevance and eligibility have not been assessed.
12 May 2026 · 6 min read · Written for general
Public trial registries are searchable and comprehensive. Searching a cancer type returns results — often a great many. This looks like a solved problem and is not.
The failure mode is not empty results. It is a list of forty trials in which the reader has no way to tell which two are worth pursuing.
What retrieval leaves undone
Scientific relevance. A trial mentioning a gene is not necessarily a trial relevant to an alteration in that gene. Different alterations in the same gene can have opposite functional consequences, and a trial designed around one may specifically exclude the other.
Eligibility, in practice. Criteria are written in prose, run to dozens of items, and include prior-therapy constraints, organ function thresholds, washout periods and disease-measurability requirements. Assessing them against a real patient record is work that retrieval does not do.
Practical feasibility. A trial that is scientifically ideal and requires weekly attendance three thousand kilometres away is not, for most patients, an option. Recruitment status in a registry is also frequently out of date.
Comparison. Three plausible trials raise the question of which to pursue first — a question about scientific rationale, burden and what each precludes. Registries do not rank.
What better matching involves
Representing the patient properly: not "lung cancer" but the histology, the specific alteration with its exact designation, the treatment sequence with responses, current disease status and measurability, organ function and performance status.
Representing the trial properly: eligibility criteria in a form that can be reasoned over rather than only read, including what the trial is scientifically testing and not merely which gene it names.
Then matching on scientific rationale, screening against the criteria that most commonly exclude, filtering by geography and burden, and ranking what remains with the reasoning attached.
The honest limits
Two things must be said plainly. Eligibility is determined by the trial investigators, not by any external assessment — a matching exercise identifies plausible candidates and nothing more. And identifying a trial is not obtaining a place on one; slots are finite, screening can exclude, and recruitment closes.
Matching that overstates either of these does the patient a disservice.
Why it matters
For a patient who has exhausted standard options, a trial may be the most substantive remaining avenue. The difference between a list of forty and a reasoned shortlist of three is the difference between information and a task.
Relevance is the whole problem. Retrieval was never the hard part.