Analysis · Public data

Trials terminated for low accrual: what the stated reasons actually say

Every stopped study on the registry has to say why. Counted across 43,563 interventional records, one reason dominates — and it is not safety, funding or futility.

Rock Enroll editorial · ClinicalTrials.gov, retrieved September 2026

Sponsors talk about enrollment failure in anecdotes. The registry lets you count it. When a study is stopped early, the record carries a Why Stopped field — free text, written by the sponsor, no standard vocabulary. It is messy, but it is contemporaneous and it is public.

We pulled every study on ClinicalTrials.gov with a status of Terminated or Withdrawn, restricted to interventional studies, and coded the stop reason with keyword patterns for accrual, funding, business/strategic decisions, safety, futility/efficacy, investigator and site problems, COVID-19, regulatory issues and supply. That is 29,768 terminated and 13,795 withdrawn records. Categories overlap — a record can say “slow accrual and funding exhausted” — so the percentages sum past 100.

Terminated studies: why sponsors stopped

Accrual is cited more than safety, futility and funding combined. That is the finding, and it is not close. A third of all early terminations in the registry are sponsors saying, in their own words, that they could not find enough patients.

Withdrawn studies: stopped before a single patient

Withdrawn means the study was registered and then never enrolled anyone. This is the category people rarely look at, and it is the more damning one: 23.4% of withdrawals name accrual, meaning the sponsor concluded before starting — or immediately after activating — that the patients were not there.

Where accrual failure concentrates

By phase

Phase 1 is the least accrual-driven, which makes sense: small cohorts, healthy volunteers in many cases, and a paid-participant model that behaves like a market. Phase 4 is the worst. Post-marketing studies have real budgets, real regulatory obligation and almost no urgency attached to them at site level — and the drug is already available, which removes the one incentive that makes a patient tolerate a trial protocol.

By sponsor type

Industry-sponsored terminations cite accrual at less than half the academic rate. Read that carefully before concluding industry recruits better. Industry studies have alternative exits an academic study does not: a portfolio decision, a strategic reprioritisation, an interim futility read. When a pharma sponsor stops a study that was also enrolling badly, the field very often says “business decision.” The accrual number for industry is a floor, not an estimate.

What this does and does not prove

It does not tell you what share of all trials miss their enrollment targets. Most under-enrolling studies do not terminate; they extend, add sites, cut the target, and are eventually marked completed. The commonly quoted 60–80% figure is measuring that broader population and is highly definition-dependent. What the registry proves is narrower and harder to argue with: among trials that die, the most common stated cause of death is that nobody could find the patients.

It also does not distinguish between the two very different failures buried inside “slow accrual”: a protocol whose eligible population barely exists, and a protocol with plenty of eligible patients that never reached them. Those need opposite remedies, and the registry field collapses them into one phrase. Separating them is the first step of any rescue — the four-stage funnel diagnosis sets out how.

And it says nothing about the method those studies were using to recruit, because the registry does not record it. Our argument about what that method usually is, and why applying AI to it does not help, is in why AI EHR parsing won't fix an under-enrolling study.