2026 Clinical Trial Enrollment Benchmarks: Site Enrollment by Therapeutic Area

Every feasibility questionnaire asks a site to commit to a number. Very few come with a reference point for what sites in that therapeutic area actually enroll. This report is that reference point.

We analyzed 10,543 industry-sponsored Phase 2 and Phase 3 trials with at least one US site and a primary completion date between 2015 and 2024, using the public ClinicalTrials.gov registry. Below are site enrollment benchmarks by therapeutic area: sites per trial, participants enrolled per site, how long trials run, and how often they stop early. The full method and its limits are at the end, so any number here can be checked.

How to Read These Benchmarks

These are planning numbers, not a scorecard. A median tells a site what a typical trial in its area looked like, which is exactly what is missing when a per-site target arrives in a feasibility packet with no context around it.

“Participants per site” is each trial's actual enrollment divided by every site location listed on its registry record, US and ex-US. Every listed site counts, including sites that activated late or enrolled no one. That makes it a trial-level average, and it is the number a sponsor is implicitly working from when it divides a study-wide target across a site network.

The “middle half” column shows the range covering the middle 50% of trials (25th to 75th percentile). If a target asks for more than the top of that range, it is asking for an upper-quartile result, and it is fair to ask what support comes with it.

The Headline Finding: More Sites, Fewer Participants Each

Across the decade, the typical trial spread across more sites, and the typical site enrolled fewer participants.

Sites and participants per site are medians across completed trials. Terminated is the share of all trials in the period. The 2019 to 2022 periods overlap the COVID-19 pandemic, and the most recent period may shift as sponsors update their registry records.

The way we read this: per-site enrollment falling while site counts rise is a structural pattern, not a statement about how well sites recruit. When the same study-wide target is spread across more locations, each site's share of the patient pool gets smaller, and more sites are drawing on overlapping local populations at the same time.

For a site, the practical point is simple. A per-site target modeled on enrollment patterns from five or ten years ago may be asking for more than the typical site in a current trial delivers. That is worth raising in the feasibility conversation, before the number is signed.

Site Enrollment Benchmarks by Therapeutic Area

Sorted by termination rate, highest first. A trial can appear in more than one area (a diabetic kidney disease trial counts under both endocrine and urogenital), so the rows do not add up to the total.

Trials, Terminated, and Terminated (enrollment cited) use all 10,543 trials. Sites, participants per site, the middle-half range, and months use the 8,148 completed trials. Scroll the table sideways on mobile.

What the Spread Tells You

Phase 2 vs. Phase 3

Phase 2 trials in the dataset run on far fewer sites (a median of 16 vs. 53), terminate more often (25.7% vs. 18.3%), and cite enrollment as the reason for stopping nearly twice as often (3.2% vs. 1.7%).

That lines up with published research. A study of 2,579 Phase 2 and 3 trials registered as closed in 2011 found that 19% either terminated for failed accrual or completed with less than 85% of expected enrollment, and that earlier phase (23% vs. 16%) and fewer research sites were both associated with unsuccessful accrual (Carlisle et al., Clinical Trials, 2015).

Why Trials Stop Early, and What the Registry Does Not Say

Of the 10,543 trials, 2,373 (22.5%) were terminated. Only 2.5% of all trials, about one in nine terminations, explicitly named enrollment or accrual as the reason in the registry's free-text “why stopped” field.

Treat that 2.5% as a floor. About 37% of terminations use sponsor, business, strategic, portfolio, or funding language, and 180 left the field blank. A sponsor decision can follow slow enrollment without saying so.

The broader research points the same way. A National Library of Medicine analysis of 905 terminated trials in the ClinicalTrials.gov results database found that insufficient accrual was the leading reason among trials stopped for reasons other than their own data: 350 of 619, or 57% (Williams et al., PLOS ONE, 2015). That analysis covered all sponsor types, not only industry.

At the site level, the most cited reference is the Tufts Center for the Study of Drug Development (CSDD): 11% of sites in a given trial enroll no patients, 37% under-enroll, 39% meet their target, and 13% exceed it (Tufts CSDD Impact Report, January/February 2013). We read that distribution as a shared industry reality with a systemic cause. Per-site targets are typically set before anyone has modeled a specific site's local patient population, so a share of capable sites will always inherit a number their catchment cannot support.

Putting the Benchmarks to Work in a Feasibility Conversation

The strongest sites we work with use benchmarks to protect their reputation with sponsors, not to argue down every target. Three checks make that practical:

For sponsors and CROs, the same benchmarks support targets that hold. A per-site number that matches the area's typical yield, with recruitment support planned in from the start, means fewer rescue sites added mid-study and a network that keeps its standing.

Cost per Randomized Patient: Our Published Case Studies

The registry does not record recruitment cost. Our own published case studies add that layer for the indications they cover. Clinical Enroll has delivered randomization across 30+ indications. The seven below are our published case studies, with full documentation, averaging $1,379 per randomized patient.

Published case study results. The Vertex figure reflects randomizations confirmed to date against investment already spent, with screening ongoing.

One operational benchmark from our own client portal data, across 30 indications: a referral contacted within the first 10 minutes of applying is 4x more likely to answer the screening call than one contacted a day later. Of everything in this report, response speed is the variable a site controls most directly.

Methodology