Where clinical trial cost data comes from, and what Open Payments misses
Every trial cost estimate you have seen is probably a survey, a model, or a 2014 mean. Here is what the only public per-trial disclosure captures.
Clinical trial cost numbers get quoted constantly and sourced almost never. When you chase the citations, nearly everything traces back to three places, and only one of them can be checked.
The three sources of trial cost numbers
The first family is surveys and models. The most-quoted study in the category, prepared for HHS in 2014, put average per-study costs around $4M for phase 1, $13M for phase 2 and $20M for phase 3. Those figures came from industry-submitted data, they are averages rather than medians, and they are more than a decade old. They still headline most blog posts on the subject.
The second family reconstructs budgets for pivotal trials. The best known, published in JAMA Internal Medicine, estimated a median $41,117 per patient across the trials behind 2015 and 2016 FDA approvals. Rigorous, but limited to pivotal studies and built on modeling assumptions you cannot audit from outside.
The third family is disclosure. CMS requires drug and device makers to report the research payments they make to US physicians and teaching hospitals, and those records carry ClinicalTrials.gov registry numbers. Nobody built those rules to measure trial costs. But match the payments to the registry and you get a spending figure for every covered trial, tied to an NCT number anyone can look up. That is the source behind our benchmarks: 5,413 US trials, $5.8B in disclosed payments.
How the match works
Each Open Payments research record names the trial it relates to. We join those records to ClinicalTrials.gov, sum disclosed payments per NCT, and divide by reported enrollment to get cost per enrollee 2020–2024 (January 2026 file release). Of the matched trials, 4,015 carry a drug-development phase; the rest are registries and device studies, which we keep out of the phase benchmarks. No modeling, no survey weights. If a number looks wrong, you can pull the same public files and check it.
What the floor leaves out
Disclosed payments are a floor on trial cost, and the exclusions are structural. CRO and vendor fees never touch a physician's ledger. Non-US sites are outside the rules entirely. So are drug supply, lab and imaging vendors, sponsor-side staff, and any funding from government or academic sources.
The undercount is also uneven. Areas that route money through academic medical centers and cooperative groups show a lower floor relative to true cost, which is a big part of why oncology, the area with the most disclosed money overall, posts below-norm per-patient figures at every phase. We cover that in oncology trial cost.
How much of the cost the floor captures
Putting our medians against the closest independent estimates makes the coverage explicit. An academic benchmark built from 10,000+ US trials puts median full study costs at $984K for phase 1, $1.46M for phase 2 and $2.52M for phase 3. Our disclosed floors for the same phases: $455K, $293K and $463K. Roughly half the estimated cost at phase 1, falling to under a fifth at phase 3.
That falling curve tells you something on its own. Early-phase spending concentrates in physicians and sites, exactly what disclosure rules capture. Late-phase spending shifts toward CROs, logistics and global operations, exactly what they miss. The gap between the floor and the estimates is the undisclosed layer of the trial economy, quantified by phase.
When a floor beats an estimate
An estimate answers "what will this cost" better than a floor does. A floor wins everywhere defensibility matters. It is current where the survey literature is a decade old. It is per-trial where estimates are averages. And it is auditable: every figure resolves to public records with an NCT number attached, which changes the conversation when a budget or a bid gets challenged.
One caution. Because the rules capture payments and not budgets, some trials match to only a few records, and the bottom percentiles sink accordingly. Benchmark from medians and upper percentiles, and read the extreme low end as reporting noise. The full benchmarks sit in what a clinical trial costs and the per-patient tables in cost per patient by phase.
Common questions
What is CMS Open Payments? A federal transparency program requiring drug and device makers to report payments to US physicians and teaching hospitals, including research payments. The data is public and republished annually.
Does Open Payments show clinical trial costs? It shows the disclosed research payments tied to each trial, which works as a verifiable floor on cost. It does not show CRO fees, non-US spending or academic funding, so it understates full budgets by design.
How current is the data? These figures cover Open Payments program years 2020 through 2024. CMS republishes the data annually.
Data notes
Sources: CMS Open Payments research payment records (openpaymentsdata.cms.gov) matched to ClinicalTrials.gov by NCT number. Coverage: 5,413 US trials, $5.8B disclosed, 4,015 with a drug-development phase 2020–2024 (January 2026 file release). Comparison estimates: Sertkaya et al. for HHS (2014 averages); Moore et al., JAMA Internal Medicine (pivotal per-patient median); clinicaltrialrisk.org (per-study medians). All DealHub figures are medians. Every figure is a floor on true cost.
This article is a market-data benchmarking resource derived from publicly available CMS Open Payments and ClinicalTrials.gov data. Reported figures are disclosed industry research payments and represent a floor on trial cost, not a complete budget, a price quote, or financial advice. Trial budgeting decisions for any specific study remain the judgment of qualified professionals. Healthcare DealHub is not affiliated with CMS or the US government.