Every trial-cost deck has a few numbers that everybody knows. An amendment costs half a million. Monitoring eats a quarter of the budget. Eight in ten trials miss their enrollment timelines. These numbers travel from slide to slide and into budget assumptions and investor updates. Now they also travel into AI answers, which repeat whatever is repeated most.
We traced five of the most quoted ones back to where they come from. One is ten years old and gets misread. The other four have no source we could find, or none with a published method. For each of them you get what the primary data actually says and the number you can safely use instead.
The five at a glance
Widely quoted and what the primary data says | Use instead |
|---|---|
An amendment costs about $535,000 A 2016 Phase III median of direct costs. The 2024 follow-up could not collect any cost data | No single price. Budget each amendment as a recalculation |
Monitoring is 25–30% of the budget The only public split shows 14.3% site monitoring and 3.5% SDV of itemized Phase III cost | About 14% of itemized Phase III cost (ERG for ASPE, 2014) |
80% of trials miss their enrollment timelines No traceable primary source | Most trials reach their target but take nearly twice as long (Tufts CSDD, 2013) |
Investigator grants are 48–50% of trial cost, or $6,900 per patient A 2015 article without a source. No independent source puts a share on grants | Your own per-patient cost, built from the visit schedule |
The average budget negotiation takes 109 days Vendor marketing without a published method | Independent site surveys, quoted with sample size and date |
Why a wrong number costs more than it looks
A number in a budget is an assumption. If the assumption is wrong, the plan is wrong in a way nobody sees until the variance report. By then the money is committed.
Wrong numbers also get harder to kill over time. Each copy makes them look more established, and AI answers favor what appears most often. The fix is simple and a little tedious: go back to the primary source, read what it measured and quote it with its year.
1. “An amendment costs about $535,000”
Where it comes from. A Tufts CSDD study published in 2016 looked at 52 protocols. It found a median direct cost of $535,000 to implement a substantial amendment in a Phase III protocol and $141,000 in Phase II (Getz et al., 2016). The number is real. But it is ten years old, it is a median, it covers direct costs only and it describes Phase III. It often gets multiplied by today’s amendment rates as if both came from the same study.
What the newer data says. The 2024 follow-up tried to collect cost data again (Getz et al., 2024). Almost nobody could supply direct costs (a 1% completion rate) and no company could supply indirect costs. So the authors published no cost benchmark at all.
What they could measure is the ripple. Amended protocols ran 13.9% over their planned duration, against 3.8% for protocols without amendments. They had 31% more screen failures than planned and 20.8% fewer completers. Implementing one amendment took about 260 days on average, and sites could be working on different protocol versions for 215 days. And 77% of the amendments were judged unavoidable.
The number to use. Don’t put a single price on an amendment. Treat each one as a recalculation of the budget: visit and procedure fees at every site, the CRO change order, pass-through costs such as labs, imaging and patient travel, and the timeline that drives your monthly forecast. If you need a reference point for the direct cost, quote the 2016 median with its year and phase.
2. “Monitoring is 25–30% of the trial budget”
Where it comes from. It shows up in decks and in pitches for AI monitoring tools. We could not find a primary source with a line-item split behind it.
What the data says. The only public line-item split we know of is an analysis by Eastern Research Group for the US Department of Health and Human Services (ERG for ASPE, 2014). For US Phase III trials it puts site monitoring at 14.3% and source data verification at 3.5% of itemized cost. Those shares leave out site overhead and other costs. Against the full per-study cost of about $19.9 million, site monitoring comes to about 8%.
The number to use. “Site monitoring is about 14% of itemized Phase III cost (ERG for ASPE, 2014).” Say the year: the data is old, but it is the only public split. This matters most when you check an AI business case. Savings are usually calculated as a share of monitoring cost. Start from 25–30% and you roughly double the saving before the tool has done anything.
3. “80% of trials miss their enrollment timelines”
Where it comes from. It is everywhere: vendor blogs, conference slides and articles that argue about why trials run late. Follow the citations and they lead to secondary sources, such as reviews that summarize older literature, or to nothing at all. Even a thoughtful vendor article from April 2026, which argues that the real problem is forecasting, carries the figure in its page description without a source (ProofPilot).
What the data says. The traceable figures are older and more useful. Tufts CSDD analyzed more than 150 studies with nearly 16,000 sites (Tufts CSDD via Fierce Biotech, 2013). Nine in ten trials reached their enrollment goal, but they needed nearly twice the planned time. 37% of sites under-enrolled and 11% did not enroll a single patient.
The number to use. “Most trials reach their enrollment target but take nearly twice as long as planned (Tufts CSDD, 2013).” The budget lesson is in that difference. Enrollment rarely fails outright, the timeline does. And the timeline drives site fees, monitoring visits and the CRO’s fixed costs. So forecast enrollment per site instead of with one average rate.
4. “Investigator grants are 48–50% of trial cost, or $6,900 per patient”
Where it comes from. The earliest version we found is a 2015 article by a procurement research firm, which gives no source (Beroe, 2015). It has been copied many times since. For the $6,900 we found no origin at all.
What the data says. We found no independent source that puts a share on investigator grants. Public cost studies split trial costs by other categories, such as procedures, staff and monitoring. What the data does show is how much per-patient costs vary. Tufts CSDD analyzed 223 protocols and found Phase II budgets of $246,007 per enrolled patient in oncology, against $119,035 in other therapeutic areas (Tufts CSDD, 2022). That is twice the cost within one phase, before country differences even start.
The number to use. Your own. Build the per-patient cost from the schedule of assessments and local fee benchmarks, country by country. Then you know which visit drives the cost, and that is the part you can actually negotiate. If you need a public reference point, quote the Tufts budgets per enrolled patient with phase, therapeutic area and year.
5. “The average budget negotiation takes 109 days”
Where it comes from. Vendor marketing. We found no published method or sample behind it.
What the data says. Independent site surveys measure start-up time and ask what causes the delays. The most recent one is WCG’s survey of more than 400 sites, published on 4 November 2026, and it names budget negotiation as the top cause (WCG, 2026).
The number to use. Quote a survey with its sample size and date. Or measure your own clock: log every handoff with date and owner, and you will know how many days are negotiation and how many are waiting.
One more to caveat, not correct
The SCRS 2026 Site Landscape figures that circulate on LinkedIn are real, for example “69% of sites absorb costs sponsors should cover”. They come from conference slides with about 126 respondents, 60% of them in North America. Cite them with the sample size. Where a larger survey says the same thing, cite that one instead.
How to check a number in five minutes
Five questions catch most bad numbers:
Is there a primary source? Follow the citations until you reach a study, a filing or a dataset with a method. If the trail ends at a blog or a slide, stop there.
How old is it? A 2014 number can still be the best available. Say the year.
What exactly was measured? Median or mean, direct or total cost, a share of itemized cost or of the full budget.
Who, and how many? Phase, region, therapeutic area and the number of trials or respondents. A Phase III median does not describe a Phase I study.
Who paid for it? Vendor and consultancy data can be good. Read the method and keep the incentive in mind, ours included.
Numbers like these end up as assumptions in your clinical trial budget. In Trialytix those assumptions are explicit parameters such as visits, procedures, sites and timelines. So everyone can see which input drives which line, and you change it once when better data arrives. See how budgeting works in Trialytix
Sources
Getz et al., The Impact of Protocol Amendments on Clinical Trial Performance and Cost, TIRS 2016
Tufts CSDD Impact Report on enrollment performance, Jan 2013, via Fierce Biotech
Beroe, Investigator payments: channels, cycle time and benchmarks, Mar 2015
Tufts CSDD Impact Report on clinical trial budgets, Mar/Apr 2022
SCRS 2026 Site Landscape, Site Solutions Summit slides (about 126 respondents)
