Clinical trial feasibility, explained.

Clinical trial feasibility explained: deciding where a study will find its patients

One Crohn's study, from protocol synopsis to a signed site list

1. The protocol sets who may joinVLT-401 needs 480 adults with moderate-to-severe Crohn's disease who have failed an advanced therapy. The target for the last patient is June 2028.
2. Countries and sites are askedQuestionnaires go to 186 sites in 14 countries on September 12. They close on October 2 with 131 back.
3. Claims are checked and everything is ranked90 questionnaires are usable. Each is compared with site history, registries and patient counts, and ranked. Most of this guide lives at this step.
4. Allocation is argued and approvedAllocation means how many participants each country gets. Should Brazil, strong but slow to approve, move from back-up to core? The study lead wants to see what it does to the date.
5. The list is signed and filedA named person signs. The reasons go to the trial master file, the study's official record, and site start-up begins.

The people, sites and figures in this example are fictional.

Why a site that never enrolls is common, and three more feasibility surprises

Questionnaire, zero-enrolling site, last patient in and P50

Fourteen terms are defined in the one-page summary.

Country feasibility, site feasibility and the enrollment forecast answer three separate questions

In shortStudy feasibility is three decisions that lean on each other. Country feasibility asks where the patients are and how fast the study can get approved. Site feasibility asks which investigators can recruit and run the study well. The forecast asks when it will end. Change one and the other two move.

One pass through all this, from questionnaires to a list, is called a feasibility round, and it shrinks a long list quickly. In the guide's example, a little under half of the sites asked end up scored at all. Nothing is published on how typical that is, because the numbers belong to sponsors, so measure your own.

No single source answers every site selection question

Good feasibility puts sources side by side because each is strong on some questions and blind on others. Site history shows what happened, though only on your own studies. Questionnaires hold the only view of what changed last month. Registries show competition, up to their reporting lag.

Feasibility, site qualification and study start-up come in that order

Three neighboring activities are often confused. From site identification to initiation, the whole process took about 31 weeks in a Tufts follow-up study, with no improvement on figures from almost ten years earlier.7

ICH E6(R3) investigator selection and 21 CFR 312.53: what sponsors must show, and since when

In shortNo rule gives a scoring formula. The rules make the sponsor responsible for selecting qualified investigators, ask each investigator to demonstrate recruiting potential and resources, and require records that can be verified later. ICH E6(R3) took effect in the EU on July 23, 2025, and FDA published it as final guidance in September 2025.

What the guideline leaves to your own site selection criteria

Read E6(R3) looking for a list of site selection criteria and you will not find one: no required weighting, no minimum enrollment rate, no limit on competing studies. A rare-disease study and a primary-care study could not share one rule. The guideline fixes the questions and leaves the answers to the sponsor, with the obligation to keep what it relied on.

The guideline also sets a courtesy that fast feasibility rounds tend to strain: section 3.7.2 asks the sponsor to give potential investigators the protocol, an up-to-date Investigator's Brochure and "sufficient time" to review them.3

How long clinical trial approval takes in the US, EU, UK, Canada, Japan, South Korea and Brazil

In shortA country with many patients can still arrive late. Regulators in the US and Canada default to about 30 days, the EU can take 60 to 106, and Brazil's law allows its regulator 90 business days. These are legal clocks. A site still needs its contract and budget, and in most of these countries its ethics review, before it can recruit, and that whole path averaged about 31 weeks in a Tufts survey.7

Geography is shifting under every feasibility plan

The European Commission says the European Economic Area's share of global clinical trials fell from 18% in 2013 to 9% in 2023, while China's rose from under 10% to nearly 30%.32 Another count, "over the past decade", gives the European Economic Area 22% falling to 12% and China doubling to 18%.33 The counts differ, and the direction agrees. Old habits about which countries are "standard" deserve a fresh look.

Protocol criteria, site history, questionnaires, registries and patient counts: what each one proves

In shortA defensible site list rests on five kinds of evidence. Each has a known trap: criteria that get paraphrased, site history that covers only your own studies, questionnaires that flatter, registries that lag, and patient counts that describe a population and say little about who will walk in the door.

Longer entry criteria shrink every country's patient pool at once

Eligibility is the input every other number inherits. In 1,197 National Cancer Institute (NCI)-affiliated Phase II and III trials started from 2008 to 2018, trials with the shortest criteria failed on accrual 11.8% of the time, and those with the longest failed 29.4% of the time. Median criteria grew from 214 unique content words in 2008 to 417 in 2018.4

These are cancer trials, so do not carry the rate across. The direction is useful anywhere. At one specialist referral center in another study, only 31.1% of patients with inflammatory bowel disease would have qualified for any of eight major trials.36

Why a country's patient funnel beats one big number

A country's eligible pool should read as a funnel, so that a reader can see where the patients go. In the fictional example, Brazil starts with 98,900 people living with Crohn's disease and ends with 11,500 a year who meet the endoscopic score. Each row is a criterion from the protocol, applied in protocol order.

The eligible pool is a ceiling. Few of these people will hear about the study, live near a site, agree and pass screening.

Four feasibility calculations worked by hand: claim check, site score, site rate and forecast range

In shortFour small calculations carry most of a feasibility decision. Compare a site's claim with its record. Combine weighted factors into a score, and keep the weights visible. Turn history into a monthly rate. Add up sites from the date each one opens, and report a range. All numbers in this section come from the fictional example.

1. Checking a questionnaire claim against what the site screened

The Duisburg site claims about 80 eligible patients a year and screened 14 in a year on an earlier study. One planning rule counts a claim at no more than twice history until the site confirms it, here 2 × 14 = 28. The cap is a sponsor's choice, and no regulation sets it. It stops an unchecked number driving a plan. Where a site has no history, the same rule set counts the claim at 60% and scores the site as neutral.

2. Building a weighted site score that a reader can argue with

A score turns six judgments into one number. The weights below are one sponsor's defaults: past enrollment 30, patient pool 20, start-up speed 15, competition 15, questionnaire quality 10, representation 10. Representation means how well a site reaches the communities the study needs. Each factor is scored from 0 to 100, multiplied by its weight, and added.

Notice the last row. A score of 85 hides a representation factor of 9, and a study team with a commitment to reach underrepresented groups would want to see it. That is the argument for showing factors beside any total.

3. Turning a site's history into a monthly rate

Compare with plan too. The Belo Horizonte site enrolled 31 against a plan of 12, and the Duisburg site enrolled 3 against a plan of 10. A careful forecast then discounts the rate for competition around the site, which is why a planning rate usually reads lower than the raw history.

4. Adding sites up from the date each one opens

A forecast then repeats this many times with rates and opening dates varied, and reads off the dates by which 10%, 50% and 90% of the runs finished. The published statistical method treats patients as arriving at each site at random around that site's average (a Poisson process), with averages that differ between sites following a gamma distribution. Researchers have found that simple ranges from that model can be too narrow, that adding more sites does not fix this, and that improved ranges exist.37 Ask how any tool builds its range.

The width is the finding. A steering group told only "May 2028" does not know the study could plausibly run a quarter late. A range lets you ask what would pull the cautious end forward.

Clinical trial enrollment statistics: zero-enrolling sites, under-enrollment and where the 80% comes from

In shortThe best-known numbers are older or softer than they sound. A 2012 Tufts report has 48% of sites falling short of target, including the 11% of ready sites that enrolled nobody. The 80% figure traces to a trade article. Peer-reviewed accrual failure, about one in five, comes from publicly funded cancer trials and measures a harder failure than lateness.

We found no public figure for how often sites overstate patient numbers, how far they overstate them, or how many questionnaires come back unusable. Your own last three studies can answer all three.

What competing trials do to site enrollment: Crohn's disease at 0.1 patients per site per month

In shortIn a crowded disease the rate decides everything. Published Crohn's recruitment fell to about 0.1 patients per site per month by 2018, while active inflammatory bowel disease trials more than doubled. At that rate a 480-participant study on 64 sites would take about six years. Our fictional example runs much faster than that, and you should read it for the method.

More trials, fewer willing patients per trial. New approved drugs reduce the appeal of a study with a placebo arm, and colonoscopies and frequent visits add burden.36 The same paper reports that most US sites enroll only one or two patients per trial in these diseases.36 One CRO's review of 34 Phase II and III ulcerative colitis studies found a median of 94 sites, some with as many active sites as target patients, and a median ratio of one site to 2.4 patients.38

The sum a Crohn's feasibility lead should run first

Why a competing study has to be the same population to count

Six Crohn's studies recruiting in a country sounds like crowding. If four of them take children, or people who have never had an advanced therapy, or people after surgery, they may not touch your patients at all.

Good competition analysis matches condition, phase, line of therapy (how many treatments a patient has already tried) and age, and counts only same-population studies against a site, within a set distance. It also checks the site's own answer, because a questionnaire can name fewer rivals than the registry shows: in the example at the top of this page, two named against four found.

Why feasibility goes wrong: optimistic claims, stale registries, lost reasons and protocols that move

In shortMost failures are ordinary. A claim goes unchecked, a registry check is months old, a rate gets averaged across sites, the protocol changes and nobody re-runs the numbers, or a site gets chosen for its famous investigator. The most expensive failure comes later, when nobody can say why a decision was made.

Protocol amendments arrive after the site list is built

In 2016 Tufts found nearly 60% of protocols had a substantial amendment, a formal change to the study plan, and sponsors judged nearly half of those avoidable. Median direct cost per substantial amendment was $141,000 in Phase II and $535,000 in Phase III.40 Amended protocols also ran about three months longer and screened and enrolled substantially fewer patients against plan.40 A 2024 Tufts study updates the picture, and in it 77% of amendments were deemed unavoidable.41

What an inspector or auditor can reasonably ask about site selection

We found no public statistic isolating site selection in inspection findings. FDA publishes bioresearch monitoring metrics in general, and a selection decision is not one of its categories. The clauses still suggest the questions. Who selected this investigator, under what agreement? On what data was recruiting potential demonstrated? Was the claim checked? Can the record behind the decision still be reached?3 A list without reasons answers only the first.

The cost of a day of clinical trial delay: $55,716 of Phase III trial cost, plus lost sales if the drug launches

In shortTufts CSDD published two kinds of number in 2024. A day of Phase II or III direct trial cost is about $40,000 on average and $55,716 in Phase III. A day of lost sales is about $500,000 in the journal article and about $800,000 in the later white paper, from one dataset. Keep them apart: one is money spent, the other is money that may be earned.

Two cautions travel with these. The sales figure is an average over a wide spread and has been falling by about $80,000 to $100,000 a year, and a product in a narrow indication should use its own forecast.42 And the trial-cost figure excludes a sponsor's indirect people and infrastructure, per the authors' method, so your own budget divided by your planned duration is the better input.43

Trial-cost days and lost-sales days do not add into one number

Trial cost is being spent for certain, on a study that is running. Sales are earned only if the medicine reaches the market. Weeks saved in feasibility count against either only if feasibility sits on the study's critical path, the chain of tasks that sets its start date. Often the protocol, drug supply or the first regulatory filing sets the first-patient date regardless. So any claim of savings is conditional, and a finance partner will test it first.

Feasibility software is sold three ways, and its market has no reliable public size

We found no reliable public figure for this market's size. Buyers can compare data vendors who sell site performance and registry data, CROs who run feasibility as a service, and clinical trial software suites. Most published claims about AI in feasibility come from those vendors. We found no independent test comparing tools on a sponsor's own finished studies, which is why Part C asks you to run one.

AI in clinical trial site selection: what it reads well, what regulators expect and where people stay

In shortSoftware reads forms, registries and exports quickly and tirelessly. It cannot tell you what a principal investigator's new referral clinic will do. FDA's January 2025 AI draft does not address tools for operational efficiency with no effect on patient safety, drug quality or study reliability, and it asks sponsors who are unsure to engage early. Either way, the sponsor stays responsible for the choice.

What record integrity asks of a machine-written rationale

E6(R3) wants source records attributable, legible, contemporaneous, original, accurate and complete, and wants processes that keep traceability so information can be verified.3 Source records are a trial's original data, and feasibility records are planning documents. The same habits still suit them. Each sentence in a rationale should open to what it came from, as it stood on that date. The study list should separate what software proposed from what a named person decided.

Four limits of AI feasibility tools, from favored sites to invented sentences

How feasibility teams work today: questionnaires, a CTMS export and a spreadsheet

In shortA feasibility lead sends questionnaires, pulls an export from the trial management system, checks registries by hand, buys patient counts and builds a spreadsheet. Country managers argue for their sites. A study lead approves allocation. The final list usually gets filed, and the reasoning usually stays in email and memory.

From the feasibility lead to the country manager: who shapes a site list

Six steps and five roles between a protocol synopsis and a signed site list

Spreadsheet, CRO service or feasibility software

VLT-401 in nine steps: Brazil moves to core, Rheinhafen stays a back-up, and the list is signed

In shortLena Ortiz, the feasibility lead, opens the study as a world map. She reads why Brazil is only a back-up, moves it to core, and watches last patient in move from May to March 2028. She sends that to the study lead. She answers a country manager's request with a written reason, signs the list and files it. When the protocol changes, the agents re-run everything before anything moves.

1

The plan opens as a map of 14 countries

VLT-401 is a Phase 3 study of Veltrimab in moderate-to-severe Crohn's disease: 480 participants, target last patient in June 2028. Questionnaires went out on September 12 and closed on October 2. Countries appear as bubbles sized by participants allocated, with the United States at 140 and Poland at 78 as lead countries, and France and Brazil held back. Other measures swap in at the top: score, eligible patients, competing trials, start-up weeks.

2

Why Brazil is only a back-up, with each reason's source

Lena clicks Brazil. It ranks 13th of 14 with a score of 60. Five factors appear, each with a value and a link to its source: past enrollment of 1.70 a site a month from site history, 11,500 eligible a year from real-world counts, one same-population trial of six recruiting, a 27-week start-up with 150 days of approval, and four questionnaires with consistent claims. Beside them sits the forecast: most likely May 2028, 7 weeks before target, with a range of February to September 2028 on 64 primary sites.

3

She moves Brazil to core, and the forecast follows

The switch reads Lead, Core, Back-up. Lena picks Core. Brazil's three ready sites open as primary and take 36 participants: 20 from the United States, 10 from Canada and 6 from Italy. The forecast is recalculated from each site's own history. Most likely last patient in becomes March 2028, two months earlier, with a range of January to July on 67 primary sites. Nothing has happened yet. It is a proposal, and it can be undone.

4

The change goes to the study lead with its evidence

One click sends it to Sam Patel, who owns the timeline and approves country allocation. It lands on the audit trail. Country manager Elena Duarte can point to Brazil's whole case: the patient funnel, the single same-population competitor among six Crohn's studies recruiting, the approval path and the sites.

5

Ninety sites, ranked on what they did

Every site with a usable questionnaire is scored on six weighted factors. Each row shows the rate a month, start-up weeks, competing trials within 25 km and the claim against history. The filter along the top finds 23 sites claiming more than three times what they screened, each counted at twice its history until it confirms. Move a weight and the agents re-propose inside each country's allocation. Decisions made by people stay.

6

A country manager pushes for a site, and the reason answers him

Tom Becker asks to make Universitätsklinikum Rheinhafen, the Duisburg hospital from the quick example, a primary site. The rationale says why it ranks 85th of 90. In VLT-208 it enrolled 3 against a plan of 10 in 12 months. Start-up took 29 weeks. Its questionnaire claims about 80 eligible a year where it screened 14. Four same-population trials recruit within 25 km, one at the hospital itself, and the questionnaire names only two. Each numbered citation opens its source in the panel on the right.

7

The site's own words, highlighted where the claim came from

Lena opens citation 3. The questionnaire appears at question 7 in the site's own words. Other tabs hold its history, its four competing trials with distances, and its investigator. She keeps Rheinhafen as a back-up that opens if a German primary site falls behind after month 6. Tom reads the reason and signs Germany, the UK and France.

8

She signs the list with the meaning of her signature

The site list shows each approval: the agents' proposal, five of five country managers, the study lead's allocation. Lena signs VLT-401 site list v3 with her password and chooses the meaning: "I approve this site list and the rationale for each site" or "I have reviewed this site list". The list, the allocation, the P10 to P90 forecast and the approval record go to the trial master file under site selection, with a rationale for each of the 76 primary and back-up sites.

9

When the protocol changes, everything re-runs before anything moves

Amendment 1 lowers the endoscopic entry score, the Simple Endoscopic Score for Crohn's Disease (SES-CD), to 4 or more (3 or more in isolated ileal disease) and allows up to four prior advanced therapies. In 3 minutes 12 seconds on this sample, the agents re-read the criteria, recount the pool (up 14% overall) and re-score all 90 sites. Three proposals change, each with its reason. Lena applies or discards them. Nothing reaches the plan until she does.

Seventeen more screens, from the country table to how well past rankings predicted enrollment

In shortThe walk above is the story. These screens show the rest: a lead country's evidence, the country table, a country's sites, drafted follow-ups, weight presets, the site's history and investigator, the decision, the list and its approvals, the filing preview, the signed state, the dashboard, and your own rules.

Select any screen to open it larger. The companies, people, sites and figures are fictional. Site rates in this sample were set to keep the example easy to follow and run well above the published Crohn's average.

What the sample dashboard says about ranking and enrollment

Six agents read, count, check and rank; country managers, the study lead and the feasibility lead decide

In shortEach agent has one job and limits it does not cross. None of them decides. Country managers review their sites and can swap one with a written reason. The study lead approves country allocation and every tier change. The feasibility lead decides each site and signs the final list. Every step goes on the audit trail.

From protocol reader to rationale writer: the six feasibility agents

Country managers review, the study lead approves, the feasibility lead signs

Each signature records name, date, time and meaning. These are controls that support your compliance with 21 CFR Part 11 and EU Annex 11; no product can be certified against them. Botminds' own credentials are SOC 2 Type II audited and information security managed to ISO 27001. Roles, separation of who proposes from who approves, and a full audit trail come as standard; registries and new answers refresh every Monday, and any site that moves more than 5 places is explained.

Lena, Sam, Elena, Tom and Maya: the people in the VLT-401 screens

Protocol, site history, questionnaires and registries in; a ranked, cited list and a live forecast out

In shortBotminds reads the protocol, your own site history, each returned questionnaire, public registries and licensed patient counts. It leaves behind a ranked list where each reason is cited, a forecast with a range, a map of competing trials, and a signed list in your trial master file. Your trial management system and questionnaire portal stay where they are.

Systems a feasibility rollout connects to

The design targets behind the 11 days and the 3%

One more design target: 3% of chosen sites enrolling nobody, against 11% in the 2012 Tufts data.2 Neither target is a result from a customer. The two-month move on Brazil is the fictional example's estimate.

What a zero-enrolling site, an under-enrolling site and a slow feasibility round cost your study

In shortStart from your own budget. Published figures frame the problem: how many sites enrolled nobody, what a Phase III day costs, and how long start-up takes. The share of your sites that under-enrolled, and what each one cost to open, are numbers only your own finished studies can give you.

Counting your own zero-enrolling sites and missed claims

Investigators who stop answering, and reasons nobody can find

What better site selection returns: feasibility hours, fewer empty sites and trial days on the critical path

In shortPut in your own numbers. The calculator counts the hours a round takes and the sites you avoid opening, against what the tool costs. It shows trial days separately, because they count only if feasibility sits on your study's critical path. Every default marked illustrative is an assumption to replace.

Hours, empty sites, days and reasons: where feasibility saves

Rounds a year, sites per round and your cost of an empty site

A sponsor running one or two studies a year probably cannot justify software on hours alone. For that sponsor, a CRO service or a disciplined spreadsheet with claims checked by hand may be the right answer.

How ready your feasibility team is: studies a year, site history, rules and who signs

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

A low score means starting with a claim check by hand

The total can flatter. A 0 on owner and budget, or on quality and AI approval, holds you at a pilot whatever else you scored, because either one can stop a rollout alone.

What to ask feasibility software or a CRO about rankings, citations, claims and forecasts

In shortAsk every vendor the same questions, us included. The answers show whose data decides the ranking, how each reason traces back, how a claim gets checked, how a forecast range is built, who decides, and how you could measure all of it on your own finished studies.

We found no published independent test of any feasibility vendor's rankings, as of November 2026. A replay on your own finished study is the fairest comparison, and the same test measures your current process.

Replay a finished study first, then run beside a live feasibility round and decide on three numbers

In shortMeasure first, then give the tool a finished study with the evidence as it stood, then run alongside a live round. Decide with three numbers: how well the ranking predicted enrollment, how many claims were caught, and team hours to a signed list. None of these steps commits you to a vendor.

1. Check claims by hand on one finished round: claim divided by screened, for every site. You will learn something whichever way you go.
2. Count your baseline from the last five studies: zero-enrolling share, share under half of plan, forecast error.
3. Choose two finished studies with their protocols, returned questionnaires and history as it stood before the study started, and an approval path to share them.
4. Run alongside a live round for one full cycle, with your country managers reviewing and your lead signing as usual.
5. Decide with three numbers: enrollment by rank band, claims caught against claims missed, and team hours to a signed list.

Rank-band enrollment, claims caught and citation accuracy on your studies

Clinical trial feasibility questions answered: site selection, questionnaires, forecasts and AI

In shortThe questions clinical operations teams and buyers search for most, each answered on its own.