Site activation essential documents, explained.

Why a hospital must prove itself on paper before its first trial participant

From a chosen hospital to its first participant, in five steps

1. The site is selectedThe sponsor sends a list of what it needs from that country. At Site 207 in Bologna that was July 8.
2. The site gathers its packetThe coordinator chases CVs, certificates and signatures from busy doctors, and waits on the ethics committee and the contract office.
3. The packet is checked against itselfDoes each person match across forms? Does each approval name the current protocol? Does every date hold up against the visit? This step is the subject of this guide.
4. Fixes go back and forthA request goes to the coordinator. Corrected documents come in and get checked again. Each extra round takes time nobody planned for.
5. Green light, filing and the visitA named person signs. Documents go to the trial master file (TMF), the study's official record, the drug can ship, and the initiation visit starts the site.

The sites and dates in this example are fictional.

Four facts about the 1572 and GCP that surprise newcomers to packet review

Form FDA 1572, delegation log, initiation visit, green light: the words to know

Fourteen terms are defined in the one-page summary.

The twelve document families in a site regulatory packet, and what each must agree with

In shortThis guide groups a site packet into twelve kinds of document, some repeated for every person. Each one proves something different. The value of a review lies in the right-hand column of the table below: what each document has to match.

ICH E6(R3), the international good clinical practice guideline, calls these essential records: records that "permit and contribute to the evaluation of the conduct of a trial". It says some "should generally be in place prior to the start of the trial", and its Essential Records Table marks those with an asterisk.7

That table covers most of the families below, leaves out US forms such as the 1572, and calls itself "not an exhaustive list".7 The table here is typical sponsor practice, built from the US rule, E6(R3)'s table, the older E6(R2) list and EU law, and your own list may differ.6,1,8

21 CFR 312.53, ICH E6(R3) and the EU Clinical Trials Regulation: what a site owes before it starts, and since when

In shortThe US Code of Federal Regulations (21 CFR) names four things a sponsor must obtain before an investigator begins. The EU asks for suitability evidence in the trial application. Canada has its own undertaking and site form. E6(R3) sets shared practice, with criteria and an open-ended table of essential records. FDA's single-IRB rule, which would change how many approvals a US study has to track, is still pending.

What FDA's single-IRB rule would change in a US packet

The Common Rule, the US rule protecting people in federally funded research, requires one IRB for non-exempt cooperative research.24 FDA proposed to do the same for the trials it regulates, in a rule listed at final rule stage with a December 2026 target.15 We found no final rule on November 3, 2026, the White House list of rules under review that day did not show it, and the target is only a target.

Sites with a central board already tend to move faster, though the study that found this warned about small samples and missing data.25 For a reviewer, one board for all US sites means one approval to read, and the check moves to whether each site's local consent wording and signature page match it.

Which expiry dates in a site packet come from a regulator, and which are your own

In shortMost of the date limits that hold up a site are sponsor rules: a CV under 24 months old, GCP training within 3 years, a license valid some months past activation. They are sensible, and an inspector will read your procedure and hold you to it. Knowing which limits you wrote yourself lets you accept a two-day gap with a recorded reason where that is right, or tighten a rule where it matters more.

Form FDA 1572, section by section: who to list in Section 6 and when a new form is needed

In shortThe form's numbered sections name the investigator, the facilities, the lab, the IRB, the team and the study. FDA's test for Section 6 is whether a person makes "a direct and significant contribution to the data". That usually takes in coordinators, and it usually leaves out on-call doctors. FDA's FAQ says a new sub-investigator does not require a new form.

Sections as printed on Form FDA 1572 (4/25).9 FDA says the investigator should sign only after receiving enough information to understand the commitments.5

Section 6 by role: what FDA's FAQ says about each person on the team

When a change does and does not require a new form

Do sites outside the US sign a 1572?

Only if the sponsor runs them under the IND. FDA's FAQ says a multinational study may have US sites under the IND and foreign sites outside it, and that foreign investigators outside the IND need not sign.5 If local law prevents signing, FDA expects the site to run as a non-IND site. Data from those sites can still support an application if the study meets 21 CFR 312.120.5 A draft revision from May 2021 covers these questions and adds a section on waiving the signature; we found no final version.20

The EU suitability forms and Canada's Qualified Investigator Undertaking, side by side with the 1572

In shortEach jurisdiction asks the same underlying question in its own form. The EU builds suitability into the trial application, per Member State. Canada splits it between a form you submit and two you keep on file. A multinational packet review needs a separate list for each country, tied to the rule behind each line.

The UK runs its own rules, amended from April 28, 2026.14 National law in any country can add to this table, and many countries have local forms this guide does not list. Check each country with your regulatory team.

Language adds its own step. An ethics opinion may arrive in five languages across five countries, and a coordinator in Bologna tends to answer faster when the request is in Italian. No source we found measures that effect, so treat it as practical experience. When software drafts a request in another language, someone who reads that language should look at it before it goes.

Measuring a GCP certificate, a CLIA certificate and a delegation date against the initiation visit

In shortA date check has three parts: when the document stops counting under your rule, the date it has to survive to, and the difference. Most trackers record only the first. Worked through on the fictional study, the most dangerous break on the page is also the smallest.

The three date tests a start-up reviewer runs, with numbers

Dr. Gallo in Bologna: completed November 2, 2023, plus three years, gives November 2, 2026. The visit is November 4, 2026. The margin is minus 2 days. Run on October 7, that is 28 days of warning.

The Austin lab's CLIA certificate ends November 30, 2026. The visit is October 21, so it gives 40 days of cover against the example's 90, a shortfall of 50 days. A US certificate of compliance lasts at most two years, so a lab near renewal is routine, and a reviewer may accept this with a note and a reminder.23

Dr. Kowalczyk in Wrocław: delegated from September 1, 2026, GCP certificate dated September 15, 2026. That gives minus 14 days. Neither document shows this alone. It appears when someone reads the two together.

Lining one person up across six documents

The second kind of check runs across forms, for each person. Names are the first trap: the same person may be "Marco Gallo, MD" on a CV and "GALLO Marco" on a suitability form. A reviewer has to treat those as one person, and also has to notice "M. Gallo" with a different register number. This table shows what a clean line-up looks like, with its single date problem.

Fictional. Tasks B, D and E are eligibility assessment, physical examination and adverse event assessment on this study's log.

Ten cross-document errors in a site packet, and why a single-document review misses them

In shortNobody publishes how often each of these happens, so count your own. They share one feature: every page involved can look fine when read alone. That is why adding more careful single-page reading does little to stop them, and why they surface late.

Essential document, delegation and training findings in EMA and FDA inspection reports

In shortEssential documents is the largest sub-category in EMA's general findings for 2024. FDA names the 1572 first among its common investigator observations. EMA's inspectors also list the specific errors this guide describes: tasks outside the delegation period, and training completed after site activation.

What EMA's inspectors wrote about delegation, training and version control

Quotations from EMA's annual report for 2024.26 The packet column is this guide's reading.

Study start-up benchmarks: eight months, five to six months and 9.4 months all measure different clocks

In shortPublished start-up figures disagree mainly because each starts and stops its clock in a different place. New sites take about ten weeks longer than repeat sites. About one site in nine that is initiated never activates. The numbers are old, or come from vendors, and none isolates document review, so measure your own.

Where documents sit in the delay, and where they do not

Documents explain only part of this delay. Tufts names contract and budget negotiation and unpredictable IRB timelines as major delays, along with complex protocols and amendments.36,33 In a 2020 vendor survey of 524 clinical operations people, 68% named contracting and budgeting as a top challenge, and 44% named essential document collection.37 A faster packet review shortens the part of the wait that you control, and it stops late surprises after the hard parts are done. A stuck contract stays stuck.

Late activation also costs enrollment. Sites run in parallel, so a site activated late has less time to enroll. In a Tufts study of 151 trials, 11% of sites enrolled no one, and 48% fell short of target, a figure that includes that 11%.38

The UK shows a public clock can move

The UK set a target of getting commercial trials running within 150 days by March 2026. The Health Research Authority reported a fall from 169 days to 122.39

The government's own report gives 122 as a median, from application to first participant, for applications from April to September 2025, with 73% (53 of 73 with known status) inside 150 days against a 95% target.35 That clock includes ethics and regulatory review, so it does not tell you about packet review. It does show that a published number with a named owner can fall.

No public figure: documents per site packet, review hours per packet, findings per packet, rounds per site or days per round. Part C explains how to measure yours in a week. One secondary source reports about 82 sites and 14 countries for a typical Phase III protocol, citing Tufts 2021.40

A Phase III trial day costs $55,716 in direct spending, and how little of that one late site explains

In shortTufts puts the direct cost of a Phase III day at $55,716 and the Phase II and III mean near $40,000. Lost sales run about $500,000 or $800,000 a day, depending on which Tufts publication you read. These price a delayed trial or launch. One site's late certificate rarely delays a whole trial by itself, so use them with care.

The lost-sales figure comes from one dataset of 645 drugs, published twice with two numbers. The journal article says about $500,000 a day.42 The August 2024 white paper says about $800,000.41 State both if you use either.

Why you should not multiply a day's cost by one site's delay

Seven days of a Phase III trial at $55,716 comes to $390,012. A reader may be tempted to call that the cost of a bounced packet. It is not a fair sum. Sites start in parallel, so one site's lost week usually means some lost enrollment at that site, and it moves the whole study only when that site was the last one needed, or when enough sites slip together. Part C gives a formula that makes the share explicit.

The costs nearer a start-up team's own work are smaller and easier to prove. A US government study in 2014 put site recruitment at $395,182 per Phase 3 study, 3.47% of costs, and IRB approvals at $114,118, or 1.00%.43 A canceled monitoring trip, a rebooked flight and a second visit are real line items you can look up in your own expense system.

What AI can read in a site packet, what regulators expect of it, and what stays with a person

In shortReading names and dates, lining up people, and measuring dates against a visit are good work for software, provided each result points to its page. Confirming a finding, sending a request and signing a green light stay with people. We found no regulator's rule written for AI in this work. General rules on electronic records and trial systems still apply.

Reading dates, lining up people, deciding readiness: what moves to software

What a packet check cannot tell you

A cross-check can tell you the papers agree. It cannot tell you the papers are true. A delegation log may say one thing while the clinic does another, and EMA lists that as a finding.26 A CV says nothing about whether a doctor has time for the trial. A forged or wrongly issued certificate can read perfectly. A scan with faint handwriting can defeat any reader, so a good system says "unreadable" and hands the field to a person.

Monitoring visits, site qualification and professional judgment all remain necessary.

Start-up specialist, CRA, coordinator, TMF owner: who handles a site packet today, and where the hours go

In shortA specialist opens each PDF, types dates into a tracker and compares names by eye. A CRA chases the site. Spreadsheets led the tools in one 2020 survey. No time study has been published, so the hours below are an estimate for you to replace.

The people between a chosen site and its green light

Three ways sponsors and CROs run packet review today

From a broken bar on the activation runway to one request in Italian: October 7 at Site 207

In shortThis is the example from the top of the page, inside the product. Lena Ortiz, the study start-up lead, opens the activation runway, a shared calendar where each site is a bar to its initiation visit. Bologna's bar breaks at November 2, two days short. She opens the break, reads the evidence, and sends one request, in English and Italian, due October 14.

1

The activation runway, with Today and First patient in marked

Twelve sites in seven countries share one calendar. Each bar runs toward a black diamond, the site's initiation visit. A red marker means expired or lapsing, purple means inconsistent, amber means missing, and an envelope means requested with a due date.

On the right, breaks are listed nearest visit first. Above them, a new packet from Austin waits: 23 documents through the site portal at 7:40 am.

2

Dr. Gallo's GCP certificate, lined up with his five other documents

Selecting the red marker opens the evidence. Certificate valid to November 2, initiation visit November 4, 2 days short of the visit. Below, the person is lined up across six documents. Names and site agree, even though one writes "Marco Gallo, MD" and another "GALLO Marco".

The only red row is the certificate. The text below says why it matters: he is delegated eligibility and adverse event assessment.

3

Every value opens the page it came from

Any row in the line-up opens its source. Here the certificate shows the course completion date highlighted, so Lena can see the date with her own eyes before she asks anyone for anything. The sample certificate itself says "Refresher recommended every 3 years", which is how such a rule often reaches a sponsor's list.

4

A drafted request to the coordinator, in English

Draft request writes to Elena Rossi, the study coordinator, with the principal investigator and the CRA in copy. It asks for one thing, a refreshed certificate dated before November 2, says why, and tells her everything else about Dr. Gallo checked and agrees.

Lena can discard it. It leaves only when she presses Send.

5

The same request, in Italian

A switch shows the same content in Italian, with the same dates. Priya Ranvelle, the CRA for the European sites, has her reason in the sample: "Bologna answers faster when the request is in Italian." A person who reads Italian should read this before it goes.

6

The break now reads "requested", with its due date

Lena sends it. On the runway the red marker becomes an envelope reading "GCP · Dr. Gallo lapses Nov 2 · due Oct 14", and the count moves to 2 to clear and 6 requested. The break closes only when the corrected document arrives and passes the re-check.

Select any screen to open it larger. Dates and times are shown in US form.

Austin's 23 documents, read in 64 seconds, four findings and a signed green light by 10:58 am

In shortA new packet from Site 104 runs through the check. Four findings come back, each citing its pages. Lena accepts one with a note, which Omar Haddad approves, and sends the other three as one request. The site uploads five corrected documents, the re-check clears them, and she signs. The whole morning is on the record. The 64 seconds is this sample run.

1

Five agents check the packet, each step shown

Lena presses Check now. Packet intake sorts 23 documents into 12 families and attaches them to 6 people. The field reader reads 61 pages and 214 fields. The people cross-checker, the version and expiry checker and the country rules checker follow, and each says what it found and how long it took.

2

Four new breaks appear on Austin's bar

The runway now shows Site 104 broken in four places, and the panel opens the first: Dr. Kwame Mensah does medical tasks but is not on the Form FDA 1572. Two rows are red, the form and the missing financial disclosure.

3

The site packet: findings by severity, with the document open

Opening the packet shows the same four findings with their reasons: 2 high, 1 medium, 1 low. Each has an "Ask for" line and a suggested decision: ask the site, accept with a note, or not an issue. The delegation log sits on the right, with each name highlighted.

4

A citation opens Section 6 of the 1572

Selecting the numbered citation opens the form at the passage. Section 6 lists Dr. Hannah Lowe and Jessica Tran. Dr. Mensah is absent, and the log gives him eligibility, consent and adverse event tasks from September 22.

The form was signed September 24, after he was delegated, so asking for a corrected form is reasonable. Under FDA's FAQ a later change of sub-investigator would not by itself require a new form.5

5

Six people, lined up across the form, the log and their certificates

The people cross-check gives one row per person: what the 1572 says, what the log says, and which of their documents are present and in date. Dr. Mensah's row breaks at Section 6 and the disclosure. Dr. Lowe's breaks at GCP.

On this screen, the sample's rule treats the coordinator and pharmacist as not needed on the statement, because they have no medical tasks. Read that as the sample's setting. FDA says a research coordinator "should usually be listed", so a real team would settle this rule with its quality group.5

6

The 42 checks that passed, each with its source

A review also has to show what it confirmed. Checks passed lists them with the page each rests on: CVs in date, licenses active, the site address on the 1572 matching the contract, normal ranges covering the 8 protocol tests.

7

One request with three items, and one note sent for approval

Lena takes the suggestions. The CLIA certificate, valid to November 30, 40 days after the visit against the sample's 90, is accepted with a note and a reminder, and goes to Omar Haddad for a second approval. The other three become one request to Mia Alvarez, with Dr. Rafael Moreno and Sam Patel in copy, due October 12.

8

Corrected documents are re-checked and the findings close

Mia uploads a revised form, Dr. Mensah's disclosure, the IRB approval of Amendment 2, consent form 3.0 and Dr. Lowe's new certificate. The re-check keeps older versions and marks them superseded. It finds Dr. Mensah in Section 6 and the approval dated October 6. Nothing is open: 46 checks done, 1 accepted with a note.

9

Lena signs, with her name and the meaning of the signature

The signature dialog names the site, the checks and the documents to file. It records the meaning, "I approve activation of this site", and asks her to confirm that the essential records are complete and current.

Name, date and time, and meaning are the three things 21 CFR 11.50 asks a signed electronic record to show.49

10

Filed to the electronic trial master file (eTMF), and the depot and the CRA are told

The signature is recorded at 10:58 am. The 24 documents are filed to the trial master file in Zones 04, 05 and 08 with their details. The depot is told it may ship VLT-305 kits to Site 104, which fits the US rule that the drug ships only to participating investigators.1 Sam Patel is asked to book the initiation visit.

The views around the activation runway: documents by site, country requirements, start-up dashboard, check rules and activity trail

In shortAround the runway sit the views a team needs across a study: every document family at every site, the site table, what each country requires, how the study is tracking, and the settings a sponsor owns. Screens show the study later that morning, with Site 104 green-lit.

1

Twelve document families at every site

Sites run down and families run across. Each cell shows one state: checked, problem, expiring soon, missing or waiting. A red cell opens the document. Filters narrow it to a region or a state.

2

The site table: stage, readiness, open findings, visit and CRA

This is the view for a weekly meeting. It can be searched and filtered by country and CRA, and exported as the start-up tracker a sponsor may still need to share.

3

What each country requires, as the sample sponsor's list

US, Canada and EU lists set out the rule each check applies. This crop shows the list and what the check looks for. The limits are the sample sponsor's: FDA does not itself require a CV to be signed, dated or under 24 months, and no regulator we read sets three years for GCP.5,2

Your lists load here, with the true source of each line, and they change only through an approved version.

4

The start-up dashboard, on sample data

Packets checked, receipt to findings, packets complete the first time, findings confirmed by people, sites green-lit against plan, what the agents found and days in each stage. The 38 minutes and 94% on this screen are design targets shown on the sample.

5

Check rules you own: CV age, GCP window, validity after activation

Change a limit, such as the GCP window, and the screen shows at once who it flags today. Part A explains why this belongs in settings: these limits are yours. A new rule can be described in a sentence and takes effect once approved. Who signs, who approves a note, and where documents arrive are set here too.

6

The activity trail for one site

Every upload, agent step, decision, request and signature, with who and when. Agents and people are marked apart, and the trail cannot be edited.

Six start-up agents, and the people who confirm findings, send requests and sign the green light

In shortSix specialist agents read, line up, date and draft. None can confirm a finding, change a document or send anything to a site. The start-up lead confirms findings and signs, a second person approves any note, and the TMF owner samples the filing.

What each start-up agent does, and where it stops

Who confirms, who approves a note and who samples the filing

Lena, Sam, Priya, Omar and Dana on VLT-305, in their own words

Packets, protocol versions and visit dates go in; a dated runway and a signed release come out

Systems a site packet review connects to

Audit trail, signatures and credentials a clinical quality team will ask about

Each upload, agent step, decision, request and signature is recorded with who, what and when, and can be exported for an audit. A signed green light carries printed name, date and time, and meaning.49 Older versions are kept and marked superseded, in line with E6(R3)'s call for version history.7 A site sees only its own packet and messages. These are controls that support your compliance with Part 11, and validating your configured installation stays with you.48,47

Botminds' information security is managed to ISO 27001 and audited under SOC 2 Type II. No product can be certified for good clinical practice or Part 11, and we make no such claim. The solution is built for enterprise scale across many studies, countries and CROs, with roles, maker-checker approvals and an exportable audit trail.

What a bounced packet, a moved initiation visit and a lapsed certificate cost a sponsor

In shortThree kinds of cost sit here: specialist hours on each packet, extra rounds with the site, and a visit that has to move. A fourth, the value of a late site to the whole trial, is the largest and the hardest to defend. Nobody publishes packet-level costs, so your own numbers are the ones to trust, and you can find most of them in a week.

Reading ten closed packets to find your own rounds, hours and moved visits

What a second round does that no tracker prices

What automated packet review returns: specialist hours, fewer rounds with sites and fewer moved visits

In shortPut in your own numbers. The calculator values hours on each packet, hours on extra rounds, and visits that no longer move. It shows the schedule value on a separate line, because that depends on a share you have to defend. Every default is illustrative, a design target or a published figure. Replace them.

Packet reading, extra rounds, moved visits and audit findings: where start-up automation saves

Sites a year, hours per packet and rounds per site, on your own figures

How ready your start-up team is: country lists, written limits, visit dates and who signs a release

In shortPick the answer closest to your team today. Eight questions score 0 to 2, and 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.

Twelve questions for any study start-up software vendor, us included

In shortMost tools collect and track documents well, and many demonstrate well on clean files. These questions test what a demonstration hides: whether the tool reads the documents, whether it tests them against one another, what it does when unsure and who can close a finding. Ask us too.

Testing packet review in ten weeks: closed packets, planted errors and one live study beside your usual process

In shortStart with packets you have closed, because you know how they turned out. Write down your rules, re-check them, plant errors to see whether the checks fail when they should, then run one live study in parallel. Decide on five numbers.

Planted errors found, false alarms and rounds per site on your own packets

Site activation questions answered: the 1572, GCP certificates, delegation logs and essential records

In shortThe questions start-up teams and buyers search for most, each answered on its own.