WorkflowClinical Development

Learn · inspection and the report

Clinical trial inspection findings in monitoring visit reports

Monitoring is a small share of GCP inspection findings and a disproportionately severe one: in EMA's 2023 numbers, 40 of 720 deficiencies, four of them critical. Almost none of them say the report was badly written. This guide counts what inspectorates publish, shows the shape a monitoring finding takes, and where Monitoring Visit Report Automation puts the same comparison before a signature instead of after one.

Updated · 13 min read

See one case, screen by screen →
Watch with transcript →

Clinical trial inspection findings about monitoring are rarer than clinical operations teams fear and heavier than they expect. In the European Medicines Agency's own count for 2023, inspectors recorded 720 deficiencies across 67 inspections, and 40 of them sat in the monitoring sub-category — 5.6% of the total, but four of those forty were graded critical, twice the critical rate of findings overall. Writing a monitoring visit report is hard because it has to agree with four other things an inspector can also read: the visit notes, the previous report's open actions, the site's own file and the study databases. Botminds' Monitoring Visit Report Automation drafts that report from those four sources with a citation on every sentence, and runs the comparison a reviewer would run before the clinical trial manager signs, not after an inspector asks.

The counts on this page come from the published reports of two inspectorates and one FDA inspection programme; the screens come from the solution's own sample study data, on fictional sponsors, sites and people. No figure here describes a Botminds customer, and none of the inspection statistics belongs to the solution: they are other organisations' inspections.

What one year of EMA-coordinated GCP inspections recorded

720

deficiencies recorded in 2023, across every category

67

inspections requested by the CHMP and conducted that year

11

findings per site inspected, on average

40

of the 720 were in the monitoring sub-category

European Medicines Agency, GCP Inspectors' Working Group annual report 2023 (EMA/INS/GCP/164938/2024), tables 3 and 4. These are CHMP-requested inspections only, a small part of the inspections EU and EEA inspectors carry out.Source: EMA GCP Inspectors’ Working Group, annual report 2023

What do inspectors look for in monitoring visit reports?

Three things, in this order: that the reports exist, that they were written and sent inside the sponsor's own timelines, and that what they say agrees with everything else in the file. FDA's compliance programme for sponsors, CROs and monitors tells its investigators to "Review the sponsor's monitoring reports and other records related to monitoring that may be located in the, trial master file (TMF), and/or clinical trial management system (CTMS) to evaluate monitoring activities" — the typographical slip is the programme's own. An instruction just before it in the same section asks whether the monitoring plan was followed, naming visits performed by the required timeline, reports completed by the required timeline and letters to sites sent by the required timeline.

The order matters because the first two tests are mechanical and the third is not. A sponsor can answer "where is it" and "was it on time" from a system. Whether a sentence in a report is true is a comparison with four other sources, and it is the test that produces findings. That is also the test a sponsor can run on itself: nothing in the rules stops a sponsor from making the comparison on every draft, and who reviews a monitoring visit report, and what they are looking for, sets out how that review is usually staffed.

How many findings does a GCP inspection produce?

About eleven, if the EMA's 2023 numbers are the guide. Across the 67 CHMP-requested inspections the working group recorded "A total of 720 deficiencies, comprising 36 critical (5.0%), 336 major (46.7%) and 348 minor (48.3%) findings … This represents an average of 11 findings per site inspected". The UK's inspectorate publishes a different cut of the same kind of work: in its 2019–2020 metrics period the MHRA carried out 36 GCP inspections, and its seven commercial-sponsor inspections averaged 0.6 critical, 2.4 major and 4.9 other findings each — close to eight per inspection, on a narrower definition of what is counted.

Neither number is a forecast for any one inspection. What both say is that findings are normal and criticals are not: in the EMA's count, one finding in twenty was critical.

Monitoring findings are graded harder than findings as a whole

All 720 deficiencies, 2023

48.3%
46.7%
5%
  • Minor · 348 · 48.3%
  • Major · 336 · 46.7%
  • Critical · 36 · 5%

The 40 monitoring findings

37.5%
52.5%
10%
  • Minor · 15 · 37.5%
  • Major · 21 · 52.5%
  • Critical · 4 · 10%
EMA GCP Inspectors' Working Group annual report 2023, tables 3 and 4. The percentages for the monitoring row are our arithmetic on the published counts of 15 minor, 21 major and 4 critical: one monitoring finding in ten was critical, against one in twenty across all categories.Source: EMA GCP Inspectors’ Working Group, annual report 2023

Critical, major and minor: what the gradings mean for a report

A grading describes impact, not craft. It asks what the deficiency could do to a participant's rights, safety or well-being, or to the reliability of the trial's results. Nothing in it rewards a well-turned sentence, and nothing punishes a clumsy one. That is why a report written in plain, dull prose can still carry a critical finding, and why a polished report that quietly omits an unresolved action is the more dangerous document.

The second thing a grading carries is an address. The EMA's working group also publishes which party each finding was attributed to: of the 36 critical findings in 2023, 23 were the sponsor's responsibility and three the investigator's, with the rest shared. Investigator sites generated most of the findings by volume — 448 of 720, against 233 at sponsor sites — but the severe ones land on the sponsor, and monitoring is one of the routes by which a site's problem becomes a sponsor's finding.

Monitoring is 40 of the 138 trial-management findings EMA recorded in 2023

Data management
48
Monitoring
40 · 4 critical
Protocol, CRF, diary and questionnaire design
17
Clinical study report
13
Document control
10
Statistical analysis
7
Audit
3
EMA GCP Inspectors' Working Group annual report 2023, table 4: the sub-categories of the trial-management category, which held 138 of the 720 deficiencies. The general category (essential documents, source documentation, training, contracts) was larger still, at 288.Source: EMA GCP Inspectors’ Working Group, annual report 2023

The shape of a monitoring finding: not written badly, but not carrying what it should

Read the EMA's own description of what its inspectors found under monitoring in 2023 and the pattern is detection, not drafting: "Assessment of AEs, eligibility, and physical examination not done by the physician was not detected by the trial monitor". The companion bullet is about procedures that failed to detect protocol deviations and to reconcile investigational product accountability. In both, the report is the place the failure became visible, not the thing that failed.

Three shapes recur across the published findings, and they need different fixes.

Three shapes a monitoring finding takes, and what each one asks of the report

The shapeWhat the inspector seesWhat would have prevented it
The monitor did not detect itSomething detectable at the visit — an eligibility assessment signed by the wrong person, product accountability that does not reconcile — appears nowhere in any report.A checklist whose every item has to become a statement, so an unanswered item cannot pass silently into a signed report.
The monitor detected it; the report does not carry itThe investigator or the site raised a concern, and the monitoring reports do not reflect it.A draft built from the visit notes with a citation on each sentence, so a fact in the notes that reached no section is visible as a gap.
The report says something the file contradictsA count, a medication statement or a closure claim in the report disagrees with the study databases or with the previous report.A comparison with the study databases and with last visit’s open actions, run on the draft, before anybody signs it.
Our grouping of the finding descriptions published by the EMA GCP Inspectors' Working Group (2023) and the MHRA GCP Inspectorate (2019–2020 metrics). The right-hand column is what each shape implies for how a report is produced, not a regulator's recommendation.

The third shape is the one most under a sponsor's control, because both sides of the contradiction are already in its own systems on the day the report is drafted. What every monitoring visit report must contain works through the sections where these statements live.

One inspection's critical finding, in the inspectorate's own words

The MHRA's 2019–2020 metrics report describes four critical findings raised at a single contract research organisation. The second concerned investigational product management. A principal investigator had told inspectors that they had raised concerns about blister wallets that did not state when each dose should be taken or the strength of each tablet. The site acted: it educated the parents and carers of its participants and halted recruitment. The sponsor and CRO did not submit an urgent safety measure or a potential serious breach. And then this sentence, which is the reason the finding belongs on this page:

"Furthermore, the PI concerns were not reflected in the monitoring visit reports."

MHRA GCP Inspectorate, inspection metrics 2019–2020

Two details are worth taking away. The first is that the finding was graded under investigational product management, not under monitoring: a count of findings filed against "monitoring" understates how often a monitoring visit report is part of the evidence for something else. The second is that the inspectors learned of the concern from the investigator at the site, and then went to the reports to see whether the sponsor had learned of it too. The report was read as a test of whether information travelled.

A concern of that kind usually has a reporting clock attached to it as well; what counts as a serious breach, and when the clock starts, sets out the test and the deadline.

What an inspector collects when monitoring is in question

FDA's compliance programme is unusually specific about what its investigators take away when they suspect monitoring failed. Its worked example is a sponsor that did not identify a site's late serious adverse event reporting; in that case "the ORA investigator should collect copies of study-specific site monitoring plan(s) and relevant standard operating procedures (SOPs), monitoring visit reports, monitoring logs, SAE reports sent to the sponsor by clinical investigators, other documentation to support and describe the SAEs, and any other supporting evidence, and document the observation in the EIR".

That list is a readiness exercise in itself. Every item is something the sponsor already holds, and the finding is written from the disagreements between them.

The documents FDA names when monitoring is in question, and what each is read against

What is collectedWhere it is usually heldWhat it is read against
Study-specific site monitoring plansThe trial master file, with its versions and effective datesWhat the reports actually did: visit types, frequency, what was verified, what triggers escalation
Relevant standard operating proceduresThe quality systemThe plan, and the timelines the reports were held to
Monitoring visit reportsThe trial master file and the trial management systemThe notes, the site file, the study databases and the previous report’s open actions
Monitoring logsThe trial management systemVisit dates and report dates against the clocks the procedures set
Serious adverse event reports sent by investigatorsSafety systems and the site fileThe dates and content of the reports that should have noticed a late one
Supporting evidence for the events themselvesSource records at the siteEverything above
The collection list is quoted from FDA Compliance Program 7348.810 (implementation date 15 September 2021). The second and third columns are our reading of where each item sits and what a disagreement between them would show.Source: FDA Compliance Program 7348.810, sponsors, CROs and monitors

There is an asymmetry in that programme worth knowing. Audit reports are protected: investigators are told in general to confirm audit certificates rather than request the reports themselves, and the agency reserves access for cases where serious non-compliance is suspected. Monitoring reports have no such shelter. They are collected as ordinary evidence, which makes them among the internal documents an inspector is most likely to read.

What happens when a monitoring report contradicts the site's records?

The inspector has both documents and no reason to prefer yours. A contradiction turns one question — what happened at this site — into three: which version is right, how the wrong one was signed, and what else the same process may have got wrong. The third question is the expensive one, because it invites a wider sample.

Contradictions come in three sizes, and in the solution's sample study data each has a different consequence. A count that disagrees — the Site 114 draft says 22 randomised where the randomisation system shows 23 — is a minor comment; it is still a sentence a signer has to decide about. A statement that disagrees is not minor: "No rescue medication" in a draft, against notes and the study database showing a participant on oral prednisone for seven days, is a must-fix, and signing stays locked until it is resolved. A closure that disagrees is the worst of the three: the Site 309 draft says consent forms were reviewed for all nine participants and all were signed on the current approved version, while the re-consent action from the previous visit has been open for 56 days.

A draft whose consent sentence disagrees with an action still open after 56 days

A monitoring visit report for Site 309, Kestrel Heart Institute, interim monitoring visit 3, with two must-fix comments open and signing locked. At the top, three actions rolled forward from the last report: a calibration certificate closed before this visit, a re-consent of participant 309-003 on ICF v3.1 marked Not in this report, and a sub-investigator's CV and GCP certificate carried forward and open 56 days. In the margin, a report reviewer comment reads: Why isn't last visit's consent issue mentioned? AI-309-007 — re-consent of 309-003 on ICF v3.1 — is still open. The consent paragraph of the draft, which states that all nine consent forms were on the current approved version, is underlined and tied to the comment by a line.
Sample study data in Monitoring Visit Report Automation; the study, site and people are fictional. The comment is raised against the sentence it questions, with the action it contradicts beside it.

Every unresolved action from the last report sits above section 01 with its fate at this visit — closed before, closed here, carried forward with its age, or not addressed — which is why a sentence claiming the matter is settled is caught by the document itself. Carrying last visit's open actions into this report follows that mechanism through a full visit.

What an enforcement letter shows about a monitor who said nothing

In August 2023 FDA wrote to a clinical investigator about three failures to follow the investigational plan, including study drug administered outside the protocol's specified body sites. The investigator's written response explained that the deviations had not been raised during the sponsor's regular monitoring visits, and that the sponsor's monitoring team had reviewed one observation without generating a query report. FDA did not accept the explanation as a defence:

"Although the monitor did not bring this observation to your attention during the conduct of the clinical trial, as the clinical investigator, you are ultimately responsible for ensuring compliance with the investigational plan."

FDA warning letter to a clinical investigator, 8 August 2023

Read from the site's chair, that is a warning that the monitor's silence protects nobody. Read from the sponsor's chair it is worse, because the same paragraph is now a public statement that months of visits to this site produced reports in which the deviation does not appear. The sponsor's own documents became the evidence for how long the problem ran.

ICH E6(R3) asks a monitoring report to summarise what was reviewed, describe significant findings, state the actions required and follow up on their resolution, including those not resolved in previous reports. An omission is therefore not a gap in a narrative. It is a requirement the report did not meet.

How do you prepare monitoring reports for an inspection?

Not by rereading them. By the time an inspection is announced the reports are signed, filed and, in most procedures, amendable only by a documented correction that is itself visible. Preparation that changes anything has to happen earlier, at the draft, and it is the same work the inspector will do: compare the report with the notes, the checklist, the study databases and the previous report's open actions.

Four habits do most of the work, and none of them needs software to state.

What two inspectorates count, and what each one implies for a sponsor's own checks

What the published numbers sayWhat it implies before signature
Volume

What the published numbers say: EMA, 2023: 720 deficiencies across 67 inspections, an average of 11 per site inspected.

What it implies before signature: Findings are normal. The aim is not zero; it is that nothing found was visible in your own file first.

Severity

What the published numbers say: EMA, 2023: 36 critical findings, 5.0% of the total — but 4 of the 40 monitoring findings, or 10%.

What it implies before signature: In that year’s count a monitoring finding was twice as likely to be critical as a finding in general. Treat the ones your own review raises accordingly.

Who answers

What the published numbers say: EMA, 2023: 23 of the 36 critical findings were attributed to the sponsor, 3 to the investigator.

What it implies before signature: Severity migrates upward. A site problem that the reports did not carry becomes the sponsor’s finding.

How often a sponsor is hit

What the published numbers say: MHRA, 2019–20: of seven commercial-sponsor inspections, four had at least one critical finding and all seven had at least one major or critical finding.

What it implies before signature: Plan on being found something. The question is whether the file shows you had already found it.

Left column from the EMA GCP Inspectors' Working Group annual report 2023 and the MHRA GCP Inspectorate metrics for 1 April 2019 to 31 March 2020 (re-issued 11 April 2023). The two count different inspection populations and grade on their own scales, so the rows compare pictures, not like with like. The right column is our reading.Source: EMA GCP Inspectors’ Working Group, annual report 2023; MHRA GCP Inspectorate, inspection metrics 2019–2020

What this changes about how a report is drafted and checked

Every finding shape above is a comparison that could have been made earlier by somebody with access to the same four sources. The constraint has never been the rule; it is that a clinical trial manager reading a queue of reports cannot reliably hold a previous report, a checklist, an enrolment table and a medication log in mind at once, for every report, inside a ten-working-day clock.

In Monitoring Visit Report Automation, the agents draft the report in the sponsor's template from the notes, the checklist, the last report and the study data, citing a source on every sentence; a reviewer then checks the draft against the monitoring plan and the site data, and leaves each finding in the margin with its evidence and a suggested replacement sentence. Findings it marks must-fix lock signing. A comment the signer disagrees with is dismissed with a reason that stays on the report, and minor comments left open are recorded as accepted as written by the person who signed. The reviewer proposes; it never signs.

What Monitoring Visit Report Automation shows a signer before the signature

The review screen of a monitoring visit report for Site 114, Riverside Dermatology Research, day 4 of 10, due 15 October, with a banner reading three must-fix comments open, signing opens when they are fixed or dismissed with a reason. A panel above section 01 lists the actions rolled forward from the last report: an IRB approval closed before this visit, an assessment of a fridge temperature excursion marked Not in this report, and a delegation log item closed at this visit. Margin comments from the report reviewer ask why the fridge excursion is not in this report, flag an opinion about the principal investigator, and note that the randomisation system shows 23 randomised where the draft says 22.
Sample study data in the solution; the study, site and people are fictional. Each comment is tied by a line to the sentence it questions, and carries the rule it rests on, the evidence and a suggested text.

What a team can measure from that is its own position, not an inspection outcome. In the sample data the reviewer raised 61 findings across 27 reports — 22 missing escalations, 14 dropped carry-forward actions, 11 statements inconsistent with the site data, 9 missing statements and 5 pieces of subjective wording — and 58 were fixed before signature. Those are the categories the published findings describe, counted on drafts while they can still be changed.

The same categories, counted on drafts rather than at an inspection

A dashboard showing how fast and how cleanly visit reports reach signature: 93 per cent signed within the clock against a 90 per cent target, 25 of 27 reports; visit to signed 6.5 median working days, was 11.8; CRA writing time 0.8 hours per report, was 3.2 by hand; 58 of 61 findings fixed before signing. A doughnut headed what the reviewer agent caught breaks 61 findings into missing escalation 22, dropped carry-forward action 14, inconsistent with site data 11, missing statement 9 and subjective wording 5. Below, open action items by site show the oldest at 74, 56, 48 and 41 days, and a writing-time bar chart compares writing by hand at 3.2 hours with confirming the draft at 0.8 and manager review at 0.2.
Sample study data in Monitoring Visit Report Automation. The clock and writing-time figures are design targets for the solution, not a measured customer result, and nothing here predicts an inspection outcome.

No tool lowers a finding rate on its own, and no honest one claims to. What a check on every draft changes is which document holds the contradiction: the file that goes to the inspector, or a comment the signer resolved on day four. To see the comparison run on a report of your own, start from Monitoring Visit Report Automation.

Questions people ask

Do inspectors read every monitoring report?

No. They sample. FDA's compliance programme directs its investigators to review the sponsor's monitoring reports and related records in the trial master file or the trial management system to evaluate monitoring activities, and the full collection — plans, SOPs, reports, logs, safety reports — is specified for the case where a monitoring failure is suspected. In practice, a suspicion raised at a site or in an interview decides which reports get read closely.

What is a critical finding in a GCP inspection?

A deficiency serious enough to have affected, or to be capable of affecting, participants' rights, safety or well-being, or the reliability of the trial's results. They are uncommon: the EMA's inspectors graded 36 of 720 deficiencies critical in 2023, 5.0% of the total. The grading is about consequence, not about how the document reads.

What share of GCP findings are about monitoring?

In the EMA's 2023 count, 40 of 720 deficiencies, or 5.6%, sat in the monitoring sub-category, of which four were critical. That understates the role of monitoring reports, because a report can be part of the evidence for a finding filed elsewhere: the MHRA's critical finding about poorly labelled investigational product, which also noted that the investigator's concerns were not reflected in the monitoring visit reports, was graded under investigational product management.

How often do sponsors get a critical finding?

Often enough to plan for. In the MHRA's 2019–2020 metrics, four of seven commercial-sponsor inspections had at least one critical finding and all seven had at least one major or critical finding. In the EMA's 2023 count, the sponsor was the responsible party for 23 of the 36 critical findings recorded, against three for the investigator.

What documents does an inspector ask the sponsor for?

When monitoring is in question, FDA names study-specific site monitoring plans and relevant SOPs, monitoring visit reports, monitoring logs, the serious adverse event reports investigators sent to the sponsor, the documentation describing those events, and any other supporting evidence. Every item is one the sponsor already holds; the finding is written from the disagreements between them.

What happens if a monitoring report contradicts the site's records?

The inspector has both, and the contradiction widens the question from what happened at this site to how the wrong version was signed and what else the process missed — which usually means a larger sample. This is the shape of finding a sponsor is best placed to prevent, because both sides of the contradiction are in its own systems on the day the report is drafted.

Monitoring Visit Report Automation.
See it on your own files.

An agentic solution from Botminds. Turns a CRA's visit notes into a cited monitoring visit report, rolls forward every open action, and a reviewer agent comments in the margin before the manager signs. Book a demo, or explore how it works.