FDA Investigator Records and Site Monitoring: What BMIS and Inspection Data Can Support

A public record is a lead for qualification

FDA's Bioresearch Monitoring Information System (BMIS) can help a sponsor investigate a clinical site's public submission history. Investigator experience, protocol suitability and monitoring intensity require direct, current evidence. FDA explicitly describes BMIS as incomplete, and submission of Form FDA 1572 itself is not required even when the sponsor must collect the form. No public match is an unknown, not proof of no IND experience. [1]

The distinction is important because the file contains a striking geographic change. In our independently recomputed parsed-record audit, non-US clinical-investigator receipt records fell from 17,847 in 2019 to 6,435 in 2025. Their share of the year's US-plus-non-US records fell from 30.0% to 12.0%. This is a change in public receipt records, with its cause unresolved in the reviewed FDA sources. [2]

Three evidence sets serve different planning purposes here. BMIS provides submission-reference records. ClinicalTrials.gov, represented through AACT, describes registered study locations. FDA's BIMO metrics and annual inspection-observation workbooks describe selected inspection outputs. Their different units and coverage determine which questions each can answer.

For a sponsor or CRO, the useful result is a documented workflow: resolve a possible public match, obtain direct evidence of qualifications and capacity, assess risks specific to the protocol, and make monitoring and issue-resolution decisions traceable. The article closes with suggested qualification and readiness workpapers. The chart-data workbook contains the eight figures' values and source notes, without investigator-level personal data.

The BMIS file needs two separate counts

The analyzed FDA download contains a text file named PUBLIC_BMIS_DATA_2026_05_29.txt, captured on 23 August 2026. We checked the original ZIP against its archived checksum and independently compared the raw text with the parsed records. The file has 1,841,747 physical data lines after the header; 1,841,422 have the expected 14 fields. The other 325 lines were excluded rather than reconstructed by guessing where embedded delimiters belonged. [2]

Among the parsed rows, 1,120,176 are designated clinical investigator, 668,556 IRB and 52,423 CRO. Another 267 carry other type values. These counts describe records, not distinct individuals or organizations. A person or organization can appear repeatedly when identified in new submissions; references to both Forms 1572 and 1571 contribute to the database. [1][2]

The all-row total also needs a date qualification. Of the parsed CI rows, 247,950 have recorded receipt dates before 1 October 2008, including historical values and obviously anomalous years. Another 872,214 fall between that date and 29 May 2026. Three are later than 29 May but no later than the August capture, and nine are dated after capture. FDA identifies at least one placeholder date, 1 January 1900, as invalid. Suspicious years remain as recorded in the all-row inventory and are excluded when outside an analysis window. [1][2]

This produces two legitimate, different summaries: an all-row inventory of the named file as recorded, and an analysis restricted to an explicit receipt-date window. It would be inaccurate to say that every one of the 1.12 million CI rows was received after October 2008, or that the filename alone guarantees a clean May 2026 cutoff.

For annual comparisons, we use full calendar years 2009 through 2025. For the later country window, we explicitly stop at 29 May 2026. That removes three US records from the partial-2026 window; the US count for January through that cutoff is 21,047. The China and non-US counts in that window are unchanged. All percentages below use the same date rule in numerator and denominator. [2]

The result is a reproducible audit of the available records. Publishing the defects alongside the counts makes the audit more useful: another analyst can distinguish the effect of changing a filter from a real change in the source file.

What changed in the non-US receipt records

Non-US CI records numbered 17,847 in 2019, 14,285 in 2020 and 7,530 in 2021. From 2022 through 2025 they remained between 6,391 and 6,904 per year. US records followed a different pattern, reaching 47,255 in 2025. The non-US share consequently fell both because its own numerator declined and because the US denominator grew. [2]

The reason for the change remains unresolved. Differences in submission practice, database capture or the mix of studies might contribute, but no reviewed source measures their contributions. Testing a policy or submission-practice explanation would require dated evidence about that mechanism and its effect on database capture.

China was the largest non-US country code in the explicit 2024 through 29 May 2026 window: 1,545 records, or 9.2% of the 16,778 non-US total. Japan contributed 1,128 and Australia 1,056. These are address-code counts. Sponsor nationality, institutional identity, regulatory destination and research quality require other evidence. [2]

The figure compares that partial window with 2019 through 2021 to show the changing distribution among selected countries. The line connects counts over unequal durations. It would be inappropriate to annualize the partial 2026 count without a model of seasonality and database updates. We leave the observed windows visible instead.

The distinction between geography and IND status also matters operationally. A foreign study may be conducted under an IND, while a foreign study outside an IND follows a different regulatory route. FDA's Form 1572 FAQ and sponsor/CRO compliance programme discuss those distinctions, including specific foreign-site signature-waiver situations. The applicable route and any waiver should be verified from study documentation. [3][4]

A study team using BMIS should consequently record what a match actually establishes: the file and date searched, the name and address fields compared, the plausible match and any ambiguity. It should then ask for direct evidence relevant to the proposed role. Current availability, staff continuity and competence with the new endpoint need direct confirmation.

Registered study locations provide a different comparison

We separately analyzed the 1 August 2026 AACT snapshot for interventional studies with registered start years 2015 through 2026. There were 299,607 such studies, of which 62,700 had an industry lead-sponsor classification. The cohort retains all study statuses and includes actual, estimated and unspecified start-date types. “Registered start year” is therefore more accurate than “studies already initiated” or “active trials.” [5][6]

The facilities table records study-location rows. A hospital participating in several trials can appear several times, while a location may be listed before recruitment or remain in a completed record. Institution-level deduplication and current-capacity assessment remain outside this analysis. This is a separate descriptive dataset, not a denominator for BMIS completeness or FDA inspection risk.

With a strict 2020 through 2026 start-year filter, the United States contributed 366,052 study-location rows, China 93,275 and France 45,251. Applying the upper bound consistently excluded 864 location rows belonging to studies with registered start years after 2026. Without that check, a script can produce plausible country totals while silently describing a broader cohort than the article states. [5]

An equal-duration comparison offers another view. The following counts use non-removed country listings on industry-led studies, rather than facility rows. Each study can contribute to several countries; actual recruitment requires further confirmation. [5]

The directions differ by country. More registered industry studies list China in the later five-year group, while Germany and the United Kingdom have fewer. This is more informative than comparing a five-year window with six years plus part of 2026 and calling the difference worldwide growth.

The relationship with BMIS remains unmeasured. The two sources have different populations, event dates and record-generating processes. Establishing a shift in IND participation would require a validated investigator-to-study linkage and evidence of the regulatory route.

For feasibility work, this separation is practical. Registry records can identify protocols and potential locations to investigate. Qualification still needs current information about people, facilities, patient access, equipment and competing commitments. A public listing gives the team a lead for that conversation.

Inspection results have three different denominators

FDA's BIMO publications distinguish conducted inspections, inspections with final classifications and Form 483 observation records. These quantities answer different questions. A Form 483 communicates inspectional observations; it is not itself the final agency determination. A final classification describes FDA's assessment of an inspection, and an Official Action Indicated classification is not the same thing as an already imposed sanction. [4][7]

Using the internally consistent historical CI series in FDA's FY2024 metrics deck, the final-classified totals for FY2020 through FY2024 were 544, 437, 504, 681 and 609. OAI counts were 5, 5, 9, 9 and 15, corresponding to shares from 0.9% to 2.5%. The series combines domestic and foreign clinical-investigator inspections. The published total covers both domestic and foreign activity. [7]

The FY2023 and FY2024 decks contain small differences in historical NAI counts. For example, the former's FY2023 component counts add to 682 while the latter's add to 681, matching the stated total. The latter series supplies every classification count and denominator in this report. [7][8]

FDA's separate CI observation-trends publications report Form 483 issuance in 142 of 693 conducted inspections in FY2023 and 156 of 692 in FY2024. The published rounded rates are 20% and 23%; FY2021's publication gives 25%. These are different cohorts from the final-classified totals. Comparing issuance with OAI would require following the same inspections through their classification process. [9][10][11]

FDA selects inspections for regulatory purposes. Estimating national risk would require country-specific outcomes and a compatible inspected-population denominator. Other FDA publications and its public classification database contain geographic information, each with its own scope; BMIS receipt volume describes a different population. [12][13]

For a qualification review, the more useful question is what a relevant inspection record says about the site, issue, timing and response. Assess the finding in its historical context, then use current evidence and the proposed study's risks to complete the assessment.

What eleven years of citation workbooks add

We independently checked the FY2015 through FY2025 FDA inspection-observation workbooks, including the original XLSX files. In each year's BIMO tab, we summed the published citation frequencies and identified the short-citation category concerning the investigational plan and the category concerning adequate and accurate case histories. Their combined share ranged from 35.6% to 53.0% of those annual citation totals. [14]

That denominator is deliberately narrow. The workbook records come from FDA's inspection-reporting tools; manually prepared 483s and some unsynchronized records are absent. The BIMO tab spans several oversight programmes, including investigator, sponsor, IRB and laboratory contexts. Each frequency counts citations within that published subset. [14]

The recurring topics are nevertheless useful prompts. Protocol execution and case histories connect participant care, source documentation, data entry and review. They should lead a team to ask where a study could lose that connection, not to assign monitor hours in direct proportion to a historical citation percentage.

FDA's risk-based monitoring Q&A organizes planning around risks to participants and critical data or processes. Study characteristics, the likelihood and consequences of errors, detectability and emerging information all matter. BMIS matches can inform further investigation; monitoring intensity follows the documented risk assessment. [15]

The following table translates inspection themes into proposed study-level questions. Teams should select the review methods and triggers appropriate to their risks.

Legal duties should retain their actual conditions when translated into a checklist. Under 21 CFR 312.64(b), investigators generally must report serious adverse events immediately to the sponsor with a causality assessment, subject to the provision's exception for serious study endpoints unless there is evidence of a causal relationship. Nonserious adverse events follow the protocol timetable. Writing simply “report all adverse events within protocol timelines” would weaken that distinction. [16]

The broader implication is that a monitoring plan should name the critical issue, its evidence source, the person responsible for review and the action the evidence can trigger. A dashboard signal can justify investigation. Confirmation and resolution each require supporting evidence.

A qualification workpaper that preserves uncertainty

For applicable IND investigations, 21 CFR 312.53 addresses sponsor selection of qualified investigators and collection of the signed investigator statement and qualifications information. Public database searches can supplement that work. Qualification for this protocol rests on the sponsor-held evidence and assessment. [17]

A defensible public-history workpaper can use three outcomes:

These are search outcomes, not investigator maturity tiers. Exact name-and-city strings leave identity ambiguity, and using an IRB chair's first name as an organization key is particularly misleading. The FDA field definitions distinguish personal-name fields from the site or organization field. Organization counts would require a validated entity-resolution method using the appropriate fields. [1]

After identity and role are established, assess the proposed work: the investigator's relevant qualifications, the delegated team's preparation, access to the necessary facilities, the endpoint workflow and current capacity. For a different assessment or population, confirm the investigator's relevant preparation. Those are study-specific judgments supported by qualification records, not facts derivable from a public receipt count.

Record the rationale for monitoring intensity in the same workpaper or link it to the controlled monitoring plan. Relevant considerations can include the consequence of an error, how quickly it could be detected, the complexity of critical procedures, prior verified issues and changes in site staffing or performance. The appropriate mix of central and on-site review follows that assessment. FDA's guidance supports adapting the approach as meaningful new information emerges. [15]

For example, a delayed data entry pattern is a signal requiring context. It could reflect a source-access problem, an unresolved workflow or an actual delay in performing an assessment. The response should establish which problem exists before treating an electronic timestamp as proof that care or protocol activity occurred late. This is an illustrative reasoning example, not an observed finding from the public datasets.

A readiness file built around retrievable evidence

A readiness file is most useful when a reviewer can follow an important decision from the requirement to what happened and how an issue was handled. The seven-part organization below is an editorial suggestion. FDA does not require this folder structure, and the documents needed depend on the applicable regulations, study and organization's controlled processes.

Use the structure to retrieve evidence across the systems and repositories where it is controlled. A CV or other statement of qualifications can satisfy the relevant qualification-information provision; the checklist should preserve the applicable alternatives. Likewise, operational summaries should preserve applicable exceptions, including protocol changes needed to eliminate an immediate hazard. [3][17]

A useful internal walkthrough starts with a critical participant record or study issue and asks the team to reconstruct the evidence path. Can it locate the applicable version, responsible role, source record, correction and follow-up? If the path breaks, the team has a specific information or process gap to resolve. This exercise is a suggested control check, not an inspection prediction or a guarantee that a future inspection will have no observations.

EClinCloud's CTMS and eTMF offerings are relevant to tracking trial activity and organizing trial documentation. Product selection should follow the required workflow, access and record controls. A system can help record actions and retain evidence, while accountable personnel assess whether the response is adequate. [18][19]

To maintain this analysis, preserve each source's own date and unit: BMIS receipt windows, AACT registered start years and FDA fiscal-year inspection outputs. Refresh and linkage decisions should preserve those distinct scopes. Use the public data to focus questions, then let verified site evidence and the protocol determine the monitoring decision.

Sources

1. U.S. FDA, Bioresearch Monitoring Information System, current scope notice and field definitions, accessed 6 September 2026.

2. EClinCloud original analysis of the FDA BMIS public download, file named 29 May 2026, captured 23 August 2026; raw ZIP and parsed records independently checked 6 September 2026. Counts and exclusions are specific to this frozen file; the public download can change.

3. U.S. FDA, Frequently Asked Questions: Statement of Investigator, Form FDA 1572, final guidance, June 2010. Distinct from the later draft revision.

4. U.S. FDA, Compliance Program 7348.810: Sponsors and Contract Research Organizations, issued 15 September 2021.

5. EClinCloud original analysis of CTTI AACT downloads, 1 August 2026 snapshot, independently recomputed 6 September 2026. Interventional studies with registered start years 2015-2026; study-location and country-listing counts use separate source tables.

6. ClinicalTrials.gov, Protocol Registration Data Element Definitions, accessed 6 September 2026.

7. U.S. FDA, Bioresearch Monitoring FY2024 Metrics, published 2025. Historical final-classified CI series used consistently across FY2020-2024.

8. U.S. FDA, Bioresearch Monitoring FY2023 Metrics, published 2024. Used to check deck-to-deck count differences.

9. U.S. FDA, FY2023 Clinical Investigator Form FDA 483 Observation Trends, published 2024.

10. U.S. FDA, FY2024 Clinical Investigator Form FDA 483 Observation Trends, published 2025.

11. U.S. FDA, FY2021 Clinical Investigator Form FDA 483 Observation Trends, data updated 24 August 2022.

12. U.S. FDA, CDER BIMO FY2023 Metrics, data updated 22 February 2024. Geographic inspection-activity information has a different scope from the all-center classification series.

13. U.S. FDA, Inspection Classification Database, coverage notice, accessed 6 September 2026.

14. U.S. FDA, Inspection Observations, FY2015-2025 annual XLSX workbooks and coverage limitations. EClinCloud independently verified all eleven BIMO sheets against the archived parsed records.

15. U.S. FDA, A Risk-Based Approach to Monitoring of Clinical Investigations: Questions and Answers, final guidance, April 2023.

16. 21 CFR 312.64: Investigator reports, current text checked 6 September 2026, particularly paragraph (b).

17. 21 CFR 312.53: Selecting investigators and monitors, current text checked 6 September 2026.

18. EClinCloud, Clinical Trial Management System, current product page, accessed 6 September 2026. Company description, not independent outcome evidence.

19. EClinCloud, Electronic Trial Master File, current product page, accessed 6 September 2026. Company description, not independent outcome evidence.