Is AI trial screening a medical device? How FDA’s definitions apply

Software that checks patient records against trial eligibility criteria generally falls outside FDA’s medical device rules, although the answer turns on intended use. FDA’s software policy lists functions that help “match patients with potentially appropriate clinical trials” among its examples of software that is not a medical device (FDA). That example concerns patient-facing tools, and we found no FDA document addressing site-facing chart review. If regulators treated a match as clinical decision support, the tool would need to meet all four statutory criteria for exclusion, and image analysis, inferred diagnoses or time-critical alerts could move it into device territory.

A tool that is not a device still falls under good clinical practice expectations for computerized systems, which FDA updated in September 2025 by adopting ICH E6(R3) (Federal Register).

The device definition turns on intended use

Section 201(h) of the Federal Food, Drug, and Cosmetic Act defines a device as an article “intended for use in the diagnosis of disease or other conditions, or in the cure, mitigation, treatment, or prevention of disease” (21 U.S.C. 321(h)). FDA’s regulation treats intent as objective and says it “may, for example, be shown by labeling claims, advertising matter, or oral or written statements” (21 CFR 801.4). The same code can be regulated differently depending on what its vendor claims.

The 21st Century Cures Act added section 520(o), which removes several categories of software from the device definition, including software “for administrative support of a health care facility” and certain clinical decision support (CDS) (21 U.S.C. 360j(o)). The administrative category gives examples such as “information about patient populations” and “population health management.” Clinical research is not among them, and FDA’s guidance on these exclusions does not discuss it (FDA).

FDA’s closest example places trial matching outside device rules

The most direct FDA statement is in its guidance on device software functions, last revised in September 2022 and still current (FDA). Its Appendix A list of “Software Functions that are NOT Medical Devices” includes software that helps “match patients with potentially appropriate clinical trials and facilitate communication between the patient and clinical trial investigators” (FDA). The example sits under patient education, whereas a site tool reads a specific patient’s notes and reports whether that patient appears to meet each criterion.

A January 2025 analysis by lawyers at Hogan Lovells (now Hogan Lovells Cadwalader) said that matching patients to studies using sources such as electronic health records “is not a use of AI we expect FDA to closely regulate,” because it “typically does not affect the design or conduct of the clinical trial” (Hogan Lovells, law firm analysis). The authors cautioned that software assigning participants to treatment groups based on disease state or severity “could potentially satisfy the definition of a medical device,” and that a tool could “venture into regulated territory by screening or diagnosing individual patients” for eligibility.

If a match counts as decision support, four criteria apply

A regulator could instead treat a match to an interventional trial as a recommendation about treatment. The CDS exclusion would then apply only if the software meets all four criteria of section 520(o)(1)(E): it must not analyze a medical image, a signal from an in vitro diagnostic device, or a pattern or signal from a signal acquisition system, must display or analyze medical information, must provide recommendations “to a health care professional,” and must enable that professional “to independently review the basis for such recommendations” so that the professional does not “rely primarily” on them (21 U.S.C. 360j(o)).

FDA revised its CDS guidance on January 6, 2026, replacing the 2022 version (Hyman, Phelps & McNamara, law firm blog), and reissued it on January 29, 2026 (FDA). The guidance defines a health care professional as an individual “licensed, registered, or certified by a State,” a test that a coordinator without such a credential would not meet. For independent review, the software or its labeling should “identify the required input medical information” and give “a plain language description of the underlying algorithm development and validation.” For a screening tool, these criteria favor showing the record text behind each criterion decision and sending every candidate to a qualified reviewer, with an investigator making the eligibility determination.

What would move a screening tool into device territory

Four kinds of function would raise the device question.

1. A tool that measures lesions on scans or reads ECG tracings would fail the first criterion if the analysis serves a medical purpose, and FDA lists ECG waveform analysis “to detect or diagnose arrhythmias” as a device function (FDA).
2. A tool that infers a diagnosis or severity classification absent from the record fits the “screening or diagnosing” scenario the Hogan Lovells analysis flagged, unlike extracting a documented diagnosis.
3. Real-time enrollment alerts for acute-care trials can resemble functions FDA treats as devices. FDA classifies software that detects “a life-threatening condition, such as stroke or sepsis” and alerts a clinician as a device function, and says functions for “a critical, time-sensitive task or decision” fail the fourth criterion.
4. A patient-facing tool that suggests a diagnosis or treatment would lose the CDS exclusion, because FDA states that functions providing recommendations to patients or caregivers “meet the definition of a device.”

Obligations that apply whether or not a tool is a device

FDA’s Part 11 regulations cover electronic records “created, modified, maintained, archived, retrieved, or transmitted” under FDA records requirements, and call for validation and “secure, computer-generated, time-stamped audit trails” (21 CFR Part 11). Whether Part 11 reaches a screening tool depends on whether the tool’s records serve as records FDA requires, such as the case histories drug-trial investigators must keep (21 CFR 312.62). FDA’s October 2024 final guidance on electronic systems in trials recommends risk-based validation that weighs a system’s potential “to affect the rights, safety, and welfare of participants or the reliability of trial results.” It also holds regulated entities that use IT service providers responsible for compliance and recommends a written agreement (FDA).

ICH E6(R3) reached Step 4 on 6 January 2025 and took effect in the EU on 23 July 2025 (EMA). It states that computerized systems used in trials “should be fit for purpose” (principle 9.3) and asks investigators to address its system requirements “proportionate to the risks to participants and to the importance of the data” for systems they deploy for trials (section 2.12.10) (ICH E6(R3)). ICH adopted Annex 2, on pragmatic and decentralized designs and real-world data, on 3 June 2026, with an EU effective date of 15 January 2027 (EMA), whereas FDA’s version remains a December 2024 draft (FDA).

FDA’s January 2025 guidance on AI to support regulatory decision-making for drugs and biologics, which FDA still lists as a draft (FDA), proposes a seven-step credibility assessment and excludes AI “used for operational efficiencies” that do not affect patient safety, drug quality, or the reliability of study results (FDA). A tool whose candidates staff review before any eligibility decision sits closer to that operational category. For multinational studies, the EU AI Act treats AI in medical devices that require third-party conformity assessment as high-risk (Regulation (EU) 2024/1689), and Regulation (EU) 2026/1744 moved high-risk obligations to 2 December 2027 for Annex III systems and 2 August 2028 for Annex I products (EUR-Lex). HIPAA rules for chart review before consent are covered in post 3.

Questions to ask a screening-software vendor

1. What is the written intended use, and does any material describe the tool as detecting or diagnosing disease?
2. Does the tool analyze images or device signals, or only text and structured data?
3. How was it validated on records like this site’s, and how are model updates revalidated?
4. Does a person review every candidate, and does each criterion decision show its source text?
5. Are there secure, time-stamped audit trails of who reviewed each match and what they decided?
6. Where are patient data processed and stored, who can access them, and what written agreement governs this?

How Centrova’s platform is designed

Centrova’s platform reviews the clinical notes and documents a site already holds and checks each patient against the inclusion and exclusion criteria of open trials. Coordinators and investigators review the candidates and make the final eligibility determination. It also includes patient tracking, communication and site coordination tools, and works across disease areas and with multiple EHR systems. Testing is free during development; contact us.

This post is general information, not legal or regulatory advice.

Sources

1. U.S. Code. 21 U.S.C. 321(h), definition of “device” (FD&C Act section 201(h)). Legal Information Institute, Cornell Law School. Accessed November 4, 2026. https://www.law.cornell.edu/uscode/text/21/321
2. U.S. Code. 21 U.S.C. 360j(o), regulation of medical and certain decision support software (FD&C Act section 520(o)). GovInfo, United States Code, 2023 edition. https://www.govinfo.gov/content/pkg/USCODE-2023-title21/html/USCODE-2023-title21-chap9-subchapV-partA-sec360j.htm
3. Electronic Code of Federal Regulations. 21 CFR 801.4, Meaning of intended uses. Up to date as of July 15, 2026. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-H/part-801/subpart-A/section-801.4
4. Electronic Code of Federal Regulations. 21 CFR Part 11, Electronic records; electronic signatures. Up to date as of July 16, 2026. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-A/part-11
5. Electronic Code of Federal Regulations. 21 CFR 312.62, Investigator recordkeeping and record retention. Accessed November 4, 2026. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-D/part-312/subpart-D/section-312.62
6. U.S. Food and Drug Administration. Policy for Device Software Functions and Mobile Medical Applications: Guidance for Industry and FDA Staff. September 28, 2022. https://www.fda.gov/media/80958/download (guidance page: https://www.fda.gov/regulatory-information/search-fda-guidance-documents/policy-device-software-functions-and-mobile-medical-applications)
7. U.S. Food and Drug Administration. Changes to Existing Medical Software Policies Resulting from Section 3060 of the 21st Century Cures Act: Guidance for Industry and FDA Staff. September 27, 2019. https://www.fda.gov/media/109622/download
8. U.S. Food and Drug Administration. Clinical Decision Support Software: Guidance for Industry and FDA Staff. January 29, 2026 (supersedes the version issued January 6, 2026). https://www.fda.gov/media/109618/download
9. Gonzalez SJ, Newberger JD. A busy day in the (CDRH) neighborhood: updates to the CDS and general wellness guidance documents. FDA Law Blog, Hyman, Phelps & McNamara, P.C. January 8, 2026. https://www.thefdalawblog.com/2026/01/a-busy-day-in-the-cdrh-neighborhood-updates-to-the-cds-and-general-wellness-guidance-documents/
10. Scott JK, Church R, Levine M, Henderson W, Lao B, Kass G, Massie L, Grey A. FDA’s evolving regulatory framework for AI use in drug & device clinical trials and research. Hogan Lovells. January 13, 2025 (now hosted on the Hogan Lovells Cadwalader website). https://www.hlc.com/en/publications/fdas-evolving-regulatory-framework-for-ai-use-in-drug-device-clinical-trials-and-research
11. U.S. Food and Drug Administration. Electronic Systems, Electronic Records, and Electronic Signatures in Clinical Investigations: Questions and Answers. Guidance for Industry. October 2024. https://www.fda.gov/media/166215/download
12. International Council for Harmonisation. ICH E6(R3) Guideline for Good Clinical Practice (GCP), Step 5 (EMA/CHMP/ICH/135/1995). Step 4 endorsement January 6, 2025; effective in the EU July 23, 2025. https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e6-r3-guideline-good-clinical-practice-gcp-step-5_en.pdf
13. European Medicines Agency. ICH E6 Good clinical practice: scientific guideline. Accessed November 4, 2026. https://www.ema.europa.eu/en/ich-e6-good-clinical-practice-scientific-guideline
14. U.S. Food and Drug Administration. E6(R3) Good Clinical Practice; International Council for Harmonisation; Guidance for Industry; Availability. Federal Register. September 9, 2025. https://www.federalregister.gov/documents/2025/09/09/2025-17311/e6r3-good-clinical-practice-international-council-for-harmonisation-guidance-for-industry
15. U.S. Food and Drug Administration. E6(R3) Good Clinical Practice: Annex 2. Draft guidance. December 2024. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-annex-2
16. U.S. Food and Drug Administration. Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products. Draft guidance. January 2025. https://www.fda.gov/media/184830/download (guidance page: https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drug-and-biological)
17. European Parliament and Council of the European Union. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. July 12, 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
18. European Parliament and Council of the European Union. Regulation (EU) 2026/1744 of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI). Official Journal of the European Union. July 24, 2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj