Does FDA accept clinical trial data from China?

The short answer: FDA’s regulations have never said “no clinical data from China.” 21 CFR 812.28 and 21 CFR 312.120 are clear: FDA can accept data from a clinical study conducted outside the US if it follows good clinical practice (GCP), is approved by a local ethics committee (IRB/IEC), protects the rights of participants, and the data are traceable.

FDA page: Acceptance of Data from Clinical Investigations for Medical Devices

The regulatory basis

21 CFR 812.28 — Acceptance of data from clinical investigations conducted outside the United States

Applies to medical devices. The core requirements: a study conducted outside the US must follow GCP, be reviewed and continuously overseen by an independent ethics committee (IEC), and obtain informed consent from participants (which an IEC may waive in special circumstances).

21 CFR 312.120 — Foreign clinical studies not conducted under an IND

Applies to drugs and biologics. The same logic: the study must follow GCP, and FDA must be able to verify the data, for example through on-site inspection.

Both regulations set out the same principle: whether data are acceptable depends on compliance and verifiability, not on where they come from.

Where the real differences lie

Diversity action plans

Since the Food and Drug Omnibus Reform Act (FDORA) took effect in 2023, FDA has had a statutory basis to require representative trial populations. Under FDORA Section 3602, FDA issued the draft guidance Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies on 26 June 2024 (Docket No. FDA-2021-D-0789; draft, non-binding recommendations).

For reference: FDA’s focus on representativeness is consistent and systematic. In August 2023 it also issued draft guidance on Postmarketing Approaches to Obtain Data on Populations Underrepresented in Clinical Trials (FDA-2022-D-2629), under which drugs and biologics with under-representative premarket populations must fill the gap through postmarketing requirements or commitments (PMR/PMC).

What this means for data from outside the US, including China: If 100% of the submitted data come from one country’s population, FDA may well find the data lack representativeness and do not meet enrollment goals, which limits how far the results can be extrapolated to the multi-ethnic US population. This is not “Chinese data are never accepted”; it is “data from a single population have limited explanatory power.” Multi-center, multi-ethnic designs can address it.

Standard of care

FDA compares routine medication and post-operative care at sites outside the US with US practice. If the context of use does not transfer, the data cannot be accepted directly.

Key step: Explain at the protocol design stage why the care pathways are comparable, and run sensitivity analyses where needed.

BIMO inspections at overseas sites

FDA is sending inspectors to overseas clinical sites under its Bioresearch Monitoring (BIMO) program noticeably more often. A single non-compliant medical record, or an informed consent form whose signing date does not match, can call the reliability of the whole data set into question.

Note: What matters is data quality, not data origin.

A way forward: Bayesian statistics

FDA’s Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials provides a legitimate, quantifiable way to borrow data from outside the US.

Section 5: Sources of prior information

Sections 5.1–5.3 list external data that can be turned into prior information:

clinical studies conducted outside the US;
historical clinical data on similar marketed products (predicate or similar devices);
data on the same product from other clinical stages or indications (prior pilot studies).

The logic: If the external data are highly similar to the planned US study in clinical design, patient population and disease definition, they can be turned into a prior probability distribution using mathematical models (power priors, hierarchical models) and carried into the US submission.

Section 4.1: Minimizing sample size and study duration

The guidance states: “If reliable prior information is available, a Bayesian study can often achieve the same statistical power with a smaller sample size than a non-Bayesian study.”

In statistical terms: posterior = prior + current data. This is the direct regulatory basis for using data from outside the US as the prior, with a small US bridging study.

Section 6: Borrowing information rigorously

Section 6.2 (borrowing via hierarchical models and power priors) describes how FDA reviews external data and discounts it where needed:

No copying and pasting: Data from China or Europe cannot simply be pooled into a US submission; dynamic borrowing is required.
Similarity sets the weight: If interim results from the small US study (current data) agree closely with the overseas data (prior data), the model gives the overseas data more weight. If they diverge, the model discounts them and borrows less.

In short: overseas data as the prior + a small US study as current data = a Bayesian posterior.

When does a 510(k) require clinical data?

Under FDA’s Recommendations for the Use of Clinical Data in Premarket Notification [510(k)] Submissions, a substantial equivalence (SE) submission generally does not need clinical data; non-clinical safety and performance data are enough. This is the least burdensome principle.

Clinical data are requested only when non-clinical data are insufficient or unreliable, in four situations:

Scenario 1 | Differences in intended use Differences in patient population, disease, anatomical site, structure or pathology; a move from a general to a specific use; an expansion of a cleared intended use; or an unknown or different benefit-risk profile for the proposed intended use.

Scenario 2 | Differences in technological characteristics Significant changes in materials, device design, energy source or other device characteristics (examples 2-A to 2-C).

Scenario 3 | Non-clinical testing cannot establish SE No analytical, bench or animal model is available; available models have limitations; models cannot predict clinical outcomes; or there are species-specific anatomical or pathophysiological issues that depend on clinical evidence.

Scenario 4 | New or increased risks identified for the predicate Sources include voluntary adverse event reports, literature, Section 522 postmarket surveillance studies, recalls and other real-world data.

Scenario 4 can go two ways:

No clinical data needed: The predicate has known failures or risks, but FDA judges that more extensive non-clinical bench and performance testing plus labeling controls (warnings, restrictions on use) adequately control the risk (example 4-A).
Clinical data required: (1) the predicate’s risks are serious and non-clinical testing cannot show they have been eliminated or reduced; (2) FDA has issued requirements for that device type, such as a Section 522 order or device-specific guidance; or (3) the defect is directly linked to serious patient harm and, after a design change, non-clinical data cannot show safety and effectiveness (examples 4-B and 4-C).

Tip: If another legally marketed predicate exists without the same new risks, FDA recommends using that one instead, which avoids the extra data requirement at the source. In every case, FDA explains in writing why it is asking for clinical data, and follows the least burdensome principle.

FAQ on accepting clinical data from outside the US

FDA’s Acceptance of Clinical Data to Support Medical Device Applications and Submissions: Frequently Asked Questions is non-binding guidance that interprets 21 CFR 812.28 and applies one standard to data from inside and outside the US.

Status: Non-binding; it reflects FDA’s current thinking and has no legal force. Companies may use an equivalent alternative approach if they justify it.
History: In 2018 FDA amended 21 CFR 807, 812 and 814 to align the rules for accepting clinical data from inside and outside the US. All clinical studies supporting an IDE, 510(k), De Novo, PMA or HDE, wherever conducted, must follow GCP.
Three elements of GCP: (1) approval by an independent ethics committee (IEC) before the study starts, with continuing review; (2) voluntary, signed informed consent from participants; (3) in certain emergency situations informed consent may be waived, with prior IEC approval.
Compliance statements by location: Studies in the US must state compliance with 21 CFR 50 (protection of human subjects), 21 CFR 56 (IRBs) and 21 CFR 812 (IDE regulation). Studies outside the US must state compliance with GCP; if not fully compliant, the sponsor must explain why in writing and describe corrective steps, or formally request a waiver.
Multinational, multi-center studies: Each site follows local regulations, and the submission explains this as a whole.

Note: The amended rule took effect one year after final publication, and applies to submissions for all types of medical devices.

Conclusion

In the regulations, the door has always been open. What holds Chinese data back is never geography. It is three things: how representative the population is, whether the standard of care is comparable, and whether the data hold up to inspection. Address all three, and data from outside the US can serve as a Bayesian prior supporting a smaller US bridging study.

Which of the clinical data scenarios above applies to your device? Contact us and we can help you assess whether you need US clinical data.

Sources

Federal regulations

21 CFR § 812.28 Acceptance of data from clinical investigations conducted outside the United States — current version
21 CFR § 312.120 Foreign clinical studies not conducted under an IND — current version

FDA guidance

The 510(k) Program: Evaluating Substantial Equivalence in Premarket Notifications [510(k)] — FDA CDRH, July 2014, final
Recommendations for the Use of Clinical Data in Premarket Notification [510(k)] Submissions — FDA CDRH, September 2023, draft
Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies — FDA OCE / CBER / CDER, June 2024, draft
Postmarketing Approaches to Obtain Data on Populations Underrepresented in Clinical Trials for Drugs and Biological Products — FDA OCE, August 2023, draft
Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials — FDA CDRH & CBER, February 2010, final

Legislation

Food and Drug Omnibus Reform Act of 2022 (FDORA), enacted as part of the Consolidated Appropriations Act, 2023 (Public Law 117-328) — signed December 2022

First published on our WeChat official account ↗

This article is general information, not regulatory or legal advice for a specific device. Requirements change; check the official sources and the date above.