NMPA CMDE Releases Two Draft Guidance Documents on Clinical Evaluation for Medical Devices

On September 16, 2026, the NMPA Center for Medical Device Evaluation released two draft guidance documents on clinical evaluation and registration review for medical devices, open for public comment until October 14, 2026. Comments should be submitted by email using the feedback form, with the subject line formatted as “[Guidance Name] (Draft for Comments) Comment Feedback + Feedback Organization Name.” Specific channels are available in the original announcement.

The Guidance on Clinical Evaluation and Registration Review of Artificial Intelligence Multi-Disease Assisted Decision-Making Medical Devices (Draft for Comments) applies to assisted triage/referral, assisted detection, and assisted diagnosis products, with classification code 21-04-02 and Class III management. Key requirements include: assisted decision-making functions generally require clinical trials, usually diagnostic clinical trials; same-variety comparison may be considered when the core algorithm remains unchanged; for independently operating multi-module combinations, existing clinical trial evidence may be cited if the core algorithm has completed compliant trials in another product of the same company. Multi-disease products need separate primary endpoints for each target disease. When imaging features are highly similar and differential diagnosis is required, differential diagnostic performance among target diseases should also be evaluated, which may be verified through well-controlled test databases. Instructions for use should note possible false negatives/positives, that the product cannot replace pathological diagnosis, and that differential diagnostic performance not verified in clinical trials should be clearly stated.

The Guidance on Clinical Evaluation and Registration Review of Implantable Deep Brain Stimulation Systems (Draft for Comments) applies to DBS systems, with classification codes 12-02-01 and 12-02-02 and Class III management. Indications typically include Parkinson's disease, tremor, epilepsy, and dystonia. These products generally require clinical evaluation through clinical trials, with separate trials for different indications. Trials are generally prospective, multicenter, randomized, double-blind, parallel-controlled, with sham stimulation as control. For additions based on predecessor products, such as directional electrodes or MRI compatibility, the same-variety pathway may be considered if non-clinical data fully demonstrate no adverse impact on safety and effectiveness. The appendices provide clinical trial considerations for Parkinson's disease, tremor, epilepsy, and dystonia, covering design type, inclusion/exclusion criteria, endpoints, follow-up time, and sample size examples.