--- title: "The Brain-Implant Threshold: Four Cases First, Two Years Later" date: 2026-11-05 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-bci-standard-03-1105 language: en --- # The Brain-Implant Threshold: Four Cases First, Two Years Later > On October 22, a notice appeared on the website of the medical device evaluation center under the National Medical Products Administration, numbered 2026 No. 29. The title is long: Technical Review On October 22, a notice appeared on the website of the medical device evaluation center under the National Medical Products Administration, numbered 2026 No. 29. The title is long: Technical Review Points for Clinical Trials of Implantable Brain-Computer Interface Systems for Hand Motor Function Compensation. It is hard to read out in one breath. The title is long, but the content comes down to two numbers: four cases, two years. The first is the usual starting sample for a feasibility trial; the second is the follow-up period recommended after the product reaches the market. Those two numbers turn the question that brain-computer interface companies have been asked most often, and answered least well, into a clause you can check against: how many people, and for how long, before you can talk about approval and hospital adoption. ## The Document Covers Only One Product Type First, what it covers. It applies to products implanted outside the dura, decoding local field potentials, with a low channel count, implanted in the hand area of the cortex, and using a pneumatic glove to assist grasping. The boundary is drawn tightly. For "low channel count," the document gives a reference of fewer than 64 channels. That parenthesis matters: it tells companies this set of requirements is tailor-made for low-channel, epidural designs, not for every brain-computer interface. It also says what it does not cover: products implanted under the dura, inside the cortex, deep in the brain or inside cerebral blood vessels; products that control wheelchairs or computer cursors; and patients outside the cervical spinal cord injury group, even when the lesion looks similar. In other words, one document maps to one technical profile. If you are developing a high-channel intracortical system, you cannot treat these review points as your passport; you need another route, or the next document. ## Four Cases First, Then Confirmation The document's wording is that a feasibility trial is usually recommended before a confirmatory trial. Note the words "usually" and "recommended": they leave room for individual cases, and for the reviewer's judgment. A feasibility trial is a scouting trip: can the electrode be placed, can signals be captured, can the patient tolerate it? Four cases is a very low floor at which regulators believe safety can begin to be seen, low enough that some companies will find it insufficient. The confirmatory trial is the real exam, where sample size, controls and endpoints follow statistics. The document gives a worked example: a target response rate of 50 percent, an expected rate of 75 percent, which yields a minimum of 32 cases. Those 32 cases rest on assumptions: a one-sided test, 80 percent power, and a 10 percent dropout rate. Change a parameter and the number changes. So it is an example, not a national cutoff, and copying it blindly invites error and misreading. ## Two Years Is a Recommendation, Not a Gate On follow-up, the text says follow-up is recommended to be no less than two years after implantation. It is a recommendation, not a mandate, and it governs post-market research rather than a pre-market gate. The two are a full registration cycle apart. Why two years? Because what electrodes fear most is not the day of surgery but six months or a year later: rising impedance, worsening signals, packaging leaks, and glial cells slowly wrapping the electrode. Any one of these can retire a system early. The document requires the post-market study report to be submitted at registration renewal. That means a certificate is not for life: it is renewed every five years, and each renewal requires the follow-up data from this cohort to be laid out again, with every missing case explained. ## How Many Trips Patients Make The document counts cases and years. For a patient, it means trains and stacks of test sheets. The primary endpoint is assessed at six months after implantation, and follow-up continues after that, at one year and two years. Follow-up is not a blood draw. Impedance is measured, the decoder is recalibrated, rating scales are administered, and imaging checks whether the electrode has moved. One visit takes half a day to a full day, a family member has to take leave, and those who live far away stay overnight. So four cases and two years are not only regulatory metrics; they are two years in the lives of four families. The document will not say this, but every clinician running such a trial knows it, and so does every family member who signed. ## There Is Also a Reference Specification Around the same time, another document was circulating: the Reference Specification for Brain-Computer Interface Technology Standards and Industrial Applications (1.0), recently released online, covering everything from terminology to scenarios. It was developed jointly by experts from four perspectives: product development, industrial investment, industrial application and academic research. It contains a terminology set, ten technical modules and seven application scenarios, running from electrodes all the way to data delivery. But to be clear: it is a reference specification, not a national standard, and not mandatory. National Business Daily flagged exactly this in its analysis, and that caveat matters more than the specification itself, because the market often describes it as a national standard. A reference specification and the review points are two different things. The first is an industry norm worked out among companies; the second is the regulator's review yardstick. One relies on self-discipline, the other is tied to a registration certificate, and conflating them leads to badly misjudged timelines. ## The Bigger Picture Pull back further. According to Xinhua, brain-computer interfaces were written into the Government Work Report for the first time this year and named as one of the future industries to be cultivated. The same report said China had approved the world's first invasive brain-computer interface medical device for market. That phrase comes from Xinhua, not from a company press release. The indication is compensation for hand grasping function. According to Economic Daily, four months later the first prescription for that certificate was written in Shanghai, and the first implantation was completed at Huashan Hospital affiliated with Fudan University. From approval to bedside took four months, which is fast. Also from Xinhua: China now has more than 200 brain-computer interface companies, and by 2040 the market is expected to reach 120 billion yuan. The second figure is a forecast, not something that has happened, and it is best cited with its source attached. The report also described a scene in Tianjin: stroke patients with hemiplegia wearing non-invasive devices for rehabilitation, and a headband in the bedroom recording brain waves to produce a sleep report in the morning. None of that requires surgery. ## After the Threshold Is Fixed Companies used to dread the question: is this a device or an experiment? Now at least one product category can point to a document and answer: four cases first, then 32, then two years of follow-up after launch. Fixing the threshold is not necessarily bad. It gives investors a ruler for comparison, hospitals a basis for ethics review, and patients a rough sense of how many gates stand behind the consent form they sign. For technology routes outside this document, the answer is equally clear: with no ready-made manual, you negotiate one door at a time. That road is slower, more expensive, and has no shortcut. **Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.** --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-bci-standard-03-1105 Free to quote with attribution and a link to the original.