Efficacy and Safety of the Felix NeuroAI Wristband for Essential Tremor: Results of a 1‐Year Open‐Label Extension Trial

Abstract

Background

Essential tremor (ET) impacts daily activities and quality of life. A pivotal study of a transcutaneous peripheral nerve stimulation (TPNS) device, the Felix NeuroAI Wristband, led to Food and Drug Administration (FDA) clearance in 2025.

Objective

The aim of the study was to report 1‐year follow‐up results.

Methods

Patients who completed the 90‐day randomized phase could continue until 1 year after the baseline visit; those in the sham group were crossed over to active TPNS treatment (delayed‐start TPNS). Outcome measures were assessed at 180 days and 1 year: modified Activities of Daily Living (mADL) subscale of TETRAS, Clinical and Patient Global Impression of Improvement (CGI‐I and PGI‐I), Quality of Life in Essential Tremor Questionnaire, and adverse device effects (ADEs).

Results

Of 93 patients who entered the extension phase (58 TPNS and 35 delayed start), 57 (38 and 19, respectively) completed 1 year of follow‐up. In the TPNS group, mADL improvement increased from 6.9 points at day 90 to 9.7 points at 1 year. Ninety days of active treatment improved mADL scores by 5.5 points in the delayed‐start TPNS group (day 180). From baseline, TETRAS ADL scores in the TPNS group improved by 6.1 ± 6.0 points at 90 days and 9.3 ± 6.0 points at 1 year. CGI‐I, PGI‐I, and quality of life improved in both groups from 90 days to 1 year. Twenty‐three new ADEs occurred (mild skin irritation, n = 22; muscle spasms, n = 1).

Conclusions

TPNS use was associated with sustained functional improvement for up to 1 year with acceptable safety. Patients who initiated active therapy earlier demonstrated greater cumulative benefit.

Essential tremor (ET) is common, but treatment options remain limited. The involuntary oscillatory movements of ET can limit many activities of daily living, including eating, writing, and hygiene activities, and can be disabling. Propranolol and primidone are among the medications typically used for first‐line treatment, but they are not tolerated by many patients, and discontinuation rates are high. ¹ , ² Surgical interventions, including focused ultrasound ablation and deep brain stimulation, can be effective but are not viable options for many patients because of concerns about the risks.

Noninvasive devices worn on the wrist can provide stimulation of peripheral nerves that alters central thalamic–cerebellar–cortical circuitry to improve tremor. We previously reported a pivotal randomized double‐blinded trial demonstrating efficacy, safety, and tolerability of a wrist‐worn transcutaneous peripheral nerve stimulation (TPNS) device, the Felix NeuroAI Wristband, that delivers stimulation driven by cloud‐based artificial intelligence (AI) and continuously makes adjustments based on patient activity and predictive analytics. The active TPNS treatment significantly improved activities related to upper limb tremor and quality of life at 90 days when compared to a sham device that produced random vibrations. On the basis of these results, the device received Food and Drug Administration (FDA) clearance. ³ Here we report the results of an open‐label extension of the pivotal trial, in which those who received active treatment were maintained on active TPNS (early‐start group), those who received the sham treatment in the randomized phase were switched to active TPNS (delayed‐start group), and both groups were followed through 1 year after the original baseline visit.

Patients and Methods

Study Design and Participants

As described in the prior publication, ⁴ a prospective, randomized, sham‐controlled, double‐blinded trial was conducted at 12 sites in the United States (8 sites) and China (4 sites) to evaluate the efficacy and safety of TPNS used to improve upper‐limb tremor in adults with ET. The protocol was approved by the Institutional Review Board of each participating site, and the trial was registered on Clinicaltrials.gov (NCT06235190). All participants provided written informed consent.

Adults (≥18 years of age) with upper‐extremity tremor and a clinical diagnosis of ET were eligible for the study. To be included, patients needed a tremor severity score of 2 or higher on at least one of the 6 performance subscale (PS) tasks (forward outstretch, lateral wing‐beating, kinetic, spiral drawing, handwriting, and dot approximation) on the Tremor Research Group Essential Tremor Rating Assessment Scale (TETRAS), ⁵ a total PS score ≥7, and familiarity with operating a smartphone and connecting to Wi‐Fi at home. See the prior publication ⁴ for additional inclusion and exclusion criteria.

After the randomized phase ended at 90 days, an open‐treatment extension phase was conducted that lasted for a total of 1 year of follow‐up, during which subjects initially assigned to sham treatment were crossed over to active TPNS. The patients were assessed at 180 and 360 days of treatment (Fig. 1), which corresponds to 90 and 270 days of active treatment for the delayed‐start group.

FIG. 1.

Study design. ADE, adverse device event; mADL, modified Activities of Daily Living subscore; CGI‐I, Clinical Global Impression of Improvement; PGI‐I, Patient Global Impression of Improvement; QUEST, Quality of Life in Essential Tremor Questionnaire; TPNS, transcutaneous peripheral nerve stimulation.

Study Device

The TPNS system (Felix NeuroAI Wristband; Fasikl Inc.) is a wearable neuromodulation device that stimulates the radial, median, and ulnar nerves. ⁶ The device records movement signals from the patient's wrist through an inertial measurement unit with a triaxial accelerometer and gyroscope. A cloud‐based AI algorithm based on a deep neural network extracts relevant parameters streamed from the device and continuously adjusts stimulation settings, such as pulse counts, stimulation sequence (between nerves), and stimulation intensity, in real time to provide personalized targeting of the individual's tremor. It is designed to be worn by the patient on the wrist throughout the waking day.

The TPNS device is placed over the radial, median, and ulnar nerves by using each individual's anatomic landmarks. At fitting, the intensity for each of the 3 electrodes is increased as tolerated but kept below levels that would be uncomfortable. The device can be worn on either hand or used on both hands with two devices; the decision of which hand(s) to treat was left to the investigator and patient.

During the initial blinded phase, participants were randomized 2:1 to receive either the TPNS device or a sham treatment, stratified by the use of ET medications and severity of tremor. The sham device was identical in appearance and weight to the active device but generated brief, random vibrations (on average once every 5 minutes) in place of electrical stimulation. Patients in the sham group wore the same electrode band and controlled the device in the same way (via the smartphone app) as those in the TPNS group. All patients and evaluators were blinded as to the treatment assignment.

Assessments and Outcomes

Daily usage was measured with an intention‐to‐treat analysis that included all days until withdrawal or end of the study.

TETRAS comprises the Activities of Daily Living (ADL) subscale and PS, which are administered by the investigator with pen and paper. ⁷ Each item of the subscales has a rating of 0 to 4, with 4 being the highest amount of tremor.

The modified ADL (mADL) score is the sum of items 2 to 11 of the TETRAS ADL subscale (feeding with a spoon; drinking from a glass; hygiene; dressing; pouring; carrying food trays, plates, or similar items; using keys; writing; working; overall disability with the most affected task) plus item 6 (Archimedes spirals, bilateral) and item 7 (handwriting, dominant hand) of the TETRAS PS.

The outcomes of the open‐access extension phase for the TPNS group were prespecified in the protocol: the change in mADL from baseline to 180 days and 1 year; the Patient Global Impression of Improvement (PGI‐I) score at 180 days and 1 year; and the Clinical Global Impression of Improvement (CGI‐I) score and Quality of Life in Essential Tremor Questionnaire (QUEST) at 1 year. For the delayed‐start TPNS group, due to the initial 90 days of sham treatment, the same outcomes were reported from baseline to 90 days (where available) and 270 days of TPNS treatment, which correspond to 180 days and 1 year in calendar days. Safety end points included the types and frequencies of device‐related adverse events (ADEs) and serious ADEs. Only those who completed the trial were included in the analyses.

Results

Participants and Treatment

The study initially randomized 125 patients, of whom 98 completed the randomized 90‐day evaluation period. Of these, 93 patients elected to continue in the open‐label extension phase (58 in the active TPNS group who stayed on TPNS therapy [“early‐start” group] and 35 in the sham TPNS group who switched to TPNS therapy [“delayed‐start” group]) (Fig. 2). Overall, 38 in the early‐start TPNS group and 19 in the delayed‐start group completed assessments at 1 year.

FIG. 2.

Subject disposition. ADE, adverse device event; SAE, serious adverse event; TPNS, transcutaneous peripheral nerve stimulation.

In the extension phase, the device was worn an average of 7.9 ± 2.1 awake hours per day in the early‐start TPNS group and 8.4 ± 1.9 awake hours per day in the delayed‐start TPNS group, similar to the duration during the randomized phase. These values did not differ meaningfully between those who did and did not complete the 1‐year follow‐up. Stimulation intensity in the early‐start TPNS group was very similar or slightly lower during the extension phase to what it had been during the randomized phase (Table 1). Stimulation intensity in the delayed‐start TPNS group was very similar to that in the early‐start TPNS group during the extension phase for all three nerves (Table 1).

TABLE 1.

Stimulation intensity during the randomized and continued‐access phases

Nerve	Stimulation (mA, mean ± SD)
	TPNS	TPNS	Delayed‐start TPNS
	Baseline to 90 days	90 days to 1 y	90 d to 1 y
Ulnar	4.5 ± 1.4	4.5 ± 1.3	4.3 ± 1.1
Radial	4.1 ± 1.3	3.9 ± 1.2	3.9 ± 1.3
Median	4.0 ± 1.3	3.7 ± 0.9	3.8 ± 1.1

Abbreviations: SD, standard deviation; TPNS, transcutaneous peripheral nerve stimulation.

Outcomes

After 90 days of active stimulation (ie, day 90 in the early‐start TPNS group and day 180 in the delayed‐start TPNS group), the decreases in mADL were 6.9 points in the early‐start TPNS group and 5.5 points in the delayed‐start TPNS group (Fig. 3).

FIG. 3.

Change in modified Activities of Daily Living (mADL) subscore after 90 days of active TPNS treatment. Decrease in mADL from baseline to 90 days of active treatment (day 90 in the TPNS group and day 180 in the delayed‐start group). TPNS, transcutaneous peripheral nerve stimulation.

In the early‐start TPNS group, mADL scores continued to improve during the extension phase, with the change increasing from 6.9 points at 90 days to 9.7 points at 1 year (Fig. 4). In the delayed‐start group, some improvement was observed during initial sham treatment, with greater improvement after sham treatment was switched to active TPNS. The magnitude of improvement after the switch was greatest between 90 and 180 days from baseline (Fig. 4).

FIG. 4.

Change in mADL over time. Change from baseline is shown, based on paired data for each patient in the study at each time. Solid blue line: TPNS, Solid orange line: Sham. Patients who received sham treatment in the randomized phase were crossed over to active treatment after 90 days. mADL, modified Activities of Daily Living; TPNS, transcutaneous peripheral nerve stimulation.

At 90 days, the TETRAS ADL subscale score of those in the early‐start TPNS group had improved 6.1 ± 6.0 points from the baseline value of 26.4 ± 6.6; the total improvement at 1 year was 9.3 ± 6.0 points. The improvement in the ADL subscale score in the delayed‐start TPNS group at 180 days, after 90 days of active treatment, was 4.3 ± 4.2 points. Improvements were seen in both groups on each individual item.

The PS changed little in the early‐start TPNS group between 90 days and 1 year (from −2.7 ± 2.3 at 90 days to –2.7 ± 2.5 at 1 year for the dominant hand total). After crossing over to active TPNS, the delayed‐start group saw small improvements on many of the individual items of the PS between 90 days and 1 year, with a decrease in the dominant hand total at 1 year of 2.6 ± 2.2. The PS was not measured at 180 days.

On the CGI‐I at 90 days, 43 of 62 (69.4%) participants in the early‐start TPNS group and 16 of 36 (44.4%) participants in the delayed‐start (sham‐treated) group showed some improvement. At 1 year, this increased to 78.9% in the early‐start TPNS group (n = 30) and 68.4% in the delayed‐start TPNS group (n = 13) (Fig. 5A).

FIG. 5.

Clinical and Patient Global Impressions of Improvement. (A) Percentages of patients showing the indicated amount of improvement on the Clinical Global Impression of Improvement at 90 days and 1 year. (B) Percentages of patients showing the indicated amount of improvement on the Patient Global Impression of Improvement at 90 days and 1 year. The percentage in each group indicating minimal, much, or very much improved is indicated at the bottom as “any improvement.” TPNS, transcutaneous peripheral nerve stimulation.

On the PGI‐I, the percentage with any improvement after 90 days of active treatment was 67.7% in the early‐start TPNS group (at 90 days), with additional improvement at 1 year. In the delayed‐start treatment group, the percentage with any improvement changed from 50.0% at 90 days to 63.2% at 1 year (270 days of active treatment) (Fig. 5B).

At baseline, the QUEST total activity score was 40.2 ± 20.2 in the early‐start TPNS group, which improved by 6.2 ± 15.2 at 90 days and 8.4 ± 20.2 at 1 year. In the delayed‐start TPNS group, the QUEST had improved by 2.0 ± 9.0 at 90 days from a baseline of 40.0 ± 21.7. At 1 year (270 days of active treatment), it had improved by 2.9 ± 9.3 over baseline.

Safety

Twenty‐three ADEs occurred during the extension phase compared to 53 during the randomized phase. Of those in the extension phase, 15 were in the early‐start TPNS group (9 participants) and 8 in the delayed‐start TPNS group (7 participants); 22 were cases of skin irritation, which required either no treatment (16 cases) or topical treatment (6 cases). As occurred during the randomization phase, most of the cases (14 events, 11 participants) occurred in China. Of the 7 other ADEs, only 1 occurred after 90 days: a case of muscle spasms in the early‐start TPNS group.

Discussion

The initial improvements in tremor‐related disability demonstrated with TPNS treatment in the pivotal 90‐day trial were maintained for up to 1 year of TPNS treatment, and patients who crossed over from sham to active treatment at 90 days demonstrated clinically meaningful improvement after initiation of active TPNS treatment. The treatment was generally well tolerated with mild skin irritation as the most common side effect.

The continued improvement over time without an increase in the stimulation intensity suggests that the electrical stimulation parameters were unlikely to have played a role in the ongoing reduction in symptoms. The findings could reflect the ongoing AI‐driven adjustments in stimulation, leading to long‐term improvement in central circuitry, or the cumulative effect of sustained stimulation. Future mechanistic physiological studies could enhance our understanding of the increasing benefit over time.

To better understand the continued improvement over time, we further evaluated the two groups from an “early” versus “delayed” TPNS start perspective. Although the delayed‐start group had marked symptomatic improvement on the mADL after 90 days (similar in magnitude to the symptomatic benefit seen in the first 90 days of treatment in the active TPNS group), these improvements did not numerically “catch up” with those of the early‐start TPNS group by 1 year, despite both groups maintaining active TPNS. However, the limitations of the study design (open‐label extension with delayed‐start crossover, no sustained randomized comparator) and the attrition rate limit our ability to interpret this finding. Prior experience with a placebo can also affect the response to a subsequent active treatment, ⁸ as has been reported in neurological studies, including those in Parkinson's disease. ⁹ , ¹⁰ However, it is notable that the sham device is not a purely inert placebo. Stimulation leading to alternative neural activity patterns in the peripheral nerve may have some effect on tremor. The wristband of the sham device can exert pressure and a tightening sensation. The device also delivers infrequent vibration, and we are not yet fully aware of how vibration frequency relates to treatment outcomes. After 90 days, if the sham provides some therapeutic effect beyond placebo, the crossover may represent a transition from a less‐optimal treatment to a more effective one, which could influence patient experience and reported outcomes.

Mild skin irritation remained the most common ADE; however, its frequency decreased substantially beyond the initial 90‐day period. The reason for the higher rate of skin irritation in the Chinese patients is not clear and may relate to factors such as skin sensitivity, environmental conditions, device placement, or site‐specific training; this warrants further evaluation.

Limitations of the study include its open‐label design in the extension phase, which increases the possibility of expectation and reporting bias. In addition, attrition over 1 year was substantial, with only 61% of those in the TPNS group and 53% of those in the delayed‐start group completing the final visit. This attrition was primarily driven by voluntary withdrawals, and only 2 patients, both in the delayed‐start group, withdrew from the extension study due to adverse events. Device usage (hours per day) did not differ meaningfully between those who did and did not complete the study; however, patients experiencing greater benefit are more likely to remain in a trial, potentially leading to overestimation of long‐term efficacy. In addition, attrition during the extension phase was higher in the delayed‐start group than in the TPNS group, possibly leading to selection biases and variability in long‐term results. Furthermore, the study was not powered to examine secondary end points such as the improvement scales, so those results must be interpreted cautiously. Finally, the study was not formally designed as a delayed‐start trial, so only descriptive data can be presented.

In conclusion, our results demonstrate that full‐day, AI‐driven TPNS safely and effectively improved ET over 1 year of use, with symptom benefit continuing to increase over time, although early use of TPNS provided greater benefit than delayed TPNS at 1 year. These findings further support the role of noninvasive neuromodulation as a therapeutic option for patients with ET. Continued optimization of the AI algorithm driven by increasing data volume and quality, as well as improvements in the underlying software and product design, could lead to further performance gains with future versions of the device.

Author Roles

(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique; (4) Funding acquisition.

S.H.I.: 1A, 1B, 2A, 3A.

Y.H.: 1B, 3B.

W.G.O.: 1C, 3B.

W.L.: 1C, 3B.

X.Z.: 1C, 3B.

Y.Y.: 1C, 3B.

A.J.E.: 1C, 3B.

D.K.: 1C, 3B.

S‐H.K.: 1C, 3B.

S.B.: 1C, 3B.

H.A.S.: 1C, 3B.

K.E.L.: 1C, 3B.

Z.Y.: 1A, 3B, 4.

Q.Z.: 2C, 3B.

Z.Z.: 1A, 2A, 2B, 3A.

R.P.: 1A, 1B, 2C, 3A.

Disclosures

Ethical Compliance Statement: The study was approved by WCG IRB and the corresponding IRB at each site: Barrow Neurological Institute, Phoenix, AZ, USA; Parkinson's Research Centers of America, Palo Alto, CA, USA; Parkinson's Disease and Movement Disorders Center of Boca Raton, FL, USA; The University of Kansas Medical Center; Parkinson's Research Centers of America—Long Island, Commack, NY, USA; Columbia University Irving Medical Center, New York, NY, USA; University of Cincinnati, Cincinnati, OH, USA; Xuanwu Hospital Capital Medical University, Beijing, China; Tianjin Medical University General Hospital, Tianjin, China; The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China; and Sir Run Run Shaw Hospital, Hangzhou, Zhejiang, China. Patients provided written informed consent at each study site. This process included a verbal discussion with the patient on all aspects of the clinical investigation that were relevant to the patient's decision to participate, such as details of clinical investigation procedures, anticipated benefits, potential risks of clinical investigation participation, and the right to withdraw at any time. The site provided the patient with the informed consent form written in a language understandable to the patient that had been approved by the center's IRB/EC. The patient had adequate time to review, ask questions, and consider participation. If the patient agreed to participate, they signed and dated the informed consent form, along with the person obtaining the consent prior to any clinical investigation‐specific procedures. The site filed the signed original in the patient's hospital or research charts and provided the other original to the patient. We confirm that we have read the journal's position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Financial Disclosures and Conflicts of Interest: Author disclosures are available in the Supporting Information.

Supporting information

Data S1: Supporting information.

Data Availability Statement

Research data are not shared.