ISSUE 011 · Oct 13, 2026 · 9 min read

What if we knew who should receive neurostimulation before another biologic?

Combine wearable autonomic signals, joint ultrasound, inflammation, and function data to estimate each patient’s benefit from implanted vagus stimulation versus the next targeted drug before surgery.

The issue in one sentence

Diagram 01

From candidacy to a validated treatment choice

01RA candidate→

027-day ECG · PPG→

03Joint ultrasound→

04Inflammation · function→

05Individual benefit→

06Shared decision→

0724-week outcome

ANM proposed model. Keep the FDA-authorized stimulation regimen unchanged and link preoperative phenotypes to randomized active-comparator outcomes. Scores remain silent in the first study.

01Clinical problem · Evidence

There is no direct evidence for choosing another drug versus nerve stimulation.

Adults with moderate-to-severe rheumatoid arthritis after inadequate response to a biologic or targeted synthetic DMARD may cycle to another drug mechanism or consider implanted vagus stimulation, authorized in 2025. The pivotal trial compared stimulation with sham—not with the next active drug or for individual treatment benefit.

02Technology stack · Proposal

Turn autonomic state and synovitis into a preoperative treatment map.

Combine seven-day single-lead ECG and wrist-PPG HRV, activity and sleep with hand/wrist power-Doppler ultrasound, hsCRP, swollen and tender joint counts, pain, fatigue, and function. A causal model estimates individual treatment difference between vagus stimulation and the next DMARD—not prognosis alone.

03Research question · Hypothesis

Which phenotype gains more from neurostimulation than from the next targeted drug?

Test whether a locked multimodal score reproduces randomized treatment-effect heterogeneity for week-24 ACR50 in external validation and adds decision value beyond clinical variables alone. The score must predict relative benefit between two options—not merely device response.

04First study · Proposal

Twenty-four centers, 480 people: vagus stimulation versus the next active DMARD.

Randomize indicated adults 1:1 on stable background therapy. One arm receives authorized implantation and one-minute daily stimulation; the other receives the next rheumatologist-selected b/tsDMARD from a prespecified formulary. Develop and lock the model in 320 participants, then validate in a separate 160. The single primary endpoint is the treatment-by-score interaction for week-24 ACR50, with success requiring p<0.01 and an absolute treatment-effect separation of at least 15 percentage points between score extremes in validation. Report surgical complications, infections, steroid rescue, CDAI, HAQ-DI, and total cost.

05Defensible data · Hypothesis

The moat is not HRV—it is paired counterfactual treatment evidence.

Link raw ECG/PPG quality and HRV, ultrasound images, inflammatory and functional trajectories, missed stimulation or charging, drug class and adherence, surgery, infection, rescue treatment, and cost to randomized assignment. Publish center-held-out validation and missingness to distinguish the dataset from ordinary responder prediction.

06Venture thesis · Assumption

Sell the right-next-treatment episode—not the implant.

Initial customers are integrated difficult-to-treat RA centers and risk-bearing payers. Bundle preoperative phenotyping, an independent choice report, an implant-surgeon network, and 24-week outcomes audit per decision episode. Contract on low disease activity, function, and total cost—not implant volume; introduce shared savings only after the active-comparator trial.

Evidence map

The treatment-selection gap that remains after approval

242 people

Randomized implanted participants in RESET-RA

35.2% vs 24.2%

Week-12 ACR20: active versus sham stimulation

1 min/day

FDA-authorized stimulation schedule

150 people

Target size of the ongoing FDA post-approval registry

Evidence

Established treatment effect—and the selection evidence still missing

Established: FDA approved the SetPoint System on July 30, 2025, for adults with moderate-to-severe RA after inadequate response, loss of response, or intolerance to at least one b/tsDMARD. Implanted on the left cervical vagus nerve, it stimulates for one minute once daily. RESET-RA randomized 242 implanted participants to active stimulation (122) or sham (120); week-12 ACR20 was 35.2% versus 24.2% (difference 11.8 percentage points, 95% CI 0.6–23.1). Related serious adverse events occurred in 1.6%, all associated with implantation or explantation procedures. FDA’s required approximately 150-person, three-year post-approval registry is uncontrolled, descriptive, and has no primary effectiveness endpoint. Relative benefit versus the next DMARD, preoperative responder selection, and cost savings therefore remain hypotheses.

FDA PMA approval record FDA safety and effectiveness summary RESET-RA trial registry FDA post-approval study status Device safety information and manuals

Failure conditions and safety, ethics, and regulatory risks

Treatment-selection failure

Do not launch individualized recommendations unless the prespecified interaction and 15-point treatment-effect separation both replicate in validation. Do not rescue failure with average effects or retrospective subgroups.

Active-comparator heterogeneity

Pooling multiple drug mechanisms may blur relative effects. Prespecify stratification by mechanism, publish sample size, effects, and missingness for each, and withhold recommendations in underpowered strata.

Surgical and stimulation harm

Independently adjudicate vocal-cord paresis and voice change, dysphagia and aspiration, bleeding, infection, nerve injury, arrhythmia, and device failure. A low score cannot alone exclude an indicated patient, and a high score cannot automatically authorize surgery.

Fairness and access

Skin tone, arrhythmia, medication, and fitness affect PPG/HRV quality, while ultrasound skill and surgical access vary by center. Report unreadable rates, calibration, and treatment effects by sex, race, age, and center; fall back to the clinical model when signals are poor.

Recommendation conflict

Revenue tied to implant volume creates an incentive to over-recommend surgery. Separate model governance and adjudication from device sales, and show patients absolute benefit, uncertainty, and drug alternatives together.

Regulatory boundary

Authorization covers a fixed one-minute daily regimen and defined indication. Treatment-selection software may require separate device review; automatic AI changes to amplitude or timing would exceed the original PMA and require supplemental authorization plus human-factors and cybersecurity validation.

Bibliography

References

Established facts are separated from research and venture hypotheses that still require validation.

01
U.S. Food and Drug Administration. SetPoint System: Premarket Approval P240039.
Primary regulatory record authorizing use after inadequate response, loss of response, or intolerance to at least one b/tsDMARD.P240039 · Approved Jul 30, 2025
02
U.S. Food and Drug Administration. Summary of Safety and Effectiveness Data: SetPoint System.
Contains the 242-person RESET-RA results, week-12 ACR20 of 35.2% versus 24.2%, 1.6% procedure-related serious events, and device details.PMA P240039 · 52 pages
03
ClinicalTrials.gov. A Clinical Trial to Assess the Safety and Efficacy of a Neuroimmune Modulation Device in Adults With Rheumatoid Arthritis (RESET-RA).
Registry for the multicenter randomized double-blind sham-controlled pivotal trial; it did not compare against the next active DMARD.NCT04539964
04
U.S. Food and Drug Administration. SetPoint System Safety and Performance Post-Approval Study.
An ongoing approximately 150-person, three-year registry describing safety and performance without a control group or primary effectiveness endpoint.P240039 / PAS001 · Protocol accepted Jan 16, 2026
05
SetPoint Medical. SetPoint System Manuals and Important Safety Information.
Manufacturer instructions detailing contraindications, MRI and energy-device precautions, and risks including infection, hoarseness, and dysphagia.Accessed Oct 13, 2026

The next issue publishes every day at 9 AM KST.All issues