Novel Biotech · interim evidence audit
Bidirectional peripheral neural interface
CONCEPT STUDY · RESEARCH ONLY · NO HUMAN TESTING · NOT AVAILABLE · NO DEVELOPMENT PROGRAM EXISTS
Historical source verdict: REVISION. Science acceptance: FAIL. Comparator narrowing may proceed; product concept visualization may not. This reading edition preserves the 12 August 2026 record and does not re-adjudicate it.
Preserved source identity
Source PDF: output/pdf/novel-biotech-bpni-evidence-audit-2026-08-12.pdfPrivate repository source
SHA-256: fd4bc686f9c2c29a285ed7ee1bff057bea542803e10c5c4b853dfdf2993eb8d2
A hash establishes byte identity, not scientific truth or present authority. All eight source pages were rendered and inspected. The substantive wording, counts, statuses, tables, and source limitations below are transcribed without modernizing the 2026 snapshot.
Source page 1 of 8 · opening decision
A visual decision brief
The source described what the human evidence supported, what remained open, and why concept visualization was still blocked.
| Observed item | Value | Source wording |
|---|---|---|
| Graph nodes | 55 | 1 search seed + 54 evidence |
| Direct nerve cohort | N=2 | 4–5 weeks in USEA report |
| Closed-loop task | N=1 | Neural + EMG forward decode |
| Claim-linked edges | 0 | 54 edge notes are null |
Evidence-audit visuals only: the source said this document was not a device concept or blueprint. Prepared for integration custody, 12 Aug 2026.
Source page 2 of 8 · route test
The route matters more than the label
Only one inspected human study used direct peripheral-nerve recording and nerve stimulation in the same participants. Muscle electromyography (EMG) systems were adjacent evidence, not substitutes.
| Route | Path | Evidence identity | Status and limit |
|---|---|---|---|
| Locked direct-ENG route | Central nervous system → peripheral nerve → direct recording → computer → controlled stimulation → return toward the central nervous system | USEA / PMID 29178940. Same-participant components: N=2; 4–5 weeks; closed-loop task: S3 only. | PASS, narrowly. Short, percutaneous, unstable; task decode combined neural and EMG signals. |
| Adjacent neuromuscular route | Muscle EMG forward → computer/controller → nerve stimulation return | iSens / e-OPRA / force-slip | UNKNOWN. Useful comparator evidence; it did not fill the direct natural ENG gap. |
Source page 3 of 8 · human evidence
Longer follow-up does not mean the same route
The longest human observations concerned muscle recordings or stimulation contacts. Direct natural peripheral-nerve recording remained short-duration evidence.
| Study | Duration | Route | Cohort / outcome | Material limit |
|---|---|---|---|---|
| USEA | 4–5 weeks | Direct ENG + stimulation | N=2 | Closed-loop S3 only |
| iSens | 15 weeks / 27 months | Muscle EMG + stimulation | N=2 | S02 infection explant |
| e-OPRA | 3–7 years | Muscle EMG + stimulation | 4 implanted / 3 outcomes | Correction is material |
| Spiral cuff | Mean 6.7 years | Stimulation contacts only | N=14 | Not natural ENG |
What the duration picture actually said
- 4–5 weeks was the direct natural ENG evidence window inspected.
- 6.7 years supported stimulation-contact function, not chronic natural ENG stability.
- OPEN: chronic fully implanted direct ENG plus nerve stimulation remained unresolved.
Source page 4 of 8 · claim gates
Four statuses. No optimistic blending.
PASS applied only to bounded author-reported endpoints. OPEN was not evidence of absence. UNKNOWN marked missing classification or comparison. FAIL marked prohibited generalization.
| Status | Claim | Source limit |
|---|---|---|
| PASS | Human direct nerve recording | N=2 / 4–5 weeks |
| PASS | Human nerve stimulation | Small-study endpoints |
| PASS | Same-participant components | USEA N=2 |
| PASS | Closed-loop task | S3 / neural + EMG |
| UNKNOWN | Pure direct-ENG-only loop | Not established |
| UNKNOWN | Simultaneous duplex | Timing not established |
| UNKNOWN | Lower-risk advantage | No matched comparator |
| OPEN | Chronic direct ENG + stimulation | Not located; not absent |
| OPEN | Chronic selectivity | Unresolved |
| OPEN | Tissue + signal stability | Unresolved |
| FAIL | Universal platform claim | Prohibited |
| FAIL | Broad novelty / readiness | Prohibited |
Historical integration rule: comparator narrowing could proceed. Concept selection, visualization, product claims, and development remained blocked.
Source page 5 of 8 · source landscape
Broad coverage. Shallow semantics.
The register contained 54 evidence rows across diverse classes. The source treated this as useful landscape coverage, while noting that the graph connected every source directly to one search seed.
| Source class | Rows |
|---|---|
| Human primary | 14 |
| Animal primary | 10 |
| Reviews | 5 |
| Patent landscape | 5 |
| Trial registry | 4 |
| Official program | 3 |
| Official regulatory | 2 |
| Animal benchtop | 2 |
| Other one-row classes | 9 |
Graph topology: 1 search seed; 54 evidence nodes; 54 one-hop edges; structural integrity marked PASS.
Read the chart correctly: more source classes improved landscape breadth. They did not automatically improve claim precision, study extraction, or traceability.
Source page 6 of 8 · provenance audit
The inventory is intact. The evidence layer is empty.
Hash parity and referential integrity proved the files were internally consistent. They did not prove that claims were linked to source-level evidence.
| Check | Observed result | Coverage |
|---|---|---|
| Manifest hash parity | 6/6 listed hashes match | 100% |
| URL set parity | 54/54 register URLs match graph | 100% |
| Referential integrity | 0 orphan edges or frontier rows | 100% |
| Node metadata coverage | 0/55 metadata objects populated | 0% |
| Edge evidence-note coverage | 0/54 evidence notes populated | 0% |
| Claim-to-source linkage | Inventory edges only | 0% |
Core gap: 54 sources led to 0 source-location evidence notes. “This is a catalog, not an audit-grade claim graph.”
Source page 7 of 8 · correction route
Three bounded packages. One owner at a time.
The evidence producer was to correct the packet; Science was to recheck it independently; only then could comparator narrowing inform a future visual concept decision.
- Product Design — repair the evidence packet. Study cards, exact route classes, claim links, and source-location notes.
- Science — reproduce the corrected gate. Check disclosures, contradictions, timing, and safe claim wording.
- Integrator — reconcile immutable evidence. Verify hashes, JSON/database parity, correction closure, and final status.
Allowed in parallel
Comparator narrowing: one named clinical problem plus one explicit lower-risk external comparator. Research activity only; no product, device, availability, safety, efficacy, or novelty conclusion.
BLOCKED: concept visualization / blueprint / selection / development program.
Source page 8 of 8 · decision
Research direction survives. Canonical evidence does not.
The source concluded that the human literature supported a narrow, short-duration two-way peripheral-nerve premise. It did not support chronic universality, direct-ENG duplex operation, broad novelty, safety, efficacy, or development readiness.
| Comparator narrowing | ALLOWED |
|---|---|
| Canonical science | FAIL |
| Concept visualization | BLOCKED |
| External release | UNAUTHORIZED |
Decisive records named in the source
USEA: pubmed.ncbi.nlm.nih.gov/29178940/iSens: pubmed.ncbi.nlm.nih.gov/40269935/e-OPRA correction: nejm.org/doi/full/10.1056/NEJMx220013Spiral cuff: pubmed.ncbi.nlm.nih.gov/28693584/NIH HORNET: grants.nih.gov/grants/guide/rfa-files/RFA-RM-21-024.htmlDARPA HAPTIX: darpa.mil/research/programs/hand-proprioception-and-touch-interfaces
Independent Science review was recorded as completed on 12 Aug 2026. The source said no files were changed by Science and labeled itself an internal evidence report, not a public copy.