DWTX / halneuron-cinp — Halneuron (purified tetrodotoxin) for chemotherapy-induced neuropathic pain

Program analysis · bpiq_drug_id 18554 · prepared 2026-08 · USD · framework v5.8.0 · Coverage: CLEARED

Company context, financials, ownership, options and the full pipeline: ../company.md.

How to read this. Written for a reader who has not studied pharmacology or finance. Terms specific to this program are in the Glossary below; general terms are defined once in framework/04-glossary.md. Tags: [VERIFIED — source] / [UNVERIFIED] / [WEB ESTIMATE — source, date].

Glossary

Executive summary

• What it is (one sentence): Halneuron is purified pufferfish toxin, given as ten small injections under the skin over four days, which blocks the sodium channels that damaged nerves use to fire spurious pain signals after chemotherapy.
• The event and when (as disclosed): Topline results from the Phase 2b HAL-CINP trial (NCT06848348, 240 patients planned, 217 enrolled). The company said on 2026-08-13 that the readout is expected “in the fall of 2026” and, in the same document, that its cash reaches “the Phase 2b readout in the fourth quarter of 2026.” BPIQ still carries Q3 2026. No source names a day or even a month; the window used throughout this document is 2026-10-01 to 2026-12-31.
• The main reason it could work: There is no approved treatment for this condition anywhere, the mechanism is real and well understood, and two independent unblinded interim reviews — at 97 and at 200 patients — reported that the drug arm separated from placebo.
• The main risk: Four controlled trials of this molecule in pain have now been run, and not one of them has met its pre-specified primary endpoint. The immediate predecessor in this exact indication (n=125) missed outright: average pain fell 1.339 points on placebo and 1.529 points on the dose that was carried forward — placebo beat two of the four drug arms. The signal that survived came from cumulative responder analyses, only one of four definitions of which cleared p<0.05, unadjusted for multiple comparisons. The current trial’s primary endpoint is a responder analysis.
• What it means for the stock: Halneuron is the whole company (../company.md C.2 records it as dominant). The modal outcome is a decline, the expected value is positive, and those two statements are not in conflict — the payoff is extremely asymmetric on a $55M enterprise value with $9.6M of cash. This is a lottery ticket with honest odds printed on it, not a directional view.

0. Program-tier coverage — CLEARED

Company-tier coverage is in ../company.md C.0.

Regulatory tier: not called at all, correctly. The catalyst is a trial readout, not a regulatory decision, and 02-connectors.md makes that tier conditional.

A. Scientific & clinical assessment

A.1 Mechanism of action and scientific rationale

Nerves carry signals as electrical pulses. Each pulse starts when tiny gates in the nerve’s outer wall — voltage-gated sodium channels — snap open and let sodium ions rush in. Platinum-based and taxane-based chemotherapy drugs damage the longest, thinnest sensory nerves first, which is why the symptoms appear in the hands and feet. Damaged nerve fibres do not simply go quiet. They over-produce and hyper-activate these sodium channel gates, and begin firing pain signals with nothing to report — a phenomenon called ectopic firing. The patient feels burning, tingling and shooting pain in a part of the body that is not being injured. That is chemotherapy-induced neuropathic pain.

Halneuron is purified tetrodotoxin, the pufferfish toxin, given at doses roughly a thousand times below the poisonous range [VERIFIED — ChEMBL CHEMBL507974, molecular weight 319.27, natural product; NCT06848348 intervention record]. It works by plugging the outer mouth of the sodium channel so the gate cannot pass sodium. The dose studied is small — 30 micrograms twice daily as a subcutaneous injection for four consecutive days, ten injections in total across a treatment cycle [VERIFIED — NCT01655823 arm descriptions; Goldlust et al., Toxins 2021, [DOI 10.3390/toxins13040235](https://doi.org/10.3390/toxins13040235)].

One correction to the company’s own description, stated plainly. Dogwood describes Halneuron as “a non-opioid, NaV 1.7 analgesic which is a highly specific voltage-gated sodium channel modulator” [VERIFIED — Form 8-K exhibit 99.1, 2026-08-13]. Tetrodotoxin is not selective for Nav1.7. It blocks six of the nine human sodium channel subtypes — Nav1.1, 1.2, 1.3, 1.4, 1.6 and 1.7 — at low nanomolar concentrations, and is largely inactive against the other three: Nav1.5 in the heart, and Nav1.8 and Nav1.9, which are themselves important pain channels [UNVERIFIED — textbook pharmacology; the ChEMBL selectivity panel across the nine subtypes was not returned by this session's compound_search call, so this is stated as established background rather than as a verified measurement]. The honest framing is that tetrodotoxin is selective for a family of channels, not for one member of it. Nav1.7 is simply the member with the strongest human genetic link to pain: people born without working Nav1.7 feel no pain at all, and Open Targets confirms SCN9A (the gene for Nav1.7) as a well-populated target record [VERIFIED — Open Targets search_entities, SCN9A → ENSG00000169432, 2026-08-13].

That distinction has two practical consequences. In the drug’s favour: sparing Nav1.5 is why a sodium-channel blocker given systemically does not stop the heart, and a dedicated thorough-QT study in 25 healthy adults found no QT prolongation at 15, 30 and 45 micrograms — a formally negative QT study [VERIFIED — Kavoosi et al., Toxins 2020, [DOI 10.3390/toxins12080511](https://doi.org/10.3390/toxins12080511)]. Against it: blocking Nav1.6 and Nav1.7 in undamaged nerves is what produces the tingling and numbness around the mouth that roughly a third of treated patients report, and that side effect is the same mechanism as the intended one, not a separate off-target liability.

The most interesting thing about this drug is also the least explained. Tetrodotoxin reaches peak blood concentration about 1.5 hours after injection and is cleared quickly [VERIFIED — Kavoosi et al., Toxins 2020, DOI 10.3390/toxins12080511]. Yet across every trial the analgesic effect is reported to last for weeks after a four-day course — a mean duration of analgesic response of 56.7 days on drug versus 9.9 days on placebo in the largest cancer-pain trial [VERIFIED — Hagen et al., Pain Res Manag 2017, [DOI 10.1155/2017/7212713](https://doi.org/10.1155/2017/7212713)]. A drug that is gone in a day and works for two months is either genuinely resetting something in the pain pathway, or is being measured against a natural history that improves on its own. Nobody has published a mechanism for the durability, and the 2007 open-label study said as much: “It may have a novel mechanism of analgesic effect” [VERIFIED — Hagen et al., J Pain Symptom Manage 2007, [DOI 10.1016/j.jpainsymman.2006.11.008](https://doi.org/10.1016/j.jpainsymman.2006.11.008)].

The exact scientific step this readout must prove. Not that sodium-channel blockade relieves neuropathic pain — lidocaine and carbamazepine settled that decades ago. Not that tetrodotoxin reaches the nerve — the pharmacokinetics are published. The step is narrower and harder: that the proportion of patients whose daily pain score falls by a clinically meaningful amount over four weeks is significantly greater on Halneuron than on placebo, in a trial where that comparison is the pre-specified primary endpoint rather than one of several analyses run afterwards. That specific thing has never been demonstrated for this molecule.

The honest scientific risk is not that the drug does nothing. It is that the effect is real but small, and smaller than the placebo response in a subjective daily pain diary. The predecessor trial’s own authors identified exactly this: mean pain scores were “not statistically different between cohorts, due to small trial size and influence of a few robust placebo responders” [VERIFIED — Goldlust et al., Toxins 2021, DOI 10.3390/toxins13040235]. A four-week self-reported pain diary in patients who have finished chemotherapy and are slowly recovering is one of the noisiest measurements in clinical medicine.

A.2 Clinical development plan, timeline, feasibility, resourcing

Trials

The predecessor’s actual numbers, because they are the single most important evidence in this document. NCT01655823 posted its results to ClinicalTrials.gov. On the pre-specified primary endpoint — change from baseline in the weekly average NPRS score at days 22 to 28, on an intention-to-treat basis — the five arms produced [VERIFIED — ClinicalTrials.gov results section, NCT01655823, read 2026-08-13]:

Read the last column. Placebo produced a larger average pain reduction than two of the four tetrodotoxin arms. The best arm beat placebo by 0.343 points on an 11-point scale, against a standard deviation above 1.8 and a commonly used threshold for a meaningful individual improvement of roughly 2 points. No statistical analyses were posted with these results, and the publication states the mean-change comparison was not significant.

What the publication reports instead is a cumulative responder analysis, in which the 30 µg twice-daily arm differed from placebo using “the maximum response at any timepoint (p=0.072), 5-day (p=0.059), 10-day (p=0.027), and 20-day (p=0.071) rolling averages” [VERIFIED — Goldlust et al., Toxins 2021, DOI 10.3390/toxins13040235]. One of those four definitions clears p<0.05, and none is adjusted for having tried four. That is the entire quantitative basis on which this program advanced.

Feasibility matrix

Resourcing sufficiency. The trial itself is funded and effectively finished: 217 patients are enrolled and the endpoint is measured four weeks after each patient’s own dosing, so the remaining cost is database lock, unblinding and statistics. Beyond that the company is not resourced. ../company.md C.3 records $9,610,033 of cash at 2026-06-30 against a recomputed burn of $1.624M a month — a runway to roughly 2026-12-24, which the company itself describes as reaching “the Phase 2b readout in the fourth quarter of 2026” and no further. A Phase 3 in this indication, which the company projects for the first half of 2027, is not funded and is not close to funded. C.3 also sets out why the ordinary financing route is closed until January 2027: the baby-shelf capacity was consumed in full by the January 2026 offering.

A.3 Target label, target product profile, strategic questions, designations

A.3a Target label. Treatment of moderate-to-severe chronic neuropathic pain in adults that is attributable to prior platinum-based and/or taxane-based chemotherapy, in patients with no actively progressing cancer.

A.3b Target product profile

A.3c Strategic Go/No-Go questions. The next decision for this asset is whether to start Phase 3, so the pre-Phase-III set applies.

Pre-Phase-III (Go-to-Phase-III / registration):

A.3d Regulatory designations.

• FDA Fast Track designation, granted for Halneuron in CINP [VERIFIED — Form 8-K exhibit 99.1, 2026-08-13]. Fast Track means the FDA has agreed the drug addresses an unmet need in a serious condition, which buys more frequent meetings and the ability to submit a marketing application in pieces as they are ready. It does not lower the evidence bar and it is not an agreement about the endpoint.
• FDA Fast Track designation is also held by IMC-1 for fibromyalgia [VERIFIED — BPIQ note field, row 18553]. That belongs to a different program and is listed here only to prevent it being read as a second designation for Halneuron.
• No Breakthrough Therapy, no Orphan Drug, no Priority Review, no RMAT designation has been disclosed for this program.

A.4 Target product profile valuation matrix

Calibration examples, not asset claims: in oncology each incremental month of overall survival is worth roughly $150–300M in peak-sales potential; oral formulations command roughly 15–25% price premiums over injectables. The second of those runs against this asset, which is an injectable competing with a generic pill.

A.5 Evidence quality and endpoints

Endpoints

An unblinding problem worth stating in its own paragraph. Tetrodotoxin produces tingling and numbness around the mouth in roughly a third of treated patients. In the predecessor trial the imbalance was large: oral paraesthesia in 11 of 26 patients on 30 µg twice daily against 3 of 25 on placebo, and oral hypoaesthesia in 10 of 26 against 3 of 25 [VERIFIED — ClinicalTrials.gov adverse-events section, NCT01655823]. A patient whose mouth goes numb within an hour of the first injection has a strong clue about which arm they are in, and the primary endpoint is that same patient’s self-reported pain diary. Quadruple masking is procedurally correct and cannot fix this: it is the drug that unblinds, not the protocol. The effect runs in the drug’s favour on the measured endpoint and against it in any regulatory review, and it is a reason to want the objective-ish secondary measures — sleep, fatigue, global impression — to move in the same direction as the diary.

A.5b Key opinion leaders.

Panel as of. 2026-08-13 — the date the CT.gov investigator search and the PubMed authorship and conflict searches below were run.

Investigators

No investigator could be named for this program’s pivotal trial, and that is itself the finding. All 25 sites on NCT06848348 are registered as “Central Recruiting Site” with no facility name, no principal investigator and no site contact; the record lists no overall officials at all, and its only central contact is a company email address (Mehran@dwtx.com, which corresponds to Mehran Kavoosi, an author on the predecessor publication while at WEX Pharmaceuticals). CT.gov search_investigators on the condition returned zero investigators across the trial. This is unusual — the predecessor trial NCT01655823 named all 23 of its sites, including academic centres such as UT Southwestern and the John Theurer Cancer Center — and it means a reader cannot check who is running this trial.

Independent voices

No independent voice on this program’s endpoint could be named this session, and the reason is structural rather than a gap in the search. The complete clinical literature on this molecule in pain is seven papers, and every one of them was generated by the sponsor’s own program: five carry a WEX Pharmaceuticals author, and the two Hagen-led cancer-pain papers are reports of WEX-sponsored trials with Walter Korz of WEX as a co-author. There is no commentary, editorial or independent replication in the indexed literature to draw a voice from. Samuel A. Goldlust (Hackensack University Medical Center) and Kenneth Deck (Alliance Research Centers) are the two non-sponsor authors on the pivotal predecessor publication, but both were site investigators on the trial they are reporting, so neither is independent and neither is recorded here — per the template’s own instruction, a person with a disclosed relationship belongs above, not in this table.

Judgement

Dissent

Empty, because Endpoint supported is UNKNOWN rather than SUPPORTIVE_CONTESTED. Recording a disagreement would require someone to be disagreeing, and no named voice on either side exists.

B. Commercial assessment

B.0 Current treatment algorithm

A patient finishes chemotherapy that included a platinum drug (cisplatin, carboplatin, oxaliplatin) or a taxane (paclitaxel, docetaxel). Weeks or months later the burning and tingling in the hands and feet has not gone away. What happens next, today:

1. During chemotherapy — dose reduction or discontinuation. This is not a treatment, it is the absence of one, and it is the most consequential step in the whole pathway. The predecessor trial’s own registry summary puts it plainly: “To improve the peripheral neuropathy, the chemotherapy dosing is often either decreased or discontinued potentially affecting tumor responsiveness, prognosis, and survival” [VERIFIED — NCT01655823 brief summary].
2. Prevention — nothing. ASCO’s guideline recommends no agent for the prevention of CIPN [VERIFIED — ASCO 2020, DOI 10.1200/JCO.20.01399].
3. First and only recommended treatment — duloxetine, an oral antidepressant of the serotonin-norepinephrine reuptake inhibitor class, taken daily. ASCO gives it a moderate-strength recommendation on intermediate-strength evidence and notes that the amount of benefit is limited [VERIFIED — ASCO 2020].
4. Everything else — tried, not recommended. Tricyclic antidepressants, gabapentin, and a compounded topical gel of baclofen, amitriptyline and ketamine may be tried, but routine use is explicitly not recommended for inadequate evidence. Exercise, acupuncture and scrambler therapy are classified as unproven but reasonable to consider [VERIFIED — ASCO 2020].
5. Opioids. Not guideline-recommended for this indication and carrying their own well-known harms, but used in practice when nothing else works.

Where Halneuron would fit. At step 3, as an alternative or an addition to duloxetine, for the patient whose pain is moderate to severe and persistent. It is a four-day injected course rather than an indefinite daily pill, which is a genuine differentiator on treatment burden and a genuine disadvantage on convenience. It does not compete with anything at steps 1 or 2, because nothing is there.

B.1 Competitive landscape and positioning

Where this asset wins, and the single fact the thesis rests on. It wins on the emptiness of the field. In an indication where the best available medicine is a generic antidepressant that the guideline itself describes as limited, a drug that reproducibly relieves pain would face almost no competition and would set its own price. That is a real and unusual advantage.

The single fact the whole thesis rests on is that the responder analysis is measuring something real rather than something that appeared because four analysis definitions were tried on a trial that had already missed. Everything else — the unmet need, the mechanism, the durability, the tolerability, the IP — is downstream of that one question, and this readout answers it.

B.2 Addressable market

The launch market is the United States, where the trial was run entirely (25 US sites) and where the pricing environment could support a specialty analgesic. The EU5 and Japan would follow an approval, and no EU trial has ever been run for this molecule.

B.3 Value and feasibility

B.3a Expected peak sales. Built bottom-up as eligible patients × annual net price × peak penetration, US only, and conditional on approval, which on the evidence in A.2 is far from assured. Every figure excludes ex-US sales, excludes SP16, and excludes any IMC-1/IMC-2 partnership proceeds — the last of which would in any case be shared with CVR holders rather than accruing wholly to shareholders [VERIFIED — Form 8-K exhibit 99.1, 2026-08-13].

Sanity check against third-party figures, and one rejection. Independent forecasts put the entire CIPN treatment market — every product, including generic duloxetine and supportive care — at roughly $1.4–2.1 billion by 2030 [WEB ESTIMATE — Strategic Market Research and Astute Analytica summaries, 2026]. The high case above therefore implies a single branded drug capturing most of a market it would have to create, which is why it is a ceiling rather than a forecast. The base case of ~$400M sits at roughly a fifth of that whole-market figure, which is demanding but not absurd for the only approved product in an indication.

The least defensible input is the patient count, and it is named here rather than buried. The incidence figure (>40% of patients on neurotoxic chemotherapy develop CIPN) is verified from the trial registry, but converting that into a prevalent pool of patients with chronic moderate-to- severe painful neuropathy required assumptions that no primary source in this session supports. The price is the second-weakest: there is no approved comparator in this indication to anchor to, and the $15,000–28,000 range is inferred from office-administered specialty analgesics generally.

B.3b Feasibility

B.4 Product-development risk

Framing questions.

1. Does the development plan support the target product profile claims? Partly. The trial is correctly designed to test the pain claim and is adequately sized for the effect the company assumes. It does not support two claims the profile leans on: durability beyond four weeks (the controlled period ends at week 4; the 12-week extension is open-label and therefore cannot demonstrate it), and any payer-facing claim about opioid use or chemotherapy dose maintenance, which is not measured at all.
2. Will the identified risks affect the target product profile? The dominant risk — that the effect is real but too small to separate from placebo on a self-reported diary — does not modify the profile, it eliminates it. There is no reduced version of this product that survives a failed pain endpoint.
3. If a risk cannot be mitigated, is the asset still differentiated from competitors? Only if the endpoint is met. Differentiation here is entirely a function of efficacy, because the field is empty: an injectable that does not reliably work is worse than a generic pill that works a little.

Readout

Sources

The modelled estimate. The registry’s primary completion date is 2026-07, and 01-rules.md rule 34’s default lag from primary completion to a topline announcement is two to four months, giving 2026-09 to 2026-11. That arithmetic is weakened here by the registry date being stale, so a second, independent construction is offered and is the one the window actually leans on: enrollment arithmetic. 200 patients were enrolled by 2026-07-20 and 217 by 2026-08-13, a rate of roughly 17 per month; the company wants “over 220”, which arrives around 2026-08-20. The primary endpoint is measured four weeks after each patient’s own dosing, so last patient out lands around 2026-09-17. Database lock, unblinding and statistical analysis on a 240-patient trial with a single four-week endpoint conventionally takes four to eight weeks, giving 2026-10-15 to 2026-11-12 [UNVERIFIED — modelled on verified enrollment figures]. The two constructions overlap in October and November, which is where the window’s likeliest edge sits.

Window

Basis. No source names a day or a month — “fall”, “Q4” and “Q3” are all periods, so the precision is PERIOD and not MONTH, however tempting the registry’s month-precision 2026-07 looks. The earliest edge is 2026-10-01 rather than a September date because both independent constructions point past September: the company’s own runway sentence says the fourth quarter, and the enrollment arithmetic cannot produce a readout before mid-October unless enrollment finished weeks earlier than the company said it had. The latest edge is 2026-12-31, the end of the fourth quarter the company named and, not coincidentally, the approximate end of its cash. Confidence is MEDIUM rather than HIGH because the same document offers two different answers, and rather than LOW because three independent constructions — company guidance, registry arithmetic and enrollment arithmetic — all land inside October–December.

Disagreement. UNRESOLVED. Three sources point three ways and none is averaged or discarded. BPIQ says Q3 2026, which ends 2026-09-30. The company says “fall of 2026” in one sentence of its 2026-08-13 release and “the fourth quarter of 2026” in another. ClinicalTrials.gov says primary completion 2026-07, a date that has already passed with no readout and no registry update since 2026-05-29. The disagreement is not cosmetic: the difference between Q3 and Q4 is the difference between a readout that arrives before the cash gets tight and one that arrives as it does.

Date slippage. Two slips across five dated statements, oldest first, parsed from the BPIQ note field and the company’s own releases.

Attribution

Status. CLEAN.

Conflicts

Empty, exactly as CLEAN requires.

Computed, not authored: lib/clustering.mjs’s attributionFor('DWTX', companyRecord, programs, 6, 18554) over this ticker’s full four-row pipeline table and this program’s own readout.window, run 2026-08-13, returned {"status": "CLEAN", "conflicts": []}.

The reason is simple enough to state without the computation, and it is worth stating because it is rare in this corpus: Halneuron is the only row on this ticker’s pipeline with has_catalyst: true. IMC-1 and IMC-2 carry TBA and no catalyst; the SP16 row carries a 2026-08-15 catalyst_date but has_catalyst: false, correctly, because a Phase 1b enrollment start is not a market-moving event. There is no second event for a price move around this readout to be confused with. A move in the window belongs to this program.

Note. (Not required — Status is CLEAN.)

Market and timing for this event

• Plain takeaway. A single-asset company with about five months of cash is about to report the only trial that matters, into a market that has already watched this ticker fall 80% and has watched it fall 22.6% on the last piece of good news it delivered.
• Months to this catalyst. 1.6 months to readout.window.earliest (2026-08-13 to 2026-10-01), and 3.1 months to the likeliest edge. readout.precision is PERIOD, so these figures are the edges of a bracket and not a countdown — the event could land anywhere in a three-month span, and the difference between its ends is material because the company’s cash runs out at the far end.
• Expected move around this event. A bracket, not a point estimate. ../company.md C.6 records the options chain as unreadable from two independent directions — BPIQ returned nothing across three attempts and the Yahoo cross-check was throttled — so no implied move is available. Built instead from this ticker’s own history in C.7, where the largest catalyst-day moves are −22.56%, +12.83% and −13.71%: a ±20% to ±45% single-day move on a binary efficacy topline is the honest bracket, wider than the historical moves because those were interim and operational news rather than a definitive readout, and on a float of roughly 5.1 million shares trading $95,000 a day.
• Nearest comparable past reaction. ../company.md C.7’s 2025-12-22, −22.56% — the positive interim Phase 2b result in this same trial. It is the closest analogue by subject and the most instructive row in the table, and it is comparable in one direction only: it shows what this market does with good Halneuron news when a raise is visible on the horizon, which is sell it. It is not comparable in magnitude, because an interim sample-size analysis is not a topline efficacy result, and the topline is the larger event.
• Materiality. DOMINANT — the exact wording recorded for this program in ../company.md C.2. This is the company: the only funded clinical readout, the destination of $3.2M of the $3.25M of Q2 research spending, and the event the company’s own cash runway is sized against.
• Date slippage. Two slips across five dated statements (see Readout, above): Q3 2026 → Fall 2026 in May, and the first appearance of “fourth quarter” in the company’s own words today. Both moves are in the same direction — later — and the second one lands the readout on top of the end of the runway.

Spot. $1.9346, the closing price on 2026-08-13, cited from ../company.md C.4.

Scenario prices

Expected value. Applying the 38% probability to the midpoint of each range: 0.38 × $6.39 + 0.62 × $0.78 = $2.91, or +50.5% against the $1.9346 spot. Method: probability-weighted midpoints of the two scenario ranges above, nothing else. This is arithmetic, not advice, and it is not a price target. It is positive while the modal outcome is a decline, because the payoff is asymmetric — see the stock-direction call below, which declines a direction for exactly that reason.

Run-up

Entry and exit

Predicted move and peak

Priority score drivers

Transcribed from lib/runup.mjs’s scoreDriver, run 2026-08-13. The two judgement drivers are read from this document; the other five are computed.

Priority score. Priority score 13 · formula_version 1.0.0

The score is low, and the reason is legible: date_confidence at 20 falls below the formula’s untradeable threshold of 30, so a hard ceiling of 15 applies over the multiplier. A run-up you cannot time to better than “sometime in the fourth quarter” is not a good run-up, whatever the float mechanics say — which is precisely what the ceiling is there to express.

Settlement. Left null at lock time. Settled only on an explicit user request, from a confirmed primary source, against the committed price cache.

Verdict

What I would do. Watch — and size any position as a lottery ticket, not as an investment.

Why. Four controlled trials of this molecule in pain have now been run and not one has met its pre-specified primary endpoint, including the n=125 predecessor in this exact indication where placebo produced a larger average pain reduction than two of the four drug arms. The signal carrying this program forward is a cumulative responder analysis in which one of four rolling-average definitions cleared p<0.05, unadjusted for multiplicity — and the current trial’s primary endpoint is that same analysis type. Against that, the unmet need is genuine and total, the company is priced at a $55M enterprise value, and the float is 5.1 million shares, so a win is worth multiples. The expected value is +50% against spot while the modal outcome is a large decline; both statements are true, and confusing the second for a reason to skip the first, or the first for a reason to ignore the second, are the two ways to get this wrong.

What would change this. To the upside: any disclosed number from either interim review. The company has twice said an independent committee saw separation from placebo and has never released an effect size, a responder rate or a p-value. A quantified interim, or a pre-specified statistical analysis plan showing a single primary responder definition fixed before unblinding, would move the 38% probability materially upward. To the downside: a financing announced before the readout. Given the baby-shelf constraint in ../company.md C.3, a pre-readout raise could only be done at a discount into weakness and would signal that management does not expect the data to fund the company — which is precisely the sequence December 2025 taught this ticker’s holders.

What to watch.

• From now to 2026-10-01 — any 8-K disclosing a securities purchase agreement, private placement or registered direct. This is the single most informative near-term observable, and it points down.
• Around 2026-08-20 — the enrollment-complete announcement. 217 patients on 2026-08-13 against a stated target of “over 220” means this is days away, and the date it lands sets the four-week clock that fixes the readout window. If it slips past August, the window’s earliest edge moves with it.
• Any time — a ClinicalTrials.gov record update on NCT06848348. The record has not been touched since 2026-05-29 and carries a primary completion date that has already passed. An update to an ACTUAL primary completion date would convert readout.precision from PERIOD toward MONTH and, with it, the run-up call’s whole tradeability.
• 2026-10-01 to 2026-12-31 — the topline itself. Read three things before the share price: whether the primary responder endpoint was met at a single pre-specified threshold, what the placebo responder rate was, and whether PGIC and the PROMIS instruments moved in the same direction as the pain diary. If the pain diary separates and the global impression does not, the unblinding problem in A.5 is the likeliest explanation.
• With the topline — whether a financing is announced alongside it. On this ticker, good news without a funded next step has already been sold once.

Locked prediction

• Outcome-direction (will the readout succeed on its primary endpoint?): miss — probability 38%, band 26–49% [UNVERIFIED — modelled]. The band sits below 50 because four controlled trials of this molecule have produced zero pre-specified primary endpoint wins, and the one signal that survived the predecessor came from a four-definition responder analysis with no multiplicity adjustment. It is as high as 38% because the trial is genuinely better powered than its predecessor (~220 versus 125), uses the dose that separated, is enriched to a clean platinum/taxane population, and has had two unblinded interim reviews report separation.
• Stock-direction (which way do the shares move?): no-edge — confidence low — window 2026-10-01 to 2026-12-31, basis: the readout.window, from its earliest to its latest edge, which is the whole period in which the topline could land and be attributed to this program. Attribution is CLEAN, so a move inside that window belongs to this readout and to nothing else on the pipeline. Materiality is DOMINANT (../company.md C.2), and the call is consistent with it in the sense that matters: a dominant program with a 38% success probability produces a large move in one direction or the other, not a small one. no-edge sits beside an expected value of $2.91, +50.5% against spot, and the two do not quietly disagree — the direction is declined because the payoff is asymmetric, not despite it. The modal outcome is a decline of 35% to 85%; the 38% case is a gain of 70% to 390%. A direction word would have to pick one and would misdescribe the distribution either way.
• Scenario prices: positive $3.28–$9.50 · miss $0.28–$1.28
• Expected value: $2.91, +50.5% against spot $1.9346 (arithmetic, not advice)
• Run-up: entry $1.9346 on 2026-08-13, exit rule T-5 trading days before readout.window.earliest — predicted move 0–30%, predicted peak +5–40% around 2026-09 — priority score 13, formula_version 1.0.0
• Settles on the company’s own topline announcement for the Phase 2b HAL-CINP trial (NCT06848348). Positive definition: the company reports that the trial met its primary efficacy endpoint — a statistically significant (p<0.05, two-sided) greater proportion of responders on Halneuron than on placebo, on the change from baseline in the weekly average of daily 24-hour pain intensity scores at week 4, on the pre-specified primary analysis population. A result described as positive on the strength of a secondary endpoint, a subgroup, a post-hoc or alternative responder threshold, an unadjusted analysis of multiple thresholds, or the open-label extension does not count as positive. Source: the company’s own press release and the corresponding Form 8-K, cross-checked against the ClinicalTrials.gov results posting when available.
• Locked: yes · Settled: no

Program data-quality flags

• ClinicalTrials.gov NCT06848348 is stale. Last updated 2026-05-29, carrying an ESTIMATED primary completion date of 2026-07 and an ESTIMATED completion date of 2026-08, both of which have passed or are passing. The registry cannot be used as a live timing source for this readout; it is used here as one of three constructions and weighted accordingly.
• The trial’s registered phase disagrees with the company’s. ClinicalTrials.gov records NCT06848348 as PHASE2; the company, BPIQ and every press release call it Phase 2b. Cosmetic for the readout, but it means a phase-based screen would not find this trial where a reader expects it.
• Enrollment figures disagree between sources on the same day. BPIQ’s row and the 2026-07-20 release say 200; the 2026-08-13 8-K says 217; the registry says 240 ESTIMATED. All three are correct for what they measure (a milestone, a current count, a target), and none was averaged.
• Open Targets returned Rate limit exceeded for client: global on two of three attempts, the documented standing throttle. The third, narrowed to a single query string, succeeded. Recorded because a run that gave up after two attempts would have recorded a BLOCKED where data was actually available.
• ChEMBL compound_search returned the molecule but not a selectivity panel. The record for CHEMBL507974 carries molecular properties and synonyms but no cross-subtype bioactivity, so the claim in A.1 that tetrodotoxin blocks six of nine sodium-channel subtypes is tagged as established background rather than as a verified measurement from this session. The HTTP 500 that blocked this tool on both tickers swept earlier the same day did not recur.
• PubMed’s author field was complete here. Recorded as a negative finding against the CNTB bug from the same day, where all 14 articles returned null authors. The author lists are what made A.5b’s independence assessment possible at all, and a run that assumed the bug would recur would have skipped the check.
• The pivotal trial names no investigator at any of its 25 sites. Every location is registered as “Central Recruiting Site” with no facility name and no contact, there are no overall officials, and the only listed contact is a company email. search_investigators returned zero. The predecessor trial named all 23 of its sites. This is unexplained, and it is why A.5b’s investigator table is empty.
• The company’s description of its own prior data conflicts with the published record, and this is a judgement flag rather than a connector flag. The 2026-08-13 release states that “Halneuron has demonstrated statistically significant and durable pain reduction in the clinical data developed to date.” The published pivotal predecessor states that changes in mean NPRS score “were not statistically different between cohorts”, and the largest cancer-pain trial reported a p-value that was “not [significant] at the prespecified two-sided 5% level”. Both statements can be reconciled only by counting post-hoc and secondary analyses as the demonstration. That reconciliation is available to the reader; it is not available silently, and it is one of the reasons the modelled probability sits at 38% rather than higher.
• The company describes Halneuron as a “NaV 1.7 analgesic”. Tetrodotoxin is selective for the TTX-sensitive sodium-channel family, not for Nav1.7 within it. Recorded in A.1 rather than smoothed over, because the distinction is load-bearing for both the safety story (Nav1.5 sparing) and the side-effect story (Nav1.6/1.7 blockade in healthy nerve).