TNX-103 cut a heart-strain marker by 49% and moved the six-minute walk by 3.5 metres. Tenax has since paused its second phase 3 trial to rewrite who gets in, using a rule drawn from the 241 patients of the first. This brief sets out the results, how the drug works, and what the new trial can prove.
TNX-103 is oral levosimendan, a new form of a heart drug long given by infusion. Tenax Therapeutics is testing it in pulmonary hypertension caused by heart failure with preserved ejection fraction (PH-HFpEF), a condition with no approved drug. Its first phase 3 trial, LEVEL, missed its primary endpoint in results reported on 10 August 2026. The gain in six-minute walk distance was 3.5 metres larger on the drug than on placebo (P = 0.63).
The same trial showed the drug doing what its pharmacology predicts. NT-proBNP, a blood marker of strain on the heart, fell 49% against placebo. An echocardiographic estimate of pressure in the lung circulation fell 3.5 mmHg. So the drug is active. What LEVEL did not show is that the activity lets patients walk further.
Tenax gave its answer on 8 October and 5 November. It paused enrolment in the second trial, LEVEL-2, and is narrowing entry to slower walkers under a rule built from LEVEL’s own data. The FDA has granted a Type C meeting. The company held $113.7 million of cash at 30 September, and the stock market valued it at about $63 million on 19 November (computed).
01What TNX-103 is: oral levosimendan for PH-HFpEF
Levosimendan is an old drug in a new job. Orion of Finland developed it as an infusion for acute heart failure. The infusion is authorised in 60 countries, though not in the United States or Canada, according to Tenax. TNX-103 is the oral formulation, taken two or three times a day. Tenax develops it under licence from Orion and says its rights are global.
The target is PH-HFpEF. In these patients a stiff left ventricle pushes pressure back into the lungs. Pulmonary hypertension from left heart disease is the most common form of pulmonary hypertension, and no drug is approved for it. Drugs built for pulmonary arterial hypertension have shown no benefit there and may do harm. That is the reading of a 2026 review (Khan et al., JACC, via PubMed).
The phase 2 result behind the programme
The case for phase 3 came from HELP, a trial of weekly levosimendan infusions published in 2021 (Burkhoff et al., JACC Heart Failure, via PubMed). HELP missed its own primary endpoint, the wedge pressure during exercise (P = 0.65). A secondary measure moved. Patients on levosimendan walked 29.3 metres further than those on placebo (95% CI 2.5 to 56.1; P = 0.033).
Two features of HELP did not carry into LEVEL. HELP gave every patient an open-label infusion first and randomised only the 37 of 44 whose exercise wedge pressure fell by at least 4 mmHg. LEVEL’s registry record describes screening and then 12 weeks of blinded treatment, with no such lead-in. And HELP used the infusion, while LEVEL used the oral drug. Two HELP authors hold senior roles in the programme. Stuart Rich is Tenax’s chief medical officer, and Sanjiv Shah was LEVEL’s principal investigator.
TNX-103 timeline: from LEVEL’s start to the Type C meeting
02TNX-103 LEVEL trial results: what Tenax’s phase 3 showed
LEVEL randomised 241 patients at 41 sites in the United States and Canada to TNX-103 or placebo for 12 weeks. The primary endpoint was the change in six-minute walk distance. The key secondary endpoint was the KCCQ total symptom score, a patient-reported measure. Both missed.
NT-proBNP and the echo estimate of lung pressure moved in the drug’s favour, and both carry nominal P values.
Most patients were already on modern heart failure drugs when they entered. Add-on trials face a ceiling when background therapy is strong, a point Witfire examined in eplontersen’s CARDIO-TTRansform miss.
The primary endpoint
At week 12 the walk distance had risen by 14.0 m on TNX-103 and by 10.4 m on placebo. The model’s estimate of the difference was 3.5 m, with a standard error of 7.3 m. Tenax’s release gave no confidence interval, so one is worked out here.
95% interval 3.5 ± 1.96 × 7.3 = −10.8 m to +17.8 m
Effect the trial was sized to detect 25 m
Shortfall 25 − 17.8 = 7.2 m
That matters because of what Tenax said eight months earlier. On 17 December 2025 a blinded review found the trial had well over 90% power to detect a 25-metre difference. With a standard error of 7.3 m the power works out at about 93% (computed), so the two statements agree. LEVEL was large enough to see a 25-metre effect, and its interval excludes one in the population it enrolled.
The symptom score
The KCCQ total symptom score rose 6.6 points on TNX-103 and 6.5 on placebo. The difference was 0.1 point with a standard error of 2.2. Both arms improved by about the same amount.
What did move
NT-proBNP fell. The ratio to placebo was 0.51 (95% CI 0.43 to 0.60), which is the 49% reduction Tenax reported. In absolute terms the mean change was −39.1 pmol/L on TNX-103 and +13.7 pmol/L on placebo. Right ventricular systolic pressure on echocardiography fell 3.5 mmHg against placebo (95% CI −6.01 to −1.10).
That result needs context. The registry and the trial’s design paper (Yaku et al., via PubMed) both list NT-proBNP among LEVEL’s secondary outcomes. The company’s release describes the analysis as prespecified and exploratory, with nominal P values. And a fall in natriuretic peptides is a long-documented effect of levosimendan itself. A 2007 pharmacology review recorded substantial falls in BNP and NT-proBNP after a single infusion (Antila et al., Clinical Pharmacokinetics, via PubMed). The marker shows the drug reached its target. It does not show that patients felt or functioned better.
TNX-103 side effects and tolerability
Five times as many patients stopped TNX-103 as stopped placebo. Tenax attributes the extra events to headache, palpitations and hypotension, which are the expected effects of a vasodilator. A similar pattern appeared in Orion’s 496-patient trial of oral levosimendan in ALS. There, increased heart rate was recorded in 33% of treated patients and 7% of those on placebo (Cudkowicz et al., Lancet Neurology 2021, via PubMed).
03The TNX-103 subgroup result cuts both ways
Tenax’s case for carrying on rests on one prespecified split. Patients were divided at the median baseline walk of 333 m. Below it, TNX-103 beat placebo by 26.3 m. The company’s October slides also give the other half.
Placebo arms 23.4 − (−5.2) = 28.6 m apart
TNX-103 arms 21.1 − 5.8 = 15.3 m apart
So the split is driven from both sides. Slower walkers on placebo lost ground and slower walkers on the drug gained. Faster walkers on placebo gained 23.4 m, and faster walkers on the drug gained 5.8 m. Tenax’s slides say the fitter patients had “little room to improve”. Yet the placebo patients in that half improved more than any other group in the table.
A post hoc cut by quartile shows the same slope. The treatment difference was +32.4 m in the slowest quarter, +21.4 m in the second, −10.7 m in the third and −27.3 m in the fastest. Tenax reports a significant interaction between treatment and baseline walk in each of 100 imputed datasets (average P = 0.006). It also says regression to the mean does not explain the pattern.
There are two ways to read a pattern like this, and each carries a cost. If the slope is real, the drug helps slower walkers and may hold back faster ones, and a label would have to say so. If it is chance, the 26.3 m gain is no more reliable than the 17.6 m loss. Either way, the half of LEVEL that supports the drug comes attached to the half that does not.
Age and symptoms narrow the claim further. The age subgroups Tenax cites (+27.1 m at 71 and over, +37.6 m above 74) are not independent support, because the slower walkers were older. Their mean age was 71.8 years against 66.5 in the faster half. And symptoms did not follow the walk test even in the favoured half. In patients below the median, the symptom score differed from placebo by 2.0 points (95% CI −4.8 to 8.7; P = 0.57).
The bar for approval is relevant too. Tenax says the FDA had agreed that one phase 3 trial with a P value of 0.01 could support a filing. The below-median result, at a nominal 0.0112, would not have cleared that bar even as a primary analysis.
04TNX-103 mechanism of action: how oral levosimendan works
The TNX-103 mechanism of action is that of levosimendan, which does two things at clinical doses. It binds troponin C and makes heart muscle more responsive to the calcium already in the cell, which strengthens contraction. And it opens ATP-sensitive potassium (K-ATP) channels in the muscle of blood vessel walls. When those channels open, potassium leaves the cell, the membrane hyperpolarises, calcium channels close and the vessel relaxes. At higher concentrations the drug also inhibits phosphodiesterase III.
For PH-HFpEF, Tenax’s argument rests on the second action, and on veins in particular. Veins hold about 70% of the body’s blood. Much of it sits in the gut circulation as a reserve that exerts little pressure. The sympathetic nervous system can squeeze that reserve into the central circulation, which raises filling pressures in the heart and lungs. Physiologists call the pressure-generating share the stressed blood volume.
In HELP, a single 24-hour infusion lowered wedge pressure and central venous pressure at rest and on exercise without changing cardiac output. The investigators concluded that levosimendan had relaxed the veins and cut stressed blood volume. They found nothing to suggest that stronger contraction played a part. A 2021 review sets out the argument (Burkhoff et al., ESC Heart Failure, via PubMed). It draws on an analysis of HELP by Brener et al. One of the review’s four authors is Tenax’s chief medical officer and another works for Orion.
Why an oral dose lasts
The parent drug has a half-life of about one hour. Its active metabolite, OR-1896, has a half-life of 70 to 80 hours. Gut bacteria reduce levosimendan to an intermediate, OR-1855, which the body then acetylates to OR-1896, the Antila review explains. OR-1896 opens the same potassium channels as the parent. So the effect of a dose outlasts the parent drug by days.
How the LEVEL data fit the mechanism
A vein-relaxing drug should lower filling pressures, and with them NT-proBNP and lung pressures. LEVEL showed both. The step the trial did not confirm is the next one: that lower pressures let patients walk further. In HELP’s 37 selected patients they did. In LEVEL’s 241 they did not, except in the slower half.
Levosimendan’s record in late-stage trials
Tenax backed LEVO-CTS, HELP and LEVEL, and each missed its primary endpoint. So did Orion’s oral trial in ALS. LEVO-CTS is reported in Mehta et al., NEJM 2017 (via PubMed). The infusion’s approvals outside the United States rest on older trials in acute heart failure, which this brief does not review.
05LEVEL-2 trial: the new TNX-103 entry rule and what it can prove
LEVEL-2 was already running when LEVEL read out. Tenax paused its enrolment on 8 October 2026 and is amending the protocol site by site. The registry now shows the trial as active, not recruiting. Its entry criteria still ask only for a qualifying walk distance and give no limits.
LEVEL-2 entry criteria: where the new rule came from
The new rule admits patients who walk 100 to 300 m at baseline. Those who walk 301 to 400 m qualify only if their right atrial volume index (RAVI) is at least 25 mL/m². Tenax derived the rule after the fact. Its November release describes a post hoc analysis in two steps. Regression models, one for each arm, estimate every patient’s treatment effect. Classification trees then pick out the baseline features that separate responders from non-responders. Among LEVEL patients who fit the rule, TNX-103 beat placebo by 34 m (95% CI 15 to 53; nominal P = 0.001). The chief executive called the observations “hypothesis-generating”.
Rule-based subgroup interval 38 m wide
Implied size 117 × (40.7 / 38)² = about 134 patients
That estimate assumes similar variability in both groups. Run on randomised patients, 119 in the subgroup and 241 in all, the same sum gives 137, again 57%. Tenax says only that more than 50% of LEVEL patients meet the criteria. Three features of the rule weaken it as evidence.
First, it was fitted and scored on the same 241 patients. Neither the slides nor the November release reports a hold-out sample or cross-validation. A rule picked for its effect in one dataset tends to show a smaller effect in the next.
Second, the upper band reaches into the part of LEVEL that went the wrong way. Quartile three (333 to 384 m) showed −10.7 m and quartile four (above 384 m) showed −27.3 m. Patients from both can enter LEVEL-2 if they walk no more than 400 m and their RAVI is at least 25. Tenax’s own slide gives a normal RAVI of 25 in men and 21 in women and sets the upper limit of normal at 30. So the threshold starts inside the normal range.
Third, the 34 m estimate comes from a 12-week trial, and LEVEL-2 measures at 26 weeks. No placebo-controlled data exist for TNX-103 in PH-HFpEF beyond week 12.
What LEVEL-2 can detect
Standard deviation 55 m (LEVEL’s planning figure)
Standard error 55 × √(1/360 + 1/180) = 5.0 m
Detectable with 90% power at P = 0.01 (2.576 + 1.282) × 5.0 = about 19 m
That is a workable target if the true effect in the enriched group is near the 26 to 34 m seen after the fact. It is tight if the true effect is half that. Tenax has not said whether the amendment changes the size of the trial.
Sites and timing
LEVEL started in January 2024, and Tenax reported its randomisation target reached on 10 March 2026. That is about 26 months, or 0.23 patients per site per month (computed). LEVEL-2 has more than three times the sites. But it now needs 540 patients from a pool the new rule cuts by roughly four in ten, on the estimate above. In December 2025 Tenax expected enrolment to finish by the end of 2027. The registry lists primary completion on 30 June 2028, and the November release revises neither date.
The cash guidance has moved alongside. Tenax’s October slides said the company was funded through the second quarter of 2028, the quarter of that registry date. The November release says through fiscal 2028, and adds that the plan may change after the FDA meeting.
06TNX-103 and the FDA: one trial, P = 0.01 and a Type C meeting
Before LEVEL read out, Tenax had a route to a filing on a single trial. Its topline release referred to an earlier agreement with the FDA. Under it, one phase 3 study with a P value of 0.01 would be enough for a new drug application. LEVEL returned 0.63.
Tenax sent the topline data to the agency in mid-August and had asked for a Type C meeting by 8 October. On 5 November it said the meeting had been granted. It has given no date. Under the FDA’s published timelines, a Type C meeting is held, or a written response issued, within 75 days of the request. If the request went in on 8 October, that points to about 22 December 2026 (computed).
Three questions will decide what the meeting is worth. Will the agency still accept one trial at 0.01 when that trial is LEVEL-2, with an entry rule drawn from a failed study? Can LEVEL count as supporting evidence when its primary endpoint missed? And is walk distance still the right primary endpoint? Merck has chosen time to clinical worsening for its phase 3 trial in an overlapping population.
In August Tenax also said it would seek scientific advice from the European Medicines Agency. The November release does not mention it.
07Tenax Therapeutics (TENX) after LEVEL: cash, share price and a class action
Tenax shares lost 89.77% on 10 August 2026, the day LEVEL read out, and have stayed near that level since. The company now holds more cash than the market says it is worth.
Market value over cash 63.3 / 113.7 = 56%
Counting pre-funded warrants 53,452,869 × $1.39 = $74.3M, or 65% of cash
Cash less all liabilities 113.73 − 8.36 = $105.4M
Cash works out at $2.50 a share, or $2.13 counting the warrants, against a share price of $1.39 (computed). A company priced below its cash is being marked down for the spending ahead, with little credit for LEVEL-2. The stock closed on 19 November within 1.1% of its close on the day of the result (computed). So the ESC presentation, the enrichment plan and the Type C grant have not moved that view.
Shares outstanding rose almost fivefold in nine months, from 9.3 million to 45.5 million. Cash rose $16.2 million over a period in which the company lost $50.2 million. Additional paid-in capital rose by $61.8 million over the same nine months (computed from the balance sheets). The release does not break that figure down.
Chardan cut its rating to neutral on 11 August and its target from $27 to $5. Evercore ISI downgraded the same day and Guggenheim followed on 31 August, according to MarketScreener’s log of analyst actions.
A shareholder class action, Katz v. Tenax Therapeutics, was filed on 26 October 2026 in the Middle District of North Carolina, the company’s quarterly report says. It covers buyers of Tenax securities between 13 November 2023 and 7 August 2026. Law firms have advertised a lead-plaintiff deadline of 28 December. The claims are allegations and no court has ruled on them. Witfire covered another one-day collapse in Unicycive’s selloff after its complete response letter.
08TNX-103 vs sotatercept: the other late-stage programme
Merck is the other company with a late-stage drug in this territory. Sotatercept, sold as Winrevair for pulmonary arterial hypertension, was tested in CADENCE. That phase 2 trial enrolled patients who have both heart failure with preserved ejection fraction and disease of the lung vessels themselves.
CADENCE met its primary endpoint (Gomberg-Maitland et al., Circulation 2026, via PubMed). At the 0.3 mg/kg dose, pulmonary vascular resistance fell by 1.02 Wood units against placebo (95% CI −1.81 to −0.23; P = 0.004). Walk distance rose 20.3 m at that dose (95% CI 1.5 to 39.1). At 0.7 mg/kg the gain was 5.8 m, with an interval that crossed zero. The two drugs aim at different parts of the problem, and the populations only partly overlap. CADENCE required pulmonary vascular disease on top of heart failure, and its patients had a median resistance of 5.2 Wood units.
The phase 3 designs differ in a way that bears on Tenax’s FDA meeting. Merck’s CHORUS trial lists time to clinical worsening as its primary endpoint and puts walk distance among the secondaries. A fall of 15% or more in walk distance counts as a worsening event. The registry gives a start date of 26 November 2026 and primary completion on 8 January 2030, about 18 months after LEVEL-2’s listed date (computed).
09TNX-103 and India: a generic infusion, and no trial sites
Levosimendan is already a hospital generic in India. Lupin sells a 12.5 mg injection as Simenda at a listed price of ₹584.34, and Samarth Life Sciences sells Levosimed at ₹296.76, according to the pharmacy Truemeds. Both are infusions for acute heart failure. The oral form is a different product, and it is not approved anywhere.
India is absent from the trial programme. LEVEL ran only in the United States and Canada. LEVEL-2 lists 140 sites in 15 countries, and its 12 Asian sites are in South Korea and Taiwan. So an approval elsewhere would arrive without Indian patients in the data.
India’s rules allow a local trial to be waived for drugs approved in the United States, the UK, Japan, Australia, Canada or the EU. That waiver is not automatic. In July 2025 the regulator’s expert committee refused it for a new use of Roche’s faricimab, Medical Dialogues reported.
With no Indian sites, the programme has brought no site work to Indian contract research firms. Medpace is the listed collaborator on LEVEL. Witfire’s brief on India’s next CRO boom looks at where that work is heading. Tenax’s slides list a US patent on oral use and a European patent, both expiring on 15 December 2040.
10Witfire Risk Score: 6.2 / 10
This score covers the TNX-103 programme as it stands before the Type C meeting.
11What to watch
- 26 November 2026. Listed start of Merck’s CHORUS trial of sotatercept (MRK). A recruiting status on the registry would confirm it.
- About 22 December 2026. The 75-day mark for a Type C meeting or written response, if Tenax’s request went in on 8 October (TENX). The company has given no date.
- 28 December 2026. Lead-plaintiff deadline in Katz v. Tenax Therapeutics.
- The registry. LEVEL-2 returning to recruiting, the new walk-test limits in the entry criteria, and any change to the 540-patient target or the 30 June 2028 date.
- First quarter of 2027. The LEVEL-2 design paper and programme update Tenax has promised, and full-year results, which came on 10 March in 2026.
- A LEVEL publication. Heart rate, blood pressure and results by sex are still unpublished, and deaths by arm are unconfirmed.
TNX-103 is an active drug that has not yet shown a benefit patients can feel. LEVEL was big enough to find a 25-metre gain, and its interval stops at 17.8.
The rescue plan asks a second trial to confirm a subgroup. That subgroup has a mirror image in the same data and a symptom score that did not move. The entry rule built on it was fitted to the patients it is scored on. LEVEL-2 also runs for 26 weeks against LEVEL’s 12, so even the favourable half of LEVEL is only a partial guide to it.
Tenax says it has the cash to run the test, and its share price says the market does not expect it to pass. The FDA’s answer on one trial at 0.01 is the next real piece of information.
FAQTNX-103: questions and answers
What is TNX-103?
TNX-103 is an oral formulation of levosimendan developed by Tenax Therapeutics for pulmonary hypertension caused by heart failure with preserved ejection fraction (PH-HFpEF). Levosimendan is a calcium sensitiser and potassium channel opener. Its intravenous form is authorised in 60 countries for acute heart failure, but not in the United States or Canada. TNX-103 is investigational and is not approved anywhere.
Did TNX-103 work in the LEVEL trial?
Not on the measures the trial was built to test. LEVEL randomised 241 patients for 12 weeks. The gain in six-minute walk distance was 3.5 metres larger on TNX-103 than on placebo (P = 0.63), and the symptom score differed by 0.1 point. NT-proBNP fell 49% against placebo and an echo estimate of lung pressure fell 3.5 mmHg, but those analyses carry nominal P values.
What is the TNX-103 mechanism of action?
TNX-103 delivers levosimendan, which opens ATP-sensitive potassium channels in the muscle of blood vessel walls and makes heart muscle more responsive to calcium. In PH-HFpEF the proposed benefit comes from relaxing veins, mainly in the gut circulation. Less blood is then pressed towards the heart and lungs, and filling pressures fall. A long-lived metabolite, OR-1896, with a half-life of 70 to 80 hours, extends the effect.
Why did NT-proBNP fall if patients did not walk further?
A fall in NT-proBNP shows that pressure and stretch in the heart went down, which is what a vein-relaxing drug should do. Levosimendan has lowered natriuretic peptides in earlier studies. LEVEL did not show that lower pressures became better exercise capacity across the trial. In the slower-walking half, walk distance improved 26.3 metres against placebo. In the faster half it was 17.6 metres worse.
What is the LEVEL-2 trial?
LEVEL-2 (NCT07288398) is the second phase 3 trial of TNX-103. It plans 540 patients at 140 sites in 15 countries, randomised 2:1, with six-minute walk distance at 26 weeks as the primary endpoint. Tenax paused enrolment on 8 October 2026 to narrow entry. The amended rule admits patients who walk 100 to 300 metres at baseline. Those who walk 301 to 400 metres need a right atrial volume index of at least 25 mL/m².
What are the new LEVEL-2 entry criteria based on?
A post hoc analysis of LEVEL. Tenax modelled individual treatment effects and used classification trees to find baseline features linked to response. LEVEL patients who fit the rule gained 34 metres over placebo (95% CI 15 to 53; nominal P = 0.001). The rule was built and measured on the same patients, and Tenax’s chief executive described the findings as hypothesis-generating.
Is TNX-103 approved by the FDA?
No. TNX-103 is investigational and has no approval in any country. Tenax says the FDA previously agreed that one phase 3 trial with a P value of 0.01 could support a filing. LEVEL’s P value was 0.63. The company has been granted a Type C meeting to discuss the programme, with no date announced. Whether LEVEL-2 alone can carry an application is the main question for that meeting.
What are the side effects of TNX-103?
In LEVEL, headache (20.8% against 14.0% on placebo), palpitations (13.3% against 7.4%) and dizziness (11.7% against 7.4%) were more common on TNX-103. Treatment was discontinued in 8.3% of patients on the drug and 1.7% on placebo. Serious adverse events were balanced at 10.8% and 10.7%.
Why did Tenax Therapeutics stock fall?
Tenax shares closed at $1.375 on 10 August 2026, down 89.77% from $13.44, after LEVEL missed its primary endpoint. They closed at $1.39 on 19 November. The company held $113.7 million of cash at 30 September 2026 and says that funds operations through fiscal 2028, subject to the outcome of the FDA meeting. Its market value on shares outstanding was about $63 million, or 56% of that cash (computed).
How does TNX-103 compare with sotatercept?
They act differently and were tested in overlapping but different patients. Sotatercept, an activin signalling inhibitor from Merck, met its phase 2 primary endpoint in CADENCE by lowering pulmonary vascular resistance. Walk distance rose 20.3 metres at the 0.3 mg/kg dose. Its phase 3 trial, CHORUS, uses time to clinical worsening as the primary endpoint. LEVEL-2 uses walk distance.
Is levosimendan available in India?
Yes, as an intravenous infusion for acute heart failure. Brands include Simenda from Lupin and Levosimed from Samarth Life Sciences. The oral form, TNX-103, is not approved anywhere, and neither LEVEL nor LEVEL-2 has an Indian site.
Author. Dr. Akhilesh Vats is Editor-in-Chief of Witfire Elite and founder of ACME Research Solutions, a pharmaceutical contract research organisation.
Disclosure: Editorial analysis, not investment or medical advice. The Witfire Risk Score is our own framework, weighting Regulatory Exposure (30%), Competitive Displacement (25%), Capital Position (20%), Evidence Integrity (15%) and Execution & Credibility (10%) on a 1–10 scale. Scores are judgments, not forecasts.
Put this framework on your own file.
The same method used in these briefs, applied privately to a decision you are actually facing — a target, a partner, a facility, a filing.
- CMC & manufacturing risk assessment
- Regulatory / CRL readiness review
- Due diligence on a target or partner
- Investor brief or market analysis
- Commissioned editorial
- Something else
The signal before the market catches up.
A low-noise briefing on the pharma events that change strategy, risk and value. No more than one a week.
