What it is, and what it is not
Tesofensine is not a peptide. It is a small molecule, a substituted phenyltropane, formula C17H23Cl2NO, with no amino acids and no peptide bond. It is in a peptide encyclopedia because people searching peptide terms search for it, and because vendor and clinic pages sell it as one. A live listing titled as tesofensine peptide capsules carries, in its own specification field for amino acid sequence, the words “Not a peptide.”
What it is instead is a stimulant. It blocks reuptake at all three monoamine transporters, at 1.7 nanomolar for noradrenaline, 6.5 for dopamine and 11 for serotonin on the 2025 cryo-EM paper in Nature Communications. That paper states that, like cocaine, tesofensine belongs to the tropane alkaloid family, and notes the 50 fold lower affinity of cocaine for the dopamine transporter than tesofensine. It is a triple reuptake inhibitor with roughly 50 times the dopamine transporter affinity of cocaine, taken as a tablet.
It began at NeuroSearch in Denmark as NS 2330, passed to Boehringer Ingelheim for neurodegenerative disease, back to NeuroSearch for obesity, then to Saniona, and is licensed for Mexico to Productos Medix. All 13 registered human trials used an oral tablet or capsule once daily. Population modelling in 320 patients found half lives of 234 hours for the parent drug and 374 for its active metabolite, so exposure keeps climbing for weeks and persists for weeks after stopping.
It was a brain drug that failed first
Tesofensine was developed for Alzheimer’s and Parkinson’s disease, and that history is missing from almost every page about it. Boehringer Ingelheim ran four 14 week phase 2 trials from 2003 to 2005, two in Parkinson’s disease and two in Alzheimer’s disease, and all four failed. SCEPTRE, in early Parkinson’s disease, found UPDRS differences of minus 0.7 to minus 1.7 across three doses, none significant. ADVANS, in advanced disease, found no clear dose response relationship. The Alzheimer’s trial, NCT00153010, randomised 430 patients across 87 sites, posted no results and was never published: everything the public can verify about its outcome is one clause in a pharmacometrics paper calling it negative.
So the obesity programme exists because of an adverse event. Astrup and colleagues pooled those four trials in 2008, 740 patients against 228 on placebo, none on a weight loss programme. Weight change at 14 weeks was plus 0.5 percent on placebo, then minus 0.5, minus 0.9, minus 1.8 and minus 2.8 percent across 0.125 to 1.0 mg. Their own sentence records the pivot: on the basis of these results, the drug was now being developed for obesity management.
The obesity trials, with the numbers labelled
TIPO-1 is the trial every page cites: 203 patients across five Danish centres, 0.25, 0.5 or 1.0 mg or placebo for 24 weeks, in the Lancet in 2008, every arm including placebo on an energy restricted diet. Two sets of its numbers are correct and pages swap them. Placebo subtracted additional weight loss, which is what the abstract reports, was 4.5, 9.2 and 10.6 percent. Absolute mean weight loss was approximately 6.5, 11.2 and 12.6 percent, or 6.7, 11.3 and 12.8 kilograms, against 2.0 percent or 2.2 kilograms on diet plus placebo. Reporting 10.6 percent as absolute understates the drug; 12.8 kilograms as placebo subtracted overstates it. Completion was 161 of 203, 79 percent, so one patient in five did not finish.
Two things commonly cited do not exist. There is no TIPO-3 in any registry, the literature or the sponsor record, the registered obesity trials being TIPO-1, TIPO-2 and TIPO-4, so a page citing TIPO-3 is citing nothing. Nor is there any trial called “Obesity Reduction and Metabolic Improvement.” TIPO-4 is real, an open label extension in 140 TIPO-1 completers over 48 weeks with long term safety as its primary endpoint, and it has no posted results and no publication.
The phase 3 evidence is a press release. One phase 3 trial has ever been run, in Mexico: 372 adults in three arms of 124 on 0.25 mg, 0.5 mg or placebo for 24 weeks. The sponsor’s parent announced that both endpoints were met at p below 0.001, with a ten per cent average weight loss and more than half of patients losing more than ten per cent. It is not registered on ClinicalTrials.gov and has never been published. There are no arm by arm figures, no placebo arm result, no numerical heart rate or blood pressure data and no dropout rates.
The pivotal paper carries an Expression of Concern
No page reviewed for this entry mentions it. The Lancet published an Expression of Concern on TIPO-1 in April 2013, after a Danish Health and Medicines Authority inspection found adverse event recording by the contract research organisation incomplete, because its staff did not consider as noteworthy the recurrence of events. The conclusion, verbatim: the overall side-effect profile that is published in The Lancet is not in accordance with the actual course of the trial. With a limit, also verbatim: data on primary and secondary end-points, and serious adverse events, were reported in accordance with the source data.
The authors’ own published response is worse than the inspection finding. They confirmed that investigators had been wrongly instructed by the monitors not to register headache, migraine, stress, and depression as adverse events if patients reported these conditions before randomisation, found no difference between inspected and uninspected sites, and concluded that listed adverse events from all centres were under-reported in the published paper. Not two bad sites, the whole trial. Two of the four named events are psychiatric, and corrected figures were never published. The efficacy stands. The adverse event table does not.
Safety
The heart rate rise is real, dose dependent, and present at the therapeutic dose in every monotherapy study. In the 968 patient pooled neurodegeneration analysis it ran minus 0.4 beats per minute on placebo, then plus 2.1, plus 4.2, plus 6.0 and plus 6.8 across 0.125 to 1.0 mg, significant from 0.25 mg upward. In TIPO-1 it was plus 7.4 at 0.5 mg. The 2018 Drugs review calls elevations in blood pressure and pulse rate the dose limiting adverse effects.
The blood pressure result needs a subtlety nobody else supplies. No significant rise was detected at 0.25 or 0.5 mg; one was detected at 1.0 mg. But a null result in a trial producing about 11 percent weight loss is not neutral, because losing that much weight should by itself lower blood pressure by several millimetres of mercury. The best framing available, and it comes from a secondary analysis rather than the trial report, is that at 0.5 mg the pressor effect ran roughly 4 to 6 mmHg above the reduction that weight loss alone would have predicted. That blood pressure did not rise is literally true and understates the pharmacology.
The strongest evidence that the signal is serious is the developer’s own behaviour. Saniona stopped developing tesofensine on its own: every trial it ran after 2014 tested Tesomet, 0.5 mg tesofensine fixed with 50 mg of the beta blocker metoprolol. A company does not build a beta blocker into its lead asset over a signal it considers trivial. Cardiovascular outcomes have never been measured in any trial, and the planned two year study was never run. Sibutramine, with a comparable haemodynamic profile, looked acceptable on surrogates for a decade and was withdrawn in 2010 after an outcomes trial.
The common events in TIPO-1 were dry mouth, nausea, constipation, hard stools, diarrhoea and insomnia, with withdrawal for adverse events near 13 percent against 6 on placebo. The best per event figures published are from the 21 patient Tesomet trial, sleep disturbance 50 percent against 13, dry mouth 43 against 0, headache 36 against 0, and those belong to the combination. It also holds the one documented treatment related psychiatric serious adverse event, an exacerbation of pre-existing anxiety. Every obesity trial excluded psychiatric illness, so the measured rates are a floor.
On abuse liability there is one human study. Schoedel and colleagues gave single doses to 52 recreational stimulant users against placebo, d-amphetamine 30 mg as a positive control, and bupropion and atomoxetine; tesofensine did not differ from placebo, and the authors concluded its abuse potential was no greater than bupropion or atomoxetine. One single dose study. There is no repeated dose, withdrawal, tolerance or post marketing data, because the drug has never been marketed. WADA lists it under S6 stimulants, prohibited in competition.
Interactions matter here because of the half lives. Protocols prohibited monoamine oxidase inhibitors for 8 weeks before randomisation and serotonin reuptake inhibitors for 6 weeks. With a 234 hour parent and a 374 hour metabolite, washout runs in weeks in both directions. On CYP3A4, the itraconazole study showed a 9 percent rise in exposure on non-compartmental analysis and a 63 percent rise on semi-mechanistic modelling of the same data.
A 2026 anti-doping paper documents tesofensine marketed online as a dietary supplement for weight management, and dosed six volunteers with 483 micrograms of material bought that way.
Regulatory status, including the Mexico correction
Tesofensine is not approved anywhere the public record confirms. It has never been approved by the FDA, has held investigational status only, has no European authorisation and no ATC code. Mexico is the correction that matters. In February and March 2023 the Comite de Moleculas Nuevas issued a favourable opinion with expert observations, and Mexican headlines used the verbs aprueba and avala, which is where the approval claim comes from. The sponsor’s own release said the favorable opinion does not represent a market authorization. On 6 November 2024 the sponsor stated that Medix has not received approval from the Mexican regulatory agency (Cofepris) for tesofensine. The sponsor said in February 2025 that a revised dossier was expected to be resubmitted that month. No confirmation that it was appears in the public record.
So there is no granted authorisation, no approved brand name, no approved dose and no approved indication text anywhere. Tesomet and Nupenta, named on some pages as approved Mexican products since 2023, could not be verified in any primary or regulatory source. Nothing is in progress either: of 13 registered trial records, 11 are completed and 2 were withdrawn without enrolling anyone.
What the record gets wrong
The route, first. A weight loss clinic markets tesofensine injections. Injection was never used in any of the 13 registered trials, all of which were oral, and the sponsor’s own framing is a tablet requiring no titration. No efficacy or safety trial has ever used an injected dose. The only parenteral human exposure on record is a pharmacokinetic study that infused 0.3 to 1.2 mg intravenously over six hours in 21 healthy subjects to model enterohepatic recirculation, which establishes nothing about a repeated injected dose and was never intended to.
Second, and this one inverts the drug. No significant difference in heart rate or blood pressure is a finding about Tesomet, tesofensine plus 50 milligrams of metoprolol, a beta blocker included to blunt exactly that effect. Quoted as a tesofensine finding, it turns the compound’s most important safety characteristic upside down. Third, the dopamine transporter potency: some secondary tabulations give 65 nanomolar where the primary structural paper gives 6.5, which looks like a decimal error.
The consumer tier is not uniformly bad. Several pages state accurately that it is not FDA approved and investigational only, and one makes the most useful observation anyone has made about its psychiatric profile: trials excluded active psychiatric disease, so the real world rate may be higher.
Every page saying tesofensine is approved in Mexico is repeating a newspaper headline about an advisory opinion that the sponsor itself said does not represent a market authorisation, and that the sponsor confirmed two years later had not produced an approval. There is no approved dose of tesofensine in any country. The only doses that exist are the ones trials administered, and the pivotal trial’s own adverse event table is an undercount its authors have admitted in print.