L-THP (Levo-Tetrahydropalmatine)

Chemical identity, and three records that overlap

Three registries hold levo-tetrahydropalmatine, and nothing in them conflicts. PubChem's entry is CID 72301, CAS 483-14-7, formula C21H25NO4, molecular weight 355.4 g/mol, InChIKey AEQDJSLRWYMAQI-KRWDZBQOSA-N, UNII 3X69CO5I79, under the IUPAC name (13aS)-2,3,9,10-tetramethoxy-6,8,13,13a-tetrahydro-5H-isoquinolino[2,1-b]isoquinoline. FDA GSRS repeats that same UNII, the same CAS and the same formula, and marks its record complete. ChEMBL indexes the molecule as CHEMBL487182, maximum development phase 2. The alkaloid itself occurs in the rhizome of Corydalis yanhusuo and in tubers of Stephania rotunda.

The racemate is a separate record. CID 5417 carries the same formula and mass but the stereochemistry-free InChIKey AEQDJSLRWYMAQI-UHFFFAOYSA-N and CAS 2934-97-6, and PubChem gives it the primary name Rotundine. That is a problem, because rotundine is also the name under which the levo enantiomer is dispensed as a medicine in China and Vietnam. A third record, CID 12220312, is tert-butylcyanoketene, an unrelated small molecule that has nothing to do with this chemistry, and its synonym list nonetheless includes Tetrahydropalmatine DL-form.

Name-based lookup can therefore land on any of the three, and the synonym lists cross-contaminate in both directions: the levo record carries Rotundine and Tetrahydropalmatine, dl- among its synonyms, and the racemate record carries the shared stereochemistry-free CAS 10097-84-4. Matching on InChIKey rather than on name is the only reliable way to establish which enantiomer a source is describing. Where a paper or a label says only tetrahydropalmatine, the stereochemistry is often not recoverable at all.

Three registered trials, one withdrawn, one unpublished

A ClinicalTrials.gov intervention search across tetrahydropalmatine, l-THP and rotundine returned four studies, one of which tests a ten-ingredient supplement rather than the compound. The remaining three all name a single sponsor, the University of Maryland, Baltimore. NCT01631383 was a phase 1 pharmacokinetic and safety study in people with a history of cocaine use, actual enrolment 20, completed June 2017. NCT02139761, a phase 2 trial in cocaine use disorder, is recorded as withdrawn with an actual enrolment of zero and the stated reason that study funding had not been established.

NCT02118610 was the largest. Sixty-three participants with schizophrenia were randomised under quadruple masking to l-THP 30 mg twice daily or placebo, 30 and 33 respectively, and 61 were analysed. The study completed in June 2019 and posted results. Brief Psychiatric Rating Scale totals moved from 41.3 (SD 8.8) to 36.6 (9.0) on the compound and from 40.3 (6.9) to 37.3 (8.7) on placebo over four weeks. The MATRICS cognitive composite moved from 28.0 to 28.8 against 27.7 to 29.0. No statistical analysis was posted alongside either outcome.

Searching PubMed for tetrahydropalmatine combined with schizophrenia returned zero records on 18 August 2026. Seven years after completion, the largest randomised trial of this molecule conducted outside China exists only as a registry results table, which is why no review of the compound discusses it. NCATS Inxight Drugs lists the highest development status as investigational, phase 2, and names no approval jurisdiction. The molecule is not an approved medicine in the United States or the European Union.

The heroin trial that most citations stop at

Yang and colleagues randomised 120 heroin-dependent inpatients at a detoxification clinic in Hunan, China, recruited between June 2000 and February 2001 and published in 2008. Participants were detoxified for seven to ten days first, then received l-THP 60 mg twice daily or placebo for four weeks as inpatients, followed by four weeks of observation and three months of follow-up with random urinalysis. The abstract reports an abstinence rate of 47.8 per cent against 15.2 per cent in the placebo group by log-rank test, P below 0.0005. That sentence is the one that travels.

Retention ran the other way. Over the first fourteen days the placebo group lost nobody, a retention probability of 100 per cent, while the treated group fell to 72 per cent (95% CI 60 to 83), and the log-rank test on dropout returned P below 0.0001 against the compound. The abstinence comparison was then computed only among those still enrolled after that fortnight, 44 on l-THP against 59 on placebo, so it conditions on a post-randomisation event whose rate differed sharply between arms. The paper states this openly. Pages quoting the 47.8 per cent figure generally do not.

Two internal inconsistencies sit in the published text. Table 1 assigns 61 participants to placebo and 59 to l-THP; the retention narrative and Table 3 assign 59 to placebo and 61 to l-THP. The abstract gives the abstinence rate as 47.8 per cent and the results section as 47.7 per cent, the latter matching 21 of 44. No registration identifier appears in the paper, and no registry entry corresponding to it was located.

Symptom scores fell in both arms. The time-dependent regression reported a larger decline on l-THP for the protracted abstinence withdrawal total (Z = -9.73, P below 0.0001), and separately for somatic symptoms, insomnia and craving, but not for mood state (Z = 0.257, P = 0.797). Among participants who terminated early, the authors recorded no reports of adverse effects and no acute hepatic toxicity attributed to the medication, and offered dopaminergic antagonism as a hypothesis for the early dropout rather than a measured cause.

Human pharmacokinetics rest on nine subjects

Hassan and colleagues reported the phase 1 study registered as NCT01631383 in 2017. Twenty-four adult men who used cocaine were randomised, 20 completed, and 19 provided complete pharmacokinetic data, of whom nine had received the compound. The regimen was 30 mg orally every twelve hours, seven doses across days 1 to 4, with an intranasal cocaine challenge of 40 mg on day 4. Geometric-mean Cmax was 42.8 ng/mL, median Tmax 1.5 h, geometric-mean terminal half-life 13.3 h, and AUC to infinity 396.1 h·ng/mL. The accumulation ratio on AUC over the first twelve hours was 1.2.

Cocaine exposure was unchanged by co-administration. AUC to infinity was 211.5 against 261.4 h·ng/mL and Cmax 83.3 against 104.5 ng/mL for the l-THP and placebo groups, and the authors reported no significant difference in heart rate, blood pressure, complete blood count or ECG. Side-effect counts were 22 (48 per cent) on the compound and 24 (52 per cent) on placebo. Every one of those figures describes nine people over three and a half days, which is what a phase 1 safety study is designed to deliver and is not a characterisation of longer exposure.

Rodent figures do not transfer. The 2022 review by Du and colleagues (PMID 35559252) cites rat half-lives of 0.44 h and 4.49 h without reconciling them, against 13.3 h in the human study, and reports that formulation moves the number again: self-microemulsifying preparations raised relative bioavailability several-fold over a plain suspension in rat and rabbit models. No human study of repeated administration beyond four weeks has been located, and no human metabolite profile for l-isocorypalmine, the demethylated metabolite that carries a different receptor profile, was found.

The receptor profile is not settled

Descriptions of what this molecule binds differ by paper. Mantsch and colleagues in 2007 described low-affinity D2 antagonism together with higher-affinity D1 antagonism and an interaction at D3. Wang and colleagues in their 2012 review (PMID 22300097) described antagonism at D1 and D2 with additional actions at D3, alpha-adrenergic and serotonin receptors. Liu and colleagues in 2019 (PMID 31172225) described D1 agonism and D2 antagonism, and reported in a mouse nerve-ligation model that a D1 antagonist abolished the antinociceptive effect while a D2 agonist abolished both the antinociceptive and the sleep effect.

Xu and colleagues screened more than 40 targets in 2013 and reported that l-THP bound only to dopamine D1 and D5 among them, at lower affinity than its metabolite l-isocorypalmine, which bound D1 and D5 as a partial agonist and D2, D3 and D4 as a moderate-affinity antagonist. Taken literally that finding removes D2 and D3 from the compound's own profile and assigns them to the metabolite, which is not how the compound is described anywhere downstream.

Curated binding data narrow the question without closing it. ChEMBL holds Ki values for this molecule drawn from four documents: D1 between 124 and 231 nM, D5 305 nM, D2 between 388 and 5000 nM, D3 1371 and 1420 nM, D4 1000 nM, and 5-HT1A 5000 nM. The D1 values agree across laboratories within a factor of two. The D2 values span more than an order of magnitude. Whether the D2 and D3 sites are engaged at concentrations a body reaches depends on which of those papers is being read, and no source reconciles them.

The Jin Bu Huan record

Three unrelated children in Colorado developed life-threatening bradycardia with central nervous system and respiratory depression during 1993 after swallowing tablets of a Chinese herbal product sold for pain. Analysis of tablets recovered from the families, by nuclear magnetic resonance and gas chromatography with mass spectrometry, found the tablets to be 36 per cent levo-tetrahydropalmatine by weight, 28.8 mg per tablet, with no other drug detected in the tablets or in the children's samples. The package identified the plant source as a different species entirely.

Adults presented with a different syndrome. Woolf and colleagues described seven previously healthy patients with acute hepatitis after a mean of 20 weeks of use, range 7 to 52 weeks, resolving in six within a mean of 8 weeks; reducing intake improved liver tests in one patient, and re-exposure in two others produced abrupt recurrence. Horowitz and colleagues in 1996 set out three children and three adults together and drew the two patterns as distinct, acute ingestion in children against prolonged use in adults, noting that the misidentified label delayed identification of the responsible alkaloid. Picciotto and colleagues (PMID 9537855) added a biopsy-proven chronic hepatitis with moderate fibrosis that resolved on discontinuation.

The NIH LiverTox monograph records more than a dozen cases, onset between 2 and 24 weeks, a hepatocellular pattern with marked aminotransferase elevation and minimal alkaline phosphatase change, re-exposure positive in two cases and negative in two, and a causality likelihood score of cannot be assessed. The implicated product was withdrawn, and the monograph records no further reports of liver injury attributed to it since 2002.

Acute overdose is still being described where the compound is dispensed without prescription. Duc and colleagues reported 37 consecutive patients admitted for acute rotundine poisoning at one Vietnamese hospital between April 2024 and June 2026: median age 27 years, 56.8 per cent female, central nervous system depression predominant, Glasgow Coma Scale below 8 in 5.4 per cent, corrected QT interval at or above 440 ms in 16.2 per cent, four patients (10.8 per cent) mechanically ventilated, and no deaths.

Product content, and one retracted paper

Luis and colleagues analysed 14 Corydalis yanhusuo supplement products in 2024, covering most of what was then offered online in the United States, quantifying 19 alkaloids by HPLC with diode-array detection after weak cation-exchange extraction. Total alkaloid content ranged from below the limit of quantification to about 11 mg/g. Five products clustered at 9.5 ± 1.6 mg/g, close to the 12.7 mg/g measured in an extract the authors prepared from rhizome. The rest sat at 1.8 ± 0.9 mg/g or below quantification. One product carried roughly 5 mg/g of tetrahydropalmatine alone, which the authors read as adulteration rather than plant content.

One paper in the animal literature carries a formal integrity flag. Wang and colleagues reported in 2018 that l-THP slowed abdominal aortic aneurysm progression in elastase-perfused rats, 36 Sprague-Dawley animals in three groups of 12 (PMID 29388563). Med Sci Monit retracted it in November 2022 over concerns about the originality of the figure images, notice at 28:e938902, PMID 36349696, and PubMed types the original as a Retracted Publication. That paper supports no claim on this page. Publication types and correction links were checked on every PMID cited here, and no other retraction, expression of concern or erratum was returned.

What is not known

No pharmacokinetic parameter for this compound in people rests on more than nine subjects, and none describes exposure beyond four days. Nothing characterises repeated administration in humans past the four weeks of the 2008 Hunan trial, and that trial's positive result is conditioned on a post-randomisation subgroup whose selection differed between arms. No prospectively registered efficacy trial has reported in a journal: the phase 2 cocaine study was withdrawn before enrolment, and the schizophrenia trial's results exist only as a registry table with no statistical analysis attached. No abuse-liability assessment of any kind has been published, so the descriptor most often applied to the molecule is the one with the least support behind it. The receptor profile is contested at D2 and D3 across a range spanning more than an order of magnitude, and no study reconciles the panels. Human metabolite data are absent, which matters because the demethylated metabolite carries a broader dopamine-receptor profile than the parent in the one screen that measured both. No carcinogenicity, reproductive or developmental study was located. The hepatotoxicity signal comes from case series attached to a withdrawn multi-ingredient product, with a formal causality score of cannot be assessed; whether the isolated alkaloid carries the same liability at the exposures a supplement delivers has not been tested, and one product survey found tetrahydropalmatine content varying from undetectable to a concentration the analysts read as deliberate adulteration.

Questions