LSD
A circulatory stimulant that failed, became a counterculture, and is now in Phase 3 for generalized anxiety.
LSD has the longest and strangest development history of any psychoactive compound: synthesised at Sandoz in 1938 in a search for a circulatory stimulant, shelved for five years, discovered psychoactive by accident, marketed as a prescription psychiatric adjunct for nearly two decades, studied in roughly a thousand papers and tens of thousands of patients, then abolished from medicine by a political reaction it had largely provoked. It re-entered controlled research in 2014. As MM120 it is now the subject of the first Phase 3 programme for LSD in any indication — with generalized anxiety disorder as the lead, and Week-12 HAM-A as the primary endpoint.
It was never designed to be psychoactive.
LSD-25 was the twenty-fifth compound in a Sandoz ergot programme hunting analeptics and uterotonics. Its psychoactivity was found five years later, by accident, and the discovery is why the compound exists at all.
The prior evidence base is enormous and unusable.
Roughly 1,000 papers and tens of thousands of patients between 1949 and 1970 — almost none to modern standards of randomisation, blinding or outcome measurement. Volume is not evidence.
Duration has a solved structural explanation.
The 2017 crystal structure showed an extracellular “lid” trapping LSD in the binding pocket, producing an extraordinarily slow dissociation rate. The 8–12 hour experience is a receptor-kinetics fact, not a metabolic one.
The Phase 2b GAD result is the strongest in the field.
MMED008: 100 µg produced a 7.7-point HAM-A advantage over placebo at Week 12 (−21.9 vs −14.2; p < 0.003; Cohen's d = 0.81), with 65% response and 48% remission sustained. That effect size is unusual for anxiety.
Functional unblinding is the central methodological problem.
You cannot blind 100 µg of LSD against inert placebo. Panorama's 50 µg arm exists specifically to confound dose guessing — the most serious attempt yet to answer the objection that sank MDMA.
Abuse liability is low; psychological risk is not.
LSD is not reliably self-administered by animals and has no established human lethal dose. The real hazards are psychiatric — acute adverse reactions, latent psychosis precipitation, and HPPD.
Eighty-eight years, in three acts
Origins:
the twenty-fifth compound
LSD came out of a rational, unglamorous industrial programme. Sandoz in Basel had spent years working the ergot alkaloids — the constituents of Claviceps purpurea, the rye fungus responsible for medieval ergotism — for cardiovascular and obstetric drugs. Arthur Stoll had isolated ergotamine in 1918. Albert Hofmann's brief was to make new lysergic acid amides and look for analeptic and uterotonic activity.1
He synthesised the diethylamide — the twenty-fifth in his series, hence LSD-25 — on 16 November 1938. Animal screening found nothing of pharmacological interest beyond some restlessness, and the compound was set aside. Hofmann's own account is that he returned to it five years later on a hunch. On 16 April 1943, during resynthesis, he absorbed a trace amount and experienced a marked, unexplained alteration of perception. Suspecting the new compound, he took a deliberate dose of 250 µg on 19 April 1943 — an amount he believed to be threshold and which is in fact several times a standard dose. The bicycle ride home from the laboratory has become the field's founding anecdote.1,2
Two things about that episode matter analytically. First, LSD's potency — active in the tens of micrograms — was itself the discovery; nothing in medicine then worked at that scale, and it is why the standard animal screens missed it. Second, the dose was a hundred-fold error in the safe direction, which set the field's initial impression of the drug's intensity.
Sandoz did what a pharmaceutical company does: it commercialised. From 1947 LSD was sold as Delysid, with two labelled uses — as an adjunct to psychoanalysis and psychotherapy, on the reasoning that it loosened repression and surfaced material; and, remarkably, as a tool for psychiatrists and trainees to induce a model psychosis in themselves in order to understand their patients.2,3 Sandoz supplied it to investigators largely free of charge, which is why the literature grew so fast.
Delysid was a medicine with a psychiatric indication, a dosing table and a package insert, for eighteen years. Everything that followed happened to a licensed drug.
This is the fact most often lost. LSD is not a compound that failed to become a medicine. It is a compound that was a medicine and was withdrawn for reasons largely external to its clinical record — which is precisely why the current programme has to rebuild the evidence from scratch rather than cite the old work.
How a medicine became a movement, and then a felony
LSD's sociocultural history is not decoration around the pharmacology. It is the direct cause of a twenty-year research blackout, and it is the reason every psychedelic programme since has had to manage political risk as a primary concern.
Three threads ran in parallel through the 1950s and 60s.
The intelligence thread. From 1953 the CIA's MK-ULTRA programme investigated LSD as an interrogation and incapacitation agent, administering it covertly to unwitting subjects including agency employees, prisoners and psychiatric patients. The programme is the origin of the field's deepest ethical injury and of a durable public association between the drug and institutional deceit; its records were largely destroyed in 1973, and what is known comes from the 1975 Church Committee and subsequent disclosure.4 Any modern discussion of psychedelic informed consent is downstream of this.
The clinical thread. Meanwhile a genuine research enterprise ran in Canada, the UK, Czechoslovakia and the US. Two paradigms competed: European psycholytic therapy used repeated low doses to facilitate conventional analysis, while the North American psychedelic model used a single high dose to induce a transformative peak experience. Osmond and Hoffer in Saskatchewan worked on alcoholism; Grof accumulated thousands of sessions in Prague. This is the body of work now cited as precedent — and it was, methodologically, of its time.3,5
The cultural thread. Timothy Leary and Richard Alpert's Harvard Psilocybin Project (1960–63) began as research and ended in dismissal, after which Leary became an advocate rather than an investigator. Ken Kesey's Acid Tests, Owsley Stanley's mass production, and the 1967 San Francisco moment moved LSD from clinic to street at scale. By 1965 Sandoz — facing an unmanageable liability position — had withdrawn Delysid. California criminalised possession in October 1966; federal Schedule I placement came with the Controlled Substances Act of 1970, and the 1971 UN Convention on Psychotropic Substances globalised it.3,6
The prohibition was not triggered by a toxicology signal. It was triggered by uncontrolled non-medical distribution and by researchers who became advocates. That is why the modern programmes are conspicuously conservative — corporate sponsors, ODT formulations, registrational endpoints, no manifestos. The discipline is a direct response to 1966.
Research did not resume meaningfully until the 1990s, and the first controlled LSD therapy trial did not publish until 2014.7 Forty-four years of a compound with a prescription history sitting untouched is the single largest discontinuity in modern psychopharmacology.
Mechanism of action
LSD is a partial agonist at the 5-HT2A receptor, and that interaction is necessary for its subjective effects. It is also strikingly promiscuous — with meaningful activity across 5-HT1A, 5-HT2B, 5-HT2C, 5-HT5A, 5-HT6, 5-HT7, and dopamine D1/D2 and adrenergic receptors. It is the least selective of the classic psychedelics.8
The 5-HT2A requirement is causally established in humans
This is a rare case in psychiatry where target attribution is not inferential. Pre-treatment with the 5-HT2A antagonist ketanserin abolishes LSD's subjective effects in healthy volunteers, and abolishes its effects on brain connectivity and on the personal meaning attributed to experience.9 Whatever else LSD does, the experience runs through 5-HT2A.
Why it lasts twelve hours — the lid
LSD's plasma half-life is only a few hours, which never explained an 8–12 hour experience. The 2017 crystal structure of LSD bound to a human serotonin receptor resolved it: extracellular loop 2 folds over the orthosteric pocket, forming a lid that traps the ligand and produces an exceptionally slow dissociation rate. Mutating the lid residue accelerated both dissociation and the decay of signalling.10 Duration is therefore a property of receptor residence time, not of clearance — an important point, because it means the duration cannot be formulated away. Any short-acting LSD would have to be a different molecule.
Functional selectivity compounds this. LSD engages both Gq and β-arrestin 2 pathways, and the structural work implicates biased signalling in the profile.10 The therapeutically relevant arm has not been isolated.
Systems level, and plasticity
Neuroimaging under LSD shows increased global functional connectivity, reduced within-network integrity of the default mode network, and desegregation of normally distinct networks — findings that underwrite Carhart-Harris and Friston's entropic brain and later REBUS accounts, in which psychedelics relax the precision of high-level priors and so permit revision of entrenched models.11 This is a framework, not a mechanism, and should be labelled as such; it generates predictions but has not been decisively tested.
The molecular convergence point is neuroplasticity. LSD promotes structural and functional plasticity — dendritic spine growth and synaptogenesis — in the psychoplastogen framework that also covers psilocin, DMT and ketamine.12 Note where this leaves the field: LSD, psilocybin, methylone and 4-OH-DiPT all now claim a plasticity mechanism, by four different receptor routes. That convergence is either the real biology or the field's current favourite explanation.
The promiscuity problem
LSD's breadth of receptor engagement relative to the tryptamine psychedelics. Wider engagement is not obviously worse — but it makes attribution harder and the cardiovascular ledger longer.
The first research era, and what survives of it
Between 1949 and 1970 LSD generated on the order of a thousand publications and was given to tens of thousands of patients across alcoholism, depression, anxiety, obsessional states, autism, schizophrenia and terminal illness.3 Almost none of it would support a modern regulatory filing. But one part of it has been re-analysed properly, and it is worth knowing.
Alcoholism. Krebs and Johansen's 2012 meta-analysis pooled six randomised controlled trials of a single dose of LSD for alcohol misuse, n = 536, conducted between 1966 and 1970. It found a significant benefit on alcohol misuse at first follow-up (odds ratio ~1.96), diminishing over subsequent months.14 That the trials were randomised at all is unusual for the era; that the effect attenuated is itself informative, and consistent with the modern single-dose durability question.
End-of-life anxiety. The most methodologically defensible modern re-entry point was Gasser's double-blind randomised trial, published 2014 — LSD-assisted psychotherapy for anxiety associated with life-threatening illness, in 12 patients. It reported reductions in state and trait anxiety with 20 µg active-placebo control, sustained at 12-month follow-up.7,15 Tiny, but it re-established that controlled LSD therapy trials could be run at all, and its indication — anxiety — prefigures the current programme.
What did not survive. The model-psychosis paradigm collapsed once it became clear that the LSD state resembles psychosis only superficially. Psycholytic therapy as a system produced no replicable controlled evidence. And the mid-century claims for schizophrenia and autism have no modern standing whatsoever. Anyone citing “forty thousand patients” as evidence of safety should be asked what was measured, and how.
The modern Swiss programme under Matthias Liechti is the bridge that made Phase 3 possible: systematic dose-response, pharmacokinetics, and placebo-controlled characterisation of the acute effects of defined LSD doses in healthy volunteers, published across the 2010s and 2020s.13 This is where the dosing rationale actually comes from — not from 1965.
MM120: LSD's first Phase 3
MM120 ODT is MindMed's pharmaceutically optimised lysergide D-tartrate, formulated with Catalent's Zydis orally disintegrating technology for faster absorption, improved bioavailability and reduced gastrointestinal side effects.16 The formulation is the proprietary asset; the molecule is not patentable.
MMED008 — the Phase 2b that started it
MMED008 was a multicentre, parallel-group, randomised, placebo-controlled, double-blind dose-ranging study in moderate-to-severe GAD, presented at the American Psychiatric Association in May 2024.17 It met its primary and key secondary endpoints and established a dose-response relationship, with 100 µg identified as optimal.16,18
The headline numbers: at Week 12, MM120 100 µg produced a 7.7-point HAM-A improvement over placebo (−21.9 vs −14.2; p < 0.003; Cohen's d = 0.81), with a 65% clinical response rate and 48% clinical remission rate sustained to Week 12 from a single dose.18,19 Adverse events were mostly mild-to-moderate, transient, confined to the dosing day, and consistent with the expected acute effects of LSD — visual changes, nausea, headache.16,20 FDA granted Breakthrough Therapy designation for GAD in March 2024.18
A d of 0.81 sustained twelve weeks after one dose is, if it replicates, a larger effect than anything in the licensed GAD armamentarium. The whole question is whether it replicates, and whether it is a drug effect.
The blinding problem, taken seriously
Nobody receiving 100 µg of LSD is in doubt about it. Functional unblinding — where participants infer assignment from perceptual effects and expectancy then drives the reported outcome — is the objection that FDA raised against MDMA-assisted therapy in 2024, and it is the single greatest threat to this programme.
MindMed's answer is design diversity across the two GAD trials. Voyage (MM120-300, NCT06741228) is a two-arm US study of 100 µg against placebo. Panorama (MM120-301, NCT06809595) adds a 50 µg arm whose explicit function is to confound participants' ability to identify their assignment, and recruits across the US and Europe.16,21 The company states this complementary-design approach aligns with FDA guidance and builds on the Phase 2b finding that the effect was not attributable to functional unblinding.16
This is a genuinely thoughtful response, and worth noting as a template. A low active dose arm is the same device Reunion used in RECONNECT with its 1.5 mg subperceptual comparator — the field is converging on it.
Status, and one open item
Voyage dosed its first patient in December 2024 with 12-week topline guided for H1 2026; Panorama dosed first patient in January 2025 with topline guided for H2 2026 and estimated completion May 2027; Emerge is a third Phase 3 in major depressive disorder.16,21,22,23 We were unable to locate a public Voyage topline readout as at 4 August 2026 despite the H1 2026 guidance. This should be checked directly before the report is circulated — if Voyage has reported, it is the most important single data point for any GAD programme and supersedes the Phase 2b figures above.
MM120 ODT clinical programme
MM120-300 · NCT06741228
MM120-301 · NCT06809595
Abuse liability and harm profile
LSD's Schedule I placement asserts high abuse potential. The pharmacological evidence for that specific claim is weak, and it is worth being precise about which risks are real.
Dependence liability is low. LSD is not reliably self-administered by laboratory animals — the standard reinforcement assay — it produces no withdrawal syndrome, and tolerance develops so rapidly that daily use is self-limiting. Compulsive use patterns are rare in epidemiological data.24 This is the basis on which petitioners have repeatedly challenged class-level “hallucinogen dependence” assertions in scheduling proceedings.
Physiological toxicity is low. No human lethal dose has been established, and the therapeutic index is extremely wide. Acute physiological effects are sympathomimetic — raised blood pressure and heart rate, mydriasis, tremor.24
The real risks are psychiatric. Three matter for a clinical programme: acute adverse psychological reactions during the 8–12 hour session, requiring trained supervision throughout; precipitation of prolonged psychosis in vulnerable individuals, which is why every modern trial excludes personal or first-degree-relative history of psychotic or bipolar disorder — Panorama's criteria do exactly this;21 and hallucinogen persisting perception disorder (HPPD), a recognised DSM diagnosis involving persistent visual disturbance after use. HPPD's incidence is poorly quantified and its risk factors are unknown, which is a genuine gap rather than a settled reassurance.24,25
The 5-HT2B question. LSD is a 5-HT2B agonist, and chronic 5-HT2B agonism is the mechanism by which fenfluramine and pergolide caused valvular heart disease. For a single supervised dose this is not a meaningful concern. For repeated low-dose regimens — the microdosing pattern — it is an unresolved theoretical risk that has never been adequately studied, and it deserves flagging whenever microdosing is discussed as though it were the conservative option.8,24
On microdosing specifically: the placebo-controlled literature has been largely disappointing. Well-controlled studies have generally failed to show the cognitive and mood benefits reported in open-label and self-report contexts, with expectancy accounting for much of the effect.26 Nothing in the MM120 programme depends on microdosing, and the two questions should not be conflated.
Where LSD sits in the commercial landscape
LSD is not being developed in isolation. A dozen psychedelic and psychedelic-adjacent programmes are in Phase 2 or Phase 3 as of August 2026, and LSD is the extreme case among them on almost every axis: longest duration, broadest receptor engagement, largest historical evidence base, heaviest cultural baggage, and — so far — the biggest reported anxiety effect size.
4-PO-DMT
same molecule
CYB003 / HLP003
lysergide D-tartrate
βk-MDMA
via luvesilocin
mebufotenin benzoate
mebufotenin
N,N-dimethyltryptamine
CYB004
midomafetamine
single enantiomer
How to read this table. Two structural facts stand out. First, almost nobody owns their molecule — psilocybin, LSD, DMT, 5-MeO-DMT, MDMA and 4-OH-DiPT are all unpatentable as compositions, so every programme competes on formulation, route, deuteration or prodrug chemistry. Second, the field has converged on shortening the session: intranasal, inhaled, buccal, subcutaneous and deuterated routes all exist to cut monitored clinic hours, which is the binding constraint on this entire class. MDMA is the cautionary entry — furthest along, and stalled on trial-conduct and blinding grounds rather than on biology.
LSD's competitive position is unusual. In GAD its only near-term rival is 4-OH-DiPT, which is a Phase behind and has no human anxiety data at all — so if Voyage and Panorama hold, MM120 could reach the indication essentially unopposed. The exposure is operational rather than clinical: every other programme in this table is engineered to shorten the session, and LSD is the one asset that structurally cannot. Its duration is a receptor residence-time property, not a formulation choice. The strategic question is therefore whether a 7.7-point HAM-A advantage justifies a twelve-hour chair.
Established, probable, speculative
Grading is ours. A — multiple independent sources, consistent. B — solid primary evidence, limited replication. C — single source, sponsor-derived, or contested.
Established · A
- 5-HT2A agonism is necessary for the subjective effects — demonstrated by ketanserin blockade in humans.9
- Duration is driven by slow receptor dissociation, structurally explained.10
- Synthesised 1938, psychoactivity found 1943, marketed as Delysid from 1947, withdrawn mid-1960s, Schedule I 1970.1,3
- Dependence liability is low; no established human lethal dose.24
- It is the least receptor-selective of the classic psychedelics.8
Probable · B
- A single 100 µg dose reduces GAD symptoms with effects persisting to Week 12.18,19
- A single dose reduced alcohol misuse in the 1966–70 randomised trials, with attenuating benefit.14
- It increases global connectivity and reduces default-mode network integrity.11
- It promotes structural neuroplasticity.12
- Placebo-controlled microdosing produces little of the benefit claimed for it.26
Speculative · C
- That the Phase 2b effect size survives Phase 3. Unknown pending Voyage and Panorama.16
- That the 50 µg arm adequately solves functional unblinding.21
- The entropic-brain / REBUS account as an explanation rather than a description.11
- HPPD incidence and risk factors in a supervised clinical population.25
- Any therapeutic role for repeated low-dose regimens, given the 5-HT2B question.24
Research gaps, in order of consequence
A Cohen's d of 0.81 from a single dose in GAD would be field-defining if it holds. Everything else about LSD's medical future is subordinate to whether Voyage and Panorama reproduce it.
The 50 µg arm is a good instrument but not a measurement. Nobody has published a validated method for partitioning drug effect from expectancy in a frankly psychoactive agent, and the field needs one.
If plasticity is the mechanism and 5-HT2A is the route, non-hallucinogenic 5-HT2A agonists should work — the premise behind RE245 and a generation of “non-psychedelic psychedelics.” No human data yet settle it, and LSD is the natural comparator when they arrive.
A DSM diagnosis with no reliable incidence estimate, no known risk factors and no treatment. It is the most under-studied adverse outcome in the entire psychedelic field.
5-HT2A mediates the experience, but LSD's therapeutic effect may draw on 5-HT1A, dopaminergic or biased-signalling contributions. Ketanserin studies address the subjective state, not the outcome weeks later.
Approval of a Schedule I substance requires rescheduling, and LSD carries more political weight than any other compound in development. No public roadmap exists for how that would be handled.
Sources
P1 peer-reviewed primary research or official instrument · P2 review, book or historical scholarship · P3 sponsor communication or trade press · P4 general reference, non-load-bearing.