• Intelligence · Clinical pipeline
The cannabinoid pipeline, measured.
• The problem
There is no register of cannabinoid trials. So we built one.
• What the data shows
Record activity, and a pipeline still thin where it counts.
Cannabinoid research did not peak and fade. Starts climbed from 55 in 2016 to a record 177 in 2025 and have held above 150 a year since 2021. But look at the gold band: Phase III has never once exceeded 27 starts in a year. The field is getting bigger, not further along.
Basis: registered start year. 543 of 2,323 records carry no start date and are excluded from this chart; 2026 counts every start registered for the year, including dates still ahead. Cannamonitor Clinical Trials database, 13 registry sources, n = 2,323. Snapshot 24 Aug 2026.
Phase II is the structural engine at 636 trials. Genuinely pivotal work — Phase III and IV together — is 368, or 16% of everything on record.
Preclinical (2) is folded into the ring but too small to render. 'Other' covers trials carrying no conventional phase label.
The largest single block is not a therapy. Substance-use research (620) plus healthy-volunteer and mechanistic work (279) accounts for 39% of all tags — research about cannabis, not medicine made from it.
Condition groups are multi-tag: a trial can carry more than one, so tags exceed the trial count. Nine of 15 groups shown.
Academia runs the pipeline; industry runs the end of it. Universities fund 1,548 trials but only 185 late-phase ones. Industry funds 442 — and 36% of all late-phase work.
Funder type is multi-tag. A trial with no recorded funder is excluded from both bars.
Yale has sponsored 71 cannabinoid trials and not one of them is late-phase. Jazz has sponsored 66 and 44 of them are. That single contrast is the whole structural problem in one row.
Sponsor as recorded on the registry entry; a trust or hospital sponsoring on behalf of a company is counted where the registry files it.
The charge that cannabinoid evidence is anecdotal does not survive the phase data. 74% of Phase III trials are placebo-controlled, against 68% at Phase II and 42% at Phase I. The closer to approval, the tighter the control.
Share of each phase's trials. These figures reproduce the June 2026 read (75 / 68 / 42%) on a base that has since almost doubled — the strongest evidence that the enlarged base is tagged consistently.
Synthetic cannabinoids have collapsed from 19% of starts to 6% while isolates nearly doubled to 35%. Whole-plant preparations held roughly half the field throughout. The field moved away from designed molecules, not toward them.
Share of trials with a registered start date in each era, by compound class (99.8% field coverage). The THC:CBD ratio field covers only 49% of the base and is not used here.
The country that runs the most trials runs the least pivotal work. 8% of US trials are Phase III, against 43% in the UK and 17% in Germany. Europe carries 186 Phase III site-tags to North America's 93 on a third of the volume.
Phase III as a share of each country's trial tags. Multinational studies are tagged in every country they run, so tags (2,187) exceed trials. 1,975 of 2,323 records carry a country.
CBD is the most-studied cannabinoid of all (1231 trials), ahead of THC (823). Behind them a minor-cannabinoid tail is now real but still small: 90 trials across CBN, CBG, THCV, CBDV, delta-8-THC, CBDA and CBC combined.
Multi-tag: a trial testing a THC:CBD product carries both. 593 trials record no specific cannabinoid.
Of 726 completed trials, only 217 have posted results — 30%. Seven in ten finished studies have never returned a public read, and only 269 records across the whole base carry a publication or DOI. The evidence gap in this field is as much a reporting failure as a funding one.
Completed trials only. 'Not recorded' means the registry entry carries no results field, not that results are absent.
The average cannabinoid Phase III enrols 175 patients. A mainstream pivotal trial runs several times that. Across all 1,388 Phase I–IV trials on record, total enrolment is 113,524 people — the whole field, ever.
Mean recorded sample size per phase, from the 1,265 phased trials that record one. Observational studies are excluded: a single registry of 200,000 people would swamp every other bar.
• Free — what your email unlocks
The live pipeline tracker. Three databases, no paywall on the view.
• The products
Eighty-seven products in development. Twelve at the final hurdle.
The product pipeline has the same shape as the trial data, one level up: 58 of 114 assets stall at Phase II. Seventeen have reached Phase III, and only 7 have ever been approved anywhere.
Furthest stage reached, including programmes since discontinued.
CBD isolates lead on count, at 38 products. But look at the gold bars: the plant-derived classes convert to a late-stage asset at close to one in five, while only 2 of the 20 synthetic small molecules have got there — and one of those two is discontinued. Designing a new cannabinoid receptor drug has been the hardest road in the field.
Composition class as recorded on the product. Five products carry no class.
Five Phase III programmes in the tracker are recorded as discontinued: Zygel (ZYN002) (Zynerba / Harmony, Fragile X syndrome), Lenabasum (JBT-101) (Corbus, Systemic sclerosis), QIXLEEF (PPP001) (Tetra Bio-Pharma, Advanced cancer pain), plus MRX2 and INSYS’s cannabidiol oral solution, both in paediatric epilepsy. A pipeline view that quietly drops its failures tells a rosier story than the field has had.
• The shelf
Seven approved products in forty years. Four of them are the same molecule.
| Product | First approved | Class | Indication |
|---|---|---|---|
| Marinol (dronabinol) | 1985 | THC isolate | CINV; AIDS-related anorexia |
| Cesamet (nabilone) | 1985 | Synthetic small molecule | Chemotherapy-induced nausea |
| Sativex (nabiximols) | 2005 | THC:CBD combination | MS spasticity |
| Syndros (dronabinol) | 2016 | THC isolate | CINV; AIDS-related anorexia |
| Epidiolex / Epidyolex | 2018 | CBD isolate | Refractory epilepsies |
| Dronabinol (generic) | — | THC isolate | CINV; AIDS-related anorexia |
| Exilby (VER-01) | 2026 | Full-spectrum botanical | Chronic low-back pain |
First approval in any market, from the Investigational Products tracker. Generic dronabinol carries no single authorisation date and is shown without one rather than guessed.
• The companies
Eighty-three companies. Six marketing authorisations.
47 of 83 companies have never got a product past Phase II. Thirteen sit at Phase III and six hold an approval — the whole industry's registration record fits in one row.
Furthest stage reached across all of a company's assets.
Nearly six in ten companies chose a single molecule or a synthetic — the route with clean composition-of-matter IP. Only 25 went after the whole-plant botanical, which is the route that has just produced the field's newest approval.
Strategic approach across 82 companies with a route recorded.
The six with a marketing authorisation
• How we build it
Why these numbers hold up.
Sources: ClinicalTrials.gov, EU CTR (EudraCT), CTIS, ANZCTR, ISRCTN, DRKS, NTR, CTRI, REBEC, IRCT, UMIN, ChiCTR and company disclosures. Cannamonitor Clinical Trials database, 13 registry sources, n = 2,323. Snapshot 24 Aug 2026.
The data behind this page
This page is a summary. Decide is the source.
• Objections
Common questions.
2,323 on our record as of 24 Aug 2026, merged from 13 registry sources. That is higher than most published counts because most published counts are a ClinicalTrials.gov search — which holds 1,549 of these and misses 774 that exist only in ANZCTR, DRKS, CTRI, ISRCTN, REBEC, IRCT, NTR, UMIN, ChiCTR or the EU registers.
Sometimes, and we would rather say so. A multinational study registered in both ClinicalTrials.gov and a national register can appear as two records — NanaBis is a live example, present under both an NCT and an ACTRN identifier. No cross-registry dedup rule is applied today. It affects Phase III counts and country tags by a small, unmeasured margin, and it is the first thing on the list to fix.
No. Nothing here is derived from a paper. Every figure is a query against a tagged database of individual trial, product and company records, and the query behind each one is recorded next to it. Re-running this page next quarter is a re-run, not a re-derivation.
Your work email opens the live pipeline tracker: the investigational-products database (114 assets), the pharmaceutical-companies database (83 developers) and the full clinical-trial analysis. No card, no trial window. The full clinical-trials database itself, the full company database and the science and studies library sit inside Decide.
Continuously. Records are added and revised as registries publish; the page is stamped with the snapshot it was built from (24 Aug 2026) and the tracker is always ahead of the page. The last record added at this snapshot landed 4 Aug 2026.
Any trial of a cannabinoid or cannabis-derived intervention in humans — THC, CBD, the minor cannabinoids (90 trials across CBN, CBG, THCV, CBDV, delta-8-THC, CBDA and CBC), whole-plant and broad-spectrum extracts, synthetic analogues such as dronabinol and nabilone, nabiximols, and endocannabinoid-system modulators (79 trials). Observational and mechanistic studies are included and labelled, not silently mixed into the therapeutic count.
• Go deeper
Got the pipeline. Here's what going deeper buys.
You have the tracker. Essentials adds the running commentary on it — a news stream, a monthly recap across 30+ markets, and insight deep-dives on the readouts and approvals that actually move a programme.
The tracker gives you the analysis. Decide gives you the record — queryable, filterable, joinable:
- The full clinical-trials database — all 2,323 records, every field, down to the registry entry
- The full company database — beyond the pharma developers, every cannabis company we track, with financials and news
- Science and studies — the published literature behind the trials
- Country and market data — where a programme has a market to launch into