A Drug Made From Marijuana Is Not Marijuana

An 820-person trial found a cannabis extract reduced back pain. Here is exactly what was in it — and why it is not what you can buy.

By Cannible Editorial

In September 2025 a large randomized trial reported that a cannabis-derived medicine reduced chronic low back pain. The finding is real. It also does not transfer to anything you can buy in a dispensary, and the reason is not a matter of interpretation, it is a matter of what was in the bottle.

Each dose in that trial contained 2.5 mg of THC, in an extract manufactured to pharmaceutical standards and standardized to a fixed cannabinoid content, swallowed by dosing syringe on a written three-week schedule of increases. That specification is the finding. Change the composition, the dose, the route or the schedule and you no longer have the thing that was studied, you have a different substance that shares an origin with it.

This matters because a headline of the form "a drug made from marijuana reduced back pain" is accurate and almost useless, and by the time it reaches a person deciding what to spend money on it has usually become "weed treats back pain." That conversion is the subject of this article. The trial is the worked example.

The headline was accurate, and it still misled

The wire story that carried this finding to a general audience opened by reporting that "an experimental medication made from marijuana successfully reduced back pain in a new study, offering further support for the drug's potential in treating one of the most common forms of chronic pain". Every clause of that is true.

The same story went further than most coverage did. It named the manufacturer, noted that the formula contains THC, and told readers that the levels were "very low, essentially a microdose compared to what's available in gummies, chocolate bars and other products sold at marijuana dispensaries in the U.S." That comparison is the wire service's own, the trial paper makes no reference to dispensary or recreational products anywhere, but the point it was reaching for is the right one, and it was in the coverage from the start.

So this is not a story about bad reporting, and Cannible is not correcting anyone. It is a story about what survives a trip. What survives a headline is the shape of the sentence: a thing made from cannabis did something good for pain. What does not survive is every number that made the sentence mean anything. The failure is structural, the accurate version and the misleading version are the same eight words, which is why the fix is a reading habit rather than a correction.

What was actually in the bottle

This section describes a clinical trial's investigational product. It is education, not medical advice, and nothing in it is a protocol, a target or a recommendation.

The drug is called VER-01. The trial paper defines it as a standardized full-spectrum extract of a patented cannabis strain, full-spectrum meaning the extract retains the plant's other cannabinoids and compounds rather than isolating one, and standardized meaning each batch is manufactured to hit the same cannabinoid content rather than whatever the plant happened to produce. The paper's own description is precise: the extract "is manufactured in a Good Manufacturing Practice (GMP) certified facility and standardized to 5% tetrahydrocannabinol (THC). Each dose unit (119 µl) of the finished product VER-01 contains 50 µl of the full-spectrum extract, delivering 2.5 mg THC, 0.1 mg cannabigerol and 0.02 mg cannabidiol, with sesame oil as excipient."

Read that as what it is: a specification. Three cannabinoids, each at a stated weight, in a stated volume of a stated oil, from a manufacturing process audited to a defined standard. The placebo was built to match it, sesame oil, with cannabis aroma and colorants, so that neither the participants nor the investigators could tell which was which.

The dose, and the way it was given

Participants took VER-01 orally, measured out with a dosing syringe, and they did not start at a full dose. The trial ran a three-week titration, a gradual increase to find a working amount, in which, in the paper's words, "the dose was increased every 3 days by one dose unit in the morning and one dose unit in the evening until the participant either experienced sufficient symptom relief or reached the maximum daily dose of 13 dose units."

Other painkillers were not permitted alongside it. Rescue medication was restricted and specified in the protocol. Every participant was under trial supervision throughout.

Who was allowed into the trial

A trial result describes the people who were in the trial, and this one was filtered. Participants were excluded, the paper states, "if they had other painful comorbidities potentially interfering with their low back pain rating, a severe mental illness (for example, psychosis, schizophrenia and bipolar disorder), or a history of alcohol, drug or medication abuse. Moreover, participants who had used cannabis-based medicinal products within 30 days before screening were excluded."

Those exclusions are ordinary trial design, and they are also the point. Several of the groups removed by that filter, people living with a serious psychiatric condition, people with a history of problem substance use, people already using medical cannabis, are among the most likely to read a cannabis pain headline and see themselves in it. The trial says nothing about any of them. Not that the drug would fail for them, and not that it would harm them: it was not measured in them, and an unmeasured group is not a reassured group.

What the trial measured, and who paid for it

The trial was phase 3, the large, late-stage kind that regulators weigh, multicenter, randomized and placebo-controlled, with 820 adults split between VER-01 (394) and placebo (426). Its first phase ran 12 weeks with neither side knowing who got what, followed by a six-month open-label phase and, later, a randomized withdrawal phase.

The headline result: the trial "met its primary endpoint in phase A, with a mean pain reduction of −1.9 NRS points in the VER-01 group (mean difference (MD) versus placebo = −0.6, 95% confidence interval (CI) = −0.9 to −0.3; P < 0.001)." The NRS is an 11-point numeric rating scale, the "rate your pain from 0 to 10" question. So the drug group improved by about two points; the difference attributable to the drug rather than to placebo was six tenths of a point.

Three further numbers belong beside it. The proportion of participants whose pain fell by at least 30% was "significantly higher for VER-01 compared to placebo (54.1% versus 39.5%), resulting in a number needed to treat to benefit (NNTB) of 6.8", meaning roughly seven people would need the drug for one additional person to hit that threshold. Adverse events were substantially more common on the drug: "83.3% versus 67.3%; P < 0.001." And the withdrawal phase, which tested whether stopping the drug brought pain back, did not work out: "phase D did not meet its primary endpoint (hazard ratio = 0.75, 95% CI = 0.44–1.27; P = 0.288)."

Whether a six-tenths-of-a-point difference matters to a person living with back pain is a real question, and this article is not going to answer it. The trial's authors argue that it does, on the strength of the responder rates and the NNTB. That argument deserves to be reported and it deserves its context: the trial "was sponsored by Vertanical, which contributed to the study design and provided support with medical writing," the company that makes VER-01, and the paper's authors declare consultancy and institutional research relationships with that company. Cannible located no independent evaluation of this trial, no editorial, no guideline assessment, no outside commentary, so the only argument on the record for the importance of that number comes from people with an interest in it. That is worth knowing, and it is not the same as saying the argument is wrong.

One more thing the comparison does not include: the drug was tested against placebo, not against any other painkiller. A result versus placebo tells you the drug beats nothing. It does not tell you how it compares with what the reader is already taking.

Why the drug and the shelf are different objects

What "an approved drug" establishes

Everything above describes a product moving through a regulatory pathway, and that pathway is the second half of the distinction.

In the United States, as the Food and Drug Administration stated on its cannabis regulation page, content current as of 16 July 2024, the date stamped on the page as read, the agency "has not approved a marketing application for cannabis for the treatment of any disease or condition." It has approved "one cannabis-derived and three cannabis-related drug products," and "these approved products are only available with a prescription from a licensed healthcare provider." What approval means, in the agency's words, is that it "has concluded that this particular drug product is safe and effective for its intended use."

Two cautions on that paragraph. First, it is a US-only statement and says nothing about other jurisdictions. Second, and more important: federal cannabis regulation has been in motion, that page carried a 2024 date when it was read, and this article did not verify the position as of its own publication. Check the agency's current page before relying on it. A dated regulatory fact is worth more than an undated one, and less than a current one.

What the distinction gives you is a category. An approved drug is a specific formulation, at specific doses, for a specific condition, that a regulator has assessed. A plant product on a dispensary shelf has not been through that assessment for any condition, which is a statement about what has been evaluated, not a claim that it does or does not work.

The trial's own authors draw the same line

The strongest version of this article's argument is not Cannible's. It is a sentence in the paper itself, written by the researchers about their own result: "Consequently, findings obtained with one cannabis extract cannot be extrapolated to others without appropriate comparative data."

Read the reach of that. The authors are not saying their finding fails to transfer to a dispensary gummy. They are saying it does not transfer to another cannabis extract, a different pharmaceutical-grade preparation, made under the same kind of controls, without a head-to-head study to bridge them. If the boundary is that tight between two laboratory extracts, the distance to an unstandardized retail product is not a matter of degree.

Five questions to run on the next headline

The distinction only becomes useful once you have watched it run. Here it is as a test, with this trial's own answers.

What exactly was given? Not "cannabis", a named, standardized full-spectrum extract with three cannabinoids at stated weights, in sesame oil, manufactured to GMP standard. If the coverage cannot tell you what the substance was, it has not told you what was studied.

How much, by what route, on what schedule? 2.5 mg of THC per dose unit, taken orally by syringe, escalated every three days over three weeks up to a stated daily maximum, with other painkillers withheld.

To whom? 820 adults with chronic low back pain, with people excluded for confounding pain conditions, severe mental illness, a history of alcohol or drug misuse, and recent use of cannabis-based medicines.

Against what? Placebo, a matched sesame-oil preparation with cannabis aroma and colorants. Not another treatment.

For how long, and what happened after? Twelve weeks double-blind, then a six-month open-label phase; and in the later randomized withdrawal phase, the trial missed its endpoint.

A headline that answers none of those five is not wrong. It is just not yet information you can do anything with.

What this explains, and what it doesn't

The model this article gives you is a boundary, and boundaries have their own edges.

It tells you that a pharmaceutical made from cannabis and a product sold on a shelf are different objects, and it tells you which specific differences did the work in this case. It does not tell you that dispensary products do not help anyone's back pain. That is a separate question, and this article carries no evidence on it in either direction, including no sourced description of what a typical dispensary product actually contains, which is itself a gap worth noticing when you are being invited to compare.

It does not settle whether the trial's effect size is clinically meaningful. It reports the number, reports the authors' argument for it, and records that no independent assessment was found.

And it does not generalize. A second phase 3 trial by the same group compared VER-01 with opioids rather than placebo, on a primary endpoint of gastrointestinal tolerability rather than pain, and the paper reports that "VER-01 was superior to opioids in terms of pain reduction over 6 months of treatment, although differences in secondary endpoints limited to week 27 alone were not significant." That is a different trial with a different design, a different population and the same commercial sponsor, and Cannible read only its published abstract, its numbers do not combine with the ones above, and nothing here should be read as an appraisal of its methods.

What the boundary does give you is durable. The next cannabis headline will arrive with a study behind it, and the study will be about something specific. The five questions are how you find out what.

Sources

View on Cannible →