Skip to main content
In a recent podcast with David Tang from CannPrescribe, How to build confidence prescribing medical cannabis, he raised a very important question as to why certain cannabis based medicines were available via the NHS, yet others weren’t. More to the point, why can the number of NHS prescriptions still be counted on only one hand, when Cannabis has been legal to prescribe in the UK since 2018?
The short answer is that legality was only the first step.
The NHS can prescribe cannabis-based medicines. Specialist clinicians can also prescribe unlicensed cannabis-based products where there is an exceptional clinical need and where, in their judgement, the likely benefits outweigh the risks. But most cannabis medicines used in the UK private sector remain unlicensed, and unlicensed CBPMs are not routinely commissioned by the NHS.
That distinction matters. The NHS is not simply asking whether a patient might benefit. It has to ask a much broader set of questions:
Does the medicine work? How certain are we? Is it safe? Is the product itself sufficiently characterised and consistent? How does it compare with existing treatment? And is it a reasonable use of finite NHS resources?
That is how medicine is supposed to work.
Legal does not mean routinely available
The 2018 rescheduling removed an important legal barrier. It did not give cannabis-based medicines automatic approval for NHS use.
Most CBPMs supplied privately are “specials”: unlicensed medicines provided to meet the clinical needs of an individual patient where an appropriate licensed medicine is not available. The MHRA makes clear that a licensed medicine has undergone regulatory assessment of its quality, safety and efficacy. An unlicensed medicine has not necessarily been through that process, and the prescriber therefore assumes greater responsibility.
For NHS clinicians, that matters enormously.
Prescribing an unlicensed medicine is not prohibited, but the clinical and medico-legal burden is higher. NHS England says decisions should normally follow established governance arrangements, often including multidisciplinary discussion, and recognises that many clinicians remain concerned about the limited long-term evidence for most products.
That is very different from saying doctors are “not allowed” to prescribe cannabis.
They are. The question is whether the evidence and governance support doing so routinely. Legal does not mean routinely available.
The 2018 rescheduling removed an important legal barrier. It did not give cannabis-based medicines automatic approval for NHS use.
Most CBPMs supplied privately are “specials”: unlicensed medicines provided to meet the clinical needs of an individual patient where an appropriate licensed medicine is not available. The MHRA makes clear that a licensed medicine has undergone regulatory assessment of its quality, safety and efficacy. An unlicensed medicine has not necessarily been through that process, and the prescriber therefore assumes greater responsibility.
For NHS clinicians, that matters enormously.
Prescribing an unlicensed medicine is not prohibited, but the clinical and medico-legal burden is higher. NHS England says decisions should normally follow established governance arrangements, often including multidisciplinary discussion, and recognises that many clinicians remain concerned about the limited long-term evidence for most products.
That is very different from saying doctors are “not allowed” to prescribe cannabis.
They are. The question is whether the evidence and governance support doing so routinely.
So why did Sativex make it through?
Sativex is a useful example because it shows that the system is capable of recommending a cannabis-based medicine when the necessary pieces come together.
Sativex is a licensed THC:CBD oromucosal spray. NICE recommends a four-week trial for adults with moderate to severe multiple-sclerosis spasticity when other pharmacological treatments have not worked, followed by continued treatment only if symptoms improve by at least 20%.
NICE did not conclude that the evidence was perfect. In fact, it described much of the evidence as low quality.
What mattered was that there was enough evidence of benefit, an identifiable patient population, a defined product and dose, and an economic model capable of assessing whether treatment represented reasonable value for the NHS. NICE concluded that, under the agreed commercial and responder arrangements, the longer-term benefits were likely to outweigh the harms and the treatment could be cost effective.
That is important. The NHS does not require absolute certainty.
It requires enough certainty to make a defensible decision.
And why did Epidyolex succeed?
Epidyolex provides an even clearer example.
It is a defined pharmaceutical preparation of cannabidiol with a marketing authorisation for specific severe epilepsies.
For Dravet syndrome, NICE considered randomised clinical trial evidence showing that cannabidiol with clobazam reduced convulsive and non-convulsive seizures compared with usual care. NICE ultimately concluded that, despite some uncertainty in the economic modelling, the treatment represented an appropriate use of NHS resources.
It reached a similar conclusion for Lennox–Gastaut syndrome and later for seizures associated with tuberous sclerosis complex.
Again, the lesson is not that CBD is somehow inherently more acceptable to the NHS than cannabis flower.
It is that Epidyolex arrived as a specific medicine for a specific indication, at a defined dose, supported by controlled trials and an economic case.
That is the language in which modern healthcare systems make decisions.
But surely there is already plenty of evidence for medical cannabis?
There is certainly a substantial and growing body of evidence.
There are observational studies, patient registries, case series, clinical experience and randomised trials across different cannabinoids and indications. “Evidence exists”, is not the same as “the evidence answers the question NICE needs answered”.
For treatment effects, NICE has a strong preference for high-quality randomised controlled trials because randomisation reduces the risk that differences between treated and untreated patients are caused by factors other than the medicine itself. NICE does consider observational and real-world evidence, and in some circumstances it can be highly valuable, but the evidential weight depends on the question being asked and the quality of the data.
For routine NHS adoption, the important questions include:
- What precisely is the intervention?
- Which patients should receive it?
- What is the appropriate comparator?
- What outcome should improve?
- By how much?
- For how long?
- What are the adverse effects?
- Does the effect persist?
- How consistent is the medicine used in the evidence?
- What does treatment cost compared with alternatives?
- What effect does it have on quality of life and wider NHS resource use?
NICE’s evaluations consider both clinical effectiveness and economic evidence, generally measuring health benefit through quality-adjusted life years and comparing costs with existing NHS care.
That is a much higher hurdle than demonstrating that a group of patients reported feeling better after treatment, as it should.
Cannabis creates an additional evidential problem
Cannabis is not one molecule. The products prescribed privately can differ in cannabinoid profile, cultivar, formulation, route of administration, manufacturing process and dose. That makes evidence generation more complicated…
A trial of a standardised cannabidiol solution tells us something very specific about that medicine. It cannot automatically be extrapolated to a THC-dominant flower, a different cultivar or another formulation.
This is where manufacturing consistency becomes part of the evidence conversation.If clinical outcomes are to be linked meaningfully to a medicine, the medicine itself has to be sufficiently well characterised and reproducible.
That is one reason why pharmaceutical quality and clinical evidence should not be treated as separate conversations.
Is real-world evidence being undervalued?
Possibly, but the position is more nuanced than sometimes suggested.
NICE already has a formal real-world evidence framework and explicitly says real-world data can help address evidence gaps, understand long-term outcomes, assess safety and determine how trial results translate into routine NHS practice.
What it does not say is that uncontrolled observational evidence should automatically replace randomised trials when trying to determine whether a treatment causes an improvement. That distinction is reasonable.
The more interesting question is whether the UK is making enough use of the extraordinary amount of real-world prescribing already taking place privately. At present, private prescribing represents most UK CBPM use, yet the national evidence infrastructure does not comprehensively capture that activity in a standardised way. That feels like a missed opportunity.
A properly governed dataset capturing indication, medicine, dose, product characteristics, adverse events, switching, persistence and validated patient outcomes could make real-world prescribing far more scientifically useful. It would not eliminate the need for trials. It could help us design better ones.
Does the system need to change?
Perhaps, but not by lowering the evidential standard. The principle behind the NHS approach is sound.
A publicly funded healthcare system should expect convincing evidence that a medicine is safe, clinically effective and represents appropriate value before routinely funding it for large patient populations. Cannabis should not be exempt from those principles simply because patients are enthusiastic about it.
There is a legitimate question about whether the pathway for generating that evidence is keeping pace with the way cannabis medicines are actually used.
NICE has already moved towards greater use of real-world evidence. The opportunity now may be to connect private prescribing, national registries, consistent product information and prospective research much more effectively.
We should also be realistic about commercial incentives.
Developing a licensed medicine requires substantial investment in product development, controlled clinical trials, regulatory submissions and health-economic evidence. That model works most naturally where a company can develop and protect a defined pharmaceutical product.
It is more difficult for medicines based on plant varieties that may already be widely available and cannot necessarily be protected commercially in the same way. That does not mean the evidence bar should disappear.
What this does mean is that policymakers should ask whether the current research infrastructure creates enough incentive to generate the evidence everyone says is missing.
The NHS is not the enemy here
It is easy to frame limited NHS prescribing as institutional resistance to cannabis. That is too simplistic.
The NHS has a responsibility to millions of patients and has to make decisions using consistent standards across thousands of medicines and treatments.
Sativex and Epidyolex show that cannabis-derived medicines can pass through that system. The more useful question is therefore not:
“Why won’t the NHS prescribe cannabis?”
It is:
“What evidence do we need to give NHS clinicians and NICE enough confidence to prescribe particular cannabis medicines for particular patients?”
That is a question the medical cannabis sector should be helping to answer. Legal access was the beginning, but for widespread NHS access, the next challenge is evidence.
Evidence requires more than belief that a medicine works. It requires a system capable of proving what works, for whom, at what dose, with what risks, and at what value to the patient and the NHS.
