A landmark JAMA review of more than 2,500 studies finds that most popular uses of medical cannabis, including chronic pain, anxiety, and sleep, lack strong clinical evidence. Only three FDA-approved cannabinoid indications cleared the evidentiary bar. On 4/20, here is what an honest reading of the science actually looks like from a physician who has worked through it with more than 300,000 patients.

Patients arrive at the clinic on 4/20 and every other day of the year carrying information they found online, most of it optimistic, much of it uncalibrated against what rigorous evidence actually shows. When a team of researchers from five major academic institutions publishes a systematic analysis in JAMA, the flagship journal of American medicine, and concludes that the evidentiary bar has been cleared for only a narrow set of indications, physicians have a responsibility to take that seriously, communicate it accurately, and avoid both overclaiming and dismissing patient experience in the process. That conversation starts with understanding what this review actually measured and what it did not.

In December 2025, JAMA published a systematic review led by Dr. Michael Hsu at UCLA Health, with contributions from researchers at Harvard, UCSF, Washington University School of Medicine, and New York University. The team analyzed more than 2,500 peer-reviewed studies published between January 2010 and September 2025, including randomized clinical trials, meta-analyses, and clinical guidelines. Their central question was straightforward: for which medical conditions does cannabis or cannabinoid therapy have reliable clinical support? More than 120 studies were prioritized based on sample size, recency, and relevance across a broad range of health conditions. (Hsu M et al., JAMA 2025; DOI: 10.1001/jama.2025.19433)

The findings were unambiguous. Clear clinical benefit was identified only for FDA-approved pharmaceutical-grade cannabinoid products used in three specific contexts: nausea and vomiting related to chemotherapy, appetite loss in the setting of HIV/AIDS, and severe pediatric seizure disorders including Dravet syndrome and Lennox-Gastaut syndrome. For chronic pain, anxiety, insomnia, and the other conditions most frequently cited by medical cannabis patients, the review found that evidence remains limited, inconsistent, or insufficiently powered to draw reliable clinical conclusions. The authors called for more rigorous research and clearer guidance from clinicians.

The Gap Is Real. So Is the Experience. A Physician’s 4/20 Reckoning With What We Know and Don’t.

I have had some version of the same conversation hundreds of thousands of times. A patient sits across from me, describes years of pain or sleeplessness or anxiety, tells me that cannabis is the only thing that has worked, and then asks me to confirm what they already believe: that the science backs them up. My job in that moment is not to agree reflexively or to dismiss the experience they just described. My job is to be accurate. And the JAMA review from Hsu and colleagues gives me another opportunity to be precise about what accurate means in this field right now.

The review is thorough. It is a serious piece of work from serious institutions, and it asks a legitimate question about where the evidence stands after 15 years of accumulation. The answer it gives, that strong clinical evidence supports only three pharmaceutical-grade cannabinoid indications, is accurate within the boundaries of what the review was designed to measure. But those boundaries matter enormously, and I want to be specific about what they are.

The review evaluated evidence by the standards that govern pharmaceutical drug development. Rigorous randomized controlled trials, adequate sample sizes, validated outcome measures, reproducible results. Those are exactly the right standards to apply when assessing whether a therapy should be approved, reimbursed, or prescribed at a population level. They are the tools that allow medicine to distinguish a real signal from noise. The problem is not the standards. The problem is that the evidence base for most cannabis indications has been deliberately stunted for decades by the very regulatory structure that would otherwise require those standards to be met. You cannot fail a trial you were never permitted to run.

I want to be careful here, because this is exactly where some cannabis advocates go wrong. Saying that the research was blocked is not the same as saying that the evidence would have been positive if it had been allowed. We genuinely do not know that. What we know is that the evidence base is thin for reasons that have more to do with policy than biology, and that thinness is not the same as absence of effect. The Schedule I classification of cannabis did not tell us that cannabis lacks therapeutic value. It told us that we were not allowed to find out. That distinction is clinically meaningful when I am sitting with a patient who is asking what to expect.

Here is what I see in practice. Patients with chronic pain who have tried multiple pharmaceutical approaches, some with significant side effects and limited success, report that cannabis changes their daily function in ways those other therapies did not. I take that seriously. Not because anecdote is the same as evidence, but because patient-reported outcomes in real clinical practice are legitimate data points, and they consistently point in directions that the existing trial literature has not yet had the opportunity to confirm or refute at scale. There are studies, including work on neuropathic pain and sleep disturbance, that show promising signals even if they fall short of the evidentiary threshold a JAMA review is designed to privilege. Those signals do not prove efficacy. They do suggest that writing off cannabis for these conditions, based on the same limited evidence base that the review correctly identifies, would be a different kind of mistake.

The honest clinical position is neither “cannabis works for everything” nor “cannabis works only for what the FDA has approved.” The honest position is: for specific indications, including severe pediatric epilepsy and chemotherapy-induced nausea, the evidence is solid. For others, including chronic pain, anxiety, and insomnia, the evidence is insufficient to draw firm conclusions, and insufficient is not the same as negative. It means we do not yet have the trial infrastructure to answer the question cleanly. Patients using cannabis for those conditions may be responding to a real pharmacological effect, a placebo effect, an improvement in adjacent symptoms like sleep or stress that then improves pain perception, or some combination of all three. Sorting those mechanisms out requires the kind of research that this field has been systematically prevented from conducting.

There is also a product problem that this review, like most large-scale analyses, cannot fully address. The cannabis products that patients are using in the real world, with varying cannabinoid ratios, delivery methods, dosing protocols, and terpene profiles, are not the same as the standardized preparations used in clinical trials. A patient using a low-THC, high-CBD tincture at bedtime for sleep is doing something quite different from what most sleep studies have examined. The heterogeneity of the exposure makes it genuinely difficult to pool results, and that difficulty is not resolved by better statistical methods. It is resolved by better product standardization and more targeted trials. Both require regulatory changes that are still in process.

What this review does that is genuinely valuable is force a recalibration of language. When clinicians or advocates say that “cannabis helps with pain” or “cannabis treats anxiety,” they are describing a claim that is not currently supported by the evidence hierarchy that governs clinical medicine. They may be right. But saying so without qualification does patients a disservice, because it sets expectations that may not be met and creates a credibility problem when the nuanced picture emerges. The evidence, right now, is what it is. Our job is to say so clearly, counsel patients on what is known and unknown, support informed decision-making, and push hard for the regulatory reforms and funding commitments that would allow the research to finally catch up to the public health reality.

On 4/20, the day most associated with cannabis culture and advocacy, the most important thing a physician can offer is not celebration or caution. It is accuracy. The JAMA review is a data point in a much larger conversation, and it is a useful one. Not because it settles anything, but because it draws the map of what remains to be done.

This is not the first analysis to identify the gap between cannabis use prevalence and clinical evidence quality. A Lancet Psychiatry review published in early 2026 examined 45 years of randomized trials on cannabis and mental health and found no evidence of benefit for anxiety, PTSD, or depression, the three conditions most frequently cited by medical cannabis patients. A separate systematic review on neuropathic pain has shown more promising signals, but those findings have not been replicated with the consistency that would support a formal clinical recommendation. What the JAMA review contributes is scale and institutional credibility. Drawing on 2,500 studies and co-authored by researchers at five major academic medical centers, it represents a formal accounting of where the evidence stands, not a fringe critique of the cannabis field.

Clinically, the most important safety considerations remain what they have been: cannabis use in adolescents and young adults carries documented risks of cognitive and psychiatric effects that are not present at the same magnitude in adult populations; high-potency THC products carry greater risk of adverse psychological effects than lower-potency formulations; and patients with personal or family histories of psychosis should be counseled carefully regardless of indication. Drug interactions with CYP450-metabolized medications, including common anticoagulants and anticonvulsants, require active monitoring. For patients currently using cannabis and asking how to interpret this review, the practical recommendation is: continue the conversation with your physician, be specific about what you are using and what you are experiencing, and do not interpret this evidence gap as permission to extrapolate beyond what the science can currently support.

This is a systematic review, which sits near the top of the evidence hierarchy. A team of researchers identified and synthesized more than 2,500 published studies using predefined criteria. They prioritized studies by sample size, recency, and relevance. The strength of this approach is its breadth and the institutional rigor behind it. The limitation is inherent to the method: a systematic review can only assess the quality and conclusions of existing studies. It cannot compensate for the structural barriers that prevented rigorous trials from being conducted in the first place. The review does not tell us that cannabis is ineffective for the conditions where evidence is lacking. It tells us that the trials needed to confirm or refute efficacy have not been adequately run.

The JAMA review is consistent with the broader scientific consensus that pharmaceutical-grade cannabinoid products have a well-defined, narrow evidence base, while dispensary-grade cannabis products used for the full range of patient-reported indications have not been subjected to the same scrutiny. This pattern is not unique to cannabis. Many therapies in common clinical use preceded the large-scale trials that would eventually confirm or refine their use. What distinguishes cannabis is the scale of public adoption in the absence of that evidence, and the decades-long regulatory structure that prevented the research from advancing in parallel. A concurrent analysis by economists at UMass Amherst published in the International Journal of Industrial Organization (Wang and Chan, 2026) found that state-level legalization tripled cannabis patents but produced no measurable increase in clinical trial activity, a finding that reflects exactly this disconnect. The evidence base remains thin not because researchers have looked and found nothing, but because the research infrastructure has not been allowed to scale to match the question.

Yes. The review applied a stringent evidentiary threshold aligned with pharmaceutical approval standards. A less restrictive meta-analysis using observational data, patient-reported outcomes, or lower-quality trials would likely surface positive signals for pain and sleep. Whether those signals reflect true therapeutic effects or methodological artifacts is exactly what better-designed trials would help determine. The review’s conclusions are correct within its stated framework. They are not the final word on whether cannabis works for these conditions, because that question has not yet been adequately tested in the trials that would answer it definitively.

Misreading 1: “This review proves cannabis doesn’t work for pain or anxiety.” It does not. It identifies that rigorous clinical trial evidence supporting those uses is currently insufficient. Insufficient evidence is not the same as evidence of inefficacy. The distinction matters.

Misreading 2: “If it’s in JAMA, it settles the debate.” A systematic review settles what the available trial literature says. It cannot speak to the trial literature that has not been conducted. It is a highly credible snapshot, not a verdict.

Misreading 3: “Patients who say cannabis helps them are wrong.” Patient-reported outcomes are not the same as RCT evidence, but they are also not meaningless. The honest clinical response is to take them seriously, communicate the evidence limitations accurately, and use both to inform a shared decision with the patient in front of you.

The JAMA review gives medicine an accurate picture of where the evidence stands. Strong, reproducible clinical evidence exists for three pharmaceutical-grade cannabinoid indications. For most of the conditions driving medical cannabis use, including pain, anxiety, and sleep, the evidence is currently insufficient by the standards that govern clinical practice. This reflects decades of research suppression more than it reflects a biological signal that has been tested and found wanting. The path forward is regulatory reform, adequately funded trials, standardized products, and clinicians who can hold both the current evidence limitations and the patient experience in the same conversation without collapsing into either dismissal or overclaim.

Q: Does this mean my cannabis use is not evidence-based?
It depends on what you are using it for and which evidence standard you apply. For a small number of pharmaceutical-grade indications, the evidence is solid. For chronic pain, anxiety, sleep, and most other common uses, the clinical trial evidence is currently insufficient to make a definitive claim either way. That is not a reason to stop, but it is a reason to stay in conversation with your physician about what you are experiencing.

Q: Why is there so little research on something so widely used?
The primary reason is that cannabis has been classified as a Schedule I substance under federal law, a classification that restricted researcher access to study-grade cannabis, made FDA/DEA approval processes unusually burdensome, and limited funding from federal sources. Those barriers have been partially addressed by the Medical Marijuana and Cannabidiol Research Expansion Act of 2022, but the research infrastructure will take years to scale.

Q: Should I stop using cannabis because of this review?
That decision belongs to you and your physician, not to a review article. What this review should change is the conversation: patients and doctors should be specific about what is known, what is not, what the patient is actually experiencing, and what risks apply to their individual situation. That conversation is better than either uncritical enthusiasm or reflexive caution.

Have thoughts on this? Share it: