Inhaled Cannabis Tied to Near-Total Opioid Cessation in Back Pain

Patients with severe lower back pain that no longer responded to opioids, anti-inflammatories, or physiotherapy almost entirely stopped using opioids after starting inhaled medical cannabis, according to a five-year study of 241 Israeli patients published in the peer-reviewed journal Biomedicines. The share of patients taking opioids fell from 100% at the start to 4.6% by year five — a result the authors call clinically meaningful but pointedly decline to call proof that cannabis caused it.

That distinction is the heart of the paper. The improvements are substantial, but the study was designed in a way that cannot fully disentangle a genuine drug effect from several other explanations, and the researchers are unusually candid about it.

What the five-year cohort showed

The patients were treated at Hasharon Hospital, part of Rabin Medical Center near Tel Aviv, between 2020 and 2025. The group was deliberately uniform: every patient had documented failure of at least a year of opioids, NSAIDs, and physiotherapy, with severe pain still present despite that regimen. Mean pain duration was 15.1 years. They were then started on inhaled cannabis — roughly 91% smoked it — with THC content ranging from 4% to 22% and CBD from 2% to 22%.

Across five years, the four pain and disability measures the team tracked all improved sharply. Average pain on a 0-to-10 scale fell from 8.1 to 2.7. At year five, 89% of patients had cut their pain by at least 30%, and 77% by at least half. The authors described the improvements as “large, sustained, and statistically robust.”

The medication changes were the most striking part:

• Opioids: 100% of patients at baseline to 4.6% at year five
• Anti-inflammatories (NSAIDs): 100% to 7.1%
• Antidepressants: 80.5% to 5.4%
• Gabapentin-type nerve-pain drugs: 38.6% to 2.5%

Side effects were almost entirely mild — dry eye, short-term mental fog, and digestive complaints — with no moderate or severe cannabis-related events recorded across 1,205 patient-years of treatment, and no deaths. Pain relief also held steady even as average monthly doses climbed over the years, which the authors read as a sign that patients were not building tolerance — a question short trials cannot answer. The result echoes earlier work tying medical cannabis to lower opioid use in chronic pain patients.

Why the study design limits what it proves

This was not a randomized controlled trial. It was a single-center, single-prescriber, observational study in which each patient served as their own “before” comparison — there was no separate group that stayed on conventional drugs to measure against. The authors write plainly that the results “cannot be interpreted causally in the absence of a concurrent randomized control arm,” and that the improvements likely reflect a mix of a real pharmacological effect, regression to the mean, and the expectancy and self-selection that come with an open-label therapy patients choose and adjust themselves. Because everyone entered at near-maximal pain, some improvement was almost guaranteed regardless of treatment.

The opioid figure is the sturdier of the two kinds of findings, the authors argue, because it is binary. Whether a patient stopped filling opioid prescriptions is a yes-or-no fact, harder to skew than a self-reported pain score that can move on expectation alone. Even so, they flag confounders: a single prescriber aware of the cannabis who may have actively encouraged tapering, and the broad push across health systems to cut long-term opioid use over the same years. The paper also reports a number-needed-to-treat near 1 for opioid cessation, but warns that the figure is measured against each patient’s own prior state, not a control group, and should not be read as the equivalent number from a randomized trial. The same caution applies to a recent Minnesota program study, where cannabis appeared to improve daily functioning more than pain itself.

How does it square with the trial evidence

The effect sizes here dwarf what controlled trials report. A 2026 systematic review in Annals of Internal Medicine pooled 25 placebo-controlled trials and found cannabinoids reduced chronic pain only slightly — roughly half a point to under a point on a 0-to-10 scale. The first phase 3 trial of a cannabis drug for back pain specifically, published in Nature Medicine in 2025, found a full-spectrum extract beat placebo by a modest margin, about 1.9 points versus 1.4. That same extract was approved for chronic back pain in Germany and Austria this year.

The Israeli cohort’s roughly five-point drop is far larger. As the authors acknowledge, that gap almost certainly reflects who was enrolled — severely affected, self-selected patients with the most room to improve — rather than a uniquely powerful drug. It is a recurring pattern in this field: observational before-and-after cohorts produce bigger numbers than randomized trials, and the realistic estimate for any individual patient sits closer to the trial figures. Population data have separately tied cannabis access to fewer opioid overdose deaths, but those studies face the same causal limits.

The authors call for the obvious next step: a randomized trial pitting inhaled cannabis against continued conventional therapy in patients who have run out of standard options. They also note their cohort was treated under Israel’s medical-cannabis system, which differs enough from US, European, and UK frameworks that the results should not be assumed to carry over. Until that trial exists, the paper frames inhaled cannabis as a reasonable option to weigh for patients who have exhausted conventional care — not a first-line treatment, and not, on this evidence, a proven cause of the improvements it documented.

References:

^{1. Robinson, Dror; Khatib, Muhammad; Lavon, Eitan; Kafri, Niv; Abu Rashed, Waseem; Murad, Hamza; and Yassin, Mustafa, “Long-Term Inhaled Cannabis Therapy for Chronic Low Back Pain: A Five-Year Retrospective Analysis of Prospectively Collected Patient-Reported Outcomes in 241 Treatment-Refractory Patients” (2026). Biomedicines, 14(6), 1255. https://doi.org/10.3390/biomedicines14061255}