Researchers searched PubMed and Embase from 1975 through 2025 for human studies evaluating cannabinoid exposure in relation to substance use disorder outcomes across opioid, alcohol, cocaine, tobacco, and methamphetamine use disorders. Two reviewers independently screened and extracted data from every included study.

Six endpoints were prespecified and mapped onto each target disorder: treatment retention, relapse, abstinence, craving, withdrawal severity, and consumption. Every study’s risk of bias was assessed with a design-matched tool: RoB 2 for randomized controlled trials, ROBINS-I for cohort studies, and JBI checklists for cross-sectional, case, and qualitative designs.

Rather than simply counting how many studies reported a benefit, the review applied a design-based weighting scheme that assigned randomized controlled trials the highest weight (1.00) and qualitative studies the lowest (0.25), following Synthesis Without Meta-analysis (SWiM) guidance. This tier-weighted approach is what lets the findings distinguish a real effect from a signal produced mostly by weaker study types.

Ninety-seven studies covering 41,954 participants contributed 195 separate endpoint instances to the final synthesis, pre-registered on PROSPERO before the review began.

Of the 195 endpoint instances, 89 (45.6%) were classified as Beneficial, 80 (41.0%) as No Significant Effect, 12 (6.2%) as Mixed or Partial, and 14 (7.2%) as Harmful or Inferior.

The large majority of Beneficial findings, 76.4%, were for short-term symptom targets such as craving, withdrawal, and consumption rather than for the sustained outcomes that define successful addiction treatment.

When the review broke down which study designs were driving the Beneficial findings for short-term symptoms, the pattern was stark: 81.5% of Beneficial craving findings, 85.7% of Beneficial withdrawal severity findings, and 80.0% of Beneficial consumption findings came from weaker study designs, not from randomized controlled trials.

This is exactly the pattern a tier-weighted review is built to catch. A raw vote-count of how many studies found a benefit would have overstated the strength of this evidence considerably.

For the outcomes that matter most in real-world addiction treatment, treatment retention, relapse, and abstinence, results were predominantly classified as No Significant Effect. This pattern was most pronounced in opioid use disorder, where the evidence base was largest and most rigorous.

The authors note that findings for opioid use disorder are the strongest in the review, meaning the most rigorously tested, not that cannabinoids were shown to work best there. Evidence for the other four disorders, alcohol, cocaine, tobacco, and methamphetamine use disorders, remains preliminary.

The study authors state directly: cannabinoids confer short-horizon symptomatic benefits but do not demonstrate efficacy for sustained abstinence, relapse prevention, or retention, most clearly in opioid use disorder, where evidence is strongest. Findings for other disorders remain preliminary.

They call for adequately powered adjunctive randomized trials with biochemically verified endpoints, rather than self-reported outcomes, as the necessary next step.

This review sits inside a longer-running debate in addiction medicine about whether cannabinoids belong in the treatment toolkit for substance use disorders. Interest has grown alongside broader cannabis legalization and a search for adjunctive options for opioid use disorder in particular, given the severity of the ongoing overdose crisis.

What this review adds to that debate is a comprehensive, quality-weighted look at the full evidence base rather than a review of a single drug, single disorder, or single study design. Its central contribution, that short-term symptom benefit and long-term outcome improvement are not the same evidence, and one does not currently support the other, is a distinction worth carrying into any future conversation about cannabinoids as an SUD treatment.

I have spent more than 20 years practicing evidence-based cannabis medicine, and this is the kind of review I want to see more of, not because it delivers a clean yes or no, but because it is honest about the difference between those two answers. A tier-weighted synthesis that actually shows its work, that the encouraging short-term findings on craving and withdrawal come disproportionately from weaker study designs, while the best-designed evidence on relapse, retention, and abstinence mostly shows nothing, is a more useful and more trustworthy result than a simple vote count would have produced.

If a patient in addiction treatment asks me whether cannabis or a cannabinoid product will help them stay off opioids, alcohol, or another substance long-term, this review tells me I cannot honestly say yes based on the current evidence. I can say that short-term symptom relief has some support, mostly from weaker studies, and that the rigorous trials we actually need, adequately powered, adjunctive, with biochemically verified outcomes rather than self-report, have not yet been run at the scale this question deserves. That is not a reason to dismiss cannabinoids in addiction care. It is a reason to be precise with patients about what we know and do not yet know, and to keep pushing for the trials that would resolve it.

A Four-Step Reading Frame