This page respects the original work's rights terms and independently introduces its bibliography, abstract, main results, and limitations. It is not a full translation.
About this resource
This page introduces a 2023 umbrella review that surveyed benefits and risks involving cannabis, cannabis-derived constituents, and synthetic cannabinoids without limiting the scope to one condition or product. Rather than reaggregating individual clinical trials directly, an umbrella review examines already published meta-analyses to assess the breadth, strength, and consistency of evidence. The authors included 50 meta-analyses of observational studies and 51 meta-analyses of randomized controlled trials, evaluating 579 associations.
The most important point is that the single word “cannabis” encompasses everyday or recreational use, products for medical purposes, products high in THC, predominantly CBD preparations, synthetic cannabinoids, and other distinct exposures. This paper does not judge all of them uniformly safe or effective. It needs to be read as a map showing that conclusions change according to population, constituent, comparison condition, and study design.
Study design
The authors searched multiple databases and extracted meta-analyses addressing exposure to cannabis or cannabinoids and health outcomes. Observational evidence covered associations involving mental health, cognition, traffic accidents, pregnancy and birth, cancer, respiratory outcomes, and other areas. Randomized trials addressed pain, multiple sclerosis, epilepsy, nausea and vomiting, sleep, adverse events, and related outcomes. Evidence credibility was graded using not only statistical significance but also effect size, heterogeneity, bias, and small-study effects.
This design enables comparison across a broad field, but it cannot exceed the quality of the underlying meta-analyses. Overlap from the same studies appearing in several meta-analyses, different exposure definitions, unclear product composition, and differences among research periods remain. The number 101 therefore indicates the breadth of evidence; it does not mean there were 101 independent, high-quality trials.
Content
The review presents benefits and harms in parallel. Mean differences favoring cannabinoid interventions were reported in certain symptom areas, while associations with disadvantages were organized across the central nervous system, mental health, cognition, traffic safety, and other domains. Much of the benefit evidence comes from short-term trials using controlled preparations, whereas much of the harm evidence comes from observational studies following real-world exposure. Ignoring this distinction could lead either to extending benefits from pharmaceutical trials to all commercial cannabis or to treating every observational association as established causation.
Age, pregnancy, psychosis risk, and frequency of use were treated as important contexts shaping outcomes. The authors particularly call for cautious preventive judgment regarding young people, use during pregnancy, frequent use, and people at risk of psychotic disorders. This is population-level risk stratification, not a criterion for diagnosing an individual.
Results
Aggregated randomized trials provided evidence that particular cannabinoid preparations were associated with improvement in some chronic pain, spasticity related to multiple sclerosis, chemotherapy-related nausea and vomiting, and specific epileptic seizures. Effect magnitude, preparation, and follow-up duration differed across areas, and adverse events such as somnolence, dizziness, and gastrointestinal symptoms need to be assessed at the same time.
Observational evidence reported associations with adverse outcomes involving psychotic disorders, traffic accidents, cognition and education, pregnancy, and birth. Evidence strength was not uniform, and some associations could not exclude confounding or reverse causation. The paper’s conclusion is neither “benefits make it safe” nor “harms make every application ineffective.” It is that preparations and populations must be specified and expected benefits and harms compared individually.
Limitations
An umbrella review is broad but does not provide a direct answer for an individual patient. In observational studies, self-report, co-used substances, socioeconomic factors, and pre-existing health status influence results. In clinical trials, participant selection, short follow-up, and pharmaceutical-grade preparations differ from real-world use. Assessment methods vary among meta-analyses, and conclusions may disagree even for the same outcome.
Recent high-THC products, concentrates, and edible products may not be adequately represented in older research. A broad category such as “history of cannabis use” does not capture THC-to-CBD ratio, frequency, age at initiation, or route of administration. The strength of a statistical association must also be distinguished from its clinical importance.
Safety
This review does not provide grounds for self-directed use or treatment changes. Pregnancy, young age, driving, a history of psychotic symptoms, and combinations with other sedating substances or medications require safety assessment beyond average research results. Clinical-trial findings for purified CBD do not guarantee the quality or safety of commercial CBD products.
It is also important not to place acute changes in perception, attention, and motor function on the same timeline as long-term health associations. This page’s rank reflects editorial comprehensiveness as a research resource, not a ranking of products or treatments.
Source and rights
Original source: Solmi M, et al. Balancing risks and benefits of cannabis use: umbrella review of meta-analyses of randomised controlled trials and observational studies. BMJ. 2023;382:e072348. DOI: 10.1136/bmj-2022-072348.
The original article is published under a Creative Commons license that includes a noncommercial condition. This page is a detailed introduction independently reconstructing bibliographic information, methods, and principal issues; it is not a full translation or republication of figures and tables. The original source takes precedence for numerical values and evidence classifications.