How Do We Rate the Level of Evidence?

Updated: October 25, 2026

In short: For every topic we give two things: how reliable the evidence is (the level of evidence) and what the research shows (the result). These are two different things. For example, in fibromyalgia the evidence is "moderate" and the result is "No effect found": in other words, we are fairly sure that CBD does not work for this condition. Our main grade is for pure CBD; we grade products that contain THC separately.

What's on this page?8 sections

Level of evidence: five grades

To assign a grade, we build on the GRADE approach, which is widely used in medicine (Guyatt 2008). This approach looks not only at the type of study but also at the risk of bias, whether the results are consistent with each other, how precise the result is, whether the study directly addresses the question being asked, and the chance that negative results have gone unpublished.

Grade	What does it mean?	Which topics today?
Strong evidence	There are several well-designed, double-blind, placebo-controlled studies and reviews that combine them; the results are consistent. A new study is not expected to change the conclusion.	Only Dravet and Lennox–Gastaut syndromes, with the approved CBD medicine.
Moderate evidence	There is at least one large, well-designed controlled study or several controlled studies, but the studies are few or there are small inconsistencies in the results. The conclusion is probably correct.	Tuberous sclerosis complex (Proven effect); CBD taken by mouth in fibromyalgia, chronic pain, and hand or knee osteoarthritis (No effect found).
Weak evidence	Controlled studies are small, short or few; results are inconsistent or have not been independently replicated. New studies could easily change the conclusion.	For example, generalized anxiety, social anxiety, insomnia, autism, Parkinson’s, neuropathic pain, and pain treated with CBD applied to the skin.
Very weak evidence	There are only very small studies or studies at high risk of bias, or uncontrolled case series. We can’t be sure of the conclusion.	For example, depression, depression in bipolar disorder, skin conditions.
Insufficient evidence	There are no, or almost no, controlled studies looking at this question in humans. All we have are cell and animal studies or observations.	For example, nausea, appetite loss and "tumor-shrinking" claims in cancer; migraine, diabetes, panic disorder; glaucoma.

The current grade for each topic is shown in the CBD Guide, at the top of each topic page.

Direction of the result: the short labels you see on our pages

Alongside the level of evidence, we sum up what the research shows with a single label:

Proven effect: Benefit was shown in controlled studies. The level of evidence next to it tells you how strong the evidence is: for example, the evidence is strong for the approved epilepsy uses of the licensed CBD medicine.
May help a little: A small effect was seen, but this does not count as proof of treatment.
Mixed results: Some studies saw a benefit; others did not. CBD has not been proven to work for this.
Not studied enough: There are not enough studies in humans; it is not known whether CBD works.
No effect found: In studies, CBD did no better than an inactive substance (placebo).
May be harmful: Studies suggest that CBD may cause harm in this condition.

"No effect found" and "Not studied enough" are not the same thing. In the first, CBD was tried and did not work; in the second, it has not yet been tried enough.

Why do we separate pure CBD from products that contain THC?

Some cannabis-derived medicines also contain THC: for example, nabiximols (a mouth spray containing THC and CBD), used for muscle stiffness caused by MS, or THC capsules tested for chemotherapy nausea. A result seen with these medicines cannot be assumed to apply to a CBD oil that contains no THC or very little.

Our main grade is for pure CBD.
Products that contain THC, where there are any, are graded separately and shown in a separate box.
Example: for muscle stiffness caused by MS, the evidence is weak for pure CBD (No effect found) and moderate for the THC-containing medicine. For chemotherapy nausea, the evidence for pure CBD is insufficient; for THC-containing medicines, there is only a weak guideline recommendation, and only for nausea that continues despite standard medicines.

We tag studies by product type: pure CBD, THC-containing, CBD-rich but with an unclear amount of THC, other cannabinoids, and mixed products.

Types of study: from most to least reliable

Systematic reviews and meta-analyses: These assess all suitable studies on a topic together. Their value depends on the quality of the studies they include.
Randomized controlled trial (RCT): Participants are assigned to groups by chance; one group gets CBD, the other a placebo. In a "double-blind" study, neither the participants nor the researchers know who is in which group.
Studies without a control group (open-label) and observations: Because there is no placebo group, there is no way to know how much of the improvement seen comes from CBD.
Case reports: One or a few patients. Especially valuable for spotting signals of side effects and drug interactions; they are not evidence of effectiveness.
Cell and animal studies: These shed light on why and how; they do not count as evidence of a treatment effect in humans.

Explanations of terms: Glossary (RCT, placebo, meta-analysis, open-label study).

Which sources do we use, and which don’t we use?

What we use:

Studies published in peer-reviewed journals indexed in PubMed, and Cochrane reviews.
Official bodies: the Official Gazette (Resmî Gazete), the Ministry of Health and the Turkish Medicines and Medical Devices Agency (TİTCK) in Turkey; the US Food and Drug Administration (FDA) and drug labels (DailyMed); the European Medicines Agency (EMA); the European Food Safety Authority (EFSA); the World Health Organization (WHO); the World Anti-Doping Agency (WADA).
Clinical guidelines from professional organizations (for example, oncology, pain and anesthesia societies).
ClinicalTrials.gov records. We flag separately studies whose results appear only in the registry and have not been published in a peer-reviewed journal.

What we don’t use:

Blogs, "research" pages and promotional texts of companies that sell products.
Commercial market reports and statistics of unclear origin.
Journals that publish for a fee without scientific review (predatory journals).
News without cited sources, social media posts and user reviews.

If a study was funded by a company that makes CBD, or its authors have ties to the industry, we say so.

How do we verify information?

We open every study from its PubMed record: the title, authors, year, journal, PubMed number (PMID) or DOI must match.
We take numbers and results from the study’s abstract or full text. If we could read only the abstract for a piece of information, we say so.
We do not publish information we could not find in its source or could not verify; in our working files we set it aside as "could not be verified."
We give the doses used in studies together with the name of the study. These are not personal dose recommendations; we do not recommend personal doses.
We use AI tools in our preparation; everything AI finds is checked against the source itself.

Updates and error reports

Every page shows the date it was last updated.
When there is an important new study, guideline or change in regulations, we update the relevant page; if the level of evidence changes, we say so on the page.
If you spot a mistake, write to us through the contact form. We check it against the source, correct it if it is right, and update the page’s date.

The limits of these grades

The level of evidence describes what research shows in general; it does not tell you what is right for you. If you have a medical condition or take medicines, make the decision with your doctor. The full picture of how we work: Editorial Policy.

Source: Guyatt GH et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ 2008;336:924–926. PMID 18436948.