The scale
5 levels · GRADE-orientedThe level does not follow from the study type alone. The type is the starting point, but methodological weaknesses pull the rating down. A meta-analysis built from many small and contradictory studies therefore lands at B despite its high-grade study type.
S Best evidence
A replicated, consistent effect across independent studies.
- Cochrane review
- Guideline
- Large double-blind RCT
N ≥ 100 · ≥ 2 independent studies
A Strong evidence
A clear quantified result with sufficient power.
- RCT
- Systematic review
- Meta-analysis (homogeneous)
N ≥ 30 · single replication
B Moderate evidence
An effect is visible, but internal or external validity is limited.
- RCT (small)
- Controlled trial
- Meta-analysis (heterogeneous)
N 15 – 100
C Weak evidence
Preliminary or contradictory findings, no randomisation.
- Cohort study
- Cross-sectional study
- Case series
N 5 – 50 · no control
D Lowest evidence
No controlled clinical proof.
- Case report
- Expert opinion
- Pilot study
N < 20 · no comparison
Oriented on GRADE (Grading of Recommendations Assessment, Development and Evaluation) and on the evidence levels of the Oxford Centre for Evidence-Based Medicine. Where the confidence measure is high enough, the study enters the public register; at medium confidence it stays behind as an internal draft until it has been reviewed; at low confidence it is not taken up. Every rating can be corrected or downgraded at any time. The evaluation is AI-assisted and is checked by a human before publication.
Up to three studies per indication are highlighted as key studies. Two things count at once: high methodological weight, for instance Cochrane reviews, large randomised trials or robust meta-analyses, and concrete findings of their own on exactly this indication. A renowned review that touches a topic only in passing therefore does not become a key study here. That way every indication shows the few papers that actually carry its evidence base.
The evidence spectrum
1169 studies · 52 indicationsThe distribution of all entries across the quality levels S to D. What is rated is quality, not study type.
The works
mechoulam is more than a list of studies. Behind it stands a system of seven steps: it finds studies, reads them, rates them, checks them and secures everything. Tap a station to open it.
Each of the 52 indications has its own search strategy, built from validated MeSH terms and boolean operators, not from a generic “cannabis + disease”.
A semantic ranking model then scores the hits by conceptual proximity to the indication: a paper in which pain is mentioned only in passing ranks lower than one in which it is the primary endpoint.
Known entries are filtered out by DOI comparison. The result is a candidate list in which every study has passed both a lexical and a semantic filter.
- MeSH terms
- 52 search strategies
- semantic ranking
- DOI comparison
For every candidate study the PICO frame is built:
• Population (who was studied)
• Intervention (which cannabinoid, dose, route of administration)
• Comparator (placebo, active control or none)
• Outcome (what was measured, in which unit, at which time point)
Study design, sample size N and the numerical main result are added. The PICO scheme enforces comparability regardless of how the authors phrased their abstract, and so makes gaps in the reporting visible.
- PICO extraction
- intervention & dose
- primary outcome
- study design
Every study is given an evidence level from S to D. The basis is the internationally established GRADE methodology together with the evidence levels of the Oxford Centre for Evidence-Based Medicine.
The study type is the starting point: a large randomised trial or a systematic review starts high, a case series or a single report starts low. From that starting point the rating deducts wherever the study is weak: high risk of bias, contradictory results, too small a sample or imprecise effect estimates.
In addition the system assigns a confidence measure, that is how certain it is of its own classification. This measure decides what happens to the study next (step 6).
- GRADE methodology
- Oxford evidence levels
- study type as starting point
- downgrade on weaknesses
- confidence measure
Before entering the public register every study has to pass several mandatory checks:
(1) The study type is unambiguously classifiable: randomised trial, controlled trial, cohort study, cross-sectional study, case series, systematic review or meta-analysis.
(2) For primary studies the sample size N is present as a number, for reviews the number of included studies k.
(3) The main result is quantified: a measured effect with a direction, no phrases such as “improvements were observed”.
(4) The entry contains no placeholders or meta comments from the evaluation step.
If a study does not pass these checks it is not taken into the public register but held back as a draft or discarded. Whether a study fits the indication at all was already ensured in step 1 by the semantic ranking.
- mandatory checks
- study type classifiable
- sample size as a number
- quantified result
- no empty phrases
The register runs under full version control: every change (entry, update, tier change, withdrawal) is logged with a timestamp, the source and the triggering process or reviewer.
DOI-based deduplication prevents the same study from appearing under several indications with contradictory ratings. Canonical format rules normalise author strings, journal names and cannabinoid nomenclature.
- version control
- DOI deduplication
- rollback
- ISO 4 normalisation
- max. 3 key studies
The system makes the first assessment, but the human keeps control over it. Studies the system is very certain about are taken straight into the register. Everything below that stays as an internal draft at first and appears publicly only after review.
Every entry can be looked at, corrected or downgraded at any time: if a study turns out to be weaker, if a well-founded critique appears, if the authors retract it or if a contradictory new review is published, the level can be lowered or the entry removed.
The register is designed to become not just larger over time, but more reliable.
- human keeps control
- high certainty → taken in directly
- otherwise draft until review
- correction at any time
- downgrade possible
01 searches The Scout
Each of the 52 indications has its own search strategy, built from validated MeSH terms and boolean operators, not from a generic “cannabis + disease”.
A semantic ranking model then scores the hits by conceptual proximity to the indication: a paper in which pain is mentioned only in passing ranks lower than one in which it is the primary endpoint.
Known entries are filtered out by DOI comparison. The result is a candidate list in which every study has passed both a lexical and a semantic filter.
- MeSH terms
- 52 search strategies
- semantic ranking
- DOI comparison
02 extracts The Reading
For every candidate study the PICO frame is built:
• Population (who was studied)
• Intervention (which cannabinoid, dose, route of administration)
• Comparator (placebo, active control or none)
• Outcome (what was measured, in which unit, at which time point)
Study design, sample size N and the numerical main result are added. The PICO scheme enforces comparability regardless of how the authors phrased their abstract, and so makes gaps in the reporting visible.
- PICO extraction
- intervention & dose
- primary outcome
- study design
03 rates The Grade
Every study is given an evidence level from S to D. The basis is the internationally established GRADE methodology together with the evidence levels of the Oxford Centre for Evidence-Based Medicine.
The study type is the starting point: a large randomised trial or a systematic review starts high, a case series or a single report starts low. From that starting point the rating deducts wherever the study is weak: high risk of bias, contradictory results, too small a sample or imprecise effect estimates.
In addition the system assigns a confidence measure, that is how certain it is of its own classification. This measure decides what happens to the study next (step 6).
- GRADE methodology
- Oxford evidence levels
- study type as starting point
- downgrade on weaknesses
- confidence measure
04 checks The Guardian
Before entering the public register every study has to pass several mandatory checks:
(1) The study type is unambiguously classifiable: randomised trial, controlled trial, cohort study, cross-sectional study, case series, systematic review or meta-analysis.
(2) For primary studies the sample size N is present as a number, for reviews the number of included studies k.
(3) The main result is quantified: a measured effect with a direction, no phrases such as “improvements were observed”.
(4) The entry contains no placeholders or meta comments from the evaluation step.
If a study does not pass these checks it is not taken into the public register but held back as a draft or discarded. Whether a study fits the indication at all was already ensured in step 1 by the semantic ranking.
- mandatory checks
- study type classifiable
- sample size as a number
- quantified result
- no empty phrases
05 secures The Foundation
The register runs under full version control: every change (entry, update, tier change, withdrawal) is logged with a timestamp, the source and the triggering process or reviewer.
DOI-based deduplication prevents the same study from appearing under several indications with contradictory ratings. Canonical format rules normalise author strings, journal names and cannabinoid nomenclature.
- version control
- DOI deduplication
- rollback
- ISO 4 normalisation
- max. 3 key studies
06 keeps control The Human
The system makes the first assessment, but the human keeps control over it. Studies the system is very certain about are taken straight into the register. Everything below that stays as an internal draft at first and appears publicly only after review.
Every entry can be looked at, corrected or downgraded at any time: if a study turns out to be weaker, if a well-founded critique appears, if the authors retract it or if a contradictory new review is published, the level can be lowered or the entry removed.
The register is designed to become not just larger over time, but more reliable.
- human keeps control
- high certainty → taken in directly
- otherwise draft until review
- correction at any time
- downgrade possible