How is the content on mechoulam produced?
mechoulam is an information service that evaluates published scientific primary literature on cannabis as medicine and presents it as structured summaries. The register is maintained in German and rendered in English on these pages. The content serves information and education only. It replaces neither medical advice nor diagnosis nor treatment.
The evaluation follows a fixed procedure in several stages. The sections below describe it from the source to the statement a person has cleared.
Which studies are included?
The basis is published scientific primary literature: original studies and controlled clinical trials, systematic reviews and meta-analyses. Systematic reviews and meta-analyses, which pool many single studies, carry more weight than one small study on its own. Sources without peer review, personal accounts and manufacturer information do not enter the register.
The collection currently holds more than 1,100 evaluated studies. For the evaluation the texts are split into many thousands of single passages and indexed by meaning. Each passage is assigned the DOI of its original paper, the globally unique identifier of a scientific publication.
How are the matching study passages found for a question?
For every medical question a selection of relevant text passages is established first. The question and its medical terminology are converted into semantic vectors, a numerical representation of their meaning, and compared against the whole corpus. This meaning-based search also captures equivalent wording, so a question phrased in German finds matching passages in English-language studies. The passages found are then re-ranked by how closely they fit the question. Only the closest reach the next step.
How is the strength of a study rated?
The load-bearing study characteristics are read out of the relevant text passages: the population under study, the intervention, the comparison and the endpoints, the PICO scheme of evidence-based medicine. On that basis the strength of the evidence is rated against established scientific frameworks, among them GRADE and the Oxford levels of evidence. Every entry also receives a category from S to D, which shows at a glance how strong the evidence behind a statement is, from high-quality controlled trials down to preliminary signals.
What is each single statement tied to?
Every statement in an evidence report leads back to a concrete text passage of a cited study. Through the assigned DOI each passage is linked to its original paper. Every statement can therefore be traced back to the passage it rests on. Unlike a freely answering chat application, each statement is tied to a concrete, cited source instead of being freely composed.
When does a statement count as supported by studies?
The evidence check holds the formulated statement against the text passage assigned to it. We tell three levels apart:
- A statement counts as supported by studies when the text passage backs it directly and verifiably.
- It counts as partly supported by studies when the passage backs the direction of the statement but not every detail, such as the exact figure.
- It counts as not supported by studies when the passage fails to back the statement or contradicts it.
Where a statement rests on a study whose text is closed to our inspection, no such comparison is possible. Statements of that kind carry the mark not verifiable.
Who clears an entry before publication?
Before publication every entry passes a confidence check and is reviewed by a person. Entries with high confidence go live after that review. Entries with medium confidence stay internal drafts until someone has checked them. Entries with low confidence never enter the register.
What can this evaluation not deliver?
What this evaluation cannot deliver. The content prepared here replaces neither medical advice nor a treatment recommendation. Where a statement rests on a study whose text is closed to our inspection, we cannot cross-check it. Statements of that kind carry the explicit mark not verifiable. The evidence shown reflects the state of published research, including every gap that exists within it. Treatment decisions always belong in a conversation with a doctor.
Where is the use of artificial intelligence disclosed?
These evidence reports and study ratings are produced from published studies with the support of artificial intelligence and reviewed by a person. The legal disclosure sits in the terms of use, which remain available in German only: Nutzungsbedingungen.