From theorem to source.
First we establish the question and the population. Then we verify the publication or official document, record the identifier, result and boundaries of the application. Each record tells you whether full text, abstract, or just metadata were checked.
For the effect of therapy, we prefer adequate RCTs and syntheses; for public policy - official document. For rare harms, observations and case reports are an important signal that an RCT may not detect.
The database is a curated review of sources, not a registered systematic review of all literature. No independent double screening or formal GRADE assessment was performed for each record. The RCT label names the study design, it does not promise high certainty.
First, comparability.
We check age, country, measurement method, end point, comparator, observation time and unit of analysis. Diopters and millimeters answer other questions. Eyes and patients are not the same. The online work and the publication in the volume may be of different years.
We do not pool two publications from the same sample as independent participants. Testing one device or glass design does not prove the effect of an entire class of products.
What we know. What we don't know. What the application can change.
In the application we show a limitation. The absence of events in a small sample is not evidence of safety without limitations. The absence of a declaration in the abstract indicates an undetermined conflict of interest - never automatically "no conflicts".
A new study of sufficient size, independent replication, longer follow-up, or a reliable signal of harm may change the assessment. The update receives an assertion ID and an entry in the change history.
Dictionary for reading evidence
- Projection
- The output of the model given the specified data and assumptions; not measuring the future.
- RCT
- Random group study; reduces some comparison errors, but does not remove all of them.
- Meta-analysis
- Quantitative pooling of research results. Its credibility depends on the quality and comparability of the input works.
- Consensus/guideline
- An agreed framework developed on the basis of evidence and expert judgment; not a single experiment.
- Observational study
- Follow-up of individuals without random assignment to intervention. The association may also reflect confounding factors.
- Cycloplegia
- Temporary pharmacological relaxation of accommodation; important for comparable measurement of refraction in children.
- Axial length
- Distance along the eye axis measured by biometry. Its changes help assess progression; are not the only decision criterion.
- OR
- Odds ratio. It does not mean that the absolute probability of the disease in a specific person is as much as possible.
- Rebound
- Acceleration of progression after completion of intervention; the extent depends on the study and method of withdrawal.
- Conflict of interest
- Financial or other relationship that may influence the rating. Disclosure allows it to be included; funding alone does not invalidate the outcome.
- Patient year
- One year of follow-up for one patient. 100 patient-years may represent 100 people for 1 year or some other sum of follow-up time.
Same content for man and machine.
The JSON registry associates persistent claim identifiers with source records. Static structured data reflects the content of pages. DOI identifies the publication; a link to a related project does not imply the identity of the entity.
Open the JSON assertion map