The GRADE system
- Concept
GRADE rates how sure we are of an effect (high to very low) and keeps that certainty apart from the strength of a recommendation.
What it is
GRADE is a system for rating the certainty of evidence for each outcome and for making recommendations. Evidence from randomised trials starts at high certainty and observational evidence at low; it is then rated down for risk of bias, inconsistency, indirectness, imprecision or publication bias, and rated up (mainly observational evidence) for a large effect, a dose-response gradient or plausible confounding that would work against the effect. A recommendation can be strong or conditional, depending on the balance of benefits and harms, certainty, values and resources.
Worked example
- Data (made up): 4 trials with 2 400 patients; pooled RR for hospital admission 0.80 (95 % CI 0.62 to 1.03).
- Starting point: randomised trials, high certainty.
- Risk of bias: 2 of the 4 trials, carrying 55 % of the weight, did not conceal allocation: rate down one level, to moderate.
- Imprecision: the CI runs from a 38 % reduction (1 − 0.62) to a 3 % increase (1.03 − 1) and crosses 1: rate down again, to low.
- Inconsistency, indirectness and publication bias: no serious problems; final certainty low.
- Comparison: a cohort with RR 0.20 for another outcome starts at low and is rated up one level for a large effect, to moderate.
- Reading: the effect on hospital admission has low certainty; the panel issues a conditional recommendation.
What lowers and what raises certainty in GRADE
| Lowers certainty | Raises certainty |
|---|---|
| Risk of bias | Large effect |
| Inconsistency between studies | Dose-response gradient |
| Indirectness (other population, other comparison) | Plausible confounding that would work against the effect |
| Imprecision (wide CI, few events) | — |
| Publication bias | — |
Markers
- GRADE
- Certainty
- Strong
- Conditional
In practice
- Clinical practice guidelines that use GRADE show the strength and the certainty of the evidence beside each recommendation.
- A conditional recommendation invites a decision with the patient: different people, with their values, would choose differently.
Common pitfalls
- Mixing up certainty of evidence and strength of recommendation: they are two separate judgements.
- Rating a study instead of an outcome: GRADE rates the whole body of evidence for each outcome.
Test yourself
Think of the answer, then tap the question to see it.
At which certainty level does GRADE start evidence from randomised trials, and observational evidence?
Trials start at high and observational studies at low.
What are the five domains that lower certainty in GRADE?
Risk of bias, inconsistency, indirectness, imprecision and publication bias.
Which factors can raise the certainty of observational evidence?
A large effect, a dose-response gradient and plausible confounding that would work against the observed effect.
Which kinds of recommendation does GRADE distinguish?
Strong and conditional (weak), for or against an intervention.
Connections
Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.
Sources: Chapter 14: Completing ‘Summary of findings’ tables and grading the certainty of the evidence (Cochrane, 2026); WHO handbook for guideline development, 2nd Edition (World Health Organization, 2014).
Spotted a mistake?We are always improving for the community. Tell us what is wrong and our team will review it.Report a mistake
Representative model, for educational purposes only.