GRADE Assessment
The Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) framework rates the certainty of evidence for each outcome in your systematic review.
Overview
GRADE classifies evidence certainty into four levels:
| Level | Meaning |
|---|---|
| ⊕⊕⊕⊕ High | Very confident the true effect is close to the estimate |
| ⊕⊕⊕◯ Moderate | Moderately confident; true effect is likely close |
| ⊕⊕◯◯ Low | Limited confidence; true effect may differ substantially |
| ⊕◯◯◯ Very Low | Very little confidence in the estimate |
Rating Domains
Start with the study design baseline (RCTs = High, observational = Low) and assess five downgrade domains:
1. Risk of Bias
Assess methodological limitations: randomization, blinding, attrition, selective reporting. Use the integrated Risk of Bias 2 (RoB 2) tool or ROBINS-I for non-randomized studies.
2. Inconsistency
Evaluate heterogeneity across studies. Consider I², direction of effects, and overlap of confidence intervals.
3. Indirectness
Assess whether the evidence directly addresses your PICO question — population, intervention, comparator, and outcome match.
4. Imprecision
Evaluate the width of confidence intervals. Consider whether the CI crosses clinically important thresholds and the optimal information size (OIS).
5. Publication Bias
Use funnel plot analysis, Egger's test, and consideration of study registrations to assess.
Upgrade Factors
Observational studies may be upgraded for:
- Large effect — RR >2 or <0.5 with no plausible confounders.
- Dose-response gradient — clear relationship between exposure level and outcome.
- Residual confounding — all plausible confounders would reduce the effect.
Summary of Findings Table
AI4Meta generates a GRADE Summary of Findings (SoF) table for each comparison:
- One row per outcome with effect estimate, CI, number of participants, and study count.
- GRADE certainty rating with footnotes explaining each downgrade/upgrade.
- Absolute effect estimates with assumed baseline risk.
- Export to Word or copy to clipboard.
Using GRADE in AI4Meta
- Navigate to the GRADE tab in your project.
- Select the outcomes to assess.
- For each domain, select the severity of concern (no serious, serious, very serious).
- Add footnotes explaining your rationale.
- The overall certainty is calculated automatically.
- Generate the SoF table from the completed assessments.