How to use GRADE (Grading of Recommendations Assessment, Development, and Evaluation)
How to Use the GRADE System
Grading of Recommendations Assessment, Development, and Evaluation (GRADE) is a transparent framework for rating the quality of evidence and strength of recommendations in systematic reviews and clinical guidelines. Below is a step-by-step guide.
1. Understand Key Concepts
- Quality of Evidence (Certainty):
- High (⊕⊕⊕⊕): Further research is unlikely to change confidence.
- Moderate (⊕⊕⊕◯): Further research may have an impact.
- Low (⊕⊕◯◯): Further research is likely to change confidence.
- Very Low (⊕◯◯◯): Evidence is very uncertain.
- Strength of Recommendation:
- Strong (“We recommend…”): Benefits clearly outweigh risks.
- Weak/Conditional (“We suggest…”): Trade-offs are uncertain.
2. Step-by-Step GRADE Process
Step 1: Start with a Study Design Baseline
- RCTs start as High quality.
- Observational studies (e.g., cohort, case-control) start as Low quality.
Step 2: Assess Reasons to Upgrade or Downgrade
Factors That Can Downgrade Evidence
| Factor | Explanation | Example |
|---|---|---|
| Risk of Bias | Serious flaws in study design (e.g., lack of blinding, high dropout). | Downgrade by 1-2 levels if most studies have high RoB (Cochrane RoB 2). |
| Inconsistency | Unexplained heterogeneity (e.g., I² > 50%). | Downgrade if results vary widely across studies. |
| Indirectness | Population, intervention, or outcome differs from research question. | Downgrade if study uses surrogate outcomes instead of patient-important ones. |
| Imprecision | Wide confidence intervals or small sample size. | Downgrade if 95% CI crosses “no effect” (RR=1). |
| Publication Bias | Suspected missing studies (e.g., funnel plot asymmetry). | Downgrade if small-study effects are likely. |
Factors That Can Upgrade Evidence (for Observational Studies)
| Factor | Explanation | Example |
|---|---|---|
| Large Effect | Strong association (e.g., RR > 2 or < 0.5). | Upgrade if treatment reduces mortality by 50%. |
| Dose-Response | Evidence of a gradient (e.g., higher dose → better outcome). | Upgrade if higher drug doses show better efficacy. |
| Plausible Confounding | All residual confounding would reduce effect. | Upgrade if smoking cessation reduces lung cancer despite confounding. |
Step 3: Determine Final Quality of Evidence
- Apply downgrades/upgrades to reach a final rating (High → Moderate → Low → Very Low).
- Example:
- Baseline: RCT (High)
- Downgrades: Risk of bias (–1), Imprecision (–1)
- Final: Moderate (⊕⊕⊕◯)
Step 4: Formulate Recommendations
- Strong recommendation: High-quality evidence with clear benefits.
- Example: “We recommend statins for CVD prevention (⊕⊕⊕⊕).”
- Weak recommendation: Lower-quality evidence or uncertain trade-offs.
- Example: “We suggest omega-3 supplements for mild depression (⊕⊕◯◯).”
3. Practical Example
Scenario: Assessing a meta-analysis of RCTs on aspirin for stroke prevention.
| Factor | Judgment | Adjustment |
|---|---|---|
| Study Design | RCTs | Start at High (⊕⊕⊕⊕) |
| Risk of Bias | Some studies had unclear blinding | Downgrade 1 (–⊕) → Moderate |
| Inconsistency | I² = 60% (substantial heterogeneity) | Downgrade 1 (–⊕) → Low |
| Indirectness | All studies used similar populations | No change |
| Imprecision | 95% CI for benefit: 0.7–1.1 (crosses 1) | Downgrade 1 (–⊕) → Very Low |
| Publication Bias | Funnel plot symmetric | No change |
| Final Rating | Very Low (⊕◯◯◯) | “Evidence is highly uncertain.” |
Recommendation:
- Weak: “We suggest considering aspirin for stroke prevention in high-risk patients (⊕◯◯◯), but shared decision-making is needed due to uncertain evidence.”
4. When to Use GRADE
✔ Systematic reviews (e.g., Cochrane reviews).
✔ Clinical practice guidelines (e.g., WHO, AHA).
✔ Health technology assessments (HTA).
5. Comparison with Other Tools
| Tool | Purpose | Output |
|---|---|---|
| GRADE | Rates evidence certainty & recommendations | High/Moderate/Low/Very Low |
| Cochrane RoB 2 | Assesses RCT bias | Low/Some/High risk |
| SIGN | Rates study quality (++/+/–) | For guideline development |
Final Tips
- Use GRADEpro (gradepro.org) for creating summary tables.
- Clearly document downgrading/upgrading decisions.
- Involve multiple reviewers to reduce bias.
Would you like a GRADE evidence profile template for your review?
