GRADE method for rating the quality of evidence
| Study design | Quality of evidence | Lower if | Higher if |
| Randomised trial | High | Risk of bias | Large effect |
| −1 Serious | +1 Serious | ||
| −2 Very serious | +2 Very serious | ||
| Inconsistency | Dose response | ||
| −1 Serious | +1 Evidence of a gradient | ||
| −2 Very serious | All plausible confounding | ||
| Indirectness | +1 Would reduce a demonstrated effect or | ||
| −1 Serious | +1 Would suggest a spurious effect when results show no effect | ||
| −2 Very serious | |||
| Imprecision | |||
| −1 Serious | |||
| −2 Very serious | |||
| Quasi randomised trial | Moderate | Publication bias | |
| Low | −1 Serious | ||
| Very low | −2 Very serious |