The GRADE system then prescribes a quality evidence rating system for each outcome reported across all relevant studies. The final rating ascribed to an outcome is labelled as high, moderate, low or very low. These ratings are the endpoints of a series of upgrading or downgrading steps that result from the presence or absence of certain quality items. For example, for randomized controlled trials (RCTs), the initial quality of evidence rating for a given outcome would start as high and then be modified upwards or downwards according to these quality items. Outcomes from an observational study would start as low, but could then be upgraded if the studies that report this outcome were of high quality and the magnitude of the treatment effect was large.
| Study design | Quality of Evidence | Downgrade if* | Upgrade if |
| Randomized trial → | High
|
Risk of bias
-1 Serious -2 Very serious
Inconsistency -1 Serious -2 Very serious
Indirectness -1 Serious -2 Very serious
Imprecision -1 Serious -2 Very serious
Publication bias -1 Likely -2 Very likely |
Large effect
+1 Large +2 Very large
Dose response +1 Evidence of a gradient
All plausible confounding +1 Would reduce a demonstrated effect or
+1 Would suggest a spurious effect when results show no effect |
| Moderate
|
|||
| Observational study → | Low
|
||
| Very low |
Case series and case reports are essentially observational studies without control groups, and thus downgrading to very low evidence is appropriate in most cases.
After evaluating the evidence for all outcome/PICO questions, the guideline authors will then synthesize the findings by determining the direction and the strength of a recommendation.