The GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) Working Group has developed criteria for grading the evidence for a clinical guideline.
Parameters:
(1) initial level of evidence based on the study design
(2) factors decreasing the quality of the evidence
(3) factors enhancing the quality of the evidence
Study Design |
Initial Evidence |
randomized trial |
high |
observational study |
low |
any other evidence |
very low |
Factors decreasing quality:
(1) limitations in study quality
(2) inconsistency
(3) uncertainty about directness
(4) data
(5) reporting bias
Decreasing Factor |
Finding |
Points |
limitations on study quality |
none |
0 |
|
mild to moderate |
0 |
|
serious |
-1 |
|
very serious |
-2 |
inconsistency |
none |
0 |
|
not important |
0 |
|
important |
-1 |
uncertainty about directness |
none |
0 |
|
some |
-1 |
|
major |
-2 |
data |
adequate |
0 |
|
imprecise |
-1 |
|
sparse |
-1 |
reporting bias |
low probability |
0 |
|
moderate probability |
0 |
|
high probability |
-1 |
decrease subscore =
= SUM(points for all 5 decreasing factors)
Factors enhancing quality:
(1) evidence of association
(2) evidence of a dose response gradient
(3) effect of all plausible confounders
Enhancing Factor |
Finding |
Points |
evidence of association |
not strong |
0 |
|
strong |
+1 |
|
very strong |
+2 |
dose response gradient |
no evidence for |
0 |
|
evidence for |
+1 |
effect of all plausible confounders |
would not reduce effect |
0 |
|
would reduce the effect |
+1 |
where:
• Very strong evidence of association has a relative risk > 5 based on direct evidence with no major threats to validity.
• Strong evidence of association has a relative risk > 2, with consistent evidence from 2 or more observational studies with no plausible confounders
enhancement subscore =
= SUM(points for all 3 enhancing factors)
total grade of evidence =
= (initial grade) + SUM(negative subscore) + SUM(positive subscore)
Final Grade |
Confidence in the Estimate of Effect |
high |
unlikely to change with further research |
moderate |
likely to change with further research |
low |
very likely to change with further research |
very low |
none to very little |
where:
• Based on the effects of the decreasing and enhancing factors there could be an argument for a very high and very, very low grades of quality.
ICD-10: ,