Criteria for the types of evidence and grades of recommendations have been proposed by Cook, Guyatt, Sackett et al. These can be helpful in evaluating clinical studies and therapeutic regimens.
Measures:
(1) levels of evidence
(2) grades of recommendations
Type of Evidence |
Level |
meta-analysis of multiple, well-designed randomized controlled studies; randomized trials are high power (low false-positive and low false-negative errors) |
I |
>= 1 well-designed randomized controlled study(ies); randomized trials are low power (high false-positive and/or high false-negative errors) |
II |
nonrandomized, controlled single group |
III |
single group pre-post comparison |
III |
cohort study |
III |
time series |
III |
matched case-control study |
III |
well-designed comparative and correlational descriptions |
IV |
well-designed case series |
IV |
case reports |
V |
clinical examples |
V |
where:
• I am a bit unsure of certain assignments in Level I and II. (a) Would a meta-analysis with low power studies be Level II?; (b) Would one large, well-designed study with high power be classified as Level I?
Recommendations |
Grade |
evidence of Level I |
A |
consistent findings from multiple studies of Levels II, III or IV |
A |
generally consistent findings from multiple studies of Levels II, III or IV |
B |
inconsistent findings from multiple studies of Levels II, III or IV |
C |
little or no systematic empirical evidence, or Level V only |
D |
where:
• It appears as if a single Level I study can trump all lower level studies.
• I am not quite sure how markedly discrepant findings (beyond simply inconsistent) would be handled.
ICD-10: ,