Table 1: System used to grade studies
Level of evidence	Description of criteria*
Individual studies
LI	Concurrent controls used (random allocation should be used wherever possible). Experimental methods proven valid (at a minimum, appropriate positive and negative controls must be described). Objective primary measure(s) or objective interpretation of measure(s). Demonstrates relevant difference between groups (not limited to statistical significance). Study conducted in humans or in an experimental model representative of human biological process(es).
LII	Concurrent controls used. All previous criteria present but major limitation(s) in the application of model to humans; small sample size or difference(s) among groups are not significant.
LIII	Control group not concurrent Major limitation in the primary outcome measure(s) or Experimental methods not reproducible or valid.
LIV	Controls not used.
Inferences from all studies
Ci	Inferences based on at least two LI studies from independent investigators. No LI studies demonstrating conflicting results.
Cii	Inferences based on one LI study with no conflicting LI or LII evidence.
Ciii	Inferences based on at least one LII study with no conflicting LII evidence.
Civ	Inferences based on theory, LIII or LIV studies or conflicting LI or LII evidence.
*Criteria are designed to grade studies evaluating biologic processes, not to evaluate outcomes from therapy. Studies evaluating outcome measures similar to clinical outcomes in the laboratory setting (e.g., a survival study in an animal model) should use rigorous experimental designs comparable to human studies (i.e., a double-blind randomized controlled clinical trial is optimal).

[Return to text]