Table 2 Criteria for recommendation grading

Ottawa Panel [11]

Evidence from one or more randomized controlled trials of a statistically significant, clinically important benefit (>15%)

Statistically significant, clinically important benefit (>15%) if the evidence is from observational studies or controlled clinical trials

Clinical importance (>15%) but no statistical significance

No clinically important difference and no statistical significance

Evidence from one or more randomized controlled trials of a statistically significant benefit favouring the control group

Canadian Consensus Conference [8] and European League Against Rheumatism [6]

Meta-analysis of randomized controlled trial or at least one randomized controlled trial

At least one controlled study without randomization or at least one quasi-experimental study

Descriptive studies, such as comparative, correlation or case–control studies

Expert committee reports or opinions and/or clinical experience of respected authorities

American Academy of Orthopaedic Surgeons [10]

Meta-analysis of multiple, well-designed controlled studies; or high-power randomized, controlled clinical trial; or consistent findings from multiple well-designed experimental studies; or low-power randomized, controlled clinical trials; or nonexperimental studies such as nonrandomized, controlled single-group, pre–post, cohort, time, or matched case–control series; or nonexperimental studies, such as comparative and correlational descriptive and case studies

Generally consistent findings from well-designed experimental studies; or low-power randomized, controlled clinical trials; or nonexperimental studies such as nonrandomized, controlled single-group, pre–post, cohort, time, or matched case–control series; or nonexperimental studies, such as comparative and correlational descriptive and case studies

Inconsistent findings from well-designed experimental studies; or low-power randomized, controlled clinical trials; or nonexperimental studies such as nonrandomized, controlled single-group, pre–post, cohort, time, or matched case–control series; or nonexperimental studies, such as comparative and correlational descriptive and case studies

Little or no systematic empirical evidence

Institute for Clinical Systems Improvement [9]

Strong design study results that are clinically important and consistent. The results are free of any significant doubts about generalizability, bias, and flaws in research design. Studies with negative results have sufficiently large samples to have adequate statistical power

Strong design study results that are inconsistent or with minor doubts about generalizability, bias, flaws in research design, or adequacy of sample size. Alternatively, evidence consists solely of consistent results from weaker designs

Strong design study results that are substantially inconsistent or with serious doubts about generalizability, bias, flaws in research design, or adequacy of sample size. Alternatively, evidence consists solely of limited results from weaker designs