Table 3

NHLBI Quality Rating of the Strength of Evidence

Type of EvidenceQuality Rating
  • Well-designed, well-executed RCT that adequately represent populations to which the results are applied and directly assess effects on health outcomes.

  • Meta-analyses of such studies.

Highly certain about the estimate of effect. Further research is unlikely to change our confidence in the estimate of effect.
High
  • RCT with minor limitations affecting confidence in, or applicability of, the results.

  • Well-designed, well-executed nonrandomized controlled studies§ and well-designed, well-executed observational studies.

  • Meta-analyses of such studies.

Moderately certain about the estimate of effect. Further research may have an impact on our confidence in the estimate of effect and may change the estimate.
Moderate
  • RCT with major limitations.

  • Nonrandomized controlled studies and observational studies with major limitations affecting confidence in, or applicability of, the results.

  • Uncontrolled clinical observations without an appropriate comparison group (e.g., case series, case reports).

  • Physiological studies in humans.

  • Meta-analyses of such studies.

Low certainty about the estimate of effect. Further research is likely to have an impact on our confidence in the estimate of effect and is likely to change the estimate.
Low

NHLBI indicates National Heart, Lung, and Blood Institute; and RCT, randomized controlled trials.

  • In some cases, other evidence, such as large all-or-none case series (e.g., jumping from airplanes or tall structures), can represent high- or moderate-quality evidence. In such cases, the rationale for the evidence rating exception should be explained by the Work Group and clearly justified.

  • “Well-designed, well-executed” refers to studies that directly address the question; use adequate randomization, blinding, and allocation concealment; are adequately powered; use intention-to-treat analyses; and have high follow-up rates.

  • Limitations include concerns with the design and execution of a study that result in decreased confidence in the true estimate of the effect. Examples of such limitations include but are not limited to: inadequate randomization, lack of blinding of study participants or outcome assessors, inadequate power, outcomes of interest that are not prespecified for the primary outcomes, low follow-up rates, and findings based on subgroup analyses. Whether the limitations are considered minor or major is based on the number and severity of flaws in design or execution. Rules for determining whether the limitations are considered minor or major and how they will affect rating of the individual studies will be developed collaboratively with the methodology team.

  • § Nonrandomized controlled studies refer to intervention studies where assignment to intervention and comparison groups is not random (e.g., quasi-experimental study design).

  • Observational studies include prospective and retrospective cohort, case-control, and cross-sectional studies.