Applying Class of Recommendations and Level of Evidence to Clinical Strategies, Interventions, Treatments, or Diagnostic Testing in Patient Care (Updated May 2019)*

CLASS (STRENGTH) OF RECOMMENDATION

CLASS 1 (STRONG) Benefit >>> Risk

Suggested phrases for writing recommendations:

  • Is recommended
  • Is indicated/useful/effective/beneficial
  • Should be performed/administered/other
  • Comparative-Effectiveness Phrases†:
    • Treatment/strategy A is recommended/indicated in preference to treatment B
    • Treatment A should be chosen over treatment B

CLASS 2a (MODERATE) Benefit >> Risk

Suggested phrases for writing recommendations:

  • Is reasonable
  • Can be useful/effective/beneficial
  • Comparative-Effectiveness Phrases†:
    • Treatment/strategy A is probably recommended/indicated in preference to treatment B
    • It is reasonable to choose treatment A over treatment B

CLASS 2b (WEAK) Benefit > Risk

Suggested phrases for writing recommendations:

  • May/might be reasonable
  • May/might be considered
  • Usefulness/effectiveness is unknown/unclear/uncertain or not well established

CLASS 3: No Benefit (WEAK) Benefit = Risk

Suggested phrases for writing recommendations:

  • Is not recommended
  • Is not indicated/useful/effective/beneficial
  • Should not be performed/administered/other

CLASS III: Harm (STRONG) Risk > Benefit

Suggested phrases for writing recommendations:

  • Potentially harmful
  • Causes harm
  • Associated with excess morbidity/mortality
  • Should not be performed/administered/other

LEVEL (QUALITY) OF EVIDENCE

LEVEL A

  • High-quality evidence‡ from more than 1 RCT
  • Meta-analyses of high-quality RCTs
  • One or more RCTs corroborated by high-quality registry studies

LEVEL B-R (Randomized)

  • Moderate-quality evidence‡ from 1 or more RCTs
  • Meta-analyses of moderate-quality RCTs

LEVEL B-NR (Nonrandomized)

  • Moderate-quality evidence‡ from 1 or more well-designed, well-executed nonrandomized studies, observational studies, or registry studies
  • Meta-analyses of such studies

LEVEL C-LD (Limited Data)

  • Randomized or nonrandomized observational or registry studies with limitations of design or execution
  • Meta-analyses of such studies
  • Physiological or mechanistic studies in human subjects

LEVEL C-EO (Expert Opinion)

  • Consensus of expert opinion based on clinical experience

COR and LOE are determined independently (any COR may be paired with any LOE).

A recommendation with LOE C does not imply that the recommendation is weak. Many important clinical questions addressed in guidelines do not lend themselves to clinical trials. Although RCTs are unavailable, there may be a very clear clinical consensus that a particular test or therapy is useful or effective.

* The outcome or result of the intervention should be specified (an improved clinical outcome or increased diagnostic accuracy or incremental prognostic information).

† For comparative-effectiveness recommendations (COR I and 2a; LOE A and B only), studies that support the use of comparator verbs should involve direct comparisons of the treatments or strategies being evaluated.

‡ The method of assessing quality is evolving, including the application of standardized, widely used, and preferably validated evidence grading tools; and for systematic reviews, the incorporation of an Evidence Review Committee.

COR indicates Class of Recommendation; EO, expert opinion; LD, limited data; LOE, Level of Evidence; NR, nonrandomized; R, randomized; and RCT, randomized controlled trial.