Skip to main content

Table 1 Levels of evidence and recommendation

From: Evidence levels in radiology: the insights into imaging approach

Levels of evidence and recommendation
Level of evidence Confidence in the evidence and recommendations grade
High Data derived from metanalyses or systematic reviews or from (multiple) randomized trials with high quality
Large retrospective observational studies or in silico clinical trials with external validation
Well defined reference standards and controlled biases
The described technique improves healthcare pathway (tests, treatment, hospitalization) or decreases costs per patient
Level is graded down to Moderate if there are limiting biases or inconsistencies between studies
Further research is unlikely to change our confidence in the estimate of benefit and risk
Strongly recommended, mainly if presumed important patient outcomes and/or acceptable costs
Wording associated with the High grade of recommendation: ‘‘must”, ‘‘should” “recommend”
Moderate Data derived from a single large randomized clinical trial or multiple nonrandomized studies
Large retrospective observational multicentre studies or large in silico clinical trials with controlled design and internal validation. Appropriate spectrum of cases
Studies on technique assessments of noninferiority, surrogate biomarkers or changes in clinical management
Level can be upgraded to High if there is a demonstrated large effect size or downgraded if the effect size is small
Further research is likely to have an impact on our confidence in the estimate of benefit and risk and may change the estimate
Recommendation is modulated to strong or weak by the presumed patient outcomes and final costs
Low Small series, non-validated results and single centre observational, experimental or technical studies
None or imperfect reference standards
No study on the validation of results
Large possible biases
Opinions, general statements, critical and educational reviews without analytical methods
Studies on either technical efficacy or diagnostic validation accuracy (reference standards)
Any estimate of effect is uncertain
Weak recommendation, mainly if not clear patient important outcomes and/or high cost
Wording associated with the Low grade of recommendation: ‘‘could”, ‘‘may”, “suggests”