Evidence level | Definition | Example of study providing this level of evidence for a therapy question | Guideline statement grade |
Ia | A good recent systematic review of studies designed to answer the question of interest | Cochrane systematic review of randomised controlled trials studying the effectiveness of flu vaccines | A+ |
Ib | One or more rigorous studies designed to answer the question, but not formally combined | Randomised controlled trial of effectiveness of a flu vaccine | A− |
II* | One or more prospective clinical studies which illuminate, but do not rigorously answer, the question | Prospective cohort study comparing pneumonia rates in patients who are and are not vaccinated against flu; non-randomised controlled trial | B+ |
III† | One or more retrospective clinical studies which illuminate but do not rigorously answer the question | Audit or retrospective case control study, comparing flu vaccination history in patients who did and did not present with pneumonia | B− |
IVa‡ | Formal combination of expert views | Delphi study of UK expert recommendations for flu vaccination | C |
IVb | Other information | Expert opinion, informal consensus; in vitro or in vivo studies on related topics | D |
*Hard to differentiate Agency for Health Care Policy and Research’s “well designed controlled study without randomisation” (level IIa) from “other type of well-designed experimental study” (level IIb).
†Major criterion is retrospective versus prospective data collection, since non-experimental designs are better suited than even randomised clinical trials for answering certain questions.
‡Distinguish formal consensus from informal consensus methods according to the Health Technology Assessment 1998 systematic review.