Levels of evidence
- Concept
The pyramid ranks studies by how well they avoid bias; design is a guide, but the quality of each study counts too.
What it is
Levels of evidence rank designs by their protection against bias for therapy questions: systematic reviews of randomised trials at the top, then trials, cohorts, case-control studies, case series and, at the base, expert opinion. Design alone is not enough: a poorly run trial can be worth less than a careful cohort, which is why systems such as GRADE judge the certainty of the evidence and not just the design.
Worked example
- Data (made up): three cohorts with 40 000 people suggest a supplement reduces heart attacks: pooled RR 0.70.
- First reading: 1 − 0.70 = 0.30, that is, 30 % fewer heart attacks in those who take it.
- But those who take the supplement exercise more and smoke less: possible confounding.
- Later, a randomised trial with 8 000 people finds RR 1.02 (95 % CI 0.90 to 1.15).
- The trial CI includes 1: no difference shown; the cohort effect was confounding by lifestyle.
- Reading: the trial (higher level) corrects the cohorts; even so, its quality (losses, blinding) is judged before accepting it.
Evidence pyramid for a therapy question
| Level | Design | Main risk |
|---|---|---|
| Highest | Systematic review of randomised trials | Quality of included trials, publication bias |
| High | Randomised controlled trial | Allocation concealment, losses, lack of blinding |
| Middle | Cohort study | Confounding, loss to follow-up |
| Lower middle | Case-control | Recall bias, choice of controls |
| Low | Case series or case report | No comparison group |
| Lowest | Expert opinion | Personal experience, the expert's own biases |
Markers
- SR
- RCT
- Cohort
- Case-control
- Expert opinion
In practice
- For a therapy question, look first for systematic reviews of trials and guidelines built on them.
- For rare or late adverse effects, large cohorts and case-control studies may be the best evidence available.
Common pitfalls
- Believing any trial beats any observational study: quality of conduct matters as much as design.
- Using the therapy pyramid for diagnosis or prognosis questions.
Test yourself
Think of the answer, then tap the question to see it.
Which design sits at the top of the pyramid for therapy questions?
The systematic review (with meta-analysis when appropriate) of good-quality randomised controlled trials.
Why does the randomised trial rank above the cohort?
Because randomisation spreads confounders, known and unknown, evenly between the groups.
Why is design not enough to judge evidence?
Because a well-designed study can be poorly run (losses, no blinding, few events) and its certainty falls.
Connections
Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.
Sources: Chapter 14: Completing ‘Summary of findings’ tables and grading the certainty of the evidence (Cochrane, 2026); Chapter 1: Starting a review (Cochrane, 2026).
Spotted a mistake?We are always improving for the community. Tell us what is wrong and our team will review it.Report a mistake
Representative model, for educational purposes only.