The levels of evidence pyramid ranks study designs by methodological quality. Systematic reviews and meta-analyses of randomized controlled trials sit at the top and expert opinion at the bottom. Use it to decide which evidence to check first, not which to accept without appraisal.
The pyramid is one part of a wider skill set. For the full toolkit, start with how to tell if a health study is reliable.
What the levels of evidence pyramid ranks
The pyramid ranks designs, not individual papers. According to an evidence-based practice toolkit from Indiana Wesleyan University’s library, levels of evidence are assigned to studies based on the methodological quality of their design, their validity, and their applicability to patient care. A Memorial Sloan Kettering library guide describes the same idea as a hierarchy ranked by the strength of the results of a clinical trial or research study.

The mechanism is bias control. Designs higher up are built to limit the ways a result can mislead you, such as who ended up in which group or who knew which treatment a person received. That is why the ranking tracks design rather than topic or journal.
Two limits matter from the start:
- Design is not the only factor. The same guide notes that the strength of results can be affected by study design, outcomes and bias, so a well-run lower-level study can be more informative than a poorly run higher-level one.
- Evidence is one input. Evidence-based practice, as a California State University San Marcos library guide defines it, integrates three things: clinical expertise, patient or client values, and the best research evidence. The pyramid only helps with the last of those.
Why systematic reviews sit at the top
Systematic reviews and meta-analyses of randomized controlled trials are considered the strongest level of evidence on which to guide practice decisions, according to the Indiana Wesleyan toolkit. The reason is what a systematic review does: it systematically searches for, appraises and summarizes all of the medical literature on a specific topic. A single study shows one result; a review weighs the whole body of research.
Two further features explain the position:
- Filtering. Systematic reviews, meta-analyses and critically appraised topics have all gone through an evaluation process, whereas uncritically appraised information is unfiltered.
- Quantitative pooling. A meta-analysis is a systematic review that uses quantitative methods to summarize the results.
A practical place to look is the Cochrane Library, which the CSUSM guide describes as a collection of databases containing different types of high-quality, independent evidence.

Trials and observational designs in the middle
Below reviews sit the primary studies they are built from. What separates them is mostly one question: was the intervention randomly allocated? In the decision tree published in the Memorial Sloan Kettering guide, an analytic study with random allocation is a randomized controlled trial, and one without is an observational study.
Randomized controlled trials
A randomized controlled trial places a randomized group of patients in an experimental group and a control group and follows them for outcomes. Trials that are randomized and double-blind avoid selection and observer bias, as the Indiana Wesleyan toolkit explains. That protection against bias is the reason trials rank above observational work, so check whether a trial actually used both features before giving it that credit.
Cohort, case-control and cross-sectional studies
Observational designs differ in timing:
- Cohort study. It identifies two groups of patients, one that received the exposure of interest and one that did not, and follows them forward to see who develops the outcome.
- Case-control study. It starts from people who already have the outcome and compares them with controls for earlier exposure. Because it relies on recall of exposure, the Memorial Sloan Kettering guide labels it a retrospective design.
- Cross-sectional study. It measures the outcome at the same time as the exposure, so it shows association at one moment rather than sequence.
If you are weighing whether a question needs an experiment or can be answered by observation, Experiment Vs Observational Study: How To Choose The Right Method In 2026 – Testpledge goes through that choice in detail.
Evidence hierarchy at a glance
| Design | What it does | Planning use |
|---|---|---|
| Systematic review or meta-analysis | Searches, appraises and summarizes the literature on a topic; a meta-analysis adds quantitative pooling | Look here first for practice decisions |
| Randomized controlled trial | Randomizes patients to experimental and control groups; double-blinding limits selection and observer bias | Best single-study answer on whether an intervention works |
| Cohort study | Follows an exposed and an unexposed group forward to the outcome | Useful when randomizing is not possible |
| Case-control study | Compares people with the outcome against controls for prior exposure | Treat as a lead, since it depends on recall |
| Cross-sectional study | Measures outcome and exposure at the same time | Shows association only; sequence is unclear |
| Expert opinion | Opinion from authorities or expert committees | Weakest level; use only when nothing stronger exists |
Design descriptions and rankings follow the Indiana Wesleyan evidence toolkit and the Memorial Sloan Kettering study-design guide, which points readers to the 2011 Centre for Evidence Based Medicine table for assigning a study’s level.
The weak end of the pyramid
The weakest level of evidence is the opinion of authorities and reports of expert committees, as the Indiana Wesleyan toolkit puts it. Opinion can be informed and still rank low, because it is not a method that controls bias. It is a reasonable starting point when you have nothing else, and a poor place to stop when better designs exist.
When the pyramid does not settle the question
The pyramid does not always apply cleanly. Higher levels of evidence may not exist for your clinical question, and fewer studies exist the higher you go. The toolkit’s advice is to move down the pyramid when a search at the top is unsuccessful. That is a sensible fallback, but the lower the design, the more caution the conclusion needs.
Rank also does not replace appraisal. Because strength of results depends on design, outcomes and bias together, a flawed trial can be less trustworthy than a careful cohort study. Reading the methods still matters.
Findings from laboratory models raise a related problem, since they are not studies of people at all. In Vivo vs In Vitro: How They Differ And When Each Matters (2026 Guide) – Testpledge explains why that distinction matters when you weigh a claim.
Using the pyramid on supplement and food claims
Consumers rarely meet the pyramid by name, but the same logic helps with product claims. A step-by-step approach:
- Find the design behind the claim. Ask whether it rests on a randomized trial, an observational study, or only on what an authority says. The design tells you where the claim sits.
- Look for a review. If a systematic review or meta-analysis exists, it outranks any single study a seller quotes. Searching a collection such as the Cochrane Library is a good first move.
- Treat observational findings as leads. A cohort or case-control result may show an association between a food and an outcome. It does not carry the bias protection of randomization.
- Discount claims resting on opinion alone. If the only support is an authority’s statement, that is the weakest level.
- Balance the evidence with your situation. Evidence-based practice combines research with expertise and personal values, so discuss a high-stakes choice with a clinician rather than relying on a ranking.
The pyramid will not tell you whether a supplement is safe or worthwhile. It tells you how much weight the supporting evidence can bear, which is often enough to decide whether a claim deserves further checking.
Frequently asked questions
What does a meta-analysis add to a systematic review?
A meta-analysis is a systematic review that uses quantitative methods to summarize the results, so it combines findings statistically instead of only describing them.
How do cohort and case-control studies differ?
A cohort study follows an exposed and an unexposed group forward to see who develops the outcome. A case-control study begins with people who have the outcome and looks back at their exposure.
What should I do if no systematic review exists for my question?
Move down the pyramid to the best available design, such as a randomized trial or a cohort study, and read it critically for bias.
Who assigns a study its level?
Levels are assigned based on design quality, validity and applicability to patient care. Library guides often use the Centre for Evidence Based Medicine’s 2011 table to work out a study’s level.



