The evidence pyramid is a triangle that ranks types of research by how well their design protects against bias. Systematic reviews and meta-analyses sit at the top, single studies in the middle, and animal work, expert opinion and case reports at the base. It is a first filter, not a verdict: the GRADE approach then adjusts the rating for the quality of each body of evidence, and clinical judgment decides what to do for the patient in front of you. This guide covers the levels, the main models, GRADE, and a practical routine. We also point to the places where ConnectedMedics helps clinicians keep up with research.
TL;DR
- The pyramid ranks study designs: syntheses of research on top, then randomized trials, cohort and case-control studies, case series and reports, and lab or animal studies and opinion at the base.
- Models differ. Some use six levels, some five (Johns Hopkins), and CEBM ranks by question type. A "level 1" in one system may not match "level 1" in another.
- GRADE rates certainty as high, moderate, low or very low. It starts with study design, then moves the rating up or down.
- A well-run observational study can outrank a poorly run trial. Design sets the starting point, not the final answer.
- Evidence informs the decision. Clinical expertise and patient preferences complete it.
Table of Contents
- What the evidence pyramid shows
- The levels of evidence, from top to base
- Why different guides rank evidence differently
- The GRADE approach: from study design to certainty
- What the pyramid does not tell you
- How to use the levels on a real clinical question
- Our perspective
- Keep evidence and career in one place
- FAQ
- Sources
What the evidence pyramid shows
The hierarchy of evidence is drawn as a triangle with the strongest evidence at the top. Research syntheses such as meta-analyses and systematic reviews sit at the top, primary studies below them, and animal and laboratory studies at the base. The logic is simple. Designs that reduce chance, bias and confounding (factors that distort the link between a treatment and an outcome) earn a higher place.
Two terms help when you read pyramid diagrams:
- Filtered information is research that someone has already appraised and combined: systematic reviews, meta-analyses, practice guidelines and critically appraised topics.
- Unfiltered information is primary research: trials, cohort studies and case-control studies.
The University of Minnesota guide places filtered information at the top, unfiltered primary literature in the middle, and background information and expert opinion at the base. A 2025 review describes the pyramid as a graphical structure for ranking the quality and dependability of research methods in medical science.
The idea is not new. Evidence-based medicine has said from the start that not all evidence is the same. That sentence is the reason the pyramid exists, and the reason you should check what kind of study stands behind a claim before you act on it.
The levels of evidence, from top to base
One widely taught version is a six-level pyramid:
| Level | Study design | What it is |
|---|---|---|
| 1 | Systematic reviews and meta-analyses | A structured search and appraisal of all relevant studies; a meta-analysis pools their results statistically |
| 2 | Randomized controlled trials (RCTs) | Participants are assigned to treatments by chance and followed |
| 3 | Cohort studies | Groups with and without an exposure are followed over time |
| 4 | Case-control studies | People with an outcome are compared with similar people without it, looking back at exposures |
| 5 | Case series, case reports, expert opinion | Descriptions of patients or authoritative views, with no comparison group |
| 6 | Animal and laboratory studies | Mechanism and early-stage research |
Why randomized trials rank so high
NYU's evidence-based medicine guide calls RCTs the gold standard for evaluating treatment effects. The reason it gives is that randomization removes known and unknown confounders, meaning factors that could explain a result other than the treatment itself. The same guide separates primary studies (original research) from secondary studies (systematic reviews and meta-analyses that synthesize primary research). Background knowledge such as textbooks and clinical guidelines sits at the base.
Why reviews sit above trials
A single trial can be small, unusual or lucky. A systematic review gathers every eligible trial, appraises each, and often pools them. The result is more precise and less dependent on any one study. A review is only as good as the studies inside it, though. A systematic review of weak studies does not become strong evidence because it is a review.
Why different guides rank evidence differently
If you have searched "is an RCT level 1 or level 2 evidence", you have met this problem. The answer depends on the system.
The Johns Hopkins five-level model
The Johns Hopkins Evidence-Based Practice Model, as summarized by the OHSU nursing library, uses five levels:
- Level I: experimental studies (RCTs) and systematic reviews of RCTs.
- Level II: quasi-experimental studies and systematic reviews that combine RCTs and quasi-experimental studies.
- Level III: non-experimental studies, qualitative studies and systematic reviews that combine them.
- Level IV: opinion of respected authorities and expert committees based on scientific evidence, such as clinical practice guidelines and consensus panels.
- Level V: experiential and non-research evidence, such as literature reviews, quality improvement evaluations, case reports and expert opinion based on experience.
Here an RCT is Level I, and a systematic review of RCTs shares that level. In the six-level pyramid above, the review is level 1 and the RCT is level 2. Neither is wrong. They answer slightly different questions about how to group the designs.
Rankings by question type
The Oxford Centre for Evidence-Based Medicine (CEBM) table ranks evidence separately for prevalence, diagnosis, prognosis, treatment benefits, treatment harms and screening. Step 1 is the strongest and Step 5 is mechanism-based reasoning. For treatment benefits, the top step is a systematic review of randomized trials.
This matters because the best design depends on what you ask. For therapy questions, one hospital-system library guide puts a meta-analysis or systematic review of RCTs first and a case series last. For diagnosis questions, the strongest evidence is a systematic review of cross-sectional studies that use consistent reference standards. A diagnostic question, such as how well a test identifies a condition, is not best answered by randomizing people to a treatment. Our own knowledge hub carries research on that side of the topic, for example the QUADAS-3 tool for assessing the quality of diagnostic accuracy studies and its explanation and elaboration guidance. For a clinician's view of how diagnostic evidence gets used in practice, the Beyond the Diagnosis podcast episode is one example of a conversation about objective evidence in diagnosis.
A U.S. primary care example
The American Academy of Family Physicians' Strength-of-Recommendation Taxonomy (SORT) uses three levels. For treatment, prevention and screening, Level 1 covers systematic reviews or meta-analyses and high-quality RCTs with consistent findings, while Level 2 includes lower-quality trials, cohort studies and case-control studies. If you read family medicine literature, expect this scale.
The GRADE approach: from study design to certainty
The pyramid ranks designs. GRADE rates the certainty of the evidence for a specific outcome. The founding GRADE paper defines quality of evidence as the extent to which one can be confident that an estimate of effect is correct, and strength of recommendation as the extent to which one can be confident that following it will do more good than harm.
GRADE works in three steps:
- Start from the design. Per a 2023 review by GRADE authors, randomized trials begin as high-certainty evidence and observational studies as low-certainty evidence.
- Rate down for problems. The five domains are risk of bias, inconsistency, indirectness, imprecision and publication bias. Reviewers rate down one level for serious concerns and two for very serious ones.
- Rate up when justified. Observational evidence can move up, mainly for a large or very large magnitude of effect. The CDC's handbook for its Advisory Committee on Immunization Practices lists strength of association, dose-response, and opposing plausible residual confounding or bias as the criteria for upgrading non-randomized studies.
The end result is one of four ratings: high, moderate, low or very low. The CDC handbook notes that ACIP has replaced its numbered levels with those four terms.
Strong and conditional recommendations
GRADE also separates certainty from the strength of a recommendation. Guideline panels make strong recommendations when all or almost all fully informed patients would choose the proposed intervention, and weak (conditional) recommendations when patients' choices would vary with their values and preferences. The same review reports that more than 110 organizations, including the World Health Organization, use GRADE.
This is why a guideline can give a strong recommendation built on moderate evidence, or a conditional one built on high-certainty evidence. The two ratings answer different questions.
What the pyramid does not tell you
The pyramid is a good starting point and a poor stopping point. The Joanna Briggs Institute cautions that levels give a ranking of the likely best available evidence and should not be used as a definitive measure of it. Its guidance adds that levels do not replace critical appraisal and clinical reasoning, and that observational evidence may sometimes be preferred over experimental evidence.
In practice, that means four checks beyond the label:
- Quality inside the level. A trial with weak blinding or heavy dropout can mislead more than a careful cohort study.
- Fit to your patient. A trial in a narrow population may not apply to an older, multimorbid patient.
- Reporting. Selective reporting of outcomes can distort even top-tier designs. Our knowledge hub lists a meta-research paper on outcome switching in cohort studies of interventions.
- Clarity of guidance. Evidence can be sound and still be turned into unclear advice. See the knowledge hub entry on how ambiguous guideline recommendations harm patient care.
One commentary on the pyramid's limits makes the same point: evidence alone does not set best practice. Clinical expertise and patient preferences carry weight too.
How to use the levels on a real clinical question
A short routine works for most clinicians, whether you are a physician reviewing a drug, a nurse updating a protocol, a pharmacist checking an interaction, or an administrator weighing a program.
- Write the question. Name the patient group, the intervention or exposure, the comparison and the outcome. State the question type: treatment, harm, diagnosis, prognosis or prevalence.
- Pick the right top design. For treatment benefit, look for a systematic review of RCTs. For diagnosis, look for diagnostic accuracy studies. Do not demand an RCT for a question an RCT cannot answer.
- Look for filtered sources first. Start with systematic reviews and current clinical guidelines. Our summary of clinical guidelines for healthcare professionals is one place to begin.
- Check the certainty rating. If the source reports GRADE, read the rating for the outcome that matters to your patient, not only the overall headline.
- Appraise what is left. Where no review exists, read the primary studies. Check randomization, follow-up, outcome reporting and whether the population resembles yours.
- Combine with judgment. Weigh the patient's values, your clinical experience and the local setting. Document the reasoning.
For more habits that make this routine quicker, see our evidence-based medicine tips for doctors and nurses. Where no trial exists, case reports still have a role. We cover how clinical case studies are applied, and a case discussion checklist helps teams weigh them.
A worked example
Suppose a nurse on a medical-surgical unit is asked whether a new dressing protocol reduces infection rates compared with the current one. The question is about treatment benefit, so the top design is a systematic review of RCTs. She finds one review. It pools several small trials and GRADE rates the evidence for infection as low certainty, because of imprecision and some risk of bias. The pyramid placed the review at the top, yet the certainty is low.
That result does not mean "ignore it". It means the protocol change is a conditional recommendation. The unit might pilot it, track infection rates, and ask patients about their preferences. The levels got her to the right evidence quickly. GRADE told her how far to trust it. Judgment set the plan.
Our perspective
Clinicians rarely lack evidence. They lack time and a trusted place to discuss it. A paper's rung on the pyramid tells you where to start reading. It does not tell you whether the result fits your ward, your clinic or your patient mix.
That is why we built ConnectedMedics around verified healthcare professionals and a knowledge hub that summarizes research and clinical trends. When a study is summarized next to peers in your own specialty, the useful questions come up quickly: Was the population like ours? Is the effect large enough to matter? What do colleagues do when the evidence is thin?
We think the best use of the evidence pyramid is as a shared language. Specialty-led communities and peer-to-peer discussion give that language a place to be used, with credentials checked so you know who is talking.
David
Keep evidence and career in one place
Reading research is one part of professional growth. Finding the next role is another. ConnectedMedics is a network for healthcare professionals with verified profiles, a knowledge hub of clinical insights and research summaries, and a global jobs board with more than 22,000 active healthcare vacancies. Members can follow specialty-led communities and discuss clinical questions with peers.
Browse the Knowledge Hub for research summaries, or join ConnectedMedics to connect with colleagues in your specialty and see healthcare jobs matched to your field.
FAQ
What is the evidence pyramid in simple terms?
It is a triangle that ranks study types by how reliable their design is. Systematic reviews and meta-analyses are at the top, then randomized trials, cohort and case-control studies, and case reports. Animal studies, lab work and expert opinion sit at the base. It is a quick guide to where to start, not a final judgment.
Is an RCT level 1 or level 2 evidence?
It depends on the system. In a six-level pyramid, systematic reviews are level 1 and RCTs are level 2. In the Johns Hopkins model, RCTs are Level I. In the AAFP taxonomy, high-quality RCTs with consistent findings are Level 1. Check which scale a source uses before comparing labels.
What is the GRADE approach to evidence?
GRADE rates how certain we can be about an effect, as high, moderate, low or very low. Randomized trials start high and observational studies start low. Reviewers then rate down for risk of bias, inconsistency, indirectness, imprecision or publication bias, and can rate up for a large effect. Recommendations are then graded as strong or conditional.
Can observational studies beat randomized trials?
Sometimes. Under GRADE, a poorly conducted trial can be rated down, and a well-conducted observational study with a very large effect can be rated up. Guidance from the Joanna Briggs Institute also says observational evidence may sometimes be preferred. Design is the starting point, not the end of the appraisal.
Does the pyramid replace clinical judgment?
No. Evidence is one of three inputs, with clinical expertise and patient preferences. The levels rank study designs, and critical appraisal and clinical reasoning still decide how to apply them to a specific patient.
Sources
- Levels of Evidence, Nursing and Physician Assistant, UC Davis library guide
- Evidence Hierarchy: What is the Best Evidence?, University of Minnesota
- Understanding the Levels of Evidence in Medical Research, 2025
- New evidence pyramid, Evidence-Based Medicine, 2016
- Appraisal and Levels of Evidence, Purdue University Northwest
- The Evidence Landscape, NYU Health Sciences Library
- Appraising Evidence, Oregon Health & Science University
- CEBM Levels of Evidence Table
- Study Design by Question Type, Advocate Health Library
- The Strength-of-Recommendation Taxonomy, AAFP
- Grading quality of evidence and strength of recommendations, 2004
- Certainty of evidence, why?, 2023
- CDC ACIP GRADE Handbook, Chapter 7
- Supporting Document for the Joanna Briggs Institute Levels of Evidence