Levels of evidence are a ranking of study designs: which kind of study is least likely to mislead you on a given question. For "does this treatment work?", the best-known schemes put a systematic review of randomised controlled trials at the top, a single trial below it, and observational studies, case reports and expert opinion further down. Drawn as a triangle, that ranking is the levels of evidence pyramid.
The ranking is a shortcut for searching, not a verdict on any paper, and the schemes' own authors say so. This guide sets out the Oxford CEBM levels, Melnyk and Fineout-Overholt's nursing levels and GRADE from their own documents, defines each design, and shows how to pull reviews or trials out of a search with the Kind of study filter in EdCitation's Find sources.
EdCitation does the finding and the citing, free, with Cite a source turning each keeper into a reference; it never writes any part of your work, so the appraisal and the argument stay yours.
What are the levels of evidence?
A hierarchy of evidence orders designs by their exposure to bias for one kind of question. The table below sets the two schemes most often named on nursing and health courses side by side, for questions about whether an intervention helps.
| Level | Study design | What it can show | Source |
|---|---|---|---|
| OCEBM 1; Melnyk I | Systematic review of randomised trials, with or without a meta-analysis | The combined answer of every eligible trial, only as sound as those trials | OCEBM table; Simmons guide |
| OCEBM 2; Melnyk II | Randomised controlled trial | Cause and effect of one intervention, with confounders spread by chance | OCEBM table; CEBM study designs |
| OCEBM 3; Melnyk III | Non-randomised controlled trial (Melnyk); non-randomised controlled cohort or follow-up study (OCEBM) | An effect, with more room for confounding | OCEBM table; Simmons guide |
| OCEBM 3 to 4; Melnyk IV | Cohort study, case-control study | Associations, risk factors, prognosis and harms | OCEBM table; Simmons guide |
| OCEBM 4 | Case series, historically controlled study | What happened to a group, with no fair comparison | OCEBM table |
| Melnyk V and VI | Systematic reviews of descriptive and qualitative studies; single descriptive or qualitative studies | Experience, meaning, description, prevalence | Simmons guide |
| OCEBM 5; Melnyk VII | Mechanism-based reasoning (OCEBM); expert opinion, narrative or literature reviews (Melnyk) | A reason to expect an effect, not a test of it | OCEBM table; Simmons guide |
| Search step | Any of the above | Find sources filters a topic search by kind of study before you read | EdCitation |
The OCEBM rows are for treatment benefits only; other questions have other ladders. A clinical guideline appears in neither scheme, because it is not a study design.
What is the difference between a systematic review and a meta-analysis?
A systematic review is the method; a meta-analysis is one statistical step that a review may or may not contain. The Cochrane Handbook describes a systematic review as an attempt to gather all the empirical evidence meeting criteria fixed in advance, using explicit methods chosen to limit bias (Lasserson et al., 2024). Its chapter on analysis defines meta-analysis as the statistical combination of results from two or more separate studies, and warns that it can mislead seriously when designs, biases and differences between studies are ignored (Deeks et al., 2024).
So a review can end without a meta-analysis when studies are too different to pool, and a meta-analysis can pool a convenient handful of studies with no review behind it. The strong form is both. A network meta-analysis goes one step further and compares several treatments at once, including pairs never tested head to head. How a review is actually carried out, stage by stage, is in how to do a systematic review.
What is an RCT, and how is it different from a cohort or case-control study?
An RCT, a randomised controlled trial, allocates participants to an intervention or a control group by a random mechanism, then follows both. The Centre for Evidence-Based Medicine (n.d.) gives the reason it ranks high: randomisation spreads confounders evenly between the groups. The costs: trials are expensive, prone to volunteer bias, and sometimes unethical to run.
Cohort, case-control and cross-sectional studies
These are observational studies: the researcher measures exposures and does not assign them. The CEBM page separates them by when the outcome is measured.
- Cohort study: people exposed and not exposed to a factor are followed, and outcomes are counted afterwards. Good for risk factors and prognosis; open to hidden confounders.
- Case-control study: people who already have an outcome are compared with people who do not, looking back at earlier exposure. Quick, cheap, and the only practical design for very rare conditions; open to recall and selection bias.
- Cross-sectional study: exposure and outcome measured at one moment. Best for prevalence and for the accuracy of a diagnostic test; it shows association at most, not cause.
Case reports, qualitative studies and guidelines
A case report describes one patient, a case series several; both are descriptive, with nothing to compare against. A qualitative study asks what an experience is like and why, which no trial answers. A clinical practice guideline, in the Institute of Medicine (2011) definition, is a set of recommendations for patient care informed by a systematic review of the evidence and a weighing of the benefits and harms of the options. It is built on the levels; it is not one of them.
What are the Oxford CEBM levels of evidence?
The Oxford CEBM 2011 Levels of Evidence are a one-page table, now version 2.1, with a five-step ladder for each clinical question: prevalence, diagnosis, prognosis, treatment benefits, common and rare harms, and screening. For treatment benefits, Level 1 is a systematic review of randomised trials or n-of-1 trials, Level 2 a randomised trial or an observational study with a dramatic effect, Level 3 a non-randomised controlled cohort or follow-up study, Level 4 case series, case-control or historically controlled studies, and Level 5 mechanism-based reasoning (OCEBM Levels of Evidence Working Group, 2011).
For "how common is the problem?", by contrast, Level 1 is a local, current random-sample survey.
Level 1 is read two ways
The OCEBM page flags a misreading of treatment Level 1: it means an n-of-1 randomised trial or a systematic review of randomised trials, not two kinds of review. A footnote lets any level be graded down for poor quality, imprecision, indirectness, inconsistency or a tiny absolute effect, and up for a large effect.
The introductory document, meant to be read with the table, is plainer. Howick et al. (2011) say the levels do not give a definitive judgement of quality, and that an observational study with a dramatic effect can outweigh a "higher level" review of a few inconclusive studies.
What are Melnyk's levels of evidence in nursing?
Melnyk and Fineout-Overholt's scale is a hierarchy of evidence nursing library guides in the United States often point to, Simmons University and Wright State University among them, with seven levels for intervention questions. The book itself could not be read for this guide; the levels below are from Simmons University Library (n.d.), which reproduces Box 1.3 of the 2023 fifth edition.
- Level I: a systematic review or meta-analysis of all relevant RCTs.
- Level II: well-designed single RCTs.
- Level III: controlled trials without randomisation; a systematic review of a whole body of evidence, higher and lower; mixed methods intervention studies.
- Level IV: well-designed case-control and cohort studies.
- Level V: systematic reviews of descriptive and qualitative studies.
- Level VI: a single descriptive or qualitative study, and EBP, quality improvement and EBQI projects.
- Level VII: the opinion of authorities, expert committee reports, and narrative or literature reviews.
Other pages word some levels differently, often from earlier editions, so check which edition your module reproduces. The same Simmons page shows Polit and Beck's eight-level scale, with non-research sources at Level VIII. How Melnyk differs from the JBI levels is in nursing students' APA citations; the rule is to use the scale your module names, and say which.
Level plus quality equals strength
The Simmons guide quotes Polit and Beck citing Melnyk and Fineout-Overholt's formula: level of evidence plus quality of evidence equals strength of evidence. Levelling a paper is half the job; appraising it is the other half.
Is the evidence pyramid wrong?
Not wrong, but too simple, say its best-known critics. Murad et al. (2016) grant that the traditional pyramid is intuitive and often right, then propose two changes. The straight lines between designs become wavy, because evidence moves up or down with its limitations whatever its design. And systematic reviews come off the top, to become a lens through which the other studies are read.
Their examples: a meta-analysis of five RCTs on tight glucose control should not rate high, because blinding and concealment were weak and the confidence interval spanned real benefit and real harm; hip replacement for disabling osteoarthritis, never tested in RCTs, is still something they are quite certain of. The new evidence pyramid is their picture of that.
What GRADE does instead
GRADE rates certainty, not designs. The GRADE Working Group (n.d.) asks users to consider each domain explicitly: risk of bias, imprecision, inconsistency, indirectness, publication bias, large effects, dose-response and residual opposing bias. In the handbook, randomised trials without important limitations start as high-quality evidence and observational studies as low, and the rating is made for each outcome across a body of evidence, not for each paper (Schünemann et al., 2013). Design sets the starting point; judgement does the rest.
The plain lesson from all three: the level of a design is not the quality of a given study. The methods section gives the level; only appraisal tells you whether to believe it.
How do I find systematic reviews and trials with the Kind of study filter?
Open EdCitation's Find sources, type the topic, and tick one or more boxes under Kind of study. The panel offers ten: Systematic review, Meta-analysis, Randomized controlled trial, Cohort study, Cross-sectional study, Qualitative study, Mixed methods, Case study or report, Literature or scoping review, and Guideline. Quick sets above it holds a one-click "Systematic reviews and meta-analyses". There is no box for case-control studies; add case-control to the search words instead.
The runs, 27 September 2026
The test topic was pressure ulcer prevention, sorted by relevance. The OpenAlex index answered every run and applied every filter.
- No kind of study ticked: 104,238 results. First: "Risk assessment scales for pressure ulcer prevention: a systematic review" (Journal of Advanced Nursing, 2006, cited 538 times), then "Pressure Ulcer Prevention and Management" (JAMA, 2003), then the 2015 Cochrane review "Support surfaces for pressure ulcer prevention".
- Systematic review and Meta-analysis ticked: 4,759 results. The 2006 review first, then the Cochrane review in its 2015, 2008 and 2011 versions, then "Support surfaces for pressure ulcer prevention: A network meta-analysis" (PLoS ONE, 2018, cited 104 times).
- Randomized controlled trial ticked: 3,233 results. The first four were Cochrane reviews and the fifth a 2002 paper in Annual Review of Nursing Research. The first trial came seventh: "Non-blanchable erythema as an indicator for the need for pressure ulcer prevention: a randomized-controlled trial" (Journal of Clinical Nursing, 2007).
What the filter matches, and what it misses
Run 3 shows how the filter works, as the panel says: "Found by what a paper calls itself in its title or abstract. No index records the kind of study." A review whose abstract says it included randomised controlled trials matches the trial box. Read the methods before you level a paper. The three Cochrane entries are versions of one review, numbered pub3 to pub5 in their DOIs: cite the version you read, and check the Cochrane Library for the current one. When the main index is busy, the page names any filter the back-up index could not apply, so an unfiltered list is never passed off as a filtered one.
Citing a result
We pasted the network meta-analysis's DOI, 10.1371/journal.pone.0192707, into Cite a source. It read the Crossref record and, in APA 7, returned:
Shi, C., Dumville, J. C., & Cullum, N. (2018). Support surfaces for pressure ulcer prevention: A network meta-analysis. PLOS ONE, 13(2), Article e0192707. https://doi.org/10.1371/journal.pone.0192707
No retraction or notice was reported. The subtitle keeps its capital after the colon, as APA 7 asks, and the journal name follows the record.
Which tool is best for finding studies by their design?
For a topic search narrowed by design, EdCitation's Find sources is the best free place to start, and the runs show why: one tick took 104,238 results to 4,759 reviews, with the matching rule printed on the panel. It searches about 300 million published works by topic or by the claim a sentence needs, and every result is a record looked up in the index, never a reference composed by a chatbot. A systematic review's own search belongs in the databases its protocol names, as how to find sources for a literature review explains.
After that, Cite a source builds each entry in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver and screens for retraction, and Verify references marks each entry of the finished list verified, "check this" or not found, and never reports one it could not check as not found. Check your paper turns a brief's rules on how many sources and how recent into a checklist. All four are free, with no account. Pro, at $8 a month, adds References from a file for a whole paper; Max, at $24 a month, adds Theoretics QA and the Library of books. For a study's own checklist, see reporting guidelines for student research; for the Source type filter, types of academic sources.
Quick questions
Is a systematic review always the highest level of evidence?
For treatment questions it heads the OCEBM and Melnyk scales, but not for every question: for prevalence, OCEBM ranks a local random-sample survey above a systematic review. Murad et al. (2016) would take reviews off the pyramid altogether and use them as a lens.
Is an RCT always better than a cohort study?
As a design for treatment questions, yes, on every scheme here. As a study, not necessarily: GRADE rates trials down for flaws and observational evidence up for large effects, so a weak trial can end below a strong cohort.
Which levels of evidence should a nursing student use?
The scale your module or placement names, usually Melnyk and Fineout-Overholt's seven levels or JBI's. Name the scale and edition in your paper and apply it to every source.
Can EdCitation find only systematic reviews or trials?
Yes. Tick the design under Kind of study in EdCitation's free Find sources, or use the "Systematic reviews and meta-analyses" quick set. It matches what a paper calls itself, so read the methods before relying on the label.
References
- Centre for Evidence-Based Medicine. (n.d.). Study designs. University of Oxford. https://www.cebm.ox.ac.uk/resources/ebm-tools/study-designs
- Deeks, J. J., Higgins, J. P. T., Altman, D. G., McKenzie, J. E., & Veroniki, A. A. (Eds.). (2024). Chapter 10: Analysing data and undertaking meta-analyses. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10
- GRADE Working Group. (n.d.). Welcome to the GRADE working group. https://www.gradeworkinggroup.org/
- Howick, J., Chalmers, I., Glasziou, P., Greenhalgh, T., Heneghan, C., Liberati, A., Moschetti, I., Phillips, B., & Thornton, H. (2011). The 2011 Oxford CEBM levels of evidence: Introductory document. Oxford Centre for Evidence-Based Medicine. https://www.cebm.ox.ac.uk/files/levels-of-evidence/cebm-levels-of-evidence-introduction-2-1.pdf
- Institute of Medicine. (2011). Clinical practice guidelines we can trust. The National Academies Press. https://doi.org/10.17226/13058
- Lasserson, T. J., Thomas, J., & Higgins, J. P. T. (2024). Chapter 1: Starting a review. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-01
- Murad, M. H., Asi, N., Alsawas, M., & Alahdab, F. (2016). New evidence pyramid. Evidence-Based Medicine, 21(4), 125–127. https://doi.org/10.1136/ebmed-2016-110401
- OCEBM Levels of Evidence Working Group. (2011). The Oxford 2011 levels of evidence (Version 2.1). Oxford Centre for Evidence-Based Medicine. https://www.cebm.ox.ac.uk/resources/levels-of-evidence/ocebm-levels-of-evidence
- Schünemann, H., Brożek, J., Guyatt, G., & Oxman, A. (Eds.). (2013). GRADE handbook for grading quality of evidence and strength of recommendations. The GRADE Working Group. https://gdt.gradepro.org/app/handbook/handbook.html
- Shi, C., Dumville, J. C., & Cullum, N. (2018). Support surfaces for pressure ulcer prevention: A network meta-analysis. PLOS ONE, 13(2), Article e0192707. https://doi.org/10.1371/journal.pone.0192707
- Simmons University Library. (n.d.). Nursing: Evidence-based practice: Levels of evidence. https://simmons.libguides.com/c.php?g=1033284&p=7490072