Contents
Academy · Research Writing · COSMOS-E Guidance
COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology
COSMOS-E is guidance developed specifically for systematic reviews that examine whether an exposure causes a disease. Observational evidence addresses this question very differently from the randomised-trial evidence for which most review methodology was designed. This article outlines the COSMOS-E recommendations and their relation to PRISMA and PROSPERO.
COSMOS-E (Conducting Systematic reviews and Meta-analyses Of Studies on causes, i.e. etiology) is methodological guidance published by Dekkers and colleagues for teams synthesising observational evidence on whether an exposure causes an outcome, as distinct from whether a treatment works. It exists because the dominant systematic review frameworks, built around pooling randomised trials, do not map cleanly onto a body of cohort, case-control, and cross-sectional studies where confounding, not chance, is usually the main threat to a valid answer.
Learning objectives
- Why etiology reviews require guidance separate from intervention-review frameworks such as Cochrane’s
- The sequence of decisions COSMOS-E recommends, from question framing through to the pooling decision
- How COSMOS-E treats confounding and bias assessment as central tasks rather than a final checklist
Common misconceptions
- “PRISMA and Cochrane intervention-review methods are sufficient for reviewing cause-and-effect observational research.” COSMOS-E was written specifically because those frameworks assume that randomisation has already addressed confounding; etiology reviews pool studies in which it has not, which alters almost every subsequent methodological decision, from eligibility criteria to whether pooling is appropriate at all.
Definition of COSMOS-E and the need for etiology-specific guidance
COSMOS-E is guidance for conducting and reporting systematic reviews and meta-analyses of observational studies that investigate etiology, meaning the causal relationship between an exposure and a disease outcome1. It was developed because the two most widely used review frameworks, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for reporting and the Cochrane Handbook for conducting intervention reviews, were designed for the synthesis of randomised controlled trials, in which random allocation is assumed to have balanced known and unknown confounders between groups25. Etiology reviews address cohort, case-control, and cross-sectional studies in which no randomisation has occurred, so confounding, exposure misclassification, and reverse causation are active threats in every included study rather than problems already resolved by randomisation. COSMOS-E does not replace PRISMA reporting standards; it complements them, addressing the conduct decisions specific to etiological, non-randomised evidence.
Guidance designed for intervention trials assumes that the principal remaining question, once a trial has been identified and appraised, is how precisely and consistently it estimates an already isolated causal effect. Etiology reviews face a prior question that those frameworks need not address: whether the comparison in any given included study is valid at all, given that the exposed and unexposed groups were never randomly formed. A review that applies the Cochrane intervention-review checklist without modification to a set of cohort studies will extract sample sizes and outcome rates while overlooking the assessment of confounder adjustment that determines whether those numbers are meaningful; this is the specific gap that COSMOS-E was written to address.
The core sequence of COSMOS-E recommendations
COSMOS-E organises the review process as a sequence of decisions rather than a single checklist, starting with framing a precise etiological question that specifies the exposure, the outcome, and the population in terms specific enough to define eligible study designs1. It then addresses the breadth of the search strategy, noting that filters designed for the randomised-trial literature will miss much of the relevant cohort and case-control evidence, and recommends registering the protocol, for instance through the International Prospective Register of Systematic Reviews (PROSPERO), before screening begins14. A distinctive feature of the guidance is that it treats the decision of whether to pool studies statistically as a genuine choice requiring justification rather than a default step, because combining etiological estimates from studies with materially different confounder adjustment can produce a pooled estimate that is precise but not valid.
Because eligible study designs for an etiology question typically span the cohort, case-control, and cross-sectional literature simultaneously, COSMOS-E recommends explicit justification of which designs the review will include and why, rather than defaulting to whichever design predominates in the initial search results. The guidance treats eligibility as a decision based on which designs can answer the specific exposure-outcome question, not as a matter of maximising inclusiveness. The breadth of the search strategy follows from the same principle: because cohort and case-control studies are indexed less consistently around a single intervention term than trials are, the guidance recommends combining controlled-vocabulary terms with free-text synonyms for both exposure and outcome, and supplementing database searches with reference-list screening of included studies and relevant prior reviews to identify literature that a database search alone would miss.
Treatment of confounding and bias assessment in COSMOS-E
COSMOS-E devotes substantial attention to confounding because it is the central threat to validity in etiological observational evidence, recommending that reviewers specify likely confounders of the exposure-outcome relationship before extraction begins and then record how adequately each included study measured and adjusted for them1. This differs from applying a single risk-of-bias score retrospectively: the guidance asks reviewers to judge whether studies are sufficiently comparable in their handling of confounders to be combined, and to conduct sensitivity analyses excluding studies with materially different adjustment sets to determine whether a pooled estimate is driven by a subset of better- or worse-adjusted studies1. Tools designed to assess the quality of non-randomised studies, such as the Newcastle-Ottawa Scale, can inform this assessment, but COSMOS-E frames the judgement as substantive rather than purely mechanical6.
A practical technique for identifying confounders systematically is the directed acyclic graph (DAG), a diagram of assumed causal relationships between exposure, outcome, and other variables, which makes explicit which variables require adjustment and which, if adjusted for, would introduce rather than remove bias. Reviewers extracting data from included studies can use a consistent DAG to check whether each study adjusted for the same core confounders, rather than accepting the covariate set selected by each study’s authors. The guidance also distinguishes effect modification from confounding, noting that a causal effect that genuinely differs across subgroups, such as by age, sex, or baseline risk, is a substantive finding to be explored through planned subgroup analysis rather than a problem to be averaged out in a single pooled estimate1. Where cohort, case-control, and cross-sectional studies, with largely independent sources of potential bias, converge on a similar estimate, COSMOS-E regards that convergence, sometimes termed triangulation, as meaningfully strengthening confidence in a causal interpretation beyond what any single design could support.
Application of COSMOS-E alongside PRISMA and PROSPERO
In practice, a team conducting an etiology review uses COSMOS-E for conduct decisions, PRISMA 2020 for reporting the completed review, and PROSPERO or a comparable registry for pre-specifying the protocol, with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach available subsequently to rate the certainty of the resulting evidence for a causal association247. The practical value of treating these as distinct, complementary tools is that each addresses a different failure: PRISMA addresses incomplete reporting, PROSPERO reduces the risk of post hoc changes to the question, and COSMOS-E addresses whether the underlying observational comparison was sufficiently valid to answer a causal question.
Fields such as nutrition and environmental epidemiology, in which randomising exposure is rarely feasible or ethical, rely particularly on this combined workflow, since a single well-conducted etiology review in these areas can inform public health guidance for years. Because the underlying observational evidence base continues to grow, a review based on COSMOS-E is also a reasonable candidate for periodic updating rather than a single exercise; the guidance recommends documenting the literature search date clearly, so that a later update can identify the evidence on which the original conclusions were based. Teams new to this type of review benefit from piloting the confounder-extraction and bias-assessment forms on two or three included studies before extracting the full set, since the judgement-based steps of COSMOS-E are easier to apply consistently once the team has calibrated its interpretation against a small number of real examples.
The final manuscript should also state explicitly where judgements were made under COSMOS-E and why, since a reader evaluating the conclusions of an etiology review needs to see not only the pooled estimate but also the reasoning behind which confounders were prioritised, which studies were excluded from pooling and on what basis, and how consistent the underlying study designs were with one another. A review that documents this reasoning transparently provides readers, and future reviewers updating the work, with the information needed to judge whether the same judgements would remain valid in light of newer evidence. This is what distinguishes an etiology review based on COSMOS-E from a less structured narrative summary of the same literature: not that its conclusions are necessarily stronger, but that every step underlying those conclusions is documented clearly enough for others to verify, replicate, or challenge.
Key points
Key points
Etiology reviews require the confounding-centred conduct guidance of COSMOS-E in addition to, not instead of, PRISMA reporting and PROSPERO registration, because randomised-trial review methods do not address the specific validity threats of observational evidence on causation.
Practical application
Practical application
- The etiological question should be framed with sufficient specificity to determine eligible study designs. The exposure, outcome, population and comparison should be defined precisely before searching, so that the eligible evidence base is clear from the outset.
- Likely confounders should be pre-specified before data extraction. Expected confounders of the exposure-outcome relationship should be listed in advance, and each included study’s measurement of and adjustment for them recorded during extraction.
- Pooling should be treated as a decision rather than a default. Sensitivity analyses excluding studies with materially different confounder adjustment should be conducted before a single pooled estimate is reported.
Frequently asked questions
No. COSMOS-E is conduct guidance focused on the decisions specific to etiological observational research, particularly confounding assessment and the pooling decision; PRISMA remains the reporting standard once the review is complete.
It is written specifically for etiological questions about exposure-disease causation; prediction and prognosis reviews are better served by frameworks such as the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) and the Prediction model Risk Of Bias ASsessment Tool (PROBAST), which address different threats to validity.
It does not mandate a specific tool but expects reviewers to assess the handling of confounders explicitly, often using instruments such as the Newcastle-Ottawa Scale as one input into that judgement.
No. COSMOS-E explicitly frames pooling as a decision requiring justification rather than an assumption, particularly when included studies differ materially in how they measured and adjusted for confounders.
PROSPERO is the most widely used registry for systematic review protocols, including etiological reviews, and registration before screening begins is consistent with the sequence recommended by COSMOS-E.
They address different stages: COSMOS-E guides the conduct of the review itself, whereas GRADE is applied subsequently to rate the certainty of the resulting body of evidence for a causal conclusion.
References
- Dekkers OM, Vandenbroucke JP, Cevallos M, Renehan AG, Altman DG, Egger M. COSMOS-E: Guidance on conducting systematic reviews and meta-analyses of observational studies of etiology. PLOS Medicine. 2019;16(2):e1002742. read 10 October 2026. View on PLOS Medicine
- Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. read 10 October 2026. View on The BMJ
- Higgins JPT, Thomas J, Chandler J, et al. (editors). Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane; 2023. read 10 October 2026. View the Cochrane Handbook
- Booth A, Clarke M, Dooley G, Ghersi D, Moher D, Petticrew M, Stewart L. The nuts and bolts of PROSPERO: an international prospective register of systematic reviews. Systematic Reviews. 2012;1:2. read 10 October 2026. View on Systematic Reviews
- von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement. The Lancet. 2007;370(9596):1453-1457. read 10 October 2026. View on The Lancet
- Wells GA, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. Ottawa Hospital Research Institute; 2021. read 10 October 2026. View the Newcastle-Ottawa Scale
- Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336(7650):924-926. read 10 October 2026. View on The BMJ
- Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Systematic Reviews. 2015;4:1. read 10 October 2026. View on Systematic Reviews
- Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109):629-634. read 10 October 2026. View on The BMJ
- Shea BJ, Reeves BC, Wells G, et al. AMSTAR 2: a critical appraisal tool for systematic reviews. BMJ. 2017;358:j4008. read 10 October 2026. View on The BMJ
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 10). COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology. Pubrica. https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/
Pubrica Academy. “COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology.” Pubrica, 10 Oct. 2026, pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/.
Pubrica Academy (2026) COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology. Pubrica. Available at: https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/ (Accessed: 15 October 2026).
Pubrica Academy. “COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology.” Pubrica. October 10, 2026. https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/
Pubrica Academy, ‘COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/> accessed 15 October 2026.
Pubrica Academy. COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/
Pubrica Academy, “COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/
@misc{/la_cosmos_e_systematic_reviews_etiology_2026,
author = {{Pubrica Academy}},
title = {COSMOS-E: Guidance for Systematic Reviews and Meta-Analyses of Observational Studies on Etiology},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/concepts-definitions/guidance-for-systematic-reviews-in-studies-on-etiology/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Etiology Review Protocol Starter
For use before registering the protocol for a systematic review of observational etiological evidence.
[ ] Exposure, outcome, population and comparison defined precisely enough to fix eligible study designs [ ] Search strategy checked against cohort/case-control indexing, not only RCT filters [ ] Likely confounders listed before extraction begins [ ] Protocol registered (e.g., PROSPERO) before screening starts [ ] Plan for assessing comparability of confounder adjustment across included studies [ ] Pre-specified criteria for whether pooling is appropriate, with planned sensitivity analyses
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica’s systematic review team applies the confounding-centred conduct framework of COSMOS-E alongside PRISMA reporting and PROSPERO registration in etiology reviews, so that the pooling decision and bias assessment are documented rather than assumed.