The STROBE statement is a reporting guideline for observational studies in epidemiology, covering the three main designs: cohort, case-control, and cross-sectional studies. STROBE stands for Strengthening the Reporting of Observational Studies in Epidemiology, and it provides a 22-item checklist describing the minimum information an author should report so that readers can assess the study’s conduct and generalisability. It governs how a study is described, not how it was designed or whether its result is valid.
Why observational research needed its own guideline
Randomised trials had CONSORT, but most published medical research is observational, and observational designs carry their own reporting pitfalls: vague case definitions, unstated eligibility windows, undefined exposure measurement, and silence about how confounding was handled. STROBE was first published in 2007 by a group of epidemiologists, methodologists, and journal editors to address exactly these gaps. The motivation parallels the one behind the CONSORT statement for randomised trials: without a complete description, a reader cannot tell whether an association is credible or an artefact of how the study was run. STROBE promotes transparent reporting; it deliberately stops short of recommending designs or judging quality.
What the 22-item checklist covers
The checklist follows the structure of a paper. The title and abstract item asks authors to name the design in the title or abstract so readers and database indexers can find it. The introduction items ask for a clear scientific background and a pre-specified set of objectives or hypotheses. The methods section is the heart of STROBE, with items on the study design, the setting and recruitment periods, the eligibility criteria and how participants were selected, the definitions of all outcomes, exposures, predictors, and confounders, the sources and measurement methods for each variable, efforts to address bias, the study size rationale, how quantitative variables were handled, and the full statistical methods including how confounding was controlled, how missing data were treated, and any sensitivity analyses.
The results items ask for a participant flow with numbers at each stage, descriptive data on participants and their exposures, the numbers of outcome events, and the main results reported as adjusted estimates with confidence intervals alongside the unadjusted figures, so a reader can see what the adjustment did. Reporting both the crude and adjusted estimates is one of the most useful STROBE requirements, because the gap between them reveals the influence of the confounders. The discussion items ask for key results, limitations including direction and likely size of any bias, cautious interpretation, and generalisability. Four items are design-specific, with separate guidance for cohort, case-control, and cross-sectional studies, since for example matching belongs to case-control work while loss to follow-up belongs to cohorts.
How to write a STROBE checklist for your study
In practice, completing a STROBE checklist is a four-step exercise. First, identify your design precisely, because the design-specific items differ. Second, work through each numbered item and record the page or line number in your manuscript where it is addressed, exactly as most journals now require on submission. Third, where an item is genuinely not applicable, state that rather than leaving it blank, so reviewers can see it was considered. Fourth, treat any item you cannot locate in your draft as a prompt to add the missing information, not as a box to leave unticked. Using the checklist while drafting, rather than after, makes it a writing scaffold for the structure of a research write-up rather than a last-minute audit.
STROBE and ROBINS-I: reporting versus bias
STROBE is frequently confused with bias assessment, but the two are complementary rather than interchangeable. STROBE asks whether the study is completely reported. A bias tool asks whether the study’s result is likely to be systematically wrong. The relevant instrument for non-randomised studies of interventions is the ROBINS-I tool for non-randomised studies, which evaluates domains such as bias due to confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result. The connection is practical: a STROBE-compliant paper that clearly states its confounders and adjustment strategy gives a ROBINS-I assessor the raw material to judge the confounding domain. A poorly reported paper forces the assessor to rate items as “no information”, which usually pushes the judgement toward higher risk. The general distinction between reporting completeness and systematic error is set out in our piece on how bias and quality assessment differ.
For reviewers, STROBE compliance in the source papers is what makes observational evidence usable. Extracting an adjusted hazard ratio or odds ratio with its interval, and knowing exactly which confounders it was adjusted for, is essential for deciding whether estimates can be pooled at all; differing adjustment sets are a major source of between-study heterogeneity. Where the reporting is complete, those estimates feed the quantitative synthesis of a review cleanly.
How to cite the STROBE statement
When you state that a study was reported in accordance with STROBE, cite the original guideline papers rather than a website alone. The primary reference is the von Elm and colleagues 2007 statement, published simultaneously in several journals, and the companion explanation and elaboration paper by Vandenbroucke and colleagues, which gives the rationale and worked examples for each item and is the better citation when you want to show you understood the guidance, not merely named it. A typical methods sentence reads that the study is reported in line with the STROBE statement, with the completed checklist supplied as a supplementary file. Citing the explanation paper signals genuine use, in the same spirit as citing the methods behind the GRADE certainty-of-evidence approach rather than name-dropping it.
Limitations and common mistakes
STROBE has clear boundaries. It is a reporting tool, so satisfying it does not make a weak study strong; a perfectly reported study with uncontrolled confounding is still confounded. It is also not a quality scale, so summing the items into a score is a misuse, the same error people make with CONSORT. Three mistakes recur in submitted observational work. The first is failing to name the design in the abstract, which buries the study and complicates retrieval during a systematic search. The second is reporting only the adjusted estimate without the crude one, hiding the effect of the adjustment. The third is vague confounder reporting, where the paper says it “adjusted for known confounders” without listing them, which makes the result impossible to compare across studies and undermines any attempt at synthesis. Extensions exist for specific contexts, such as molecular epidemiology and genetic association studies, so authors in those fields should follow the matching variant. Used honestly, STROBE makes observational research auditable, which is the precondition for it contributing to the wider family of evidence-synthesis review types.