CONSORT, PRISMA, STROBE: Which Checklist and When

A paper reports that a new fixation technique produced better outcomes. It does not say how patients were allocated, how many were lost to follow-up, whether the outcome assessor knew which group they were in, or how the sample size was chosen.

The study might be excellent. There is no way to tell — and a study whose conduct cannot be assessed cannot be used, which means the work was wasted regardless of how well it was done.

Reporting guidelines exist to prevent that. They are widely treated as a box-ticking exercise completed the night before submission, which is precisely the use that extracts no value from them. This post covers which applies to what, and why the timing matters more than the checklist. It sits within our research integrity series.

Which guideline for which design

The EQUATOR Network maintains the definitive library of these and their extensions. When you are unsure, that is where to look — and it is also where to check whether the version you are using is current, which matters more than people assume.

Use the current version — they do change

Reporting guidelines are periodically revised, and using a superseded version is a small but avoidable error.

CONSORT 2025 replaced CONSORT 2010, published on 28–29 April 2025 simultaneously across JAMA, The Lancet, The BMJ, Nature Medicine and PLoS Medicine. It comprises a 30-item checklist plus the participant flow diagram. Seven new items were added, three revised, one deleted, several items from key extensions integrated, and the checklist restructured with a new section on open science.

PRISMA 2020 replaced the 2009 statement, with a 27-item checklist, an expanded version, a separate checklist for abstracts, and flow diagrams for both new and updated reviews.

STROBE remains the standard for observational studies.

Accurate as of July 2026. Guidelines are revised periodically — check EQUATOR for the current version before you start, not after.

That CONSORT update is worth noticing for a reason beyond version numbers: the addition of an open-science section signals where expectations are heading. Registration, protocol availability, data sharing and code availability are becoming reporting requirements rather than virtues.

The three you will actually use

CONSORT — randomised trials

The flow diagram is the part that matters most and the part most often done badly. It must account for every patient: assessed for eligibility, excluded and why, randomised, allocated, lost to follow-up, analysed.

The numbers must reconcile. A diagram where randomised does not equal analysed plus lost is the commonest and most revealing error in surgical trial reporting, because the discrepancy is usually where the attrition bias lives.

Other items surgeons routinely under-report: how the allocation sequence was generated, how allocation was concealed, who was blinded (in surgical trials often only the assessor — say so explicitly rather than writing "blinded"), and how the sample size was determined.

STROBE — observational studies

The guideline most orthopaedic authors need, because most of what we publish is a retrospective series.

The items most often missed: how participants were selected and why, how confounders were addressed, how missing data were handled, and — the big one — an honest limitations section that names the biases rather than gesturing at them.

STROBE will not make a retrospective series into a trial. It will make it interpretable, which is all anyone can ask of it.

PRISMA — systematic reviews

The flow diagram again does the heavy lifting: records identified, duplicates removed, screened, excluded with reasons, full texts assessed, studies included.

Also expected: the full search strategy for at least one database reproduced verbatim, the protocol registration (PROSPERO or equivalent), how risk of bias was assessed, and how certainty of evidence was rated.

A systematic review without a registered protocol and a reproducible search is a literature review with a more impressive name.

The timing point, which is the whole argument

Open the checklist before you design the study, not before you submit it.

Used at the end, a reporting guideline can only reveal what you failed to record — and by then it is too late. Used at the start, it functions as a design specification: it tells you what you will need to report, which tells you what you need to capture.

Concretely, reading CONSORT before starting a trial prompts you to decide in advance how the allocation sequence will be generated, who will be blinded, and how you will handle protocol deviations. Reading STROBE before a retrospective series prompts you to record the eligibility criteria you actually applied, rather than reconstructing them a year later.

The same logic applies to SQUIRE for quality improvement work — reconstructing context and rationale after the fact is much harder than recording it as you go.

This is also the cheapest guard against selective outcome reporting there is. A checklist that requires you to report all pre-specified outcomes makes it awkward to drop, without comment, the six that did not reach significance.

What journals actually require

Most reputable journals now require the relevant checklist at submission, with page or line numbers indicating where each item is addressed. Many require the flow diagram as a figure. Some check compliance at peer review; some do not.

Two practical notes:

Completing the checklist honestly is more useful than completing it completely. If an item does not apply, say so and why. If you did not do it, do not claim you did — reviewers check, and a checklist claiming blinding that the methods section does not describe is worse than no checklist.

A checklist is not a quality guarantee. It certifies that things were reported, not that they were done well. A thoroughly reported underpowered retrospective series is still an underpowered retrospective series. Reporting completeness and methodological quality are different axes, and conflating them is a common misreading.

The recurring orthopaedic errors

• Calling a retrospective series a cohort study and reporting it as though it were prospective.
• A flow diagram whose numbers do not reconcile.
• "Patients were blinded" in a trial where blinding a surgical patient was impossible. Say who was actually blinded.
• Loss to follow-up unexplained — a real problem in arthroplasty studies with long follow-up, and the reader cannot judge the result without it.
• No sample size justification, or one calculated retrospectively.
• Outcome measure switched between protocol and paper, without acknowledgement — see PROMs Explained.
• Limitations written defensively, listing trivial issues while omitting the design's actual weakness.

That last one is worth dwelling on. A limitations section that names the real problem — "this is a single-surgeon retrospective series with 22% loss to follow-up, and the comparison group differed systematically in age" — builds far more confidence than one that says "further research is needed."

Practical steps

1. Identify your design honestly, before anything else.
2. Find the current guideline on EQUATOR.
3. Read the checklist before designing the study.
4. Build the data collection around it, so you capture what you will need.
5. Draft to the checklist, using its items as a structure.
6. Complete it honestly at submission, with line numbers.
7. Include the flow diagram where the guideline has one, and check the arithmetic.

The summary

CONSORT for trials — now the 2025 version, 30 items and an open-science section. STROBE for observational work, which is most of what orthopaedics publishes. PRISMA 2020 for systematic reviews. EQUATOR when you are unsure.

Open the checklist at the protocol stage, not at submission. Used at the end it is paperwork; used at the beginning it is a design specification that prevents the omissions which make otherwise sound work unusable.