The most prominent result in a paper is easier to interpret when you can compare it with the question recorded before the results were known. Registration, protocols and analysis plans provide different pieces of that history. None should be treated as a ceremonial link at the end of the article.
- Match the outcome’s name, measurement method and time point.
- Check dates and versions, not only the latest registry page.
- A documented amendment can be legitimate; its timing and explanation matter.
Give the study a stable identity
Begin with the trial identifier, then match the title, interventions, population and investigators across records. One trial may generate a main paper, follow-up, subgroup report and methods paper. Four publications do not necessarily mean four independent participant groups.
For this guide, all documents and dates in the example are fictional. Imagine a programme trial registered on 1 January 2026 before the first participant enters on 15 January. Its primary outcome is a specified sleep score at week 12.
Keep that identifier and outcome on a reading card. They let you connect later reports without relying on similar titles or a press release’s description.
Match three parts of the outcome
WHO’s registration data set asks for an outcome’s name, measurement method and time point. Each part matters. “Sleep” at six weeks is not automatically the same endpoint as a particular sleep score at twelve weeks.
In our fictional record, write all three parts before opening the article’s results. Now suppose the publication highlights a fatigue score at week 6. That may be a valid secondary or amended outcome, but it is a different question from the one on your card.
Do not accuse the researchers based on a mismatch alone. First look for the protocol, amendments and analysis plan explaining how the endpoint acquired its role.
Source 1 ↗An amendment needs its own timeline
Add a fictional amendment dated 1 March that changes the primary outcome to fatigue at week 6. The key questions are why it changed, who approved it and what information was available when the decision was made. The date by itself does not tell us whether outcome results had been inspected.
A reason such as a measurement problem may be legitimate, but the reader needs the supporting account. Conversely, a late edit to a registry should not be assumed to reflect a decision made at the same late date if an earlier dated plan documents it.
Our method is to record the earliest inspectable evidence for each decision and leave unknown information marked unknown.
Find the results that lost prominence
Suppose the fictional July paper leads with fatigue and gives little space to the original sleep outcome. Look for the sleep result in the main tables, supplement, registry results or another linked report. An unfavourable result should not become invisible merely because another endpoint is more interesting.
Cochrane distinguishes missing evidence from an ordinary negative finding and explains how selective non-reporting can distort an evidence synthesis. A reader’s first contribution is to notice which planned results can and cannot be found.
If a source cannot be accessed, say “not available in the material checked.” That is more precise than claiming the outcome was never reported anywhere.
Source 2 ↗Read the analysis plan as a map, not a certificate
An analysis plan can specify the primary comparison, handling of missing data, subgroup analyses and testing order. Its existence does not guarantee that the final analysis followed it or that every choice was optimal.
For our fictional study, compare a few consequential decisions: the endpoint, time point, analysis population and treatment of missing scores. Note whether deviations are disclosed and whether additional analyses are clearly labelled exploratory. You do not need to reconstruct every line of statistical code to identify these differences.
This reading aid is not a formal audit or a validated bias score. It is a way to make the questions visible before an article turns one favourable result into the study’s entire story.
What a better evidence trail would look like
The next report from our fictional programme could link the dated protocol, show the amendment’s rationale, give the original and revised outcomes and explain which analyses were planned. A follow-up could add durability information while retaining the original trial identifier.
Open documentation would make the reasoning easier to inspect; it would not guarantee a positive treatment effect. A transparent null result can add more useful knowledge than a dramatic result whose selection history cannot be reconstructed.
End your reading card with two sentences: “The original question was…” and “The published claim is…” If they differ, explain the documented reason or identify the missing explanation. That creates a traceable comparison without turning uncertainty into an accusation.
A fictional document trail to investigate
| Document and date | What it says | What to check next |
|---|---|---|
| Registration: 1 January 2026 | Primary: specified sleep score at week 12 | Match the identifier and initial protocol |
| First enrolment: 15 January 2026 | Participants begin entering | Was the original plan in place beforehand? |
| Amendment: 1 March 2026 | Primary changed to fatigue at week 6 | Reason, approval and access to outcome data |
| Publication: July 2026 | Fatigue leads the results | Amendment disclosed? Original sleep result available? |
Entirely fictional timeline. A discrepancy prompts investigation; it is not evidence of wrongdoing. WHO supplies the outcome-field concepts, not these documents or dates.
Your questions, answered
Does registration prove that a trial is well designed?
No. It records information about the planned study and can improve transparency. The design, conduct, analysis and reporting still need appraisal. A registration number is a starting point, not a quality seal.
Is any change of primary outcome misconduct?
No. Changes can be justified. Inspect the reason, timing, documentation and information available to the decision-makers. The important issue is whether the final claim honestly represents that history.
Why compare old versions of a record?
The latest version may not show what was recorded before enrolment or before results were known. A dated history can help reconstruct changes, though it may still leave the reasons or data access unclear.
What if the original outcome is absent from the main paper?
Look in accessible supplements, registry results and linked reports. If it remains unavailable, state the scope of your search. Absence from one article is not proof that it was never reported anywhere.
Are a follow-up and a subgroup report independent studies?
They may report different analyses or time periods from the same participants. Track the trial identifier and population overlap so that multiple publications are not counted as independent replications.
What should a future report make easy to verify?
The planned question, dated amendments, primary analysis and results for the important original outcomes. Readers should be able to distinguish planned confirmation, changed plans and exploratory findings without guessing.
Limits of this interpretation
- This is a selected educational explanation, not a systematic review, validated appraisal instrument or personal care recommendation.
- Numerical examples are hypothetical. Their deliberately simplified assumptions must not be transferred to a real study without checking its methods.
- An AI source check can miss errors; source access and the absence of independent human review are stated explicitly.
Sources & transparency
- WHO: Trial Registration Data Set
Public HTML data-set descriptions checked by AI, including primary and secondary outcomes and study dates. No private registry or unpublished protocol was accessed. · Accessed 27 Sep 2026
- Cochrane Handbook, chapter 13: Assessing risk of bias due to missing evidence in a meta-analysis
Public HTML, selected discussion of selective non-reporting and missing evidence checked by AI. The document timeline in this guide is fictional. · Accessed 27 Sep 2026
Prepared and source-checked with AI on 27 September 2026. Press-news Team is the publication’s collective byline, not a claim of medical credentials or human review. No human editorial or clinical review has taken place. Source access is described below each reference. Worked examples are invented for education and do not report a clinical trial or predict an individual outcome.
Source check: AI source check — selected methods references and worked examples
Clinical review: No human editorial or clinical review
Suggest a correction