A result can be numerically correct while its place in the original research plan remains unclear. Read the plan and the paper together.
- Match the measure, timing and primary or secondary status.
- Look for a dated explanation when an outcome changes.
- A reporting discrepancy is a reason to investigate, not proof of invented data.
- Study unit
- Published trial reports
- Sample
- 67 trials; five journals
- Period sampled
- 19 October–30 November 2015
- Report published
- 2019
Original publication: 14 Feb 2019 · The date above refers to this brief.
Reports with discrepancies under the project’s rules
Number of trial reports; 67 in total
The document trail
COMPare checked trial reports against protocols or registry histories dated before recruitment. Its coders accepted disclosed outcome changes as correctly reported. Of 58 submitted correction letters, 23 were published. The project examined both reporting and the response to attempted correction; its short, selected journal sample is not a current prevalence estimate for all medicine.
Source 1 ↗What counts as a meaningful change?
The historical CONSORT guidance used in this project asked authors to define primary and secondary outcomes, including timing, and explain changes after a trial began. A changed endpoint can have a defensible reason. Knowing what changed and when lets readers judge the claim. A registry’s latest version alone may hide the sequence of amendments.
Source 2 ↗Source 1 ↗An original outcome-matching exercise
Imagine an invented protocol that names hospital admission by day 90 as primary. The paper highlights a symptom scale at day 30. Make two rows, one for each measure. Record whether the original outcome appears anywhere, whether the new one was planned, and whether the change has a dated explanation. Do not label the trial invalid merely because the abstract emphasises something different. First identify the missing document or passage.
Why the reply is part of the evidence
A companion qualitative study examined responses from 20 research teams. That is a selected set of respondents, shaped by which letters journals published. It offers context about disputes over reporting, not an independent prevalence survey or a psychological diagnosis of authors. When reading a correction exchange, preserve the question, response and final documented change.
Source 3 ↗What would a better reporting system make visible?
Our proposed improvement is a short outcome map beside each article: planned measure, reported measure, amendment date and explanation. Its value would need testing. A future evaluation could measure unresolved discrepancies and reader comprehension separately. Simply publishing more forms would not itself demonstrate better understanding.
A practical document check
| Evidence | What it contributes | What remains open |
|---|---|---|
| Pre-trial plan | Defines the original question | Was the available version dated before recruitment? |
| Amendment log | Explains a change in the plan | Was the timing and reason recorded? |
| Results paper | Shows what was analysed and reported | Can each planned outcome be located? |
| Correction or response | Adds a later clarification | What actually changed in the record? |
Original reading worksheet informed by the cited historical reporting standard; it is not a formal risk-of-bias assessment.
Your questions, answered
Is every changed outcome a problem?
No. The critical questions are whether the change is documented and why it happened. A disclosed amendment and an unexplained substitution are different situations.
Does a discrepancy mean fraud?
No. The project’s reporting classification should not be converted into an allegation about intent. Check documents before drawing conclusions.
Where should a reader begin?
Start with the registration number in the paper. Locate the original entry and amendments, then match the exact outcome and time point.
Is a correction letter a new clinical trial?
No. It may clarify how an existing trial was reported. Keep it attached to that trial rather than counting it as an independent set of participants.
Limits of this interpretation
- Selected high-impact journals and a six-week sampling window.
- This article does not independently adjudicate the individual trial disputes.
- The historical CONSORT source is included to explain the project’s rules, not to replace current reporting guidance.
Sources & transparency
- Goldacre et al. (2019): COMPare prospective study of outcome reporting
Public full-text HTML: sample, coding rules, results and limitations. Individual trial coding sheets were not independently recoded. · Accessed 27 Sep 2026
DOI: 10.1186/s13063-019-3173-2 - CONSORT 2010 explanation: defining outcomes and explaining changes
Public HTML, items 6a and 6b. Historical reporting standard used by COMPare, not a claim that this is the latest checklist. · Accessed 27 Sep 2026
DOI: 10.1136/bmj.c869 - COMPare (2019): qualitative analysis of trialists’ responses
Indexed full-text passages: abstract, sample and limitations. Correspondence files were not reanalysed. · Accessed 27 Sep 2026
Prepared with AI using the source passages described below. No human editorial or clinical review has been completed. Reading exercises and future research questions are our interpretation, not additional study findings. This is an educational case study about research reporting, not a treatment recommendation.
Source check: AI source check — primary-source passages and reported numbers; 27 September 2026
Clinical review: Not applicable to this educational guide
Suggest a correction