Non-prospective trial registration persisted despite editorial screening at The BMJ
A large observational analysis of clinical trials submitted to The BMJ shows how often registration problems survive beyond journal screening and still reach publication in high-profile venues. The cohort centered on 239 submissions from 2019 to 2023 that were flagged by editors as potentially not prospectively registered, paired with 239 prospectively registered trials drawn from the same submission period. The pattern was not subtle: some trial characteristics appeared to protect against non-prospective registration, while others were linked to higher odds of it. Even more striking, many of the trials rejected by The BMJ later appeared in journals with strong reputations and stated adherence to ICMJE registration rules.
Prospective registration is not a bureaucratic nicety. It is the mechanism that helps prevent outcome switching, selective reporting, and post hoc reshaping of a trial’s story after the results are known.
Trial features linked to lower odds of registration failure
Several submission-level characteristics were associated with reduced odds of non-prospective registration in an ICMJE-accepted registry. Larger trials were less likely to fall into the non-prospective category: trials with 101 to 500 participants had lower odds than those with 1 to 100 participants, with an odds ratio of 0.43 and a 95% confidence interval of 0.22 to 0.84. A larger author group also tracked with lower odds, with 10 authors compared with 5 associated with an odds ratio of 0.71 (0.59 to 0.87). The presence of Consolidated Standards of Reporting Trials mention was associated with still lower odds, with an odds ratio of 0.22 (0.06 to 0.67), suggesting that stronger reporting discipline may travel with better trial governance from the outset.
Funding mattered as well. Trials with funding, including non-profit support, were less likely to be non-prospectively registered than trials with no funding or no funding information; the odds ratio was 0.20 (0.09 to 0.41). More recent submissions from 2021 to 2023 also showed lower odds than those from 2019 to 2020, with an odds ratio of 0.63 (0.42 to 0.95), hinting at some improvement over time in submission practices or editorial awareness.
Geography entered the picture too. Corresponding authors based in Oceania had lower odds of non-prospective registration than those based in Europe, with an odds ratio of 0.35 (0.14 to 0.82). Authors based in Asia had higher odds than those in Europe, with an odds ratio of 1.75 (1.07 to 2.89). These findings do not establish causation, and they should not be read as markers of scientific quality. They do suggest that registration compliance remains uneven across regions, sponsor structures, and trial sizes.
Retrospective registration was common, and delays were long
Among the 176 trials not prospectively registered in an ICMJE-accepted registry and submitted between 2019 and 2021, retrospective registration was the dominant pattern. A total of 146 trials, or 83%, were eventually registered after the fact, with a median delay of 193 days. Twenty-three trials, or 13%, remained unregistered. Seven trials, or 4%, were registered in a registry that was not accepted by ICMJE.
Those delays matter because retrospective registration weakens the very safeguard trial registration is designed to provide. Once a trial has already begun, the registry entry can no longer function as a pre-specification of the study question, outcomes, and analysis plan in the way prospective registration does. In practice, a delayed entry can make it difficult for readers, peer reviewers, and editors to determine whether the reported endpoints were planned before data inspection or assembled later to fit the observed results.
Publication often followed rejection, frequently with no disclosure
Most of the non-prospectively registered trials did not disappear after rejection by The BMJ. Of the 176 trials in this subgroup, 155, or 88%, were later published. The downstream journals were often well established: 138 of those publications, or 89%, appeared in journals with an impact factor, with a median impact factor of 5.39. Even more concerning, 96 publications, or 62%, appeared in journals that claimed adherence to ICMJE recommendations.
The median time from the initial submission to The BMJ to eventual publication was 12 months. During that interval, the trial moved through a different editorial pipeline, and the registration deficiency frequently disappeared from view. Only about one sixth of the published papers explicitly disclosed the registration problem. In effect, the flaw was often visible to one journal but invisible to the next audience of clinicians, methodologists, and policy makers.
This is where the biology of trial credibility meets the mechanics of publication. Registration errors do not directly alter a molecule or a patient’s blood pressure, but they can alter what gets counted as success, what gets omitted from the final paper, and how much trust a clinician can place in a result when choosing a therapy or designing a guideline.
Authors often claimed prospective registration when it was absent
The study also tested how authors described their own trials on submission. Of 72 authors who responded about prospective registration at the time of submission, 60, or 83%, incorrectly claimed adherence. That gap between declaration and documentation is not a minor paperwork mismatch. It speaks to a persistent disconnect between what investigators believe they have done, what the registry shows, and what editors must verify before publication.
Such inaccuracies can arise from misunderstanding, poor administrative oversight, or deliberate misrepresentation; the study does not separate those possibilities. What it does show is that self-report cannot be treated as a reliable substitute for registry confirmation. Editorial systems that depend on author assertion alone will miss a substantial share of non-prospective registrations.
Editorial enforcement is doing some work, but the downstream system still leaks
The findings point to a familiar weak spot in trial publishing: one journal may refuse a trial on registration grounds, yet another may later publish it with little or no mention of the original deficiency. The result is a publication record that can look clean while preserving a hidden trail of protocol slippage. For clinicians reading the final paper, that omission can matter as much as the registration defect itself because it obscures how much confidence should be placed in the outcome hierarchy and the trial’s claims of rigor.
The analysis leaves a narrow and practical message. Registration enforcement at submission is necessary, but it is not sufficient when later journals publish the same trial without clear disclosure. Better registry checking, tighter inter-journal communication, and explicit reporting of retrospective registration would help close the gap between trial conduct and trial publication. Until those steps become routine, a trial can still move from rejected manuscript to polished publication without the reader ever being told how late the registration really was.