When a corrigendum is not sufficient

Summary

We present the reasoning behind a retraction noting that even small, honest errors can result in significant changes in findings.

Research integrity is paramount in the pursuit and dissemination of medical knowledge. This requires the combined efforts of researchers, authors and all those involved in publishing such as editors and editorial boards, peer reviewers and research institutions and learned societies. There are many parties involved but ultimate responsibility rests with two groups. The editorial team must assess in detail the article before its acceptance, and the authors must verify their research findings and how they are presented in their article, to ensure accuracy prior to submission and then again when they review proofs prior to publication, to avoid/minimise any need for corrective actions. However, honest errors can occur.

Retraction versus corrigendum

When it comes to light that research integrity has been compromised by an article that has already been published, editors have to then consider what corrective action is most appropriate. For example, should the article now be retracted, or should it remain in the public domain and a corrigendum be published alongside?

The Committee on Publication Ethics (COPE) retraction guideline summary specifies two key points: (a) ‘Journal editors should consider retracting a publication if: they have clear evidence that the findings are unreliable, either as a result of misconduct (e.g., data fabrication) or honest error (e.g., miscalculation or experimental error)’ and (b) ‘Journal editors should consider issuing a correction if: a small portion of an otherwise reliable publication proves to be misleading (especially because of honest error)’.¹

The article by Davis and colleagues (Mental health in UK Biobank: development, implementation and results from an online questionnaire completed by 157 366 participants) posed a quandary for us as to which corrective action should be chosen.² The authors voluntarily submitted a corrigendum that acknowledged that a data analysis coding error for Alcohol Use Disorders Identification Test (AUDIT) resulted in decreased alcohol use disorders prevalence and a resultant decrease in total psychiatric disorder prevalence. Although this is a relatively minor coding error, it led to a major and important change in the findings. For example, the current published abstract refers to ‘Thirty-five per cent (55 750) of participants had at least one defined syndrome, of which lifetime depression was the most common at 24% (37 434)’ whereas with the corrected AUDIT coding this should be ‘Forty-five per cent (70 892) of participants had at least one defined syndrome, of which lifetime depression was the most common at 24% (37 434)’. This error is noted in the Results section and associated comorbidity tables.

The proposed corrigendum was reviewed by the Senior Editorial Committee with further communications with the authors. The authors were very responsive to all the questions put to them, which focused on their research oversight and were intended to ensure research integrity through the correction of any errors. Nonetheless, the key concern remained unchanged – publishing a corrigendum would leave the original paper to be cited with quite inaccurate data. Thus, to ensure research integrity, it has been decided that a retraction of the original article with subsequent publication of a corrected paper is the best course of action. The authors are aware and accept this editorial decision, and have been invited to submit a fully corrected version of the article for further peer review.

Conclusions

In sum, we feel the integrity of the publication processes we abide by in our journals has been upheld. We thank the authors for volunteering their honest error, as errors should be reported as soon as possible with correction of the scientific record, but, at the same time, note that even small errors can have significant effects on key findings that a corrigendum alone cannot correct, and may therefore not be sufficient – necessitating a full retraction. We feel that a retraction is neither punitive nor even an action of last resort, but rather the necessary and logical step required to maintain the quality and honesty of the scientific record on which further research is based, especially as the findings may influence policy and practice.

We continue to strive for research honesty and integrity in our journals and welcome comments from our readers to help us maintain the highest possible standards.