Part publication: When it is ethical and when it is not.

Many previous editorials in this journal have addressed ethical issues in publication and research.[1234] In this editorial, we examine an issue that has troubled the journal editors in recent years; the subject was also raised in eJournal Club India, a collaborative initiative from the Indian Psychiatric Society and the Department of Psychopharmacology at the National Institute of Mental Health and Neurosciences, Bengaluru.[5] The issue relates to part publication of data from a completed research study.

Consider the situation where a team of investigators conducts and completes a study. The investigators, for manifold reasons, wish to publish as many papers as possible from the data that were collected. It occurs to them that they could analyze the data from about half of the sample and publish the results in one journal, and then analyze the data from the entire sample and publish the results in another journal. Is this ethical?

The short answer is no, it is absolutely unethical to publish an analysis of data from only a part of the sample after a study is complete.

WHEN PART PUBLICATION IS UNETHICAL

Why is part publication unethical? One reason is that the purpose of communicating medical research findings is to advance the cause of science and to benefit patients. Publishing an analysis of data from only part of the sample could generate inconclusive or even misleading results because the analysis would be underpowered, or the subsample could be nonrepresentative. This could deceive a reader who does not see the second paper (or who sees the second paper much later) that presents the analysis of the entire dataset.

Additional deceptions are that it would seem that two studies have been conducted, and perhaps that two studies support the conclusions of the authors when, in reality, there is only one study. Finally, meta-analyses would count the two publications as separate studies, thereby duplicating patient data and inflating the contribution of that study to the meta-analysis.

WHEN PART PUBLICATION MAY BE PERMISSIBLE

There could indeed be situations when publishing part of the data is acceptable. These situations can arise when the study is ongoing, and there is a specific reason why a preliminary analysis, followed by the publication of the preliminary results, is deemed necessary. An example is to present pilot data that carry an important message or to present the results of the preliminary analysis of data from an adaptive study design where the results of the preliminary analysis will influence the methods or design of the rest of the study. Such a preliminary analysis is usually preplanned, stated in the study protocol, and approved by the Ethics Committee or the Institutional Review Board.

There is nothing unethical in conducting a study and presenting part of the results in one paper and the rest of the results in another paper as long as each paper is based on the full sample and each paper contains sufficient information for a standalone publication. For example, in a study of antipsychotic efficacy in schizophrenia, it could be reasonable to present psychopathology ratings, global ratings, assessments of quality of life, and assessments of adverse effects in one paper, and magnetic resonance imaging findings correlated with neuropsychological test battery results in another paper. The second paper should reference the first so that readers realize that both papers were based on the same sample.

SALAMI PUBLICATION

Publishing independent and unduplicated data from a study in multiple papers may sometimes be borderline unethical or outright unethical even when the analyses are based on the entire sample. As an example, authors might present the results of Positive and Negative Syndrome Scale and global ratings in one paper, the results of Brief Psychiatric Rating Scale and quality of life ratings in another paper, the results from half of the neuropsychological battery in a third paper, and so on. Salami publication makes it difficult for readers to locate all of the findings because they tend to be scattered across journals. Salami publications waste journal space. Salami publications are outright unethical when the authors do not cross-reference the different papers, thereby making it seem that the different papers are based on different studies.

REANALYSIS

It could be acceptable to reanalyze old, previously published data using a new method of analysis provided that there is sufficient justification for it and provided that the new analysis contributes something substantial that is different from the original analysis. Data can also be reanalyzed from different perspectives. For example, after the publication of the outcomes of a randomized controlled trial, the data can be reanalyzed to examine outcomes in patients with different initial severity of illness; in patients with different subtypes of illness; in patients who received different doses of the study drug; and so on. Again, readers should be informed that the results represent a reanalysis and not an original study; the original study should be referenced.

FINAL NOTES

It goes without saying that in all cases editors and reviewers need to be informed about prior publication of data from the study on which the submitted manuscript is based.