In a letter to the editor, Gray (Gray et al 2020) raised an issue regarding two publications in BMC Genriatics (Chan et al 2019 ChiCTR-IOR-14005490) and International Journal of Nursing Studies (Kao et al 2018 ChiCTR-IOR-17013812), which appeared to be reports of the same study but which reported different primary outcomes and had different trial registration numbers. A further paper reporting results from the trial citing registration ChiCTR-IOR-14005490 has also been published with a different primary outcome (Chiu et al 2018). The authors responded: “The integrated project was complicated, and the measurements of outcome indicators were difficult to digest within the confines of the space. Selective outcome reporting did not occur as it was pointed out in our previous work (IJNS) in the Chan et al. 2019 BMC Geriatric paper (page 3, Appendix doc. Number: 3) that this was a series study so the outcomes have all been reported in the two articles mentioned, not just a selection of the study” “Our two separate publications regarding different research questions and sub-studies but using the same original participant sample ideally should have used the original registration (trial identifier ChiCTR-IOR-14005490) but we were not able to transfer the registration elsewhere or edit the registry which is why we tried to rectify the situation with re-registration (trial identifier ChiCTR-IOR-17013812). There were no intentions of selective outcome reporting that would distort the evidence base.” The authors wish to clarify that all three papers draw on data from the integrated project with multiple different interventions. The primary outcome stated in both protocols was ‘cognitive function' composed of multiple domains. Distinct aims, different objectives, essential questions, and specific hypotheses were explored further in each separate paper. Therefore the description of these outcomes as primary outcomes has the potential to confuse, as does the presentation of power calculations in each paper, which implies that the sample size was determined specifically to detect pre-specified differences in the primary outcome identified in each paper. In fact, the power calculations presented in these papers were ‘post hoc' and were not used to determine the required sample size in advance. Editors' Note: Transparent reporting and adherence to prespecified protocols (or explicit description of deviation from protocols) is an important safeguard against selective outcome reporting and other practices that can mislead readers. Similarly, it is important that readers are able to properly identify when data in several papers arise from the same trial, because the outcomes reported are not fully independent. Post hoc power calculations should not be presented as if they were used to determine the sample size required. We are grateful to Gray and colleagues for alerting us to these papers and giving the authors the opportunity to clarify the relationship between these three studies. The authors assure us that there was no attempt to deliberately mislead readers, but the use of the term primary outcome in each paper and the lack of transparency caused by multiple registrations has that potential.
ASJC Scopus subject areas