Transparent Reporting at The Journal of Infectious Diseases

(See the Major Article by Zavalis et al on pages 227–34.)

Zavalis and colleagues [1] assessed transparency in reporting in the 9 most-cited infectious diseases journals, using PubMed Central Open Access articles published in 2019 or 2021 (a total of 5340 articles). The text-mining R package, rtransparent was used for the analyses. Features evaluated included code sharing, data sharing, study registration, conflict of interest (COI) disclosures, and funding disclosures. The authors report that rates of code sharing, data sharing, and study registration, in particular, were found to be low (each reported overall for <10% of evaluated articles). In contrast, COI and funding disclosures were identified for the majority of the articles (79% and 91% of articles, respectively). However, all evaluated measures differed substantively between journals (8%–93% differential reporting between journals).

Overall, the editors of The Journal of Infectious Diseases (JID) applaud the article by Zavalis et al [1], presenting their thoughtful and systematic reporting and discussion of the transparency outcomes evaluated. We note that the same criteria were applied to each journal evaluated, although each has different publication visions and goals. As discussed in a recent editorial [2], JID focuses on investigative, experimental, and translational, science—that is, from “bench to bedside.” Publication of the article by Zavalis and colleagues [1] affords us the opportunity to review how JID performed by these metrics across the 383 articles published during the study period (7% of all articles evaluated), to put these results into context for our readership, and to discuss “action items” that are under consideration by the editors in response to the findings.

Consistent with the overall results, COI and funding disclosures were reported for 99% and 97% of JID articles, whereas code sharing, data sharing, and registration were reported infrequently (in 3%, 5%, and 12% of articles, respectively). As described in our latest instructions to authors, posted on the JID website, we now offer initial format-free submissions that require a funding disclosure and a statement of data availability at early stages of article evaluation. If the submitted article moves toward publication, we require a COI disclosure by each author using the ICMJE form, and article production does not proceed until the COI requirement is met. At this stage in publication, JID now requires links to all analytical code and publicly available data. Specific sections within the JID instructions to authors provide guidance to reporting funding sources and COI. Thus, even before publication of the analyses by Zavalis et al, JID had strengthened our code and data sharing requirements.

One area of uncertainty in those analyses surrounds the metric of study registration, which was reported in approximately 8% of articles [1]. While this percentage may seem low, Zavalis et al explain that only clinical trials involving human subjects currently require study registration (eg, in ClinicalTrials.gov or similar registries in other countries). In their manually curated data set, the authors found that registration was performed in 100% of clinical trials and 31% of systematic reviews but in <2% of all other types of studies. Thus, journals that publish relatively few clinical trials (especially those like JID that publish a large number of translational and laboratory-focused studies) would be expected to report lower rates of study registration. While Zavalis et al suggest that even nonclinical trials consider prospective registration, it is not clear that the rigidity required of clinical trials analysis is practical or desirable for cohort, retrospective, or translational studies. Additional discussion will be needed as to whether the benefits outweigh the burdens of this proposed approach.

The article by Zavalis et al [1] should be interpreted with some additional limitations in mind. First, the authors validated their data manually and reported a false-negative rate with their automated software, including missing COI disclosures in 3% of articles and funding disclosures in 6%. They attempted to adjust for the manual validation results in the corrected estimates. With respect to code sharing, their report does not provide a definitive answer as to the percentage of studies that should have reported code but did not. A low rate of code sharing in a journal could reflect a relatively low rate of published articles with a computational component.

In response to the findings of Zavalis et al, we are taking a series of steps. We are now updating on an ongoing basis sections of our instructions to authors that appear to be unclear to authors. We will shortly add a section that specifically focuses on guidance for code reporting and data sharing, as relevant, and that will require specific statements within published articles on each of these topics. The Infectious Diseases Society of America (IDSA) has meetings scheduled for mid-2023 to begin to plan for and to proactively and coordinately respond in all IDSA journals (JID, Clinical Infectious Diseases, Open Forum Infectious Diseases) to the White House Office of Science and Technology statement on transparency in scientific reporting. The JID editors are sensitive to the increasing demands on our investigators and, at this time, will encourage authors to consider registration of human subject investigations (when feasible or practical) and systematic reviews/meta-analyses (eg, in the PROSPERO registry). We will continue to monitor emerging reporting guidelines and revise our approach as new data emerge. We welcome the input of our readership and authors on the important topic of transparency in scientific reporting.

References

Author notes