ICMJE on data sharing/ Not so random RCTs? Positive results bias/ What’s next for peer review? Ethics of predatory publishing/ Is the Impact Factor stochastic?

DATA SHARING

ICMJE statement on data sharing, published June 5, 2017, in the ICMJE journals:

“1. As of July 1, 2018 manuscripts submitted to ICMJE journals that report the results of clinical trials must contain a data sharing statement as described below

2. Clinical trials that begin enrolling participants on or after January 1, 2019 must include a data sharing plan in the trial’s registration…If the data sharing plan changes after registration this should be reflected in the statement submitted and published with the manuscript, and updated in the registry record. Data sharing statements must indicate the following: whether individual deidentified participant data (including data dictionaries) will be shared; what data in particular will be shared; whether additional, related documents will be available (e.g., study protocol, statistical analysis plan, etc.); when the data will become available and for how long; by what access criteria data will be shared (including with whom, for what types of analyses and by what mechanism)…Sharing clinical trial data is one step in the process articulated by the World Health Organization (WHO) and other professional organizations as best practice for clinical trials: universal prospective registration; public disclosure of results from all clinical trials (including through journal publication); and data sharing.”

Taichman DB, Sahni P, Pinborg A, Peiperl L, Laine C, James A, et al. Data Sharing Statements for Clinical Trials: A Requirement of the International Committee of Medical Journal Editors. PLOS Med. 2017.14(6): e1002315. https://doi.org/10.1371/journal.pmed.1002315

 

RESEARCH REPRODUCIBILITY AND MISCONDUCT

• Not so random?

Randomization in an RCT confers an advantage over other study designs because random sampling means that any differences in variables between comparison groups occur at random (rather than due to confounding). However, some researchers have identified RCTs that do not appear to have been randomly sampled–a clue that the methodology may have been different from what authors are reporting.

Carlisle “analysed the distribution of 72,261 means of 29,789 variables in 5087 randomised, controlled trials published in eight journals between January 2000 and December 2015…Some p values were so extreme that the baseline data could not be correct: for instance, for 43/5015 unretracted trials the probability was less than 1 in 1015 (equivalent to one drop of water in 20,000 Olympic-sized swimming pools).”

Carlisle JB.  Data fabrication and other reasons for non-random sampling in 5087 randomised, controlled trials in anaesthetic and general medical journals , Anaesthesia, 2017.72: 944–952. doi:10.1111/anae.13938

• In another study, Carlisle et al applied the same approach and concluded that “The Monte Carlo analysis may be an appropriate screening tool to check for non-random (i.e. unreliable) data in randomised controlled trials submitted to journals.”

Carlisle JB, Dexter F, Pandit JJ, Shafer SL, Yentis SM. Calculating the probability of random sampling for continuous variables in submitted or published randomised controlled trials. Anaesthesia, 2015.70: 848–858. doi:10.1111/anae.13126

• Bolland et al used Carlisle’s method to analyze RCTs published by a group of investigators “about which concerns have been raised” and found:

“Treatment groups were improbably similar. The distribution of p values for differences in baseline characteristics differed markedly from the expected uniform distribution (p 5 5.2 3 10282). The distribution of standardized sample means for baseline continuous variables and the differences between participant numbers in randomized groups also differed markedly from the expected distributions (p 5 4.3 3 1024, p 5 1.5 3 1025, respectively).”

Mark J. Bolland, Alison Avenell, Greg D. Gamble, and Andrew Grey
Systematic review and statistical analysis of the integrity of 33 randomized controlled trials. Neurology 2016 : WNL.0000000000003387v1-10.1212/WNL.0000000000003387.

• Is this approach yet another type of manuscript review for busy editors to apply, assuming the calculations are not too daunting? In Retraction Watch, Oransky comments, “So should all journals use the method — which is freely available online — to screen papers? In their editorial accompanying Carlisle’s paper, Loadsman and McCulloch note that if that were to become the case, ‘…dishonest authors could employ techniques to produce data that would avoid detection. We believe this would be quite easy to achieve although, for obvious reasons, we prefer not to describe the likely methodology here.’ Which begs the question: what should institutions’ responsibilities be in all this?

From: Two in 100 clinical trials in eight major journals likely contain inaccurate data: Study (Retraction Watch)

• In other news, PubPeer announces PubPeer 2.0. From Retraction Watch: “RW: Will the identity changes you’ve installed make it more difficult for scientists to unmask (and thereby seek recourse from) anonymous commenters? BS: Yes, that is one of the main motivations for that change. Once the transition to the new site is complete our goal is to not be able to reveal any user information if we receive another subpoena or if the site is hacked.”

Meet PubPeer 2.0: New version of post-publication peer review site launches today (Retraction Watch)

• Retraction Watch announces the DiRT (Do the Right Thing) prize and its first awardee

 

RESEARCH BIAS

Addressing bias toward positive results

• “The good news is that the scientific community seems increasingly focused on solutions…But true success will require a change in the culture of science. As long as the academic environment has incentives for scientists to work in silos and hoard their data, transparency will be impossible. As long as the public demands a constant stream of significant results, researchers will consciously or subconsciously push their experiments to achieve those findings, valid or not. As long as the media hypes new findings instead of approaching them with the proper skepticism, placing them in context with what has come before, everyone will be nudged toward results that are not reproducible…For years, financial conflicts of interest have been properly identified as biasing research in improper ways. Other conflicts of interest exist, though, and they are just as powerful — if not more so — in influencing the work of scientists across the country and around the globe. We are making progress in making science better, but we’ve still got a long way to go.”

Carroll AE.  Science Needs a Solution for the Temptation of Positive Results (NY Times)

• But replication leads to a different bias, says Strack: “In contrast, what is informative for replications? Not that the original finding has been replicated, but that it has been ‘overturned.’ Even if the editors’ bias (Gertler, 2016) bias [sic] is controlled by preregistration, overturned findings are more likely to attract readers’ attention and to get cited…However, there is an important difference between these two biases in that a positive effect can only be obtained by increasing the systematic variance and/or decreasing the error variance. Typically, this requires experience with the subject matter and some effort in controlling unwanted influences, while this may also create some undesired biases. In contrast, to overturn the original result, it is sufficient to decrease the systematic variance and to increase the error. In other words, it is easier to be successful at non-replications while it takes expertise and diligence to generate a new result in a reliable fashion..”

Track F.  From Data to Truth in Psychological Science. A Personal Perspective. Front Psychol, 16 May 2017 | https://doi.org/10.3389/fpsyg.2017.00702

 

PEER REVIEW

What’s next for peer review?

From the London School of Economics blog, reproduced from “SpotOn Report: What might peer review look like in 2030?” from BioMed Central and Digital Science:

“To square the [peer reviewer] incentives ledger, we need to look to institutions, world ranking bodies and funders. These parties hold either the purse strings or the decision-making power to influence the actions of researchers. So how can these players more formally recognise review to bring balance back to the system and what tools do they need to do it?

Institutions: Quite simply, institutions could give greater weight to peer review contributions in funding distribution and career advancement decisions. If there was a clear understanding that being an active peer reviewer would help further your research career, then experts would put a greater emphasis on their reviewing habits and research would benefit.

Funders: If funders factored in peer review contributions and performance when determining funding recipients, then institutions and individuals would have greater reason to contribute to the peer review process.

World ranking bodies: Like researchers, institutions also care about their standing and esteem on the world stage. If world ranking bodies such as THE World University Rankings and QS World Rankings gave proportionate weighting to the peer review contributions and performance of institutions, then institutions would have greater reason to reward the individuals tasked with peer reviewing.

More formal weighting for peer review contributions also makes sense, because peer review is actually a great measure of one’s expertise and standing in the field. Being asked to peer review is external validation that academic editors deem a researcher equipped to scrutinise and make recommendations on the latest research findings.

Researchers: Researchers will do what they have to in order to advance their careers and secure funding. If institutions and funders make it clear that peer review is a pathway to progression, tenure and funding, researchers will make reviewing a priority.

Tools In order for peer review to be formally acknowledged, benchmarks are necessary. There needs to be a clear understanding of the norms of peer review output and quality across the myriad research disciplines in order to assign any relative weighting to an individual’s review record. This is where the research enterprise can utilise the new data tools available to track, verify and report all the different kinds of peer review contributions. These tools already exist and researchers are using them. It’s time the institutions that rely on peer review got on board too.”

Formal recognition for peer review will propel research forward (London School of Economics)

PREDATORY/PSEUDO-JOURNALS

Biochemia Medica published a cluster of papers on predatory journals this month, including research by Stojanovski and Ana Marusic on 44 Croatian open access journals, which concludes: “In order to clearly differentiate themselves from predatory journals, it is not enough for journals from small research communities to operate on non-commercial bases…[they must also have] transparent editorial policies.” The issue also include ethical issues of predatory publishing (for which I am a coauthor, by way of disclosure) and an essay by Jeffrey Beall.

IMPACT FACTOR

“…more productive years yield higher-cited papers because they have more chances to draw a large value. This suggests that citation counts, and the rewards that have come to be associated with them, may be more stochastic [randomly determined] than previously appreciated.”

Michalska-Smith MJ, Allesina S. And, not or: Quality, quantity in scientific publishing. PLOS ONE. 2017.12(6): e0178074. https://doi.org/10.1371/journal.pone.0178074

 

ACCESS

• The American Psychological Association raised the ire of some authors after requesting that links to free copies of APA-published articles (“unauthorized online postings”) from authors’ websites be removed.

Researchers protest publisher’s orders to remove papers from their websites (Retraction Watch)

• Access challenges in a mobile world 

Bianca Kramer at the University of Utrecht in the Netherlands studied Sci-Hub usage data attributed to her institution and compared it with holdings data at her library. She found that “75% of Utrecht Sci-Hub downloads would have been available either through our library subscriptions (60%) or as Gold Open Access/free from publisher (15%).” While these data are not comprehensive, nor granular enough for certainty, she concluded that a significant component of usage of Sci-Hub was caused by problems of access and the desire for convenience by users.

Failure to Deliver: Reaching Users in an Increasingly Mobile World (Scholarly Kitchen)

• Resource: How to access research literature from developing countries (from INASP), for librarians

__

Newsletter #11: Originally circulated June 18, 2017. Sources of links include Retraction Watch, Health Information for All listserve, Scholarly Kitchen, Twitter. Providing the links does not imply WAME’s endorsement.