New Phillips-Burstyn-Carter working paper on the failure of peer review in public health

by Carl V Phillips

Our new working paper is available at EP-ology, “The limited value of journal peer review in public health: a case series of tobacco harm reduction articles”, Carl V Phillips, Igor Burstyn, and Brian L Carter (all of the authors are affiliated with CASAA, for those who may not know).

Abstract:

Background: A widespread belief holds that the journal peer-review process has magical powers to ensure that published claims are correct. While this misperception has limited consequences in many fields, in public health it results in consumer, clinical, and policy decisions being based on blind faith in the accuracy of published claims. At best, the review process is merely a couple of readers — perhaps, but not necessarily, highly expert — reading through a paper to ensure the research and presentation are reasonably sound. In reality, even this is often not accomplished.

Methods: We conducted reviews of 12 articles that focused on tobacco harm reduction published in a mainstream public health journal, BMC Public Health, consecutively during 2012-15. We each wrote a reviewer report of the manuscript version that was sent to the journal reviewers, as if we were writing a review for a journal. We then compared these to the reviews written by the journal reviewers. Additionally, we reviewed the changes made to the papers as a result of the journal reviews.

Results: Almost all the papers in the dataset suffered from major flaws, most of which could have been corrected, but none were corrected by the journal review process. The journal peer reviews were almost all inadequate and many contained no substantive comments. Those that contained substantive observations still did not identify most of the blatant major flaws that we noted. In the single case where a journal reviewer identified many of the major flaws, the comments were basically ignored by the authors and the paper was published with no substantive changes. Other than cosmetic improvements, the journal review process was about as likely to make the published version worse than the submitted manuscript, rather than better. Papers with no apparent value were published by the journal and the potential value of other studies was lost because serious flaws in the paper were ignored. Unreported conflict of interest was common among both authors and reviewers.

Conclusions: Faith in the journal peer-review process is misplaced. Even at best, the process cannot promise that a published claim is correct, but in reality it does not even ensure that patent major flaws are not present. In public health, the phrase “according to a peer-reviewed journal article” seems to mean little more than “I read this somewhere.”

It should be evident from the abstract that the primary study aim is not about THR. However, readers of this blog may be interested for several reasons. Most obviously, the case studies are based on articles about THR. But also, the idolatry of journal peer review is one of the more important causes of the persistence of anti-THR lies. Analyzing the reviews of the papers, not the papers themselves, was the purpose of the research, but that required assessing the papers en passant, which means that readers interested in that aspect will should find a fair bit of the content interesting (particularly delving into a few of the appendices).

Some of the material has already been covered here. The previous post was basically written as a footnote for the paper. The post about the terrible paper by Hughes (“Associations between e-cigarette access and smoking and drinking behaviours in teenagers”, by Karen Hughes, Mark A Bellis, Katherine A Hardcastle, Philip McHale, Andrew Bennett, Robin Ireland, and Kate Pike) was basically an excerpt from that research project. The extensive analysis of the Popova-Ling travesty was incorporated as part of the analysis in the paper.

There are few other papers in the analyzed case series that also fall solidly within the anti-THR lies category. There is this one (“Portrayal of electronic cigarettes on YouTube”, Chuan Luo, Xiaolong Zheng, Daniel Dajun Zeng, Scott Leischow) whose value lies entirely in it being unintentional comedy. Strangely, despite being a useless, silly, and badly conducted study, that was then written up as a political broadside that had nothing to do with the study results, it was probably only the 10th worst of the 12 papers in our case series.

Two of the papers were written to try to vilify snus. This one (“Snus user identity and addiction. A Swedish focus group study on adolescents”, Ingrid Edvardsson, Margareta Troein, Göran Ejlertsson, Lena Lendahls) is mostly just uninteresting in terms of what it claims. This one (“Predictors of smoking among Swedish adolescents” Junia Joffer, Gunilla Burell, Erik Bergström, Hans Stenlund, Linda Sjörs, Lars Jerdén) relates more closely to recurring themes from this blog. In particular, it makes the naive gateway claim that fails to distinguish between association and causation. This is particularly pathetic in this case because the paper is ostensibly about predictors and not causes.

Comments on the working paper are welcome, either in the blog comments at EP-ology or via email.