The Editors' Notepad
The goal of this blog is to help EPIDEMIOLOGY authors produce papers that clearly and effectively communicate their science.

Wednesday, September 21, 2011

Epidemiologist, Heal Thyself - A guest blog by Jay Kaufman
Scientists often portray themselves as the noble but hapless victims of sensationalism and exaggeration in the popular media [1]. But are scientists in fact sometimes complicit in these abuses, hyping their work in media interviews, making claims that would not survive peer review in the published articles? If so, this constitutes an important ethical violation that deserves further scrutiny, since communication with the public is at least as socially consequential as communication between scientists. Public opinion plays a long-term role in funding levels for competing research programs, for example, which makes exaggeration in news stories a serious abuse of the power granted to the scientist by a credulous and trusting media and public.
Here’s one example I came across recently which may fit this description. Nature Genetics published a meta-analysis by Dara Torgerson and colleagues in their September issue. The authors pooled North American genome-wide association studies of asthma that included over five thousand cases, including individuals of European, African and Latino ancestry [2]. They reported a number of susceptibility loci, most of which showed similar associations across ethnic populations and had been previously described. But one variant was novel, and the association was described as being specific to individuals of African descent. Table 2 of the paper reported a SNP near the gene PYHIN1 on chromosome 1 with an odds ratio (OR) among African Americans and Afro-Caribbeans of 1.34 (95% CI: 1.19-1.49). In a replication data set, this association remained substantial (OR=1.23), although at a slightly different locus.  For European Americans, the corresponding association for this SNP was reported as “NA”, which a footnote defined as “not available (the SNP was not polymorphic).” As noted by the authors, this finding is potentially interesting and important because of the substantial racial/ethnic disparity in asthma prevalence in the US (7.7% in European Americans versus 12.5% in African Americans). 
Although the main text of the paper reports only the odds ratios and their confidence intervals, Table 1 on page 18 of the electronic supplement details the allele frequencies by group. Surprisingly, it is the minor allele, which was not observed in European Americans, that is associated with lower risk. The major allele had reported prevalences of 77.0% and 71.9% in African-origin cases and controls, respectively. There is no association in European Americans because 100% have the major allele. If this SNP is taken to be causal, therefore, the pattern for this variant would be opposite of the observed disease phenotype prevalences, with 100% of European Americans having the high risk variant. Under the more likely interpretation that the SNP is a marker in linkage disequilibrium with a causal variant in the gene PYHIN1, however, the data have nothing at all to say about PYHIN1 and asthma in European Americans. The authors would have a basis to consider the unknown variation in PYHIN1 as explaining some cases of asthma within the African-origin population, but no claim to this being relevant in any way to racial/ethnic disparities.  European Americans might have more or less of the high risk version of this gene; the data are completely silent on this issue.
It came as a surprise, therefore, to see the news reporting on this publication. For example, the Reuters story published on July 31st began "U.S. researchers have discovered a genetic mutation unique to African Americans that could help explain why blacks are so susceptible to asthma." [3] The story seemed to portray the SNP as the causal variant itself:  
"But because the study was so large and ethnically enabled the researchers to find this new gene variant that exists only in African Americans and African Caribbeans. This new variant, located in a gene called PYHIN1, is part of a family of genes linked with the body's response to viral infections, Ober said. "We were very excited when we realized it doesn't exist in Europe," she said."
How can one make sense of this text in relation to the published paper? If the reported SNP is by some great stroke of luck the causal variant itself, then it cannot explain the observed racial/ethnic disparity since it would lower risk in some blacks in relation to whites. If, on the other hand, the SNP is merely a marker for a causal variant somewhere nearby, presumably in PYHIN1, then it is nonsense to say of this unknown variant that it “doesn’t exist in Europe.” The data reveal nothing at all about the distribution of this variant in European Americans since no marker for this gene was found in that population. Either way, therefore, the news story did not seem to reflect the data that were reported in the article.
Thinking that this was an example of the press being irresponsibly sensationalistic, and misrepresenting the peer-reviewed article, I sent a letter on August 2nd to the Reuters science reporter and editor, signed by myself and about a dozen colleagues. We also sent a copy of the letter to the corresponding author of the article, the University of Chicago statistical geneticist Dan Nicolae.
The Reuters editor sent a detailed response without delay. She reviewed the statistical significance of the association measure and the proposed biological mechanism for how the PYHIN1 gene might affect asthma risk, and noted that the science reporter’s text was supported by interviews with two of the researchers as well as from a contact at the National Institutes of Health. To document this, she attached an e-mail from two of the authors, Dan Nicolae and Carole Ober, in which they affirmed their approval of the coverage their work had received. “First let us say that we think that the article is very well written and we have no major issues with it. We do not understand the issues raised in Dr. Kaufman's letter,” they wrote. They went on to note that perhaps the Reuters title might “slightly overstate the conclusion of our study”, but that it was a “subtle distinction” at best. “We thank you for helping us promote our science,” they concluded.
I then wrote to Dan Nicolae directly, asking him how the Reuters text could be construed to be consistent with the information in the paper. “I understand that race is a sensitive issue, subject to many debates,” he responded. “My research is on understanding molecular mechanisms of complex diseases, with the hope that this will lead to better treatments. It has nothing to do with this debate. On the Reuters news item, let me state that there are several scenarios where our data would fit with that headline. I will not discuss these scenarios here because I am convinced they will produce other discussion, and I prefer to use my time on my research projects.”
Apparently, Dr. Nicolae was comfortable that the Reuters reporting did not reflect the content of the paper because he believed that there were theories, not explored in the published article, which could make the news story valid.  On the basis of his reply, I came to believe that this incident was not the result of a science reporter misunderstanding the published paper. Rather, it seemed to be the case of the scientist providing a speculative interpretation that was not vetted by the reviewers or the editors of the journal. Dr. Nicolae offered that my confusion may have arisen from ignorance, and recommended that I read up on tag SNPs, differences in linkage disequilibrium patterns between Europeans and Africans, and association signals produced by interactions. “These will lead you to these scenarios I am referring to,” he concluded.
While it is possible for a risk factor to operate in different directions across two populations, this entirely sidesteps my concern, which is that the reporting strayed from what could be said based on the content of the published article. There could be no evidence of effect measure modification presented for this variant, since there was no exposure variation in the European Americans, and therefore no association measure could be estimated in that group. Dr. Nicolae did not appear to disagree with me on this point, but seemed to view the media interview as an opportunity for presenting his research program as relevant to racial disparities in a way that could not be directly derived from the published data. This is surely a fine line, because journalists often want scientists to give their expert opinions on the broader interpretation of the published work. But how far should authors go in describing what they might speculate to be true, rather than what they actually found? The impetus for the news story was the publication of an article in a respected scientific journal. Are there really no constraints on how far authors can extend their interpretation while claiming to be referring to the article? Should they clearly indicate that they are speculating - and should they also present at the same time the potential contrary or skeptical view? With so much attention and funding riding on efforts to understand and reduce minority excess burden of disease, the authors’ speculation risks the appearance of being self-serving. If scientists sometimes disparage science reporters as the source of popular misinformation, the fair reply might therefore be “Cura te ipsum!”

Jay S. Kaufman, Ph.D

If you like to comment, Email me - or in this case Dr Kaufman -  directly at or submit your comment via the journal which requires a password protected login. Unfortunately, published comments are limited to 1000 characters.

[2] Torgerson DG, et al. Meta-analysis of genome-wide association studies of asthma in ethnically diverse North American populations. Nat Genet. 2011 Jul 31;43(9):887-92. doi: 10.1038/ng.888.