JAMA Oncology has retracted a 2018 paper by a group from Sweden and the UK after the researchers discovered critical errors in their analysis that significantly altered the conclusions of the study.
The paper, “Risk of esophageal adenocarcinoma after antireflux surgery in patients with gastroesophageal reflux disease in the Nordic countries,” purported to find that
Medical and surgical treatment of GERD were associated with a similar reduced esophageal adenocarcinoma risk, with the risk decreasing to the same level as that in the background population over time, supporting the hypothesis that effective treatment of GERD might prevent esophageal adenocarcinoma.
But shortly after publication the authors, led by Jesper Lagergren, of King’s College London and the Karolinska Institutet in Stockholm, discovered a “major” problem. As they explain in the retraction notice:
This article reports the results of a population-based cohort study that included 48 414 participants with gastroesophageal reflux disease (GERD) who underwent antireflux surgery and 894 492 participants with GERD treated with medication only in Denmark, Finland, Iceland, Norway, and Sweden. We concluded that the risk of esophageal adenocarcinoma decreased over time after antireflux surgery to reach a similar level as that in the background population and was similar when comparing individuals who underwent surgery with those who did not.
Unfortunately, we realized after publication that there was a major problem with the analysis in our study. One of our group’s statisticians, who was not an author of this article, recently started analyzing the risk of pharyngeal and laryngeal cancer in the same cohort and identified a problem with the original statistical analyses of the follow-up categories after antireflux surgery and medication. For each follow-up category, only those people with the total study time within the category were kept in the analysis. If an individual was included, the individual’s entire follow-up was used for the calculations. This led to 2 problems: in earlier follow-up categories, the person-years were underestimated, and in later follow-up categories, the person-years were overestimated.
After re-analyzing their data, Lagergren and his colleagues found that
the reported pattern of a decreased risk of esophageal adenocarcinoma over time after antireflux surgery and medication is incorrect. Instead, no decrease was found after the correct analyses were conducted. The same mistake was present in the Cox regression analysis, but because both exposed and nonexposed were treated in the same way, the resulting hazard ratios are similar to those initially reported. Finally, some minor corrections of the cohort members led to some small differences in the numbers in the cohort, but these did not change the results.
However, the study’s conclusion changes significantly and in a clinically important manner to the following: “Medical and surgical treatment of GERD were not associated with any reduced esophageal adenocarcinoma risk, compared with the background population over time, indicating that treatment of GERD might not prevent esophageal adenocarcinoma.” Thus, we have requested that the journal retract the article.
The change has important clinical implications. GERD affects as much as 20 percent of the U.S. population and is the leading risk factor for Barrett’s esophagus, a precancerous condition linked to esophageal cancer.
‘A rare commodity’
In a brief editorial accompanying the retraction, Mary Disis, of the University of Washington, and Lee Ellis, of MD Anderson, who edit JAMA Oncology, praised the authors for their “honesty”:
Scientific integrity seems to be a rare commodity these days. In this issue of JAMA Oncology, we have a retraction of an Article due to an analysis error brought to our attention by the authors themselves. We commend Jesper Lagergren,MD, PhD, and colleagues for discovering this mistake after their research was published and immediately contacting the journal to ensure that corrected information in the analysis and outcome was rapidly made available to our readers. Ensuring that accurate data are accessible to others for the foundation of future studies is the basis of scientific publishing. The honesty of authors, such as Dr Lagergren and colleagues, serves as an excellent example of scientific integrity.
Ellis told us the retraction was the journal’s first.
Like Retraction Watch? You can make a tax-deductible contribution to support our growth, follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up for an email every time there’s a new post (look for the “follow” button at the lower right part of your screen), or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
Exemplary!
Please don’t slander all of science – the vast majority of colleagues that I interact with regularly have plenty of integrity!
That said, I work in the very basic biochemical sciences, not translational medicine. I can’t speak for other fields.
First, it wouldn’t be slander it would be libel, and second you should find out just what both mean before commenting.
RW: Exactly what purpose does it serve for you to publish a comment in which a poster tells a woman she is too ignorant to participate in the discussion?
I see that S.D.’s criticism is pointed and probably unnecessary and unconstructive, but what does gender have to do with it? S.D. did not mention gender, only you did.
Further, are you suggesting Retraction Watch should police the gender of commenters in order to shelter women commenters from potential criticism? Are women commenters too fragile to withstand or rebut criticism themselves? Are you assuming S.D. is a different gender than the original commenter because he/she/they made a critical comment?
Your remark strikes me as containing sexist thinking more than anything commenter S.D. or Retraction Watch have done here.
I have a problem with any commenter that criticizes another commenter over some triviality totally irrelevant to the discussion and assumes they have the right to police the space by telling that person whether or how they should post. I’m surprised RW allowed SD’s comment as it is a hostile and offensive response to a perfectly legitimate post. It’s not “probably unnecessary”. It is unnecessary, and destructive to the civil discussion RW normally encourages. I’ve been reading RW for quite a while, and I don’t recall another such post being published.
And yes, I noticed that SD directed this comment towards a woman, and I have a particular problem with that. It’s a sport I’ve seen played for a long time. That’s what gender has to do with it.
To suggest that my objection to SD’s demeaning, condescending, and offensive post means that I want RW to “police the gender of commenters in order to shelter women from potential criticism” and is “sexist thinking” is, I hope, disingenuous and not really as obtuse as it sounds.
How many other statistical/data errors are out there that haven’t been discovered and perhaps may never be? This one was only found out because one of the statisticians started to analyze a different outcome in the same cohort, otherwise it would never have been noticed.
K. F, I agree. There seems to be a worrying lack of quality control too many places.
In this case it looks as if the problem is not with the numbers but with the logic. And that ought to be thought through by the people who designed the study, if no later then at the time they wrote the paper.
And a plug for consulting your biostatistician and epidemiologist *early* in the study design, catching these sorts of errors is kind of our job.