In March 2016, researchers in Switzerland and Canada published a meta-analysis in The Lancet, exploring the optimal painkiller and dose for treating pain in knee and hip osteoarthritis. Soon after, the authors were informed of an error that would change “all numbers” in a paper that may influence clinical practice.
The authors contacted The Lancet immediately, in July 2016, to inform them of the issue. Sven Trelle, the paper’s corresponding author, also told us:
…it was clear very early on/quickly that the changes will not affect the conclusions but only insignificant numbers.
But given “the extent of the changes necessary,” The Lancet editors opted to retract the article and publish a corrected version, after the paper had been analyzed and reviewed again.
The process took a year. In that time, the uncorrected paper—which, according to Trelle, has the potential to change practice—has been cited 23 times, making it a Highly Cited Paper (HCP), according to Clarivate Analytics’ Web of Science. The HCP designation means that the paper received enough citations to place it in the top 1% for its field and publication year.
Still, The Lancet did not issue an expression of concern or otherwise notify readers of the problem. A spokesperson for The Lancet explained:
An expression of concern was not thought to be needed in this case as these are usually only used as an interim position, often while we await the outcome of an institutional investigation, which in this case wasn’t indicated – what we did need was further analysis and further review. As the Comment says, “the overall message remains the same”.
On July 8, 2017, The Lancet published a retraction and replacement notice as well as the revised manuscript. Here’s the retraction and replacement notice:
On March 17, 2016, The Lancet published online a network meta-analysis of the effectiveness of non-steroidal anti-inflammatory drugs for pain in knee and hip osteoarthritis, and the Article was published in print on May 21, 2016.1 On July 6, 2016, the authors drew our attention to two missed trials2, 3 and a duplicate publication.4, 5 Lancet editors discussed the corrections that were needed in the paper, and decided, in accordance with the Committee on Publication Ethics’ guidelines, that because of the extent of the changes necessary, the previous version of the Article should be retracted and a corrected version republished after reanalysis and rereview.
Today we retract the previous version and republish online the corrected version of the Article,6 in which the findings are slightly changed—ie, confidence intervals around the effects have changed slightly, mainly in the second digit after the decimal point, and the test for a linear dose effect is now significant for only one preparation (but was for three in the previous publication). The overall message remains the same. The previous version of the Article has been added to the appendix in the new version and is marked retracted.
“Effectiveness of non-steroidal anti-inflammatory drugs for the treatment of osteoarthritis pain: a network meta-analysis” received 22 of its 23 citations after July 2016.
Trelle, who is based at the University of Bern in Switzerland, also told us:
…we were contacted by an external group that informed us that we had missed 2 trials and included 1 duplicate. We decided to consider these trials. Not unsurprisingly, all numbers change after changing the dataset and we decided to be fully transparent about the changes. However, none of the changes were clinically relevant or changed the conclusions/interpretation.
Still, the extent of the changes made it necessary for The Lancet to classify the correction as a retraction-republication. The accompanying comment is transparent in this respect. Personally, I consider the classification as retraction unfortunate – but this is obviously the opinion of an author and I acknowledge that others might have another view.
Although the authors “immediately informed The Lancet about the issue,” Trelle explained that:
Re-running everything took a little bit of time (especially because we re-checked the whole selection process again also) but it was clear beforehand (given the small amount of changes in the data) that none of the conclusions would change.
A spokesperson for The Lancet explained:
… further analysis was needed to be done by the authors, and then the editors sought further review of the paper, hence the delay.
The spokesperson added:
We have published the corrected version of the paper, and the previous version remains attached to the appendix for full transparency, so that others can compare the findings.
In the past few years, we have seen an uptick in the use of retraction and replacement, especially in The Lancet and JAMA. According to several editors, retract and replace is a good option for correcting honest mistakes that may be too extensive for an erratum.
Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up on our homepage for an email every time there’s a new post, or subscribe to our new daily digest. Click here to review our Comments Policy. For a sneak peek at what we’re working on, click here.