The British Medical Journal has retracted an article examining when unexpected weight loss could be a warning sign of cancer after the authors found an error in their work. The journal published an updated version of the analysis with different conclusions, which the authors think could influence patient care.
The retracted paper, “Prioritising primary care patients with unexpected weight loss for cancer investigation: diagnostic accuracy study,” appeared Aug. 13, 2020. The researchers, led by Brian D. Nicholson, a general practitioner and associate professor in the Nuffield Department of Primary Care Health Sciences at the University of Oxford, England, used electronic health records data to look for people diagnosed with cancer within six months after a recording of unexpected weight loss.
The authors were attempting to replicate their results in another dataset when they found “some differences in the study findings and study population that we could not easily explain,” Nicholson told Retraction Watch. He continued:
For many months we tried hard to understand what could have caused these differences. As part of our investigation, we reviewed the processes that had been used to select the study population included in the analysis and identified that it was likely that an unanticipated selection bias had been introduced whilst the dataset had been prepared for the analysis.
Nicholson and his colleagues contacted the BMJ when they felt they could “articulate what we thought had happened,” he said. The group “found the journal very responsive to our approach for advice about what we should do next,” he added.
On March 25, the journal published an expression of concern:
The authors of the paper recently identified an error in their approach to the research. Some patients were excluded from the study because their healthcare records contained a code that was not associated with weight loss. Sometime later, however, some of these patients are likely to have had a code included in their healthcare record that was associated with unintended weight loss and cancer. This created an unintended selection bias in the work; some patients were excluded from the study who should have been included in the study. Rectifying this error would result in substantially more patients being included in the study. Reanalysis changes the study’s key results and messages. It corrects an underestimate of the likelihood of cancer for some men, and a larger group of women. BMJ is working with the authors to review a new version of the paper and to determine what post-publication action is suitable in this case. An update will be provided when our evaluation has concluded.
The journal ultimately decided to publish an updated version of the article, and the authors were “relieved,” Nicholson said.
The retraction notice, published October 16, stated the authors and journal agreed to retract the flawed paper, and linked to the updated version. The notice was largely the same as the expression of concern, but elaborated on the effect of excluding the patients who should have been included. The exclusions “caused the study to underestimate the prevalence of cancer in the general primary care population,” the notice stated.
The original study included nearly 64,000 patients with unexpected weight loss recorded in their medical records between 2000-2012, of which 1.4% ended up receiving a cancer diagnosis within six months. In the update, the authors analyzed records from about 326,000 patients with unexpected weight loss between 2000-2019. This time, 4.8% had received a cancer diagnosis half a year later.
The results also changed. The first analysis found unexpected weight loss indicated a high enough risk of cancer to warrant follow-up investigation only for men at least 50 years old who had smoked and people with additional symptoms. The re-analysis found unexpected weight loss in all men at least 50 years old and women at least 60 years old, as well as younger patients with additional symptoms, indicated “the risk of cancer warrants referral for invasive investigation.”
The BMJ “has contacted those who have cited the work to alert them to the new results in the updated version of the paper,” according to a spokesperson for the journal. The authors and The BMJ have also contacted the UK’s National Institute for Health and Care Excellence “because the paper could inform recommendations about which individuals should be referred to specialist services with unexplained weight loss in primary care.”
NICE did not respond to our request for comment on whether the organization would need to update any recommendations.
The retracted article has been cited 15 times, according to Clarivate’s Web of Science, with 4 citations since the March expression of concern.
The revised results have not yet been reported as widely as the first article, according to Altmetric, which tracks the online attention articles receive.
“Our updated results report the risk of cancer in patients attending their GP with unexpected weight loss,” Nicholson said. “We hope these results will be taken into account when our national guidelines, and other international clinical guidelines, are updated about who should be urgently referred for cancer investigation.”
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, 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].