Corrections and Retractions in Health Research

Papers can change after they are published

A journal article is not frozen. Publishers can correct it, flag it with a warning, or withdraw it. These notices are part of how science repairs itself, and they matter to anyone who reads or cites research. Each one means something different. Early drafts have their own issues, covered in Preprints vs Peer-Reviewed Papers, and the wider map of source types is in Press Release vs Conference Abstract vs Peer-Reviewed Paper.

The main notice types

Why papers are retracted

Journals may retract an article because of pervasive error or unsubstantiated or irreproducible data. [1] The reasons vary widely. A review of 2,047 biomedical and life-science articles indexed as retracted in PubMed on May 3, 2012 found that only 21.3% of retractions were attributable to error, while 67.4% were attributable to misconduct, which included fraud or suspected fraud, duplicate publication and plagiarism. [2] That analysis is from 2012, so treat the figures as a snapshot, not today's rates.

One catch: PubMed does not separate honest error from misconduct, because it links to what the journal published. [1] So the wording of the retraction notice itself is worth reading.

How to check a study before you rely on it

• Look for the label. PubMed gives the retracted article the publication type Retracted Publication. [1] It also links retraction notices to the original articles. [1]
• Read the journal page. Editors are advised to link the retraction and the original article both ways and to label the retracted article as retracted in every form. [3]
• Search independent trackers. Retraction Watch follows retractions and runs a database of them. [4]
• Check related work. The validity of earlier work by the author of a fraudulent paper cannot be assumed. [3]

Step-by-step searching is in How to Find Emerging Research and Trial Records.

What a retraction does and does not mean

• It means that specific paper should not be relied on. It does not by itself show that the idea, the treatment or the whole field is wrong. Other, independent studies may exist, and they should be judged on their own.
• A missing retraction does not mean a result is solid. Findings can fail without any notice. Replication effects in one large psychology project were half the size of the original effects. [5] Among highly cited clinical studies, 16% were later contradicted. [6]
• Not every concern ends in a retraction. The ICMJE editors' guidance notes there may be circumstances in which no misconduct is proven. [3]

How we handle it on this site

We check for notices before using a paper. If a source we used is later retracted, we remove or rewrite the claim it supported and change the page's editorial date. See How We Write Study Explainers and Research Updates. We do not claim to have scanned every paper in a field for notices, only the ones we cite.