On 2014 Aug 26, Joyce E B Backus commented:

Readers of this analysis should be aware of the background of retractions in PubMed and, in particular, the peculiarities of the examples cited. Publishers often submit a new record for a retraction so that the original citation and retraction notice citation can be linked together but remain separate. In this case, the publisher chose to make a change to the record for the original paper. The retraction notice Anonymous, 2011 (featured in Table 2), is the record upon which the authors base their finding that the “number of retraction notices became stable 35 months after the retraction” but this record has some unusual circumstances PubMed.

The original paper was published electronically ahead-of-print on 2008 Jul 23; then the paper was retracted (for the first time in 2009), and was never published in print. The publishers substituted the retraction notice online in place of the original paper, but the electronic date of publication remained that of the original paper even though most of the record is related to the retraction notice. The situation is further complicated because the additional retraction notice for the original paper, Lindstrom A, 2009, was published on 2009 Mar 15, and correctly lacks the electronic date of publication of the article. Neither of these retraction notices was published in 2008; Lindstrom A, 2009 was not part of the study and Anonymous, 2011 should also not have been part of the study. Another retraction notice example listed in Table 1 that has a similar publishing practice involves Wolfort RM, 2011 (which incorrectly has the October 24, 2008 electronic publication date of the original article) and Anonymous, 2011. In these cases, the records carry publication dates that do not pertain to the retraction notice itself because the publisher chose to send a replacement record which only partially replaced the record for the original paper.

Also, these few examples can’t be used to draw broad conclusions about the timing of retraction notices, which can be issued at any time, sometimes many years after publication. For example, a 2014 PubMed record Anonymous, 2014 retracts a 2005 article, and a 2013 PubMed record Van Le TS, 2013 retracts a 2004 article.

Retractions are explained in detail in the NLM Fact Sheet, Errata, Retractions, Partial Retractions, Corrected and Republished Articles, Duplicate Publications, Comments, Updates, Patient Summaries, and Republished (Reprinted) Articles Policy for MEDLINE.”

Joyce Backus, Associate Director for Library Operations, National Library of Medicine

On 2014 Aug 26, Joyce E B Backus commented:

Readers of this analysis should be aware of the background of retractions in PubMed and, in particular, the peculiarities of the examples cited. Publishers often submit a new record for a retraction so that the original citation and retraction notice citation can be linked together but remain separate. In this case, the publisher chose to make a change to the record for the original paper. The retraction notice Anonymous, 2011 (featured in Table 2), is the record upon which the authors base their finding that the “number of retraction notices became stable 35 months after the retraction” but this record has some unusual circumstances PubMed.

The original paper was published electronically ahead-of-print on 2008 Jul 23; then the paper was retracted (for the first time in 2009), and was never published in print. The publishers substituted the retraction notice online in place of the original paper, but the electronic date of publication remained that of the original paper even though most of the record is related to the retraction notice. The situation is further complicated because the additional retraction notice for the original paper, Lindstrom A, 2009, was published on 2009 Mar 15, and correctly lacks the electronic date of publication of the article. Neither of these retraction notices was published in 2008; Lindstrom A, 2009 was not part of the study and Anonymous, 2011 should also not have been part of the study. Another retraction notice example listed in Table 1 that has a similar publishing practice involves Wolfort RM, 2011 (which incorrectly has the October 24, 2008 electronic publication date of the original article) and Anonymous, 2011. In these cases, the records carry publication dates that do not pertain to the retraction notice itself because the publisher chose to send a replacement record which only partially replaced the record for the original paper.

Also, these few examples can’t be used to draw broad conclusions about the timing of retraction notices, which can be issued at any time, sometimes many years after publication. For example, a 2014 PubMed record Anonymous, 2014 retracts a 2005 article, and a 2013 PubMed record Van Le TS, 2013 retracts a 2004 article.

Retractions are explained in detail in the NLM Fact Sheet, Errata, Retractions, Partial Retractions, Corrected and Republished Articles, Duplicate Publications, Comments, Updates, Patient Summaries, and Republished (Reprinted) Articles Policy for MEDLINE.”

Joyce Backus, Associate Director for Library Operations, National Library of Medicine