Jumbled analysis leads to retraction of cancer study — but also another paper

The authors of a study on cancer incidence and survival in the Dutch migrant community have retracted it after realizing they’d made some errors that significantly affected the results.

But in what seems like an appropriate reward for coming forward, the newly analyzed data, with additional information, will be part of a forthcoming paper in another journal by the same authors.

The original retrospective study came out in May 2011 in the European Journal of Cancer Prevention and found that risk and survival of breast and stomach cancers differed depending on the person’s mother country. The problem was that the study jumbled up many of the participants’ homelands during the analysis.

The authors issued the following retraction:

The authors of the article ‘Breast and stomach cancer incidence and survival in migrants in the Netherlands, 1996–2006’ (European Journal of Cancer Prevention, 20, 150–156) recently found errors in the coding of country of birth that forced them to reconsider the handling of missing values in this study. Re-examination of the results of breast and stomach cancer incidence and survival revealed the presence of inaccuracies. The re-analysis with correctly coded and updated data till 2009 led to more accurate results, especially when some further stratification by menopausal status of breast cancer was included. For the survival analysis, the authors included only cases from three out of nine regional registries with high completeness of country of birth, covering approx. 40% of the Dutch population, and recalculated the estimates. The authors found that survival from breast cancer was slightly lower in premenopausal and slightly higher in postmenopausal migrant women when compared to Dutch natives. However, hazard ratios for death were higher in premenopausal Moroccan and Surinamese patients only. The incidence analysis was again carried out based on nationwide data: it confirmed the, already found, significantly reduced risks for breast cancer in migrant women (except Antilleans) and strongly elevated risks for non-cardia stomach cancer in all migrant groups (except Indonesians).

The authors have therefore decided to formally retract the below named paper.

Arnold M, Aarts MJ, Siesling S, van der Aa M, Visser O, Coebergh JW (2011). Breast and stomach cancer incidence and survival in migrants in the Netherlands, 1996–2006. Eur J Cancer Prev 20:150–156.

The paper has been cited just once, according to Thomson Scientific’s Web of Knowledge.

The senior author, Jan Willem Coebergh, an epidemiologist at Erasmus MC in Rotterdam, Netherlands, gave Retraction Watch some details on how his team found the error and how they turned the retraction into a new manuscript.

The error was found by ourselves when a similar analysis was carried out by our group on two other cancers, which has in the meantime been published in the Eur J of Public Health. When we reported this to the editor of the Eur J Cancer Prevention, we offered him to replace the original article by a new version. We were then asked to prepare a short retraction note on the reasons for our withdrawal. We briefly explained to the scientific audience what happened and in what way the error influenced the results. We were in fact then rewriting the paper, also including more recent data and using slightly different methods. We also made a distinction between premenopausal and postmenopausal breast cancer. However, this new version  was rejected by the Eur J Cancer Prevention (pity  but true) but subsequently submitted and accepted and currently in press at Acta Oncologica.

All in all, we believe that this approach was the most sensible and honest way to handle this situation.

We’d have to agree. We’d file this under “transparent.”

The editor-in-chief of the European Journal of Cancer Prevention did not respond to requests for comment.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.