“Development and validation of algorithms to differentiate ductal carcinoma in situ from invasive breast cancer within administrative claims data” by Hirth JM, Hatch SS, Lin Y, Giordano SH, Silva HC, Kuo Y.

Abstract

The above article, published in Wiley Online Library (wileyonlinelibrary.com) on 18 April 2018, has been retracted by agreement of the authors; the journal Editor‐in‐Chief, Fadlo R. Khuri; Wiley Periodicals, Inc.; and the American Cancer Society.

After publication, the authors determined that an error was made in the article, for which they sincerely apologize. They first notified Cancer of the error on May 16, 2018, and after providing additional information requested by the journal on July 13, 2018, and August 15, 2018, the authors and Cancer editors determined that the article should be retracted. There was no institutional investigation related to this study. No other papers by the author group are affected by this retraction. Cancer has determined that the mistakes were honest error.

It was decided that retraction of this article was necessary because 297 of the 1244 ductal carcinoma in situ (DCIS) cases used as the gold standard reference in the study sample were other types of carcinoma in situ of the breast. Therefore, at the last step of algorithm 1 in the training dataset, 514 (77.2%) of the 666 original true positives had been diagnosed with DCIS in the Texas Cancer Registry (TCR) data. For algorithm 2, 339 out of the original 424, or 80.0%, had been diagnosed with DCIS in the TCR data. Unfortunately, the ICD‐9 code 233.0, which the authors misstated as DCIS rather than as carcinoma in situ of the breast, does not differentiate between the different types of carcinoma in situ of the breast. Therefore, the algorithms in the original publication actually captured all cases of in situ breast cancers and would be better described as algorithms to determine any in situ cancer of the breast. When the authors discovered their error, they ran the algorithm with DCIS cases only, and there was a significant drop in positive predictive value of their algorithm to detect only DCIS. As a result, the authors determined their algorithms do not differentiate DCIS from other breast carcinoma in situ cases well. This could cause overestimations of undertreatment for DCIS using these algorithms, as other types of carcinoma in situ of the breast have different treatment recommendations. Therefore, the authors determined, along with the editors of Cancer, that this paper should be retracted.

REFERENCE

Hirth JM, Hatch SS, Lin Y, Giordano SH, Silva HC, Kuo Y. Development and validation of algorithms to differentiate ductal carcinoma in situ from invasive breast cancer within administrative claims data. Cancer 2018 Jul 1;124(13): 2815‐23. https://doi.org/10.1002/cncr.31393.

http://bit.ly/2DflbZR