Incorrect P Values in Two Articles by Ke-Xuan Liu, M.D., Ph.D., et al.

We have read with interest the editorial by Drs. Kharasch and Houle¹  regarding an algorithm developed by Carlisle to detect nonrandom sampling in clinical trials.²  Two of our previous articles^3,4  published in Anesthesiology were mentioned in Carlisle’s publication, and we were asked by the journal to verify the correctness of reporting in these articles. Upon review, it is apparent that the P values in tables 1 and 2 of our 2014 article³  and in table 2 of our 2013 article⁴  are systematically incorrect, given the descriptive statistics. We are unable to locate the primary data to recalculate the P values in these tables. We have located the primary data for table 3 of the 2014 article and verified that the descriptive data presented in the article are correct. We believe that the reported descriptive statistics, including tables 1 and 2 in both articles, are correct, and that the reported P values are incorrect as the result of a systematic error. We believe this has been a systematic copy-and-paste error when data were transferred from one set of analysis to the tables. We have recalculated the P values using the descriptive statistics with simplified t tests, and we have found that all new P values are systematically smaller than the reported erroneous P values. Thus, we believe the conclusions of our studies are not affected by the error.

The authors declare no competing interests.