Error in a Study of the Comparison of Platelet Function Tests in Predicting Clinical Outcome in Patients Undergoing Coronary Artery Stent Implantation

To the Editor: We are writing to inform the readers and editors of JAMA that we have discovered errors in our article entitled “Comparison of platelet function tests in predicting clinical outcome in patients undergoing coronary stent implantation,”¹ published in the February 24, 2010, issue of JAMA (2010; 303[8]:754-762), caused by a flaw in our statistics program (R).

Because of the negative concordance of 3 of in-total 8 tests studied in the Popular study (the platelet function analysis [PFA] collagen/adenosine diphosphate [ADP] cartridge, the Innovance PFA P2Y, and the IMPACT-R ADP test), the initially calculated receiver operating characteristic (ROC) curves were mirrored, leading to the incorrect optimal cutoff value. Consequently, the cutoff values we reported for these 3 tests were incorrect. This error was brought to our attention by a staff member of Siemens Healthcare Diagnostics Products (Marburg, Germany) who had access to the data set related to the Innovance PFA P2Y, because this cartridge was under development at the time of inclusion in the Popular study. Because Siemens used another statistics program (SAS; SAS Institute, Cary, North Carolina) in which the ROC curve was mirrored prior to the calculation of the cutoff value, another cutoff value for this cartridge was derived. After careful recalculation, we confirmed that our cutoff value was erroneous and that the same error was made in calculating the cutoff values of the IMPACT-R ADP as well as the PFA COL/ADP. We regret to report that this error was overlooked despite careful review by the principal investigators, the statistician, and the coauthors.

This specific error affects multiple numbers, percentages, and associated odds ratios with confidence intervals in several paragraphs, tables, and figures in our article. The correctly calculated cutoff values do not have any effect on the predictability of the IMPACT-R test and the PFA COL/ADP; however, with the corrected cutoff values, the Innovance PFA P2Y predicted the primary end point (where it did not with the incorrectly calculated cutoff values in the original article).

In addition, we observed an error in Figure 2 concerning the PFA-100 system. The closure times of the PFA COL/ADP are correct, but for the Innovance P2Y the lower line indicates patients with an end point and should not be dotted.

A Article appears in this issue of JAMA, and the article has been corrected online. We sincerely regret the errors in the initially calculated cutoff values and subsequent interpretation of the data and apologize for the confusion caused by publication of this incorrect information.