| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
Stanford University; Editor-in-Chief; Anesthesia & Analgesia; steven.shafer{at}stanford.edu
In Response:
As noted in the prior two letters (1,2), the manuscript by Basran et al. (3) contains errors in the tables. Specifically, the odds ratios are not the geometric means of the upper and lower bounds of the confidence intervals.
To respond to the concerns of Rothmann, Braun, and Ng, the authors attempted to return to their data analysis and figure out which was incorrect, the odds ratio or the confidence intervals. However, the study was done 5 years ago, when the authors were at Columbia University. Since then the authors have left Columbia. None of the authors can locate the original data set. What should we do?
First, let’s consider the tables. As Rothmann, Braun, and Ng point out, the tables are internally inconsistent. Indeed, in every line of Tables 2–4, the odds ratio is not the geometric mean of the upper and lower bounds of confidence interval. Which is correct, the odds ratio, or the confidence interval? If the confidence intervals are correct, then every line in Tables 2–4 contains three separate errors: 1) an error in the odds ratio, 2) an error in the P value, and 3) an error in the calculation of the confidence interval that offsets the error in the odds ratio (remember that confidence intervals are calculated from the odds ratio). Occam’s razor suggests that there is a single error in the calculation of the confidence intervals, rather than three offsetting errors yielding a correct confidence interval about an incorrect odds ratio.
|
|
|
Second, what to do about the manuscript itself? The authors have lost the data. They cannot defend their results. How embarrassing! The authors considered retracting the manuscript, as did the Editorial Board of Anesthesia & Analgesia. However, retraction implies that the findings are not to be believed. Turning again to Occam’s razor, the most simple explanation is that the files were simply lost as people moved about, and that the only problem with the manuscript is the incorrectly calculated confidence interval. If this is the case, then the findings of the manuscript still stand, albeit weakly. After much deliberation, all parties decided that the manuscript would not be retracted, but that the author’s inability to find the original data should be publicly stated.
The findings of the manuscript are important. The authors, and other laboratories, need to replicate these findings. The lives of our patients depend upon our knowing whether there is increased risk of transfusing blood that is nearing its expiration date.
Footnotes
This article has supplementary material on the Web site: www.anesthesia-analgesia.org.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
|
