ACCURACY OF STATISTICAL-METHODS IN TRANSFUSION - A REVIEW OF ARTICLES FROM JULY/AUGUST-1992 THROUGH JUNE-1993

Background: Statistical errors have been noted in a large percentage of articles appearing in medical journals. Their incidence in a blood banking journal, however, has not been studied. Study Design and Methods: Original articles appearing in the journal TRANSFUSION from July 1992 through June 1993 were analyzed for correctness of statistical methods. Each article was reviewed by both a transfusion medicine physician and a statistician. Results: There were 122 original articles, of which 59 contained statistical methods and were reviewed. Of these, 23 (39%) contained an error in the statistical description of their data, 47 (80%) failed to describe clearly the statistical tests performed, and 44 (75%) involved an incorrect statistical test or contained an error in test calculation or interpretation. As a result of these errors, 13 (22%) of the 55 articles analyzed reported conclusions not supported by the data. Conclusion: Misuse of statistical methodology may not be uncommon in transfusion medicine research, and it would be prudent to give more attention to statistical methodology in such research.