BACKGROUND: A problem when studying adverse events of blood transfusions is that patients have usually received transfusions from several donors while only one of these donors is the actual cause. This will result in underestimation of the effect of donor-related risk factors if not adequately corrected for. We encountered this problem when studying transfusion-related acute lung injury (TRALI) and describe four methods to overcome this problem.
STUDY DESIGN AND METHODS: Simulated data are used to illustrate the results of six different approaches: not correcting for the number of donors, using standard correction methods, and four newly proposed methods. Donor sex is used throughout as an example. The first two new methods apply restriction of the study to cases who have received a transfusion from a single donor or from donors who are all of the same sex. In both restriction designs the sex of the causal donors is known and can be compared to the expected value from a reference population. The other two new methods apply statistical correction for the number of donors, either by standardization or by maximum likelihood methods.
RESULTS: If not corrected for, or if corrected for by standard methods, increasing numbers of donors per patient result in decreasing estimates of the effect of risk factors. All four newly proposed methods yield valid estimates.
CONCLUSIONS: It is clear that the problem of multiple transfusions requires specialized correction methods. All four newly proposed methods yield on average good estimates of the underlying true value.