Donor sex, prior pregnancy not linked to mortality risk among transfusion recipients

The results from this paper are in stark contrast with those from two of our previous studies and several publications from other authors. As Edgren and colleagues correctly noted, this could be due to differences between patient or donor populations or blood banking practices, or methodological differences between the studies.

Differences in populations and banking practices clearly exist. For example, donors are much younger in the United States, where high school and college blood drives are common. Also, prevalence of many diseases differs widely among countries. It has been suggested that patients with cardiovascular disease, which is relatively common in the United States, are insensitive to the negative effects of blood from previously pregnant donors. Further, the investigated patient population consists only of transfusion recipients, which creates a strong selection bias. Because of the much more liberal transfusion policies in both the United States and Scandinavia, patients will be transfused with diagnoses that would not have resulted in a transfusion in the Netherlands.

There are also major differences in blood product preparation techniques. In the United States, the standard method is the “soft spin,” whereas in Europe it’s the “hard spin,” which could lead to a completely different composition of the red cell pellet, including how many and which types of leukocytes end up polluting that pellet.

All these differences could provide important clues about the biological mechanism behind the association we have repeatedly observed.

Differences in methodology also are plentiful, but difficult to fully appreciate in a short text. One major difference is that Edgren and colleagues focused their primary analyses on in-hospital mortality, which covers a very short time span. We previously observed a difference in mortality that kept growing for at least 3 years. Also, the methods used to correct for confounding by number of transfusions are not entirely clear from the short text that is allowed in a JAMA publication, but it appears these could leave room for residual confounding. Similarly, the very short description of the imputation model seems to suggest this method could introduce random noise, which would result in effect dilution.

Finally, in both our previous studies we analysed a “single transfusion cohort,” which cannot suffer from confounding by number of transfusions. We could therefore present a Kaplan-Meier survival graph, which suggested one in 16 men aged younger than 50 years who receives a single transfusion with blood from a previously pregnant female donor will die within 3 years, whereas he would not have had he received that transfusion from a male donor. Although Edgren and colleagues analyse a single transfusion cohort, they do not present such an insightful Kaplan-Meier graph.

Edgren and colleagues did some great work that raises important new suggestions for future lines of research. However, these results should not detract from the problem we observe in several other countries, including the Netherlands. Further research should aim to elucidate the biological mechanism, which would allow a much more targeted approach to increasing the safety of transfusion recipients. Currently, our best options seem to be matching transfusions by sex of the donor and recipient or permanently deferring all previously pregnant donors. Both of these are logistically demanding, and information that would allow for more targeted approaches is clearly needed.