The authors address an interesting topic that has not been described earlier, and they had the opportunity to use a large dataset to answer their research question, “Is there a dose-response relationship between transfusion volume and mortality or composite morbidity?” ICD codes were used to score pre-hospital and in-hospital (co-)morbidity, and the Charlson comorbidity index was calculated and used as a potential confounder in a multivariable model. The authors conclude that patients receiving high- or very-high-dose transfusion are at especially high risk for hospital-acquired infections and thrombotic events.

In my opinion, more information is warranted for better interpretation of these results. Categorisation of the surgical population is needed: what were the results for elective surgical patients compared to those undergoing acute surgery (trauma patients, etc.)? What were the exact reasons for admission (sometimes difficult to identify using ICD codes) and what was the exact indication for transfusion (also difficult to determine using ICD codes)? The term “dose-response relationship” is rather confusing, since it implies a causal relationship that could not be established. The reader should be aware that the results only show associations, as the authors recognise in the discussion section.

Furthermore, the data were collected in the USA from a hospital billing database, therefore the results may not be easily extrapolated to countries with different insurance systems. The main limitation, however, remains confounding by indication: sicker patients receive higher-dose transfusions (also described as residual confounding). Although this study provides interesting data, we first need to await other studies before changing clinical practice.

– Cynthia So-Osman