Red blood cell transfusion in acute pulmonary embolism

To the Editors:

Wong et al.1 reported that in patients with acute pulmonary embolism (APE), red blood cell (RBC) transfusion is associated with higher mortality, independent of haemoglobin level.

However, there must be concern that this conclusion may be biased by factors, such as inappropriate comparisons. While the sample size of this retrospective cohort study was relatively large (1376 patients), the rate of RBC transfusion was low (5.3%). In our view, the baseline characteristics between these two groups were significantly different, as patients in the transfusion group were more severely ill. Thus, the overall conclusion could be affected by selection bias. RBC transfusion is not a routine procedure in the management of APE,2 but it is recommended in the small subgroup of patients who have major bleeding or severe anaemia. As the aim of this study was to investigate the association between RBC transfusion and mortality, in our view, patients without any indication for RBC transfusion should have been excluded from analyses.

Finally, only significant variables with P < 0.05 were included in the final model. In logistic model building, variable selection is an important step to improve the fitness of the model, with various methods such as stepwise and best subset approaches. However, both statistically significant and clinically important variables should be considered as confounders.3, 4 For the current study, factors such as severe blood loss and haemodynamic status (including vasopressor use, blood pressure etc.) should be included in the model as these factors are important reasons for RBC transfusion in clinical practice. Moreover, based on the imbalance between groups, a propensity score matching approach may have been a better option to minimize the risk of bias.