Pre‐operative anaemia and mortality: addressing time bias and model robustness

We read with interest the study by Groenewegen et al. 1 and wish to highlight three considerations that could refine interpretation and guide future work. First, transfusion was treated as a binary exposure (any vs. none), which may obscure clinically relevant gradients in transfusion burden. In the TRACS randomised trial, the number of transfused units was an independent risk factor for 30-day complications or death, supporting modelling transfusion burden beyond an ‘any vs. none’ indicator 2. Incorporating transfusion quantity (continuous or categorical burden) into mediation frameworks could clarify whether the mediated effect is concentrated among patients receiving higher volume transfusion and may better align inference with bedside decision-making. Second, residual confounding by patient vulnerability and the underlying causes of anaemia may account partly for the anaemia–mortality association attributed to transfusion. Frailty, which is often unmeasured in registry data sets, was associated with substantially higher surgical and mid-term mortality after cardiac surgery even after adjustment in included studies 3. If frailty and related factors such as malnutrition or inflammatory phenotypes influence transfusion likelihood and mortality risk, incomplete adjustment could overstate the causal role of transfusion in mediation analyses. Future registry linkages or prospective cohorts that capture frailty measures and key biological correlates of anaemia could strengthen causal interpretation. Third, timing likely matters. Groenewegen et al. defined transfusion as any transfusion exposure during hospital stay, yet intra- and postoperative transfusions may reflect different clinical scenarios. In a coronary artery bypass graft cohort, postoperative (but not isolated intra-operative) transfusion was associated with higher odds of 30-day readmission and heart failure, suggesting that late transfusion contexts may carry distinct prognostic implications 4. Stratifying analyses by transfusion timing (intra-operative vs. postoperative) may help distinguish potentially modifiable surgical haemostasis issues from downstream ICU complications and could identify more actionable intervention points. Groenewegen et al. provide important evidence that transfusion explains a substantial proportion of the anaemia–mortality association after cardiac surgery. Modelling transfusion burden, improving adjustment for frailty and related patient factors and accounting for transfusion timing may yield more clinically actionable and causally robust insights for optimising peri-operative blood management.