Adverse blood transfusion outcomes: establishing causation

Area of Science:

• Transfusion Medicine
• Medical Statistics
• Clinical Causation

Background:

• Allogeneic red blood cell (RBC) transfusions are standard medical practice, based on assumptions of efficacy and safety.
• Growing evidence questions the efficacy and highlights risks of RBC transfusions in various clinical settings.
• Causation of transfusion-related adverse outcomes has not been definitively established through randomized clinical trials (RCTs).

Purpose of the Study:

• To critically analyze the evidence for causation of adverse clinical outcomes following RBC transfusions.
• To explore methodologies for establishing transfusion-related harm, particularly when RCTs are not feasible.
• To advocate for a precautionary approach to RBC transfusion due to accumulating risk data.

Main Methods:

• Examination of the nature and definition of causation in transfusion medicine.
• Application of deductive deterministic methods (e.g., modified Koch's postulates) for established transfusion hazards.
• Utilizing inferential probabilistic methodology (e.g., modified Bradford Hill criteria) for observational data linking RBC transfusions to adverse outcomes.

Main Results:

• While some transfusion hazards are well-established, causation for others, particularly immunomodulation and storage lesion effects, relies heavily on observational data.
• Modified Bradford Hill criteria can strengthen the case for RBC transfusion as a causal factor in adverse outcomes when RCTs are lacking.
• The evidence suggests RBC transfusions may be causal for adverse clinical outcomes in specific contexts.

Conclusions:

• A critical evaluation of existing evidence is needed to establish causation for RBC transfusion-related harms.
• Inferential probabilistic methods applied to observational studies are crucial for understanding transfusion risks.
• A greater precautionary principle should guide RBC transfusion practices, as RCTs may not always be feasible or ethically justifiable.