UCSF

There is good evidence that blood transfusion practice varies by time and geography, and often departs from evidence‐based guidelines. On the other hand, interventions to change physician practice around the ordering of blood transfusions can be successful. The challenge is therefore to develop reliable evidence to support rational transfusion practice. Both observational studies and randomized clinical trials have strengths and limitations. Observational studies are cheaper and usually larger with concomitant improved statistical power, and may be more representative by including a wide range of patients and hospitals in the routine setting unaffected by a research protocol. Weakness of observational studies include confounding by unmeasured factors; indication bias, namely overestimation of adverse outcomes by the fact that transfused patients tend to be sicker than similar patients who did not receive transfusion; and the fact that available data may not suit the research question. Randomized clinical trials have the advantage of minimizing the effects of confounding and bias by virtue of the randomization process. Limitations of randomized trials include their expense which reserves them for specific research questions and may limit statistical power; and ethical problems in study design such as inability to utilize a placebo by withholding transfusion. Finally, inclusion and exclusion criteria can limit the generalizability of conclusions by restricting enrollment to a set of patients that differ from those seen in general medical practice. In this review, I shall attempt to illustrate the complementary contributions and limitations of observational epidemiological studies and randomized clinical trials in understanding several transfusion‐related adverse outcomes: mortality; TRALI; TACO; and the RBC storage lesion. I conclude that evidence for the efficacy and adverse outcomes of blood transfusion needs to be obtained from a balanced combination of observational studies and clinical trials.