Perioperative blood transfusionsBMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h3153 (Published 15 June 2015) Cite this as: BMJ 2015;350:h3153
- Stavros G Memtsoudis, clinical professor of anesthesiology and public health; attending anesthesiologist and senior scientist
- 1Department of Anesthesiology, Hospital for Special Surgery, Weill Cornell Medical College, 535 East 70th Street, New York, NY 10021, USA
- Correspondence to: S G Memtsoudis
The debate regarding the impact of blood transfusions on outcomes spans decades, but many questions, especially concerning any associated cardiovascular risks, remain unanswered. Despite several high profile studies suggesting that a restrictive transfusion strategy might be superior to more liberal approaches,1 2 3 firm consensus is far from established. For example, the most recent meta-analysis published on this topic, with 31 trials and 9813 patients, found no difference in overall morbidity and mortality between liberal and restrictive strategies.4 Importantly, the statistical model in that study was unable to reach the “required information size” for the analysis of mortality and myocardial infarction, a consistent limitation of rare event research, even when data are pooled from a relatively large number of clinical trials.
It is within this context that a linked article by Whitlock and colleagues shows the value of large population based datasets.5 Analyzing data from over 1.5 million patients admitted for surgery to 346 hospitals in the United States, the authors found that receipt of as little as one unit of blood was associated with a 2.33-fold increase in the odds of developing a stroke or myocardial infarction. Although the study is unable to prove a causal relation between transfusions and adverse outcomes, the findings add new information as they raise the possibility of a link on a population level. The authors’ conclusions are further supported by a plausible mechanism of injury connecting blood transfusions to the reported adverse events.5
Results from previous analyses of administrative data are congruent with the new study. One study, which used data from the National Surgical Quality Improvement Program on over 6 000 000 surgical patients, raised the possibility of harm among those who required more than four units of blood.6 Danninger and colleagues reported an association between blood transfusions and adverse events in 530 089 patients undergoing major joint arthroplasty within the Premier Perspective dataset.7 They concluded that the use of blood transfusions in this setting could explain 9.8% of all complications.
So, what can population based datasets add to our knowledge and where do they fall short? Use of large national datasets has increased over the past decade, encouraged by methodological advances and the large amount of patient information available in these databases, to help evaluate rare or moderately rare outcomes. Cited advantages include the use of “real world information” collected from various practice settings, which are not restricted by the often artificial inclusion and exclusion criteria of randomized controlled trials. As a result, population based analyses often have a higher external validity than randomized controlled trials.
Database research, however, is limited by several factors.8 Causal relations cannot be established; only associations can be identified. Important clinical details are often missing from databases, including patients’ hemodynamic parameters, hemoglobin concentrations, symptoms, use of transfusion triggers, and the actual clinical context, which might include ongoing blood loss. All of these factors are relevant to the association between transfusion and outcomes. Additional uncertainties include the contribution of comorbidities and surgical invasiveness. Comorbidities and procedures defined by the international classification of diseases, ninth revision-clinical modification (ICD-9-CM) coding systems rarely account for subtle but clinically important differences between individual patients in the same categories.
Whitlock and colleagues made valiant efforts to overcome these limitations, to minimize bias, and to measure the likely extent of residual confounding.5 But we still cannot conclude with total confidence that transfusion of a single unit is responsible for serious cardiovascular complications. The possibility still remains that any transfusion is simply a marker for an unmeasured (or unmeasurable) variable such as a clinical instability, more extensive surgery, or blood loss or even a clinician’s “concern threshold” for giving a transfusion to a patient with certain comorbidities or characteristics.
The message from this and other studies is one of caution and restraint when it comes to the use of perioperative blood transfusions. Even though definitive evidence remains elusive, the research to date supports the working hypothesis that a liberal transfusion strategy is associated with worse outcomes than a more restrictive strategy. Studies such as this add incremental but important evidence. They imply that when consistent results from clinical trials suggesting a negative effect of blood transfusions are tested in large population based datasets, the conclusions point in the same direction. Even without knowing if blood loss, anemia, or subsequent transfusions are to blame for adverse cardiovascular outcomes, clinicians can and should focus on developing, testing, and implementing measures to reduce blood loss and anemia and so minimize the need for perioperative transfusions. Effective and evidence based interventions already exist, including the use of antifibrinolytics and other blood management techniques.9 10
As neither population based datasets nor clinical trials give definitive answers, we might look instead to a third way: specialized data collection constructs such as multi-institutional registries that combine the benefits of large numbers of patients with detailed and relevant clinical information.
In the meantime, prudent clinicians should practice restraint when contemplating blood transfusions and also use strategies to decrease the risk of bleeding and anemia. Finally, we must avoid a dogmatic approach to transfusions. There will always be patients and situations when transfusions are clearly needed, and the risks and benefits might be much more dependent on individual factors than is apparent from available studies.
Cite this as: BMJ 2015;350:h3153
Competing interests: I have read and understood the BMJ Group policy on declaration of interests and declare the following interests: I am an unpaid consultant for B Braun.
Provenance and peer review: Commissioned; not externally peer reviewed.