In their article, Prediction of Intraoperative Transfusion Requirements During Orthotopic Liver Transplantation and their Influence on Postoperative Patient Survival, Cywinski et al.1 present the results of a retrospective analysis of 804 adult liver transplants completed between 2001 and 2010 at the Cleveland Clinic. The objectives of the study were 2-fold: (1) To determine models to predict the volume of blood (both transplanted and salvaged) required by patients undergoing a liver transplant; and (2) to evaluate the impact of the volume of blood transfused on patient morbidity and mortality. Like multiple previous studies,2–6 the authors were able to demonstrate a relationship between increasing patient mortality and increased transfusion. Of course, it is impossible to determine whether sicker patients simply required more blood or whether transfusing more blood actually increased the risk of death. Importantly, despite a detailed statistical analysis of blood consumption to analyze both total transfusion requirements and markers for patients who might be expected to require large volumes of blood, the resulting models had weak predictive power. In other words, the study produced “negative” results.
Why should a study that fails to accomplish its primary goal (i.e., is unable to predict how much blood a patient will require during a liver transplant) be published? It is well known that scientific publication has a positive bias. Studies that report significant results are more likely to be published than those that find no experimental effect.7 Failing to observe an effect, of course, does not prove that an effect does not exist. For this reason, negative studies are often viewed as underpowered or otherwise less convincing than those that show an identifiable effect. In this article, the fact that the authors were unable to develop a model that reliably predicted transfusion requirements is what is important! This article has important implications for the clinical and the operational practices of managing blood products during a liver transplant.
The authors conducted 2 sets of analyses to predict transfusion requirements for liver transplant patients. They built a multivariate regression model to explain variation in the total demand for blood based on a set of preoperative recipient, donor, and procedure characteristics. Despite being statistically significant, the resulting predictive model explained only 22% of the total variability in transfusion requirements, which the authors found to be clinically insignificant. The authors then employed logistic regression to identify factors that would lead to large transfusion requirements (either >20 or >30 units). These models, while successful at finding factors associated with greater bleeding, had similarly weak predictive power (32% and 11% of patients, respectively) to categorize individuals within the dataset. A secondary analysis, focused on factors that lead to death following transplantation, demonstrated that increased blood use led to increased mortality shortly after surgery. However, since causality could not be determined, the question remains as to whether sicker patients required more blood or more blood made patients sicker.
The most important message is the authors’ conclusion, based on analysis of the 804 surgeries and a review of the existing literature, that there is no “reliable algorithm for prediction of intraoperative blood product requirements, even (within) the setting of a single institution.” This means that since it is impossible to predict ahead of time which patients are going to require massive transfusions, it is important to treat all patients as if they will require massive transfusion. From a practical standpoint, this means that there must be sufficient blood products to resuscitate patients who require large transfusions (i.e., patients the “high” end of the demand distribution). Indeed, the study data in the Cywinski et al.1 article shows that a reserve of 20 units would be sufficient only for the 80th percentile of patient demand. Twenty percent of the patients in their study required >20 units red blood cells (RBC), and 10% required >30 units RBC, which is more than twice the median demand of 14 units. Their study also shows that even though cell salvaging can provide approximately 40% of the total RBC requirements, a robust transfusion service, an effective protocol for massive transfusion, and clear lines of communication between transfusion centers and their blood product supplier are crucial factors in ensuring patient safety and maximizing the potential for a positive outcome.
“Hope for the best, but prepare for the worst” appeared frequently in letters one of our grandfathers wrote from the trenches of Belgium and France during the First World War. The Cywinski et al.1 article provides the same message to patients, clinicians, and health care managers involved in liver transplantation: Since it is impossible to determine ahead of time which patients will require large transfusions, the only safe strategy is to treat all patients as if they will require massive transfusion. Everyone involved in the surgery, and those providing the blood products to support the surgery, must be prepared for things to go awry, regardless of how reassuring the preoperative characteristics of the recipient or the donor may appear. For researchers, the Cywinski et al.1 article provides a challenge. Are there other factors, perhaps not part of the currently collected dataset, which might help to explain why some patients require more intraoperative blood products than others? Sadly, for patients, the Cywinski et al.1 suggests that there is little that can be done to help understand their individual risk preoperatively, other than to indicate that the more blood required during surgery, the poorer the outcomes are likely to be.
As the grandfather who survived the war noted, hope is more likely to prevail when prepared for the worst.
Edward C. Nemergut is the Section Editor of Graduate Medical Education and Transplantation Anesthesiology for Anesthesia & Analgesia. This manuscript was handled by Dr. Steven L. Shafer, Editor-in-Chief, and Dr. Nemergut was not involved in any way with the editorial process or decision.
Name: John Blake, PhD.
Contribution: This author helped write the manuscript.
Attestation: John Blake approved the final manuscript.
Name: Edward C. Nemergut, MD.
Contribution: This author helped write the manuscript.
Attestation: Edward C. Nemergut approved the final manuscript.
1. Cywinski JB, Alster JM, Miller C, Vogt DP, Parker BM. Prediction of intraoperative transfusion requirements during orthotopic liver transplantation and their influence on postoperative patient survival. Anesth Analg. 2014;118:428–37
2. de Morais BS, Sanches MD, Ribeiro DD, Lima AS, de Abreu Ferrari TC, Duarte MM, Cançado GH. [Association between the use of blood components and the five-year mortality after liver transplant]. Rev Bras Anestesiol. 2011;61:286–92
3. Benson AB, Burton JR Jr, Austin GL, Biggins SW, Zimmerman MA, Kam I, Mandell S, Silliman CC, Rosen H, Moss M. Differential effects of plasma and red blood cell transfusions on acute lung injury and infection risk following liver transplantation. Liver Transpl. 2011;17:149–58
4. Warner P, Fusai G, Glantzounis GK, Sabin CA, Rolando N, Patch D, Sharma D, Davidson BR, Rolles K, Burroughs AK. Risk factors associated with early hepatic artery thrombosis after orthotopic liver transplantation-univariable and multivariable analysis. Transpl Int. 2011;24:401–8
5. Hendriks HG, van der Meer J, de Wolf JT, Peeters PM, Porte RJ, de Jong K, Lip H, Post WJ, Slooff MJ. Intraoperative blood transfusion requirement is the main determinant of early surgical re-intervention after orthotopic liver transplantation. Transpl Int. 2005;17:673–9
6. Dunn LK, Thiele RH, Ma JZ, Sawyer RG, Nemergut EC. Duration of red blood cell storage and outcomes following orthotopic liver transplantation. Liver Transpl. 2012;18:475–81
7. Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR. Publication bias in clinical research. Lancet. 1991;337:867–72