Time for a Fresh Approach to Examining Factors Associated With Red Blood Cell Transfusion Outcome

Red blood cell (RBC) transfusion can be a life-saving intervention in hemorrhaging or severely anemic patients. However, RBC transfusion is also associated with adverse outcomes, including hemolytic and febrile nonhemolytic transfusion reactions, transmission of viral and bacterial pathogens, transfusion-related acute lung injury (TRALI), transfusion-associated circulatory overload, and transfusion-related immunomodulation, a term that describes the immunosuppressive properties of RBCs, which may predispose to hospital-acquired infection.1 In addition, RBC transfusion is associated with organ dysfunction and increased mortality in surgical patients and the critically ill.2 These risks have created a difficult decision-making paradigm for physicians who increasingly view RBC transfusion as a “double-edged sword.” Several putative mechanisms contribute to the adverse effects of RBC transfusion. These include scavenging of recipient nitric oxide in the microcirculation, alterations in normal recipient immune responses, and increased production of inflammatory cytokines in the recipient including: tumor necrosis factor α, interleukin-6, and interleukin-8.1 A great deal of attention has been paid to the RBC “storage lesion” because RBCs are approved by the Food and Drug Administration to be stored for 42 days with refrigeration. The storage lesion is now well described, with RBCs showing reduced deformability, increased swelling, and loss of normal geometry during storage. RBCs also have progressively decreased concentrations of adenosine triphosphate and 2,3-diphosphoglycerate and increased degradation with release of microparticles and free hemoglobin into the supernatant. Finally, damaged RBCs are cleared within 1–2 hours by the recipient’s reticuloendothelial system, which leads to increased levels of nontransferrin bound iron.1,3 The majority of studies that have examined factors affecting RBC transfusion outcome have been observational. This is in part because serious harm from RBC transfusion is relatively uncommon, and large numbers of transfused patients are needed to study outcomes like mortality. One of the first observational studies to examine the impact of RBC storage duration on clinical outcome was from Koch et al,4 who found increased mortality and higher complication rates in cardiac surgery patients who received “old” RBCs. The median storage duration of old RBCs in this study was 20 days. A 2012 meta-analysis that examined 21 studies (18 observational and 3 randomized control trials [RCTs]), further supported this finding, noting that old RBCs were associated with an increased risk of death when compared with “fresh” RBCs; odds ratio, 1.16 (95% confidence interval [CI], 1.07–1.24).5 This analysis included more than 400,000 patients across a variety of settings, including both children and adults. These observations instigated several RCTs designed to study the impact of RBC storage duration on patient outcome. In 2015, 2 RCT groups, age of transfused blood in critically ill adults (ABLE) and effects of red-cell storage duration on patients undergoing cardiac surgery (RECESS), published their findings. In both studies, there was no difference in mortality between patients receiving old and fresh RBCs.6,7 A follow-up meta-analysis, which included 5000 patients from 12 RCTs, similarly did not show any difference in mortality between patients receiving old and fresh RBCs.8 The largest RCT to date, effect of short-term versus long-term blood storage on mortality after transfusion (INFORM), included more than 20,000 patients at 6 clinical sites in 4 countries and again found no difference in mortality between patients receiving old and fresh RBCs.9 The main criticism of these RCTs is that they did not compare truly old RBCs (35–42 days) against fresh RBCs, and also there was substantial heterogeneity in how old and fresh were defined.10 In ABLE, fresh RBCs had a mean storage duration of 6 days, whereas old RBCs had a mean storage duration of 22 days; and in RECESS, fresh RBCs had a mean storage duration of 8 days, while old RBCs had a mean storage duration of 28 days. In our opinion, it is unlikely that a future RCT will compare transfusion of truly old RBCs (35–42 days) against either fresh or moderate-aged RBCs because of serious ethical concerns that would preclude this type of trial. There are a number of important lessons that can be taken from the RBC storage duration debate. First, there appears to be no clinical benefit to transfusing fresh (<10 days) RBCs. Second, observational studies are a good starting point for exploring clinical questions because they are inexpensive and relatively easy to conduct; however, they are also prone to information bias, residual confounding, and frequently, their results can be skewed by a small number of individuals in a cohort with a particular outcome.11 For these reasons, RCTs remain the gold standard in clinical research, and providers should be prudent to change their practice even when numerous observational studies support a particular finding. Recently, several observational studies examined the effects of donor age and sex on RBC transfusion outcome. Age and sex are potentially important donor variables because even with leukoreduction, packed RBCs contain approximately 1 million donor leukocytes and some donor antibodies, and donor age is associated with solid organ transplant outcomes.12 The Table summarizes these studies and their main findings. In 2011, Middelburg et al13 published a retrospective cohort study from a single center in the Netherlands. This study included 11,211 patients who received allogeneic blood products, with RBCs accounting for the majority of transfused products (73,293 of 96,009 units). In this cohort, 3806 patients received blood products from a single-sex donor. After adjusting for confounders with propensity score analysis, the authors found that male recipients 1–55 years of age who received female donor blood products had increased mortality at several follow-up intervals; hazard ratio (HR) = 2.9 (95% CI, 1.3–6.5) for 90-day mortality and HR = 3.1 (95% CI, 1.5–6.3) for 6-month mortality. When isolating their analysis to male patients 1–55 years of age who received only female donor RBCs, the association with mortality remained significant, with an HR of 2.6 (95% CI, 1.1–6.2) for 90-day mortality. The authors hypothesized that antibodies against male HY antigens could explain worse outcomes for males who received female donor RBCs.Table.: Recent Observational Studies Examining the Impact of Donor Sex and Age on RBC Transfusion OutcomeTwo subsequent studies that focused exclusively on cardiac surgery patients were published in 2016 with conflicting results. Desmarets et al14 investigated a cohort of 2715 cardiac surgery patients who received RBCs at 2 hospitals in France from 2007 to 2011. In this study, patients transfused with sex-mismatched RBC units had increased 1-year mortality; HR = 2.28 (95% CI, 1.67–3.12). However, the effect was no longer significant after adjusting for confounders with multivariate regression; HR = 1.22 (95% CI, 0.81–1.82). In a single-center Swedish cohort study of 9907 cardiac surgery patients with 12 years of follow-up, there was increased mortality in patients who received sex-mismatched RBCs; HR = 1.083 (95% CI, 1.028–1.140).15 In contrast to the findings of Middelburg et al,13 who reported increased mortality at an early follow-up interval, the association between sex mismatch and mortality was not significant within the first year after surgery; HR = 1.11 (95% CI, 0.99–1.29) in this study. This discrepancy highlights the importance of determining an appropriate follow-up interval for future studies examining donor-specific variables. RBC donor age has also been explored as a risk factor for transfusion outcome in several observational studies. One retrospective cohort study of trauma patients compared transfused RBCs from old and young donors and examined the incidence of TRALI.16 In this study, old donors were defined as >63 years of age. Of the 212 patients included in the analysis, 48% received at least 1 RBC unit from an old donor. After multivariate regression analysis, which took into account injury severity and the overall number of transfused RBCs, the authors found no increased incidence of TRALI in patients receiving RBCs from old donors; odds ratio for TRALI = 0.98 (95% CI, 0.5–1.89). The generalizability of this study was limited by its small sample size and the fact that many patients received blood from both old and younger donors. A retrospective cohort study published by Vasan et al17 in 2016 matched patients from a Scandinavian database into 3 cohorts: those who received blood exclusively from donors <25 years of age, those who received blood from donors 25–50 years of age, and those who received blood from donors >50 years of age. This study included 136,639 patients, who were matched for the total number of RBC transfusions. The authors found no difference in 30-day or 1-year mortality across the 3 groups: HR for 30-day mortality = 1.02 (95% CI, 0.97–1.08) for donors <25 years of age, HR for 30-day mortality = 1.00 (95% CI, 0.97–1.04) for donors >50 years of age, with donors 25–50 years of age as the reference. However, as the authors noted in their discussion, the study design and matching process limited the analyzed patients to those who received few total transfusions, with the median number being 1 unit (range 1–5 units). Thus, any potential dose response may have been missed. Chassé et al18 performed a retrospective cohort study at 4 Canadian sites over a 7-year period from 2006 to 2013, examining the effects of both RBC donor sex and age on mortality. This study, published in 2016, included 30,503 patients from 80,755 donors and tracked survival from the time of RBC transfusion to the end of the study period. All-cause mortality for patients who received RBC transfusion during the study period was 43%. Unlike the Vasan et al17 study, the Chassé et al18 study analyzed patients who received RBCs from both male and female donors. This design allowed for greater evaluation of the cumulative effects of RBC transfusion, with a median number of 3 transfused units (IQR = 2–6) for the analyzed cohort. Interestingly, the authors found associations between both donor sex and age and mortality. RBCs from female donors were associated with increased mortality when compared with male donors; HR = 1.08 (95% CI, 1.06–1.09). Also, patients who received RBCs from younger donors had increased mortality compared with those who were transfused with RBCs from old (40–49 years of age) donors; HR = 1.08 (95% CI, 1.06–1.10) for donors 17–19 years of age and HR = 1.06 (95% CI, 1.04–1.09) for donors 20–29 years of age. Finally, Edgren et al19 studied >950,000 patients between 2003 and 2012 using the Scandanavian Donations and Transfusions database to determine the impact of donor sex and age on RBC transfusion outcome. In this study, the authors created several Cox proportional hazards models with the exposure variables (sex-mismatched transfusion and transfusion from different-aged donors) modeled in a number of ways. Critically, the study authors controlled for the total number of transfusions, which the Chassé et al18 study did not. Using this methodology, the authors found no association between donor sex mismatch or donor age and mortality, and they concluded that the Chassé et al18 study was confounded. The confounding was related to the fact that patients in the Chassé et al18 study who had received more transfusions were also more likely to receive sex-mismatched transfusion or transfusion from donors with extremes of age. To date, no RCT has compared sex-matched and sex-mismatched RBC transfusion or RBC transfusion from different-aged donors. This type of study is logistically challenging because of RBC inventory issues and because many patients have received previous transfusions from donors of the opposite sex or different ages, which could confound findings. The Edgren et al19 study was sufficiently large and used robust methodology to control for confounding, but it also may not be generalizable to all patient populations, and further investigation of donor sex and age is warranted in our opinion. A translational approach might also be helpful in answering questions about donor factors affecting RBC transfusion outcome. One worthwhile line of investigation may be to identify specific harmful molecules present in the supernatant. This approach is attractive because it could allow for “bad” RBC units to be identified during storage or at the time of collection and to be eliminated from the blood supply. Alternatively, RBC units with high concentrations of harmful molecules could be preferentially filtered or washed before transfusion. The molecular content of RBC supernatant has only recently been explored in detail. RBC supernatant contains approximately 400 proteins, many of which change in concentration during the 42-day storage duration and depend on the degree of leukoreduction.20 Several proteins in the supernatant have already been implicated as potentially contributing to lung injury, including peroxiredoxin 6, which oxidizes phospholipids.20 There are no large-scale proteomic or lipidomic studies evaluating RBC supernatant from multiple donors at this time, but as these methods become more cost efficient, it may become feasible to correlate levels of particular proteins or lipids in the supernatant with clinical outcomes in transfused patients. In summary, harmful effects of RBC transfusion have been increasingly recognized, but identification of donor- or unit-specific factors that are associated with harm are lacking. Thus far, observational studies have failed to convincingly identify modifiable risk factors that improve RBC transfusion outcome. We believe that a translational approach will be helpful in elucidating specific mechanisms by which stored RBCs cause harm. Further, future observational studies examining donor- or unit-specific risk factors should include very large numbers of patients and robust controls for confounding so that the mistakes of the past can be avoided. DISCLOSURES Name: Michael Mazzeffi, MD, MPH, MSc. Contribution: This author helped design, analyze, and write the article. Name: Evan Chriss, MD. Contribution: This author helped design, analyze, and write the article. Name: Kenichi Tanaka, MD, MSc. Contribution: This author helped design the study, analyze the data, and write the article. This manuscript was handled by: Marisa B. Marques, MD.