Thank you for the opportunity to respond, and warm thanks to Drs. Gross, Shander, Waters, and Roth for their interest in our study1and their valuable comments. We emphasize that we did not study the short-term effects of blood transfusion and so are not able to respond to comments relating to short-term effects. We believe that it is important not to confuse short- and long-term effects.
It is an overstatement to suggest that all doses of blood, under all circumstances, have been consistently associated with long-term harm. It is contradicted by two of the references that Gross et al. provided.2,3The 2002 study by Engoren et al. was indeed a landmark, but it showed no relationship between total 5-yr mortality and intraoperative transfusion (relative risk, 1.2; 95% CIs, 0.6-1.7, P = 0.53). It was only postoperative transfusion or a combination of intraoperative and postoperative transfusions that had a significant association. More recently, Engoren et al. 4have studied 993 patients undergoing cardiac valve surgery and concluded that blood transfusion was not associated with reduced long-term survival. They also studied 2,213 intensive care patients5and found that blood transfusion was associated with improved late survival. Surgenor et al. 3studied 3,254 cardiac surgery patients; between 6 months and 5 yr after surgery “there was no association between exposure to 1 or 2 units of blood and survival” (hazard ratio, 1.06; 95% CIs, 0.91-1.24, P = 0.431).
We make no apologies for designing our study to investigate the dose-dependent effects of allogeneic transfusion on life expectancy. We went into the study with an open mind. We observed that the transfusion of 1-2 units of allogeneic blood products was associated with a 1.00 relative risk of death, with 95% CIs between 0.7 and 1.4. The relative risk for patients who received 3-6 units was 0.98 (0.6-1.4). We concluded that there was not a strong association between moderate transfusion of blood or blood products and long-term mortality. It would be scientifically invalid to conclude otherwise.
To consider the power limitations of our study, it may be worthwhile to estimate the effect on absolute mortality that our data allow us to exclude. We observed an absolute mortality rate of 1.8% per year; our CIs allow us to exclude a 40% increase in relative risk or a 30% decrease in relative risk. Therefore, our study allows us to estimate that a 1- to 2-unit transfusion is associated with not more than a 0.7% increase, or 0.5% decrease, in the annual mortality rate.
Neither fresh frozen plasma nor platelets are truly acellular. We included them in our count of donor exposures because both products contain leukocytes, are allogeneic, and are associated with serious acute transfusion-related immune-mediated effects. Fresh frozen plasma is implicated in the pathogenesis of transfusion-related lung injury, the most common cause of transfusion-related death in the United States.6Presumably, Gross et al. are not suggesting that these products can be considered free of long-term risk. Nevertheless, it is a proper scientific scepticism to question this aspect of our study design. Consequently, we have further analyzed our data, counting only red cell containing transfusions: the relative risk of long-term mortality in patients receiving 1-2 units of red cells was 0.98 (95% CIs, 0.7-1.4).
Gross et al. may be underestimating the importance of anemia as a predictor of long-term survival. The Atherosclerosis Risk in Communities study7of 14,410 people between 45 and 64 yr old in the general United States population showed that anemia was associated with a 1.65 relative risk of mortality. This is a considerable risk factor; it is of similar magnitude to that associated with diabetes or heart failure. This magnitude of risk has been demonstrated in many other studies. van Straten et al. 8studied 10,025 patients undergoing coronary artery graft surgery and demonstrated a relative risk for late mortality of 1.64 in patients with mild anemia. Riva et al. 9studied 7,536 Europeans demonstrating an association between mild anemia and late mortality of 1.86. We suggest that any study of long-term survival that does not take baseline anemia into account may be flawed. We note that some studies on the association between long-term mortality risk and blood transfusion been published without taking preoperative anemia into account. We would be interested in seeing a reanalysis of their results after this factor has been included.
We thank Gross et al. for recognizing an error in our results. One hundred seventy-three patients, not 183, died who had been transfused and who did not have a history of cancer. This typographical error was the authors' responsibility not the publisher. Nevertheless, the preponderance of cancer-related deaths in the nontransfused group remains unlikely to have occurred by chance. We note this simply as an observation; diagnoses recorded on death certificates were never part of our original hypothesis and cannot be used in the inferential process. We do not agree with Gross et al. that it is scientifically valid to include a long-term outcome (death with a cancer-related diagnosis) in a predictive model based on the perioperative physical status of the patient.
Dr. Roth's suggestion that blood transfusion and extracorporeal circulation may have synergistic adverse effects is not unreasonable. Certainly, this is one explanation for the observation that cardiac surgery is the only clinical setting where adverse transfusion-related immunomodulation effects have been conclusively demonstrated.10Our study data does not help to clarify this suggestion, however, as most of our transfused patients were exposed to extracorporeal circulation.
None of this is any reason to promote the use of blood. The short-term risks are compelling, as are the huge costs and the ever diminishing resource. However, we should not be misinforming patients who have survived more than 2 months after coronary artery grafts that they have a serious risk of premature death as a result of a moderate transfusion of blood products.
*Sir Charles Gairdner Hospital and University of Western Australia, Perth, Western Australia. email@example.com