Distribution of survivors and nonsurvivors according to the amount of blood transfused.
Velmahos GC, Chan L, Chan M, Tatevossian R, Cornwell III EE, Asensio JA, Berne TV, Demetriades D. Is There a Limit to Massive Blood Transfusion After Severe Trauma?. Arch Surg. 1998;133(9):947-952. doi:10.1001/archsurg.133.9.947
To examine the hypothesis that the futility of short-term care for trauma patients requiring emergency operation can be determined based on the number of units of blood transfused and associated risk factors.
A 4-year retrospective review of a cohort of critically injured patients who underwent an emergency operation.
A large-volume, academic level I, urban trauma center.
One hundred forty-one consecutive patients received massive blood transfusions of 20 U or more of blood during preoperative and intraoperative resuscitation (highest, 68 U). There were 43 survivors (30.5%) and 98 nonsurvivors (69.5%).
Main Outcome Measures
The number of blood units transfused did not differ between survivors and nonsurvivors (mean±SD, 31±11 vs 32±10; P=.52). Stepwise multiple regression analysis identified 3 independent variables associated with mortality: need for aortic clamping, intraoperative use of inotropes, and intraoperative time with a systolic blood pressure of 90 mm Hg or less. However, blood usage was not different among the subgroups of patients who had 1 or more of these risk factors. When patients were stratified according to the amount of massive blood transfusion (20-29, 30-39, 40-49, and 50-68 U), the incidence of risk factors was not different across the 4 subgroups. Survival in the presence of risk factors was not affected by the amount of blood transfused.
Although mortality among critically injured patients requiring operation and massive blood transfusion can be correlated with independent risk factors, discontinuation of short-term care cannot be justified based on the need for massive blood transfusion of up to 68 units.
MASSIVE BLOOD transfusion (MBT) is most commonly defined as complete replacement of a patient's blood volume within a 24-hour period.1 Despite significant complications and logistical problems associated with MBT, results of many studies have demonstrated significant survival rates that justify its use in trauma.2,3 However, none of these studies focused on the early postinjury period spent in the emergency department and operating room (OR). With advances in life support and rapid prehospital triage, a greater number of critically injured patients are transferred to the OR and are offered MBT. It is not unusual for trauma surgeons to exhaust the blood bank supplies in a seemingly futile attempt to resuscitate critically injured patients in the acute stage.1 In trying to save these patients, it is possible that the care of other patients is compromised by depleting valuable resources.
In this study, we evaluated the initial critical period after injury in patients requiring surgery. Intraoperative surgical decision making occasionally includes whether resuscitation efforts should be continued in the face of ongoing exsanguination. Such decisions are expected to come under increasing scrutiny in a managed care–driven medical system. Limits on resource-depleting MBT may be justified in critically ill patients if a point of futility can be identified.4
Our anecdotal clinical experience indicates that survival is not uncommon with blood transfusion of less than 20 U. For this study, we defined MBT as the use of 20 U or more of blood during the preoperative and intraoperative period. We tried to answer the following questions: (1) Should there be an absolute limit on MBT based on subsequent outcome? (2) Are there any associated risk factors that can reliably predict mortality in patients undergoing MBT?
Included in this study were 141 consecutive (from January 1993 to January 1997) trauma patients who required an emergency operation and MBT. A total of 56 data elements addressing demographic and injury severity characteristics, hemodynamic and laboratory values, interventions, and outcome variables were collected from 3 sources: medical records, blood bank records, and the trauma registry. One unit of blood was defined as 1 U of whole blood or 1 U of packed red blood cells. Transfusion was analyzed only during the emergency department and OR period. Massive blood transfusion was defined as transfusion of more than 19 U of blood during that period.
Comparison between survivors and nonsurvivors was performed by the 2-sample Student t test for mean values of continuous variables and by χ2 or Fisher exact test for categorical variables. Variables with M.10 and complete data in more than 90% of cases were entered into stepwise logistic regression analysis to identify independent risk factors associated with survival. Subsequently, the study population was divided into 4 groups according to the level of MBT: 20 to 29, 30 to 39, 40 to 49, and 50 to 68 U of blood. There were 70, 45, 15, and 11 patients in these groups, respectively. The relationship of the predetermined risk factors with the amount of MBT across the 4 groups was tested by the Student t test for 2-group comparisons and by analysis of variance for comparisons of 3 or more groups.
Of 141 patients, 43 (30.5%) survived and 98 (69.5%) died. The survival rates in the 4 subgroups of transfusion were 33%, 29%, 27%, and 27%. The highest levels of MBT among survivors were 68 U in 1 patient and 64 U in 2 other patients. Demographic and injury characteristics of the whole group are given in Table 1. Among these variables, only age was significantly different for survivors and nonsurvivors.
Comparing physiologic status and incidence of interventions during resuscitation in the emergency department, only Glasgow Coma Scale score was shown to be different between patients who lived and those who died (Table 2). Twenty percent of survivors, but 42% of nonsurvivors, had Glasgow Coma Scale scores less than 12 (P=.02). When hemodynamic and laboratory values and incidence of therapeutic interventions were evaluated during the time in the OR, 11 variables were significantly different between survivors and nonsurvivors (Table 3): aorta clamping for control of blood pressure (BP), use of inotropic agents, amount of time with a systolic BP of 90 mm Hg or less, time in the OR, temperature below 34°C, urine output, pH level of 7.0 or less, PO2 to fraction of inspired oxygen of 150 mm Hg or less, PCO2 of 50 mm Hg or more, serum potassium level of 6 mmol/L or more, and serum calcium level of 2 mmol/L or less (≤8 mg/dL). However, the incidence of missing values for the latter 8 variables was more than 10%, and, therefore, they were not selected for the multivariate analysis. The difference in OR time was believed to be because of the expected shorter OR stay for patients who died soon after admission; for that reason, OR time was not selected either.
Age, Glasgow Coma Scale score, aorta clamping, use of inotropes, and time with BP less than or equal to 90 mm Hg were variables entered into the stepwise logistic regression analysis. The continuous variables were transformed to dichotomous variables. Age was divided into 55 years and older vs 54 years and younger; Glasgow Coma Scale score into 12 or more vs less than 12; and duration of hypotension into 90 minutes or less vs more than 90 minutes. A total of 119 patients with complete data for the selected variables were used in the analysis. Aorta clamping, use of inotropes, and time with a BP of 90 mm Hg or less were found to be independent risk factors for mortality.
Figure 1 presents the distribution of survivors and nonsurvivors according to the amount of blood transfused. A comparison of mortality and its associated risk factors among the 4 groups, defined by amount of MBT, is shown in Table 4. None of these variables were different among groups. Furthermore, the mean±SD amount of MBT was surprisingly consistent among subgroups defined by the presence or absence of 1 or more risk factors (Table 5). Survivors underwent transfusion with 32±10 U compared with 31±11 U for nonsurvivors. Such close similarity in blood usage was shown across all risk subgroups. Table 6 shows the mortality rates in the 4 MBT subgroups when 1 or more risk factors are present. In the presence of risk factors, mortality is not affected by the amount of blood transfused.
Because of exploding health care costs in the United States, there has been considerable interest in determining futility of care.4- 6 A Consensus Statement of the Society of Critical Care Medicine suggested that communities have a legitimate interest in allocating medical resources by limiting inadvisable treatments and sought to define treatments as futile when they will not accomplish their intended goal.7 The disproportionate economic impact of injury makes the trauma setting a fertile ground for such discussion. Because blood is one of the most precious resources maintained in trauma centers, limits on MBT may be justified in critically injured patients if a point of futility can be determined.
Many studies have evaluated mortality after MBT.2,3,8 Although survival rates that approximate 40% have been reported, evaluation and comparison of these data are difficult because the amount of MBT is different from study to study and is calculated during periods ranging from 1 day after trauma to the total hospital stay.1- 3,8- 10 When blood transfusion is given for a long time, a good outcome is expected. However, when MBT is required during the first few hours after admission, the outcome may be dismal. In this study, we focused on the preoperative and intraoperative period that follows emergency admission for critical injuries requiring surgical interventions. The surgeon in the OR is frequently faced with the dilemma of continuing or abandoning aggressive resuscitation and MBT for patients in extremis. Identifying limits of MBT and risk factors beyond which survival is not expected could assist the surgeon in making such decisions and could avoid unnecessary expenditure of valuable resources.
In seeking to identify a reasonable limit to MBT, our data point to some interesting conclusions. First, there is no amount of blood required before and during an operation for trauma that can by itself preclude survival. The 3 survivors with MBT higher than 50 units support this observation. Second, even in the presence of risk factors for mortality, the extent of MBT is not associated with survival. Table 6 demonstrates that patients with 1 or more risk factors had similar chances to survive across all 4 subgroups of MBT. There was a striking consistency of mean units of blood transfused among survivors and nonsurvivors when stratified for the 3 independent variables that were associated with mortality. Among all these 3 groups or any combination of them, survivors and nonsurvivors had a consistent average 31- to 32-U MBT requirement. Stated differently, this finding means that although need for aortic clamping, requirement for inotropic support, and prolonged intraoperative shock predicted mortality, patients with these factors had an equal probability of dying regardless of the amount of blood transfused. The only combination that seemed to be invariably lethal was the presence of all 3 risk factors in the face of MBT of more than 30 U (Table 6). Although the number of patients in these groups was small (13 patients total), mortality was 100%. One may assume that under such conditions further resuscitative efforts should be abandoned, but a larger number of patients is desirable to make this conclusion.
This study has the limitations inherent in a retrospective analysis, the most disturbing being the inability to extract complete data regarding our selected variables for the whole study population. Although univariate analysis indicated variables such as hypothermia, acidosis, electrolyte abnormalities, decrease in urine output, and problems in oxygenation and ventilation were important for prediction of survival, the high incidence of missing values precluded their entry in the multivariate analysis. Had they been taken into consideration, the sample size would decrease to only 36 patients with complete information for all variables. We strongly emphasize the need for better recording of physiologic parameters in the OR to enhance monitoring of patient status and to develop strategies for critical decision making.
In conclusion, the results of this study support the notion of pursuing aggressive resuscitation at the acute stage after life-threatening injuries. Although blood is valuable, surgeons should not base decisions regarding continuation of treatment on the amount of blood that is required. Even in the presence of established risk factors for mortality, blood should not be denied at any point during the early hours after trauma.
Gail T. Tominaga, MD, Honolulu, Hawaii: This paper addresses a question asked by many caregivers treating patients with exsanguinating traumatic injuries. It is not uncommon for the operating surgeon to be asked if additional blood should be given when a critically ill, exsanguinating patient requires ongoing transfusions. When is it futile to give additional blood transfusions? The answer to this is not always evident. The authors address this issue and analyze 56 data elements in an attempt to define risk factors for mortality.
In the 1970s, mortality rates over 90% were reported in patients receiving massive transfusions. Over the next 2 decades, improvements in blood banking, resuscitation, and operative technique increased survival rates to 48% to 66% for patients receiving massive transfusions. These previous studies looked at transfusions given anywhere from 12 to 24 hours following trauma to the entire length of hospitalization. These are not comparable. In addition, anecdotal cases of survivors following transfusions of greater than 100 U of blood have been reported. The current study is unique from previous reports since it concentrates on transfusions given during a very short time frame, namely, the preoperative and intraoperative periods. The average number of units transfused in this study was 32, with a range of 19 to 68 U. The overall mortality rate was high at 70%, which would be expected in the severely injured patients requiring massive transfusions over a period of a few hours.
This study confirms what surgeons have long thought. The total blood loss and amount of transfused blood are far less critical than the duration and severity of shock. The current study found no significant difference in mortality rate in patients receiving 20 to 29 U of blood from those receiving 50 to 68 U of blood. The authors found 3 independent risk factors for mortality. These include aorta cross-clamping, use of inotropes, and more than 90 minutes of hypotension (defined as systolic blood pressure <90 mm Hg). Every surgeon would agree that these factors reflect a state of severe shock. The presence or absence of these risk factors did not correlate with the amount of blood transfused. In addition, the presence of all 3 risk factors did not preclude survival in this study. However, on close examination of one of the tables that was presented, a 100% mortality rate was seen in 13 patients who had all 3 risk factors present and received 30 U or more of blood.
Is it futile to resuscitate such patients, ie, should patients who receive over 30 U of blood who require aortic cross-clamping, use of inotropes, and are hypotensive for over 90 minutes receive continued resuscitation? Or, do you think the sample size of this group of only 13 is too small to draw such a conclusion?
As would be expected, age and Glasgow Coma Scale score were found to be significantly different between survivors and nonsurvivors. The "older" nonsurvivors had a mean age of only 32 years. In Hawaii, we have a large number of elderly trauma patients. I would like to ask the authors: How old was the oldest survivor in your study? Do you think that there is an age that precludes survival following injuries requiring massive transfusions?
In trauma care, it has become very clear that the triad of metabolic acidosis, hypothermia, and coagulopathy leads to death if not reversed. Jurkovich and colleagues in 1987 found 100% mortality in severely injured patients with core temperatures of 32°C or less. Phillips in 1987 reported a 77% mortality in patients who developed coagulopathy. In 1988, Feliciano and coworkers reported on 300 abdominal gunshot wound patients and concluded that acidosis, hypothermia, and coagulopathy contributed to 85% of deaths. In the current study, it is concerning that temperature and arterial blood gas data were not consistently documented.
I would like the authors to address the following: Do you have set protocols for hypothermia prevention in the emergency room and OR? What measures do you use to reverse hypothermia? Do you think the presence of acidosis, hypothermia, and coagulopathy contributed to the deaths in your patients who received massive transfusions? In 1983, Stone and colleagues described a method of managing major coagulopathy with onset during laparotomy which has led to the concept of "damage control." A recent 20-year review reported a mortality rate of 52% in 961 severely injured patients treated with the damage control approach. My last question is: Did any of the patients in your study have a damage control laparotomy with delayed definitive operation?
William P. Schecter, MD, San Francisco, Calif: Most of the people in the room who take care of high blood loss cases have had a number of patients who have survived with blood losses in excess of 60 U. I think the authors' conclusion fits with at least my experience. However, the equal mortality rates do not compute with what I would intuitively think would be the result. I would like to ask the authors if they looked at the time of death in the different groups. In other words, did the patients who died with 20 U of blood loss die earlier in their operative course as a result of technically challenging problems and therefore increased their mortality rate to the level of the patients with 60 U of blood loss who may have died of multiple organ failure later in their course?
Dr Cornwell: I will make some general comments and then answer the questions. With 12 million units of blood being transfused annually, and one fourth of that (3 million units of blood annually) being transfused just specifically for emergency surgical patients, the question may well come up as to the utility of this resource. In his presidential address 2 days ago, Dr Vetto discussed things that we should want and cited as number 8 that we be able to treat our patients according to our best judgment. He stated that "guidelines are important adjuncts to therapy," but they are sometimes taken as absolute.
The Ethics Committee of the Society of Critical Care Medicine is in search of the definition of futility of care and sought to define it as that care which will not achieve its intended goal. So we see this paper as somewhat of a preemptive report, if you will, to address the type of concerns that are raised regarding the use of these resources. Specifically, as stated in the conclusions, we could identify an upper limit on the number of units of blood transfused.
Dr Tominaga raised the points very insightfully that among the 13 patients who had those 3 risk factors that we cited on multivariate analysis (ie,  the need for inotrope support,  prolonged hypotension, and  aortic cross-clamping), and received greater than 30 U of blood, there were no survivors. Indeed, that same table makes the point, when you look at those patients who had none of the 3 risk factors, that there was a 75% survival across all levels of transfusion, including 5 of 5 patients who had greater than 30 U of blood transfused. So it would seem that if you have those risk factors, mortality is imminent regardless of how many units, greater than 30 or less than 30. Again, the number of units of blood transfused did not correlate with survival.
Dr Tominaga asked about the oldest survivor. Our oldest survivor was a 55 year old who received 41 U of blood. Age was a risk factor on the univariate analysis, but on the stepwise logistic regression analysis it did not come out as an independent risk factor. The reason for this is that we have small numbers of patients older than 55 years, so we were unable to suggest an age above which there should not be MBTs.
Regarding the prevention of hypothermia, for all trauma cases going to the OR, the ambient room temperature is turned up, all patients are placed on a BAHR hugger, and receive transfusion with warm fluids and warmed inspired gases. We attempt, of course, to prevent the hypothermia coagulopathy acidosis syndrome, and, indeed, only 19% of these patients with massive transfusions experienced that syndrome. Once the patients develop hypothermia, we do attempt to treat, we add to those modalities instillation of 40°C fluid in the peritoneal cavity, although Gentillelo and associates at Washington showed that that is not a very efficient way of reversing the syndrome.
Along those same lines, we do, in fact, employ damage control techniques. Indeed, over 60% of the survivors in our series had multiple procedures for one reason or another. So yes, we do believe strongly that the hypothermia acidosis coagulopathy syndrome contributed to deaths in this series. The reason why we stopped short of identifying that in this presentation is again that it did not come out on multivariate analysis. Our statistician tells us that in order to appropriately do a multivariate analysis, every patient who is entered must have that data available for each of the variables. For 2 of those variables, hypothermia and acidosis, over 10% of the patients did not have all of the data available. Had we used those variables and included only patients who had all of the information available, our patient sample size for multivariate analysis would have dropped from 141 to 36. So, while we strongly believe that that syndrome does contribute to death, we stopped short of putting that statement in the conclusions for reasons of statistical validity.
Dr Schecter asked a very insightful question regarding OR times in survivors vs nonsurvivors and exactly when they were dying. All of these patients went to the OR. There was a significant difference in the mean operating time; the nonsurvivors had a shorter operating time, suggesting that perhaps they had the more exsanguinating injuries, and they died earlier in the course of the attempted treatment of their injuries.
Presented as a poster at the 69th Annual Session of the Pacific Coast Surgical Association, Maui, Hawaii, February 18, 1998.
We acknowledge the contribution of J. M. Nelson, MD, from the Blood Bank at Los Angeles County+USC Medical Center, Los Angeles, Calif, in accumulating the data on massive blood transfusion.
Reprints: G. C. Velmahos, MD, PhD, Los Angeles County+USC Medical Center, 1200 N State St, Room 9900, Los Angeles, CA 90033 (e-mail: firstname.lastname@example.org).