Hemorrhage is a leading cause of death after trauma, and identification and management of hemorrhage is at the core of the American College of Surgeons Advanced Trauma Life Support (ATLS) curriculum. 1 Conventional emergency department protocols and ATLS call for rapid fluid resuscitation in all hemorrhaging trauma patients, beginning with the administration of up to 2 L of crystalloid and continuing with packed red blood cells and plasma as needed to maintain a normal systolic blood pressure.
This approach has been challenged by a number of authors, on the grounds that aggressive fluid administration in animal models leads to increased bleeding because of increased arterial and venous pressure, dilution of clotting factors, and decrease in blood viscosity. 2–10 Models of uncontrolled hemorrhage in swine, 3–6 dogs, 7 sheep, 8 and rats 9,10 have demonstrated increased hemorrhage when a normal systolic blood pressure is used as the target for fluid resuscitation. The majority of these studies have also documented a decrease in survival in animals targeted to a normal systolic blood pressure. 3–7,9,10 Several trials have identified a decrease in tissue oxygen delivery (largely because of hemodilution) when hemorrhaging animals are resuscitated to normal baseline blood pressure. 3,5,10
Clinical study of deliberate hypotension in the resuscitation of trauma patients has been confined to one prospective trial completed in Houston in the early 1990s. 2 Hypotensive victims of penetrating torso trauma were randomized in the field to either receive intravenous fluids or not, and this therapy was continued until the end of the patient’s stay in the emergency department. Although this study showed a survival advantage in the no-fluid group, it was subject to a number of statistical and methodologic shortcomings. The results were limited to penetrating trauma, although the majority of hemorrhagic shock seen in the United States is the result of blunt injury. Second, limited resuscitation occurred only during prehospital and emergency department care; all patients were aggressively resuscitated in the operating room, even if still hemorrhaging. Finally, the “all-or-none” nature of the protocol ignored the titration of fluid administration to the patient’s vital signs and clinical condition, the normal standard of care.
The Houston study sparked controversy but has done little to change the standard practice of resuscitation in hemorrhaging trauma patients. Further human research in this area was limited by the Food and Drug Administration from 1993 to 1995, because of changes in the process for obtaining informed consent for trials of emergency therapy. With the release of new guidelines in 1996, 11 we undertook to repeat and extend the findings of the Houston investigators, with the hope of clarifying this important issue.
We report the results of a clinical comparison of two different protocols for fluid resuscitation during active hemorrhage in injured humans. Our hypothesis was that fluid administration directed to a systolic blood pressure (SBP) of 70 mm Hg would lead to increased survival compared with conventional fluid administration directed to an SBP > 100 mm Hg.
PATIENTS AND METHODS
The Fluid Resuscitation in Trauma (FRT) study was conducted from 1996 to 1999 under a policy of “delayed informed consent,” in accordance with standards established by the Food and Drug Administration, 11 and with the approval of the University of Maryland Institutional Review Board. Eligible patients were enrolled in the study at the time of presentation to the trauma center. All patients or their guardians were then informed of their participation as soon as possible after study enrollment, and permission was obtained for continued data collection and inclusion of these data in our analysis. Patient and family acceptance of this process for obtaining informed consent in emergency settings has been shown to be high. 12
Patients were eligible for inclusion if they presented directly from the scene of a traumatic injury, had evidence of ongoing hemorrhage, and had an SBP < 90 mm Hg recorded at least once within the first hour after injury. Patients were excluded if they were pregnant, had a central nervous system injury impairing their level of consciousness or motor function, were older than 55, or had a previous medical history of diabetes or coronary artery disease.
Eligible patients were prospectively randomized to one of two groups during the period of active hemorrhage: fluid administration titrated to a “conventional” SBP > 100 mm Hg, or to a “low” SBP of 70 mm Hg. Blood pressure below the target level was treated with administration of crystalloid or blood products, as appropriate to elevate the SBP to the target level while maintaining a hematocrit of at least 25%. Sustained SBP above the target level was managed by restriction of fluids and administration of appropriate doses of anesthetic or analgesic medication. Figure 1 shows the algorithm used for fluid management during the period of active hemorrhage.
The end of active bleeding was determined in each case by the trauma surgeon and anesthesiologist, on the basis of one or more of the following criteria: visible control of hemorrhage in the operating room, stable blood pressure not requiring fluid administration for support, tolerance of a normal level of analgesia and sedation, and diagnostic studies such as computed tomographic scan or angiography showing no evidence of ongoing hemorrhage.
After the end of active bleeding, resuscitation was completed in all patients in accordance with ATLS guidelines and the existing protocols of the trauma center. Clinical targets for final resuscitation included the following: normal systolic blood pressure and heart rate, hematocrit > 25%, urine output > 0.5 mL/kg/h, arterial lactate level < 2 mg/dL, and normal arterial base deficit.
The clinical outcome of each study patient, Injury Severity Score (ISS), and predicted probability of survival were recorded directly from the medical record and the trauma registry. Blood pressures during active hemorrhage were abstracted from the medical record. The length of active hemorrhage was calculated from the time of admission until the end of active hemorrhage, as determined by the patient’s attending surgeon and anesthesiologist.
Results are presented as mean ± SD. Statistical analysis included two-tailed comparison of independent variables with χ2 analysis. Results were considered statistically significant at a level of p < 0.05. The probability of survival statistic was calculated using the TRISS methodology. 13
One hundred ten patients were enrolled over 20 months, 55 in each group (Table 1). The study cohort had a mean age of 31 years, 79% were male patients, and 51% of the patients had sustained penetrating trauma. There was a significant difference in SBP observed during the study period (114 mm Hg vs. 100 mm Hg, p < 0.001). The duration of active hemorrhage was not different between groups (2.97 ± 1.75 hours vs. 2.57 ± 1.46 hours, p = 0.20) Overall survival was 92.7%, with four deaths in each group.
There was no difference in the degree of anatomic injury between groups: the ISS in the conventional-pressure group was 19.65 ± 11.84, compared with 23.64 ± 13.82 in the low-pressure group (p = 0.11). There were 17 patients in the conventional-pressure group with an ISS > 24, and 23 in the low-pressure group. There was no statistically significant difference between groups in predicted probability of survival, on the basis of the TRISS methodology.
Figure 2 shows the therapeutic paths followed in the study population, illustrating the flexible nature of clinical practice in acute trauma. Of the 110 patients studied, 85% underwent at least one surgical procedure and 15% underwent angiographic embolization, with this technique as the sole means of hemorrhage control in 4%. Thirteen percent of patients achieved hemostasis spontaneously, the majority after tube thoracostomy or nonoperative management of a low-grade splenic injury.
Demographic characteristics of the 110 patients enrolled in the trial are summarized in Table 2. There were no significant differences between groups in the sites of hemorrhage (Table 3). There were no significant differences in the number of patients in each group who underwent surgery, angiography, or nonoperative management.
Table 4 summarizes demographics of the eight study patients who died. Two of four patients in the conventional-pressure group died while still actively hemorrhaging, as compared with three of four patients in the low-pressure group. The two patients in the conventional-pressure group who did not die from acute hemorrhagic shock in the operating room died from multiple organ failure on hospital days 7 and 5, respectively. The first patient was a morbidly obese man who had suffered an upper abdominal gunshot wound, with injury to the liver and pancreas. He developed acute respiratory distress syndrome postoperatively, then renal failure, and then cardiac failure from which he could not be resuscitated. The second patient had suffered evisceration and massive lower abdominal soft tissue loss in an industrial explosion, complicated by 30% third-degree burns. Although hemorrhage was controlled initially, the extent of his injury made surgical reconstruction impossible.
The one mortality in the low-pressure group who survived the initial hemorrhage had a grade V blunt liver injury, and required a near total resection and embolization of the liver. This patient died from fulminant hepatic failure on hospital day 9 while undergoing assessment for transplantation.
Deliberate hypotensive resuscitation during active hemorrhagic shock has been demonstrated to improve survival in a large number of animal trials. 3–10 In these models, limiting fluid resuscitation facilitates control of hemorrhage, leading to better preservation of oxygen-carrying capacity and a reduced incidence of rebleeding. In the series of blunt and penetrating trauma patients we report, however, targeting volume resuscitation to a lower than normal blood pressure during active hemorrhage did not improve survival. We believe there are a variety of reasons why this clinical trial did not replicate the laboratory results.
Clinical reality precludes the degree of sophistication of monitoring and observation that can be attained in the laboratory. Blood pressure is well known to be a poor surrogate for tissue oxygen delivery, especially in younger patients with well-preserved vasoconstrictive reflexes. The choice to base our protocol on blood pressure was driven in large part by necessity: blood pressure is the measure that is most reliably available during early resuscitation, and blood pressure is the most consistent driver of fluid therapy in actual practice. Although better markers of perfusion have been described, such as mixed venous oxygen saturation 14 or gastric tonometry, 15 they are not readily available in the first hours after patient admission, particularly in the setting of ongoing hemorrhage. Patients like those in our study are typically undergoing multiple diagnostic and therapeutic procedures during this time, and continuous hemodynamic monitoring is limited to that which can be quickly applied and easily shifted with the patient from the emergency department to the operating room and/or the angiography suite.
The selected endpoint of our study is also subject to criticism. In-hospital mortality is easy to determine and clinically relevant, but may be too broad an endpoint to discriminate subtle differences in outcome between groups. Other outcome measures that might have been considered included surrogate markers of resuscitation such as lactate or base deficit, hospital length of stay, or the incidence of multiple organ system failure. However, each of these measures suffered from either subjectivity in interpretation or sensitivity to the patient’s initial injury. In the end, we selected hospital mortality as the primary outcome variable precisely because of its relevance and lack of ambiguity. The TRISS methodology was then used as a secondary analysis, in an effort to adjust for differences caused by differing initial injuries.
It is possible that our results were affected by a failure to achieve the proposed methodology. Targeting a lower than normal systolic blood pressure for resuscitation (70–80 mm Hg) resulted in an average pressure during active hemorrhage of 100 mm Hg, as compared with 114 mm Hg in the group targeted to SBP > 100 mm Hg (p < 0.001). The reason for the discrepancy between the target and the result is twofold. First, the dynamic interaction between fluid administration, anesthetic agents, and the patient’s own autoregulatory mechanisms make a stable blood pressure unlikely during active hemorrhage. When fluids are administered in small, titrated amounts in response to decreases in blood pressure, the result is an oscillation of blood pressure in the vicinity of the target. Second, as hemorrhage slows or stops, the blood pressure will rise spontaneously toward normal, even in the absence of infused fluids. Figure 3 is an illustration of this effect in a single patient from the low-pressure group, who underwent “damage control” packing of a grade V liver injury, followed by angiographic embolization, followed by an immediate return to the operating room to achieve complete hemostasis. The data points represent the patient’s SBP at 15-minute intervals.
This effect was also observed in the Houston study. 2 In that trial, hypotensive patients enrolled in the field arrived at the emergency department with equivalent blood pressures, whether they had received intravenous fluids or not. Indeed, the spontaneous slowing of hemorrhage because of the preservation or encouragement of native hemostatic mechanisms is thought to be an important advantage of low-volume resuscitation. 2,3
As with any human clinical trial, there are questions of patient selection that must be considered when interpreting our results. The observed overall mortality in our study was significantly less than that in the Houston trial (8% vs. 34%), possibly as a result of bias in selecting against patients already very near death on arrival to the trauma center. Although the physicians caring for the patient did not know the group assignment until after the patient was randomized, their initial decision to include the patient for study may have been biased, and may have affected the overall outcomes of the study population such that the group actually studied was healthier than intended. We believe this would lead to less observed difference in outcome between groups, because the expected difference in survival will be lower in healthier patients. There is also a possibility that our results were skewed by a Hawthorne effect, wherein both groups received more attentive care than usual (and less infused crystalloid) because they were enrolled in a study requiring close observation of administered fluids.
Patient heterogeneity is important in considering the results of this study, as it represents one of the most important differences between laboratory and clinical research. The trauma patients in our study presented with a common clinical syndrome—hemorrhagic shock—but a variety of underlying anatomic injuries. It is likely that deliberate hypotensive management is of more value when treating some injuries than others. In particular, it would seem that keeping the blood pressure low would be of the greatest value in lesions that are not readily accessible surgically, such as posterior pelvic fractures, or in areas of the body where hemostasis is difficult to achieve, such as the liver.
The impact of deliberate hypotensive management on other injuries is hard to estimate. We specifically excluded patients with traumatic brain injury (TBI) because of the substantial clinical literature supporting the absolute prevention of hypotension in TBI patients, 15 although one recent laboratory report demonstrated improved survival in a rat model of hemorrhage in combination with TBI. 16 The degree to which brain perfusion must be preserved in the face of ongoing hemorrhage remains an important open question in the management of severely injured blunt trauma patients, as current recommendations call for avoidance of hypotension at all costs. 17
Although it is possible that a difference in outcomes between groups would emerge if the study were allowed to continue for a longer period of time—particularly in light of the variance in injury severity between groups—we believe that this would introduce other methodologic errors at least as substantial. On the basis of the results to date, more than 500 additional patients would have to be enrolled to demonstrate a statistical difference in survival, if overall survival in the cohort remained similar and the (presently nonsignificant) difference in TRISS-predicted outcome continued. The extension of this study for several more years would introduce a large potential for bias arising from changing surgical practice, with a substantial risk that overall mortality would decrease even further. Just within the period studied, our practice has evolved to include increased use of focused abdominal sonography for diagnosis of abdominal hemorrhage, angiographic alternatives for hemorrhage control, and greater intraoperative use of thrombogenic packing materials. We have elected therefore to close the study at this time, and to present the existing data.
Deliberate hypotensive management of the actively hemorrhaging trauma patient, as described herein, has no greater impact on mortality than conventional therapy. Further studies in this area should focus on specific patient populations most likely to benefit from deliberate hypotensive resuscitation, and on the development of better markers for assessing tissue perfusion and ischemic risk. Future studies are feasible from the standpoint of emergency consent, 12 but will likely require larger numbers of patients, and therefore multiple investigative sites, to yield clinically relevant results.
We thank Mehrunissa Owens, Elizabeth Kramer, Sharon Boswell, and the surgeons and anesthesiologists of the shock trauma center for their specific assistance in completing the FRT study.
1. American College of Surgeons, Committee on Trauma. Advanced Trauma Life Support Program for Physicians. Chicago, IL: American College of Surgeons; 1993.
2. Bickell WH, Wall MJ, Pepe PE, et al. Immediate versus delayed resuscitation
for hypotensive patients with penetrating torso injuries. N Engl J Med. 1994; 331: 1105–1109.
3. Stern A, Dronen SC, Birrer P, Wang X. Effect of blood pressure on haemorrhagic volume in a near-fatal haemorrhage model incorporating a vascular injury. Ann Emerg Med. 1993; 22: 155–163.
4. Kowalenko T, Stern S, Dronen S, Wang X. Improved outcome with hypotensive resuscitation
of uncontrolled hemorrhagic shock in a swine model. J Trauma. 1992; 33: 349–353.
5. Bickell WH, Bruttig SP, Millnamow GA, O’Benar JO, Wade CE. The detrimental effects of intravenous crystalloid after aortotomy in swine. Surgery. 1991; 110: 529–536.
6. Riddez L, Johnson L, Hahn RG. Central and regional hemodynamics during fluid therapy after uncontrolled intra-abdominal bleeding. J Trauma. 1998; 44: 1–7.
7. Burris D, Rhee P, Kaufmann C, et al. Controlled resuscitation
for uncontrolled hemorrhagic shock. J Trauma. 1999; 46: 216–233.
8. Sakles JC, Sena MJ, Knight DA, Davis JM. Effect of immediate fluid resuscitation
on the rate, volume, and duration of pulmonary vascular hemorrhage in a sheep model of penetrating thoracic trauma. Ann Emerg Med. 1997; 29: 392–399.
9. Capone A, Safar P, Tisherman S, et al. Treatment of uncontrolled haemorrhagic shock: improved outcome with fluid restriction [abstract]. J Trauma. 1993; 35: 984.
10. Smail N, Wang P, Cioffi WG, Bland KI, Chaudry IH. Resuscitation
after uncontrolled venous hemorrhage: does increased resuscitation
volume improve regional perfusion? J Trauma. 1998; 44: 701–708.
11. Federal Register.
12. Dutton RP, Mackenzie CF, Scalea TM. Clinical research in emergency patients: experience with delayed consent at a Level 1 trauma center. Anesthesiology. 2000; 93: 3A.
13. Champion HR, Copes WS, Sacco WJ, et al. The Major Trauma Outcome Study: establishing national norms for trauma care [abstract]. J Trauma. 1990; 30: 1356–1365.
14. Abou-Khalil B, Scalea TM, Trooskin SZ, Henry SM, Hitchcock R. Hemodynamic responses to shock in young trauma patients: need for invasive monitoring. Crit Care Med. 1994; 22: 633–639.
15. Ivatury RR, Simon RJ, Havriliak D, et al. Gastric mucosal pH and oxygen delivery and oxygen consumption indices in the assessment of adequacy of resuscitation
after trauma: a prospective, randomized study. J Trauma. 1995; 39: 128–136.
16. Novak L, Shackford SR, Bourguignon P, et al. Comparison of standard and alternative prehospital resuscitation
in uncontrolled hemorrhagic shock and head injury. J Trauma. 1999; 47: 834–844.
17. Chestnut RM, Marshall LF, Klauber MR, et al. The role of secondary brain injury in determining outcome from severe head injury. J Trauma. 1993; 34: 216–222.