Endpoints of Resuscitation of Critically Injured Patients: Normal or Supranormal?

Abstract

Objective

To evaluate the effect of early optimization in the survival of severely injured patients.

Summary Background Data

It is unclear whether supranormal (“optimal”) hemodynamic values should serve as endpoints of resuscitation or simply as markers of the physiologic reserve of critically injured patients. The failure of optimization to produce improved survival in some randomized controlled trials may be associated with delays in starting the attempt to reach optimal goals. There are limited controlled data on trauma patients.

Methods

Seventy-five consecutive severely injured patients with shock resulting from bleeding and without major intracranial or spinal cord trauma were randomized to resuscitation, starting immediately after admission, to either normal values of systolic blood pressure, urine output, base deficit, hemoglobin, and cardiac index (control group, 35 patients) or optimal values (cardiac index >4.5 L/min/m², ratio of transcutaneous oxygen tension to fractional inspired oxygen >200, oxygen delivery index >600 mL/min/m², and oxygen consumption index >170 mL/min/m²; optimal group, 40 patients). Initial cardiac output monitoring was done noninvasively by bioimpedance and, subsequently, invasively by thermodilution. Crystalloids, colloids, blood, inotropes, and vasopressors were used by predetermined algorithms.

Results

Optimal values were reached intentionally by 70% of the optimal patients and spontaneously by 40% of the control patients. There was no difference in rates of death (15% optimal vs. 11% control), organ failure, sepsis, or the length of intensive care unit or hospital stay between the two groups. Patients from both groups who achieved optimal values had better outcomes than patients who did not. The death rate was 0% among patients who achieved optimal values compared with 30% among patients who did not. Age younger than 40 years was the only independent predictive factor of the ability to reach optimal values.

Conclusions

Severely injured patients who can achieve optimal hemodynamic values are more likely to survive than those who cannot, regardless of the resuscitation technique. In this study, attempts at early optimization did not improve the outcome of the examined subgroup of severely injured patients.

The quest for ideal endpoints of resuscitation after severe trauma is incomplete. Higher-than-normal values (traditionally called supranormal or optimal) of global hemodynamic and oxygen transport parameters are associated with improved outcome. ^1,2 Frequently used optimal values are a cardiac index (CI) greater than 4.5 L/min/m², an oxygen delivery index (DO₂I) greater than 600 mL/min/m², and an oxygen consumption index (VO₂I) greater than 170 mL/min/m². ^1,2 However, it is not clear whether these values should represent endpoints of resuscitation or are simply markers of physiologic reserve and predictors of outcome.

Ten prospective randomized studies have examined this issue. ^3–12 Five of them ^3–7 found no difference between patients who were randomized to receive resuscitation aimed at optimal values versus patients who had only normal values as endpoints of their resuscitation. Four trials ^8–11 showed that patients randomized to the optimal group had better outcome. One trial ¹² showed a higher death rate among patients randomized to optimal values. Most of these studies examined heterogeneous populations, randomized after severe sepsis or organ failure was diagnosed. Only one study ¹⁰ evaluated trauma patients exclusively.

A major critique of the existing studies is that resuscitation to optimal goals was not started early in the evolution of the critical illness. ^13,14 It has been argued that once organ failure is established, reversal is difficult, irrespective of the resuscitation technique used. ¹³ Optimization is best at improving survival when it is used to prevent organ failure, not treat it. ^14–16 However, hemodynamic and oxygen transport values are monitored after pulmonary artery catheterization. Because in most institutions a pulmonary artery catheter is not inserted before the patient arrives in the intensive care unit (ICU), invasive hemodynamic monitoring cannot start while the trauma patient is in the emergency department, radiology suite, or operating room. Often, many hours have elapsed and multiple diagnostic tests and therapeutic interventions have been completed before the trauma patient is transferred to the ICU and receives a pulmonary artery catheter.

In this randomized controlled study, we examined the appropriateness of resuscitating critically injured patients to optimal goals by starting immediately after admission. To obtain hemodynamic measurements before ICU admission, we used thoracic bioimpedance to calculate the CI and transcutaneous oxygen tension measurement to assess tissue perfusion during the pre-ICU resuscitation. Both techniques recognize accurately, in our experience, circulatory and perfusion deficits, and the derived values are associated with final outcome after critical injury. ^17,18

METHODS

Inclusion and Exclusion Criteria

The study was approved by the institutional review board. During a 2-year period from May 1997 to May 1999, we included consecutive trauma patients who were admitted at the Los Angeles County+University of Southern California Medical Center and had severe blunt or penetrating injuries requiring ICU admission, shock on arrival (systolic blood pressure <100 mmHg and heart rate >100 beats/min), bleeding as the cause of shock, and need for emergent surgery. Patients were excluded if they were younger than 18 or older than 70 years or pregnant; had severe intracranial injury (abbreviated head injury score 3–5), spinal cord injury with complete spinal cord deficit, shock that was easily reversed by administration of fluids in the emergency department, or hyotension from causes other than bleeding; needed emergency department thoracotomy; or died within 48 hours of admission.

The inclusion and exclusion criteria were purposefully restrictive to allow us to capture only severely injured patients who had major physiologic compromise. The patient sample size became as homogeneous as possible by not including patients with injuries (e.g., intracranial or spinal cord) that might affect the hemodynamic condition by mechanisms not related directly to blood loss, or who might deteriorate after excessive fluid resuscitation (e.g., selected head injuries). By excluding patients requiring emergency department thoracotomy, we eliminated patients with a very low likelihood for survival regardless of the method of resuscitation. The Injury Severity Score (ISS) and Trauma Related Injury Severity Score (TRISS) methodology were not used to select patients because these scores are not easily calculated soon after admission and have significant limitations in describing the severity of injury or physiologic compromise in certain subgroups of patients. ^19,20

Identification and Randomization of Patients

Our level 1 trauma center is staffed by seven full-time trauma surgeons and one critical care physician. The trauma team, including senior and junior surgical residents and a trauma attending, provides 24-hour coverage. Attending trauma surgeons participate in all trauma team activations and supervise all major resuscitations. The trauma attending was responsible for identifying patients for the study within 30 minutes of patient arrival. An on-call research team was responsible for connecting the patient to noninvasive monitors and recording the monitored values. Randomization was done by the research fellow opening a sealed envelope containing a card labeled “optimal” or “control.” The cards were produced by a computer-generated randomization program in blocks of six to avoid the same assignment more than three times in a row. The treating physicians were not masked to the assignment of each patient.

Endpoints of Resuscitation

Optimal Group

Patients randomized to the optimal group received standard monitoring (blood pressure, electrocardiography, oxygen saturation [SaO₂]) and were also connected to monitors measuring cardiac output by thoracic bioimpedance and transcutaneous oxygen tension. Values were recorded every 3 minutes, and therapy was given to achieve the resuscitation goals, which were as follows:

• Systolic blood pressure more than 100 mmHg

• Hematocrit more than 30%

• Urine output 1 mL/kg/hour or more

• Base deficit on the arterial blood gases less than −3

• CI 4.5 L/min/m² or more

• Ratio of transcutaneous oxygen tension to fractional inspired oxygen (P_tcO₂/FiO₂) more than 200.

If after ICU admission the patient had a pulmonary artery catheter, two additional resuscitation goals were set: DO₂I more than 600 mL/min/m² and VO₂I more than 170 mL/min/m². The P_tcO₂/FiO₂ cutoff point of 200 was selected because it correlates with survival, according to our previous experience. ^21–23

Control Group

Patients who were randomized to the control group were also connected to the noninvasive cardiac index by bioimpedance (CI_bi) and P_tcO₂ monitors, but the screens were covered and the treating physicians did not have access to the information collected by the research fellows from these monitors. Therefore, therapy was given based on information provided by routine methods of monitoring. The endpoints of resuscitation in this group were systolic blood pressure more than 100 mmHg, hematocrit more than 30%, urine output 1 mL/kg/hour or more, and base deficit less than −3. If after ICU admission the patient had a pulmonary artery catheter, additional resuscitation goals were set: DO₂I more than 450 mL/min/m² and VO₂I more than 130 mL/min/m².

Thus, the patients in the optimal group were treated within minutes after admission to optimal hemodynamic goals, whereas the patients in the control group were treated by routine methods to normal values. Study of resuscitation and monitoring was terminated at 24 hours after admission. Resuscitation was offered by the same methods in both groups and included fluids (in a 3-to-1 crystalloid-to-colloid ratio), blood, and inotropic support by dopamine, dobutamine, or both, if the hemodynamic goals were not achieved. Epinephrine or norepinephrine was used if the predetermined endpoints had not yet been reached. Only patients who reached and maintained the endpoints of resuscitation in each group were considered adequately resuscitated.

Methods of Monitoring

Cardiac output was monitored continuously and noninvasively by thoracic bioimpedance (IQ, Renaissance Technologies, Newton, PA). We have previously reported good correlation between the bioimpedance method and the thermodilution method of cardiac output monitoring. In a multicenter study of acutely ill emergency patients, ²⁴ the correlation coefficient between the two methods was r = 0.85, r² = 0.73, and the precision and bias was −0.124 ± 0.75 L/min/m². In two studies of all blunt trauma ¹⁶ and severe blunt trauma ¹⁷ patients, we found excellent correlations (r = 0.91 , r² = 0.83, precision and bias 0.36 ± 0.59 L/min/m²) between the two methods. Measurements of bioimpedance cardiac output values were recorded every 3 minutes. Measurements of thermodilution cardiac output values (if the patient had a pulmonary artery catheter) were recorded every 2 to 4 hours. If both techniques were available, thermodilution cardiac output measurements were selected for analysis over bioimpedance measurements.

P_tcO₂ measurements were recorded every 3 minutes throughout the observation period as an indicator of tissue perfusion. Previous studies have demonstrated the ability of P_tcO₂ to reflect the delivery of oxygen to the local area of the skin and parallel the mixed venous oxygen tension. ^25,26 Blood gas analysis was obtained every 2 to 4 hours. Blood pressure was measured noninvasively and by intraarterial catheterization in the operating room or ICU.

The noninvasive methods of monitoring were applied on the patient within minutes of arrival. It took approximately 5 minutes to connect the patient and 20 minutes until the first reliable measurement of P_tcO₂ was taken. To standardize reporting of these values, our first measurements of systolic blood pressure, CI_bi, and P_tcO₂/FiO₂ started at 30 minutes after patient arrival.

Data Collection

Data about demographics, mechanism and severity of injury, physiologic status, diagnostic and therapeutic interventions, and final outcome were collected prospectively by the research team (not the treating physicians). Although all injuries were recorded, the presence of colon or major vessel injury was noted separately because colon injury is associated with sepsis and major vessel injury with significant bleeding. The primary outcome was in-hospital death. Secondary outcomes were organ failure, sepsis, and ICU and hospital stay.

The following definitions were used ^27–29 :

• Adult respiratory distress syndrome: intrapulmonary shunt more than 20%, PaO₂/FiO₂ less than 200, pulmonary artery occlusion pressure less than 18 cm H₂O, diffuse infiltrates on chest radiograph

• Respiratory failure: mechanical ventilation for more than 48 hours for lung-related reasons

• Circulatory failure: systolic blood pressure less than 90 mmHg in the presence of adequate volume load

• Renal failure: serum creatinine more than 2 mg/dL in the absence of prerenal causes or preexisting renal disease

• Hepatic failure: serum bilirubin more than 3 mg/dL and serum transaminase levels more than twice normal

• Sepsis: three or more of the following: temperature more than 38°C, white blood cell count more than 12,000/mm³ or less than 4,000/mm³, need for antibiotics, or positive cultures.

Statistical Analysis

Baseline and intervention characteristics and outcome measures were compared between the optimal and control groups using the chi-square test or the Fisher exact test for categorical variables and the Student t test for continuous variables. Comparison of the same characteristics was done between patients who achieved optimal values from both groups versus patients who did not. Dichotomous values were created from continuous variables across clinically significant cutoff points. Univariate analysis identified significant differences on the tested variables. Variables with P < .2 were selected for stepwise logistic regression to identify independent predictors of the ability to reach optimal values and survival. Relative survival and the relative ability to reach optimal values (as well as the odds ratio for death and the inability to reach optimal values) and their 95% confidence intervals were derived. Significance at P < .05 was maintained for all comparisons. SAS statistical software (version 6.12, SAS Institute, Cary, NC) was used.

We estimated our sample size by power analysis. To detect a 10% difference in death rate, from 30% to 20% (a one-third risk reduction) between the control and optimal groups at a .05 level of significance with 80% power, the required sample size was 626 patients (313 per group) for a two-sided hypothesis and 502 patients (251 per group) for a one-sided hypothesis.

Recruitment

The recruitment of patients was slower than expected. During a 2-year period, we entered 107 patients into the study. Of these, 32 were excluded: 16 because of inappropriate inclusion and 16 because they died within 48 hours of entry into the study.

Inappropriate inclusions occurred because the study called for early identification and randomization of potential candidates to achieve the predetermined study endpoints as soon as possible. All but one of these patients were inappropriately included during the first few months of the study (one was inappropriately included later) because the admitting trauma surgeon thought the patient was severely injured. When diagnostic tests revealed no major injuries, the patients were removed from the study (typically 1–5 hours after inclusion) and treated per routine. Seven of them were initially randomized to the optimal group and nine to the control group.

The 16 patients who died within 48 hours of inclusion were accounted for in an intent-to-treat analysis (8 in the optimal group and 8 in the control group). None of the remaining patients dropped out for any other reason.

Because at the end of the 2 years the sample size was smaller than expected, we decided to perform a preliminary analysis to determine whether there were significant trends that would justify continuation of the study.

RESULTS

Seventy-five patients satisfied all inclusion criteria: 40 in the optimal group and 35 in the control group. Optimal values were reached intentionally in 28 patients (70%) in the optimal group and spontaneously in 14 patients (40%) in the control group (P = .009). In all, organ failure developed in 35 patients (47%) and 10 (13%) died.

Optimal Group

There were no differences in baseline characteristics between the 28 patients who achieved optimal values compared with the 12 who did not. These subgroups had a similar incidence of male gender (86% vs. 100%, P = .297), penetrating mechanism of injury (82% vs. 58%, P = .133), colon injury (25% vs. 17%, P = .697), and major vessel injury (21% vs. 50%, P = .130) and similar ISS (22 ± 14 vs. 20 ± 8, P = .508) and systolic blood pressure on admission (112 ± 29 vs. 97 ± 28 mmHg, P = .124). However, they differed in age (27 ± 8 vs. 47 ± 16 years old, P = .0008). As expected, the patients who did not achieve optimal values were given more fluids during the first 24 hours (19 ± 7 vs. 12 ± 6 L, P = .008) and more units of blood (18 ± 15 vs. 8 ± 6 units, P = .046), and more of these patients required inotropic support (75% vs. 14%, P = .00003). These treatment differences reflect our efforts to optimize patients who could not achieve the predetermined optimal values.

There were strong differences in all outcome parameters between the two subgroups. Patients who achieved optimal values had a lower incidence of organ failure (21% vs. 75%, P = .0003), sepsis (21% vs. 83%, P = .00004), and death (0% vs. 50%, P = .00002). They also had a shorter ICU stay (8 ± 9 vs. 31 ± 34 days, P = .045) and showed a trend toward a shorter hospital stay (21 ± 17 vs. 35 ± 34 days, P = .08).

Control Group

There were no differences in baseline characteristics between the 14 patients who achieved optimal values spontaneously and the 21 who did not. The two subgroups were similar for gender (86% vs. 81% male, P = 1.0), mechanism of injury (78% vs. 52% penetrating, P = .162), colon injury (50% vs. 19%, P = .073), major vessel injury (36% vs. 33%, P = 1.0), ISS (24 ± 16 vs. 21 ± 11, P = .618), and systolic blood pressure on admission (112 ± 31 vs. 109 ± 29 mmHg, P = .768). They differed in age (25.5 ± 8 vs. 38 ± 17 years old, P = .008). Because there was no attempt to achieve optimal values for patients in the control group, the two subgroups had similar volumes of fluids and units of blood administered during the first 24 hours (14 ± 5.5 vs. 12 ± 6 L, P = 0.396, and 10 ± 7 vs. 11 ± 8 units, P = 0.449) and a similar frequency of inotrope use (7% vs. 24%, P = .366). Although there were clinical trends in favor of patients who spontaneously achieved optimal values in terms of organ failure and death, none of the outcome parameters was significantly different between the two subgroups: organ failure (43% vs. 67%, P = .163), sepsis (50% vs. 52%, P = 1.0), death (0% vs. 19%, P = .133), ICU stay (13 ± 13 vs. 15 ± 12 days, P = .644), and hospital stay (24 ± 15 vs. 28 ± 24 days, P = .543).

Optimal Group Versus Control Group

The two groups were similar in baseline characteristics and resuscitation offered, although there was a clinical trend toward greater use of inotropes in patients in the optimal group (Table 1). There was no difference in any of the outcome parameters between the two groups (Table 2). Survival rates were almost identical between the two groups—85% for the optimal group and 89% the control group. There was a clinical trend toward fewer organ failures in the optimal group.

Table 1. BASELINE CHARACTERISTICS AND RESUSCITATION

ISS, Injury Severity Score; SBP, systolic blood pressure.

Values are expressed as mean ± standard deviations for continuous variables and proportions of events for categorical variables.

Table 2. OUTCOME PARAMETERS

ICU, intensive care unit.

Results are given as mean ± standard deviations for continuous variables and proportions of outcome for categorical variables.

Patients With Versus Without Optimal Values

Comparison of the 42 patients from either group who achieved optimal values (intentionally or spontaneously) with the 33 patients who did not failed to show any differences in baseline characteristics except for age and mechanism of injury. Patients with optimal values were younger and had penetrating injuries more frequently than patients without optimal values (Table 3). Although both groups received the same volume of fluids in 24 hours, patients with optimal values received more units of blood and more frequent inotropes, as expected. There were significant differences in outcome: patients with optimal values had a lower incidence of organ failure, sepsis, and death, and shorter ICU and hospital stays (Table 4).

Table 3. BASELINE CHARACTERISTICS AND RESUSCITATION

ISS, Injury Severity Score; SBP, systolic blood pressure.

Values are expressed as mean ± standard deviation for continuous variables and proportions of events for categorical variables.

Table 4. OUTCOMES BETWEEN PATIENTS WHO ACHIEVED AND DID NOT ACHIEVE OPTIMAL VALUES

ICU, intensive care unit.

Results are given as mean ± standard deviations for continuous variables or as proportions of outcomes for categorical variables.

Predictors of Optimization

Multivariate analysis identified only one variable that independently predicted the patient’s ability to achieve optimal values: age younger than 40. Patients 40 years or older were less likely to achieve optimal values (odds ratio 0.29, 95% confidence interval 0.08–0.95, P = .046). Of 10 patients older than 50 years, none achieved optimal values (Fig. 1). The presence of blunt trauma also decreased the likelihood a patient would achieve optimal values, but the association did not reach statistical significance (odds ratio 0.27, 95% confidence interval 0.08–0.87, P = .088).

Figure 1. Effect of age on the ability to achieve optimal hemodynamic values.

Predictors of Survival

Patients who died were older (48 ± 19 vs. 31 ± 12 years old, P = .017) and had a marginally higher ISS (28 ± 10 vs. 21 ± 12, P = .062) than patients who survived. Age 40 or older decreased the likelihood of survival, but this association did not reach significance (odds ratio 0.13, 95% confidence interval 0.01–1.07, P = .072). The relative survival between patients in the optimal and control group was 0.96 (95% confidence interval 0.80–1.14, P = .742). The relative survival between patients who did and patients who did not achieve optimal values was 1.43 (95% confidence interval 1.15–1.80, P = .0001).

Intent-to-Treat Analysis

Of the 16 patients who died within 48 hours of admission, 8 were in the optimal group and 8 in the control group. There was no difference in baseline characteristics or resuscitation offered between the two groups. These 18 patients had a higher mean ISS (31 ± 11 vs. 22 ± 12, P = .009) and a lower systolic blood pressure (88 ± 34 vs. 109 ± 29 mmHg, P = .014) compared with patients who did not die within 48 hours, but no other differences were identified. The intent-to-treat analysis showed no difference in death rates between the optimal and control groups when these 18 patients were added (29% vs. 28%, P = 1.0).

Organ Failures and Sepsis According to Time to Optimization

Of the 42 optimal or control patients who achieved optimal values, organ failure developed in 12 and sepsis in 13. The average time to optimization was 6 ± 6 hours in both groups. There was no correlation between the development of these complications and the time to optimization (Fig. 2).

Figure 2. Effect of the time required to achieve optimal values on the development of organ failure or sepsis among optimized patients. Numerals above data points indicate number of patients.

Subanalysis of Nonoptimized Patients

The death rate for patients in the optimal group who despite all efforts did not achieve optimal values was 50% compared with 19% for patients in the control group who did not spontaneously reach optimal values (P = .11). The two subgroups were similar in all baseline characteristics. Nonoptimized patients in the optimal group received more fluids, blood, and inotropes than nonoptimized patients in the control group.

DISCUSSION

This study is the first to examine the role of optimal values as endpoints of resuscitation in critically injured patients, when the effort to achieve this goal starts immediately after the patient arrives at the hospital. Supporters of optimization ¹⁴ have explained the failure of some studies to find improved outcomes after optimization by arguing that its benefits disappear if it is started too late. It is only reasonable to assume that, like many other forms of therapy, optimization would produce better outcomes if used before the development of organ failure. Once the circulatory system decompensates and vital organs fail, it is difficult to reverse the damage by any therapy. Therefore, according to some authors, optimization should be used to prevent rather than treat organ failure. ¹³

In the literature, there is only one prospective randomized trial of trauma patients who had pulmonary artery catheters inserted within 12 hours of admission. ¹⁰ The group randomized to optimal values had better survival rates than the control group. In our study, the attempt to optimize was started on admission. Because pulmonary artery catheter insertion is difficult under emergency room conditions, and monitoring of data derived from it is almost impossible while the patient is being transported to different areas of the hospital for diagnostic and therapeutic interventions, at the early stages after admission we used noninvasive systems to guide our attempts to achieve optimal goals. These systems provide continuous, real-time monitoring of many hemodynamic parameters, including CI_bi and P_tcO₂ values, which have been validated as reliable indicators of physiologic status. ^16–18

There are two main findings from this study. First, the survival rates were identical between the optimal and the control group. Second, patients from either group who achieved optimal values had improved survival rates compared with patients who did not. These findings support the argument that achieving optimal values is an indicator of physiologic reserve rather than an endpoint of resuscitation. Patients who have the inherent ability to respond to trauma by increasing their cardiac output and oxygen transport capacity beyond normal levels are more likely to eliminate the existing oxygen deficit, avoid organ failure, and survive. Age seems to play the most important factor, determining the physiologic reserve required to compensate after severe trauma. Penetrating trauma may also be a predictor of optimal response because it is usually associated with specific injuries that can be addressed surgically, unlike blunt trauma, which is usually caused by diffuse impact with multisystem effects that cannot always be controlled by surgery.

Being able to predict which patients can respond to the efforts to raise their hemodynamic values is important. Patients who could not achieve optimal values in the optimal group had a higher death rate (50%) than similar patients in the control group (19%). This may indicate that attempts to optimize patients who do not have the necessary physiologic reserve may be detrimental.

Although initially a larger study was planned, our preliminary analysis showed that such a study would be extremely unlikely to detect significant differences in the death rate between the two randomized groups because the death rates in our study were almost identical. A reason for the relatively small sample size was the adoption of strict inclusion and exclusion criteria. We wanted to identify a homogeneous trauma population with severe injuries and bleeding as the cause of shock, and avoid patients with confounding injuries, such as severe brain or spinal cord trauma, which may create hemodynamic abnormalities by separate pathophysiologic mechanisms. For this reason, our results should be restricted to this very specific population—that is, patients with severe trauma and shock from bleeding, and without severe head and spinal cord injuries.

Optimal values can be reached by various methods. The amount and type of fluids or inotropic, vasoconstrictive, or vasodilating agents vary from study to study. 3,6,10,12 Because the debate on crystalloids versus colloids is unsolved, we elected to use both in a ratio of 3 to 1. Inotropes were used only after the patient failed to achieve the predetermined values by fluid challenges. It is possible that if we had used different methods to achieve optimal values, we might have found different outcomes.

Optimization to predetermined values (CI > 4.5 L/m²/min, DO₂I > 600 mL/min/m², VO₂I > 170 mL/min/m²) ignores the variability in human physiology and each patient’s individual response to trauma. These criteria were used to comply with the existing body of literature. However, we believe that the design of oxygen delivery/oxygen consumption curves and identification of the point that each patient becomes flow-independent (i.e., the oxygen consumption does not increase by further increasing the oxygen delivery) might have been a more appropriate method to define optimal resuscitation goals. This study remains to be done.

In summary, we have shown that the deliberate attempt to increase certain hemodynamic values to supranormal endpoints is not associated with increased survival in this subgroup of critically injured patients, even if it starts in the immediate posttraumatic phase. Patients who have the necessary physiologic reserve to overcome the oxygen debt by reaching optimal values will do so regardless of the endpoints of resuscitation or methods of monitoring and are accordingly more likely to survive. Age seems to be the most important factor for the ability to achieve optimal values.

Discussion

Dr. Gill Cryer (Los Angeles, California): I would like to congratulate the authors on what is an extremely difficult study to do. Decision for enrollment was made by attending surgeons in the emergency room when the patient first arrived and they had research teams who applied the monitoring system to the patients in a fashion that allowed the results to be blinded in the control group. You can all imagine that this would be very difficult to do over a several-year period of time.

They hypothesize that optimization of resuscitation to achieve supranormal hemodynamic values beginning in the emergency room would lead to improved outcome. The study is prospective and randomized. It had a power of 80% to detect a 10% difference in mortality, i.e., a reduction from 30 to 20%, if 626 patients were enrolled.

The study design was actually very good. It required so many patients because a large number of patients in both groups will achieve supranormal hemodynamic values on their own with the standard resuscitation techniques. This fact has been documented by several previous studies and occurred in this study as expected. Therefore, my first question is, why did you quit after only 75 patients were enrolled? The real question to be answered in this study is, is there a group of patients who do not achieve supranormal values on their own but can be made to achieve supranormal values by using an advanced monitoring and resuscitation technique, and will there be an improved outcome in these patients?

So my second question is, how many patients in the optimization group did not achieve supranormal values initially but were able to achieve them with additional resuscitation efforts? If such patients existed, they obviously had a zero mortality rate, because that was the mortality rate in the optimized patient group as a whole. There is a strong likelihood that your conclusions are the result of a type 2 error, and I would encourage to you keep going with the study.

Presenter Dr. George C. Velmahos (Los Angeles, California): Your first question, why quit before the end-goal of our analysis? The recruitment of patients was considerably slower than we had expected. Essentially, we decided to sacrifice sample size in favor of homogeneity, because in other studies, the heterogeneity of patients included was a significant problem. To achieve the goal of our power analysis, we should include 650 patients, which meant, by the recruitment standards we had during these 2 years, that we should be carrying on with the study for 15 years. This fact also reveals the importance of multicenter cooperation for these types of studies. The second reason for discontinuing the study is that there were no significant clinical trends in mortality; therefore, we were somehow discouraged about carrying on with the study any further.

The second question about patients not initially optimized but optimized subsequently by additional techniques, and the ability to recognize a specific subgroup of patients requiring additional techniques, I would answer that this is very difficult to assess. The average time to optimization was about 6 hours. So whoever was able to reach optimal values, reached optimal values relatively early. And definitely a type 2 error with this sample size cannot be excluded. A supplemental analysis showed that patients who were not optimized in the control group had a mortality rate of 11%, patients who could not be optimized in the optimal group, despite us pushing them with additional methods, had a mortality rate of 50%. Therefore, we came to conclude that probably this is not the ideal thing to do.

Dr. Charles E. Lucas (Detroit, Michigan): This is an excellent study, which by definition means that it supports all of my biases based on our own studies that we have performed in Detroit over the past many years. My only question concerns the use of colloid. As you know, the Cochrane group in England, using metaanalysis, in 1997 concluded that colloids when added to resuscitation are bad. Following their lead, the FDA in 1998 issued a warning that colloids should be only used when they are part of a prospective randomized study testing the efficacy of colloids. Two questions: What colloid did you use and who were you trying to please when you added the colloids? Certainly, your institution is not known for supporting colloid-supplemented resuscitation for injured patients.

Dr. Velmahos: Obviously, the fight between colloid and crystalloid usage is not unique to your institution. We also have diverse views in our own institution. At least one of the investigators is in favor of colloids; other investigators on the other side are in favor of crystalloids. We tried to satisfy all of them. In this kind of study we had to reach a consensus—particularly when there are seven dedicated trauma and critical care surgeons who will admit those patients. That is why we adopted the intermediate solution and used both colloids and crystalloids in the 1-to-3 ratio. As colloids, we used Hespan or albumin, and that was left to the physician’s discretion.

Dr. Philip S. Barie (New York, New York): Taking your data in the context of the large multicenter trial published by Gattinonni in the New England Journal of Medicine in 1995, would you concur that the hypothesis supporting the notion of enforced oxygen transport resuscitation to supraphysiologic levels is essentially a discredited hypothesis at this point and that we should move on? If so, do you believe the problem has been that the hypothesis has been fundamentally flawed or that we simply don’t have resuscitation strategies adequate to the task? If the latter, how might you suggest that we could improve our resuscitation strategies in the future?

My second question specifically relates to the finding that your elderly patients were essentially nonresuscitatable. How many of them did you have? Are there any specific strategies that we should use for our elderly patients, who are increasingly populating the emergency rooms at major trauma centers, and would a study specifically looking at resuscitation strategies in elderly patients be of value at this point?

Dr. Velmahos: With regard to Gattinonni’s study, as with many similar studies, it included a very heterogeneous population, patients who were septic, medical patients, surgical patients, trauma patients. We tried to avoid this bias of including a mixed type of patients under the same hypothesis. He showed—in a slightly different design than ours—that these endpoints of resuscitation are not valid. So we do agree with the study. But as I said before, we tried to examine a very homogeneous population to at least give a confirmative answer for a specific group of patients.

Whether we should use other endpoints, we didn’t examine that. But I am sure that other endpoints like Dr. Ivatury has used with gastric tonometry are quite valid and are awaiting randomized controlled trials.

Should we study the elderly patients separately? I believe, yes. Dr. Yu’s studies have shown that trying to optimize elderly patients results in different outcomes depending on the age of the patient. So, one has to play with the data and the populations that are at hand in order to find the appropriate place for optimization.

Dr. Rao Ivatury (Richmond, Virginia): A couple of questions for you. You mentioned transcutaneous oxygen monitoring. You did not comment how it impacted on your study. My concern also is that your inclusion criteria are a little vague. Did you look at specific anatomic and physiologic injury severity to see whether the groups are comparable in terms of ISS RTS? Third, we had presented a study where we showed that the use of PA catheter and measurement of cardiac index, oxygen consumption, et cetera, are not necessary as long as the other endpoints (base deficit, lactate) are moving in the right direction, which is basically what you are saying now. So, would you comment on whether we can do away with the PA catheter as well as the measurement of cardiac index, et cetera?

Dr. Velmahos: With regard to the transcutaneous oxygen monitoring, we have done previous studies that have proven it to be a good surrogate of oxygen tissue perfusion. And we actually have found that transcutaneous oxygen to FiO₂ ratio of more than 200 is a so-called optimal value that indicates good tissue perfusion. This is why we use it. It is another noninvasive and continuous monitor of tissue perfusion.

With regard to the second question about whether we should abandon the PA catheter, I don’t think we can do away with it. There is definitely a place for the PA catheter. The particular values derived by the PA catheter are not useful in every occasion, but this doesn’t mean that the tool as such should never be used. I do agree that our endpoints of resuscitation should target the cell instead of targeting global oxygen tissue perfusion values. But this still remains to be seen and evaluated in further trials.

Dr. Frank R. Lewis (Detroit, Michigan): I enjoyed this paper a great deal and applaud the authors for their willingness to effectively reverse their previous positions of many years. However, I would urge that you consider one step further, because the correlation seen with optimization may still lead one to believe that optimization is a critical variable. It may in fact be a purely coincidental variable that has nothing to do with the underlying cause of improved survival.

For example, if I tried to correlate the number of units of transfusion with mortality in trauma patients, it would be a fairly easy thing to do. But no one would contend that it was the transfusion of the blood that caused the mortality—rather, it was the underlying severity of the injury. I think exactly the same may be true here. This may be a totally coincidental variable that has no relationship to the real underlying cause. That should be considered.

The second point I would like you to address is that optimization by blood volume replacement at increased preload is totally different in its effects than optimization by the use of inotropic agents. It is quite likely that the latter is harmful. You did not attempt to differentiate this in the manuscript in any way, and I would ask how often you actually used inotropes and if you considered that those two factors have very different physiologic effects.

Dr. Velmahos: We showed that whether it is a coincidental variable or a variable that should not be used as an endpoint of resuscitation are probably different ways of stating the same conclusion. So I agree with your first statement.

In terms of whether we should use more or less inotropes and more or less fluid, again I come back to the need for consensus when a variety of physicians admit a variety of patients. We standardized our resuscitation scheme to satisfy everybody and to include fluid and blood and inotropes and vasopressors in a stepwise manner during the resuscitation. We used inotropes in anywhere between 40% to 50% of our patients, and did not examine whether the optimization by inotropes was harmful compared to optimization by fluids.

Dr. Erwin F. Hirsch (Boston, Massachusetts): I congratulate you for your efforts and for 2 years of battling the emergency department. That in itself is a reportable case.

Many of us believe that optimizing or finding endpoints of resuscitation are based by acid-base determinations, the magnitude of the base deficit, not so much in an individual determination but the time it takes with which to return the patient to normal acid-base balance. My question to you is, in your group of patients that were either optimized or failed to optimize, have you seen a change in speed or change in trend and acid-base metabolism? Have those patients who failed to optimize showed delay in normalization of their acid-base metabolism compared to the other ones? Were you able to correlate your hemodynamic and other variables to that arterial blood gas over a period of time?

Dr. Velmahos: I had exactly the same thought. We have not done it yet. But we are trying to reexamine this population and to correlate the base deficit with the ability to achieve or not achieve optimal values. So I wouldn’t have an answer to your question. But we are planning to reanalyze our results to that effect.

Dr. Basil A. Pruitt, Jr. (San Antonio, Texas): Since bioimpedance is influenced by changes in volume status, how reliable and reproducible were the measurements that you made by bioimpedance during resuscitation? Since the fluid volumes of the outcome groups were about the same, is the only difference due to the more frequent pharmacologic intervention in the optimal value group? Third, what about the incidence of complications of fluid excess in the intent-to-treat group, not just the optimal value outcome group but in all those whom you treated to optimize values? Were there more instances of need for fasciotomy or for surgical release of the abdominal compartment syndrome? I think we have moved to a period in which we see problems due to excess resuscitation rather than problems of inadequate resuscitation.

Dr. Velmahos: Thoracic bioimpedance is a relatively new method measuring cardiac output. We have published a few studies that show that values monitored by bioimpedance correlates well with the same values monitored by thermodilution through a PA catheter.

So with correlations that range anywhere between .8 and .9 we feel pretty safe that cardiac output bioimpedance is a reliable tool. However, there are specific circumstances, such as patients with emergency room thoracotomies, where the measurements by thoracic bioimpedance could be unreliable. This is an additional reason we excluded these patients.

With regard to your other question, the optimal group received much more inotropes than the control group and yet they had no difference in outcome. Actually, offering too much fluid in our attempt to optimize created some problems. On some occasions, it may have caused abdominal hypertension. However, I think we didn’t find any statistical differences between the two groups on such complications. And we looked for it. I can tell you anecdotally, as I was looking one by one at these patients, that we may have flooded some of them as we endlessly pushed fluids and inotropes to optimize them despite the fact that they clearly could not achieve these optimal values.