The majority of what is known about the physiology of hemorrhagic shock has been learned from studies with animal models—primarily dogs and rats. Determination of the degree to which experimental results from studies with animal models can be extrapolated to the human shock condition must consider several factors, including (1) the species, (2) the initial experimental condition of the animal, (3) the shock model (i.e., how do you go about inducing the blood loss and what do you do after the initial blood loss is induced?), and (4) the dynamic character of the shock process itself. For example, anesthetics have been shown to play a vital role in affecting the way that an animal responds to the shock process; that is, the type of anesthesia influences central nervous system responses as well as peripheral vasomotor responses to blood loss.
Choice of species can have a major impact on outcomes. For example, greater blood loss has been exhibited in studies of isobaric models of hemorrhagic shock in species with a large splenic red blood cell reserve, such as the dog. This variable can also affect the severity of the shock when the blood loss is based on percent body weight in fixed-volume models of hemorrhagic shock. The pig, although noted for the similarity of its cardiovascular system responsiveness to that of the human cardiovascular system, is quite dissimilar with respect to the hemoglobin P50 (the partial pressure of oxygen where hemoglobin is 50 percent saturated with oxygen) which is substantially higher than that for the human hemoglobin (e.g., 38 versus 28-30 millimeters of mercury [mm Hg]). Rats demonstrate a very small oxygen extraction reserve in skeletal muscle beds compared with that for dogs, which have an extraordinarily large extraction reserve. The magnitude of this reserve affects the degree of tolerable flow decrement during hypotension, which in turn affects the magnitude of the systemic vascular resistance response observed during hemorrhage.
Canine models typically develop portal hypertension during shock and bloody diarrhea as a postresuscitation complication that other species do not exhibit. Baboons appear to exhibit a more robust compensatory response to hemorrhagic shock compared with the responses of humans and most other species, and are able to tolerate a mean arterial blood pressure of 40 mm Hg for over 8 hours before they exhibit signs of decompensation.
Models of hemorrhagic shock are of two basic types: controlled and uncontrolled with respect to the manner in which ongoing blood loss is allowed to proceed. Controlled models are either of the fixed-volume or constant-pressure varieties and are generally more reproducible than uncontrolled bleeding models, which usually involve laceration, puncture, or transection. When an experimental hypothesis includes the treatment effect on homeostasis, uncontrolled bleeding models are relevant. Since shock is a dynamic process involving dramatic changes in cardiovascular and metabolic states that vary with time, species, laboratory, and even investigator, it is important to evaluate the effectiveness of various resuscitative regimens not simply on the basis of time but also on the basis of a more complete definition of the stage of physiologic compensation at the time that treatment is initiated. It has been shown that the type of anesthesia, state of hydration, nutrition status, core temperature, and use of heparin all affect the time course and degree of compensatory capacity of the animal in response to blood loss.
Effects of Extent of Hypotension and Rates of Hemorrhage on Immune Function in Mouse Models
Available information indicates that even transient hypotension in the absence of significant tissue trauma is sufficient to produce marked suppression of both specific and nonspecific immune responses. This appears to be the case irrespective of whether the model of hemorrhage used is one of a fixed-pressure versus a fixed-volume bleed, a bleedout over a relatively brief period (less than 5 minutes of cardiac stick/fixed-volume model), an intermediate period (approximately 5 to 15 minutes of hemorrhage to fixed pressure), or a protracted period (tail vein laceration or uncontrolled hemorrhage models), as well as whether the animal is anesthetized or unanesthetized. Also, although many of these models are typically nonlethal, they render the animal highly susceptible to subsequent lethal septic challenge. This appears to be the case irrespective of the administration of standard fluid resuscitation (with or without blood), although it can be modified by the rate of fluid administration and by the nature of the resuscitation fluid (e.g., lactated Ringer's solution versus hypertonic saline).
Nonetheless, trauma appears to provide a modulatory effect in the sense that it can be additive or prolongs the duration of immune aberrations encountered in these models. Another significant modifier frequently present in many models is an anticoagulant, such as heparin. Studies indicate that this agent can reduce the severity of insult produced in these models. Interestingly, microbial
translocation and the release of endotoxin do not appear to be significant mediators of these immune suppressive effects in short-term (up to 1.5-hour) models but may be a factor in longer-term (more severe) models.
Mechanistically, the process by which cell, immune, or organ function is altered appears to be biphasic. The initial early phase is characterized by acute cellular changes (metabolic dyshomeostatis, which consists of a decrease in adenosine triphosphate [ATP] levels and pH, increases in calcium ion [CA2+]I, etc.) associated with systemic proinflammation (an increase in tumor necrosis factor [TNF], interleukin-1 [IL-1], and IL-6 levels, etc.), which accounts for the early depression in the cellular response. This transitions over time to a chronic phase of mediator-induced (an increase in anti-inflammatory cytokine, nitric oxide [NO], and oxygen [O2] levels, etc.) or endocrine system-induced (prolactin, androgens versus estrogens, etc.) sustained immune system or organ dysfunction.
The advantage of mouse models in studying both the pathologic changes that occur during shock and the physiologic responses to acute hemorrhage is that inbred strains of mice have greater uniformity at the species level. Also, transgenic or gene-knockout strains provide animal models that are deficient in various mediators and that therefore add important information that often is not available by traditional pharmacologic approaches.
Swine Models of Combined Hemorrhage and Injury
Research conducted by Proctor (1998) relies on models in which tissue injury is superimposed on hemorrhagic shock, followed by administration of the same type of fluid resuscitation that would be available in the combat or civilian setting. The strength of the model is that injury is always associated with hemorrhage. The addition of tissue injury activates the inflammatory process (neurohumoral factors), which alters the response to hemorrhage.
Trauma and shock produce whole-body ischemia, and resuscitation produces a reperfusion injury. During the traumatic insult, there is decreased blood flow and oxygen delivery and breakdown of ATP to adenosine. Proctor (1998) is investigating in a swine model the roles of increased adenosine levels and altered neutophil activation. The data collected in that research, however, are relevant only to those who survive to the point of first aid. Nonetheless, Proctor has found that there is, in fact, secondary injury caused by the activated white blood cells, which can be affected by altering CD 18 cells or granulocyte colony-stimulating factor (GCSF). The only way that these changes can occur, however, is by administering a secondary insult. These studies have also shown that liquid ventilation is able to produce the same protective effects as positive end expiratory pressure without the negative hemodynamic actions.
Value of Animal Models
A great deal of literature exists with regard to canine, swine, and rodent models, but somewhat less information regarding simian or monkey (baboon) models is available. There are some significant interspecies variations in terms of biochemical responses, focus on organ injury, immune response, and neural-element (sympathetic nervous system) involvement. In the canine model there appears to be a significant bowel and splanchnic organ focus of injury, with a combination of endotoxin-like and toxic peptide autodigestive components contributing to the irreversible nature of shock. There is also an early and prominent sympathetic nervous system and adrenal cortical and medullary response. In the canine model splanchnic response to sympathetic nervous system activation and the increase in plasma catecholamines levels is augmentation of the circulating blood volume by autotransfusion from a large splenic reservoir, which does not exist to the same extent in other species.
The appearance of pulmonary injury occurs in a more delayed fashion. Some investigators have preferred the swine model and believe that the hemodynamic responses and cardiovascular reserve capacity more closely relate to the human pathophysiology of hemorrhagic shock. The rodent models provide easier access to large amounts of data, although the technical difficulties of complete cardiovascular monitoring are greater because of the size of the animals and limits to the technology for obtaining certain measurements in small animals. One advantage of the rodent model is the availability of a large number of genetic variants and gene-knockout models, making possible assessment of various tissue mediators, hormones, and neural components of the shock response. The use of simian models has been based on the fact that the animals' anatomy and physiology more closely approximate those of humans and their pathophysiologic responses to hemorrhagic hypotension or shock more closely resemble the human response. The simian models exhibit a more robust response, and the animals usually sustain a more prolonged survival and better outcome with exposure to the same level of hemorrhagic stress to which humans are exposed.
Two principal models of hemorrhagic shock exist: the controlled hemorrhage model and the uncontrolled hemorrhage model. The controlled hemorrhage model uses either bleeding to a predetermined pressure or a predetermined volume as a percentage of blood volume and body weight. In the Wiggers model (Wiggers, 1950), hemorrhage to a mean arterial blood pressure of 40 mm Hg is maintained for a predetermined period of time with measurement of the shed blood volume, various hemodynamic attributes, regional blood flow, biochemical markers of organ function, sympathetic nervous system activity, and the circulating concentrations in plasma of catecholamine and various hormonal indicators of stress such as cortisol. After the predetermined period the animals
are reinfused with the shed blood or various resuscitation fluids, and survival as well as physiologic measurements can be monitored at intervals to determine responses and rates of survival versus death. A number of criticisms have indicated the shortcomings of this model. These shortcomings stem from such things as pretreatment with agents that interfere with the sympathetic nervous system (e.g., ganglionic blockers or adrenergic antagonists) or the use of potent general anesthetics that decrease the level of the shed blood volume at which the desired mean arterial pressure is reached and that lower the catecholamine and hormonal responses, indicating a lower level of hemorrhagic stress. In other words, the mechanics of the experimental design may influence the endpoints of research that uses this model.
When a controlled hemorrhage by percentage of blood volume or body weight is used, the indicators of the response rather than the shed blood volume are the blood pressure and other hemodynamic measurements as well as other measurements of stress, as indicated above. The animals are held for a period of time and are then reinfused with shed blood or the fluid resuscitation being tested, or a combination of both. This technique also permits measurements of various response parameters and can be used to obtain survival data. The common criticism of both of these controlled hemorrhage models is that they do not mimic actual shock conditions in humans.
The uncontrolled hemorrhage model can be totally uncontrolled hemorrhage from either a catheter or a rent in the aorta or major vessel. In this model, controlled rates of bleeding are sometimes used or controlled rates of bleeding are combined with a partially controlled bleeding model with further hemorrhage being uncontrolled. Various hemodynamic measurements as well as other organ function measurements can be made as described above for the controlled hemorrhage model. It has been stated that the model that most closely approximates battlefield casualty conditions is one that uses the uncontrolled hemorrhage in the absence of anesthesia in subjects who were previously dehydrated and exposed to various stresses. Although this might be a closer approximation to battlefield conditions, the ability to measure responses and outcomes would appear to be better with controlled hemorrhage models, simply because additional experimental variability is introduced by uncontrolled hemorrhage.
The combat injury is usually a combination of hemorrhage and soft-tissue injury. The soft-tissue injury component is most often a penetrating injury that produces a number of subsequent reactions that can influence the outcome in different ways. The release of cytokines and other substances from injured tissue contributes significantly to the organ function disorders associated with shock. The animal model that most closely approximates the battlefield injury should include not only an acute hemorrhage but also some aspects of tissue injury such as a penetrating or crush injury to an extremity. Penetrating injury of visceral organs may introduce other complicating factors such as septic shock introduced by penetrating injuries of the gut and other abdominal visceral organs. Penetrating wounds of the chest may add additional injuries to the heart or lungs,
resulting in earlier cardiovascular collapse because of cardiac injury or early respiratory decompensation because of lung injury.
Role of Anesthesia
All of the currently used inhalational anesthetics including halothane, enflurane, isoflurane, desflurane, and sevoflurane may exhibit depressant effects on the cardiovascular system when they are administered at doses of between 1.0 and 2.0 the minimum alveolar concentrations. The cardiac depressant effects of halothane, enflurane, isoflurane, desflurane, and sevoflurane have been demonstrated in isolated heart preparations and in the unanesthetized chronically instrumented dog (Pagel et al., 1991a,b). These studies have indicated that these anesthetics have negative inotropic effects and impair ventricular diastolic function (Pagel et al., 1991c). Among the intravenous anesthetics, barbiturates, the benzodiazepines, etomidate, and propofol, all produce dose-dependent cardiovascular effects (Merin, 1996). Greater controversy exists with regard to ketamine, which may produce transient stimulating effects on the heart and peripheral circulation, but these effects are eliminated in animal models when the sympathetic nervous system is blocked or inhibited (Pagel et al., 1992). The minimal depressant effects of newer narcotics such as fentanyl and its derivatives have been confirmed by studies with animals and humans, but such efforts are found only at higher narcotic levels if ventilation is maintained since respiratory depression is a common finding at higher doses (Merin, 1996).
Among the physiologic responses to acute hemorrhage that occur while receiving potent inhalational anesthetics and many intravenous anesthetics are depression of (1) both high-pressure and cardiac low-pressure reflexes, (2) the chemoreceptor reflex, and (3) sympathetic nervous system responses. Studies with human volunteers have demonstrated inhibition of carotid sinus and aortic sinus reflexes by halothane (Duke et al., 1977), enflurane (Morton et al., 1980), and isoflurane (Kotrly et al., 1984). Other studies with humans have demonstrated that halothane has an inhibitory effect on cardiopulmonary reflex regulation of limb and vascular resistance (Kotrly et al., 1985). The potent anesthetics halothane and isoflurane have been shown (Seagard et al., 1983, 1985) to inhibit the carotid sinus reflex at multiple sites including the central nervous system, sympathetic pre- and postganglionic sites, and neuroeffector junctions in the heart and in the arteries and venules. Direct inhibitory effects of halothane on sympathetic ganglia have also been demonstrated (Bosnjak et al., 1988).
Inhibition of chemoreflex regulation of the cardiovascular system by potent volatile anesthetics has been demonstrated in animal models during halothane (Stekiel et al., 1992) and isoflurane (Stekiel et al., 1995) anesthesia. These effects were demonstrated for arterial resistance vessels as well as venous capacitance vessels.
In summary, direct depressant effects of potent volatile anesthetic agents and many intravenous anesthetics on the heart and peripheral blood vessels as a
baseline effect have been demonstrated. Of equal or greater concern are the depressant effects of most general anesthetics on the major physiologic regulatory systems that are activated in response to acute hemorrhage. The depressant effects have been described for high-pressure baroreflexes, cardiopulmonary baroreflexes, chemoreflexes, and the sympathetic nervous system. These anesthetic-induced alterations in basic physiologic control mechanisms result in subsequent alterations in blood flow, oxygen delivery, tissue oxygenation, and even survival in hemorrhaged animals (Longnecker and Sturgill, 1976; Longnecker et al., 1982; Seyde and Longnecker, 1984; Seyde et al., 1985; Weiskopf et al., 1981). Furthermore, these responses vary by organ and by species (Longnecker and Seyde, 1986), and thus, no single anesthetic regimen can be recommended for all studies involving traumatic shock. Rather, the choice of anesthesia will depend on the experimental design, the animal species, and the measured outcome variables. However, the ubiquitous nature of these effects must be taken into account both in the design of studies and in the analysis of the data. Animal protocols that do not require anesthesia are preferred, when feasible, for they most closely mimic the clinical scenario of battlefield trauma. When anesthesia is required, agents and techniques that are reproducible and that minimize the effects on cardiopulmonary control systems that are activated by acute hemorrhage should be selected.
Animal models should be selected for specific reasons on the basis of the research questions being asked. If survival is an outcome measurement, survival for 5 to 7 days,, not just 12 to 24 hours, should be the endpoint. Experimental designs should look realistically at establishing a repeatable 50 percent lethal dose for the control group. If animal models are chosen to develop a consensus for application of novel treatment regimens and agents to potential human trials, then these interventions should be applied in an identical animal model with identical experimental protocols with defined endpoints or outcomes.
Role of Clinical Trials in Development of Therapies
At the outset of this discussion, the committee differentiates between clinical research and clinical trials. Clinical research has recently been considered to be a term applicable to three major areas of biomedical investigation (National Institutes of Health Director's Panel on Clinical Research, 1997):
Patient-oriented research, that is, research conducted with human subjects (or with material of human origin, such as tissues, specimens, and cognitive phenomena) and in which an investigator directly interacts with human subjects. This area of research includes (a) mechanisms of human disease, (b) therapeutic interventions, (c) development of new technologies, and (d) clinical trials.
Epidemiological and behavioral studies.
Outcomes research and health services research.
Clinical trials, which are listed as a subset of clinical research, have been defined as prospective studies comparing the effect and value of an intervention(s) in human subjects against those in a control (Friedman et al., 1985, p. 2). This definition implies that the intervention may be prophylactic, therapeutic, or diagnostic, but it is a necessary component. Without active intervention (e.g., follow-up of subjects over a period of time), the study is merely observational—no experiment is performed and the study would not properly be termed a clinical trial.
Clinical trials often provide the most definitive proof of the safety and efficacy of diagnostic and therapeutic interventions in humans. Although studies with animals and uncontrolled clinical observations contribute to the understanding of a clinical entity, such studies usually cannot definitively demonstrate whether a new treatment has made a difference in clinical outcome. The well-designed clinical trial, in which the treatment group is comparable in every way to the control group except for the intervention studied, provides the means for such definitive demonstration.
Endpoints and Indications
Because of the power of controlled clinical trials, they are an integral part of the regulatory approval system for drugs, biologics, and medical devices. The pertinent federal regulations require that the safety and efficacy of new agents be demonstrated by controlled clinical trials. Regarding efficacy, the regulations (21 CFR 601.25(d)(2)) state:
Effectiveness means a reasonable expectation that, in a significant proportion of the target population, the pharmacological or other effect of the product ... will serve a clinically significant function in the diagnosis, cure, mitigation, treatment or prevention of disease.
Although this regulation pertains to approvals by the U.S. Food and Drug Administration (FDA), the underlying concept is generally accepted. Clinically significant function means that the intervention under evaluation must be shown to benefit clinically the patient population under study. Put another way, the primary endpoint of the pivotal clinical trial that is intended to support approval must be a direct measurement of the clinical benefit of the intervention. A limited number of types of endpoints satisfy the definition given above:
- increased survival of the study population,
- measurable symptomatic relief to the study population, or
- prevention or slowing of the progression of a disease process.
In many cases, it is not possible to measure such direct endpoints, and investigators seek to substitute a more readily measured entity as a surrogate for a clinical endpoint. Such surrogate endpoints in clinical trials are usually laboratory measurements or physical signs that are substitutes for and that are expected to correlate with a clinically meaningful endpoint that directly measures how a patient feels, functions, or survives. There are two principal risks involved with use of surrogate endpoints:
- The clinically meaningful endpoint may not actually correlate with the proposed surrogate.
- The surrogate may correspond to a real benefit, but the intervention may have serious undesirable consequences, complicating evaluation of the true risk-benefit ratio.
In general, surrogate endpoints can be considered when there is sufficient knowledge of the disease entity and of the intervention under investigation, when the feasibility of performing meaningful clinical trials is poor, and when the overall risk-benefit situation justifies the use of such endpoints. Hypertension is a surrogate marker for hypertensive cardiovascular disease, on the basis of extensive research on and experience with these entities. A decrease in the blood pressure of hypertensive patients has been used as a surrogate endpoint in the clinical trial of antihypertensive agents, and this is an example of a validated surrogate endpoint.
The endpoints chosen for a clinical trial are a critical element of the protocol, for they will determine the duration, complexity, and perhaps the success or failure of a trial. Selection of endpoints for trauma trials is especially difficult, and when a clinical endpoint is desired, mortality rate is often chosen. At a recent conference on blood substitutes and oxygen therapeutics (BCI Conference on Blood Substitutes and Oxygen Therapeutics, November 19 and 20, 1998, Bethesda, Md.), a speaker from FDA stated that mortality will be the endpoint of choice for clinical trials on hemorrhagic shock or exsanguinating hemorrhage. If a resuscitation solution studied in a clinical trial on trauma is not anticipated to improve the rate of mortality associated with trauma, then the ability of such a product to improve a major cause of morbidity can be used to demonstrate the efficacy of the product, whereas a solution that does not worsen the mortality rate but that results in major irreversible morbidity in the survivors would not be judged to be effective. Again, the validation of these endpoints will provide a challenge to investigators.
The use of surrogate endpoints in clinical trials on trauma is not as acceptable as is the case with the example of hypertension given above. The clinical conditions associated with trauma and the interventions of interest are more complex and heterogeneous and the relationship between them is less well understood. The use of surrogate endpoints (such as blood pressure, lactate levels, base deficit, or organ functional assessments, among others) must be validated as correlating with survival in hemorrhagic shock or exsanguinating hemorrhage
before use in lieu of a mortality endpoint. Scoring systems developed to evaluate the severity of an injury, such as the APACHE score, have been proposed as the basis for endpoints in trials on trauma. The applicability of these scores as surrogate endpoints has been the subject of considerable debate.
Evaluation of Resuscitation Protocols
The existing trauma indexing systems have contributed a great deal to the triage of trauma patients and to the development of systems for assessment of quality of care (Champion et al., 1996, 1990, 1989, 1983, 1981; Copes et al., 1990, 1988; Gennarelli et al., 1994, 1989; Sacco et al., 1988, 1984); however, the current trauma indexing systems are inadequate for use in the evaluation of future research (Brenneman et al., 1998; Demetriades et al., 1998; Hoyt, 1998; Roorda et al., 1996; Rutledge and Osler, 1998; Rutledge et al., 1998). The injury severity score (ISS) does not accurately stratify patients according to injury because it was not designed to evaluate penetrating injury and is inaccurate in its ability to categorize head injury. The ISS was developed to categorize blunt injuries sustained in motor vehicle accidents (American Association for Automotive Medicine, 1985). In addition, there may be reason to believe that injuries as defined by ISS do not correlate with the actual demand for resources. There are also problems with the ability of physiologic indexing systems, such as the Revised Trauma Score (RTS), to predict resource need. Casualties with scores that imply a minor injury may have penetrating abdominal injuries that will nonetheless require surgery.
There are also problems with the trauma and injury severity score (TRISS) assessment methodology, which has become the benchmark for the evaluation of trauma care. The TRISS model functions as follows:
The probability of survival Ps is computed by the following equation:
where bn are regression coefficients obtained from the large multithousand patient database that the American College of Surgeons (Baker et al., 1974; Champion et al., 1980a,b, 1981; Flora, 1978; Walker and Duncan, 1967) has collected for over 15 years and analyzed by regression analysis.
Comparison of the predicted outcome with the realized outcome (Z) is calculated by:
where Pt is the probability of survival for each patient i. As calculated above, Qt, equals 1 - Pt (the probability of death of patient i), and Dt equals the actual number of patients who died. The methodology essentially compares the number of patients who the regression process would predict would die on the basis of national experience of mortality from those injury severities versus the number who actually died.
There are two problems with this methodology. First, there is no matching of injury. Although one can run a Mantel-Haenszel statistic to demonstrate that the distribution of the Ps between the national patient database and a comparative group is statistically the same, there is no way actually to match the types of injuries themselves. For example, if one is testing a new fluid resuscitation protocol with a group of patients using the TRISS methodology to measure a difference in effect, the following could take place: 10 percent of the national group might have had closed head injuries with a Glasgow Coma Score of less than 8 along with other minor injuries, whereas 40 percent of the experimental group might have closed head injuries with a Glasgow Coma Score of less than 8. Since it is known that patients with closed head injuries and Glasgow Coma Scores of less than 8 have a significantly higher mortality rate than patients with minor head injuries and the same ISS, there will be a major bias against the experimental group by TRISS. The TRISS method achieves standardization of injury by use of the RTS and ISS. Although the Glasgow Coma Score affects RTS and the score has a significant influence on the probability of survival, it may still not be sufficient to control for this confounding. In addition, the American College of Surgeons Penetrating Injury database comprises data for civilian penetrating trauma victims who have mainly sustained knife or gunshot wounds. As noted in Chapter 1, the majority of penetrating injuries on the battlefield are due to shrapnel from explosive munitions. This may make the regression coefficients obtained from data for civilians invalid for the latter group.
The second problem is that as many as 90 percent of all patients who are initially triaged as major trauma victims, at least in the civilian sector, do not, as it turns out, have major life-threatening injuries (Bellamy, 1995). The analysis of performance gained by use of the TRISS methodology is statistically based on the outcomes for the remaining 10 percent of patients. The TRISS methodology is troubled because of the sigmoid nature of the distribution of the Ps of trauma patients. The mathematics of the method are such that it is overly affected by either the death or the survival of a very few of these severely injured patients. The use of this type of analysis to evaluate the effectiveness of a therapy is fraught with difficulty because one or two patients can drastically change the results.
Evaluation of the effectiveness of these measures requires the development of a new model to assess injury severity. One approach may be to use some of the newer statistical paradigms that are based on parallel distributed processing that use nonlinear statistics and that appear to be able to take into account multiple confounding effects (Armoni, 1998; Dombi et al., 1995; Forsstrom and Dalton, 1995; Obana and Fukui, 1996; Rutledge et al., 1998). Finally, trauma
outcome research should strive to compare like patients with similar injuries. A standardized data collection system should also be developed for all human studies that encompassed standardized definitions and standardized datum points. The Utstein system (American Heart Association, 1998), designed by investigators in cardiopulmonary resuscitation research, could be used as a template for this development.
Unique Problems of Clinical Trials of Trauma
Heterogeneity of Subject Population
The heterogeneity of the subject population has been alluded to before in this report, but it is a central issue and warrants attention. The patients who would be studied in a clinical trial designed to test an intervention for hemorrhagic shock might have suffered blunt or penetrating trauma or a combination of these; they may have a discrete injury to a single organ or multiorgan lesions; the trauma might be a small lesion in a major vessel or a large lesion affecting a diffuse vascular bed. Comparison of such diversity will demand an understanding or at least a recognition of some unifying principles within the study groups.
In a clinical trial the subjects or their authorized representatives must be clearly and completely informed of the risks and benefits associated with their participation in the trial. Because trauma victims may not be capable of giving meaningful consent, investigators faced a special problem with clinical trials with this population. To correct this situation, the federal regulation pertaining to the protection of human subjects (21 CFR, Part 50) has recently been amended to provide a narrow exception to the requirement for obtaining and documenting informed consent from each human subject or his or her authorized representative in certain situations. The exception would apply to a limited class of research activities involving human subjects who are in need of emergency medical intervention but who cannot give informed consent because of their life-threatening medical condition and who do not have a legally authorized person to represent them (Federal Register, 1996). It should be noted that concern over this change in regulations has been expressed (Moreno et al., 1998). The issues that cause concern include the concept and implementation of community consent, the ability of the system to exert oversight, and the appropriate use of consent from authorized representatives.
The regulations require the study sponsor to provide information about the trial to the community before and after the study is performed. These regulatory changes apply to the civilian population, and investigators using military populations for trauma studies must obtain permission directly from the patient or from an authorized representative.
Applicability of Civilian Clinical Data to Military Needs
Examples supporting or refuting the applicability of civilian clinical data relevant to fluid resuscitation of trauma victims to military needs could be given. The primary difference between the battlefield and civilian settings is that the vast majority of combat injuries are penetrating, whereas those encountered in civilian practice are blunt. Furthermore, not only are the majority of lethal injuries in the battlefield penetrating, but many are caused not by bullets but by shrapnel from explosive munitions.
Transportation and logistics may also differ, in that evacuation of a casualty in a combat setting as opposed to that in the civilian environment is frequently not rapid and the casualty is not routinely brought to a setting where definitive care can be administered. It is therefore not clear whether the results of studies performed under relatively controlled conditions (i.e., in a modem civilian trauma service) could be extrapolated to military field conditions. On the other hand, there are questions for which data are simply lacking, irrespective of the circumstances. Even though fundamental differences exist between combat and civilian settings, it is noteworthy that trauma is the leading cause of morbidity and mortality among teenagers and young adults in Central Europe, and in the United States data from U.S. trauma centers indicate that for approximately 40 percent of fatal trauma cases, death is due to exsanguination or its sequelae. For such clinical situations, data collected in civilian trauma services would be an advance over the current data availability situation, and such data could later be complemented by experience gained in military situations. In this regard, it is noteworthy that civilian trauma facilities are now used for the training of military medical personnel.
Clinical Research and Clinical Trials in Trauma Centers
The United States has regional trauma centers that are mostly self-designated and has accredited and verified trauma centers that are very organized. Research is done at such centers, and multi-institute trials are of interest to several professional organizations (including the American Association for Surgery in Trauma, Multi-Institute Trial Committee; the Eastern Association for Surgery in Trauma; and the Western Trauma Society). These groups have formulated trials for voluntary participation. The multicenter study on blunt aortic trauma recently reported by Fabian and colleagues (1997) is an example of such a study.
No ongoing, formal, funded groups for clinical studies on trauma analogous to those on, for example, oncology, currently exist. For example, at present, three major cooperative groups perform multi-institute trials on various diagnostic and therapeutic approaches to cancer. These are funded by the National Cancer Institute and by the pharmaceutical industry, foundations, and donations from individuals. The groups are the Cancer and Leukemia Group B, the Eastern Cooperative Oncology Group, and the Southwest Oncology Group. Studies
are performed at medical centers and clinics throughout the country, and the results of those studies are constantly reported in the open literature. A group or groups that would organize trauma studies suitable for multicenter implementation at existing trauma centers have the potential to address important questions and generate valuable data. A funding mechanism would be needed to cover the costs of the research that would be above and beyond the costs of care already expended for these patients.
Conclusions and Recommendations
Military-civilian research and education opportunities should be expanded and facilitated. The Civilian Level I and Level II Trauma Centers have the potential to evaluate outcomes and costs, to transport trauma patients, and to score the severity of injuries. Although severe limitations on the comparability of the civilian and military situations exist, the best available models are the civilian trauma centers. Progress in the civilian sector, for example, has had military applications in the following areas:
- preservation of blood,
- magnitude and type of fluid therapy,
- helicopter transportation, and
- treatment of burn injuries.
Civilian trauma centers should be used as an educational resource for military residency programs as well as for continuing medical education (CME) for career officers.
The committee found that much of the earlier work in the field of traumatic shock has been tainted by the failure to recognize the differences between pure hemorrhagic shock and traumatic shock associated with tissue injury, the failure to standardize animal models with regard to anesthesia, and the failure to observe subjects for longer-term survival. Clinical research has been hampered by the lack of an organized national approach to trauma research that takes advantage of the considerable clinical material and research expertise among the regional trauma centers. Advances in the treatment of traumatic shock will be enhanced significantly by improved approaches to research performed in studies with both animals and humans.
More specifically, the committee found that animal models in shock research have been broadly selected for convenience or availability rather than specific species-related reasons. It also found that the period of observation has been too short to justify the drawing of any conclusions about survival or mortality. Furthermore, although the use of anesthesia is appropriate for invasive protocols with animals, there is strong evidence that inhalational and intravenous anesthetics as well as many related drugs produce alterations in baseline cardiovascular functions. In addition, they inhibit the physiologic responses to hemorrhage in a significant way.
Many different technical models for the evaluation of hypovolemic shock are available. However, the two major technical models, the controlled and the uncontrolled hemorrhage models, differ significantly in their reproducibilities. Furthermore, many shock protocols are not able to define the beneficial effects of therapies since they have not established clear-cut and reproducible mortality rates. The majority of experimental hypovolemic shock studies deal solely with blood loss and do not address the problem of coincidental tissue injury, which is more typical of the injury sustained on the battlefield. Clinical trials often provide the most definitive proof of the safety and efficacy of diagnostic and therapeutic interventions for humans and are an integral part of the regulatory approval system for new agents. There is a large volume of civilian trauma in the United States, and many of the trauma patients are treated in well-organized trauma centers. Although there are identifiable differences between civilian and military trauma, there are also basic questions that could be approached in the civilian setting to obtain data useful to the military.
Finally, approaches to both current treatment and future research are hampered by inadequate methods for classification of the severity of trauma; such classifications are essential to evaluations of the efficacies of new treatment protocols that involve modifications in fluid formulations or novel therapies. Current trauma indexing systems are inadequate for use in future trauma research. Therefore, the committee makes the following recommendations.
Recommendation 6.1 Laboratory research should be reproducible and relevant to the clinical scenario. For fluid resuscitation research, the experimental design of animal research should be guided by the following principles:
- when feasible, the experimental model should include soft-tissue injury in addition to hemorrhage;
- controlled hemorrhage protocols are preferred to uncontrolled hemorrhage models;
- when feasible, protocols that do not require anesthesia are preferred. If anesthesia is required, the depth of anesthesia should be reproducible, and the anesthetic agent should be selected to minimize alterations in the physiologic responses to hemorrhage;
- experimental animal species should be selected on the basis of clinical relevance, and will vary depending on the research question;
- if survival is an endpoint, mortality should be measured for at least 5 days; and
- the experimental design should establish a reliable 50 percent lethal dose (LD50) for the control group.
Recommendation 6.2 A national study group should be convened to develop and implement clinical research, including multicenter clinical trials on selected topics at existing regional trauma centers.
Federal agencies, including the U.S. Department of Defense, the U.S. Department of Veterans Affairs, and the National Institutes of Health, and national professional organizations, should collaborate with each other and with the private sector in this activity.
Recommendation 6.3 A new system for categorizing injury in trauma care should be developed.