In a noninjured, nonseptic, healthy state, oxygen consumption (VO₂) is a closely regulated process because oxygen serves as the critical carbon acceptor in the generation of energy from a wide variety of metabolic fuels. Post-traumatic hemorrhage leads to a hypovolemia in which blood flow and consequently oxygen delivery to vital organs are decreased. When oxygen delivery is decreased to a degree sufficient to reduce VO₂ to below a critical level, a state of shock occurs, producing ischemic metabolic insuffiency 123. This degree of restriction in VO₂ can also be produced by cardiogenic or vasodilatory shock, in which oxygen delivery is restricted by low flow. When this critical level of oxygen restriction is reached, an oxygen debt (O₂D) occurs. In the literature, the terms 'oxygen debt' and 'oxygen deficit' are used interchangeably and are defined as the integral difference between the prehemorrhage/pretrauma resting normal VO₂ and the VO₂ during the hypovolemic, hemorrhage period 456789. For purposes of simplification, the term O₂D ('oxygen debt') is used in this review. The presence and extent of an O₂D is further highlighted by an increase in the unmetabolized metabolic acids generated by the anaerobic processes. It is the close congruence of O₂D and related metabolic acidemia that permits precise quantification of the severity of the ischemic shock process in both animals and humans.

The aim of this review is to demonstrate the quantitative similarity between experimental O₂D shock and that induced in humans by post-traumatic or severe hemorrhagic, hypovolemic conditions. It also examines the use of metabolic correlates of O₂D as indices of the severity of the shock process in two mammalian species and in humans, and the value of these correlates as guides to the adequacy of volume-mediated resuscitation.

This review is based on a search of the Medline and Cochrane Library databases from 1964 to December 2004. The search terms 'oxygen debt or deficit', 'base excess or deficit', 'lactate', 'hemorrhagic shock' and 'multiple trauma' were used. These terms were mapped to Medline Subject Headings (MESH) terms, as well as being searched for as text items. The following combinations were studied: 'oxygen debt' or 'oxygen deficit' and 'hemorrhagic shock', 'lactate' and 'multiple trauma', as well as 'base excess' or 'base deficit' and 'multiple trauma'. No language restrictions were applied.

That post-traumatic shock is initiated by acute volume loss was first noted by Cannon 10 and later demonstrated by the experimental studies conducted by Blalock 11. Subsequently, Wiggers 12 and Guyton 13 developed a variety of animal models based on controlled hemorrhage. Other models involving uncontrolled bleeding 1415, fixed volume loss 1617181920, or a defined level of hypotension 1619202122 have been used. In previous studies, the severity of shock was defined by the degree and duration of the resulting hypovolemia. Thus, attempts were made to quantify the effectiveness of resuscitation by assessing the improvement in blood pressure or perfusion occurring in response to different volumes of electrolyte, colloid, or blood-containing fluids, which are administered to prevent death during the immediate postshock period.

In the clinical arena, this issue became acute during World War II, when fluid transfusion and use of blood and blood products as a means of effectively restoring blood volume became a realistic possibility. Consequently, volume infusion and blood or blood product transfusion were used extensively for the first time during the North African Campaign by US and UK forces 23, and was a primary modality for treatment of shock in the Korean War 24. These clinical advances led to extensive efforts to elucidate human hypovolemic shock and to establish experimental models that emulate clinical shock. The most extensive series of clinical/physiologic studies were performed in postoperative 2526 and posttrauma 27 shock patients, in whom the response to volume infusion was evaluated. These and other studies 2829 of resuscitation after hypovolemic shock demonstrated the fall in VO₂ associated with the decrease in cardiac output, and demonstrated the arterial vasoconstriction that occurred in an attempt to compensate for the fall in blood pressure. They also demonstrated the postresuscitation hyperdynamic state, in which cardiac output rises to permit an increase in VO₂, apparently compensating for and even exceeding the initial fall in VO₂ 1226. These data appeared to validate in humans the 'oxygen deficit' concept initially enunciated by Crowell and Smith 4 based on experimental findings. Nevertheless, in spite of these animal and clinical physiological studies, controversy remains with regard to the optimal nature and magnitude of postshock volume resuscitation. Options include massive isotonic fluid replacement 3031, use of intravascular colloid containing fluids 32, and substitution with small volume hypertonic saline after hemorrhage 33.

Recently, however, a new resuscitation concept has emerged for application when the degree of autogenous vascular control is uncertain, namely permissive hypotension; this is achieved by administering small volumes of resuscitation fluid, permitting only minimal increase in perfusion until full vascular control of hemorrhage can be achieved by surgical intervention 3435. Although the statistical validity of the initial human studies 34 has been questioned 36, the concept appears to have some utility, provided that sufficient levels of tissue VO₂ can be achieved to prevent the acute consequences of cellular ischemia 37. These issues focus on the need for accurate and easily measured correlates of O₂D that can quantify the severity of O₂D and that can be monitored on a continuing basis during resuscitation.

A large number of animal models have been developed to simulate the critical end-points of hemorrhagic shock. Deitch 38 divided these models into three general categories: uncontrolled bleeding, controlled bleeding volume, and controlled decrements in blood pressure.

A more physiologically relevant animal model is needed because of the clinical requirement to progress beyond the traditional end-points of volume loss and subsequent blood pressure levels 39. Furthermore, such a model is needed to determine why a state of hyperdynamic cardiovascular compensation develops after hypovolemic shock 2540. Also, numerous clinical studies have shown that hypovolemic trauma patients can remain in a state of shock, with evidence of inadequate tissue perfusion and metabolic acidosis 294142, even if the traditional end-points have been normalized 122540. This is reflected in the present definition promulgated by the American College of Surgeons: 'Shock is an abnormality of the circulatory system that results in inadequate organ perfusion and tissue oxygenation' 3. This understanding of the relationship between shock and inadequate perfusion has led to the development of a possibly more clinically relevant fourth general category of experimental hemorrhagic shock models, based on the concept of repayment of shock-induced O₂D. Table 1 summarizes the historical development of hemorrhagic shock models with O₂D as an end-point. It is based on a systematic Medline/Cochrane Library literature search using the terms 'oxygen debt or deficit' and 'hemorrhagic shock'. From 52 suggested articles, only 13 that strictly dealt with defined O₂D in a hemorrhagic shock model are included.

Thus, development of models of hemorrhagic shock must follow current knowledge and must consider indices of inadequate organ perfusion and tissue oxygenation, which are more meaningful end-points in the clinical setting 4. Up to the 1990s O₂D was used as a secondary end-point in pressure-controlled or volume-controlled models of hemorrhagic shock (Table 1); in contrast, Dunham and coworkers 9 described a canine model of hemorrhagic shock in which O₂D was used as the independent predictor of the probability of death and organ failure. This canine model, which was validated in subsequent studies 3743, follows the hypothesis that the total magnitude of O₂D reached during hemorrhage is the critical determinant of survival, and that this variable and its metabolic consequences of lactic acidemia and base deficit better reflect the severity of the cellular insult than do traditional variables such as bleeding volume and blood pressure. This hypothesis was also verified in a pig model of O₂D hemorrhagic shock 44.

In healthy young men, the resting VO₂ has been shown to average 140 ml/min per m². If this VO₂ is decreased by reduced blood flow with restriction in organ and tissue perfusion, a critical level of ischemia is induced, with a disparity between the oxidative requirement mandated by the level of metabolism and the level of oxygen delivery – an O₂D occurs. Physiologically, if resuscitation is performed before a fatal metabolic debt is incurred then there is rapid repayment of the O₂D, with VO₂ overshoot as the unmetabolized acids are oxidatively metabolized during the reperfusion period. This is effected by an increase in oxygen delivery mediated by a rise in cardiac output – the 'hyperdynamic state' 126. However, as the O₂D accumulates the likelihood of cellular injury increases, with reduction in cellular membrane integrity and consequent cell swelling as intracellular water increases. Later in the process intracellular organelles become damaged, cellular synthetic mechanisms cease, and finally lysosomes are activated, which results in cell necrosis and death 45. Even at less severe O₂D levels, mechanisms that initiate later apoptosis are activated 46. Depending on the extent and severity of the cellular injury, specific features of multiple organ failure (MOF) are initiated. Cells with the greatest oxidative requirements (e.g. brain, liver, kidney, myocardium and immunologic tissues) appear to be most vulnerable to O₂D-induced injury or cell death.

Although evidence of cellular and organ failure often appears at various time points after recovery from O₂D, it has long been known that the relationship between O₂D and acute death can be quantified. Crowell and Smith 4 were the first to describe the effect of O₂D in terms of a lethal dose (LD) effect. In their canine studies, O₂Ds of 100 ml/kg or less were not lethal; O₂Ds of 120 ml/kg led to an LD₅₀ (i.e. a dose sufficient to kill 50% of the population studied); and O₂Ds of 140 ml/kg or more were invariably fatal. A more precise quantification of the probability of death with increasing O₂D in the same animal species was conducted by Dunham and coworkers 9, who established a complete probability of death function (Fig. 1a). Their studies noted an exponential relationship between probability of death and O₂D, such that although the LD₂₅ was at an O₂D of 95.5 ml/kg, the LD₅₀ lay at 113.5 ml/kg and the LD₇₅ was at 126.5 ml/kg. This relationship has repeatedly been confirmed in dogs by more recent studies 3743. Studies in pigs 44 have found a nearly identical relationship, although the LD₅₀ for the pig is at a slightly lower O₂D/kg (95 ml/kg; Fig. 1b), corresponding with values calculated by Hannon and coworkers 19 in the same species. This difference between the two animals appears to reflect the greater percentage of adipose tissue in the pig as compared with the much leaner hound dog over the same range of body weight.

To understand better the concept of hemorrhage-induced O₂D accumulation and its repayment by volume infusion, experimental animal responses were recently studied by Siegel and coworkers 37. In that study 40 dogs were bled to achieve an O₂D of 104 ± 7.6 ml/kg at 60 min after initiation of hemorrhage (estimated probability of death: 35.7% 9; actual death rate: 40%; shed blood volume [SBV]: 71.0 ± 6.8% of the animals' estimated total blood volume 37). Following hemorrhage, the animals were either given no initial resuscitation for 2 hours and then fully resuscitated with a volume of 5% colloid equivalent to 120% of their SBV. Alternatively, they were randomly assigned to initial resuscitation (R1) with a predetermined percentage of their SBV (again by infusing 5% colloid) equivalent to 8.4%, 15%, 30%, or 120% of their SBV. Then, after a 2 hour delay period in which no further volume resuscitation was given, the animals were given the remaining portion of the calculated 120% of the SBV lost during hemorrhage (delayed resuscitation: R2). This made the final quantity of volume replacement in each animal equal to 120% of the SBV. It is important to note that in those animals given no initial resuscitation, the O₂D accumulation rate continued to rise either at the same (or slighly lower) rate as during the hemorrhage to or above the 90% mortality level, even though no further blood loss occurred. However, in all instances of R1, the O₂D also continued to rise slightly until a critical quantity of R1 was given (at least 30% of the SBV), but the initial rate of recovery from the hemorrhage-induced O₂D level was increased proportionately to the increase in R1. This relationship between the magnitude of initial resuscitation and the rate of O₂D decrease was highly significant (P < 0.001) and predicted later evidence of cell death and organ failure in 7-day postshock survivors 37. In contrast to the significant discrimination provided by O₂D level, the simultaneously measured mean blood pressure responses were not found to be significant predictive of the adequacy of resuscitation 37.

These canine data and parallel data obtained in the O₂D pig model 44 demonstrate that quantity of blood volume loss or replacement, blood pressure, and even cardiac output response do not reflect well the severity of shock or the effectiveness of volume resuscitation. However, these end-points are well defined in a quantitative manner by the magnitude of the O₂D and by its rate of resolution during the resuscitation period, independent of species.

A considerable body of evidence has accumulated that strongly suggests, both in the animal setting 9374344 and in humans 2941474849, that metabolic acids in blood or plasma are indices that reflect the degree of tissue hypoxia associated with hypovolemic ischemia. In this review the strict definition of base deficit (BD) – namely, a negative base excess – is used 294950, with a decrease in base excess with increasing metabolic decompensation implying progressively negative values (e.g. -6 mmol/l to -10 mmol/l). However, because BD implies a negative base excess, only positive values of BD (without the minus sign) are used in the present review.

As the concept of O₂D as the key process determining outcome evolved, one of the major goals of experimental studies was to examine the relationship between lactate or BD and hemorrhage-induced O₂D 9374344. This significant relationship was repeatedly demonstrated in progressive hemorrhage, with increases in BD or lactate being paralleled by increases in O₂D 9374344. Similar significant relationships were noted between decreases in these metabolic variables; the O₂D fell during volume resuscitation, regardless of whether the fluid was crystalloid, hypertonic saline, carbonate/gelatine, colloid, or whole blood 937434451. The rate of decline in O₂D (and BD and lactate) was significantly more rapid when an oxygen-carrying solution of recombinant hemoglobin was employed for resuscitation 43. The relationship between BD and O₂D tended to reflect better the effectiveness of increases in initial volume resuscitation, whereas lactate reflected the overall trend in effectiveness of resuscitation but with less discrimination 37. Very similar, albeit more variable, significant relationships (P < 0.0001) for the two metabolic correlates of O₂D were also noted in the pig model 44. The greater variability found in the pig may reflect a closer similarity to broad range of adipose tissue found in humans. Nevertheless, BD and lactate appear to correlate best with O₂D in experimental hemorrhagic shock. This relationship is significant across species 944.

Finally, the relationship between O₂D and BD can be used to address the problem of quantifying the effectiveness of small volume resuscitation during permissive hypotension. In other words, the paramedic, surgeon, or intensivist could resuscitate a hypovolemic patient to a level at which perfusion will yield a reduction in O₂D that will allow critical organ oxidative metabolism to be maintained, at a blood pressure that will not encourage further hemorrhage until all open vessels are Although it is generally not practical to measure O₂D in humans, a model for this approach using BD can be derived from animal data. In a canine O₂D shock model, Siegel and coworkers 37 demonstrated that animals that were effectively volume resuscitated moved progressively down the O₂D/BD regression line to lower values compatible with a reduced probability of death 37. However, those animals that received inadequate volume resuscitation, particularly those that died during the 2 hour postshock period, moved to progressively higher points in the O₂D/BD relationship. A similar but less quantifiable relationship was found for the O₂D/lactate relationship.

In this respect attention must be paid to the recent development of hemorrhagic shock models with a target endpoint of metabolic acidosis 525354. Schultz 52 and Powell 53 and their groups studied bacterial translocation and restoration of central venous oxygen saturation after BD-guided hemorrhagic shock in rats. Also, DeAngeles and coworkers 54 studied resuscitation from BD/lactate guided hemorrhagic shock with diaspirin cross-linked hemoglobin, blood, and hetastarch in sheep. Thus, the use of BD and lactate as clinically useful surrogates for O₂D is strongly supported by experimentation in numerous animal species.

The search for identifiable and easily measured metabolic correlates of shock that could be used to quantify the severity of human circulatory failure began with the pioneering work of Huckabee 55, Weil and Afifi 56 and Harken 57. These studies confirmed that the circulating level of lactate provided an indication of the anaerobic component induced by the shock process. Bakker and coworkers 58 reported evidence that the dependency on oxygen supply to body tissues was associated with increasing lactate levels.

Table 2 provides a summary of literature on lactate as an outcome predictor in adult multiple trauma patients based on a systematic Medline/Cochrane Library literature search, using the terms 'lactate' and 'multiple trauma'. Of 59 originally retrieved articles, 27 are specifically noted in the present review because they strictly deal with lactate as an outcome predictor in multiple trauma patients. In almost 3000 multiple trauma patients lactate was shown to predict outcome following postoperative complications, intracranial pressure, infection, sepsis, adult respiratory distress syndrome (ARDS), MOF, injury and hemorrhage severity, and survival.

Clinically, however, it is important to note that not all cases of hyperlactatemia are accompanied by acidosis, and neither are all cases of hyperlactatemia caused by O₂D. Other metabolic dysfunctions may also be associated with hyper-lactatemia 59 and can confuse assessment of the O₂D effect, as can excessive alcohol intake and acute cocaine use. The most prominent group of patients with increased lactate levels in the absense of hypovolemia are patients with severe sepsis 28. However, diabetic patients with keto-acidosis have increased lactate, and in patients with impaired hepatic function lactate uptake may be reduced and lactate levels may rise. Of specific importance in patients resuscitated from hemorrhagic shock is that administration of large quantities of exogenous lactate (e.g. via mass infusion of Ringer's lactate) has been shown to increase lactate to levels significantly greater than those expected to result from the shock process alone 60. This clearly may distort interpretation of lactate levels as a clinical diagnostic tool. Furthermore, the reduction in oxygen delivery that induces O₂D also causes other metabolic acids to accumulate in the extracellular/intravascular components, and so plasma lactate levels may not always quantitatively reflect the O₂D process. Thus, the origin of a hyperlactatemia is clinically important and has direct implications for treatment choice.

Although the use of lactate-free resuscitation fluids may become routine in the future 60, the current widespread use of Ringer's lactate may be a further reason why it remains unclear whether the lactate level on hospital admission is prognostically significant in multiple trauma patients. Several studies have noted the predictive value of the initial lactate level 586162, but others have shown other variables to be equivalent 63 or even better 29 in outcome prediction. In contrast, more than one study found no significant correlation between initial lactate level and posttrauma outcome 49646566. Nevertheless, in a study of 375 trauma patients admitted directly from the scene of injury to a level I trauma center 62, simultaneously obtained arterial and peripheral venous lactate levels were shown to be highly correlated, and a lactate threshold level of > 2 mmol/l appeared to predict the likelihood of the Injury Severity Score (ISS) being 13 or greater with a high degree of accuracy. Thus, lactate appears to represent a good triage tool.

The data reported above strongly indicate a need to add quantitative estimates of the effectiveness of perfusion and VO₂ to hemorrhagic shock studies. Currently, indirect measurement techniques that reflect cellular oxygen utilization and perfusion either systemically (lactate and BD) or locally (gastric intramucosal pH and microdialysis 78) predominate. It would be ideal to measure O₂D at the cellular level as an end-point of experimental and clinical hemorrhagic shock. The muscle beds, subcutaneous tissue, and even skin have been advocated as sites at which perfusion may be more directly measured at the tissue level. Hartmann and coworkers 79 found good correlations between subcutaneous and transcutaneous partial oxygen tension (PO₂), and with gastric tonometry in pigs. Subcutaneous PO₂ tissue probes have been used in the experimental setting 80 as well as in severely injured patients 8182. McKinley and coworkers 83 studied skeletal muscle PO₂, partial carbon dioxide tension, and pH using fiberoptic technology in hemorrhaged dogs, and Knudson and coworkers 84 examined the posthemorrhage and resuscitation oxygen tension response in muscle and liver in pigs. Another technique that holds promise for the future is that of near infrared spectroscopy 85. All of these techniques, along with others currently being developed, may move the endpoints of hemorrhagic shock models to the organ, cellular, and subcellular levels. However, at present these newer technologies require the use of relatively complex and expensive or invasive methodologies, whereas relatively inexpensive handheld devices now exist for rapid field, emergency room, or ICU determinations of lactate and BD 86.

However, it is clear from experimental 87 and clinical studies 8188 that some vascular beds may be more vasoconstricted than others; the skin, subcutaneous and muscle tissue, and intestinal perfusion are sacrificed to preserve cardiac, central nervous system, renal, and hepatic perfusion. Consequently, probes placed in the physiologically expendable tissues may not reflect the true total body situation, and especially vital organ O₂Ds. This contention is supported by the findings reported by Siegel and coworkers 37, which showed that adequate resuscitation with a volume of 30% of SBV could preserve essential organ histology and physiologic function from an LD_35–40 of O₂D without increasing the cardiac index above control preshock levels. Only when the remaining volume of delayed full resuscitation was given did the cardiac index and oxygen consumption rise to hyperdynamic levels, suggesting that a large percentage of this hyperdynamic state is devoted to repayment of the O₂D in less essential organs, which collectively represent the greater portion of body cell mass.

Nevertheless, both animal and clinical data strongly suggest that the overall O₂D and/or its metabolic correlates (BD and lactate) better reflect the severity of shock than do currently available measures of local tissue or organ perfusion. This is shown by the probability of death curves for individual species, and by the relative consistency of LD₂₅ and LD₅₀ points for BD across species and especially in humans, when adjusted for GCS and other significant variables.

We require a more precise technique for assessing total body O₂D, or at least that of critical organs, that can easily and repeatedly be applied in the clinical setting. Until such a technique becomes available the use of BD, either alone or in combination with GCS score and other significant variables of high predictive accuracy (e.g. prothrombin time and age), represents the best present system for clinical assessment of shock severity and success of resuscitation. These variables may be used to obtain information rapidly on a patient's level of compensation in response to posttrauma or hemorrhagic shock either by immediate reference to a predetermined graph (Fig. 3) or by entry of data into a handheld computer for computation of an estimate of probability of death using the regression equation shown above. This would facilitate clinical decision making at the bedside, in the emergency room, or in the ICU.

In conclusion, the data examined in this review strongly indicate that there is a need to add quantitative estimates of O₂D and resulting metabolic acidosis to clinical studies, and that these variables should be considered in the management of patients sustaining severe hemorrhagic shock. The data also suggest that evaluation of metabolic correlates of the total body O₂D (BD and, to a lesser extent, lactate) may be more useful in quantifying the responses of trauma or nontrauma patients to hemorrhage than are estimates of blood loss, quantitative measurements of volume replacement, or blood pressure and heart rate. Finally, we believe that further research based on the parameters of oxygen utilization and O₂D will achieve even better, clinically suitable variables by which to assess the magnitude and severity of human stress physiology and to quantify the effectiveness of resuscitation therapies in the multiple trauma patient or the patient with life-threatening hemorrhage from a gastrointestinal lesion.

ARDS = acute respiratory distress syndrome; BD = base deficit; GCS = Glasgow Coma Scale; ICU = intensive care unit; ISS = Injury Severity Score; LD = lethal dose; MOF = multiple organ failure; O₂D = oxygen debt; PO₂ = partial oxygen tension; ROC = receiver operating characteristic; SBV = shed blood volume; TRISS = Trauma and Injury Severity Score; VO₂ = oxygen consumption.

The author(s) declare that they have no competing interests.

Both authors (DR and JHS) made substantial contributions to the conception and design of this review, and to the acquisition, analysis, and interpretation of data. Furthermore, both authors were involved in drafting the article and revising it critically for important content, and gave final approval of the version to be published.