A critique of fluid bolus resuscitation in severe sepsis

Resuscitation of septic patients by means of one or more fluid boluses is recommended by guidelines from multiple relevant organizations and as a component of surviving sepsis campaigns. The technique is considered a key and life-saving intervention during the initial treatment of severe sepsis in children and adults. Such recommendations, however, are only based on expert opinion and lack adequate experimental or controlled human evidence. Despite these limitations, fluid bolus therapy (20 to 40 ml/kg) is widely practiced and is currently considered a cornerstone of the management of sepsis. In this pointof-view critique, we will argue that such therapy has weak physiological support, has limited experimental support, and is at odds with emerging observational data in several subgroups of critically ill patients or those having major abdominal surgery. Finally, we will argue that this paradigm is now challenged by the findings of a large randomized controlled trial in septic children. In the present article, we contend that the concept of large fluid bolus resuscitation in sepsis needs to be investigated further.

If a clinician investigator enters the terms 'sepsis' AND 'fl uid resuscitation' AND 'humans' AND 'clinical trials' into the PubMed electronic reference library system, he or she should be able to identify 84 publications. Th is result might initially suggest that there is a treasure trove of experimentally robust evidence to guide clinicians in this fi eld and support the notion that fl uid resuscitation leads to better outcomes in septic patients. When the search is narrowed to randomized controlled trials, however, the list rapidly shortens to 47 papers -and disappointment sets in when these studies are analyzed in detail, as only two papers remain that use more than physiological outcomes as end points [1,2]. As of 25 May 2011, no studies existed comparing the initial resuscitation of patients with sepsis or septic shock with fl uid boluses versus no fl uid boluses; or indeed comparing large fl uid boluses with moderate fl uid boluses combined with vasopressor drugs. Finally, the only multicentre randomized control led studies available were designed to address the type of fl uids to be used for such resuscitation [1,2], not whether the fl uids should have been given in the fi rst place. Additionally, only two studies report data on mortality [1,2]. Most of the human studies available focus on the physiological eff ects of fl uid boluses [3][4][5] rather than on patient-centered outcomes.
Despite the limited evidence, such is the hold of the fl uid bolus resuscitation paradigm on the mind of critical care physicians that no one has ever considered challenging this prevailing dogma. Now that the FEAST study has created such a challenge [6], it seems important to refl ect on why we believe fl uid bolus resuscitation should be a cornerstone of the management of severe sepsis or septic shock.

The philosophical fallacy: bolus fl uid resuscitation is self-evidently benefi cial
When fl uid boluses are given to septic patients, clinicians are typically confronted with an unwell, hypotensive patient. Th e low blood pressure is often associated with an increased blood lactate concentration, a low urinary output, and other signs and symptoms of physiological distress. Clinicians then make a compelling patho physiological inference and reason that there must be inadequate vital organ blood fl ow (typically described by the loose term perfusion) to explain this clinical picture. Th is inference is necessary because vital organ blood fl ow cannot be accurately measured in acutely septic patients.

Abstract
Resuscitation of septic patients by means of one or more fl uid boluses is recommended by guidelines from multiple relevant organizations and as a component of surviving sepsis campaigns. The technique is considered a key and life-saving intervention during the initial treatment of severe sepsis in children and adults. Such recommendations, however, are only based on expert opinion and lack adequate experimental or controlled human evidence. Despite these limitations, fl uid bolus therapy (20 to 40 ml/ kg) is widely practiced and is currently considered a cornerstone of the management of sepsis. In this pointof-view critique, we will argue that such therapy has weak physiological support, has limited experimental support, and is at odds with emerging observational data in several subgroups of critically ill patients or those having major abdominal surgery. Finally, we will argue that this paradigm is now challenged by the fi ndings of a large randomized controlled trial in septic children. In the present article, we contend that the concept of large fl uid bolus resuscitation in sepsis needs to be investigated further.
An additional inference is then made that that the patient's cardiac output must be low because the vasodilatation, which is responsible for the patient's low blood pressure, must have decreased venous return and thereby preload. Th is is despite the fact that no multicenter case series exists to demonstrate in unresuscitated sepsis that the cardiac output is typically below the normal reference range. Nonetheless, once these inferences are made, then rapid bolus fl uid resuscitation becomes physio logi cally logical and the need for its administration becomes self-evident. Such fl uid bolus becomes the best means by which cardiac output can be increased, organ blood fl ow restored and arterial blood pressure improved.
Th e above paradigm rests, of course, on the untested assumption that the above increases are both necessary and life-saving and that using vasopressor drugs or moderate fl uid therapy with vasopressor drugs to achieve them would lead to inferior patient outcomes. We can fi nd no evidence of consideration being given to the concept that increasing the delivery of cytokine-rich, potentially toxic blood to the organs may, in fact, be deleterious and increase the statistical risk of immunemediated organ injury and failure. Similarly, we can fi nd no evidence of consideration being given to the possibility that, to a degree, this hypotensive state may have evolved as a protective mechanism through millions of years of mammalian evolution and that therefore caution should be applied when administering therapies such as intra venous fl uid boluses that have not been part of such an evolutionary armamentarium.
Given this loose construction of post hoc inferences and teleological argument, many fundamental questions require an answer. For example, does fl uid bolus resuscitation in patients with severe sepsis actually increase vital organ blood fl ow or even cardiac output? If it does, what is the magnitude of its eff ect on both the blood fl ow and cardiac output for a given dose? How long does this eff ect last? What is the physiological price (positive fl uid balance, worse gas exchange, organ edema, acid-base changes) paid to achieve this eff ect? What evidence do we have to answer some of these questions in man?
Unfortunately, there are no convincing answers to any of the above questions. No human studies have measured cardiac output at presentation to the emergency department in septic patients before and after fl uid resuscitation, let alone made attempts to measure vital organ blood fl ow. No human studies have measured how much such cardiac output increases when a recommended fl uid bolus [7][8][9] of 20 to 30 ml/kg over 30 to 60 minutes is given. No human studies have reported how long such an increase might last. No human studies have reported the magnitude and duration of the eff ect of fl uid bolus resuscitation on the mean arterial blood pressure, and so on (the fact that many of these patients are admitted to intensive care to receive vasopressor therapy must, at the very least, imply that fl uid boluses do not reliably restore target blood pressure). Additionally, no human studies have measured what happens to organ blood fl ow (perfusion) before and immediately after fl uid bolus resuscitation. Finally, until recently, no study had asked the ultimate question: are the physiological gains (if they exist) worth the physiological costs (as listed above)? We now know that the answer in children in resource-poor countries is a resounding 'no' . It seems open to question that a therapy associated with close to a 50% increase in the risk of death in African children would instead decrease the risk of death in adults in resource-rich countries instead.

The experimental fallacy: animal experiments prove it
Sometimes interventions cannot be easily studied in a controlled fashion in humans. Animal experiments then provide the only scientifi c knowledge that guides clinical practice.
In this fi eld one might reasonably expect a large body of research to consistently show that fl uid resuscitation increases survival in models of sepsis compared with no fl uid resuscitation. In fact, if the clinician investigator using the above search terms applied to human studies now confi ned them to animal studies, he or she would fi nd 284 papers. Unfortunately, however, of these only four assess the impact of fl uid bolus resuscitation compared with no fl uid bolus resuscitation or limited fl uid therapy on mortality.
One study in 90 mice found increased survival with fl uid bolus resuscitation [10]; another study in rats also found improved survival [11]. A third study in 48 pigs compared fi xed moderate fl uid resuscitation (10 ml/kg) with higher volume fl uid resuscitation (20 ml/kg), accompanied by boluses of 50 ml hydroxyethyl starch in response to low cardiac fi lling pressures and/or oliguria. Th is study found increased mortality in animals allocated to fi xed higher volume resuscitation [12]. A fourth study in mice also found that fl uid therapy increased survival compared with control. However, such therapy was given as subcutaneous boluses every 6 hours, making inferences concerning intravenous fl uid boluses in human sepsis problematic [13]. Th is level of experimental evidence is hardly overwhelming.
Yet, one might argue -if experimental evidence showed that septic models typically, consistently and reliably induce a low cardiac output, low organ blood fl ow state -that such evidence might be enough to justify our inferences about humans and fl uid bolus therapy. Unfortunately, this is not the case because several models show the opposite (high cardiac output and increased organ blood fl ow) [12][13][14][15]. More over, even in severe septic shock, magnetic resonance technology demonstrates continued preservation of renal ATP [16], suggesting no bio-energetic defi cit to justify the need for increased perfusion. Yet even if this was the case, evidence in normal and septic animals that fl uid therapy consistently leads to major (>50%) and sustained (hours to days) increases in organ blood fl ow would surely provide some justifi cation for fl uid boluses in sepsis. Unfortunately, animal studies show that the eff ects of bolus fl uid resuscitation typically dissipate within 45 to 60 minutes for crystalloids and somewhat longer for colloids [17,18]. What should one then do for the next 23 hours? Repeat the bolus every hour until the patient is 10 liters of fl uid overloaded? Given the association between a positive fl uid balance and increased mortality in critically ill patients with sepsis and acute kidney injury [19], or with sepsis and acute respiratory distress syndrome (ARDS) [20], this approach seems illogical and possibly dangerous [21].

The analogy fallacy: we should give fl uids because they help in other conditions
Although there are no controlled human studies supporting fl uid bolus resuscitation in sepsis and there are also no convincing supportive animal studies, clinicians could justify such treatment by analogy: fl uid resuscitation makes other really sick patients better, so it should make septic patients better too. Th is argument appears to have merit as it seems self-evident that fl uid therapy has saved and still saves millions of patients with cholera and other forms of bacterial or viral diarrhea leading to severe fl uid depletion. Yet these patients have clearly lost large amounts of isotonic fl uids and the analogy argument stops right there. Fluid resuscitation, on the other hand, worsens outcomes in bleeding patients with penetrating torso injury [22], positive fl uid balances increase morbidity in patients with acute kidney injury [19] and ARDS [20], and fl uid liberal approaches increase morbidity in patients receiving colorectal surgery [23]. Once again, this is not exactly overwhelming evidence that fl uid boluses in sepsis are clearly justifi ed, even by analogy with other, similar, medical conditions.

The phenotype fallacy: the abiding power of association
Perhaps the strongest driver to fl uid bolus resuscitation in sepsis has been the association between a hyper dynamic cardiac output state and better outcome in patients with critical illness [24]. Th is association has led people to believe that, if only we could change patients with a normal cardiac output or a low-normal cardiac output or a minimally increased cardiac output into patients with a markedly hyperdynamic (high or very high cardiac output) circulation, just like the patients who seem to do better, then all would be well. Th is notion has been challenged by multiple studies showing that increasing cardiac output or mixed venous oxygen saturation with or without fl uids or inotropic drugs does not achieve improved outcomes and may in fact worsen them [25,26].
Moreover, it is a formal logical fallacy to reason that 'survivors have a hyperdynamic circulation; therefore, if a patient can be made to have a hyperdynamic circulation, he or she will survive' . It is also a pathophysiological fallacy that one can take patients with a hemodynamic phenotype associated with a statistically less favorable outcome and, with fl uid loading or inotropic drugs or both, change them into the cardiovascular phenotype of those patients who seem to do better and then expect such patients to achieve better survival. Th ere are several potential reasons why these patients display a nonhyperdynamic phenotype in the fi rst place (underlying co-morbidities, diff erent physiological reserve, diff erent bacterial load, diff erent genetically-driven responses, and so on). Th ese forces are the likely major drivers of their outcome. Changing one aspect (cardiac output) of their phenotype with fl uids will not alter these driving forces and is unlikely to alter prognosis. To use an analogy: repainting the façade of a crumbling building will no doubt change its 'phenotype' , but the building will still collapse.

Where are we now?
So, up to 26 May 2011, we had no controlled human data and no understanding of the physiological eff ects of fl uid bolus resuscitation in humans presenting with severe sepsis compared with just simple fl uid replacement therapy. Instead, we used untested inferences, had patchy and contradictory animal data, relied on weak clinical analogies and were enticed by the phenotype association fallacy. Such a situation alone should have demanded a serious reassessment of our faith in fl uid boluses for the resuscitation of septic patients anyway. Now, however, we have a large randomized controlled trial in more than 3,000 subjects (children) with no comorbidities and no artifi cial confounders (rescue mechanical ventilation, rescue vasopressor therapy, rescue dialysis, rescue extracorporeal membrane oxygenation) [6]. Th is trial found that saline or albumin fl uid resuscitation increased mortality by almost 50%. Th e study has many strengths, but it also has many characteristics that make it open to criticism in terms of its relevance to patients seen in resource-rich countries. Key concerns are that only a minority of children had severe hypotension and all such severely hypotensive children received fl uid bolus resuscitation; one-third of children had hemoglobin <5 g/dl, suggesting that fl uid-induced hemodilution in itself is dangerous; and 80% had respiratory distress, making fl uid loading particularly problematic in the absence of rescue mechanical ventilation. Th ese observations make these patients perhaps more closely related to those in the FACTT study, where fl uids were shown to adversely aff ect the outcome of patients with the combination of ARDS and sepsis [20]. Th eir relevance to septic patients presenting to emergency departments of resource-rich countries is therefore also open to challenge.
Th ese concerns highlight the fact that one should be cautious in using this FEAST trial as a clear-cut analogy for how sepsis resuscitation should be managed in resource-rich countries. Given the fi ndings and limitations of the FEAST study, however, and the context discussed above, it remains legitimate to ask: what is the role of fl uid bolus resuscitation in human sepsis? While in resource-rich countries the above diff erence in mortality may almost entirely disappear as rescue therapies are applied to attenuate any potential adverse eff ect of fl uid overload (ARDS, pulmonary edema, acute kidney injury), there is now concern that such therapy may not be as benefi cial as previously believed.
Given the risks and the frequency of use, the time has come to conduct randomized controlled trials in septic children and adults in developed countries to attempt to answer this important question. Fortunately, several studies are currently underway in the USA (PROCESS), in Australia and New Zealand (ARISE), and in the United Kingdom (PROMISE) that address the issue of whether early goal-directed resuscitation increases survival in severe sepsis. Th ese and other such studies will also provide crucial information on the association between fl uid bolus resuscitation and outcome, and will be an important additional step in the reassessment of this therapy.
In the meanwhile, the authors believe that a moderate position might be reasonable until further information is available. Th e treatment of patients with severe sepsis/ septic shock in resource-rich countries need not yet change. While clinicians in developed countries are unlikely to ever consider managing these patients without fl uid resuscitation, however, it is likely that there will be renewed interest in applying a strategy that moderates the amount of fl uid given to perhaps 20 to 25 ml/kg over the fi rst 6 hours and promotes the earlier use of vasopressor drugs. Th is approach to resuscitation is already widespread in Australia and New Zealand [27] and is associated with excellent outcomes.