Background

In patients with septic shock, hemodynamic stabilization using intravenous fluids remains a major therapeutic challenge as numerous questions remain regarding the type, dose and timing of fluid administration. In these patients, fluids play an important role beyond hemodynamic stabilization and resuscitation. Intravenous fluids should be prescribed as any other drug we give to our patients: we should take into account the indications and contraindications for different types of fluids [2,3,4,5,6,7,8]. We should only prescribe fluids when they are clearly indicated and should balance the risk of not administering enough with the increasingly apparent risks of too much fluid.

In this review, we will expand on the concept of the “four D’s” of fluid therapy (drug, duration, dosing and de-escalation). We will also focus on the recent concept defining four different phases in the time course of septic shock (resuscitation, optimization, stabilization and evacuation). Each phase requires a different therapeutic attitude regarding fluid administration. Taking into account both of these concepts in combination with other suggested ideas may promote more rational fluid administration aimed at avoiding both too little and too much. In analogy to the way we handle antibiotic usage in the critically ill, it is now time for fluid stewardship.

The risk of fluid overload

Treating a patient with septic shock inevitably results in some degree of salt and water overload. First and foremost, this is the result of the initial fluid resuscitation with the aim of restoring intravascular volume, increasing cardiac output, augmenting oxygen delivery and improving tissue oxygenation. Salt and water overload can also result from the administration of large volumes of fluid as drug diluents, artificial nutrition and maintenance fluids. The capillary leak that is inherent to sepsis promotes the extravasation of large amounts of fluid, inducing relative central hypovolemia that often requires further fluid administration, despite interstitial oedema. Capillary leak represents the maladaptive, often excessive, and undesirable loss of fluid and electrolytes with or without protein into the interstitium that generates anasarca and end-organ oedema causing organ dysfunction and eventually failure [9]. Fluid overload should be avoided in this setting.

The four D’s of fluid therapy

When prescribing fluids in patients with septic shock, we must take into account their composition and their pharmocodynamic and pharmacokinetic properties. In practice, we should consider the “four D’s” of fluid therapy: drug, dosing, duration and de-escalation (Table 1) [5]. Many clinicians already use these four D’s for the prescription of antibiotics (Table 1).

Table 1 Analogy between the four D’s of antibiotic and fluid therapy.
Full size table

Drug

We should consider the different compounds: crystalloids versus colloids, synthetic versus blood derived, balanced versus unbalanced, intravenous versus oral. The osmolality, tonicity, pH, electrolyte composition (chloride, sodium, potassium, etc.) and levels of other metabolically active compounds (lactate, acetate, malate, etc.) are all equally important. Clinical factors (underlying conditions, kidney or liver failure, presence of capillary leak, acid–base equilibrium, albumin levels, fluid balance, etc.) must all be taken into account when choosing the type and amount of fluid for a given patient at a given time. Moreover, the type of fluid is different depending on the reason why they are administered. There are only four indications for fluid administration, namely resuscitation, maintenance, replacement and nutrition, or a combination.

Resuscitation fluids

Resuscitation fluids are given to correct an intravascular volume deficit in the case of absolute or relative hypovolemia. In theory, the choice between colloids and crystalloids should take into account the revised Starling equation and the glycocalyx model of transvascular fluid exchange [17]. When capillary pressure (or transendothelial pressure difference) is low, as in hypovolemia or sepsis and especially septic shock, or during hypotension (after induction and anaesthesia), albumin or plasma substitutes have no advantage over crystalloid infusions, since they all remain intravascular. However, the glycocalyx layer is a fragile structure and is disrupted by surgical trauma-induced systemic inflammation or sepsis, but also by rapid infusion of fluids (especially saline). Under these circumstances, transcapillary flow (albumin leakage and risk of tissue oedema) is increased, as is the risk to evolve to a state of global increased permeability syndrome (GIPS) [17].

Maintenance fluids

Maintenance fluids are given, specifically, to cover the patient’s daily basal requirements of water, glucose and electrolytes. As such, they are intended to cover daily needs. The basic daily needs are water, in an amount of 25–30 mL/kg of body weight, 1 mmol/kg potassium, 1–1.5 mmol/kg sodium per day and glucose or dextrose 5 or 10% 1.4–1.6 g/kg (to avoid starvation ketosis) [31].

Some specific maintenance solutions are commercially available, but they are far from ideal. There is a lot of debate whether isotonic or hypotonic maintenance solutions should be used. Data in children showed that hypotonic solutions carry the risk of hyponatremia and neurologic complications [32, 33]. However, studies in adults are scarce and indicate that administration of isotonic solutions will result in a more positive fluid balance as compared to hypotonic solutions [34]. This was confirmed in a recent pilot study in healthy volunteers showing that isotonic solutions caused lower urine output, characterized by decreased aldosterone concentrations indicating (unintentional) volume expansion, than hypotonic solutions. Despite their lower sodium and potassium content, hypotonic fluids were not associated with hyponatremia or hypokalemia [24].

Replacement fluids

Replacement fluids are administered to correct fluid deficits that cannot be compensated by oral intake. Such fluid deficits have a number of potential origins, like drains or stomata, fistulas, hyperthermia, open wounds, polyuria (salt-wasting nephropathy, cerebral salt wasting, osmotic diuresis or diabetes insipidus) [4].

Data on replacement fluids are also scarce. Several recent guidelines advise matching the amount and composition of fluid and electrolytes as closely as possible to the fluid that is being or has been lost [35, 36]. An overview of the composition of the different body fluids can be found in the NICE guidelines [35]. Replacement fluids are usually isotonic balanced solutions. In patients with fluid deficit due to a loss of chloride-rich gastric fluid, high-chloride solutions, like saline (0.9% NaCl), might be used as replacement fluid.

Nutrition fluids

Often overlooked, it is about time to consider parenteral nutrition as another source of intravenous fluids that may contribute to fluid overload. Likewise, nutritional therapy in the critically ill should be seen as “medication” helping the healing process. As such, we might consider also the four D’s of nutritional therapy in analogy to how we deal with antibiotics and fluids [5]: drug (type of feeding), dose (caloric and protein load), duration (when and how long) and de-escalation (stop enteral nutrition and/or parenteral nutrition when oral intake improves) [37].

Combination of fluids

A combination of different types of fluids is often justified. For example, numerous combinations may be used in daily practice with regard to resuscitation fluids: blood and crystalloids (trauma), crystalloids early (post-operative hypovolemia), albumin late (sepsis). Similarly, maintenance fluids are often a combination of enteral and parenteral nutrition, other glucose-containing solutions, saline and/or balanced crystalloids to dissolve medications.

Duration

The longer the delay in fluid administration, the more microcirculatory hypoperfusion and subsequent organ damage related to ischaemia–reperfusion injury. In patients with sepsis [38], Murphy and colleagues compared outcomes related to early adequate versus early conservative and late conservative versus late liberal fluid administration and found that the combination of early adequate and late conservative fluid management carried the best prognosis [38] (Fig. 4). Combined data from other studies confirm that late conservative is maybe more important than early adequate fluid therapy [39,40,41].

Fig. 4
figure 4

Impact on outcome of appropriate timing of fluid administration. Bar graph showing outcome (mortality %) in different fluid management categories. Comparison of the data obtained from different studies: hospital mortality in 212 patients with septic shock and acute lung injury, adapted from Murphy et al. (light blue bars) [38], hospital mortality in 180 patients with sepsis, capillary leak and fluid overload, adapted and combined from two papers by Cordemans et al. (middle blue bars) [40, 41], 90-day mortality in 151 adult patients with septic shock randomized to restrictive versus standard fluid therapy (CLASSIC trial), adapted from Hjortrup et al. (dark blue bars) [39]. See text for explanation. EA: early adequate fluid management, defined as fluid intake > 50 mL/kg/first 12–24 h of ICU stay. EC: early conservative fluid management, defined as fluid intake < 25 mL/kg/first 12–24 h of ICU stay. LC: late conservative fluid management, defined as 2 negative consecutive daily fluid balances within first week of ICU stay. LL: late liberal fluid management, defined as the absence of 2 consecutive negative daily fluid balances within first week of ICU stay

Full size image

Dosing

As Paracelsus nicely stated: “All things are poison, and nothing is without poison; only the dose permits something not to be poisonous” Like other drugs, it is the dose of fluids that make them poisonous. As stated before, the risk of excessive fluid overload is well established.

Similar to other drugs, choosing the right dose implies that we take into account the pharmacokinetics and pharmacodynamics of intravenous fluids (Table 1). Pharmacokinetics describes how the body affects a drug resulting in a particular plasma and effect site concentration [42]. Pharmacokinetics of intravenous fluids depends on distribution volume, osmolality, tonicity, oncoticity and kidney function. Eventually, the half-time depends on the type of fluid, but also on the patient’s condition and the clinical context (Table 2). When administering 1 L of fluid only, 10% of glucose solution, versus 25–30% of an isotonic crystalloid solution, versus 100% of a colloid solution will remain intravascularly after 1 h, but as stated above the half-life is dependent on other conditions (like infection, inflammation, sedation, surgery, anaesthesia, blood pressure) (Fig. 3) [29, 43].

Table 2 Overview of half-life (T1/2) of Ringer’s, glucose and colloid solutions as reported in different studies.
Full size table

Volume kinetics is an adaptation of pharmacokinetic theory that makes it possible to analyse and simulate the distribution and elimination of infusion fluids [29]. Applying this concept, it is possible, by simulation, to determine the infusion rate that is required to reach a predetermined plasma volume expansion. Volume kinetics may also allow the quantification of changes in the distribution and elimination of fluids (and calculation of the half-life) that result from stress, hypovolemia, anaesthesia and surgery [43].

Pharmacodynamics relates the drug concentrations to its specific effect. For fluids, the Frank–Starling relationship between cardiac output and cardiac preload is the equivalent of the dose effect curve for standard medications. Because of the shape of the Frank–Starling relationship, the response of cardiac output to the fluid-induced increase in cardiac preload is not constant [44]. The effective dose 50 (ED50), in pharmacology, is the dose or amount of drug that produces a therapeutic response or desired effect in 50% of the subjects receiving it, whereas lethal dose 50 (LD50) will result in death of 50% of recipients. Translated to IV fluids, this would be the dose of fluid that induces, respectively, a therapeutic response or death in 50% of the patients. The problem is that the therapeutic response varies from one patient to another. Fluid administration can be toxic (or even lethal) at a high enough dose, as demonstrated in 2007 when a California woman died of water intoxication (and hyponatremia) in a contest organized by a radio station (http://articles.latimes.com/2007/jan/14/local/me-water14). The difference between toxicity and efficacy is dependent upon the particular patient and the specific condition of that patient, although the amount of fluids administered by a physician should fall into the predetermined therapeutic window. Unanswered questions remain: what is an effective dose of IV fluids? What is the exact desired therapeutic effect? What is the therapeutic window? In some patients, volume expansion increases the mean systemic filling pressure (the backward pressure of venous return), but it increases the right atrial pressure (the forward pressure of venous return) to the same extent, such that venous return and, hence, cardiac output do not increase [45]. Hence, venous congestion and backward failure may even play a more important and currently underestimated role [46]. The probability of the heart to “respond” to fluid by a significant increase in cardiac preload varies along the shock time course, and thus, pharmacodynamics of fluids must be regularly evaluated. At the very early phase, fluid responsiveness is constant. After the very initial fluid administration, only one half of patients with circulatory failure respond to an increase in cardiac output [47].

De-escalation

As we will discuss below, the final step in fluid therapy is to consider withholding or withdrawing resuscitation fluids when they are no longer required [1, 14, 15].

Like for antibiotics (Table 1), the duration of fluid therapy must be as short as possible, and the volume must be tapered when shock is resolved. However, many clinicians use certain triggers to start, but are unaware of triggers to stop fluid resuscitation, increasing the potential for fluid overload. As with duration of antibiotics, although there is no strong evidence, there is a trend towards shorter duration of intravenous fluids [39].

The four phases of fluid therapy

Not only are the characteristics of fluids important, but also the strategy for their administration. This strategy fundamentally changes along with the time course of septic shock. Recently a three-hit, or even four-hit model of septic shock was suggested trying to answer four basic questions, in which we can recognize four distinct dynamic phases of fluid therapy [40]: resuscitation, optimization, stabilization and evacuation (de-resuscitation) (the acronym ROSE) (Table 3, Fig. 5). The four questions that will be discussed in the next section are “When to start intravenous fluids?”, “When to stop intravenous fluids?”, “When to start de-resuscitation or active fluid removal?” and finally “When to stop de-resuscitation?”

Table 3 The ROSE concept avoiding fluid overload.
Full size table
Fig. 5
figure 5

The different fluid phases during shock. Adapted from Malbrain et al. with permission [1]. a Graph showing the four-hit model of shock with ebb and flow phases and evolution of patients’ cumulative fluid volume status over time during the five distinct phases of resuscitation: resuscitation (1), optimization (2), stabilization (3) and evacuation (4) (ROSE), followed by a possible risk of Hypoperfusion (5) in case of too aggressive de-resuscitation. See text for explanation. b Graph illustrating the four-hit model of shock corresponding to the impact on end-organ function in relation to the fluid status. On admission patients are hypovolemic (1), followed by normovolemia (2) after fluid resuscitation, and fluid overload (3), again followed by a phase going to normovolemia with de-resuscitation (4) and hypovolemia with risk of hypoperfusion (5). In case of hypovolemia (phases 1 and 5), O2 cannot get into the tissues because of convective problems, in case of hypervolemia (phase 3) O2 cannot get into the tissue because of diffusion problems related to interstitial and pulmonary oedema, gut oedema (ileus and abdominal hypertension). See text for explanation

Full size image

First phase: Resuscitation

After the first hit which can be sepsis, but also burns, pancreatitis or trauma, the patient will enter the “ebb” phase of shock. This life-threatening phase of severe circulatory shock can occur within minutes and is characterized by a strong vasodilation leading to a low mean arterial pressure and microcirculatory impairment (Table 3). It may be accompanied by high (hyperdynamic circulatory shock as seen in sepsis, burns, severe acute pancreatitis, liver cirrhosis, thiamine deficiency, etc.) or low cardiac output (e.g. septic shock with severe hypovolemia or septic shock with sepsis-induced cardiomyopathy).

At this initial phase, usually during the first 3–6 h after the initiation of therapy, fluid resuscitation is commonly administered according to an early, adequate, goal-directed, fluid management strategy. The modalities of fluid administration at this early phase have been a matter of great debate. In the study by Rivers et al. [49], a protocol-based fluid management called early goal-directed therapy (EGDT) was associated with a significant reduction in mortality compared to standard care. Since this publication, similar outcome benefits have been reported in over 70 observational and randomized controlled studies comprising over 70,000 patients [50]. As a result, EGDT was incorporated as a “resuscitation bundle” into the first 6 h of sepsis management adopted by the Surviving Sepsis Campaign. As such, it has been disseminated internationally as the standard of care for early sepsis management. Recently, a trio of trials (ProCESS [51], ARISE [52] and ProMISe [53]), while reporting an all-time low sepsis mortality, showed no improvement in outcomes with EGDT, questioning the need and pointing towards the potential dangers of protocolized care for patients with severe and septic shock [54, 55]. A recent study employing a combined Bayesian and frequentist methodological approach to evaluate 12 randomized trials and 31 observational studies found that EGDT was potentially harmful in the patients with the highest disease severity [56]. In addition, although conducted in sub-Saharan Africa, three recent trials have demonstrated worse outcomes when administering fluid boluses for resuscitation in patients with septic shock [57,58,59]. What remains from the EGDT debate is that the rapidity of fluid administration and of the achievement of hemodynamic goals for initial resuscitation is important, even though this aspect has also recently been called into question [60].

In fact, rather than infusing a predefined given amount of fluid, the goal should be individualized for every patient, based on the evaluation of the need for fluids and on the patient’s premorbid conditions [16, 55, 61,62,63,64]. In this phase, on an individual basis for each patient, we try to find an answer to the first question: “When to start fluid therapy?”

At the very initial phase of septic shock, answering the question is easy: fluid administration will significantly increase cardiac output in almost all cases. Nevertheless, after the first boluses of fluid, the likelihood of preload unresponsiveness is high. Therefore, at this stage, fluid administration should be conditioned to the positivity of indices and tests predicting fluid responsiveness. However, it must be noted that the state of responsiveness can only be determined a posteriori (after the intervention with administration of fluid bolus) and when a hemodynamic monitoring device is in place to estimate or calculate cardiac output. Therefore, we advocate the use of specific tests to increase the a priori probability and likelihood for a favourable event/outcome, as fluid administration should be limited to responders.

Second phase: Optimization

The second hit occurs within hours and refers to ischaemia and reperfusion (Table 3). At this phase, fluid accumulation reflects the severity of illness and might be considered a “biomarker” for it [70]. The greater the fluid requirement, the sicker the patient and the more likely organ failure (e.g. acute kidney injury) may occur [71, 72].

In this phase, we must try to find an answer to the second question: “When to stop fluid therapy?” avoiding fluid overload. Indices of fluid responsiveness are again of utmost importance, since fluid administration should be stopped when these indices become negative [73]. Second, the clinical context must be taken into account. Obviously, more fluid is needed in septic shock from peritonitis than from pneumonia. Third, the decision to refrain from fluid administration should be based on indices that indicate the risk of excessive fluid administration. The presence of lung impairment is the condition that is most likely to be associated with the worst consequences of fluid overload. To estimate the pulmonary risk of further fluid infusion, one may consider the pulmonary artery occlusion pressure measured with the Swan–Ganz catheter. Nonetheless, this does not take into account the degree of lung permeability, which is a key factor in the mechanisms of pulmonary oedema formation [48]. Extravascular lung water measured by transpulmonary thermodilution, as well as the pulmonary vascular permeability index which is inferred from it, might reflect the pulmonary risk of fluid infusion more directly [40, 48, 74]. Intra-abdominal hypertension is also a potential consequence of too much fluid administration [40]. The intra-abdominal pressure should be cautiously monitored in patients at risk [75].

Third phase: Stabilization

With successful treatment, stabilization should follow the optimization phase (homoeostasis), evolving over the next few days (Table 3). It is distinguished from the prior two by the absence of shock or the imminent threat of shock. As previously described, the focus is now on organ support and this phase reflects the point at which a patient is in a stable steady state [1, 76] (Table 3).

Fluid therapy is now only needed for ongoing maintenance in the setting of normal fluid losses (i.e. renal, gastrointestinal, insensible) and replacement fluids if the patient is experiencing ongoing losses because of unresolved pathologic conditions [1, 76]. Since persistence of a positive daily fluid balance over time is strongly associated with a higher mortality rate in septic patients [11, 77], clinicians should also be aware of the hidden obligatory fluid intake, as it may contribute more than a litre daily [78].

Fourth phase: Evacuation

After the second hit, the patient may either further recover, entering the “flow” phase with spontaneous evacuation of the excess fluids that have been administrated previously, or, as is the case in many critically ill patients, the patient remains in a “no-flow” state followed by a third hit, usually resulting from global increased permeability syndrome with ongoing fluid accumulation due to capillary leak [17, 79]. In any case, the patient enters a phase of “de-resuscitation” (Table 3). This term was first suggested in 2012 [41] and finally coined in 2014 [1]. It specifically refers to late goal-directed fluid removal and late conservative fluid management.

Late goal-directed fluid removal involves aggressive and active fluid removal using diuretics and renal replacement therapy with net ultrafiltration. It is characterized by the discontinuation of invasive therapies and a transition to a negative fluid balance [40]. Late conservative fluid management describes a moderate fluid management strategy following the initial treatment in order to avoid (or reverse) fluid overload. Recent studies showed that two consecutive days of negative fluid balance within the first week of the intensive care unit stay is a strong and independent predictor of survival [1].

In this de-resuscitation phase, we try to find an answer to the third and fourth question: “When to start fluid removal?” and “When to stop fluid removal?” in order to find the balance between the benefits (reduction in second and third space fluid accumulation and tissue oedema) and risk (hypoperfusion) of fluid removal. To answer these questions, testing preload responsiveness may still be useful. Indeed, if no preload responsiveness is detected, it is reasonable to assume that fluid removal will not induce a reduction in cardiac output [80]. On the opposite, positive indices of preload responsiveness might indicate the limit of fluid removal and could even be a target to reach when removing fluids.

Obviously, the risk at this phase is to be too aggressive with fluid removal and to induce hypovolemia, which may trigger a “fourth hit” for hemodynamic deterioration and hypoperfusion (Fig. 5). If fluid is needed at this phase, the use of albumin seems to have positive effects on vessel wall integrity facilitates achieving a negative fluid balance in hypoalbuminemia and may be less likely to cause nephrotoxicity [81].

This four-phase approach should be better characterized by some epidemiological studies. Its prognostic impact might be significant, because it may lead to a reduction in the cumulative fluid balance, which by itself is clearly associated with poor prognosis (Fig. 4). Similar principles have also been suggested by others, confirming the need for a multicenter prospective clinical trial with a biphasic fluid therapy approach, starting with initial early adequate goal-directed treatment followed by late conservative fluid management in those patients not transgressing spontaneously from the ebb to the flow phase [14, 15, 70, 76, 82,83,84,85,86]. The RADAR (Role of Active De-resuscitation After Resuscitation) trial may help to find such answers (http://www.hra.nhs.uk/news/research-summaries/radar-icu/).

Conclusions

There are only four major indications for fluid administration in the critically ill: resuscitation, maintenance, replacement and nutrition (enteral or parenteral). In this review, a conceptual framework is presented looking at fluids as drugs by taking into account the four D’s (drug selection, dose, duration and de-escalation) and the four phases of fluid therapy within the ROSE concept (resuscitation, optimization, stabilization and evacuation). The four hits model is presented herein. This will provide answers to the four basic questions surrounding fluid therapy: (1) When to start IV fluids? (2) When to stop fluid administration? (3) When to start fluid removal and finally (4) When to stop fluid removal? In analogy to the way we deal with antibiotics in critically ill patients, it is time for fluid stewardship.