Clinical review: Volume of fluid resuscitation and the incidence of acute kidney injury - a systematic review

Intravenous fluids are widely administered to maintain renal perfusion and prevent acute kidney injury (AKI). However, fluid overload is of concern during AKI. Using the Pubmed database (up to October 2011) we identified all randomised controlled studies of goal-directed therapy (GDT)-based fluid resuscitation (FR) reporting renal outcomes and documenting fluid given during perioperative care. In 24 perioperative studies, GDT was associated with decreased risk of postoperative AKI (odds ratio (OR) = 0.59, 95% confidence interval (CI) = 0.39 to 0.89) but additional fluid given was limited (median: 555 ml). Moreover, the decrease in AKI was greatest (OR = 0.47, 95% CI = 0.29 to 0.76) in the 10 studies where FR was the same between GDT and control groups. Inotropic drug use in GDT patients was associated with decreased AKI (OR = 0.52, 95% CI = 0.34 to 0.80, P = 0.003), whereas studies not involving inotropic drugs found no effect (OR = 0.75, 95% CI = 0.37 to 1.53, P = 0.43). The greatest protection from AKI occurred in patients with no difference in total fluid delivery and use of inotropes (OR = 0.46, 95% CI = 0.27 to 0.76, P = 0.0036). GDT-based FR may decrease AKI in surgical patients; however, this effect requires little overall FR and appears most effective when supported by inotropic drugs.

decreased surgical complications [3,8,9] and reduced risk of postoperative AKI [10]. GDT strategies involve the use of intensive monitoring and specifi c haemodynamic targets to guide early aggressive fl uid and/or vasoactive drug administration. While individually persuasive, collect ively these two lines of evidence appear diffi cult to reconcile. On the one hand, fl uid overload appears harmful overall, and to the kidney in particular; on the other hand, goal-directed FR reduces mortality, morbidity and renal dysfunction. Th e evidence base for both of these conclusions is strongest in surgery, and consequently perioperative resuscitation was chosen as the setting to best explore this issue.
Th is seeming confl ict in evidence may be the cause of controversy surrounding the extent of FR in patients at risk of AKI. Yet, without knowledge of the relative timing of onset, duration and quantity of fl uid given during GDT-FR compared with the current standard of care, it is impossible to gauge whether the apparent benefi ts of more conservative FR and those of goal-directed haemodynamic therapy are truly in opposition. Accordingly, we conducted a systematic literature review and metaanalysis of controlled trials of goal-directed resuscitation protocols in the settings of surgery. Our focus was on AKI and the quantity of fl uid administered. We hypothesised that the benefi cial eff ects of GDT on subsequent renal function are related to the amount of additional fl uid given during GDT compared with control patients, treated according to the standard of care.

Aims
We aimed to identify and review studies describ ing haemo dynamically targeted FR during GDT for perioperative optimisation. In particular, we aimed to determine whether such protocols have a benefi cial eff ect on renal function and to what extent this treatment involves increased fl uid administration in comparison with control groups treated according to the standard of care.

Electronic search strategy
We interrogated the MEDLINE electronic reference database using a combination of search terms ( Figure 1). In addition, bibliographies of known systematic reviews and retrieved articles were examined for references of potential relevance. Our wide search terms defi ned a set of records containing randomised controlled trials (RCTs) of GDT in surgery; these abstracts were then manually examined to identify a subset of studies of potential relevance for manuscript retrieval. Th ese publications were examined manually for accordance with our inclusion criteria. Additional checks were made to ensure none of these publications had been subject to retraction [11].

Study inclusion criteria Population
We considered only patients undergoing major surgery. Studies involving mixed populations of critically ill patients were excluded.

Intervention
GDT was defi ned as monitoring of physiological measure ments related to adequacy of cardiac output, systemic oxygen delivery or ventricular preload and providing specifi c therapy to target normal or supranormal values by FR alone or in combination with inotropic drugs. GDT was required to commence before or within 8 hours from commencement of surgery.

Between-group comparisons
We considered studies that reported the eff ects of GDT versus standard haemodynamic therapy, with special emphasis on the comparisons of cumulative volumes of FR. We excluded RCTs with no diff erence in, or no description of, fl uid management strategies between groups, RCTs comparing diff ering GDT protocols targeting the same physiological goal, or studies without predefi ned end points. A specifi c requirement was that studies needed to report a volumetric quantifi cation of FR. We did not consider studies that examined the eff ect of diff ering maintenance fl uid infusions in our main analysis.

Outcome
Studies were required to report a renal outcome measure that defi ned AKI -this could be an absolute or relative change in serum creatinine or urea, a defi ned period of oliguria, a need for renal replacement therapy for AKI or some combination of these.

Data collection
We collected data on the study participants, the setting, the timing and duration of the intervention, the defi nition of AKI used, the method of physiological monitoring and the GDT protocol. We additionally examined indicators of study quality and reporting, including documentation of the method of randomisation, the extent of blinding and description of drop-outs based on the score proposed by Jadad and colleagues [12]. Finally, we documented the recorded volume of fl uid administered and whether there were statistical diff erences in these quantities.
To assess the volumes of fl uid administered in the treatment and control groups we calculated the diff erence in the mean or median total fl uid volume delivered between groups. When a total was not reported, we summed the mean or median volumes of crystalloids, colloid and blood products given during the protocol period in the GDT and control groups to estimate the total fl uid volume delivered in each group. We defi ned studies employing a larger volume of FR during GDT as those demonstrating a statistically signifi cant diff erence between one or more types of fl uid administered between groups during GDT. In addition, we excluded from our defi nition any studies where the diff erence in mean or median fl uid adminis tration during the GDT protocol period was <500 ml, an arbitrary defi nition of a clinically insignifi cant diff er ence in fl uid administration between GDT and standard of care.

Statistical analysis
Meta-analysis was performed using R: A language and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria [13]), utilising the packages meta, metafor and rmeta. As studies compared diff ering interventions in diff ering settings with diff ering measures of renal function, a random eff ects model was employed for all analyses. Th e eff ect of GDT in the metaanalysis of pooled studies was assessed by calculation of the odds ratio (OR) for AKI: OR <1 favoured GDT over the control group. Th e 95% confi dence intervals (CIs) of the ORs and two-sided P values were calculated. A statistical diff erence favouring or disfavouring GDT was considered to occur if the 95% CI failed to cross the OR = 1 line (equivalent to P <0.05).
Statistical heterogeneity [14] was assessed by calcu lation of the I 2 statistic and the Q-test [15,16]. Heterogeneity or inconsistency was considered signifi cant if I 2 was >40% or if the P value for the Q-test was <0.1 [17]. When heterogeneity was present we assessed for outlier studies using a Galbraith plot [18] and repeated the analysis excluding the most signifi cant outlier [19]. Our metaanalysis was then repeated with subdivision by diff erence in volume of FR between the GDT and control groups and by use or nonuse of inotropes in GDT protocols.

Results
Our search strategy on GDT in surgery identifi ed 3,797 abstracts or articles from which we examined 53 potentially relevant manuscripts, fi nding 24 articles that met our inclusion criteria   (Figure 2). All studies were in adult patients. Among the surgical studies not meeting the inclu sion criteria, three recent studies examined a more restric tive perioperative maintenance fl uid infusion com bined with the use of GDT in one or both groups to facilitate a conservative maintenance fl uid regimen [44][45][46]. While these studies were not directly comparable with standard GDT trials, they are of relevance to a discussion of the inter action between GDT and fl uid balance and we elected to examine these investigations in a separate analysis.
Th e study details, population, size and quality indicators are presented in Table 1; 15 of 24 studies had a Jadad score of 3. Blinding is inherently diffi cult when using a treatment intervention triggered by invasive monitoring, and most studies were unblinded or partially blinded; while the randomisation method and dropouts were generally well recorded. As most studies were of similar quality, we did not subdivide our analysis based on quality.
Analysis of details of GDT interventions and renal outcome measures are presented in Table 2 and fl uid adminis tration during GDT is presented in Table 3; further details of individual fl uids administered in these studies are provided in Additional fi le 1. Diagnostic criteria for AKI varied widely from a defi ned increment in serum creatinine from baseline or oliguria over a specifi ed duration to a need for renal replacement therapy ( Table 2); for two studies [37,38] renal outcome data not in the original trial publication were reported in a previous meta-analysis [10]. Given the large variation in defi nition of AKI for studies of GDT in surgery, we performed a further sensitivity analysis considering only studies where AKI was defi ned as an elevation of creatinine from baseline of ≤100% or an absolute rise in serum creatinine of ≤200 μmol/l, representing studies using creatinine criteria for AKI thus including patients with less severe postoperative renal dysfunction (approximat ing RIFLE risk or injury AKI).

Goal-directed therapy in surgery
Of the 24 included studies, 16 reported a statistically signifi cant greater quantity of colloid, crystalloid or total fl uid administered during GDT. In two studies the diff erence in fl uid administration was statistically signifi cant but the total diff erence in FR between the GDT and control groups during the study period was small (+241 ml and +450 ml), so these studies were reallocated to the equivalent volume of FR with GDT group [32,39]. Fourteen studies were thus deemed to involve a greater volume of FR during GDT, their median diff erence in FR between the GDT and control groups being +1,247 ml (range 500 to 11,099 ml). In the 10 studies categorised as having no greater volume of total FR during GDT, the median diff erence in FR was 10 ml (range -600 to +529 ml). Among all 24 studies, the median diff erence in FR was +555 ml.
Overall, 13 studies employed inotropic drugs in some or all patients in the GDT group, 10 studies did not use inotropic drugs, and one study had two treatment armsone arm involving infusion of dopexamine in addition to stroke-volume-directed fl uid boluses, and the other arm involving stroke-volume-guided therapy alone [31] ( Table 2). In the meta-analysis this study was included as a single study in the overall analysis and in the analysis based on fl uid therapy, but was treated as two studies when dividing studies on the basis of inotropic use ( Figure 3 and Table 4).
Overall, GDT in surgery was associated with a significantly lower incidence of AKI (OR = 0.59, 95% CI = 0.39 to 0.89, P = 0.013; 24 studies, n = 2,763 patients) ( Figure 3 and Table 4). When subdividing by larger versus equivalent fl uid administered during GDT, studies involving equivalent overall fl uid therapy were associated with a lower incidence of AKI (OR = 0.47, 95% CI = 0.29 to 0.76,    (Figure 3 and Table 4) whereas studies categorised as involving larger fl uid administration during GDT were not associated with lower incidence (OR = 0.70, 95% CI = 0.35 to 1.41, P = 0.32; 13 studies, n = 1,462 patients); one study was excluded for heterogeneity [34].

Goal-directed therapy and inotropic drug use in surgery
Inotropic drug use in GDT patients was associated with a reduction in the incidence of postoperative AKI (OR = 0.52, 95% CI = 0.34 to 0.80, P = 0.003; 14 studies, n = 1,634 patients) ( Figure 3 and Table 4) compared with studies not involving inotropic drugs (OR = 0.75, 95% CI = 0.37 to 1.53, P = 0.43; 10 studies, n = 1,016 patients); one study was excluded for heterogeneity [34] (Figure 3 and Table 4). Th e lowest incidence of postoperative AKI occurred in patients with equivalent total fl uid delivery in a protocol involving inotropic drugs (OR = 0.46, 95% CI = 0.27 to 0.76, P = 0.0036; eight studies, n = 1,033 patients) (Figure 3 and Table 4); however, only a small number of studies (three studies, 280 patients) without use of inotropic drugs had equivalent total fl uid delivery in the treatment group, so the eff ect in this group was diffi cult to assess.

Goal-directed therapy involving restrictive fl uid strategies in surgery
Th ree studies involving GDT and more restrictive fl uid management were identifi ed ( Table 1). Two of these studies involved GDT in both groups with diff ering maintenance fl uid rates during surgery, while one study Approximate mean or median fl uid volume per patient was approximated from data provided in individual studies (see Addition File 1 for further information). Any statistically signifi cantly greater administration of a component of fl uid resuscitation in goal-directed therapy (GDT) over the control group is highlighted in bold. a Extra volume above maintenance fl uids 500 ml/m 2 /day documented. b Statistically signifi cant greater fl uid administration in GDT deemed clinically insignifi cant (diff erence <500 ml). c Cumulative fl uid balance in study through to end of surgery.
involved GDT only in the more restrictive fl uid infusion group (Table 2). Th e median diff erence in fl uid delivered was 2,208 ml lower (range -1,733 to -2,622 ml) in the treatment group (Table 3). In these three studies there was no clear statistical association between the volume of fl uid therapy delivered and AKI (OR for AKI in treatment group = 1.68, 95% CI = 0.69 to 4.12, P = 0.26) (Figure 4 and Table 4).

Key fi ndings
In this review we have examined the evidence for the use of FR in improving renal outcomes during major surgery. We found that, on average, protocolised GDT did not result in signifi cantly larger volumes of fl uid therapy being administered when compared with standard management. Furthermore, most of the protection from AKI seemed to occur in studies reporting equivalent total quantity of fl uid administration between groups. From this evidence, it is diffi cult to ascribe benefi ts of GDT simply to the provision of greater volumes of intravenous fl uids to hypovolaemic patients, suggesting any eff ect actually derived from other aspects of care. Our fi ndings were similar when only studies using more sensitive creatinine-based defi nitions of AKI were examined in a sensitivity analysis. Use of inotropes seemed to be associated with lower incidence of AKI with GDT and with the delivery of equivalent amounts of FR.

Relationship to previous studies
Conventionally, a reduction in RBF and renal ischaemia have been regarded central to the pathogenesis of AKI in critical illness [47]; however, the situation is likely to be more complex [48]. In animal models, AKI has been reported when cardiac output was preserved or elevated [49], and subtotal (90%) occlusion of the renal arteries did not lead to prolonged AKI [50]. Furthermore, cardiac arrest and warm renal ischaemia in humans only lead to signifi cant AKI in the presence of severe post-resusci tation disease and cardiogenic shock [51]. Th ese obser vations infer that local and systemic infl ammatory res ponses, Statistical analysis -random eff ects model. AKI, acute kidney injury; CI, confi dence interval. a One study presented data on goal-directed therapy (GDT) with and without inotropic therapy [31]. Data from these groups were considered as separate studies, dividing on the basis of use/nonuse of inotropes in GDT. changes in intra-RBF distribution, microcircu latory dysfunction and glomerular haemodynamics may all account for loss of kidney function, even in the presence of maintained or increased RBF.
Fluid administration has been shown to augment both cardiac output and RBF without an overall increase in renal oxygen delivery [52], while FR in an experimental model of haemorrhagic shock can restore blood pressure and cardiac output without recovery of renal tissue oxygen tension [53]. Finally, even when renal oxygen delivery is reduced, increased oxygen extraction may main tain renal oxygen consumption [54] -suggesting that increasing renal oxygen delivery by increasing cardiac output may not avert AKI.
Measurements of cardiac output are rarely made until critical illness is established, RBF measurements are not clinically available, and fl uid therapy is routinely targeted to the treatment of arterial hypotension. Absolute or relative hypotension is considered a signifi cant risk factor for the development of AKI [55,56], and lower blood pressure has been associated with the progression of AKI in sepsis [57]. However, hypotension may be related to low cardiac output, systemic vasodilation or a combination of both; and in the setting of systemic hypotension, RBF may be reduced, elevated or unchanged. As hypotension will only respond to FR if a meaningful increase in cardiac output can be obtained, this response may only be partial or transient in the context of systemic vasodilatation. Th e RBF eff ect of FR for hypotension might thus be signifi cant or negligible depending on the context. Bedside physiological measurements such as blood pressure and urine output will be even more indirectly related to the pathogenesis of AKI, although they are routine triggers for FR.

Signifi cance of study fi ndings
Any potential benefi ts of FR must be weighed against the consequences of fl uid accumulation. In acute illness, exogenous fl uid replacement almost inevitably leads to positive fl uid balances. Additionally, the critically ill are particularly susceptible to fl uid overload [5]. Physiologically, fl uid overload results in tissue oedema that may contribute to progressive organ dysfunction. Liberal perioperative fl uid strategies, involving large positive fl uid balances, have been associated with increased complications after surgery [3,4,58]. Importantly, as an encapsulated organ the kidney is itself particularly aff ected by fl uid congestion, because raised venous and intra-capsular

Favours Fluid Restriction
Favours Standard Care pressures cause signifi cant decreases in RBF and the glomerular fi ltration rate [59]. Studies examining liberal versus more conservative fl uid management during surgery have failed to demonstrate an increase in AKI in patients receiv ing less fl uid. Instead, recent analysis of results from the Fluid and Catheter Treatment Trial has shown that pursuing an even fl uid balance strategy in patients with acute lung injury was associated with a lower incidence of AKI after allowing for the dilutional eff ect of positive fl uid balances on serum creatinine concentration [60]. GDT may thus be benefi cial not only by providing extra fl uid when it is specifi cally indicated, but also by allowing earlier and guided use of fl uids while preventing the delivery of unnecessary fl uid when specifi c and objective haemodynamic targets are met. Seen in this light, appropriate use of GDT is in harmony, rather than in confl ict, with more restrictive approaches to peri operative fl uid therapy.
We identifi ed three studies that fell outside the criteria for inclusion in our main analysis where GDT was used in conjunction with fl uid restriction [44][45][46]. Th ese studies moved beyond fl uid optimisation using GDT to actively minimise perioperative maintenance fl uid adminis tration, resulting in signifi cantly smaller fl uid administration than standard care (Table 3). No clear association was found between postoperative AKI and more restrictive fl uid balances. A possible trend toward increased incidence of AKI with fl uid restriction was seen, but there was signifi cant heterogeneity between a small number of studies. In addition, relative haemodilution in control groups might be expected to artefactually lower the incidence of AKI, which was defi ned by small rises in serum creatinine. Given that avoidance of fl uid overload may improve wider postoperative outcomes, and that GDT appears to protect against AKI when compared with conventional fl uid management, the use of GDT strategies to facilitate more restrictive perioperative fl uid therapy while avoiding covert hypovolaemia merits further examination in larger trials.
Low-dose inotropes have been associated with better postoperative outcomes when used in perioperative GDT [61]. In our meta-analysis, use of inotropes in GDT was associated with a decreased risk of postoperative AKI -a fi nding described previously [10]. Without specifi c trials, however, it is diffi cult to dissect the relative contributions of inotropy and fl uid management on renal outcomes. We found that trials using inotropes in GDT and similar volumes of fl uid therapy to controls were associated with the best renal outcomes, while excess fl uid and no inotropes were associated with the worst renal outcomes. It is diffi cult to assess the relative contributions of FR and inotropes, but a synergistic eff ect is likely; so that goaldirected FR may allow optimal use of inotropic drugs, while ino tropic drugs may maximise the haemodynamic eff ect of fl uids and reduce the risk of fl uid overload. Notably, an important physio logical consequence of increased cardiac contract ility is a decrease in right atrial pressure, potentially a benefi cial eff ect limiting the increase in venous pressure with FR.
Are these fi ndings generalisable beyond the context of major surgery? Sepsis is the leading cause of AKI in the ICU and shares many features with systemic infl ammatory response after major surgery. In septic shock, even though patients tend to have high cardiac outputs, insensible or distributive fl uid losses are common and FR may be required acutely. However, increased capillary permeability makes septic patients particularly prone to generalised tissue oedema and the adverse eff ects of fl uid overload. Historically, approaches targeting supranormal levels of cardiac output and oxygen delivery in patients with established critical illness (including many patients with severe sepsis) have been nonbenefi cial [62] or harmful [63]. More recently, treatment protocols immediately after presentation have demonstrated that targeting central venous oxygen saturation may improve survival in septic shock [64]. However, the specifi c eff ect of goal-directed resuscitation on renal function was not explicitly examined.
Applying a similar search strategy to that employed in the present systematic review to locate studies examining GDT for the acute treatment of septic shock, we only identifi ed two articles documenting a renal outcome [65,66] -only one of which was in adult patients, precluding a useful meta-analysis in this population. Both of these studies showed a benefi cial eff ect of GDT on renal as well as other out comes. Signifi cantly, in both studies GDT was associated with earlier FR in at a higher fl ow rate than standard care, but the overall quantity of FR did not diff er between GDT and standard care -suggesting that, in septic shock, early, targeted, therapy may restore haemodynamics and spare later fl uid require ments. Th e same eff ect -early GDT actually being associated with similar fl uid balances to controls over a longer period of treatment -was also observed by Rivers and colleagues in their landmark, although controversial, single-centre study of early GDT in septic shock [64]. By contrast, in the recent FEAST study of fl uid management in children presenting with severe sepsis in Kenya, Tanzania and Uganda, rapid fl uid bolus volume expan sion was associated with signifi cant harm [67]; importantly, however, fl uid therapy in the FEAST study was not guided by any form of invasive haemodynamic monitoring. Th erefore, in the setting of septic shock as well as surgery, goal-directed haemodynamic resuscitation may be benefi cial to the kidney and other organs by targeting FR only to those patients that require it, and avoiding FR in those that do not. At present, however, evidence to guide choices in FR for septic shock remains very limited.

Strengths and limitations of review fi ndings
Th e meta-analyses performed in this review prov ide a basis for discussion of the benefi ts of FR on renal function in the critically ill. Our search was confi ned to the Pubmed electronic database and it is possible that relevant studies were missed; however, similar metaanalyses with a wider database search identifi ed a similar set of publications [8,10], while inclusion in Pubmed publications should ensure original studies were subject to independent peer review. Studies varied widely in populations, outcome measures and methodology, and conclusions from these data can only be regarded as hypothesis forming -although our results were essentially unchanged when considering only studies with more similar defi nitions of AKI. GDT protocols are complex and it is diffi cult to attribute benefi t to specifi c components of each strategy. We can hypothesise that GDT results in timelier FR for patients who require it, even if the total volume of FR within a trial does not diff er from the standard of care. While this is probably true considering the nature of GDT, data provided in the surgical trials we considered generally did not include timing of fl uids within the study period, so we had no capacity to examine this important question in our analysis.
We did not assess the formulation of intravenous fl uids in these studies. Th is is a complex issue of considerable clinical signifi cance. However, none of the studies we examined set out to specifi cally compare diff ering formulations of intravenous fl uid, and a variety of intravenous fl uids were used in the treatment and control groups in these studies. Individual patient data would be required to explore the eff ect of fl uid formulation on renal outcomes in these studies. It is of course possible that higher volume fl uid administration is associated with a larger volume of colloid exposure, potentially associated with harm to the kidney countering the benefi ts of GDT, but assessing the size of any eff ect will be very diffi cult. Major multicentre randomised trials comparing the eff ect of colloid versus crystalloids for FR in intensive care are nearing publication [68,69] and high-quality data regarding the formulation of FR and outcomes should shortly be available. When these data are available we will be in a much better position to assess the interaction between the volume of FR, the composition of intravenous fl uids and AKI.
Our assessments of overall fl uid volumes administered between groups in these studies are based on the mean or median values provided in the articles; they are therefore imprecise and do not provide information on the fl uid balance, timing of fl uid administration or rate of fl uid administration. However, even though the exact fl uid quantities and their clinical signifi cance could be argued, we feel the general conclusion that perioperative GDT does not, overall, result in much additional intravenous fl uid being administered is supported by the studies identifi ed. We considered fl uid administration rather than fl uid balance, because fl uid balances during GDT were only reported in a small number of studies. Th ree large multi centre RCTs examining the eff ect of GDT in sepsis (ProCESS [70], ARISE [71] and ProMISe [72]) and another RCT examining perioperative GDT in high-risk surgery (OPTIMISE [73]) are ongoing. Our level of evidence to guide fl uid therapy in complex, acute unwell patients should therefore increase in the next 3 to 4 years. In the interim, current evidence suggests that fl uid management that rapidly recognises and treats hypovolaemia while avoiding inappropriate fl uid loading may be optimal for renal function and wider clinical outcomes.

Conclusions
Goal-directed haemodynamic therapies involving guided administration of intravenous fl uids are associated with decreased incidence of AKI in the setting of major surgery. Evidence in sepsis is presently inconclusive. Crucially, the benefi t of these targeted haemodynamic interventions does not seem to be dependent on the amount of fl uid administered. On the contrary, GDT seems to achieve better kidney protection when delivering equivalent amounts of FR, perhaps by providing fl uid earlier, faster and with the support of inotropic drugs, thus minimizing the harmful eff ects of fl uid overload.