What is the impact of the fluid challenge technique on diagnosis of fluid responsiveness? A systematic review and meta-analysis

Background The fluid challenge is considered the gold standard for diagnosis of fluid responsiveness. The objective of this study was to describe the fluid challenge techniques reported in fluid responsiveness studies and to assess the difference in the proportion of ‘responders,’ (PR) depending on the type of fluid, volume, duration of infusion and timing of assessment. Methods Searches of MEDLINE and Embase were performed for studies using the fluid challenge as a test of cardiac preload with a description of the technique, a reported definition of fluid responsiveness and PR. The primary outcome was the mean PR, depending on volume of fluid, type of fluids, rate of infusion and time of assessment. Results A total of 85 studies (3601 patients) were included in the analysis. The PR were 54.4% (95% CI 46.9–62.7) where <500 ml was administered, 57.2% (95% CI 52.9–61.0) where 500 ml was administered and 60.5% (95% CI 35.9–79.2) where >500 ml was administered (p = 0.71). The PR was not affected by type of fluid. The PR was similar among patients administered a fluid challenge for <15 minutes (59.2%, 95% CI 54.2–64.1) and for 15–30 minutes (57.7%, 95% CI 52.4–62.4, p = 1). Where the infusion time was ≥30 minutes, there was a lower PR of 49.9% (95% CI 45.6–54, p = 0.04). Response was assessed at the end of fluid challenge, between 1 and 10 minutes, and >10 minutes after the fluid challenge. The proportions of responders were 53.9%, 57.7% and 52.3%, respectively (p = 0.47). Conclusions The PR decreases with a long infusion time. A standard technique for fluid challenge is desirable. Electronic supplementary material The online version of this article (doi:10.1186/s13054-017-1796-9) contains supplementary material, which is available to authorized users.


Background
Intravenous fluid is one of the most commonly administered therapies for critically ill patients and is the cornerstone of haemodynamic management of patients in intensive care units (ICUs) [1]. The rationale for volume expansion is to increase the cardiac output (CO) and oxygen delivery to ultimately improve tissue oxygenation. The gold standard for assessing fluid responsiveness to guide fluid administration in critically ill patients is to perform a fluid challenge. This involves the infusion of a specific amount of intravenous fluid to assess ventricular preload reserve and subsequent systemic haemodynamic effects [2]. The volume of fluid infused must be sufficient to increase right ventricular diastolic volume and subsequently stroke volume (SV) as described by the Frank-Starling law [3]. Fluid responsiveness is conventionally defined as an increase of at least 10% to 15% in SV in response to a fluid challenge, which is a reflection of the limits of precision of the technology used [4,5]. Patients who reach this threshold are considered 'fluid responders'. Clinical studies have demonstrated that approximately 50% of critically ill patients who are deemed to have inadequate CO are fluid responders [6]. However, fluid responsiveness is neither a binary nor a static condition, because it depends on dynamic interaction between intravascular volume, vascular tone and ventricular function. Furthermore, fluid responsiveness may also depend on the particularities of the fluid challenge, including the type and volume of fluid as well as the administration rate.
Administration of a fluid challenge is not a standardised technique, with varying volumes, infusion rates, fluid types and durations of response. The use of different methods to estimate SV is a further confounder. Whilst different clinical conditions may require different fluid challenge techniques, there is heterogeneity in practice for the same clinical condition [6].
We hypothesise that the technique of fluid challenge affects fluid responsiveness. This may result in different clinical decisions. Either inadequate or excessive fluid administration has adverse clinical consequences, and a better understanding fluid administration is likely to improve patient management and outcome. The objective of this study was to describe the different fluid challenge techniques used in clinical trials by assessing fluid responsiveness and how the proportion of patients deemed 'fluid-responsive' varies according to the technique used.

Studies
This study was conducted following a pre-defined protocol (Additional file 1: Appendix 1). No ethical approval or patient consent was necessary for the present study. We included studies meeting the following inclusion criteria: use of a fluid challenge as a test of cardiac preload or as part of a clinical algorithm, studies performed in ICUs or operating theatres with adult patients, studies including a full description of the fluid challenge technique (volume, infusion rate, type of fluid used and timing of assessment of the haemodynamic response), studies which included a clear definition of fluid responsiveness, and studies where the numbers of responders and non-responders to the fluid challenge were stated. Only studies published as full-text articles, published in English and in an indexed journal were included. Reviews, case reports and studies published in abstract form were excluded. We excluded studies involving pregnant women and children, studies where more than one fluid challenge was performed in the same patient, studies involving passive leg raising without use of a fluid challenge technique, studies where more than one fluid type was used whilst reporting a single result, studies using a continuous infusion of fluid, and studies where the fluid responsiveness was assessed only after a period of 60 minutes or more following completion of fluid challenge. Studies reporting more than one type of fluid challenge with a full description of results for each type of fluid challenge used were included for analysis as two separate studies. Studies reporting more than one type of fluid challenge (i.e., colloids and crystalloids) without a full description of results for each type of fluid challenge were excluded from the relevant part of the analysis (i.e., type of fluid).

Search strategy and data extraction
Three of the authors (LT, DA and DB) conducted a computerised search of the MEDLINE and Embase databases in February 2016. The terms included for the research were used in the following Boolean operators: 'fluid challenge' OR 'fluid bolus' OR 'fluid therapy' OR 'fluid responsiveness' OR 'fluid resuscitation' AND 'intensive care' OR 'critical care' OR 'operative theatre' OR 'anaesthesia' AND 'stroke volume' OR 'cardiac output' OR 'cardiac index' OR 'stroke volume variation' OR 'pulse pressure variation' OR 'stroke pressure variation'. The search was filtered by language, the age of participants (adults) and the availability of full-text articles using the native filter function of each database used.
Titles and abstracts of the trials identified in the search were independently reviewed and pooled for further screening. The full text of each trial identified was analysed, and each reviewer compiled a list of studies that met the inclusion criteria. Each review author's list was compared, and any disagreement was resolved through discussions until a consensus was reached among all review authors.
The following data were extracted from each study: volume of fluid used in the fluid challenge, duration of the infusion, type of fluid used, definition of fluid responsiveness, methodology used for the fluid responsiveness assessment, characteristics of the patients enrolled in the study, clinical environment in which the study was performed, number of patients included in the study, and percentage of 'fluid responders'. Data were extracted independently by three authors (LT, DA and DB) and verified by another author (HDA).
The identification, screening and inclusion of studies in this review are summarised in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram in Fig. 1. A PRISMA checklist is also reported in Additional file 1.

Statistical analysis
Data were examined graphically and statistically (Shapiro-Wilk test) to understand the distribution and nature of each variable. Data are presented as mean and 95% CI when normally distributed or as median and IQR for nonparametric data. Not all the studies reported the data required for the analysis of all the outcomes. Whenever any data were missing, only those studies with the data reported for the relevant analysis were included. Not imputation technique was applied.
The primary outcome of the study was the difference in means of proportion of fluid responders (PR). The included studies were grouped into three categories on the basis of volume used for the fluid challenge: <500 ml, 500 ml and >500 ml. Studies were grouped into three categories for the duration of the fluid infusion: <15 minutes, between 15 and 30 minutes and ≥30 minutes. Cut-off values for the duration and volume of fluids infused were defined following review of the literature. The types of fluid used were grouped into two categories: colloids and crystalloids.
Two-way independent analysis of variance (ANOVA) was conducted to compare means and variances between groups using as second variable (the setting of the study: ICU vs theatre), given the potential different pathophysiology of these two groups and the potential impact on the PR. Bootstrapping was conducted using 1000 samples and bias-corrected and accelerated. When assumptions for two-way independent ANOVA were not met, one-way independent ANOVA results are reported. Post hoc test results are reported with Bonferroni correction for multiple comparisons. Statistical significance was considered at a p value <0.05. Statistical analysis was performed using IBM SPSS Statistics version 24 software (IBM, Armonk, NY, USA).

Results
A total of 363 titles were identified through PubMed, and 163 were identified through Embase. After removal of duplications, 404 titles were collected for the analysis (Fig. 1). Screening by title and abstract excluded 233 studies, and 171 studies were selected for full-text assessment. Three studies were identified by snowballing. Eighty-five studies were selected for the final analysis. Two different sets of data were extracted from three studies because two different fluid challenge techniques were reported with the respective proportions of responders and non-responders. In total, 88 sets of data extracted from 85 studies with an aggregated 3601 patients were analysed ( Table 1).
The definition of positive response to a fluid challenge varies substantially across studies (Additional file 1: Figure S1). Physiological parameters used to assess fluid response include cardiac index (47.5%), CO (17.1%), SV (11.0%) and stroke volume index (24.3%). The increment from baseline measurements in physiological parameters deemed to have a positive response to a fluid challenge was either 10% (25.5% of studies) or 15% (74.5% of studies). The most frequent definition of a positive response to a fluid challenge was an increase in cardiac index of at least 15% from baseline (n = 33 [40.2%]). CO was estimated using several different technologies (Additional file 1: Figure S2), with pulse index continuous CO (PiCCO; PULSION Medical Systems, Feldkirchen, Germany) used most frequently (31.7% of studies), followed by the pulmonary artery catheter (PAC; 22% of studies) ( Table 1). There was a higher percentage of responders in studies performed in the operating room (63.4%, 95% CI 58.3-68.4) than in the ICU (51.5%, 95% CI 48.2-54.8, p < 0.001).

Duration of infusion
The time of infusion was <15 minutes in 24 studies (27.3%), between 15 and 29 minutes in 26 studies (29.5%), and ≥30 minutes in 29 studies (33%). Nine studies (10.2%) did not report duration of infusion. Where the fluid challenge was administered for <15 minutes, between 15 and 29 minutes, and >30 minutes, the proportions of patients deemed to be fluid responders were 59.2% (95% CI 54.2-64.1), 57.7% (95% CI 52.4-62.4), and 49.9% (95% CI 45.6-54) respectively. The duration of the fluid infusion affects the proportion of fluid responders [F(2,73) = 3.63, p = 0.03] (Fig. 2). The PR to a fluid challenge given in ≥30 minutes was lower than the PR when the fluid challenge was given in <15 minutes (p = 0.045). The proportion of patients responding to a fluid challenge that was administered in <15 minutes and between 15 and 30 minutes was similar (p = 1.0).

Timing of assessment
The assessment of response to a fluid challenge was at the point of administration (n = 50 [58.1%]

Discussion
We demonstrate that the duration of the fluid infusion in a fluid challenge has a significant influence on fluid responsiveness. This confirms our hypothesis that the proportion of patients deemed to respond to a fluid challenge is influenced by the characteristics of a fluid challenge technique, in addition to intravascular filling, vascular tone or ventricular contractility. Other aspects of the fluid challenge, including the volume, type of fluid or assessment time, do not affect the proportion of patients who are fluid responders. Currently, no consensus exists on how to perform an effective fluid challenge. This study highlights the need for a standardised technique for research and clinical purposes.
Fluid challenge is one of the commonest interventions in critical care medicine. A recent international observational study [6] including 2279 patients from 311 centres highlighted the variability in this intervention. In contrast to our results, crystalloids were more frequently used (74.0%), with balanced solutions used in most of cases (53.3%). The study was undertaken following the publication of large, randomised clinical trials advocating the use of crystalloids over colloids [7][8][9][10]. Up to two to three times as much crystalloid as colloid may be required to maintain intravascular volume, owing to differences in intravascular half-life [11]. Fluid challenges consisting of colloids compared with crystalloids are associated with a more linear increase in cardiac filling and SV compared with crystalloids [12]. However, the theoretical benefits of colloids over crystalloids in critically ill patients with altered endothelial permeability have not been borne out in clinical trials. Starch-based solutions are associated with increased rates of acute kidney injury and coagulopathy compared with crystalloid solutions [7,8,13]. Human albumin solution is associated with a poorer prognosis in patients with traumatic brain injury [14] and is not associated with any survival benefit compared with colloids in patients with sepsis [15]. We did not find any difference in PR by the type of fluid used for a fluid challenge. If the time of assessment of fluid responsiveness is immediately after fluid infusion or in the first minutes, it is unlikely that the type of fluid would make any difference, because in both cases (colloids/crystalloids) it is likely that a big proportion of the volume infused will remain in the intravascular compartment. If the assessment of fluid responsiveness were performed later, it would be possible to observe some differences because theoretically colloids remain longer in the intravascular space than crystalloids do. This would require further investigation.
Consistent with a recent large observational study [6], the most common volume of fluid used for a fluid challenge was 500 ml. However, there was significant variability in the volume of fluid used. The total volume of fluid administered to determine fluid responsiveness varies widely, from 4 to 20 ml/kg or 100 to 1000 ml. Whilst fluid challenge with larger volumes may have serious clinical consequences, such as pulmonary oedema, very small volumes may not represent a cardiovascular challenge. The clinical challenge lies in determining the optimal volume of fluid required to optimise cardiac performance and tissue perfusion. The effect of the volume of fluid challenge was recently investigated by our group [16]. Eighty patients were administered four different volumes as fluid challenges (1, 2, 3 and 4 ml/kg of crystalloids) over 5 minutes. Pmsf-arm, a surrogate of the mean systemic filling pressure (Pmsf ), was measured. Pmsf itself is a measure of effective intravascular filling independent of cardiac function [17]. This technique has been shown to be precise for a change of 14% from baseline [18]. The minimal volume required to achieve an increment of 14% was 4 ml/kg. Importantly, the dose of fluids used affects the change in CO and consequently the proportion of patients considered to be responsive to a fluid challenge. Differences in the volume of fluid required to achieve a positive fluid response between this study and other studies in this metaanalysis may be explained by the heterogeneity in the methods used for estimating CO, thresholds defining a positive response, patient case mix and illness severity.
The optimal rate of fluid infusion is unknown. The researchers in the Fluid Challenges in Intensive Care (FENICE) study [6] reported a median infusion time of 24 minutes to administer a fluid challenge. Our results suggest that the duration of the fluid infusion has a significant effect on observed fluid responders. An infusion time <30 minutes is more effective in detecting fluid responders than infusion times >30 minutes. These results are consistent with our understanding of cardiovascular physiology, where a rapid intravenous fluid bolus will rapidly increase venous return to increase right ventricular end-diastolic volume. A slower rate of infusion,  Pooled data in this meta-analysis indicate that the timing of assessment of a fluid challenge does not have a significant impact on detecting a positive response. This is in contrast to previous work by our group in which the haemodynamic effect of a 250-ml crystalloid fluid challenge was almost completely dissipated after 10 minutes from the end of the fluid challenge [19]. In this meta-analysis, many studies used PAC as a method to estimate CO, which cannot accurately detect immediate changes in SV. This makes it more challenging to study the immediate physiological effect of the fluid challenge on SV. A more sustained response would intuitively be clinically favourable. However, this is likely to be influenced by the patient's underlying pathophysiology in addition to the fluid challenge technique itself. In this study, it is possible to comment only on the physiological effect of the fluid challenge, because the clinical effect is beyond the scope of this review. Another possible explanation for the discrepancy in results is the distribution of studies between categories of the assessment time: only five studies reported a time of assessment after 10 minutes, which is the time point at which we have previously observed complete dissipation of the haemodynamic effect of the fluid challenge.
As with all retrospective observational studies, the data presented must be interpreted in the context of its limitations. There is likely to be significant heterogeneity in the patient case mix, illness severity and overall management. Different permutations of the rate of fluid administered, the type and volume of fluid, method of haemodynamic assessment, threshold for definition of responsiveness, and the time of assessment of fluid challenge does not allow any strong conclusions to be made. Furthermore, we have not accounted for the different methods of haemodynamic monitoring used. However, we highlight the heterogeneity in practice of this commonly applied technique and the need for further investigation to elucidate the clinical effect of the different aspects of a fluid challenge.

Conclusions
The proportion of patients who respond to a fluid challenge is dependent on the particularities of the technique used. A rapid infusion of fluid volume increases the proportion of patients with a positive response. However, the type and volume of fluid or the time of assessment does not appear to have any effect on the detection of fluid responders. This study highlights that standardisation of the fluid challenge technique is needed for contextualisation of clinical trial data and patient management.