Search Strategy

Our systematic review was conducted in accordance with the PRISMA guidelines for reporting systematic reviews and meta-analyses [15].

Three databases – MEDLINE via PubMed, EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL) – were searched independently and in duplicate by 2 reviewers (SH, NF) from inception to February 29, 2012 with no geographical or language restrictions using a compound search strategy detailed in the pre-defined protocol (File S1). We additionally searched bibliographies of relevant reviews and contacted experts in the field in an attempt to identify relevant studies. Data extraction was done independently and in duplicate by two reviewers (NF, SH). The review sought randomized trials, quasi-randomized trials, and controlled before-after studies assessing children with septic shock and/or shock and severe infection (as defined by the studies) in which at least one group was treated with bolus fluids. Studies that only addressed non-infectious causes of shock, neonatal shock, or patient populations with severe dehydration, were excluded consistent with previous systematic reviews [7]. Studies in which >30% of participants were considered to have septic shock were included, but outcomes were not pooled. The primary outcome was mortality at 48 hours. Secondary outcomes included mortality at 4 weeks and adverse clinical events. Results were pooled according to cause of septic shock.

Assessment of Methodological Quality

Individual studies were rated according to three main indicators of methodological quality of randomized trials: allocation concealment, loss to follow up <20%, and reporting of adverse events. For each category of septic shock, assessment of methodological quality followed GRADE which rates evidence according to four criteria: limitations, inconsistency, indirectness, and imprecision. Publication bias was considered as a potential limitation of the systematic review overall.

Data Analysis

Relative risks (RRs) and 95% confidence intervals (CI) were calculated and data pooled using fixed-effects method, in which the weight assigned the estimated treatment effect from a given trial is proportional to the amount of information provided by that trial. The robustness of this analysis was explored in sensitivity analysis using the random-effects method [16]. Data were pooled according to pre-defined subgroups depending on cause of sepsis given differences in prognosis, and heterogeneity estimated by the I² statistic. Point estimates and 95% CIs were calculated for the frequencies of adverse events. All analyses were conducted using Stata, version 12 (StataCorp LP, College Station, Texas, USA) and GRADE Pro (www.gradeworkinggroup.org).

Study Inclusions

The search strategy yielded 342 articles that were screened by title and abstract. An additional 14 articles were identified through bibliographic searches and contact with experts. In total, 13 studies met the inclusion criteria and were taken through for full review [8], [17]–[28]. Study inclusions as well as final exclusions are detailed in Figure 1. Studies were done in populations with malaria (4 studies), dengue fever (4 studies), and mixed causes of septic shock (4 studies). In addition, one study was done in children with severe malnutrition, in which over a third (34%) were determined to have hypovolemic shock secondary to sepsis. Baseline characteristics of included studies are summarised in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Study inclusions and exclusions.

https://doi.org/10.1371/journal.pone.0043953.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study characteristics.

https://doi.org/10.1371/journal.pone.0043953.t001

Assessment of Methodological Quality

The assessment of methodological quality of individual studies is summarised in File S2. Overall, the methodological quality of trials was high: most studies used allocation concealment (10/14) and reported adverse events (12/14), and all had low rates of loss-to-follow up. The GRADE evidence assessment for each category of shock trial is summarised in File S3. For septic shock the quality of the evidence was rated as high; this rating was driven largely by the contribution of the FEAST trial. For malaria the quality of the evidence was rated high if the FEAST trial data was considered, otherwise it was rated as moderate, mainly due to imprecision. For Dengue the quality of the evidence was rated as moderate; this was due to the rating of serious imprecision driven by the low event rate in trials. For malnutrition the quality of evidence was graded as low because there was only a single, small trial.

Primary Outcomes

Primary outcomes of mortality at 48 hours are summarised in Figure 1.

Four trials assessed interventions in children in septic shock [8], [17]–[19]. Overall mortality across all studies was 10.7%. These studies provide evidence on four comparisons: no bolus vs colloid bolus (2094 patients), no bolus vs crystalloid bolus (2091 patients), colloids bolus vs crystalloid bolus (2157 patients), and different formulations of crystalloid bolus (160 patients). The only significant effect was found in the FEAST trial, the only large trial to compare no bolus vs bolus; in this trial the no bolus group (control) had a lower mortality compared to the bolus group (RR 0.69; 95% CI 0.54–0.89 [Figure 2]). There was no other difference in mortality comparing crystalloid vs colloid (RR 1.01; 95%CI 0.80–1.28; I² 0%; p = 0.9 [Figure 3]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Forest plot for the outcome of mortality comparing no bolus and bolus.

https://doi.org/10.1371/journal.pone.0043953.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Forest plot for the outcome of mortality comparing colloids and crystalloids.

https://doi.org/10.1371/journal.pone.0043953.g003

Four trials (378 patients) assessed interventions in children with shock associated with malaria infection [20]–[23]: a further 1793 children in the FEAST trial had malaria. Overall mortality across all studies was 16.4%. The studies provide evidence comparing bolus and no bolus (2005 patients), colloids vs crystalloids (118 patients), crystalloids vs maintenance therapy (133 patients), and different formulations of colloid (167 patients). For this subgroup of patients, no bolus was found to decrease mortality compared to bolus (RR 0.64; 95%CI 0.45–0.91 [Figure 2]). Heterogeneity was low (I² 0%; p = 0.7). This finding was unchanged using the random-effects method (RR 0.65; 95%CI 0.46–0.92). No statistically significant difference was found for any other comparison (Figure 3).

Four trials (811 patients) assessed colloids vs crystalloids for the treatment of children in dengue shock [24], [26]–[28]. Overall mortality was low at 1.3%. There was no difference in treatment effects across arms (pooled RR 0.61; 95%CI 0.11–3.48). Heterogeneity was low (I² 0%; p = 0.7 [Figure 3]).

One trial was identified assessing the role of fluids in children with severe acute malnutrition (61 patients, of whom 21 had sepsis) [25]. This trial had high mortality (50.8%), but found no difference between study groups which compared an isotonic crystalloid (Ringer’s lactate) against two hypotonic crystalloids (human albumin solution [HAS] or HSD/5D; RR 0.98; 95%CI 0.42–2.28 [Figure 3]).

Although the FEAST trial excluded children with severe malnutrition, 70 (2%) children had a mid-upper-arm circumference (MUAC) ≤11.5 cm (indicating severe acute malnutrition). The effect of bolus fluids was not significantly different in children with a mid-upper arm circumference of >11.5 cm (p = 0.96) [29].

Secondary Outcomes

Only one study, the FEAST trial, reported mortality at four weeks. Overall, no bolus was protective against mortality compared to bolus (RR 0.69; 95%CI 0.54–0.87). Adverse clinical events, reported by all studies, were low, irrespective of intervention, and ranged from 0% to 11.1% (95%CI 4.2–22.6).

Interpretation

The majority of all randomized evidence to date comes from the FEAST trial, which found that fluid boluses were harmful compared to no bolus. The 2008 Surviving Sepsis Campaign Guidelines, informed by a modified Delphi process, graded the current paediatric recommendation (20 mL/kg boluses over 5–10 minutes up to 60 mL/kg) as 2C, indicating a weak recommendation with low quality of evidence [30]. Although a single trial, the evidence provided by FEAST that fluid bolus are harmful compared to no bolus is of high quality and sufficient precision, suggesting that the withholding of bolus fluids should be considered for populations similar to those enrolled in FEAST. The important question is the extent to which these results are applicable to other populations. The FEAST trial excluded patients likely to have fluid loss either through bleeding/burns or dehydration due to gastroenteritis. Malnourished children were also excluded. The study population did not include neonates nor children with dengue fever. Caution must be taken in extrapolating the findings beyond populations similar to those included in this trial.

Much of the debate around the validity of the results of this trial have focused on the applied definition of septic shock [9], [14]. Part of the difficulty rests on the fact that there are no uniform definitions for septic shock and many guidelines lack stringent criteria. A re-analysis of the trial data found the results to be robust to the application of different definitions of shock, and while only 65(2%) of children fulfilled the strict WHO definition of shock, even in this small subset there was a significant excess risk associated with boluses with an absolute risk difference of 28% (95%CI 3.4–52.5) [29].

Across all populations, there is no evidence that colloids are superior to crystalloids. A previous systematic review of this question published in 2010 reported outcomes from nine studies (1230 children) and concluded that the evidence base is limited [7]. This systematic review adds to the findings of the previous review principally by including data from the FEAST trial, which increases confidence in this conclusion within the limits of external validity. Of note, more than half of patients enrolled in FEAST had malaria, and overall mortality rates were similar in FEAST to other malaria trials. For other specific populations such as Dengue shock and shock associated with malnutrition, the evidence-base remains limited. Nevertheless, the high rates of survival demonstrated by the Dengue trials provides moderate evidence to support fluid resuscitation for these patients.

There are several strengths and limitations to note. Strengths include the restriction of inclusion of comparative trials and the large meta-analytic dataset compared to previous reviews. Limitations of the evidence-base include the small sample sizes resulting in poor precision for specific populations, in particular malnourished children. In addition, there is inconsistent reporting of secondary outcomes, which limited analysis, although the most important outcomes (mortality and adverse clinical events) were reported by all studies. Few trials included a control group making it impossible to assess the impact of the bolus itself in most trials. Subgroup analyses carry the risk of spurious findings, but these analyses were limited in number and pre-specified in the research protocol. Another potential limitation of this review is the search strategy (only 3 databases searched). Attempts were made to limit the possibility of having missed studies by using a highly sensitive search strategy and consulting with experts in the field. Publication bias is an ever present risk of any systematic review. We were unable to assess publication bias formally due to the limited number of studies identified for review, but the results do not appear to suggest publication bias. Finally, as highlighted by the debate generated following the publication of the FEAST trial, the lack of a standardised definition for shock is an important caveat to consider when comparing different studies. Nevertheless, the results of the FEAST trial were found to be robust to a range of sensitivity analyses that applied different definitions of shock [29].

The most important direction for future research is the applicability of the findings of the FEAST trial to other populations and settings. Simple algorithms are needed to support health-care providers in the triage of patients to determine who could be potentially be harmed by the provision of bolus fluids, and who will benefit. Further studies are urgently needed in the area of shock associated with malnutrition, because the evidence base is scant and mortality high. Work is needed to establish a uniform definition of shock to assist in the comparability of future studies and the application of practice guidelines. The FEAST trial used relatively modest fluid boluses in both rate and time of infusion compared to the other trials included and in comparison to current guidelines. Despite these conservative volumes and rates, bolus therapy was still found to be harmful. Finally, more research is needed to determine exactly what role for fluid therapy in this population beyond the use of bolus therapy.

Conclusions

The majority of all randomized evidence to date comes from a single trial, which found that fluid boluses were harmful compared to no bolus. While this finding cannot be applied broadly, it does provide for the first time strong evidence on which to base guidelines for management of paediatric septic shock. In the subpopulation of children with haemorrhagic dengue fever, there is moderate evidence to support fluid resuscitation. A priority for future operational research, therefore, is the definition of practical guidelines and algorithms that will allow health-care providers to distinguish between those groups of children likely to benefit from fluid bolus, and those children who could be harmed by this intervention.