Skip to content

Awakening the FEAST: Why did fluid boluses kill?


In paediatric medicine, it sometimes feels like we talk about sepsis more than anything else. We’ve all become familiar with the “sepsis bundles”, including a generous portion of intravenous fluid given as a rapid bolus in resuscitation.

This is why the FEAST trial blew everyone’s minds when it was published in 2011. This large RCT in East Africa of over 3000 children was stopped early when it demonstrated an increase in mortality for children given a fluid bolus of either 0.9% saline or 5% albumin instead of just maintenance fluids. It sparked huge controversy and various groups trying to explain away this effect with limited success. Guidelines however were not changed on the basis of this trial, because of concerns that it wasn’t applicable to a high resource setting, because the trial included children without measurement of blood pressure, and partially because no-one could come up with a good explanation for why this would be the case (see this breakdown in The Lancet). Why would giving a fluid bolus increase mortality in sepsis?

This is the question a new paper in The Lancet Respiratory Medicine tries to answer, using data from the FEAST trial (and others), and including members of the original FEAST team.

Levin M, Cunnington A, Wilson C, Nadel S, Lang H, Ninis N et al. Effects of saline or albumin fluid bolus in resuscitation: evidence from re-analysis of the FEAST trial. Lancet Respir Med. Epub June 2019

This study is a little complicated, but we can break it down to make it a bit easier to digest. Let’s start with the most important thing:

Research Question

We want to know why a fluid bolus increased mortality in FEAST.

To get to that, there are 2 questions asked in this study:

  1. Are changes in cardiovascular, neurological or respiratory function, or blood oxygen carrying capacity or biochemistry associated with adverse outcomes in children with sepsis?
  2. In the FEAST trial, were fluid boluses associated with changes in these parameters?

In order to address these 2 questions, the populations analysed were a mix from several studies of childhood sepsis, including:

3170 patients from the FEAST trial (children with sepsis from East Africa)

502 patients from the UK with meningococcal sepsis

448 from Malawi with cerebral malaria

61 children from South Africa with presumed sepsis/gastroenteritis (unpublished)

18,863 children attending a UK emergency department (unpublished)


The authors developed novel physiological scores for cardiovascular, neurological and respiratory function which could be assessed in a low/middle income setting. They didn’t use existing scores, as these are not single organ specific and require monitoring or tests which are not available in these settings, so could not be applied to the FEAST population

The scores were:

The cohorts other than FEAST were essentially just used to test their scoring system. They performed logistic regression to see if an increasing score (i.e. worsening on function) correlated with an “adverse outcome”. Basically testing if their score had any clinical relevance.

In the FEAST population, physiological scores and Hb, lactate, base excess, pH and electrolytes were measured at baseline (prior to any fluids) and then at several time points after fluid administration (either bolus or maintenance). Distributions of these scores were compared in the fluid bolus vs no bolus groups in FEAST to see if this could account for the increased mortality.

Other analyses

To cut a long study short – there are a huge number of different comparative analyses performed in this study. Feel free to spend days wading through the data (I know I will), but for now we’ll stick to the headlines which are the answers to our 2 big questions.

1.    Are changes in cardio, neuro or resp function, or blood oxygen carrying capacity or biochem associated with adverse outcomes?

The answer is yes.

In the FEAST trial, for every 10 units increase (worsening) in score, the odds ratio for death was; respiratory 1.09, neurological 1.26 and cardiovascular 1.09. So they’ve developed a new score, and looks like it tells is what we want.

In the other cohorts analysed, worsening of almost every physiological parameter was also associated with adverse outcomes (see pg 22 supp appendix)

2.    In the FEAST trial, were fluid boluses associated with changes in these parameters?

Yes they were.

Giving a fluid bolus increased (worsened) respiratory score by 3.45 at 1hr and neurological score by 2.64 at 1hr. These differences disappeared at 12h.

Fluid bolus decreased (improved) cardiovascular score by 2.17 at 1h. This difference disappeared at 12h.

Fluid bolus decreased Hb by 0.33g/dL at 8hrs

Fluid bolus did not change lactate

Fluid bolus decreased bicarbonate by 0.96mmol/L and decreased base excess by 1.41mEq/L

Additional points of interest

It made no difference if they received albumin or saline boluses, which is expected as there was no mortality difference either.

Higher volume fluid boluses (≥30mL/kg) were associated with worse respiratory and neurological scores at 4h then lower volume boluses, but high or low volume bolus made no difference in cardiovascular scores.

Significantly, in a post hoc principle component analysis (stay with me – this is not complicated I promise) – differences in physiological scores at 1h explain all the differences in mortality observed between bolus and no-bolus groups of the trial. The major determinants were neurological score, base excess and respiratory score. This is important as although differences between the groups were only apparent at 1h, this was enough to completely account for the differences in mortality.

There is some exciting cluster analysis (1. Mild derangement of scores, 2. Severe anaemia, high lactate and CV score, 3. High resp and neuro scores), but we will ignore this for now as the answer to our main questions didn’t differ among any of the clusters.

So how does a fluid bolus increase mortality in sepsis?

Whilst this question cannot be definitively answered through this study, the authors have presented a bioplausible explanation of their results. They have summarised it in an amazing picture, but if you like bullet points…

A fluid boluses causes:

  1. Haemodilution, causing anaemia, reducing O2 delivery and worsening metabolic acidosis
  2. Hyperchloraemia and bicarbonate dilution, worsening metabolic acidosis
  3. Worsening of respiratory function (partially through increased acidosis), causes decreased CO2 excretion, which both worsens metabolic acidosis and causes cerebral vasodilation, increasing intracranial pressure
  4. Cerebral oedema, causing raised intracranial pressure.

Limitations of the study

Should we all stop giving fluid boluses to children with sepsis now? Some limitations to consider from this study include:

  • This resource limited setting without intensive care may not be directly applicable to high income settings, as mechanical ventilation and neuro intensive care may be able to mitigate some of the adverse effects of bolus
  • Their scores are not independently validated and were not initially derived from the data
  • The influence of large volume fluid boluses may suffer from confounding by indication
  • Acid-base data was only available at 24hrs, by which time most deaths had occurred. Some analysis was performed using imputations.
  • Some of the statistics are almost impenetrable to us mere mortals


In the FEAST trial of children with sepsis in East Africa, a fluid bolus of either 0.9% saline or 5% albumin was associated with a worsening of respiratory and neurological function, anaemia and metabolic acidosis, which in turn was associated with increased mortality. There was a transient increase in cardiovascular function.

The implications of this according to the authors:

Particular care should be taken if considering giving fluid boluses to children with sepsis and respiratory or neurological compromise at presentation

Resuscitation with buffered solutions may be preferable in severely unwell children to avoid worsening of metabolic acidosis (trials in adults currently underway to assess this)

Where next?

As the authors state, it is unlikely ethical approval will be given for further trials of fluid boluses in a low/middle income setting following the increased mortality shown in FEAST. Conversely, trials in higher income countries are difficult because fluid boluses are seen as standard of care for sepsis, and as demonstrated recently in the FISH trial, sepsis has become so rare that trials of fluid management are difficult, even approaching unfeasible. We may find people become generally more cautious when implementing fluid bolus resuscitation as time goes on, perhaps with 10mL/kg boluses becoming preferable to the historically recommended 20mL/kg.

For further reading, check out the great accompanying editorial here, and check out references 3 – 14 in the paper for some of the heated controversy surrounding FEAST.


  • Alasdair Munro is a Paediatric registrar in the UK, currently working as a Clinical Research Fellow in Paediatric Infectious Diseases. His interests include evidence based medicine, diagnostics and antimicrobial resistance. @apsmunro | Ally's DFTB posts


No data was found

Leave a Reply

Your email address will not be published. Required fields are marked *

1 thought on “Awakening the FEAST: Why did fluid boluses kill?”

  1. Think it’s also important to read the paper by Kathryn Maitland responding to the Levin paper
    The article by Michael Levin and colleagues1 rekindles the discussion about adverse effects of boluses in children with febrile illness and shock, as reported in the FEAST trial.2 However, unfortunately, results of this re-analysis of FEAST data fall very short of providing mechanisms by which boluses increase mortality; differences in physiological scores and biochemistry results between groups by no means imply causality. Furthermore, some re-analyses and statements about our previous papers are erroneous and contradict work by the FEAST team aimed at understanding these mechanisms.
    We firmly stand by our previous analyses of terminal clinical events (TCEs) in FEAST showing that the main reason for excess deaths with boluses was cardiovascular collapse, and not respiratory or neurological causes
    Unfortunately, the statistical methods used have major errors which invalidate their conclusion that “Hyperchloraemic acidosis and respiratory and neurological dysfunction induced by saline or albumin bolus explained the excess mortality due to bolus”