FOAM Eye-Catchers 4

Time for a Rethink on IV Fluids in Sepsis?

Earlier this year in March, I was fortunate enough to pilgrimage to the inaugural FOAMite mecca – SMACC (Social Media And Critical Care) Conference in Sydney. Quite frankly, it was the best conference I’ve ever attended. Some of the highlights included talks by Prof John Myburgh, in particular one on IV fluids in Sepsis, available for free via SMACC talk archive.

Recently Emcrit released a paradigm shifting lecture by Dr Paul Marik on fluids in sepsis that builds on some of the theme’s of Prof Myburgh’s talk. Not long after, Scott Weingart released his response to Dr Marik’s lecture.

All 3 talks are well worth listening to in full. However I’ve summarised some of the key concepts below as well as the FEAST study which is a landmark study discussed in both talks:

  • Albumin is safe to use as demonstrated in the SAFE trial compared with normal saline. The only exception was in the subgroup of traumatic brain injury where there was increase in short and long term mortality as it leaks across the damaged blood brain barrier worsening oedema. There was a suggestion of benefit of albumin in the subgroup of severe sepsis patients that was not (but was almost) statistically significant (RR reduction in death 0.87, p= 0.06).
  • As a colloid, albumin was far less fluid sparing than expected. Instead of colloid-crystalloid ratios of 1 to 4 or 1:5 as theorised, the actual ratio was only 1 to 1.4.
  • Colloid starches increase the risk of renal failure, renal replacement therapy and mortality and should not be used
  • Normal Saline is not physiological. There was an error in the original experiment to determine an isotonic solution, resulting in the development of a hypertonic, hyperchloraemic solution. There is an argument to consider the use of balanced solutions instead but this needs further study.
  • Reduce your use of fluids in sepsis. Only give what you’ve lost – if you can’t see it leaking out [via a drain or catheter] don’t give it.
  • The FEAST study showing increased mortality from fluid boluses in sepsis “poses the question that we need to think very very carefully over how much fluid we give and over what period of time” (see below for more on FEAST).


  • Sepsis is not a volume depleted state
  • Giving fluid may worsen septic shock by causing vasodilation, leaky capillaries, and myocardial depression
  • The Starling mechanism (hydrostatic pressures v protein oncotic pressures) of fluid movement between compartments is wrong because it fails to take into account the Endothelial Glycocalyx (EG). Despite its initial discovery 40 years ago using electron microscopy, the EG is a relatively new revelation in understanding blood vessels and fluid movements that is well worth spending the time to get your head around to more fully understand fluid management.  A great starting point is this LITFL post which includes a 5 minute easily understandable video, and I’d also recommend checking out the references at the bottom as well. 
  • Damage to the EG in sepsis is a primary reason for the movement of protein and fluid into the interstitial space
  • Fluid resus causes peripheral vasodilatation and increased atrial pressure. This results in ANP and BNP release which cause shedding of the EG and resultant leaky capillaries.
  • Human studies have shown that excess fluid administration is an independent predictor of increased mortality with dose-response relationship demonstrated. Similarly negative fluid balance predicts survival in patients with septic shock. However these are not proven causal associations.
  • In animal studies of sepsis there were causal associations between increased fluids and increased mortality despite expected improvement in haemodynamic parameters occurring.
  • The landmark FEAST study demonstrated a causal association between giving fluid bolus in septic shock and increased mortality (see below for more on FEAST).
  • “Give patients as much as they need and not a drop more”. Only given fluid to patients who are are fluid responsive measured by lifting the legs and measuring cardiac output (not pressure) via an appropriate monitor within 3 minutes. This apparently has 95% accuracy at predicting fluid responsiveness. In his ICU this test of fluid responsiveness is determined by the nurses.
  • In general in sepsis he gives repeated fluid boluses up to a maximum of 30ml/kg (eg 2L in average adult) and then starts early use of vasopressors.
  • Vasopressors tense the venous side increasing stressed volume, preload and cardiac output.
  • On fluid choice: avoid “AbNormal” Saline due to its adverse effect on renal perfusion and acidosis. Starches should not be used. Albumin has theoretical advantages in sepsis as it binds to and maintains the EG.


Scott Weingart, while agreeing with the premise that we should be cautious with fluids in sepsis argues against several points including but not limited to:

  • Much of the data was in reference to “fluid balance” and this does not equal “fluid administration”.
  • He explains that most of the data cited by Marik showed non-causal associations or referred to animal studies, including data associating fluids, ANP and BNP with worse outcomes.
  • In the key FEAST study showing causal association of fluid bolus with mortality, he suggested this may be due to the high chloride content of the administered fluid rather than the fluid volume itself (discussed below).


The FEAST Study deserves its own page to examine so I’ve discussed it here in detail. However a brief summary is that it showed in febrile kids in Africa with sepsis a 20ml/kg fluid bolus increased mortality compared with just giving maintenance fluid. It was a landmark study performed with high quality in difficult conditions and must inform our current practice. Many of the criticisms of the original study have since been debunked by the recently released additional analysis by the authors and the associated analysis by Myburgh and Finfer.


In summary there seems to be a controversial paradigm shift away from our fluid liberal approach to sepsis management where we tend to resuscitate patients to short term haemodynamic parameters such as blood pressure, as long as they are fluid tolerant, towards a more cautious approach giving smaller volumes at slower rates to those who we predict to be fluid responsive based on cardiac output. There is an absence of high quality evidence supporting our aggressive approach to fluid resuscitation in sepsis and the emerging evidence questions this approach. Our focus on short term haemodynamics may be misguided as despite improvements with fluid therapy this does not necessarily correlate with the patient’s outcome and may possibly adversely affect it.

In short, we should extend the same due care and consideration when administering IV fluids as we do when we administer any other medication.