The colloid vs crystalloid debate in the resuscitation of critically ill patients has come up in previous exam papers, but has not appeared in the last ten years or so. Question 29 from the second paper of 2007 asked the candidates to compare albumin with gelatin (Haemaccel and Gelofusine). In contrast, the more thoughtful Question 1 from the second paper of 2001 asked the candidates to choose one example each of colloid and crystalloid, and then to debate their relative merits in the context of resuscitating a patient with biliary sepsis. This 2001 question needs to be viewed in the context of the windblown limbo which occupied the time between the 1998 Cochrane review of albumin (discussed in greater detail below) and the 2004 SAFE study. In addition, an excellent resource (perhaps also helpful in the dark times of the CICM primary exam) is a 2014 paper by Langer et al (where Luciano Gattinoni hides quietly in the et al).
Both questions could have been answered in a tabulated format, and these tables can be found in the Appendices at the end of the chapter. In order to prepare for the possibility of future SAQs, what follows is a formulaic answer to the question, "critically evaluate the use of colloid as a resuscitation fluid in intensive care".
In summary, crystalloid is currently “in”, and colloid is “out”
- Recent guidelines (eg. Scandinavian guidelines) recommend crystalloid over colloid.
- This reflects the broad trend in critical care over the last 20 years.
- However all evidence (eg. meta-analysis with n=16,000) points to no difference in mortality.
- Theoretical benefits in the use of colloid (eg. defence of the glycocalyx with albumin) have thus far not translated into measurable hard outcome advantages.
- Practical benefits observed with colloid infusion (eg. a more rapid achievement of haemodynamic goals) have also failed to improve mortality.
Specific problems with colloids:
- Some colloids have a propensity to cause renal failure (eg. hydroxyethyl starch)
- Some colloids have a propensity to cause coagulopathy (eg. gelatins)
- Albumin is expensive and may have no survival benefit
Specific problems with crystalloid:
- It takes more crystalloid volume to achieve the same resuscitation endpoints
- A more positive fluid balance has been associated with worse outcomes in certain conditions (eg. sepsis)
During the Second World War, all manner of gelatins and branched sugars were used to make up for the shortage of human plasma, and for the perils of storing and transporting whole blood. This was a time of much experimentation, and people were trying all manner of weird stuff (eg. intra-arterial injection of resuscitation fluid, or whole blood as by Case et al, 1953). One front-runner which seemed promising was polyvinylpyrrolidone (PVP). Introduced first in 1943, this synthetic colloid ended up being given to about three hundred thousand casualties on the Russian front in Germany. Unfortunately, as it turns out PVP is retained (permanently!) in the reticuloendothelial system (Altemeier et al, 1954).
Of the non-synthetic colloids, dried plasma was a popular choice. Opposite, one may see images of the original packaging for a unit of dried plasma (stolen shamelessly from the US Army Office of Medical History). "The components of the unit, consisting of a flame-sealed glass vial of dried plasma, a glass bottle of distilled water, an intravenous needle, and rubber tubing, were rather loosely packaged in a cardboard box." The unit supplied the dried content of approximately 300ml of plasma.
An account of the contemporary resuscitation practice is offered by Allen Gold's "Resuscitation of the Severely Wounded" (1946). Gold was a captain in the Canadian Army and served with the Field Transfusion Unit in the North-West Europe campaign from 1944 to 1945. A representative case was selected to stimulate discussion ("...diagnosis: shell wounds, multiple, both legs, thighs and buttocks, compound fracture left femur"). The practice was to transfuse 1 unit of plasma for every 2-6 units of whole blood; the mentioned advantages of the plasma solution were reduced viscosity (allowing a flow rate of 100ml/minute) and easier availability (as whole blood suffered lesions in storage).
During the subsequent decade, the favoured plasma expander was 6% dextran. It was discovered by Swedish chemists who were performing research on sugar, sponsored by the sugar industry. Frazer Gurd’s “Current trends in the treatment of shock” (1955) reports that 6% dextran was the obvious favourite, waving aside concerns about renal damage. The product referred to in this article was probably Macrodex, a dextran with an average molecular weight of around 75,000Da. However, this was also the time during which isolation and storage of albumin was being perfected, and many looked upon it as “the ideal resuscitation colloid”.
Colloids entered their main period of popularity after the publication of studies such as Shoemaker et al (1981). The investigators observed a greatly reduced required volume of resuscitation fluid, as well as “greater increases in hemodynamic and oxygen transport variables “ after 5% albumin when compared to lactated Ringer's solution. Up to 4 times the volume of crystalloid was required to achieve the same resuscitation goals.
As a consequence, colloids (particularly synthetic colloids such as hetastarch and gelatin) became popular, as a cheap alternative to albumin which was no less effective. “Plasma volume must be maintained to prevent a decreased blood flow to vital organs such as the kidneys. Blood or colloid solutions, not crystalloid solutions, should be used for this purpose, since the latter are distributed throughout the whole extracellular space and are less effective in maintaining plasma volume”, wrote Twigley and Hilman in 1985 (“The End of the Crystalloid Era”). The enthusiasm for their use was such that many expected a complete abandonment of crystalloid as resuscitation fluid (only 5% dextrose would remain, used to contribute free water ). From that point forth , it seemed all patients everywhere would get colloids, for any indication.
Enthusiasm for the wanton use of colloids lasted until the mid-to-late 1990s, when meta-analysis of RCTs demonstrated an apparent trend in the direction of increased mortality from colloid solutions. For one example, Schierhout et al (1998) found a 4% increase in the risk of death when only trials with adequate concealment of allocation were included. “This systematic review does not support the continued use of colloids”, the authors concluded.
It got worse. The death of the Colloid Age is chronicled in the excellent article by Wiedermann et al (2015) In short, Wiedermann describes how a Cochrane review of human albumin (Cochrane Injuries Group, 1998) killed albumin, to such an extent that Europeans switched almost completely to starch. The Cochrane Injuries Group called for a ban on the use of 5% albumin on the basis of a 6% increase in mortality (numbers needed to kill = 17). Mortality in the albumin patients approached 70%, suggesting some sort of horrific risk.
The use of albumin declined steeply after this publication, in spite of the study’s many flaws, or the fact that the 1998 Cochrane findings were never again reproduced by other meta-analysis studies. The authors did not suggest any sort of plausible explanation as to why albumin was killing people, or explain why their findings were dramatically different to the known very low adverse event rate associated with albumin (about 5.24 events per 100,000,000 doses).
It seems the 1998 Cochrane analysis was poisoned by small-trial bias. The authors included numerous small trials with sizes ranging from n=219 to n=14. Wilkes and Navickis (2002) performed a more statisctially sound analysis in 2002 and found no mortality difference, just like every other colloid vs crystalloid study ever since. However, European use of albumin remained low even after the SAFE study (Finfer et al, 2004) demonstrated its safety. Even twelve years later after the Cochrane publication, Finfer et al (2010) found the use of albumin was almost non-existent in Europe (the graph above is shamelessly stolen from Finfer’s paper, with no permission whatsoever).
The European anxieties about albumin had also percolated into the local environment. Hammond et al (2015) surveyed units in Australia and New Zealand between 2007 and 2012, finding that the rate of crystalloid use had changed from 28.9% in 2007 to 50.5% in 2013. The decline use of colloids and the increase in the use of crystalloid was attributed to the increased availablility and more widespread use of "buffered salt" solutions, which have fewer biochemical disadvantages when compared to normal saline.
After the death of starch in 2012, the modern era of colloid vs crystalloid reserach is characterised by negative studies. At the time of writing, the most recent efforts at meta-analysis (eg. Qireshi et al, 2016) can identify around sixty eligible studies, containing over 16,000 patients (of which 7000 come from the SAFE study alone). Again, no mortality difference is seen. Unfortunately, this is not very good for guideline authors, who need to make some sort of recommendations. The Scandinavian Guidelines (Perner et al, 2015) have thus far erred on the side of caution, and recommended crystalloid instead of any colloid for resuscitation in critical illness.
The following is a list of the more famous landmark studies which have been influential in the course of this controversy, and which have helped shape current intensive care practice.
This thing is probably the most famous ICU study next to the ALVEOLI trial. It was certainly one of the biggest, with 6997 patients enrolled. With such numbers, SAFE was powered to detect a 3% absolute reduction in mortality. On comparing 4% albumin and normal saline, the investigators found no difference whatsoever in their primary endpoint (28 day mortality). A lot was made of post-hoc analysis of subgroups, some of which found statistically significant influences on mortality (worse for head injury, better for sepsis) none of which were meaningful because the subgroups weren’t powered to detect them, and because of the post-hoc nature of the analysis.
Criticm of the SAFE study included several points made by subsequent authors:
The 6 S's stand for " Scandinavian Starch for Severe Sepsis/Septic Shock ". Like chest, 6% HES was studied, but this time in the specific setting of severe septic shock. On a smaller scale than CHEST, Perner et al randomised about 800 severly septic patients to receive HES 130/0.42 or Ringer's acetate. The oucomes were a bloodbath, strongly discouraging the use of starch: at 90 days, there was 51% mortality in the colloid arm, versus 43% mortality in the crystalloid arm (a 20% mortality increase!). There was also increased risk of renal replacement therapy, a trend for increased bleeding, and increased blood product transfusion in the starched group. Ringer's acetate was also the carrier vehicle for the trial starch fluid, which makes the study a well-designed comparison, isolating one specific component of the resuscitation fluid.
Some criticsm does exist. There was no control for co-interventions. Also, 69 patients had protocol violation- 28 patients in the starch group and 41 patients in the Ringers acetate group were resuscitated with larger volumes of trial fluid than was permitted (there was a calculated weight-based maximum) - theoretically, exceeding the maximum dose of strach places one at greater risk of acute kidney injury, and this may have skewed the data in favour of acetate. Also, though the study reports on incidence of renal replacement therapy as one of the secondary outcomes, there was no specific criteria for initiation of RRT (it was left up to the clinician) and concievably starch-hating clinicans may have pulled the trigger on dialysis earlier in the patients whom they suspected were poisoned with trial starch. Lasly, the authors were unable to offer a compelling explanation of the mechanism by which HES contributed to mortality or acute kidney injury (though others have, eg. Hartog et al, 2012).
This study pretty much killed starch. The Surviving Sepsis people have quoted these data to make strong recommendations against the use of starch in their bundle. Estrada and Murugan (2013) described its publication as the end of the Starch Era (if there was even such an era). This study, combined with CHEST and with the humiliating public retraction of Joachim Boldt's starch studies had resulted in the withdrawal of starch fluid from most formularies.
The CHEST study is another one of those humongous ANZICS CORE monoliths. 7000 generic ICU patients were randomly assigned to receive 6% HES (with saline carrier) or saline. Of these, 18% of the starched and 17% of the salted patients ended up dying at 90 days, which did not add up to a statisctically significant difference. These patients were much less sick than the 6S group, and this may account for the lack of mortality difference (as it is possible to infuse just about any random material into healthy people without killing them). However, a disturbing trend towards worse renal outcomes was again detected ( 7.0% of HES patients vs. 5.8% of saline patients).
CRISTAL was a largely European multicenter randomized controlled trial; n= 2857. Colloid was compared to crystalloid as a resuscitation fluid. For maintenance, anything was permitted. The study colloids were also quite diverse, ranging from dextran and hydroxyethyl starch to 20% concentrated albumin.
At 28 days, there was no mortality difference; only at 90 days was there any difference (30.7% dead in the colloid group vs. 34.7% in the crystalloid group). The colloid group was also better at staying off the ventilator and off vasopressors.
However, the colloid group was quite diverse: anything vaguely colloidal was lumped together as “colloid”, which means results from patients receiveing 20% albumin were analysed in the same group as patients receiving the hideously toxic hydroxyethyl starch. The nephrotoxicity of the aforementioned starch is may have made the results unfairly skewed in favour of crystalloid (or rather, the results of CRISTAL could have been even more in favour of colloids). However, between the two groups there was no statistically significant difference in the incidence of starchy renal failure.
And recruitment was not completed in time, (the goal of n=3010 was not achieved; it was required to power the study to detect a 5% mortality difference at 28 days). The recruitment period was overly long, lasting from 2003 until 2012, and therefore spanning an entire era of fluid resuscitation philosophy. Over the course of this time we saw the publication of the SAFE study, the starch papers, and the rise and fall of early goal-directed therapy for sepsis. So, they ran out of time and money, and their study was not powered to detect their primary endpoint.
This was an Italian multicentre randomised controlled trial, enrolling 1818 severely septic patients to either receive albumin (20%) or crystalloid. Albumin appeared to improve mortality of septic shock patients once hemodynamic stability has been achieved, and was associated with some haemodynamic advantages – which, on close inspection, appeared ridiculous. For instance, the increase in MAP was by 1-2mmHg in the albumin group, which reached statistical significance and was listed as an advantage of albumin, but which would not be viewed as a great haemodynamic victory by any sane person.
Below is a table of colloid solutions (from Question 29, second paper of 2007)
|Property||Albumin (20%)||Gelofusine 4%||Dextran (10%)||Hydroxyethyl starch 6%|
|Drug class||Endogenous protein||Succynylated bovine gelatin||Branched polysaccharide||Amylopectin derivative|
|Molecular weight||69 000 Da||5 000 - 15 000 Da||14 000-18 000 Da||70 000 Da|
|Plasma halflife||24 hours||2.5 hours||12 hours||5 days|
|Elimination||Degradation by reticuloendothelial system||Renally excreted||Renally excreted||Some renally excreted,
some metabolised by the reticuloendothelial system
|Plasma expansion as a percentage of infused volume||200-400%||70-80%||100-150%||~100%|
Free radical scavenging effects
Protection of glycocalyx
Relatively safe in renal failure
No limits on infused volume
Decreases the viscosity of blood, improving microcirculation
No risk of CJ disease
Large maximum allowable volume
No risk of CJ disease
Lowest risk of anaphylactoid recations among non-albumin colloids
Risk of CJ disease
Interference with ABO crossmatch
Renal failure (ATN)
Increase in amylase
Below are three tables of intravenous fluid composition, presented here for rapid reference. These tables include the familiar selection of fluids known in Australia, as well as some exotic Euroweirdness ( Sterofundin?). They were "borrowed" with no specific permission from Langer et al (2014), in an attempt to also borrow some accuracy and credibility. It is important, when asked "where'd you get those numbers from", to be able to point at a scholarly publication. Unfortunately, borrowing from this specific scholarly publication lends neither accuracy (the potassium content of a bag of saline is given as 154 mEq/L) nor credibility (as does not give references to where their numbers come from). There are three possible sources for their data, in an order of increasing unlikelihood:
With that caveat, here's the data.