There are no set guidelines for fluid administration. A 2006 article by JL Vincent attempts to bring some sort of order into the lawless Mad Max wasteland of fluid resuscitation practice; another attempt was made in 2011 by Cecconi et al. The latter paper forms the basis of the summary which follows.
This topic came up in Question 24 from the second paper of 2012; the college wanted "general guidelines" used to administer a fluid challenge, and a list of parameters used to predict fluid responsiveness. Judging from the college "model answer" it sounds like the examiners were prepared to tolerate a range of wacky responses. Again, in Question 24 from the second paper of 2016 the examiners wanted a definition of fluid responsiveness (for 1 mark) and some details about the physiological basis and limitatuions of CVP, passive leg raise and pulse pressure variation.
Definition of fluid responsiveness
What exactly is "fluid responsiveness", anyway? What are we assessing when we assess "fluid responsiveness", and how can one rate the accuracy of an assessment method if one cannot agree on what precisely is being assessed? This is very frustrating. For example, Paul Marik, an intensivist with worldwide rockstar-like fame, suggests that a response to fluids is "an increase of stroke volume of 10-15% after the patient receives 500 ml of crystalloid over 10-15 minutes". Others have used measures like a 10% increase in cardiac output. Stroke volume seems like the most sensible measure, because stroke volume is the main variable which changes in response to changes in preload.
Of course, all of this makes little sense. Let us picture ourselves at the bedside. Let us say we find a parameter which suggests the patient is in need of fluids, like reduced urine output, or a decreased cardiac index, or an increase in the "swing" of the arterial line. So, we give fluids. That parameter then either changes (labelling the patient as "fluid-responsive" or it doesn't change ("fluid-unresponsive"). Does the patient get better? Are they somehow less sick because of this manipulation?
Who can say. The LVEDA might look better, or the IVC might look more plump, or the noradrenaline infusion rate might have decreased by 2ml/hr. But the gut might have become that little bit more oedematous, the clotting factors that little bit more diluted; the space between capillaries in the boggy tissues has increased, making it more difficult for oxygen to diffuse across. So perhaps overall there was no benefit, but we have fooled ourselves into associating the patient's wellbeing with the diameter of one of their vessels, so we feel as if our actions have had some sort of positive effect.
Ultimately, the gold standard of fluid responsiveness testing is a fluid bolus. Nothing says "I might respond to fluids" better than somebody actually responding to fluids.
Methods of predicting fluid responsiveness
Fluid boluses are usually a reflex response to hypotension or hypoperfusion. However, after one has flogged the tenth litre of crystalloid into one's patient, one tends to reflect on the utility of further litres.
How does one figure this out?
There are two ways you can go about this.
The Luddite purists will urge you to lay your hands on the patient and make an assessment on the basis of non-invasive bedside manoeuvres and clinical examination findings. The gadget freaks and company representatives will encourage the use of attractively shiny devices which produce complicated hemodynamic indices.
Thus, let us separate the methods of establishing fluid responsiveness into clinical and parameter-driven.
By "clinical" in this context I mean "anything which does not require performing thermodilution measurements, raping the pulmonary artery, PhD-level pulse contour calculus or sophisticated echosonography skills".
In brief summary:
Measures of fluid responsiveness:
|Method||Physiology or rationale||Limitations|
SVV becomes invalid in the following situations:
|Passive leg raise autotransfusion||
In some detail:
Physical signs as predictors of fluid responsiveness
These are findings discovered during your examination of the shocked patient, to help you decide whether they would benefit from another fluid bolus. Certainly, the ICU patient is covered in grossly obvious physical signs, which are their way of signalling to us their physiological needs. Its really a matter of deciding whether any of these signs are saying "I need a bag of crystalloid". Right?
Well, as it turns out we clinicians are completely useless at predicting fluid responsiveness without the aid of some sort of apparatus. And we have some good evidence to support that statement.
A study performed among shocked malaria patients is available; they used GEDVI to define intravascular volume.
I will paraphrase their abstract.
- JVP does not correlated with intravascular volume (unsurprisingly, as most senior clinicians cant agree where the JVP is. The external jugular may be a little more useful, but again - these only tell you about the CVP, and we all know how completely useless this is as a predictor of fluid responsiveness).
- Dry mucous membranes do not correlate well with intravascular volume.
- Poor capillary return DOES correlate with intravascular volume.
- MAP was unrelated to fluid responsiveness.
So, what's the point of performing these examinations?
The presence of features of shock is probably the useful part.
Something about your cold, clammy, hypotensive patient has given you the impression that their tissue perfusion is inadequate. The combination of your physical examination findings and history has resulted in a decision to treat the shock.
Assessment of fluid responsiveness in the absence of technology thus rests on the observation of these features of shock, and on the observation of the changes that take place in response to fluid boluses. An intubated patient will not suffer terribly from small amounts of fluid; you still have PEEP and FiO2 to titrate in case of error. If the features of shock appear to retreat, one might surmise that the stroke volume has improved.
This is an application of the "gold standard" assessment for fluid responsiveness, which is an actual fluid challenge.
Central venous pressure as a predictor of fluid responsiveness
In short, the response of central venous pressure to a fluid challenge was once thought to be predictive of right heart preload. The rationale for this is as follows:
- A patient who is turly or relatively hypovolemic will be expected to have a low CVP
- That patient's CVP should increase in response to fluid challenge
- If the patient remains relatively hypovolemic in spite of the fluid challenge, the change in CVP ("delta CVP") will be relatively small.
- A patient who is "well filled" will have a large increase in their CVP, as the elastic central vessels approach the limits of their stretch.
Some thought that this "delta CVP" might have some sort of predicitve value when it comes to fluid responsiveness. Surely, if you keep dumping fluid into your patient and the CVP fails to change dramatically, that must mean they have plenty of space left - plenty of "central venous reserve" - and thus more fluid boluses are required.
However, it turns out the CVP is really bad at predicting the patients' responsiveness to fluid challenges.
There are too many variables governing central venous pressure; it is never a perfect picture like this graph, where central venous compliance is predictable and constant. This has become evident from some high-quality evidence, and it has been known for some time. Indeed, so obvious the uselessness of CVP in this scenario, and so entrenched the practice of its use, that prominent authors have described a recent meta-analysis as a plea for common sense.
The limitations of CVP for assessment of fluid responsiveness are discussed in greater detail in the chapter on utility of CVP measurement in the ICU. In brief:
- CVP does not always correspond to RA pressure or RV transmural pressure
- The pressure read by the transducer is not the RA filling pressure (measured at the onset of the c-wave)
- CVP does not correlate well with cardiac index (Ishihara et al, 2000)
- CVP does not correlate well with stroke volume index (Michard et al, 2003)
- CVP correlates poorly with ITBVI and LVEDVI. (Diebel et al, 1992)
- Changes in CVP correlate poorly with changes in stroke volume in response to volume loading (Gödje et al, 1998)
- Probably the best summary of these arguments comes from the recent meta-analysis by Marik and Cavallazzi (2013); "This approach to fluid resuscitation should be abandoned", they said.
Pulmonary artry wedge pressure as a predictor of fluid responsiveness
The rationale for the use of PAWP as a measure of fluid responsiveness is as follows:
- In a capillary in Wests' Zone 3, there is continuous column of blood to the LA.
- With an inflated wedge balloon, the flow in that capillary is reduced to zero.
- With flow at zero, the PAWP should be similar to LA pressure
- LA pressure should be similar to left ventricular end-diastolic pressure (LVEDP) which should (ideally) represent LV preload
- Thus, if PAWP is low, the LV preload is also low, and this predicts fluid responsiveness (whereas if it were high, there would be little point in giving more fluid).
The limitations are:
- It is confused by many situations in which the PAWP is not equal to LV end-diastolic pressure:
- It is higher than LVED when there mitral stenosis or regurgitation, left-to-right shunt, COPD, positive pressure ventilation, atrial myxoma, pulmonary venous hypertension or simply poor catheter placement.
- It is lower than the LVEDP when there is LV failure, high PEEP, a poorly compliant LV (eg. in HOCM) or whenever there is aortic regurgitation
In 2004, Kumar et al published an influential paper which laid waste to the concept of PAWP as a predictor of fluid response. Indeed, they found that neither PAWP nor CVP correlated even slightly with end-diastolic volume or stroke volume variation. Nor did PAWP or CVP change appreciably after a fluid bolus (Whereas LVEDI and SVI certainly changed, suggesting that these derived variables probably do mean something). This lack of relationship between PAWP and preload was consistently observed among both critically ill patients and in the normal population.
Stroke volue variation as a predictor of fluid responsiveness
The rationale is as follows:
- The lower on the Frank-Starling Curve you are, the more stroke volume will vary depending on the phase of ventilation.
- Decrease in preload due to mechanical inspiration results in a decrease in ventricular wall stretch
- This results in a decrease in stroke volume
- Thus, patients who have decreased filling are going to have more difference between their inspiration and expiration stroke volumes.
You aim for an SVV of under 10%; any greater variation than this warrants a fluid bolus.
Stroke vlume is difficult to measure directly, but it can be inferred from performing a cardiac output calibration on a pulse-contour cardiac output monitor such as the PiCCO. In this way, the shape of an arterial line pulse waveform can be related to a cardiac output value, and therefore to a measured stroke volume.
We have a 2011 meta-analysis which neatly summarises the results of 23 studies. Overall, SVV was found to be a good predictor of fluid responsiveness, with a sensitivity of 81% and specificity of 80%. The patients in whom this would be most accurate are those being ventilated with large volumes (8ml/kg or higher).
Overall, SVV appears to be a good predictor of fluid responsiveness. The patients in whom this would be most accurate are those being ventilated with large volumes (8ml/kg or higher). This rests on the premise that certain predictable conditions for validity are met, eg. the patient is not breathing spontaneously nor is suffering atrial fibrillation, and so on.
SVV becomes invalid in the following situations:
- spontaneously breathing patient
- cardiac arrhythmia
- valvular heart disease, especially aortic
- cardiogenic shock (with poor LV function)
- intracradiac shunts
- severe peripheral vascular disease
Passive leg raise autotransfusion
According to the French, this manoeuvre is a reliable predictor of fluid responsiveness, irrespective of the fact that the patient might be mechanically ventilated or breathing spontaneously, or having some sort of arrhythmia.
The elegant simplicity of this test rests in the premise that your leg right now contains about 250ml of venous blood, and that to raise both your legs up to 45° will result in an "autotransfusion" of venous blood into the central veins. The bonus is that this is an already body-size adjusted fluid challenge, and it is completely reversible (just drop the legs!)
In the French study the stroke volume change was measured by oesophageal Doppler measurements of aortic blood flow. This is not a bad surrogate of stroke volume, if done correctly. The legs were raised to 45° using the mechanical bed, while the rest of the body remained supine.
So. The method of doing this correctly at the bedside, in order to replicate the study, is as follows.
1) Drop the patient's torso to supine position
2) Raise both legs to 45°
3) Keep them up for 1 minute
So, after 1 minute of this sort of thing, you should be able to confidently expect whatever measure of cardiac output or stroke volume to increase, depending on what you're using. At least as far as oesophageal Doppler is concerned, the French found this technique to have a sensitivity of 97% and a specificity of 94%.
From 2006 onwards, the passive leg raise attracted a considerable amount of attention. So much so that in 2012, there was enough data for a nice meta-analysis. With all the data (from 9 studies) pooled, the sensitivity and specificity ended up at 89% and 91%, respectively. Particularly, an increase in cardiac output measurement (rather than a reduction in pulse pressure variation) correlated better with a positive passive leg raise test.
In short, this manoeuvre is not a bad one.
It has some limitations:
- You need a patient with both legs intact
- You rely on an intact pelvis, so this excludes a lot of messy trauma patients (in whom it would be very useful)
- It can't be done if you have a balloon pump in situ, or post angiography (because you need to lie flat) - and thus a lot of low-cardiac-output cardiogenic shock patients are excluded, which is a pity
- It can't be done if you are even slightly concerned about your intracranial pressure.
Global End-Diastolic Volume (GEDV)
GEDV relates well to preload, but somehow not to fluid responsiveness.
In his 2009 review, Marik et al trashed GEDVI as a means of predicting fluid response - it was no better than CVP or PAWP, which is to say "next to useless". Additionally, GEDV loses its correlation with preload in early sepsis.
So, in short, GEDV and GEDVI should not be used to guide volume resuscitation.
Inferior vena cava ultrasonography
This is another natural extension of understanding the haemodynamic effect of mechanical ventilation.
Basically, one relies on the idea that the right heart filling is the dominant variable affected by the intrathoracic positive pressure, and that the increase in intrathoracic pressure means an increase in the extrathoracic central venous volume. So one can think of the IVC as a blood-filled manometer.
Criticalecho.com has a wonderful resource on practical IVC measurement using M-mode, and any attempt to paraphrase or summarise it would be insulting to the authors. It is perfect as it is. However, for my own reference, I will comment that there does not seem to be any agreement as to where the best spot is to measure the IVC, or indeed whether it matters very much.
The key trick is to put an M-mode Doppler through the IVC, and observe what happens over the course of a ventilator breath. A underfilled IVC will fluctuate in diameter; a well-filled IVC will remain solemnly distended.
In 2004 Feissel et al found that the M-mode measurement of IVC diameter in ventilated patients had a positive predictive value of 92%, which seems pretty good. In general, because these patients were ventilated with mandatory breaths, their breath-to-breath variation was minimal, and the measurements were improved by this. In 2010 Moretti and Pizzi confirmed that among SAH patients, the change in IVC diameter was a reliable way of identifying fluid responders.
Conflicting opinions emerged. People were unimpressed with the lack of consensus of what the threshold for diameter variability was (it varies form 12% to 40% among studies), where to measure it, and how to train the measurer. Most recently, in 2014 a systematic review of the existing studies was forced to conclude that among critically ill patients, point of care IVC diameter measurements are probably very useful.
Guidelines for fluid challenge administration
- Colloid and crystalloid are equivalent in terms of mortality (SAFE study)
- Of the colloids, there is insufficient evidence to recommend one over another in terms of mortality (though albumin may have non-oncotic ancillary effects which may be beneficial in sepsis)
- Of the crystalloids, "balanced" fluids eg. Plasmalyte-148 are associated with improved mortality (at least in sepsis) when compared to isotonic saline.
How much fluid?
- At least in sepsis, perhaps less fluid is better (FEAST study).
- Conventional teaching recommends 500-1000ml of crystalloid, or (in other sources) 10-20ml/kg. This convention may be closely related to the usual fluid bag content.
- Rate of administration may be more important than the amount and type of fluid.
- There is no scientific consensus as to how fast is fas enough.
- Surviving Sepsis people recommend the fluids be given over 30 minutes.
- Another technique is SV maximisation - a process where 250ml boluses are given over 5-10 minutes until stroke volume (as measured by invasive hemodynamic monitoring) stops increasing by 10-15% with each bolus
When to stop?
- Though not based in any firm evidence, resuscitation endpoints have historically included the following parameter theresholds:
- MAP > 65mmHg
- CVP >8mmHg, or a change of over 7mmHg in response to the bolus
- PAOP change of over 5mmHg in response to the bolus
- Normal lactate (<2.0mmol/L)
- Urine output >0.5ml/kg/hr
- ScvO2~ 75mmHg
- Resolution of clinical features of hypovolemia which had given rise to the decision to administer the fluid bolus.