Our primary concern in the management of the critically ill patient is the optimisation of tissue oxygen delivery. Insufficient intravascular loading in the early resuscitation of acute sepsis results in tissue underperfusion, organ dysfunction, and acidosis. Excessive fluid administration has also been shown to be detrimental in the perioperative setting and in acute lung injury, prolonging both time on mechanical ventilation and time in intensive care [
It has been reported that as few as 40 percent of critically ill patients thought to be intravascularly deplete gain an improvement in cardiac output after a standard fluid bolus, exposing more than half of patients to the risks of excessive fluid administration [
Knowledge of static measures of preload such as central venous pressure, pulmonary artery wedge pressure, end-diastolic volumes, and intrathoracic blood volume has not translated into patient benefit [
Contemporary investigation has therefore focussed on the search for clinical markers which predict a useful response to a fluid bolus. These “dynamic” markers make use of provoked cardiac reaction assessed without the need for a fluid bolus, instead utilizing either the consequences of heart-lung interaction during ventilation or the response to postural change to mimic the effect of a fluid bolus on stroke volume.
Firstly, in mechanically ventilated patients who have no spontaneous respiratory effort, the change in intrathoracic pressure has a cyclical effect on both the left and right heart (as shown in Figure
The physiological explanation for the changes in stroke volume and IVC diameter caused by mechanical ventilation. RV right ventricle, LV left ventricle,
Secondly, in the spontaneously ventilating subject, negative intrapleural pressure during inspiration results in a reduction in the diameter of the abdominal inferior vena cava (IVC). The degree of collapse during tidal volume breaths is known to reflect the right atrial pressure with reasonable accuracy in health [
Thirdly, raising the legs from the horizontal position to 45 degrees causes the gravitational movement of lower limb venous blood towards the heart. This provides a transient volume load of between 150 and 300 millilitres to the central circulation, lasting for a few minutes [
Modern intensive care is increasingly concerned with the avoidance of unnecessary invasive procedures which contribute to patient morbidity either directly or more often through the associated risk of catheter-related bloodstream infection [
Transoesophageal echocardiography may provide superior image quality in some cases and is increasingly utilised for cardiovascular monitoring on intensive care units. Nonetheless, it requires equipment, time, and skills that are less abundant on many intensive care units. It is contraindicated in some patients with upper airway or oesophageal surgery and also usually necessitates sedation which is not always achieved without adverse consequence.
Accordingly, the objective of this review is to systematically evaluate the literature examining the use of transthoracic echocardiography in the assessment of dynamic markers of preload used to predict fluid responsiveness in the critically ill patient.
An electronic literature search was carried out using Medline, EMBASE, CINAHL, and the Cochrane database of systematic reviews. The search terms used were ((fluid) OR (volume) OR (preload) OR (filling)) AND ((respons*) OR (status) OR (assess*)) AND ((echocardiograph*) OR (echog*)). The search was limited to “human” and “English language.” Figure
Citation filtering process.
Transoesophageal echocardiography studies were excluded, as were those in which the sample group, the equipment used and the reference test cut-off criteria implied the conclusions were not applicable to the critically ill patient in a high dependency of critical care environment.
“Fluid responsiveness” refers to a predefined rise in stroke volume or cardiac output after rapid fluid loading with a predetermined volume of fluid. Between investigators, there are inevitable differences in the choice of stroke volume or cardiac output, the volume of fluid given, the duration over which the fluid load was given, and the type of fluid given.
The diagnostic test or equivalent of the index test in these studies is defined as the echocardiographic test done to give a prediction of fluid responsiveness. In this review, this test will be termed the “predictive test.” The test done to assess the response to a fluid bolus once given is similar to a reference test but for the purposes of this review will be called the “response test.”
Responders are those patients in whom the cardiac output or stroke volume rises by the threshold amount after a given bolus of fluid.
The Standards for Reporting of Diagnostic Accuracy (STARD) initiative developed a guide for assessing the quality of reporting of studies of diagnostic accuracy [
Modified STARD criteria assessment [
Criteria | Specific question |
---|---|
1 | Was the study population described (inclusion and exclusion criteria included)? |
2 | Is there a description of the sampling (e.g., consecutive patients, if not why not?)? |
3 | Is it clear whether the tests were done prospectively or retrospectively? |
4 | Is there a description of the response test (including fluid bolus)? |
5 | Is there a detailed description of the equipment and techniques used in the tests? |
6 | Is the rationale for cut-offs and ranges given? |
7 | Is there detail of the operators in terms of number and training? |
8 | Is there detail of what information was available to the readers of the response ? |
9 | Were the statistical methods for comparing diagnostic accuracy detailed? |
10 | Are there details of tests of reproducibility? |
11 | Are the patient demographics and comorbidities shown? |
12 | Is there detail of those meeting inclusion criteria but not undergoing either test? |
13 | Was there detail of the interval between predictive and response tests? |
14 | Is there a report cross-tabulating predictive and response test results? |
15 | Is diagnostic accuracy described, including likelihood ratios or data to calculate them? |
16 | Is there mention of how missing values were dealt with (i.e., unobtainable values)? |
17 | Are the estimates of accuracy variability between operators/readers included? |
18 | Are there estimates of reproducibility? |
19 | Is the clinical applicability of the study findings discussed? |
The results of the selected studies were not meta-analysed due to the heterogeneity of methodologies, as well as the differences in patient selection, modes of ventilation and definition of fluid response. There was insufficient data for the construction of summary receiver-operator characteristic (SROC) curves or for the calculation of
Of the 3138 articles identified through the search terms, eight studies were included for review
Characteristics of studies selected.
Study | Technique | Patient group | Selection | Ventilation | Rhythm | Volume and type | Time (min) | Response criteria |
---|---|---|---|---|---|---|---|---|
Barbier et al. [ | IVC DI | Mixed ICU | Shock (sepsis) and acute lung injury | All mand | Any | 7 mL/kg colloid | 30 | >15% CO TTE |
Feissel et al. [ | Medical ICU | Shock (sepsis) | All mand | Any | 8 mL/kg colloid | 20 | >15% CO TTE | |
Lamia et al. [ | PLR | Medical ICU | Shock (sepsis or hypovolaemia) | All spont | Regular SR, or AF | 500 mL crystalloid | 15 | >15% SV TTE |
Maizel et al. [ | PLR | Mixed ICU | Shock (unspecified) | All spont | Regular SR | 500 mL crystalloid | 15 | >12% CO TTE |
Biais et al. [ | PLR | Surgical ICU | Shock (sepsis or haemorrhage) | All spont | Any | 500 crystalloid | 15 | >15% SV TTE |
Biais wt al. [ | SVV | Surgical ICU | Post-operative (liver surgery) | All mand | Regular SR | 20 mL/kg/m2 colloid | 20 | >15% CO TTE |
Thiel et al. [ | PLR | Medical ICU | Shock (unspecified) | Mixed | Any | 500 mL crystalloid or colloid | Unspec | >15% SV TTE |
Préau et al. [ | PLR | Medical ICU | Shock (sepsis or acute pancreatitis) | All spont | Regular SR | 500 mL colloid | <30 | >15% SV TTE |
Selection: inclusion criteria summary, PLR: passive leg raising, spont: spontaneous respiratory effort whether or not on mechanical ventilation, mand: ventilator giving mandatory breaths only and patient fully adapted to ventilator, SR: sinus rhythm, AF: atrial fibrillation, TTE: transthoracic echocardiography, SV: stroke volume, CO: cardiac output,
Collated results of all included studies.
Study | Number of tests | Predictive test | Threshold | Resp % | Intra-obs % | Inter-obs % | AUC (ROC) | Sens | Spec | PLiR | NLiR | PPV | NPV | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Lamia et al. [ | 24 | PLR SVI or CO rise | ≥12.5% | 54 | 77 | 99 | 77 | 0.23 | 0.79 | |||||
Maizel et al. [ | 34 | PLR CO rise | ≥12% | 50 | 63 | 89 | 5.73 | 0.42 | 85 | 76 | 0.75 | |||
PLR SV rise | ≥12% | 69 | 89 | 6.27 | 0.35 | 83 | 73 | 0.57 | ||||||
Biais et al. [ | 34 | PLR SV rise | ≥13% | 67 | SI | 100 | 80 | 5.00 | 0.00 | |||||
Thiel et al. [ | 102 | PLR SV rise | ≥15% | 46 | SI | 81 | 93 | 11.57 | 0.20 | 91 | 85 | |||
Préau et al. [ | 34 | PLR SV rise | ≥10% | 41 | SI | 86 | 90 | 8.60 | 0.16 | 86 | 90 | 0.74 | ||
PLR dVF rise | ≥8% | 86 | 80 | 4.30 | 0.18 | 75 | 89 | 0.58 | ||||||
Biais et al. [ | 30 | SVV | ≥9% | 47 | SI | 0.95 | 100 | 88 | 8.33 | 0.00 | 0.80 | |||
Barbier et al. [ | 23 | IVC DI | ≥18% | 41 | 90 | 90 | 9.00 | 0.11 | 0.90 | |||||
Feissel et al. [ | 39 | ≥12% | 41 | SI | 93 | 92 | 0.82 |
Threshold: cut-off between responders and nonresponders, Resp: proportion responding to fluid load, Intra-obs: intraobserver variability, Inter-obs: interobserver variability, AUC(ROC): area under the receiver-operator curve, Sens: Sensitivity, Spec: Specificity, PLiR: positive likelihood ratio, NLiR: negative likelihood ratio, PPV: positive predictive value, NPV: negative predictive value,
Five studies used transaortic stroke volume variation to predict fluid responsiveness using passive leg raising to mimic a fluid bolus [
Important differences between studies were evident in the study protocols. Maizel et al. [
The stages of the two different methods of passive leg raising. CO cardiac output, SV stroke volume. aMeasurements at this stage were not taken in one study (Maizel).
All studies showed good sensitivity (77 to 100 percent) and specificity (88 to 99 percent) using a threshold of 10 to 15 percent increment of stroke volume or cardiac output.
Strikingly, stroke volume change with PLR predicted the correct response to volume expansion in 16 of the 18 patients with arrhythmia [
A single study by Biais et al. looked at the use of stroke volume variation for prediction of fluid responsiveness [
The area under the receiver operator characteristic (ROC) curve was used to ascertain a threshold of nine percent stroke volume variation as being the most useful for discerning responders from nonresponders. Using this cut-off, there was excellent sensitivity and specificity (100 and 88 percent, resp.).
Two studies by Barbier et al. and Feissel et al. used respiratory variation of the diameter of the IVC to predict fluid responsiveness [
Barbier et al. used a “distensibility index” calculated by
This review shows that TTE is a highly discriminative test for the prediction of the stroke volume or cardiac output response to volume loading in critically ill patients, thus highlighting the potential for expansion of its role in quantitative assessment.
Importantly, TTE techniques appear useful in patients with spontaneous respiratory effort and those with arrhythmias: this is in contrast to many of the techniques that involve invasive monitoring which have been shown to be inaccurate in these situations [
Although TTE does not provide continuous monitoring which can be managed by nursing staff at the bedside, in reality, most clinical questions regarding fluid management arise intermittently. With equipment close at hand the time taken for a focussed TTE assessment rarely takes more than few minutes [
The techniques of IVC diameter assessment, transaortic stroke volume variability with respiration and stroke volume increment with passive leg raising all provided strong predictive ability for response to a fluid bolus. The area under ROC curves was greater than 0.9 in all articles that presented the statistic. Although a clear threshold value for discriminating responders from nonresponders seems intuitively advantageous, clinicians are adept at coping with non-discriminatory results and using them to inform decisions made on the basis of the whole clinical picture.
None of the three TTE techniques is convincingly the best and if possible all three should be used to minimize the impact of their limitations. On occasion, this may not be achievable for a number of reasons. Local pain or delirium may preclude all or part of a TTE exam in a small minority of cases. In the 260 scans attempted within the studies selected, just 13 could not be performed for these reasons making this a well-tolerated procedure in the main. Thoracic or abdominal wounds may sometimes make views impossible to achieve. Obesity or rib prominence can also make TTE acoustic windows difficult to obtain but it is rare that at least a single usable view cannot be obtained in an individual. In the reviewed studies, only nine of the 260 attempted scans were abandoned due to difficulty with anatomy. Additionally, the applicable techniques will depend on the presence or absence of mechanical ventilation or dysrhythmias. For example, in a patient with atrial fibrillation who is fully ventilated, transaortic Doppler assessment is inaccurate but subcostal measurement of the IVC diameter variation can be safely used.
The concept of “wet” and “dry” intensive care units has long been debated. The apparent benefits of goal-directed aggressive fluid resuscitation in the early stages of sepsis must be balanced with evidence for reduced morbidity when “restrictive” fluid regimes are used [
It is important to recognize that this review neither allows assumptions about the longevity of the response to fluid, nor the value of a continuous fluid infusion thereafter. It also follows that a forecast suggesting the patient will be fluid responsive in no way guarantees the safety of a delivered bolus in terms of increasing extravascular lung water or worsening regional organ oedema and function.
The literature contains a growing body of work on optimising haemodynamics using other echocardiographic parameters, beyond simple measures of contractility and structural pathology. Patterns of flow across the mitral valve and tissue velocity of the annulus have proved useful, principally when assessed in combination. Tissue velocity, particularly that measured close to the mitral valve annulus, assessed using Doppler imaging (TDI) provides an accurate estimation of diastolic function of the left ventricle irrespective of preload changes [
Although detailed examination of the heart requires an experienced echocardiography practitioner, there is an increasing acceptance of the value of focussed echocardiographic assessments to answer common clinical questions arising in critical illness. This has arisen in tandem with the emergence of a number of courses and training programmes centred on evaluation of the critically ill patient by those less experienced in echocardiography. Jensen showed that with only limited training, a diagnostic transthoracic window was achieved 97 percent of the time when used in the evaluation of shock [
This review was restricted to the specific question of fluid response. In reality, echocardiographic assessment of the critically ill aims to gain as complete a picture as possible of the cardiovascular state. Ideally, this should also involve a full structural study in addition to inspection of left ventricular filling state and perhaps even ultrasonic examination of the lungs.
Furthermore, studies using transoesophageal echocardiography (TOE) were not selected for this review and, although it would seem intuitive that flow or diameter measurements techniques taken with one kind of echocardiography could be safely extrapolated to another, this ignores the differing technical restrictions of each technique. Transoesophageal echocardiography has its own growing evidence base for its application in intensive care and clearly where it is available provides invaluable haemodynamic information to inform clinical decisions.
A significant limitation of this review is the small size of the study groups since only a single study included more than 40 patients [
It is conspicuous that only one article reported on the time between the initial predictive test and the subsequent assessment of a response to a fluid bolus [
The amount of fluid used, the type used, and the rate at which it was given all impact upon the response test in these studies. Unfortunately, there is no agreed formulation for a standard fluid load although almost all studies use approximately the same formulation.
Although no specific details were given about the qualifications of the echocardiography operator or reader most studies inferred they were experienced. Furthermore, blinding of the operator or reader, to the measurements taken after volume loading was rare and this is, therefore, a source of observer bias within the data.
Intraobserver variability was considered by the majority of studies and attempts were made to measure it with variable success. An intuitively more useful measurement of reproducibility was achieved by examining the variability of repeated measurements of distensibility by Feissel et al. [
Of note, whilst the effects of varying tidal volumes on echocardiographic parameter assessment are minimal, the impact of raised intra-abdominal pressure and of different positive end expiratory pressure is largely unstudied [
The clinical question that was not addressed in any of the articles was that of the “real-world” value of echocardiographic approaches to assessing fluid responsiveness. The studies reviewed do not provide us with information about translation into effects on morbidity or mortality, nor is there yet such a current evidence base in the literature. This evidence may well originate in the context of future investigation into the dilemma of conservative versus liberal fluid management.
Transpulmonary microsphere contrast has already been shown to dramatically improve volumetric assessment and its use in the critically ill would intuitively improve the clinical utility of the modality still further [
Transthoracic echocardiography is becoming a powerful noninvasive tool in the daily care of the critically ill. This review brings together the evidence for employing TTE to predict fluid responsiveness. Assuming there is equipment and local expertise TTE is a repeatable and reliable method of predicting volume responsiveness in the critically ill.
Transaortic stroke volume variation with the respiratory cycle, stroke volume difference following passive leg raising, and IVC diameter changes with respiration all provide good prediction of the likelihood of a response to a fluid bolus. The techniques can be used individually to address the needs of different patients and in combination to triangulate clinical information where uncertainties may occur.
The studies reviewed form a robust platform of physiological data on which to base further studies involving larger numbers of patients which engage with clinically relevant outcomes, such as inotrope use, blood pressure, length of stay, and time to weaning from mechanical ventilation.
Improved access to clinician-echocardiographers through a defined training process will facilitate such clinical studies and give patients access to accurate noninvasive information in answer to the daily clinical conundrum of fluid responsiveness.