Echocardiography-guided fluid resuscitation in critical illness: from fluid responsiveness to fluid tolerance

Background

Fluid resuscitation is among the most frequent therapeutic interventions in the intensive care unit, yet it is also one of the easiest to overuse. The traditional assumption that hypotension, oliguria, or a low central venous pressure should prompt more intravenous fluid has been repeatedly challenged (4). Only a proportion of critically ill patients are preload responsive at any given time, and even when stroke volume rises transiently, this does not guarantee that additional fluid is biologically desirable or clinically safe (1, 5–7).

This distinction has become increasingly important as critical care has shifted from early liberal resuscitation towards a more nuanced strategy that considers the phase of shock, the mechanism of circulatory failure, and the risk of fluid accumulation. Recent work has also emphasised that a patient may be both fluid responsive and already congested, meaning that the mere ability to increase cardiac output does not automatically justify fluid administration (2, 3, 19–21).

Echocardiography is uniquely positioned in this context because it can answer several clinically relevant questions at the bedside within minutes. It can identify pericardial tamponade, severe left or right ventricular dysfunction, acute cor pulmonale, major valvular disease, and septic cardiomyopathy; estimate whether further fluid is likely to augment stroke volume; and help detect signs that the patient is unlikely to tolerate more fluid (1, 2, 10, 18, 19). An important limitation must, however, be stated from the outset: echocardiography enables effective haemodynamic monitoring only when assessments are performed serially and frequently. A single snapshot in time is insufficient to guide dynamic fluid management in critically ill patients who may deteriorate or improve rapidly. Another important limitation is that echocardiography cannot replace continuous cardiac output monitoring, which remains warranted, and is recommended by clinical guidelines, in the most severe or haemodynamically complex patients (3). This is partly because passive leg raising appears to predict fluid responsiveness more accurately when the induced haemodynamic response is tracked continuously, for example with pulse-contour cardiac output monitoring calibrated by transpulmonary thermodilution, rather than assessed using a single intermittent echocardiographic measurement (12, 27). This does not diminish the value of echocardiography, but rather defines its role. The two approaches are complementary rather than interchangeable: continuous cardiac output monitoring provides high-frequency haemodynamic trends, whereas echocardiography provides direct anatomical and functional information, including valvular regurgitation, right ventricular function, pericardial tamponade and venous congestion patterns such as the Venous Excess Ultrasound score (VExUS), which cannot be obtained from pulse-contour or thermodilution devices alone (1, 2, 10, 18, 22).

To frame this narrative review, we performed a structured PubMed search covering January 2000 to January 2026 and hand-searched major review articles, guidelines, and reference lists. We focused on adult ICU, perioperative, and emergency-care studies evaluating echocardiographic indices or echocardiography-assisted functional tests for predicting fluid responsiveness, while also incorporating more recent literature on venous congestion and fluid tolerance published in 2024–2026. Because of heterogeneity in populations, thresholds, reference standards, and outcome definitions, we present a critical narrative synthesis rather than a PRISMA-based systematic review. Where possible, each assertion is supported by a reference to the original research article from which it derives, rather than by review articles alone. A practical echocardiography-guided framework integrating fluid responsiveness and fluid tolerance is presented in Figure 1.

Main text

Static echocardiographic variables: useful for phenotype, poor for prediction

Static echocardiographic measurements remain clinically valuable, but mostly because they describe the haemodynamic context rather than predict the effect of a future fluid bolus. A small, hyperdynamic left ventricle may suggest fluid responsiveness, whereas a dilated, poorly contractile ventricle should raise concern about fluid intolerance. Similarly, marked right ventricular enlargement or septal flattening should immediately prompt caution, because worsening RV distension during fluid loading may further reduce left-sided output (1, 2, 10, 18).

It is worth emphasising that the ability of echocardiography to distinguish the causes of haemodynamic instability, differentiating hypovolaemia from vasodilation, myocardial dysfunction, or right-sided limitation, is universally useful and applies across all clinical settings, not only those discussed in a later section of this review.

Inferior vena cava (IVC) diameter and collapsibility are widely used at the bedside because they are rapidly obtainable. However, IVC behaviour is influenced by spontaneous inspiratory effort, tidal volume, positive end-expiratory pressure, abdominal pressure, obesity, and probe position. As a consequence, reported diagnostic performance of IVC metrics varies widely across studies and ventilation conditions, and these indices should not be used in isolation to decide on fluid administration (5, 8, 10). The role of the IVC in assessing fluid tolerance, rather than fluid responsiveness, should be considered separately. In the VExUS grading system, an IVC diameter greater than 2.0 cm is required to diagnose significant venous congestion (22). However, an IVC diameter below this threshold does not fully exclude the risk of fluid-induced harm, particularly in patients with altered vascular permeability such as sepsis, ARDS, or burns, in whom venous congestion and interstitial oedema can develop despite a non-dilated cava (21).

The superior vena cava (SVC) has been proposed as an alternative to the IVC in mechanically ventilated patients. Its diameter and respiratory variation are less affected by intra-abdominal pressure and probe compression than IVC-based indices, which may offer a practical advantage. However, SVC assessment typically requires a transoesophageal approach, limiting its routine bedside applicability.

Diastolic indices, especially E/e' and left atrial size, are better viewed as markers of filling pressure and fluid tolerance than as predictors of fluid responsiveness (9, 18). Several observational studies have shown that an elevated E/e’ ratio is associated with higher left-sided filling pressures and carries prognostic implications for fluid tolerance (9), although its ability to predict fluid responsiveness per se has not been established in original research. In practice, these indices can help identify patients in whom a further volume challenge carries a higher risk of pulmonary congestion, particularly when interpreted together with lung ultrasound findings and the broader clinical context including the patient's clinical presentation, biological markers, and haemodynamic trajectory (2, 9, 18).

Dynamic assessment: LVOT VTI as the core echocardiographic signal

The main strength of echocardiography in fluid management lies in the ability to track changes in stroke volume surrogates during a controlled preload-modifying manoeuvre. Pulsed-wave Doppler measurement of the left ventricular outflow tract velocity-time integral (LVOT VTI) is the most practical and widely used approach. Although absolute stroke volume calculation requires accurate LVOT diameter measurement, relative changes in VTI can be obtained rapidly and are more reproducible for serial decision-making (11–14). An important caveat, however, concerns the accuracy of this measurement. Subaortic VTI is susceptible to variability in angle correction, sample volume positioning, and signal quality; small measurement errors can meaningfully affect the magnitude of the calculated change in cardiac output during a fluid responsiveness test. This limitation is particularly relevant when using LVOT VTI to detect modest changes in stroke volume, and threshold values must therefore be interpreted with this measurement uncertainty in mind (14). This limitation is best formalised through the concept of the least significant change (LSC), defined as the smallest change in a measurement that exceeds inherent measurement variability and can therefore be interpreted as a true haemodynamic change rather than random error. Studies estimate the LSC of TTE-derived LVOT VTI at approximately 10%–15% (14), which substantially overlaps with the diagnostic thresholds commonly used for the end-expiratory occlusion test and the mini-fluid challenge, often around 5%–10% for changes in cardiac output or VTI.

The diagnostic performance, practical strengths and limitations, and bedside integration of these approaches are summarized in Tables 1–3.

Table 1. Typical diagnostic performance of commonly used echocardiographic approaches for predicting fluid responsiveness.

AUROC, area under the receiver operating characteristic curve; CI, confidence interval; EEOT, end-expiratory occlusion test; IVC, inferior vena cava; LV, left ventricle; LVOT VTI, left ventricular outflow tract velocity-time integral; NA, not applicable; PLR, passive leg raising. Point estimates are taken from the key reference cited for each row. Because the corresponding studies do not report 95% confidence intervals in a fully consistent manner, readers should consult the original articles for the confidence interval around each estimate before applying a specific threshold at the bedside. Diagnostic thresholds are specified in the Method column for each test.

Table 2. Practical strengths and limitations of the main echocardiography-assisted functional tests.

Table 3. Examples of bedside questions that echocardiography can help answer before another fluid bolus is prescribed.

Accordingly, some changes classified as diagnostic by these tests may approach, or even fall within, the measurement noise of TTE-derived VTI. This should temper the interpretation of studies reporting very high diagnostic accuracy and supports cautious bedside use, ideally based on averaged, repeated measurements. For the same reason, transoesophageal echocardiography may be preferable in selected settings where the expected change in VTI is small, such as during EEOT or mini-fluid challenge, because its greater image stability and reproducibility may offer a practical advantage. This potential benefit must, however, be balanced against the invasiveness of TEE.

In controlled mechanical ventilation, respiratory variations in LVOT VTI can predict fluid responsiveness under specific conditions, especially when tidal volume is not excessively low and spontaneous breathing efforts are absent. However, their usefulness decreases with lung-protective ventilation, arrhythmias, right ventricular dysfunction, or marked alterations in thoraco-pulmonary compliance. These limitations explain why contemporary practice increasingly favours functional tests over reliance on passive respiratory variation alone (10, 15–18).

Functional tests integrated with echocardiography

Before considering any functional test for fluid responsiveness, three important conceptual points deserve emphasis. First, fluid administration should only be considered when clinical or biological indicators, such as haemodynamic instability, signs of organ hypoperfusion, or an elevated lactate, suggest a potential benefit from increasing cardiac output. Fluid responsiveness testing is not warranted in the absence of such indicators, even when the patient appears haemodynamically borderline. Second, fluid responsiveness is not a binary positive-or-negative state. All diagnostic studies have used a threshold to calculate sensitivity and specificity, but physiologically, responsiveness is a continuous variable: a patient can be highly responsive, marginally responsive, or not responsive at all, and the clinical interpretation must account for the magnitude of the response, not only its direction relative to a cut-off. Third, testing for fluid responsiveness is not exclusively a tool to guide fluid administration; it has also been proposed as a means of assessing the safety of fluid removal, a clearly non-responsive state supporting de-resuscitation strategies (2, 3, 19–21).

Passive leg raising (PLR) remains the best validated bedside test because it produces a reversible endogenous fluid challenge without administering intravenous fluid. When PLR is paired with real-time LVOT VTI monitoring, it provides high diagnostic accuracy, is applicable in spontaneous breathing and controlled ventilation, and remains interpretable in arrhythmias. For many clinicians, PLR plus LVOT VTI is the preferred echocardiographic test when technically feasible (12–15, 18). Several procedural details are essential for reliable bedside implementation. PLR is contraindicated or of uncertain reliability in patients with intra-abdominal hypertension, elevated intracranial pressure, or morbid obesity, in whom the manoeuvre may fail to generate an effective preload challenge or may be poorly tolerated. These situations should therefore be regarded as relative contraindications (12, 29). The response should be assessed using VTI rather than heart rate or blood pressure, as the latter are insensitive and non-specific markers of preload-induced changes in stroke volume. They may remain unchanged despite a clinically meaningful increase in cardiac output, or change for reasons unrelated to preload. At baseline and during PLR, VTI should be averaged over 3–5 cardiac cycles in patients in sinus rhythm, and over a larger number of cycles, up to 10–12, in those with arrhythmias, to reduce beat-to-beat variability.

The commonly used diagnostic threshold is an increase in VTI of at least 10%–12% within one minute of leg elevation, although some studies have used a 15% cut-off. Reported 95% confidence intervals around these thresholds vary across studies, which should be considered when applying a single cut-off at the bedside (12–14).

Finally, the technique used to perform PLR strongly influences the magnitude of the haemodynamic response. PLR should be performed by moving the patient from the semi-recumbent to the supine position while simultaneously elevating the legs, preferably by adjusting the bed mechanism rather than manually lifting the legs. The key requirements are a complete postural change, minimal stimulation of the patient, and real-time measurement of the maximal haemodynamic response. Clinicians should therefore follow detailed technical guidance when using PLR to guide fluid decisions (29).

Mini-fluid challenge is a pragmatic alternative when PLR is impractical or contraindicated. A small bolus of 100–150 mL as described in the original studies validating this approach is administered while the operator immediately reassesses LVOT VTI (13–15). The main drawback is that some fluid is still given to non-responders, and repeated testing may cumulatively contribute to overload. Nevertheless, this method can be clinically useful when decision-making is time-sensitive and acoustic windows are adequate (13–15). Like the PLR test, the mini-fluid challenge can be used in spontaneously breathing patients as well as in those receiving mechanical ventilation, which extends its applicability considerably (15).

The end-expiratory occlusion test (EEOT) is particularly attractive in invasively ventilated patients receiving protective ventilation, because it transiently increases cardiac preload, specifically right ventricular filling and right atrial pressure, without delivering fluid. The term “venous return” is sometimes used in this context but is physiologically imprecise; venous return equals cardiac output by definition, and what the EEOT actually increases is cardiac preload, not venous return per se. The test's main limitations are that it requires controlled mechanical ventilation without significant spontaneous effort, close coordination at the bedside, and real-time Doppler acquisition during a short time window (15, 16, 18).

From fluid responsiveness to fluid tolerance

A major conceptual advance in recent years is the recognition that haemodynamic efficacy and biological tolerance are not equivalent. A patient may still show an increase in stroke volume after PLR or a mini-fluid challenge while simultaneously exhibiting worsening B-lines, venous Doppler abnormalities, right ventricular overload, or elevated filling pressures. This coexistence of responsiveness and congestion, if present, would be clinically important, because it identifies the very patients in whom the risk-benefit balance of a fluid bolus is most uncertain. However, the supporting literature remains limited and should be interpreted cautiously rather than presented as an established finding.

Among the studies most often cited in this context, one is a position paper rather than an original clinical study (19). Another is a prospective study in which fluid responsiveness was defined by a positive preload-responsiveness test, rather than by the effect of an actual fluid bolus. In that study, a positive venous congestion score, assessed using VExUS, was observed in both preload-responsive and preload-unresponsive patients, although scores were higher in the latter group; the clinical implications therefore remain uncertain (20). A further study tempers the concept even more directly, as it found that fluid challenge did not worsen venous congestion in most fluid responders, suggesting that the coexistence of fluid responsiveness and congestion may be less common than often assumed (21).

Another study is particularly informative. In preload-responsive and fluid-tolerant patients undergoing cardiac surgery, a substantial proportion developed transient venous congestion within minutes of a fluid bolus, and this was associated with worse renal outcomes. Notably, this transient congestion could be predicted before fluid administration by the change in portal vein pulsatility index induced by passive leg raising (28). Taken together, these data suggest that the coexistence of fluid responsiveness and congestion is biologically plausible and clinically relevant. However, its true prevalence, its dependence on the definitions used, and its implications for bedside fluid decision-making remain uncertain and require dedicated prospective validation.

Accordingly, echocardiography-guided fluid management should be paired with markers of fluid tolerance. Lung ultrasound can rapidly detect interstitial oedema and is especially useful when repeated over time. Right ventricular size, septal motion, and tricuspid regurgitation patterns provide clues to pressure or volume overload. From a pathophysiological standpoint, the right ventricle is often the first chamber to reflect the consequences of excessive volume loading or increased afterload. The clinical meaning of RV dilatation, however, depends on its time course. Acute RV dilatation, caused by a sudden rise in afterload, acute volume overload, pulmonary embolism, ARDS-related pulmonary hypertension, or inappropriate ventilator settings, may indicate poor fluid tolerance and should prompt caution. By contrast, chronic RV enlargement, as seen in chronic cor pulmonale, may represent a baseline adaptive state and does not necessarily indicate current fluid intolerance. Distinguishing acute from chronic RV dilatation is therefore important, using prior imaging when available, together with assessment of RV wall thickness, contractility and septal motion.

Dynamic changes in diastolic parameters during preload-modifying manoeuvres may provide further information. Variations in E/e', e' velocity, or the E/A ratio during PLR or a fluid challenge, sometimes termed a “load challenge”, may provide insight into dynamic fluid tolerance, although this approach remains insufficiently studied and represents an important direction for future research. E/e' and left atrial enlargement help contextualise left-sided filling pressures, while VExUS or other venous Doppler signals offer complementary information about systemic venous congestion (2, 18–22).

It is also useful to distinguish left-sided from right-sided fluid intolerance, as these reflect different pathophysiological states with different management implications. Left-sided intolerance is primarily associated with pulmonary congestion, reflected by elevated left atrial pressure, an increased E/e' ratio, and B-lines on lung ultrasound. Right-sided intolerance relates more to systemic venous congestion and organ oedema, manifested by RV dilatation, septal dyskinesia, abnormal hepatic vein flow, and VExUS abnormalities. Although these phenotypes frequently coexist, this distinction can help clinicians interpret echocardiographic and ultrasound findings more discriminatively at the bedside.

The bedside decision is therefore not simply whether stroke volume can rise, but whether the expected haemodynamic benefit outweighs the risk of pulmonary or systemic congestion. This framework aligns with recent discussions advocating stepwise or personalised fluid management rather than automatic repetition of fluid boluses during shock resuscitation (2, 3, 18–21). Assessment of fluid tolerance is inherently multimodal and should not rely on cardiac ultrasound alone. It requires integration of lung ultrasound, venous Doppler or VExUS, and the broader clinical and biological context. Particular caution is warranted in patients at high risk of fluid intolerance, including those with pre-existing heart failure, whether with preserved or reduced ejection fraction, chronic kidney disease, ARDS, intra-abdominal hypertension, severe capillary leak syndromes, or high-dose vasopressor requirements. In these patients, even a technically positive fluid-responsiveness test should be interpreted cautiously before fluid is administered.

A further important limitation is that the clinical value of echocardiography depends on repeated, serial assessment throughout the patient's course. A single examination only captures one point in time, whereas the haemodynamic status of a critically ill patient evolves rapidly. Echocardiography should therefore be integrated into a structured monitoring strategy rather than used as a one-off decision tool.

Clinical settings

While echocardiography's ability to distinguish the causes of haemodynamic instability, whether hypovolaemia, vasodilation, myocardial dysfunction, or right-sided limitation, applies universally across all critically ill patients and clinical contexts, certain specific settings warrant additional discussion.

In septic shock, echocardiography should not be reduced to a yes/no fluid responsiveness test. It first helps exclude alternative or additional mechanisms of shock and identify septic cardiomyopathy or significant right ventricular dysfunction. After the initial emergency phase, repeated echocardiography can help determine when vasopressors, inotropes, or de-resuscitation should replace further fluid administration (3, 18, 23).

In perioperative and post-operative care, echocardiography is particularly useful when the cause of hypotension is unclear. It can distinguish hypovolaemia from vasodilation, myocardial dysfunction, or right-sided limitation. Meta-analytic data support goal-directed strategies incorporating echocardiographic stroke volume assessment, although the benefit likely reflects the entire protocol rather than echocardiography alone (24).

Trauma, haemorrhagic shock, and patients with acute respiratory distress syndrome require additional nuance. In trauma, the imperative to control bleeding and deliver blood products may supersede repeated functional testing, but echocardiography remains valuable for detecting RV strain, tamponade, or unexpected cardiogenic components. In ARDS or severe RV afterload mismatch, even a fluid-responsive state may not justify fluid expansion if right ventricular failure, lung water, or venous congestion are already prominent (16, 18, 19, 21).

A practical bedside approach

The following sequence is proposed as a practical framework for bedside decision-making. It reflects the authors' physiological interpretation of the available evidence discussed above, rather than a validated or mandatory protocol, and has not been evaluated against patient-centred outcomes. At the bedside, a pragmatic sequence of this kind may be more useful than reliance on any single variable. First, identify the dominant mechanism of shock and rule out major structural or obstructive causes. Second, select the most appropriate functional test for the patient's condition, with PLR plus LVOT VTI as the default option when feasible. Third, simultaneously assess fluid tolerance using lung ultrasound, right ventricular findings, filling-pressure surrogates, and venous congestion markers. When fluid administration is not appropriate because of fluid intolerance or a non-responsive state, other methods exist to restore haemodynamic stability: these include optimising vasopressor dosing and adding inotropic support. Reducing positive end-expiratory pressure or lightening sedation to permit spontaneous respiratory effort may occasionally be considered, but these measures are highly context-dependent and may be harmful in patients with ARDS, patient-ventilator dyssynchrony, an excessive respiratory drive, or increased oxygen consumption. Lightening sedation to restore spontaneous effort can precipitate patient self-inflicted lung injury through vigorous inspiratory effort (31) and pendelluft (33), particularly when lung injury is already present; because a higher PEEP renders spontaneous effort less injurious, reducing PEEP at the same time as allowing spontaneous effort may compound rather than mitigate this risk, even though a lower PEEP can independently improve venous return and right ventricular afterload (32). These interventions therefore require simultaneous evaluation of their respiratory and haemodynamic consequences and should not be regarded as routine strategies for managing fluid intolerance. Finally, re-evaluate after each intervention rather than assuming persistent benefit from repeated boluses (2, 3, 18–21). We reiterate that this sequence is intended as a framework for organising bedside reasoning rather than a directive strategy. Although its individual components are supported by published evidence, the sequence as a whole still requires prospective clinical validation.

This approach has two advantages. It avoids unnecessary fluid administration to clear non-responders, and it also reduces the risk that a technically “positive” responsiveness test will be interpreted without regard to the patient's congestion phenotype. In other words, echocardiography is most useful not as a binary test but as an integrative haemodynamic method.

Implementation challenges and research priorities

Despite broad physiological appeal and extensive observational support, echocardiography-guided fluid management is not synonymous with outcome-proven care. Many studies are single-centre, include selected populations, use different thresholds for response, and rely on surrogate haemodynamic endpoints rather than patient-centred outcomes. Operator skill and image acquisition quality also remain important determinants of reliability (1, 10, 12–18, 24–26). This heterogeneity of populations, thresholds, and reporting practices, together with the limited number of high-quality outcome studies, remains a major barrier to comparing studies and translating findings into practice. It should therefore be acknowledged as a key limitation of the current evidence base. In this regard, the PRICES statement, an ESICM expert consensus on the design and reporting of critical care echocardiography research, provides a relevant methodological framework. Wider adoption of these recommendations in future studies would improve the quality, consistency and comparability of the evidence supporting echocardiography-guided fluid management (30). Throughout this review, efforts have been made to reference individual assertions to the original research articles from which they derive, rather than relying predominantly on other review articles.

Future work should therefore address three priorities: first, stronger trials evaluating patient-centred outcomes; second, standardised training pathways that improve reproducibility of Doppler-based measurements, particularly given the variability inherent in LVOT VTI acquisition; and third, greater integration of cardiac ultrasound with lung and venous ultrasound so that haemodynamic efficacy and congestion can be interpreted together rather than in parallel silos (2, 3, 18–21, 25, 26).

Conclusions

Echocardiography has become central to bedside haemodynamic assessment, but its greatest value in fluid resuscitation does not lie in static measurements alone. Rather, it lies in integrating shock phenotyping, dynamic assessment of the expected stroke-volume response, and identification of fluid intolerance. Among currently available strategies, passive leg raising combined with LVOT VTI monitoring remains the most practical and best-validated echocardiographic approach for predicting fluid responsiveness. Nevertheless, echocardiography is operator-dependent, and its reliability in guiding fluid decisions is closely linked to the skill and experience of the clinician performing and interpreting the examination. This remains a major practical limitation, should temper enthusiasm for indiscriminate bedside use, and reinforces the need for structured training. In the most complex patients, echocardiography should not replace continuous haemodynamic monitoring, but should be used alongside it as a complementary tool.

The field is now moving beyond the binary concept of fluid responsiveness. The clinically relevant question is whether a patient is both likely to benefit from fluid and likely to tolerate it. A stepwise echocardiography-based approach, ideally combined with lung and venous ultrasound and repeated serially throughout the patient's course, provides a physiologically coherent framework for this decision and may help avoid unnecessary fluid loading in already congested patients. It should be stressed, however, that this framework remains a physiologically grounded proposal rather than a validated clinical strategy. No randomised trial has yet shown that an echocardiography-guided approach to fluid therapy, whether based on fluid responsiveness alone or combined with markers of fluid tolerance, improves patient-centred outcomes such as mortality, organ failure or length of stay. Until such evidence is available, echocardiographic findings should inform bedside fluid decisions, not dictate them.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Raiko Blondonnet, Private Hopital La Châtaigneraie, France

Reviewed by: Julien Gotchac, Centre Hospitalier Universitaire de Bordeaux, France

Wenliang Song, The First Affiliated Hospital of Sun Yat-sen University, China

Abbreviations ARDS, acute respiratory distress syndrome; AUROC, area under the receiver operating characteristic curve; EEOT, end-expiratory occlusion test; E/e', ratio of early transmitral inflow velocity to early diastolic mitral annular velocity; IVC, inferior vena cava; LSC, least significant change; LV, left ventricle/left ventricular; LVOT VTI, left ventricular outflow tract velocity-time integral; PLR, passive leg raising; RV, right ventricle/right ventricular; SVC, superior vena cava; TEE, transoesophageal echocardiography; TTE, transthoracic echocardiography; VExUS, venous excess ultrasound.