Assessment of volume responsiveness during mechanical ventilation: recent advances

Predicting which patients with acute circulatory failure will respond to fluid by a significant increase in cardiac output is a daily challenge, in particular in the setting of the intensive care unit (ICU). This challenge has become even more crucial because evidence is growing that administering excessive amounts of fluid is a risk factor in critically ill patients, in particular in patients with lung injury. However, some tests and indices allow prediction of fluid responsiveness before intravenous fluids are infused.


The conc ept of predicting fl uid responsiveness
Volume expansion is the fi rst-line treatment in the majority of cases of acute circulatory failure. Fluid is administered with the expectation that it will increase cardiac preload and cardiac output to a signifi cant extent. Nevertheless, this can occur only if cardiac output is dependent upon cardiac preload, i. e., if both ventricles operate on the ascending limb of the cardiac function curve [1] (Fig. 1). If this is not the case, volume expansion may only exert adver se eff ects without having any hemodynamic benefi t. An important point is that excessive fl uid administration has been demonstrated to increase mortality during septic shock [2,3] and to prolong mechanical ventilation during acute respiratory distress syn dr ome (ARDS) [4]. In the same co ntext, the amount of extravascular lung water (EVLW), i. e., the volume of lung edema, has been demonstrated to be related to mortality in critically ill patients [5] and, more recently, to be an independent prognostic factor during ARDS [6]. Th us, fl uid responsiveness should be detected before deciding to administer volume expansion, especial ly in patients in whom fl uid overload should be particularly avoided, i. e., patients with septic shock and/ or ARDS.
For this purpose, 'static' markers of cardiac preload have been used for many years. Nevertheless, a very large number of studies clearly demonstrate that neither pressure nor volume markers of preload can predict fl uid responsiveness [7,8]. Th is fi ndin g is mainly because a given value of preload can correspond to either a large or a negligible response of cardiac output to fl uid administration, depending upon the slope of the Frank-Starling curve, which cannot be a priori determined in a given patient ( Figure 1). Th is is the reason why a 'dynamic approac h' has been developed for assessing volume responsiveness [9]. Th e concept is to assess preloaddependency by observing the eff ects on cardiac output of changes in cardiac preload induced by various tests.

Pulse pressure variation
Th e fi rst application of this 'dynamic' concept consisted of quantifying the variations in stroke volume induced by positive-pressure ventilation [10]. As a result of heartlung interactions, each mechanical insuffl ation decrea ses venous return and, if the right ventricle is preloaddependent, reduces the right ventricular (RV) outfl ow. Increase in RV afterload induced by increased lung volume contributes to this reduction in RV outfl ow. In turn, this results in a decrease in left ventricular (LV) preload, which occurs after a delay of a few cardiac cycles, required for the blood to transit through the lungs. In the case of conventional ventilation, this eff ect should occur at expiration. If the left ventricle is also preload-dependent, the LV stroke volume transiently decreases. As a result, a cyclic variation in stroke volume under mechanical ventilation indicates the existence of preload-dependen cy of both ventricles [10].
Th e arterial pulse pressure (systolic minus diastolic arter ial pressure) as a surrogate of stroke volume has been proposed to predict fl uid responsiveness through its respiratory variation [10]. A number of studies conducted in various clinical settings have repeatedly demonstrated that PPV actually predicts fl uid responsiveness [11]. Among the d iff erent indicators of fl uid responsiveness, PPV is supported by the highest level of evidence. Importantly, PPV is calculated automatically and displayed in real-time by the most recent bedside hemodynamic monitors.

Other surrogate s of stroke volume
Almost all indices that provide a beat-to-beat estimation of stroke volume have been investigated for their ability to test fl uid responsiveness by their respiratory variation: Aortic blood fl ow measured by esophageal Doppler [12]; subaortic peak velocity measured by echocardiography [13]; stroke volume estimated from pulse contour analysis [14]. Some recent studies have investigated non-invasive estimations of stroke volume, such as the non-invasive arterial pulse pressure estimated by the volume-clamp method [15,16] or the amplitude of the plethysmographic waveform [17]. All the se non-invasive methods still require confi rmatory studi es, but they may be of great interest for the perioperative management of low-risk surgical patients, when there is no need for an invasive monitoring device.
Finally, analysis of respiratory variations of the diameter of the inferior vena cava (using transthoracic echocardiography) [18] or of the superior vena cava (using trans esophageal echocardiography) [19] can also be used to assess fl uid responsiveness in mechanicall y ventilated patients.

Limitations of the respiratory variation in stroke volume for predicting fl uid responsiveness
Th e use of respiratory variation of stroke volume or surrogates to predict fl uid responsiveness has some limita tions that are now clearly identifi ed. Th e fi rst and most important limitation is the presence of some spontaneous breathing activity [20]. When a patient has some breathing eff orts under mechanical ventilationand even more when the patient is not intubated -the variation in intrathoracic pressure is not regular, neither in rate nor in amplitude, such that the variation in stroke volume can not relate to preload-dependency [21][22][23]. Th us, using the respiratory variation of stroke volume to test the preload sensitivity is valid only in cases of coma or deep sedation during mechanical ventilation, a condition that has become less frequent in the ICU.  A second limitation of using the respiratory variation in stroke volume is the presence of cardiac arrhythmias as in these cases, the variation in stroke volume is obviously more related to the irregularity of diastole than to heart-lung interactions.
A third limitation refers to conditions in which the variations in intravascul ar pre ssure induced by mechanical ventilation are of small amplitude. In the case of low tidal volume, the small variations in intrathoracic pressure may not be suffi cient to trigger signifi cant preload variations, even in cases of preload responsiveness. Some studies have actually shown that PPV loses its predictive value in the case of low tidal volume [24,25]. Th e changes in intravascular pressure induced by mechanical ventilation may also be reduced if the transmission of changes in alveolar pressure to the pressure of the intrathoracic structures is attenuated, e. g., in the case of low lung compliance. A recent clinical study demonstrated that if the compliance of the respiratory system is less than 30 ml/cm H 2 O , th e value of PPV for predicting fl uid responsiveness is dramatically reduced, independently of tidal volume [26].
A fourth limitation is the use of high frequency ventilation. If the ratio of heart rate to respiratory rate is low, e. g., if the respiratory rate is elevated, the number of cardiac cycles per respiratory cycle may be too low to allow respiratory stroke volume variation (SVV) to occur [27]; nevertheless, this applies to respiratory rates as high as 40 breaths/min [27].
A fi fth limitation is the presence of increased abdominal pressure [28,29]. In such cases, a higher PPV cut-off value must be considered for the prediction of fl u id responsiveness [28]. Finally, open-chest surgery is another situation where the ventilation-induced variation of hemodynamic signals loses its predictive value for fl uid responsiveness [30].
Except for open-chest condit ions, the limitations of PPV and surrogates as predictors of fl uid responsiveness mainly concern the intensive care setting and not the operat ing room where these heart-lung interaction indices are fully relevant. During recent years, alternative methods have been developed for predicting fl uid responsiveness in critically ill patients (F igure 2).

The end-expiratory occl usion test
Th is test is another method that takes advantage of heartlung interactions to predict fl uid responsiveness in ventilated patients. During mechanical ventilation, each insuffl ation increases the intrathoracic pressure and impedes venous return. Th us, interrupting the respiratory cycle at end-expiration inhibits the cyclic impediment in venous return. Th e resulting increase in cardiac preload may thus help to test preload responsiveness (Fig. 1). Indeed, it was demonstrat ed that if a 15-sec ond endexpiratory occlusion test increased the arterial pulse pressure or the pulse contour-derived cardiac output by more than 5 %, the response of cardiac output to a 500 ml saline infusion could be predicted with good sensitivity and specifi city [31]. Noticeably, all patien ts of the latter study were arrhythmic or had mild spontaneous breathing activity. Th ese initial results were recently confi rmed [26].
Beyond its simplicity, the main advantage of the endexpiratory occlusion test is that it exerts its hemodynamic eff ects over several cardiac cycles and thus remains valuable in c ase of cardiac arrhythmias [31]. Also, the end-expiratory occlusion test can be used in patients with spontaneous breathing activity, unless marked triggering activity interrupts the test. Another limitation is that the eff ects of the end-expiratory occlusion t est, which must be observed over 15 s, are much easier to observe on a continuous display of cardiac output than on the arterial pulse pressure because the value of the latter is not continuously calculated and displayed by bedside monitors.

The 'mini' fl uid challenge
Obviously, the easiest way to test preload responsiveness is to administer fl uid and to observe the resulting eff ect on cardiac output. Nevertheless, the disadvantage of the 'classical' fl uid challenge is that it consists of administration of 300-500 ml of fl uid [32]. Because it is not reversible, such a fl uid challenge may contribute to fl uid overload, especially when it is repeated several times a day [33].
In this regard, a new method has been proposed for performing a fl uid challenge [34]. It consists of administe ring 100 ml of colloid over 1 min and observe the eff ects of this 'min i' fl uid challenge on stroke volume, as measured by the sub aortic velocity time index using transthoracic echocardiography (Figure 1). In a clin ical study, an increase in the velocity time index of more than 10 % predicted fl uid responsiveness with a sensitivity of 95 % and a specifi city of 78 % [34].
Th e fi rst advantage of this test over the classical fl uid challenge is obviously that such a small volume of fl uid is unlikely to induce fl uid overload [33], even if repeated several times a day. Another advantage is that it is easy to perform and that it can be assessed in a non-invasive way [34]. Nevertheless, a strong limitation is tha t, even in cases of preload-dependency, such a small volume infusion will unavoidably induce only small changes in cardiac output. Th is test, therefore, requires a very precise technique for measuring cardiac output. Whether non-inva sive techniques, such as echocardiography, are pre cise enough for this purpose is still uncertain. Finally, this method cannot be used in the presence of cardiac arrhythmias. In the study by Muller et al., one fourth of patients were excluded because of this limitation [34].

The passive leg-raising test
In a l ying subject, raising the legs from the horizontal position passively transfers a signifi cant volume of blood from the lower part of the body toward the cardiac chambers: Passive leg-raising (PLR) partially empties the venous reservoir and converts a part of the unstressed blood volume to stressed volume. In this connection, PLR increases right [35][36][37] and left [22,35] cardiac pre load. Eventually, the in crease in left cardiac preload results in an increase in cardiac output depending upon the degree of preload reserve of the left ventricle. Th e increase in cardiac preload induced by PLR totally reverses once the legs are returned back to the supine position [22,35,36]. In summary, PLR acts like a rev ersibl e and short-lived 'self ' volume challenge [38] (Fig. 1). As a clinical applicat ion of this sim ple physiological concept, several studies reported that the increase in cardiac output induced by the PLR test enables prediction of fl uid responsiveness; many of these studies were included in a meta-analysis [39]. Th ese fi ndings have contributed to establish PLR as a reliable and easy way to predict fl uid responsiveness at the bedside [40]. Interestingly, since the test ex erts its eff ects over several cardiac and respiratory cycles, it remains a good predictor of fl uid responsiveness in patients with spontaneous breathing activity (even non-intubated) or cardiac arrhythmias [22,31].
Th e postural change used for perfor ming the PLR test is important [38]. If PLR is started from the 45° semirecumbent position, the induced increase in central venous pre ssure is larger than if started from the supine position [36]. In fact, starting PLR from the semirecumbent position mobilizes blood coming not only from the inf erior limbs but also from the large splanchnic compartment. As a consequence, starting PLR from the semi-recumbent posture is more sensitive than starting from the horizontal posture to detect fl uid responsiveness [36], so that this method should be considered as a standard.
Another important point concerns the method that can be used for measuring the changes in cardiac output during PLR [38]. A real-time measurement able to tr ack hemodynamic changes in the time frame of PLR eff ects, i. e., 30-90 s, must be used. Indeed, the increase in cardiac output during PLR is not sustained when the leg elevation is prolonged. Th is aspect is particularly true in septic patients in whom capillary leak may account for an attenuation of the PLR eff ects after one minute, as already described [22]. Th is is why clinical studies that ha ve tested the value of PLR to predict volume responsiveness used real-time hemodynamic measurements, such as aortic blood fl ow measured by esophageal Doppler [22,41], pulse contour analysis-derived ca rdiac output [16,26,42], cardiac output measured by bio reacta nce [4 3,44] or endotracheal bioimpedance cardi ograph y [45], subaortic blood velocity measured by echocardiography [46][47][48], ascending aortic velocity measure d by s uprasternal   [49] and, more recently, end-tidal carbon dioxide [50,51].
Beyond its reliability and ease o f use, the PLR test has some limitations [38]. First, its eff ects cannot be assesse d by observing the arterial pressure. PLR-induced changes in arterial pulse pressure are less accurate than PLRinduced changes in cardiac output or stroke volume, as found by several studies [39]. Th is fi nding is explained by the fac t that arterial pulse pressure is only a rough surrogate of stroke volume. Th is means that correct performance of a PLR test requires a device that allows a more direct estimation of cardiac output. Second, the PLR test cannot be used in instances in which mobilizing the patient is not possible or allowed, e. g., in the operating room or in the case of head injury [52].

Eff ects of volume expansion on cardiac ou tput
If the decision is taken to infuse fl uid in case of preload reserve, the ensuing question is whether the fl uid actually exerts its expected benefi cial eff ects. What is expected from fl uid administration is a signifi cant increase in cardiac output. In this regard, it has been recently shown that changes in arterial pressure are relatively imprecise to estimate the eff ects of fl uid infusion on cardiac output [53,54]. In 228 patients who received a sta ndardi zed saline infusion, fl uid-induced changes in arterial pulse pressure were weakly correlated with the simultaneous changes in cardiac output (r = 0.56) [54]. Consequently, the changes in pul se press ure induced by volume expansion detected a positive response to fl uid (i. e., an increase in cardiac output ≥ 15 %) with a specifi city of 85 % but with a sensitivity of only 65 %; in other words, 22 % of cases were false negatives, meaning that in these patients, fl uid administration signifi cantly increased cardiac output whereas the arterial pulse pressure did not change to a large extent. Pierrakos et al. confi rmed these results in 51 septic shock patients [54], in whom there was no signifi cant correlation betw een fl uid-induced changes in arterial pulse pressure and fl uid-induced changes in cardiac output. Th ese results are explained by the fact that arterial pulse pressure is physiologically related to stroke volume but also inversely correlated with arterial compliance [55], which may diff er among patients and may change ov er time in the same patient. Moreover, the proportionality between pulse pressure and stroke volume is physiologically expected at the aortic level, but not at the peripheral arterial level because of the pulse wave amplifi cation phenomenon.
Another important point emphasized by the above cited studies [53,54], is that the fl uid-induced changes in cardiac o utput are not refl ected at all by the fl uid-induced changes in mean arterial pressure (MAP). Physiologically, the changes in MAP are dissociated from the changes in cardiac output because of the sympathetic modulation of the arterial tone, which tends to maintain MAP constant while cardiac output varies. Th ese results suggest that precise assessment of the eff ects of volume expansion should not rely on simple blood pressure measurements but should rather be based on direct measurements of cardiac output. Th is may be particularly important in patients at risk of fl uid overload during sepsis and/or ARDS.

Eff ects of volume expansion on tissue oxygenation
Wha t is really expected from volume expansion in a patient with acute circulatory failure is not only an increase in cardiac output but also an improvement in tissue oxygenation. Th is can only occur if oxygen consumption depends on oxygen delivery (DO 2 ). However, this is not always the case. In a recent study, we observed that volume expansion administered in patients with acute circulatory failure increased cardiac output ≥ 15 % in 49 % of cases ("volume-responders") [56]. Although DO 2 signifi cantly increased in "volumeresponder s", a signifi cant increase in oxygen consumption occurred in only 56 % of these patients. When looking at variables that could identify patients who would benefi t from volume expansion in terms of oxygen consumption, we found that markers of global tissue hypoxia, such as blood lactate, were more relevant than the central venous oxygen saturation (ScvO 2 ). Th is illustrates the diffi culty of using ScvO 2 for assessment of fl uid therapy [57] because ScvO 2 cannot detect preload responsiveness [58] or identify among volume-responders those who will benefi t from fl uid infusion in terms of tissue oxygenation [56]. Importantly, in our study, DO 2 decreased with volume infusion in the "volume-non-responders" (51 %), owing to development of hemodilution [56]. Th is latter fi nding reinforces the message that identifying volume r esponders before any fl uid administration is crucial since volume expansion in patients without preloaddependency can be deleterious not only on lung function but also on peripheral oxygenation.

Conclusion
Th ere is growing evidence suggesting that overzealous fl uid administration is del eterious in critically ill patients, particularly in cases of sepsis and/or lung injury. During recent years, several tests have been developed to detect volume re spon siveness before administering fl uid. Th ese tests can also serve to detect volume unresponsiveness, which could be helpful at any moment of fl uid resuscitation to better assess the benefi t/risk ratio of continuing such a strategy [59]. Th e analysis of respiratory variation in stroke volume has received the largest level of evidence, but cannot be used in cases of spontaneous breathing activity, cardiac arrhythmias, low tidal volume or low lung compliance. Some more recently developed tests, such as the end-expiratory occlusion test, the 'mini' fl uid challenge and the PLR test can be used as alternative methods, solving the problem of prediction of volume responsiveness in cases of spontaneous breathing activity and/or cardiac ar rhythmias. Th e ideal management of fl uid therapy should also include a precise a ssessment of the eff ects of volume expan sion on cardiac output and tissue oxygen consumption.