Clinical review: Does it matter which hemodynamic monitoring system is used?

Hemodynamic monitoring and management has greatly improved during the past decade. Technologies have evolved from very invasive to non-invasive, and the philosophy has shifted from a static approach to a functional approach. However, despite these major changes, the critical care community still has potential to improve its ability to adopt the most modern standards of research methodology in order to more effectively evaluate new monitoring systems and their impact on patient outcome. Today, despite the huge enthusiasm raised by new hemodynamic monitoring systems, there is still a big gap between clinical research studies evaluating these monitors and clinical practice. A few studies, especially in the perioperative period, have shown that hemodynamic monitoring systems coupled with treatment protocols can improve patient outcome. These trials are small and, overall, the corpus of science related to this topic does not yet fit the standard of clinical research methodology encountered in other specialties such as cardiology and oncology. Larger randomized trials or quality improvement processes will probably answer questions related to the real impact of these systems.

instance, the pulse oximeter, which has been evalu ated in randomized controlled trials conducted in more than 20,000 patients in the anesthesiology setting, has never been shown to improve patient outcome [12,13]. In the same vein, there is no large randomized controlled trial showing that transesophageal echocardiography can improve patient survival even in the cardiac surgery setting [14]. Do we interpret this to mean that these devices should not be used in the clinical setting?
A few studies, especially in the perioperative period, have suggested that hemodynamic monitoring systems coupled with treatment protocols can improve patient outcome. Th ese trials are small and, overall, the corpus of science related to this topic does not yet fi t the standard of clinical research methodology encountered in other specialties such as cardiology and oncology. Larger randomized trials, quality improvement processes, and comparative eff ectiveness research studies will probably answer questions related to the real impact of these systems.
Th e term hemodynamic monitoring system is very broad and many concepts could be included under this terminology. Th e defi nition could range from microcircu lation and mitochondrial function monitoring to arterial pressure and heart rate monitoring. In the present article, we will focus our discussion on systems monitoring cardiac output (CO) and functional hemodynamic parameters. Several review articles have been published recently detailing the diff erent hemo dynamic monitoring systems available, and we refer the readers to these manuscripts for an in-depth technological understanding of these systems [3,4,8,15].
Th e present paper is divided into three parts. In the fi rst part we will describe the evolution of hemodynamic monitoring in the critical care environment during the past 50 years (from the pulmonary artery catheter (PAC) to the most recent functional hemodynamic monitoring). In the second part, we will analyze how these systems have been evaluated in the clinical practice (accuracy for CO monitors and predictive value for functional hemodynamic parameters) and we will analyze the impact of these systems on patient outcome. Lastly, in the third section, we propose a plan for the use of hemodynamic monitoring systems in critical care settings based on the specifi c clinical situation, the protocol to be applied, and on the patient. Th is approach is based on the model of clinical pathways and quality improvement processes implementation.
Hemodynamic monitoring in the critical care setting: from past to present

Pulmonary artery catheter
Intermittent thermodilution obtained through the PAC has been considered the gold standard for CO monitoring in the clinical setting since the late 1960s [16][17][18]. Th is system was widely used until the 1990s [19], when it started to dramatically decrease in all settings [20] secondary to a shift in philosophy, replacement by newer technologies [21], and also probably due to the widespread use of transesophageal echocardiography. Most studies focusing on the PAC and outcome have shown no positive association between PAC use for fl uid management and survival in the ICU [22][23][24] or in the high-risk surgery patient [25]. Th is conclusion combined with the extreme level of invasiveness, advanced level training for placement, and incorrect parameter interpretation have lead to declining use of this system [26]. However, the PAC still holds utility in the assessment of right ventricular CO, pulmonary arterial pressures [6,8], and mixed venous oxygen saturation monitoring [8,27]. Th e lack of positive impact of the PAC on patient outcome does not preclude its use in a selected sample of patients by adequately trained physicians. In addition, most new CO monitoring systems are still evaluated against the intermittent thermodilution technique in the clinical research practice [8].

Esophageal Doppler
Simultaneous to the decline of PAC use, development of less invasive hemodynamic monitoring systems started in the 1990s. One of the fi rst systems to be described and developed was the esophageal Doppler system allowing for non-invasive monitoring of CO [28]. Th is approach was described in the mid-1970s [29,30] and gained popularity in the 1990s after several small studies demonstrated a positive impact on postoperative outcome in patients undergoing high-risk surgery [31][32][33][34]. Th e waveform is highly dependent on correct positioning and requires frequent adjustments of depth, orientation, and gain to optimize the signal [35] and, while esophageal Doppler has demonstrated utility in aiding the assessment of the hemodynamic status of critically ill patients, this technology has been slow to be adopted [36]. Th is system has the most evidence regard ing improvement of outcome in patients under going high-risk surgery and therefore should be strongly con sidered in such a setting [37][38][39][40][41]. Th e National Health Service in the UK has recently recommended the use of this device during high-risk surgery [42,43].

Transpulmonary thermodilution
One of the most successful systems to be described has been transpulmonary thermodilution together with the concept of pulse contour analysis. Th is system was developed in the 1990s by a German company, Pulsion, who commercialized the PiCCO system (Pulsion, Munchen, Germany) [44]. Th is system, which requires the use of a dedicated arterial line (preferentially femoral) and a central venous catheter in the superior vena cava, achieved reasonable acceptance in Europe but is still relatively unknown in the USA [36]. Recently, Edwards Lifesciences released a similar trans pulmonary thermodilution system (Volume View; Edwards Lifesciences, Irvine, CA, USA) [45,46]. CO measure ments obtained using these systems correlate well with the PAC measurements [47][48][49][50]. Th ese systems allow for continuous CO measurements using the calibrated pulse contour analysis method. While interest ing, this method remains invasive [51] and requires frequent recalibration when the vasomotor tone changes [52][53][54][55]. Today, these systems have gained popularity in the ICU but are still rarely used in the operating rooms. Additionally, few outcome studies have been published using this system [56] and it is not clear whether their benefi ts outweigh the risks associated with femoral line placement.

Mini-invasive and non-invasive cardiac output monitoring systems
More recently, mini-invasive and non-invasive hemodynamic monitoring systems have been developed and deployed [57]. Most mini-invasive systems rely on the analysis of the arterial pressure waveform combined with the patients' demographics [58]. Th e systems are not user dependent and are very easy to use (plug-and-play). Th e main drawback of these systems is that they are highly dependent on vasomotor tone and on vascular compliance. Any acute change in these variables impacts the accuracy of these systems [54,59,60]. Th ere are limited, but positive, out come studies using these systems [61,62].
Finally, totally non-invasive systems have been developed. Bioimpedance and, more recently, bio reactance systems are back in the game (bioimpedance was actually developed very early on, before esophageal Doppler, but was never quite successful in the critical care setting) [63][64][65][66][67][68]. Ultrasound techniques such as the USCOM device have been proposed in the intensive care and emergency department settings (USCOM, Sydney, Australia). More innovative, uncalibrated and noninvasive CO measurements obtained through the analysis of a non-invasive arterial pressure waveform have also been released recently [69,70]. However, we need more evidence regarding the accuracy of these systems in order to use the output variables for clinical decision-making. Table 1 summarizes the diff erent CO monitoring systems clinically available today.

Functional hemodynamic monitoring parameters
Apart from the technological development of CO monitoring systems, most of the recent medical literature focusing on hemodynamic monitoring in the critical care setting has focused on the so-called functional hemo dynamic monitoring parameters [1]. Instead of monitoring a given parameter, functional hemodynamic moni toring assesses the eff ect of a stressor on a hemodynamic parameter. For the assessment of preload dependence, the stress has been established as a fl uid challenge and the parameter monitored is the stroke volume or one of its surrogates (for example, arterial pressure) [71]. Th e eff ects of positive pressure ventilation on preload and stroke volume are used to detect fl uid responsiveness in mechanically ventilated patients under general anesthesia [5]. If mechanical ventilation induces prominent respira tory variations in stroke volume [5], systolic pressure [72][73][74] or in arterial pulse pressure (PPV) [75], the patient's heart is more likely to be working on the steep portion of the Frank-Starling relationship and is thus preload dependent. Th ese eff ects can also be assessed by measur ing the variability on the inferior [76,77] or superior [78] vena cava diameter. When these parameters cannot be used because one limitation is present (spon taneous ventilation [79], arrhythmia [80], tidal volume <7 ml/kg [81], open chest conditions [82]), the eff ects of passive leg raising on the stroke volume can be used to detect preload dependence instead [83].
Monnet and colleagues have also described the eff ects of a tele-expiratory occlusion on arterial pulse pressure to predict the eff ects of volume expansion on CO [84]. Th e tele-expiratory occlusion test consists of a 15-second end-expiratory occlusion in patients under mechanical ventilation. Patients presenting a more than 5% increase in pulse pressure (systolic pressure -diastolic pressure) are more likely to be responders to volume expansion (in these patients, the increase in venous return induced by the decrease in intrathoracic pressure induces an increase in stroke volume and pulse pressure because the patient's heart is working on the steep portion of the Frank-Starling curve). Th ese maneuvers (passive leg raising and tele-expiratory occlusion tests) are more appropriate for the ICU setting while stroke volume variation, PPV or respiratory variations in the plethysmographic waveform [85] are more suited for the operating room setting.
Th ese dynamic parameters have consistently been shown to be superior to static parameters for the prediction of fl uid responsiveness [75,86]. Moreover, some studies have suggested that minimization of the respiratory variations in the arterial pressure or in the plethysmo graphic waveforms could improve fl uid management and postoperative outcome [10,87,88]. Table 2 presents the various functional hemodynamic parameters available for the prediction of fl uid responsiveness as well as the monitors available for their display.

Evaluation of hemodynamic monitoring systems and how to choose among them?
Now that we have discussed the diff erent hemodynamic systems available, the question remains as to whether or not it matters which of them is used for the management of critically ill patients. Practically speaking, if we want to reasonably answer this question, we need to defi ne what we expect from these systems, to identify the setting in which the system is going to be used, and to evaluate whether or not these systems accurately achieve what we expect from them.

What do we expect from a hemodynamic monitoring system?
Th is question depends on the monitor. At the very least, we expect a CO monitoring system to measure CO accurately and we expect a fl uid responsiveness monitor to predict fl uid responsiveness accurately.

Evaluating the accuracy of a cardiac output monitoring system
Th e last two decades have seen an explosion in the numbers of manuscripts published aiming at evaluating the accuracy of a hemodynamic monitoring system to measure CO accurately. Dozens of manuscripts have tried to defi ne the methodology that should be used to clearly demonstrate whether or not a monitor can measure and track CO in the clinical setting. After more than a decade of research on this methodology alone, a consensus seems to have been reached [89][90][91][92]. Th e fi rst step is to evaluate the agreement between the new method and the gold standard (most studies still use intermittent thermodilution as the gold standard). For this purpose, Bland-Altman analysis -originally described to assess the agreement between two methods of clinical measurement -should be used [93,94]. Th is analy sis provides the users with a bias and limits of agreement. Unfortunately, little is known with regard to what is considered acceptable or not. Th e second step is to calculate the mean percentage error, which is defi ned as the ratio between the range of the limits of agreement and the mean CO of the gold standard [95]. Th e idea is that narrow limits of agreement may mean that the system is very precise, but one must acknowledge that this may also only indicate the study was conducted in a sample of subjects presenting with very low CO values. According to Critchley and Critchley, a mean percentage error >30% should allow the conclusion of the new method as being inaccurate [95]. Once again, one must remember that this approach depends on the intrinsic precision of the gold standard [96]. Th e third step is then to test the concordance between the new method and the gold standard [97,98]. Basically, this approach aims at evaluating whether or not the two techniques follow the same direction when CO is modifi ed. For some, a weak accuracy may not be a problem as long as the ability to track changes is accurate. Th is is theoretically true since most goal-directed therapy protocols for fl uid management rely on relative changes in CO, as opposed to absolute values. However, a perfect trending ability together with a weak accuracy would essentially indicate that further calibration would solve the problem. Unfortunately, it is more likely that the bias and the limits of agreement drift over time. According to Critchley and colleagues, concordance <92% should be considered inacceptable.

Evaluating the predictive value of a functional hemodynamic parameter
Th e use of functional hemodynamic monitoring in the clinical practice is, in a way, simpler to evaluate. Th e goal of most functional hemodynamic monitoring parameters is to predict fl uid responsiveness in critically ill patients. For this purpose, the methodology is quite straightforward and has for a long time relied on the use of receiver operating characteristics curve analysis [99]. Th is analysis results in a single threshold value associated with a high sensitivity and specifi city for the prediction of fl uid responsiveness. For example, it was shown initially that PPV >13% in septic patients was able to predict fl uid responsiveness with sensitivity and specifi city >90% [100]. However, this methodology is very old and may not refl ect the actual clinical setting where such polarized situations do not often exist. Signifi cant improvements have recently been made in the methodology used for the evaluation of a biomarker or diagnostic tool [101]. For instance, the gray zone approach has been proposed to avoid the binary constraints resulting from the black-or-white nature of the receiver operating characteristics curve that often does not fi t the reality of clinical or screening practice [101]. Th e gray zone technique proposes two cutoff values that constitute the borders of the gray zone. Th e fi rst cutoff allows the practitioner to exclude the diagnosis (fl uid responsiveness in the present case) with near-certainty (that is, privilege sensitivity and negative predictive value), whereas the second cutoff is chosen to indicate the value above which the selected diagnosis can be included with near-certainty (that is, privilege specifi city and positive predictive value) [101]. Intermediate values included in the gray zone correspond to a prediction value not precise enough for a diagnostic decision [102]. Th is approach has recently been applied to test the ability of PPV to predict fl uid responsiveness in the perioperative setting, and it was shown, in more than 400 patients, that the gray zone for PPV is between 8 and 13% and that about 25% of the patients are within this gray zone. If this approach was used in the ICU setting, the majority of patients would more likely be within this gray zone. Th is type of approach should help to better defi ne the clinical application of these functional hemodynamic parameters. In any case, these dynamic parameters have consistently been shown to be the best predictors of fl uid responsiveness [75,86].

Evaluating the impact on outcome
Finally, the ultimate test is to evaluate whether or not the use of a monitor to guide hemodynamic management can improve patient outcome.
Th e problem is that none of the CO monitoring systems available today consistently present with <30% mean percentage error, >92% concordance, and positive outcome studies. Most widely used CO monitoring systems demonstrate a mean percentage error of around 40 to 45% [103] and most of these devices present with concordance <92%. Interestingly, despite these very disappointing results, these systems have still been tested in clinical outcome studies and some have demonstrated positive results [37,104]. Let us stress this point and be a little bit provocative: it is surprising to observe that a professional discipline such as medicine is able to conduct clinical studies using devices that have been consistently demonstrated to be inaccurate. One may argue that the methodology used to evaluate these systems (mean percentage error, concordance) is not appropriate and presents intrinsic limitations. However, would any other industry dealing with life and death situations accept such a shortcoming? Would an altimeter be used on a commercial passenger plane despite the fact that it has been demonstrated to be inaccurate according to the most commonly accepted standards from the Federal Aviation Administration? Why would we, as physicians, accept what other industries would clearly consider unacceptable?
Th e reason for this shortcoming is probably related to the fact that human physiology and physiopathology is an incredibly complex model. Th is explains why it is so diffi cult to reliably measure physiological variables, and it also explains why it is so diffi cult to make good clinical decisions. As a matter of fact, when dealing with complex situations, medical decision-making can be completely diff erent from one physician to the next [26]. Th is lack of standardization in patient management is probably one of the major factors infl uencing patient outcome and, coincidentally, one of the only factors that we can infl uence for the improvement of patient care. Th is has been beautifully demonstrated during the past 10 years by studies in the critical care setting focusing on protocol implementation and quality improvement processes aiming at standardizing patient care. Goal-directed therapy protocols or checklist implementations exemplify this type of approach [105][106][107]. Studies such as those conducted by Rivers and colleagues in septic patients showing the impact of standardizing hemodynamic management on survival have opened the fi eld to such approaches [9]. Most recently, studies have demonstrated that applying a simple checklist in the ICU and in the operating room can signifi cantly impact outcome [108,109]. Th ese studies are repeatedly concluding that decreasing the variability of care can save lives.
Concerning hemodynamic monitoring systems, the same approach could be applied. Clearly, despite the lack of precision of most CO monitoring systems available, some positive outcome studies have been published, especially in the perioperative setting with patients undergoing high-risk surgery. Th ese studies have shown that the optimization of fl uid administration based on CO monitoring can decrease postoperative morbidity and decrease the length of stay in the hospital and in the ICUs (Figure 1) [37,104]. Since fl uid and hemodynamic management have been shown to impact postoperative outcome and because these two major focuses of our fi eld have been shown to be widely nonstandardized [110,111], it would then be reasonable to assume that using a CO monitoring system (even if not a perfect one) to guide fl uid administration in a standardized way in the perioperative period has the potential to improve post operative outcome. Th is approach consists of titrating fl uid, based on CO, until it reaches the plateau of the Frank-Starling relationship (Figure 1), which has been shown in several small clinical studies to improve patient outcome.
Th e evidence has been considered strong enough by the National Health Service in the UK to universally endorse this practice in the high-risk surgery setting [42,43], even though it is has created some heated discussion [112,113]. Widespread acceptance of this concept in other countries will probably take longer due to the relative infrequency of large clinical studies. Such evaluations are strongly needed in the perioperative period [11,114]. Th is is exemplifi ed by Devereaux and colleagues in an editorial recently published in Anesthesiology [114]: 'Unlike cardiology, large clinical studies remain uncommon in perioperative medicine [115,116]. Further, there has been a tendency to believe the results of small peri operative clinical studies, especially when they demon strate statis tically signifi cant results. Th is position is supported by the fact that perioperative guideline committees recom mended β-blockers to patients undergoing noncardiac surgery for a decade based upon the results of small trials demonstrating implausibly large treatment eff ects' . One should also mention that large qualityimprovement programs and comparative eff ectiveness research studies could also be utilized as an alterna tive to this approach [40,117].
In addition, while functional hemodynamic parameters can be used as diagnostic tools to answer whether a patient needs fl uid or not, another approach consists of using these parameters to guide fl uid optimization during high-risk surgery [10]. As a matter of fact, the concept of CO maximization during surgery could be achieved by applying the concept of respiratory variations in arterial pressure or in the plethysmographic waveform minimization ( Figure 2) [10]. Conducting CO maximization using CO monitors that have >40% mean percentage error [103] could theoretically be easily achieved by conducting PPV minimization. Th is would be a cheap and free-of-right way to optimize hemodynamics in the perioperative period. PPV minimization has been suggested and recently described [10,118], and the method could be of major importance in countries or institutions where the use of CO monitoring systems cannot be reasonably expected for all patients undergoing high-risk surgery but where fl uid optimization still has the potential to dramatically infl uence patient outcome [119]. Of course, it may be diffi cult to determine the clinical eff ect of minimizing PPV without CO trending ability. However, recent studies strongly suggest that changes in PPV induced by volume expansion refl ect changes in CO with excellent sensitivity, specifi city, and a very narrow gray zone [120].

Which hemodynamic monitoring system? For which patient? When? How?
Several parameters must be considered before deciding which hemodynamic monitoring system should be used because, yes, it does eventually matter. Ideally, this decision should be made at the institutional level. Most departments throughout the world cannot aff ord to buy all of the market-available systems. At the same time, no system available today can eff ectively be used in all the diff erent sectors of a hospital. Depending on the patient's specifi c history and course through the hospital, one hemodynamic monitoring system may be more appropriate than the other. Defi ning a set of systems available that will be adaptable to the various patient populations and clinical pathways will then be essential. Th is approach has been recently proposed by Alhashemi and colleagues ( Figure 3) and has been described as an integrative perspective for the use of CO monitoring systems [3]. Th e defi ning approach takes into account the setting (ward, emergency depart ment, operating room, and ICU) as well as the integration of CO monitoring with or without other hemodynamic variables.

The decision should be an institutional decision and should integrate all clinical pathways existing within the institution
Institutions containing emergency departments, operating rooms, and ICUs should have non-invasive, miniinvasive and invasive hemodynamic monitoring systems available to the clinicians and his or her patients. Likewise, when the institution performs cardiac surgery, it is still highly recommended to have PACs available. Th is system is well fi tted for patients with low ejection fraction (<30 to 35%), moderate to severe pulmonary hypertension, sepsis (endocarditis), and for cardiac transplantation. If the patient spends more than 72 hours in the ICU after surgery, it is recommended to switch from the PAC to a transpulmonary thermodilution system. Of course, transesophageal echocardiography should be available in all institutions performing cardiac surgery. However, this system is not a monitoring system per se and does not substitute for a continuous hemo dynamic monitoring system.
An important consideration is that patient management is a continuum of care. Consequently, it is essential to maintain compati bility among hemodynamic monitoring technologies between diff erent departments within the institution and to favour systems able to adapt to various clinical pathways. For example, some patients will enter the hospital through the emergency department, will then go to the operating room, and then to the ICU. Ideally, the evolution in hemodynamic monitoring should be made available on an identical platform that will adapt to the changes in hemodynamic status of the patient as well as to the clinical scenario in these diff erent departments. Today, technological platforms allowing for a con tinuum of care from a totally non-invasive hemodynamic monitoring system to a mini-invasive one and then to an invasive one (or vice versa) are just emerging. For an institution to work within a given system that would be fl exible and allow any kind of patient throughout the hospital to be eff ectively monitored would make perfect sense. Once again, this kind of platform is just starting to emerge and most institutions still have to purchase diff erent systems, from diff erent companies, in order to monitor diff erent patients.

The systems should be paired with clearly defi ned protocols
As mentioned earlier, the only way to impact patient outcome is to pair the monitoring system with a therapeutic protocol. Th is approach has been shown to improve perioperative outcome in several small clinical studies and in some quality improvement processes employing the esophageal Doppler [40]. Such standardization of patient care is the only way to change current practice and to pragmatically and positively impact clinical decision-making. Standardization guidelines should also include the indications for hemodynamic monitoring and which hemodynamic monitoring system should be used for what patients (based on the integrative approach described above; Figure 3). Once again, the National Health Service in the UK has exemplifi ed this through its release of National Institute for Health and Clinical Excellence guidance regarding hemodynamic monitoring and optimization during high-risk surgery [42,43]. Th ese guidelines are clear and easy to apply and can easily be applied in any institution.

The system should be adapted to the patient
Of course, the fi nal choice of a hemodynamic monitoring system is patient and pathology dependent. Additionally, whenever possible, a non-invasive system should be used. However, at this stage, non-invasive systems may not be as reliable as invasive ones. Th ere is no doubt that noninvasive systems will eventually take the lead in the future [121], but we are still contemplating the eff ective length of a development phase [122]. For example, non-invasive systems based on pulse oximeter waveform analysis have been shown to be able to provide useful information regarding fl uid responsiveness in healthy patients under general anesthesia [85]. However, these systems may not be reliable in the ICU in septic shock patients [123]. Th at being said: who would consider managing the hemodynamic status of a septic shock patient based solely on the plethysmographic waveform alone? On one hand, the risk of using a non-invasive technique in a challenging setting is that it will lead to inappropriate clinical decisions. On the other hand, it is unacceptable to expand the indications for invasive monitoring when their risks outweigh their benefi ts. We should always keep this in mind when choosing the most appropriate hemodynamic monitor for our patients.

Conclusion
Hemodynamic monitoring and management has greatly improved during the past decade. Technologies have evolved from very invasive to non-invasive, and the philosophy has shifted from a static approach to a functional approach. However, the critical care community still has potential to improve its ability to adopt the most modern standards of research methodology in order to more eff ectively evaluate new monitoring systems and their impact on patient outcome. Today, despite the huge enthusiasm raised by new hemodynamic monitoring systems, there is still a big gap between clinical research studies evaluating these monitors and clinical practice. A few studies, especially in the perioperative period, have shown that hemodynamic monitoring systems coupled with treatment protocols can improve patient outcome. Unfortunately these trials are small and, overall, the corpus of science related to this topic does not yet fi t the standard of clinical research methodology encountered in other specialties such as cardiology and oncology. Larger randomized trials, quality improvement processes, and comparative eff ectiveness research studies are probably needed. However, some innovative professional societies have considered that this evidence was strong enough to release recommendations regarding hemodynamic monitoring and management during high-risk surgery. For this purpose, strictly speaking, the eso phageal Doppler is the device that currently presents with the most positive evidence.
Finally, the choice of the hemodynamic monitoring systems available should be a widespread institutional process and all departments involved should be consulted (emergency department, ward, ICU, and operating room). At the end of the day, the choice depends on the available expertise, on the patient population, and on the clinical pathways. For institutions who cannot aff ord a proprietary hemo dynamic monitoring system for their patients, fl uid optimization can be achieved eff ectively by monitoring respiratory variations in the arterial pressure or in the plethysmographic waveform depending on the clinical context. Abbreviations CO, cardiac output; PAC, pulmonary artery catheter; PPV, pulse pressure variation.

Competing interests
DR is a consultant and speaker for Edwards Lifesciences. MC is a consultant for Masimo Corp., Edwards Lifesciences, Covidien, Draeger, Philips Medical Systems, Fresenius Kabi, and a speaker for ConMed. MC received research grants from Edwards Lifesciences and Masimo Corp. MC is co-inventor and co-owner on US and International patent applications related to closedloop fl uid administration and hold an equity positions in Sironis, a company developing closed-loop medical devices. BA declares that he has no competing interests.