Clinical review: Respiratory monitoring in the ICU - a consensus of 16

Monitoring plays an important role in the current management of patients with acute respiratory failure but sometimes lacks definition regarding which 'signals' and 'derived variables' should be prioritized as well as specifics related to timing (continuous versus intermittent) and modality (static versus dynamic). Many new techniques of respiratory monitoring have been made available for clinical use recently, but their place is not always well defined. Appropriate use of available monitoring techniques and correct interpretation of the data provided can help improve our understanding of the disease processes involved and the effects of clinical interventions. In this consensus paper, we provide an overview of the important parameters that can and should be monitored in the critically ill patient with respiratory failure and discuss how the data provided can impact on clinical management.


Introduction
Monitoring plays an important role in the current management of patients with acute respiratory failure. However, unlike monitoring of other organs and functions, monitoring of respiratory function in the critically ill sometimes lacks defi nition regarding which 'signals' and 'derived variables' should be prioritized as well as specifi cs related to timing (continuous versus inter mittent) and modality (static versus dynamic). In this consensus paper, we summarize current modes of respira tory monitor ing and their potential practical appli ca tions (Table 1).
Th e amount of text devoted to each modality varies according to perceived familiarity with the technique: more text is dedicated to novel strategies and those with newer indications.

Gas exchange Pulse oximetry and transcutaneous carbon dioxide monitoring
Pulse oximetry is widely used in anesthesiology and intensive care and, in intensive care unit (ICU) patients, has a bias of less than 1% and a good to moderate precision [1]; accuracy decreases in hypoxemia (oxygen saturation as measured by pulse oximetry, or SpO 2 , of less than 90%). Among the intrinsic limitations of pulse oximetry are that it is insensitive to changes in arterial partial pressure of oxygen (PaO 2 ) at high PaO 2 levels and cannot distinguish between normal hemoglobin and methemo globin or carboxyhemoglobin. Nail polish may aff ect the measure ment by about 2% (not really clinically relevant) [2], and pulse oximetry can slightly underestimate arterial oxygen saturation (SaO 2 ) in patients with darkly pigmented skin [3]. Altered skin perfusion and carboxyhemoglobin can also lead to inaccurate pulse oximetry readings. Th e type of probe can make a diff erence, and accuracy is usually better for fi nger than for earlobe probes [4]. False alarms are common, usually because of motion artifacts, particularly in the pediatric population.
Pulse oximetry readings should be used to provide an early warning sign, decreasing the need for blood gas measurements. In a randomized controlled trial in more than 20,000 surgical patients [5], pulse oximetry was not associated with decreased postoperative complications or mortality, but 80% of the anesthesiologists felt more secure when a pulse oximeter was used! Transcutaneous partial pressure of carbon dioxide (PCO 2 ) monitors have also been developed with probes generally placed on the earlobe. Precision of trans cutaneous PCO 2 measurements has improved as tech nology has advanced, and devices have become smaller but still need regular recalibration [6]. Th eir place in the respiratory monitoring of ICU patients has not yet been defi ned.

Volumetric capnography and dead space calculation
Th e expiratory capnogram provides qualitative information on the waveform patterns associated with mechanical ventilation and quantitative estimation of expired CO 2 . Capnography tracings show three phases (Figure 1) [7]: phase I contains gas from the apparatus and anatomic dead space (airway), phase II represents increasing CO 2 concentration resulting from progressive emptying of alveoli, and phase III represents alveolar gas. Phase III is often referred to as the plateau and its appearance is relatively fl at or has a small positive slope; the highest point is the end-tidal PCO 2 (PetCO 2 ). Th e almost rectangular shape of the expired capnogram depends on the homogeneity of the gas distribution and alveolar ventilation. Lung heterogeneity creates regional diff erences in CO 2 concentration, and gas from high V/Q regions fi rst appears in the upper airway during exhalation. Th is sequential emptying contributes to the rise of the alveolar plateau; the greater the V/Q heterogeneity, the steeper the expired CO 2 slope. Accordingly, the slope of the alveolar plateau has been shown to correlate with the severity of airfl ow obstruction [8].
Physiologic dead space (Vd phys ) can be easily calculated from the Enghoff modifi cation of the Bohr equation by using arterial partial pressure of carbon dioxide (PaCO 2 ) with the assumption that PaCO 2 is similar to alveolar PCO 2 : Vd phys / V T = (PaCO 2 − P E CO 2 ) / PaCO 2 , where V T is the tidal volume and P E CO 2 is the partial pressure of CO 2 in mixed expired gas and is equal to the mean expired CO 2 fraction multiplied by the diff erence between the atmospheric pressure and the water vapor pressure. Vd phys is increased in acute respiratory distress syndrome (ARDS), and a high dead space fraction represents an impaired ability to excrete CO 2 because of any kind of V/Q mismatch. Several authors [9,10] have demonstrated that increased Vd phys values are independently associated with an increased risk of death in these patients. Since Vd phys /V T measures the fraction of each tidal breath that is wasted on alveolar dead space (Vd alv ) and airway dead space (Vd aw ), the Vd aw must be subtracted from Vd phys /V T to obtain the Vd alv /V T [11]. By using the PetCO 2 instead of P E CO 2 in the equation, the Vd alv can be calculated. Equating the alveolar PCO 2 to the arterial PaCO 2 is, however, valid only in healthy subjects. In patients with high right-to-left shunt, PaCO 2 is higher than alveolar PCO 2 because of the shunted blood with high PvCO 2 (partial pressure of carbon dioxide in mixed venous blood). Without correction [12], it must be remembered that a high dead space also includes the shunt eff ect. PaCO 2 can be grossly estimated by PetCO 2 . Monitoring PetCO 2 can also help to track PaCO 2 when changes in PaCO 2 are to be avoided (especially in critically ill, neurological patients with normal lungs). Th e gradient between PaCO 2 and PetCO 2 widens in ARDS and correlates across the diff erent levels of Vd phys [13]. Th e diff erence between PaCO 2 and PetCO 2 is reduced by using the maximal values of PetCO 2 over time [14].
In patients with sudden pulmonary vascular occlusion due to pulmonary embolism, the resultant V/Q mismatch produces an increase in Vd alv . When volumetric capnography is used as a bedside technique, the association of a normal D-dimer assay result plus a normal Vd alv is a highly sensitive screening test to rule out the diagnosis of pulmonary embolism [15]. Volumetric capnography has also been shown to be an excellent tool for monitoring thrombolytic effi cacy in patients with pulmonary embolism [16].
When the application of positive end-expiratory pressure (PEEP) results in global lung recruitment, physio logic and alveolar dead space decrease [17]; the reverse is true when PEEP application results in lung overdistension [18]. Th erefore, volumetric capnography may also be helpful to identify overdistension or better alveolar gas diff usion [19].
In summary, volumetric capnography has important potential for monitoring the diffi cult-to-ventilate patient. Volumetric capnography needs sophisticated equipment and this has limited its widespread use.

Blood gases
Th e PaO 2 /inspired fraction of oxygen (PaO 2 /FiO 2 ) ratio is still the most frequently used variable for evaluating the severity of lung failure and is included in the current defi nition of acute lung injury/ARDS [20]. Th e PaO 2 /FiO 2 ratio is often a curvilinear (U-shaped) relationship, being at its lowest for moderate ranges of FiO 2 , depending on the shunt level, the hemoglobin value, and the arteriovenous diff erence in O 2 content [21][22][23]. For a given PaO 2 /FiO 2 ratio, the higher the FiO 2 , the poorer the prognosis [24]. In patients with ARDS, the PaO 2 /FiO 2 ratio is dependent on the PEEP level and can be a surrogate, though imperfect, marker of recruitment [25]. Hemodynamic status (via the mixed venous oxygen tension, or PvO 2 ) and intracardiac shunt (patent foramen ovale) also infl uence the PaO 2 /FiO 2 ratio [26]. Despite its limitations, this ratio remains the most commonly used means of assessing severity of lung disease. Th e oxygen index ([mean airway pressure × FiO 2 × 100]/PaO 2 ) accounts better for the infl uence of ventilator pressures on oxygenation value [27].

Extravascular lung water
Extravascular lung water (EVLW) is a quantitative measure of pulmonary edema and is correlated, in multiple patient populations, to mortality [32]. Normal values are 5 to 7 mL/kg (indexed to predicted body weight), and quantities above 10 mL/kg are associated with adverse clinical outcomes [33].
Indicator dilution techniques for measuring EVLW are available for bedside use in critically ill patients. Th e single-indicator technique is now well validated and off ers the additional value of simultaneously measuring cardiovascular performance (cardiac output, fl uid responsive ness, and fi lling volumes). Current technology uses an injection of cold saline into the right atrium and assesses transpulmonary thermodilution in the arterial system by using a femoral or brachial catheter. Phase I contains gas from the apparatus and anatomic dead space (airway), phase II represents increasing carbon dioxide concentration resulting from progressive emptying of alveoli, and phase III represents alveolar gas. The highest point of phase III is the end-tidal partial pressure of carbon dioxide (PetCO 2 ). PaCO 2 , arterial partial pressure of carbon dioxide; PCO 2 , partial pressure of carbon dioxide. Limita tions of the technique include requirements for good indicator mixing without loss and for constant blood fl ow and temperature. EVLW can be assessed only in perfused areas of the lung [34].

Exhaled
EVLW measurements may be used in combination with other cardiovascular and pulmonary parameters to diagnose pulmonary edema. Complementary information from indicator dilution techniques, such as cardiac fi lling volumes, helps to diff erentiate between hydrostatic/cardiogenic pulmonary edema and permeability edema [35]. Although repeated measures could be used to assess response to interventions [35], it is unclear how fast the response time is and whether this technique can be used as a tool to guide therapy.

Respiratory mechanics Compliance and resistance
Monitoring airway pressures can provide important information. In fl ow (volume)-controlled mode, peak airway pressure is determined by both resistance and compliance -a high peak pressure with a much lower plateau pressure indicates a high resistance related either to the patient (bronchospasm) or to the equipment (small-diameter endotracheal tube [ETT] or narrow or obstructed tubing). Plateau pressure measurement requires a pause at end-insuffl ation of at least 200 to 500 ms with a quasi-steady pressure. Longer pauses may be required to precisely estimate lung homogeneities or pendelluft phenomena but their clinical role is unclear. Patients must be relaxed during this measurement.
Compliance is easily calculated as the ratio between V T and plateau pressure minus PEEP ( Figure 2). Elastance is the reverse of compliance (how much pressure we need for a given volume). A low compliance-high elastance refl ects mainly a small aerated lung available for ventilation. High plateau pressure may be related to either low compliance or high end-expiratory pressure (fl ow limitation or dynamic hyperinfl ation). Peak airway pressure is very sensitive to changes in respiratory mechanics; performing end-inspiratory and end-expiratory pauses may allow the exact cause of a high peak airway pressure to be determined. Compliance is not easily assessed on pressure-predetermined modes, especially when the expiratory phase starts before fl ow stops. In such cases, an end-inspiratory occlusion test should be performed to assess plateau pressure, even in time-cycled pressurelimited modes.
Measurements of respiratory mechanics are simple to perform and provide useful and relevant information for severity assessment and ventilator management. Th ey are really reliable only in passive conditions of ventilation, in which plateau pressure monitoring is essential for adequate ventilatory management.

Pressure/volume curves
Th e study of lung mechanics is particularly helpful in patients with ARDS. Study of pressure/volume (P/V) curves requires insuffl ations at very low fl ow to avoid the infl uence of the resistive components [36][37][38]. Th e accent has often been placed on identifi cation of the lower (LIP) [39] and upper (UIP) [40] infl ection points on the P/V curve, but this approach has limitations. First, identifying the LIP or UIP is sometimes diffi cult. Second, recruitment takes place throughout the P/V curve [41,42], and recruitment and overdistension can occur at the same time. Th ird, application of an optimal PEEP level should, ideally, be assessed from the expiratory rather than the inspiratory limb of this relationship.
Interpretation of P/V curves is diffi cult in the presence of altered chest compliance [43]. Chest wall compliance may be decreased in cases of increased abdominal pressure, thoracic trauma, large pleural eff usions, obesity, and so on. Measuring esophageal pressure (surrogate of pleural pressure) allows pressure dissipated through the chest wall to be diff erentiated from pressure distending the lungs (transpulmonary pressure). In medical patients, the chest wall has little to modest impact on respiratory pressures [43]; whether this is diff erent in patients with abdominal surgery or obesity needs further study. Nevertheless, the concept remains that ventilating down to too low a pressure may result in so-called atelectrauma (open ing and closing the alveoli repeatedly), and infl ating the lungs too much when most of the recruitment has already occurred may result in overdistension. The diff erence between peak or maximal pressure (P max ) and plateau pressure (P plat ) defi nes the resistive pressure, whereas the diff erence between P plat and positive end-expiratory pressure (PEEP) defi nes the elastic pressure. Analysis of the airway pressure shape during the phase of constant fl ow infl ation (removing initial and fi nal parts) can be used to calculate the stress index (arrow).
Th e diff erence between the inspiratory and expiratory parts of the P/V curve are related, in part, to hysteresis [44], which refl ects whether PEEP should be increased or not. If the two limbs of the curve are superimposed, increasing PEEP will not help; if there is a large diff erence in volume between inspiratory and expiratory portions, PEEP may help ( Figure 3). Quantifi cation of recruitment requires multiple P/V curves [45], and, although P/V curves are now more frequently available on commercial ventilators, the lack of an estimate of recruitment still limits clinical usefulness. Th e P/V curve technique has thus been used mainly as a research tool.
During constant fl ow insuffl ations, a stress index ( Figure 2) can be calculated from the shape of the airway pressure-versus-time curve (which is essentially the opposite of the P/V curve since during constant fl ow time equals volume) [46]. If there is down ward concavity, compliance improves over time (stress index of less than 1), refl ecting tidal recruitment of collapsed alveoli; if the curve is straight (stress index of 1), compliance is constant, refl ecting ventilation of the normal lung; and if there is upward concavity (stress index of greater than 1), it means that compliance is decreasing over time during insuffl ations, refl ecting overinfl ation. A stress index of less than 1 may suggest a need to increase PEEP; a stress index of greater than 1 may suggest a need to reduce V T [47]. Th e same limitations described for the P/V curve (that is, recruitment and overdistension) apply to this kind of analysis. Th e clinical place and reliability of the stress index are still debated [48].

Diaphragmatic function
Mechanical ventilation has been associated with ventilator-induced diaphragmatic dysfunction [49]. Diaphragmatic function can be altered early and is related to the duration of mechanical ventilation [50]. Th e transdiaphragmatic pressure diff erence (gastric minus esophageal pressure) refl ects diaphragmatic func tion but only in patients who have spontaneous ventila tory breaths and who can cooperate. Magnetic phrenic stimulation can be used to assess diaphragmatic function [51] as a noninvasive method in sedated and non-sedated patients but remains a test of respiratory muscle function rather than a monitoring tool and is used mainly in research.
Measurements of diaphragmatic electrical activity are now possible and have been used to drive the ventilator during neurally adjusted ventilatory assist [52]. Although it does not provide absolute values, monitoring diaphragmatic electrical activity may be of potential interest to detect patient-ventilator asynchrony.

Pressure and fl ow monitoring to assess asynchrony
A considerable amount of information can be obtained from pressure and fl ow time curve analysis [53]. Th e airfl ow trace can reveal the presence of auto-PEEP, when fl ow does not return to zero at the end of expiration ( Figure 4). Dyssynchrony can be caused by poor or delayed ventilator triggering or cycling or both. Excessive levels of pressure support may result in ineff ective triggering because they are associated with long inspiratory times and intrinsic PEEP [54], and insuffi cient assistance (for example, because of a short inspiratory time during assist/control ventilation) can also result in dyssynchrony. Auto-cycling, which results in excessive assistance and can be due to excessive triggering sensitivity or leaks, is diffi cult to detect. It may be revealed by reducing trigger sensitivity during a short series of 'test' breaths. Decreasing levels of pressure support and increasing expiratory trigger are the most eff ective solutions for ineff ective eff orts, whereas applying some PEEP may help but does not always work [55].
Recognizing dyssynchrony is important because it can indicate dynamic hyperinfl ation and may lead to excessive ventilatory assistance [55] and induce delays in weaning from mechanical ventilation [56] and severe sleep disruption [57]. Th ere is no automatic method to detect dyssynchrony. Because of the clinical importance of dyssynchrony, one must learn how to recognize it from traces on the ventilator (this can be relatively easy, at least for gross asynchronies) [56] (Figure 5), and improved bedside training of curve reading is needed. Electromyo graphy can also be of use in determining the presence of dyssynchrony but is rather complex for clinical use [58].

Work of breathing
Work of breathing (WOB) represents the integral of the product of volume and pressure. It represents the energy associated with a given V T at a given pressure (spontaneous, mechanical, or both) [59]. Th e airway pressure is the pressure of the whole respiratory system (lungs plus chest wall); the transpulmonary pressure is the pressure needed to distend the lung parenchyma (airway pressure minus the pleural pressure); fi nally, the pleural pressure is the pressure needed to distend the chest wall. In the clinical/physiological setting, esophageal pressure is used as a surrogate for pleural pressure. 'Work' is not the same as 'eff ort' -eff ort without volume generation will not result in increased WOB. Normal WOB values range between 0.2 and 1 J/L.
In paralyzed patients with mechanical ventila tion, plots of airway pressure versus V T indicate the total amount of work needed to infl ate the respiratory system (that is, the work done by the ventilator on the whole respiratory system and the ETT). Th is is not the amount of work performed by the respiratory muscles, for which eso phageal (pleural) pressure measurements are needed. One also needs to know the slope of the passive P/V curve of the chest wall (which denotes the relaxation of the respiratory muscles). Th e surface encom passed within the passive P/V curve of the chest wall and the negative esophageal pressure swing during an inspiratory eff ort is shown in the so-called Campbell diagram [60,61]. Finally, the two components of work (that is, elastic and resistive) can be split by joining the zero fl ow points at the beginning and the end of inspiration ( Figure 6). Th e Campbell diagram allows the true work performed by the respiratory muscles to be estimated under diff erent clinical conditions, even when auto-PEEP is present [60,61].
Calculation of WOB may also be useful in understanding weaning failure. Jubran and colleagues [62] showed that the esophageal pressure trend during a spon taneous breathing trial (SBT) complemented the pre dic tion of weaning outcome provided by the frequency/ V T index measured during the fi rst minute of an SBT. Monitoring WOB can also theoretically help in titrating ventilator support. It could also be used to evaluate the eff ects of diff erent ventilatory modes, understand the mechanisms of disease (weaning failure, acute asthma, and exacer bations of chronic obstructive pulmonary disease (COPD)), and evaluate the eff ects of therapeutic interven tions (for example, bronchodilators [63]) and the infl uence of ventilator performance (triggering, fl ow delivery, and so on). Because it requires esophageal pressure measurement, this technique has been reserved largely for clinical research. It has potential for clinical use but few monitors provide bedside calculations.

Occlusion pressure (P 0.1 )
Th e occlusion pressure, also referred to as P 0.1 , refl ects the respiratory drive to breathe and is correlated to WOB for a given patient. Measurements of P 0.1 , now automatically provided on ventilators, may be useful to assess the patient's response to titration of ventilator settings (that is, fl ow rate, PEEP, and so on) and could be used as a surro gate of WOB to help titrate pressure support or external PEEP in cases of intrinsic PEEP [60,64]. A P 0.1 of less than 2 cm H 2 O is considered normal. Th is measure has been restricted largely to research. However, because the P 0.1 is now more widely available in the ICU and is an extremely simple and rapid way to estimate central respiratory drive, its potential clinical role needs to be evaluated.

Pressure-time product
Th e pressure-time product is the integral of the pressure performed by the respiratory muscles during inspiration or expiration and time or both. Th e pressure-time product is an alternative to WOB and has some theoretical and practical advantages over WOB calculations. Th e pressure-time product is asso ciated with oxygen consumption by the respiratory muscles [65] and could be considered a surrogate to quantify the metabolic expense of respiratory eff ort. Since it is independent from the ability of the patient to generate volume, the pressuretime product is relevant in situations in which there is a disconnection between eff ort and volume (for example, during asynchrony) [66]. Normal values for the pressuretime product range between 60 and 150 cm H 2 O/second per minute [67].

Transpul monary and esophageal pressure
Transpulmonary pressure is the diff erence in pressure between the inside (alveoli) and the outside (pleural space) of the lung. Variations in transpulmonary pressure where P El,lung is transpulmonary pressure and El, lung is the elastance of the lung.
In static conditions (that is, no fl ow), the pressure inside the lung can be easily estimated from P aw (airway pressure), but the pressure outside the lung (that is, the pleural pressure, or P pl ) is not easily measurable and must be estimated from the esophageal pressure (P es ): P El,lung = P aw -P es .
For any change in lung volume, the higher the elastance of the chest wall (El cw ), the greater the contribution of the P pl change to the total P aw change. Th e chest wall and the lung contribute to the change in airway pressure in proportion to their elastance: ΔP aw = ΔP El,lung + ΔP es = ΔVol × El, lung + ΔVol × El cw = ΔVol × (El, lung + El cw ) P El,lung is the real eff ector of lung volume changes and of potential ventilator-induced lung injury (VILI) and is of major importance in the setting of mechanical ventilation [68].
Measurement of P es is not always straightforward, in particular when absolute values are used [69], and some clinicians fi nd P es diffi cult to use. Chiumello and colleagues [70] recently explored the concept of specifi c elastance, which refl ects the intrinsic elastic properties of the lung parenchyma and which relates stress (transpulmonary pressure) and strain (change in volume relative to functional residual capacity (FRC)). Specifi c elastance is rather constant among patients with ARDS (and even healthy subjects) and thus the measurement of endexpiratory lung volume (EELV) could allow an eff ective evaluation of the P El,lung change caused by V T ; that is, once the strain is measured, it is possible to infer the stress. Th e concept that the risk of VILI can be related to the ratio between V T and EELV has been suggested by positron emission tomography studies in patients with ARDS [71].

Inspiratory efforts triggering the ventilator
Specifi c problems arise from the fact that the ARDS lung is non-homogeneous, and some areas, possibly entire lobes, are not exposed to airway pressure because of collapse, whereas the boundary regions between ventilated and collapsed areas may be exposed to high distending pressures, potentially causing VILI. Attention should also be paid when applying these concepts to assisted rather than controlled breathing conditions [72]. Th e pressure developed by the inspiratory muscles and by the diaphragm may cause negative swings in pleural pressure, bringing the transpulmonary pressure to levels well beyond the VILI threshold. Th e eff ects of huge inspiratory eff orts have only occasionally been investigated in patients with ARDS [73] but are known to cause lung edema in the experimental animal and in airwayobstructed patients. Decreased pleural pressure has been shown to be associated with cardiovascular failure during weaning [74], possibly because of increased transdiaphrag matic pressure and right heart overload [75].
In clinical practice, transpulmonary pressure estimate poses two problems. First, like WOB and the pressuretime product, it requires correct placement of an esophageal probe. Th e availability of nasogastric feeding tubes with esophageal balloons should greatly facilitate the use of these techniques. Second, ensuring the validity of the absolute value of esophageal pressure in a supine patient for estimating end-inspiratory or end-expiratory transpulmonary pressure values remains diffi cult.

Abdominal pressure
Increased intra-abdominal pressure (IAP) can decrease compliance of the lung and chest wall and increase dead space and shunt fraction (Qs/Qt). Increased IAP reduces the impact of trans pulmonary pressure as the driving force for alveolar opening and prevention of closing. Th ere is some relation ship between abdominal and pleural pressures. IAP can be assessed fairly simply by using a bladder catheter [76], and given that high IAPs can have conse quences in terms of diagnosis and management, more regular measurement of IAP is recommended.

Lung volumes Direct measurement of end-expiratory lung volume
ARDS is associated with a marked reduction in lung volume [77]. Monitoring of FRC can provide information to assess pulmonary function. When the closed dilution technique is used, the patient breathes in a fi xed concentration of helium or methane mixed with oxygen and the concentration in the expired breath can be used to calculate the FRC. Th is technique is used for research purposes. An alternative approach is a washout/washin technique using nitrogen or oxygen. Olegard and colleagues [78] reported that, by changing the FiO 2 abruptly by as little as 0.1, the FRC could be calculated by using standard gas-monitoring equipment. Th e precision of this method seems acceptable, and the method can be used even in the most severely hypoxemic patients [79,80].
FRC is sex-, length-, and age-dependent. Ibanez and Raurich [81] showed that FRC decreased by 25% after changing the position from sitting to supine in healthy volunteers. Bikker and colleagues [82] found a reduction of 34% in mechanically ventilated patients with 'healthy' lungs and attributed this to the loss of muscle tension with sedation in ICU patients. In critically ill patients receiving mechanical ventilation and diff erent levels of PEEP, it is better to speak of EELV [83]. Application of PEEP leads to increased EELV values as a result of recruit ment or further distention of already ventilated alveoli. To diff erentiate between recruitment and distention, EELV changes can be combined with compliance values [82]. From the compliance calculation, one can determine the expected change in EELV for a given change in PEEP. If application of PEEP leads to a higher EELV, this method can be used to estimate alveolar recruitment at the bedside [84]. Measurement of EELV has been made available recently for routine use. Although we still have limited experience with this technique, it has considerable potential, at least in the management of patients with ARDS.

Chest ultrasonography and computed tomography
Chest ultrasonography can be useful at the bedside for early identifi cation of edema as well as other abnormalities like pneumothorax or pleural eff usion [85,86]. However, this technique requires training. Recently, it was shown that lung ultrasonography can be used to estimate alveolar reaeration in patients treated for ventilator-associated pneumonia [87] and to estimate PEEP-induced lung recruitment [88]. Th is is a relatively new but promising and non-invasive technique that could have important clinical applications in the ICU.
Computed tomography (CT) scanning can be useful to identify ongoing pathology. CT images can be used to compute average lung density and quantitate the respective amounts of air and tissue, but this approach is currently restricted to research [42,89]. CT could potentially have roles in guiding protective mechanical ventilation in ARDS and in appropriately setting V T and PEEP [90]. Th e major limitations are the need to transfer the patient to the CT scanner and the complex processing needed for analysis.

Electrical bioimpedance tomography
In electrical bioimpedance tomography (EIT), a current is applied via 16 electrodes positioned around the thorax. A scan of the impedance to fl ow in a slice of the thorax refl ects changes in aeration but gives no information on EELV and measures only relative impedance without providing absolute values. Images can be subdivided into several regions and can be used to monitor regional ventilation. EIT can be used to show whether a recruitment maneuver has been successful and document the eff ects of positioning and of PEEP application [91,92].
Th e caudal thoracic level above the diaphragm is of particular importance because atelectasis due to mechanical ventilation can be expected at this level. Various studies have described ventilation distribution change maps to evaluate lung collapse or overdistension [93][94][95]. Costa and colleagues [93] introduced an algorithm for estimating recruitable alveolar collapse by using EIT. Bikker and colleagues [94] clearly visualized improvement or loss of ventilation in dependent and nondependent parts of the lung by using EIT measurements in ICU patients. When EIT was used at multiple levels in mechanically ventilated patients, ventilation distribution was shown to shift from the dorsal to ventral region but also from the caudal to cranial level during a decremental PEEP trial [95].
Th e increase in airway pressure during normal inspiration is followed by a continuous redistribution of gas from non-dependent to dependent regions. Th us, during the initial phase of inspiration, most of the inspiratory gas goes to the non-dependent lung, and during the last parts of inspiration, the most dependent parts receive the inspiratory fl ow, especially in patients with a positive response to a recruitment maneuver. Th is technique may thus represent a means of identifying responders and non-responders to recruitment during normal tidal ventilation, enabling one to avoid exposing non-responders to high-pressure recruitment maneuvers. Additional clinical work is now needed to delineate the place of EIT in the ICU as a qualitative tool to visualize ventilation distribution or as a quantitative technique to estimate the eff ects of interventions.

Cardiopulmonary interactions Hemodynamic monitoring
Th e use of hemodynamic monitoring in the unstable critically ill patient was reviewed recently [96]. Hemodynamic monitoring is particularly helpful during difficult weaning processes to separate cardiac from pulmonary aspects of failed weaning. Th e heart may not be able to meet the increased oxygen demand during weaning and then cardiac fi lling pressures generally increase and the mixed venous oxygen saturation (SvO 2 ) (or central venous oxygen saturation) decreases. SvO 2 is a rather non-specifi c but sensitive kind of monitoring, as its changes refl ect a change in one or more of the major homeostatic systems (respiration, circulation, and energy demand). During the weaning process, a reduction in SvO 2 may be expected as spontaneous breathing represents a form of exercise, but a decrease in SvO 2 may refl ect the inability of the heart to face the increased oxygen demand, especially if arterial hypertension is present [97].
Echocardiographic evaluation may be helpful in acute respiratory failure, simply to identify a dilated right ventricle (RV) or RV failure, which may necessitate a decrease in PEEP or V T (or both) to reduce RV afterload. Measurement of pulmonary artery pressure can also be reliably obtained by Doppler measurements [98]. Echography can be particularly helpful just before a suspected diffi cult SBT and sometimes during the SBT. Patients at risk of weaning failure were identifi ed as having decreased ejection fraction and increased fi lling pressures before SBT [99]. Echo can help to estimate pulmonary artery occlusion pressure elevation during SBT. Th e limitations of echocardiography are that it requires some training and is time-consuming, but it is an increasingly useful tool for cardiorespiratory monitoring in the ICU.
Cardiac biomarkers, like B-type natriuretic protein (BNP) or N-terminal prohormone of BNP (NT-proBNP), may be useful for diagnosing heart dysfunction but also for monitoring purposes, especially during weaning from mechanical ventilation [100,101].

Lung infl ammation Bronchoalveolar lavage studies
Bronchoalveolar lavage (BAL) can be used to assess hemorrhage and measure neutrophils, eosinophils, hyaline membranes, lipid inclusion, and cancer cells (although this requires a careful cytologic examination of the alveolar fl uid sampled). BAL fi brocyte levels are elevated in ARDS and related to outcome, and levels higher than 6% were observed in non-survivors [102]. BAL fl uid analysis may also help to identify patients who may respond to steroids. Many studies have shown higher levels of infl ammatory mediators (cytokines and so on) in BAL fl uid of non-survivors than survivors of ARDS. Measuring cytokines or phosphorylation products may help to identify VILI, but there is a high signal-to-noise ratio at present. Among the limitations of BAL fl uid analysis are that it needs an endoscope and requires some training, and there is no standardized technique (depends on the volume instilled and amount of fl uid returned). A potential complication is hypoxemia. Detailed BAL analy sis is suitable for bacteriological purposes, but detailed cytological or biomarker assessment is often reserved for clinical trials.

Respiratory monitoring in acute respiratory distress syndrome
As already mentioned, assessment of the severity of ARDS should include not only oxygenation but also dead space estimate and lung mechanics. Monitoring of plateau pressure, as a refl ection of the maximal alveolar pressure, is essential. Potentially important tools for the most severe patients include esophageal pressure and lung volume measurements. P/V curves with assessment of recruitability could also be potentially useful. EVLW estimates may help in diff erential diagnosis.

Respiratory monitoring in chronic obstructive pulmonary disease/asthma
In the initial phase of mechanical ventilation, a detailed assessment of respiratory mechanics, including plateau pressure and auto-PEEP, is essential to characterize the patient. When patients are switched to triggered breaths, detection of asynchrony is very useful in titrating ventilator assistance.

Respiratory monitoring in non-invasive ventilation
Respiratory monitoring in patients receiving noninvasive ventilation (NIV) should start with a full clinical assessment: dyspnea, respiratory muscle function, comfort, mental alertness, and gastric distention are important signs. Th ese clinical indicators should then be combined with additional objective variables for a full evaluation of respiratory status. Pulse oximetry is essential but does not provide information about PCO 2 . Expired CO 2 can be measured (from the mask or helmet), but leaks often make these measurements unreliable. Transcutaneous capnometry off ers a continuous and noninvasive method of monitoring alveolar ventilation [103]. Arterial blood gases remain important in assessing the response to therapy.
Patient-ventilator asynchrony occurs in almost one half of patients and may be related largely to leaks [104]. Clinical evaluation (tachypnea, accessory muscle activity, agitation, and lack of cooperation) or waveform analysis can be useful in assessing the presence of patientventilator dyssynchrony. Electromyography tracings are cumbersome, but airway pressures and fl ows can be used to monitor patient eff orts and identify 'autotriggering' , premature cycling, or ineff ective triggering [105]. Inspired and expired V T values can help to identify air leaks.
In the hypoxemic patient, respiratory monitoring should also help to identify when to intubate the trachea (not too late!). Shock, including measurement of blood lactate levels, should also be looked for since the presence of acute circulatory failure is an exclusion criterion for NIV [106].

Respiratory monitoring in the neurological patient
Neurological dysfunction is one of the most frequent reasons for initiating mechanical ventilation [107]. Outcomes of critically ill neurological patients are driven mainly by the underlying neurological pathology [108,109], and the infl uence of extracerebral organ dysfunction and ventilatory management on outcomes in this group of patients is not well established [110].
In mechanically ventilated neurological patients, no consensus has been reached about optimal V T , PEEP, PaO 2 , or PaCO 2 levels [111], largely because these patients have been universally excluded from randomized trials of lung-protective ventilation because of concerns about potential intracranial pressure (ICP) increases due to hypercapnia or increased thoracic pressures. Moreover, owing to persistently decreased levels of conscious ness, typical weaning and liberation techniques used in medical-surgical ICU patients may not apply to this group [109,112]. Finally, tracheostomy is commonly imple mented as part of the management of these neurological patients, but the technique of choice and timing are controversial [109].
Pulmonary hyperventilation rapidly reduces ICP by reducing cerebral blood fl ow (CBF). Th e eff ect on ICP is not sustained, whereas CBF may remain low, raising risks of ischemia [113,114]; hyperventilation should, therefore, be avoided during the fi rst 24 hours after injury, when CBF is often already low [115]. Th e CBF level at which irreversible ischemia occurs has not been clearly established, but ischemic cell change has been demonstrated following traumatic brain injury and is likely to occur when CBF decreases to less than 15 to 20 mL/100 g per minute. Hyper ventilation should, therefore, be used only for short-term management of raised ICP and only in patients with life-threatening intracranial hypertension [115,116]. If hyper venti lation is used, jugular venous oxygen saturation (SjO 2 ) or brain tissue oxygen tension (PbrO 2 ) monitoring should be used to evaluate oxygen delivery where possible [115]. Th e ideal value for PaCO 2 is one that keeps ICP to less than 20 mm Hg and cerebral extraction of oxygen (CEO 2 ) to between 24% and 42% to avoid brain ischemia.
Moderate levels of PEEP (for example, less than 15 cm H 2 O) can be safely used in patients with cerebral lesions, mainly in those with low pulmonary compliance [117]. Hence, although care should be taken when applying PEEP in patients with neurological injury, it should not be withheld if needed to maintain adequate oxygenation [114]. Multimodal brain monitoring, including brain tissue oxygen tension, CBF measurement, and intracerebral microdialysis (with measurement of lactate, pyruvate, glutamate, glycerol, and infl ammatory mediators), may be useful to optimize mechanical ventilation in patients with severe brain injury [118].

Conclusions and perspectives
Although most of the clinical interventions applied to the respiratory system of the critically ill are relatively simple, they are often misused, largely as a result of a poor understanding of the physiology under lying a specifi c intervention or of its consequences on the patho physiology of respiratory disease or both. We need to encourage increased training in these techniques guided by a better knowledge of underlying mechanisms and appropriate use and correct interpretation of the data provided by available monitoring techniques.
Importantly, two key concepts can be highlighted: 1. Multiple variables need to be integrated: Many critically ill patients with respiratory failure have complex pathologies, and the respiratory failure is part of a broader multiple organ failure. Management will, therefore, require the consideration of monitored data from several diff erent systems. Th is is one reason why it is diffi cult to develop protocols for the use of mechanical ventilation. 2. Monitoring of solitary static values provides limited information; evaluation of dynamic changes in variables over time is more important. Th e future is likely to see advances in neuro-ventilatory coupling to limit the occurrence and adverse eff ects of patient-ventilator asynchrony. Biomarker panels will be developed to determine the risk of ARDS and VILI and to guide therapy. Chest ultrasonography and other, less invasive techniques will also be used more as they become more readily available at the bedside and training is improved.

Competing interests
Over the last fi ve years, L Brochard's laboratory has received research grants for specifi c research projects from the following companies: Dräger (SmartCare; Lübeck, Germany), Maquet (NAVA; Rastatt, Germany), Covidien (PAV+; Dublin, Ireland), General Electric Company (FRC; Fairfi eld, CT, USA), Fisher & Paykel (Optifl ow; Auckland, New Zealand), and Philips Respironics (NIV; Murrysville, PA, USA). GSM and FJB serve on the medical advisory board for Pulsion Medical Systems (Munich, Germany). L Blanch holds €1,001 to €5,000 in stock ownership in Better Care S.L., a spinoff of Corporació Sanitària Parc Taulí (Sabadell, Spain). JM has received research grants from Covidien and General Electric Company. J-CR has received research grants from Dräger, Covidien, and General Electric Company. DG is a member of the critical care medical advisory board of GE Healthcare (Madison, WI, USA) and has received speaking fees from Dräger and GE Healthcare. AV-B has received research grants from General Electric Company and Maquet. The research laboratory of SJ has received research grants from Dräger Medical France and Maquet France (Ardon, France). OS has received lecture fees and travel expenses from Dräger Medical and GE Healthcare and has received payment for a patent application from Dräger Medical. The other authors declare that they have no competing interests.