Resolution of a patient’s distress with the onset of bagging indicates that the problem lies within the ventilator or its external circuit (see Fig. 53-1). Malfunction of the ventilator may arise from tubing connected to a wrong outlet; a poor fit of connections; uncoupling of connections; defective material; obstruction of the circuit secondary to kinks, intraluminal fluids, or a malfunctioning valve; or malfunction of microprocessor controls. If malfunction is suspected, the ventilator should be replaced while each component is checked against the schematic circuit diagram. When a patient’s delivered VT is adequate and distress is relieved by manual ventilation, a fault in the setting of the fractional inspired oxygen concentration should be suspected. This problem can be verified by obtaining an independent direct measurement of inspired oxygen concentration.
External Ventilator Circuit
The ventilator circuit primarily consists of the tubing and connectors that form inspiratory and exhalation limbs; the humidifier; the exhalation valve assembly and flow measuring devices; adaptors placed within the circuit for monitoring or delivering medications; and inline filters.13 Problems within this circuit that can cause patient distress include leaks, disconnects, accumulation of condensate, and inline nebulizers.
A leak may arise in parts of the circuit assembly that screw together. Leaks or uncoupling of connections will cause sounding of the low pressure and low VT alarms. The Y-connection to the tracheal tube is the most common site of a disconnection. Other locations of disconnections and leaks include the humidifier system, an incompetent exhalation valve assembly, disconnection of the proximal line that connects the pressure tap at the Y-piece to the ventilator manometer, or small ruptures in the tubing.
In patients ventilated with pressure support, a leak at any location in the circuit, including at the cuff of the endotracheal tube, predisposes to a unique problem.94 During pressure support, the ventilator strives to maintain a preset level of pressure throughout inspiration. A leak, however, will tend to cause the airway pressure to fall. To prevent the fall in airway pressure, the ventilator increases inspiratory flow. The algorithm employed by many ventilators for terminating the time of lung inflation is a fall in inspiratory flow to an absolute value of 5 L/min (or a fall in flow of 75% from the peak value). The increase in flow being delivered by the ventilator means that inspiratory flow never falls to the threshold required for termination of inflation. This phenomenon results in the unremitting application of positive pressure, which is relieved by correcting the leak.
Gas supplied to the inspiratory limb is typically warmed to 32°C (89.6°F) to 34°C (93.2°F) and fully saturated with water vapor. The gas cools as it passes through the inspiratory tubing, causing condensate to form. The condensate accumulates in a U-loop of the circuit. If enough condensate accumulates, movement of the water can cause the ventilator to trigger. During assist-control ventilation, excessive triggering can lead to barotrauma or hemodynamic compromise. During pressure-support or pressure-controlled ventilation, the resistance caused by the condensate may cause a decrease in achieved VT for a set airway pressure.
The insertion of a continuous-flow nebulizer between the patient and the pressure sensor within the ventilator can lead to hypoventilation during pressure support or intermittent mandatory ventilation (IMV).95 When a patient’s mean inspiratory flow rate is less than the flow of gas used to power the nebulizer (6 to 10 L/min), airway pressure will not fall sufficiently to trigger the ventilator. Moreover, the continuous flow from the nebulizer creates a bias flow, which the ventilator interprets as forming part of the patient’s minute ventilation. Consequently, the low minute volume alarm will not sound.
Patient–ventilator dyssynchrony can lead to considerable patient distress, and it also impedes the effectiveness of the ventilator in decreasing respiratory work. For the most effective unloading of the inspiratory muscles, the ventilator should cycle in synchrony with the patient’s central respiratory rhythm. For perfect synchronization, the period of mechanical inflation must match the period of neural inspiratory time (the duration of inspiratory effort), and the period of mechanical inactivity must match the neural expiratory time.96,97 The interplay between the ventilator and the respiratory neuromuscular apparatus is complex, and problems can arise at several points in the respiratory cycle: the onset of ventilator triggering, the remainder of inspiration after triggering, the switch from inspiration to expiration, and the end of expiration.42
Triggering of the Ventilator
Patients reach the set sensitivity by activating their inspiratory muscles. For a given set sensitivity, the delay between onset of patient inspiratory effort and onset of ventilator assistance is a function of a patient’s respiratory motor output. When a patient’s respiratory drive is low, assistance may not begin until well into the patient’s inspiratory time, thereby causing the ventilator to cycle almost completely out of phase with the patient’s respiratory cycle. When the threshold to open the demand valve is reached and the ventilator starts to provide positive-pressure assistance, the inspiratory neurons do not simply switch off, and a patient may expend considerable inspiratory effort throughout the machine-cycled inflation.98 The level of patient effort during this post-trigger phase is closely related to a patient’s respiratory motor output at the point of triggering.99 As such, measures that decrease respiratory drive may enhance respiratory muscle rest during mechanical ventilation.
If respiratory motor output at the point of triggering is important, one might expect that effort during the time of triggering would determine patient effort during the remainder of inspiration.100 To investigate this issue, Leung et al99 applied graded levels of pressure support in eleven critically ill patients. They achieved a fourfold reduction in overall patient effort. Yet patient effort during the time of triggering did not change. The constancy of effort during the trigger phase was probably secondary to different factors becoming operational as the level of ventilator assistance was varied (Fig. 53-15). Thus, increases in the level of ventilator assistance do not substantially decrease patient effort during the time of triggering.
Graded increases in pressure support produced a decrease in total pressure-time product (PTP) per breath (blue symbols), although PTP during the trigger phase (red symbols) did not change (left panel). The constancy of PTP during triggering probably resulted from different factors becoming operational at different levels of assistance (right panel). At low levels of pressure support, respiratory motor output (dP/dt green symbols) and intrinsic positive end-expiratory pressure (PEEPi) were high but triggering time was short (brown symbols), resulting in a large change in pleural pressure over a brief interval. At high levels of pressure support, dP/dt and PEEPi were low but triggering time was long, resulting in a smaller change in pleural pressure over a longer time. (Based on data from Leung et al.91)
At high levels of mechanical assistance, up to one-third of a patient’s inspiratory efforts may fail to trigger the machine (see Fig. 53-12).99,101,102 The number of ineffective triggering attempts increases in direct proportion to the level of ventilator assistance.99 Surprisingly, unsuccessful triggering is not the result of poor inspiratory effort. In a study of factors contributing to ineffective triggering, a decrease in the magnitude of inspiratory effort at a given level of assistance was not the cause; indeed, effort was 38% higher during nontriggering attempts than during the triggering phase of attempts that successfully opened the ventilator valve.99 Significant differences, however, were noted in the characteristics of the breaths before the triggering and nontriggering attempts. Breaths before nontriggering attempts had a higher VT than did the breaths before triggering attempts, 486 ± 19 and 444 ± 16 mL, respectively, and a shorter expiratory time, 1.02 ± 0.04 and 1.24 ± 0.03 seconds, respectively. An abbreviated expiratory time does not allow the lung to return to its relaxation volume, leading to an increase in elastic recoil pressure. Indeed, PEEPi was higher at the onset of nontriggering attempts than at the onset of triggering attempts: 4.22 ± 0.26 versus 3.25 ± 0.23 cm H2O. Thus, nontriggering results from premature inspiratory efforts that are not sufficient to overcome the increased elastic recoil associated with dynamic hyperinflation.99
When triggering fails as a result of dynamic hyperinflation, the nontriggering on that breath allows the lungs to more completely empty in preparation for the next breath. Occasionally, two or three failed triggering attempts take place before triggering becomes successful. Because there is no lung inflation during a failed triggering attempt, mechanical exhalation continues for a longer time and end-expiratory volume continues to fall until triggering becomes successful. This pattern may occur repeatedly such that it resembles the Wenckebach pattern of atrioventricular block on an ECG.103
In addition to an increase in elastic recoil pressure,99 an elevated PEEPi can also result from an increase in expiratory muscle activity. Parthasarathy et al97 investigated the relative contributions of these two factors to ineffective triggering in healthy subjects receiving pressure support and in whom they induced airflow limitation with a Starling resistor. Nontriggering was linked to the fraction of PEEPi caused by elastic recoil but not to the fraction caused by expiratory effort. This observation suggests that external PEEP might be clinically useful in reducing ineffective triggering.
Although the magnitude of expiratory effort does not appear to influence the success of triggering attempts, the time that expiratory efforts commence in relation to the cycling of the ventilator is an important factor. Parthasarathy et al97 quantified the relationship between the onset of expiratory muscle activity, measured with a wire electrode in the subject’s transversus abdominis, and the termination of mechanical inflation by the ventilator. At pressure support of 20 cm H2O, mechanical inflation was found to continue for a longer time into neural expiration in the breaths preceding nontriggering attempts. Continuation of mechanical inflation into neural expiration counters expiratory flow, and also decreases the time available for unopposed exhalation. Consequently, elastic recoil increases. In turn, a greater inspiratory effort will be needed to achieve effective triggering. In this way, the time that a patient commences an expiratory effort (in relation to cycling-off of mechanical inflation) partly determines the success of the ensuing inspiratory effort in triggering the ventilator.
Some patients exhibit two mechanical inflations within a single neural inspiration, a phenomenon known as double triggering (Fig. 53-16). With assist-control ventilation, double triggering is likely when the set mechanical inspiratory time is substantially less than a patient’s neural inspiratory time. In this situation, mechanical inflation terminates while the patient is still making an inspiratory effort. After a brief period, the ventilator may trigger again, resulting in a second inflation within the same neural inspiration and, thus, greater alveolar distension than with the delivery of a single tidal volume.103
Four incidents of double triggering, each indicated by an arrowhead. Airway pressure (Paw) and esophageal pressure (Pes) in a patient with COPD and pneumonia who was receiving assist-control ventilation at the following settings: tidal volume 600 mL, inspiratory flow 60 L/min, trigger sensitivity −2 cm H2O, and positive end-expiratory pressure 5 cm H2O. The duration of neural inhalation of the double-triggered breaths, roughly equivalent to the width of the associated swings in esophageal pressure, was substantially longer than the neural inhalation of the normally triggered breaths.
With pressure-support ventilation, double triggering is likely when the time constant is short (low resistance, high elastance) and patient neural inspiratory time is relatively long (a slow, spontaneous respiratory rate).104 Following the loss of ventilator pressure after a first triggering attempt, volume decreases because the pressure exerted by the inspiratory muscles (Pmus) alone is not sufficient to sustain elastic recoil. During this phase, the persistence of neural inhalation will cause a progressive increase in Pmus, while elastic recoil continues to decrease. If the patient’s neural inspiratory time is sufficiently long, a point is reached where Pmus will exceed elastic recoil and flow will become inspiratory and double triggering will occur.105
Randomized clinical trials have revealed that use of a high tidal volume (12 mL/kg) is associated with an increased mortality in patients with the acute respiratory distress syndrome.106 Consequently, it has become standard practice to lower the delivered tidal volume in order to minimize alveolar overdistension; for example, to keep the inspiratory pressure during a pause at the end of inspiration (plateau pressure) to 32 cm H2O or lower.106 A low tidal volume is typically accompanied by a short mechanical inspiratory time and thus these patients are especially susceptible to double triggering. Accordingly, conscious attempts to lower tidal volume can paradoxically result in greater alveolar distension than occurs with conventional tidal volume settings (Fig. 53-17). An additional (and often unrecognized) factor that may produce alveolar distension is the tachypnea-associated increase in intrinsic PEEP that accompanies lowering of tidal volume.107
Volume stacking caused by double triggering. Flow (top panel), volume (middle panel), and esophageal pressure (Pes; lower panel) in a patient with COPD receiving assist-control ventilation. During first breath, esophageal pressure remains positive indicating that the patient did not trigger the inflation. During the second breath, esophageal pressure becomes negative indicating active inspiratory effort, which lasts more than 1 second; the duration of mechanical inflation is 0.6 second. The longer duration of neural inspiration as compared with mechanical inflation causes the ventilator to deliver a second breath before there is time for exhalation. As a result, end-inspiratory lung volume increases (breath stacking) with a consequent increase in elastic recoil. The increase in elastic recoil is responsible for the higher peak expiratory flow on the second breath as compared with the first breath.
Setting of Inspiratory Flow
When a patient is first connected to a ventilator, inspiratory flow is set at some default value, such as 60 L/min. Many critically ill patients, however, have an elevated respiratory motor output and the initial flow setting may be insufficient to meet flow demands. As a result, patients will struggle against their own respiratory impedance and that of the ventilator (Fig. 53-18). Consequently, the work of breathing increases. Clinicians sometimes increase flow so as to shorten the inspiratory time and increase the expiratory time. But an increase in flow causes immediate and persistent tachypnea; as a result, expiratory time may be shortened.108 In healthy subjects, Laghi et al109 found that increases in inspiratory flow from 30 L/min to 60 and 90 L/min caused increases in the respiratory rate of 20% and 41%, respectively.
Influence of ventilator flow setting on patient effort. Flow (inspiration directed upward), airway pressure (Paw), and esophageal pressure (Pes) in a patient with respiratory failure who is receiving assist-control ventilation; inspiratory flow is set at 60 L/min in the left panel and at 90 L/min in the right panel. At an inspiratory flow of 60 L/min (left panel), the pronounced negative deflection in airway pressure (patient effort to trigger the ventilator) together with subsequent extensive scalloping signifies that the inspiratory flow delivered by the ventilator is insufficient to meet the high demand. At an inspiratory flow of 90 L/min (right panel), the small negative deflection in airway pressure together with the subsequent smooth convex contour signifies that the delivered flow satisfies the patient’s respiratory drive. Accordingly, the flow of 90 L/min achieved greater unloading of the respiratory muscles, as signaled by the shorter duration of inspiratory effort and the smaller swings in esophageal pressure.
A main reason that clinicians increase inspiratory flow is to decrease inspiratory time, in the hope of allowing more time for expiration and thus decrease PEEPi, especially in patients with COPD. Because increased flow usually leads to an increase in rate, the expected shortening of expiratory time might actually increase PEEPi. Laghi et al110 studied this phenomenon in ten patients with COPD (Fig. 53-19). As with healthy subjects, an increase in flow from 30 to 90 L/min caused the respiratory rate to increase from 16.1 ± 1.0 to 20.8 ± 1.5 breaths/min. Despite the increase in rate, PEEPi fell from 7.0 ± 1.3 to 6.4 ± 1.1 cm H2O. The decrease in PEEPi arose because of an increase in expiratory time, 2.1 ± 0.2 to 2.3 ± 0.2 seconds, which allowed more time for lung deflation. Why did expiratory time increase? An increase in inspiratory flow is usually achieved by shortening mechanical inspiratory time. The shortened inspiratory time combined with time-constant in homogeneity of COPD will cause overinflation of some lung units to persist into neural expiration. Continued inflation during neural expiration causes stimulation of the vagus nerve, which prolongs expiratory time.111,112
Continuous recordings of flow, esophageal pressure (Pes), and the sum of rib cage and abdominal motion in a patient with COPD receiving assist-control ventilation at a constant tidal volume. As flow increased from 30 to 60 and 90 L/min (from right to left; i.e., the opposite of the usual presentation), frequency increased (from 18 to 23 and 26 breaths/min, respectively), PEEPi decreased (from 15.6 to 14.4 and 13.3 cm H2O, respectively), and end-expiratory lung volume also fell. Increases in flow from 30 L/min to 60 and 90 L/min also led to decreases in the swings in Pes from 21.5 to 19.5 and 16.8 cm H2O, respectively. (Used, with permission, from Laghi et al.110)
Mode-Specific Effects of Inspiratory Unloading
Pressure support and IMV are sometimes combined in a given patient. In an international survey of mechanical ventilation,113 this combination tied with assist-control ventilation as the most commonly used mode of ventilation in North America (34% for each). The rationale for combining the two modes is unclear. Presumably, clinicians use pressure support to overcome the work imposed by the endotracheal tube and demand valve during the non-mandatory breaths.
Examining the response of the respiratory centers to this combination of modes provides useful insight into patient–ventilator interaction. A decrease in the number of mandatory breaths produces a decrease in the average VT,99 with inevitable increase in the ratio of dead space to VT. To avoid a decrease in alveolar ventilation, the patients increased respiratory motor output, inspiratory effort, and rate. Adding pressure support of 10 cm H2O caused a decrease in effort at any given IMV rate. The decrease in effort during the mandatory ventilator breaths was related to the decrease in respiratory motor output during the intervening breaths (r = 0.67) (Fig. 53-20).99 In other words, the reduction in motor output during the intervening breaths achieved by adding pressure support was carried over to the mandatory breaths, facilitating greater unloading. Combining IMV and pressure support provides a sometimes useful means of achieving a high level of assistance; the combination has a clinical advantage when it is difficult to achieve a high inspiratory flow in the assist-control mode, as with the Siemens 900C ventilator (Siemens Corporation, New York, NY), although few of these machines are likely to be in current use.
The change in pressure-time product per breath (PTP/br) during mandatory breaths (of IMV) consequent to the addition of pressure support of 10 cm H2O to a given level of IMV was related to the change in respiratory motor output (dP/dt) effected by pressure support during the intervening breaths (r = 0.67; p < 0.0001). The more that pressure support decreased respiratory motor output during the intervening breaths, the greater was the reduction in patient work during the mandatory ventilator breaths delivered during IMV. (Used, with permission, from Leung et al.99)
In studies of interactions between patient effort and mechanical ventilation, remarkably little attention has been paid to the switch between inspiration and expiration. The most common mode of ventilation is some form of volume assistance,113 such as assist control or IMV. “Cycling-off” of mechanical inflation, however, may be based only indirectly on volume. Instead, inspiratory flow is commonly preset and the ventilator adjusts inspiratory time to achieve a given VT. This system is more precisely termed time-cycled ventilation. Inflation time is constant with a time-cycled machine, but patients invariably display considerable breath-to-breath variability in inspiratory time.114 Accordingly, a patient’s neural inspiratory time may be shorter or longer than the inflation time of the machine. If the machine delivers the set VT before the end of a patient’s neural inspiratory time, ventilator assistance will cease while the patient continues to make an inspiratory effort—with double triggering (two ventilator breaths for a single effort) a likely consequence.104
During assist-control ventilation, when a patient’s neural inspiratory time is short, ventilator inflation may continue into neural expiration and thus decrease the time available for lung emptying. The sense of being unable to empty the lungs may cause patients to activate the expiratory muscles with the result that the patient appears to fight or buck the ventilator (Fig. 53-21). The decrease in time for emptying also increases the likelihood for dynamic hyperinflation and thus inspiratory efforts that fail to trigger the ventilator.
Recordings of flow, airway pressure (Paw), and transversus abdominis electromyography (EMG) in a critically ill patient with COPD receiving pressure support of 20 cm H2O. The onset of expiratory muscle activity (vertical dotted line) occurred when mechanical inflation was only partly completed. (Used, with permission, from Parthasarathy et al.97)
With pressure support, the algorithm used for “cycling-off” of mechanical inflation varies among brands (see Chapter 8). On most ventilators, the termination of inspiratory assistance during pressure support is set at a default threshold, such as 25% of the peak inspiratory flow or inspiratory flow of 5 L/min. That is, when inspiratory flow falls to 25% of the peak value or below 5 L/min, inspiratory pressurization switches off. Such algorithms can be problematic in patients with COPD because increases in resistance and compliance produce a slow time-constant of the respiratory system. The longer time needed for flow to fall to the threshold value can cause mechanical inflation to persist into neural expiration. In twelve patients with COPD receiving pressure support of 20 cm H2O, five recruited their expiratory muscles while the machine was still inflating the thorax.115 The patients who recruited their expiratory muscles during mechanical inflation had an average time constant of 0.54 seconds, compared with an average of 0.38 seconds in the patients who did not exhibit expiratory muscle activity. The persistence of mechanical inflation into neural expiration is very uncomfortable, as is well recognized with use of inverse-ratio ventilation. Algorithms that achieve better coordination between the end of mechanical inflation and the onset of a patient’s expiration may lessen this form of patient–ventilator asynchrony.116,117
The transition between sleep and wakefulness can lead to respiratory distress in the ventilated patient. More surprisingly, the transition between wakefulness and sleep can also cause distress. Specifically, the selection of ventilator mode and settings can provoke sleep disruption. Ventilated patients experience considerable sleep disruption, with as many as seventy-nine arousals and awakenings per hour.118,119 Sleep disruption can adversely affect patient outcome.120
In eleven critically ill patients, Parthasarathy and Tobin121 studied the interaction between ventilator mode and sleep. Sleep fragmentation was greater during pressure support than during assist-control ventilation: 79 versus 54 arousals and awakenings per hour (see Fig. 53-9). Six of the eleven patients developed central apneas during pressure support, but not during assist-control ventilation. VT was 8 mL/kg during assist-control; pressure support was titrated to achieve the same VT. The level of pressure support was 16.8 ± 1.5 cm H2O in patients with apneas and 19.6 ± 2.6 cm H2O in patients without apneas; thus, apneas did not result simply from a higher level of pressure support.
Sleep fragmentation, measured as the sum of arousals and awakenings, was greater during pressure support than during assist-control: 79 ± 7 versus 54 ± 7 events per hour. Disturbed sleep during pressure support was related to the development of central apneas (r = 0.57), which, in turn, was significantly related to the difference between during resting breathing and the patient’s apnea threshold (end-tidal CO2[ΔPETCO2]) (r = −0.83) (Fig. 53-22). ΔPETCO2 was the most important determinant for the development of apneas: As ΔPETCO2 grew wider, the number of central apneas increased. The addition of 100 mL of dead space to the ventilator circuit in the six patients who developed apneas produced a 4.3 mm Hg increase in end-tidal CO2, decreased the frequency of central apneas, from fifty-three to four apneas per hour, and the frequency of arousals and awakenings, from eighty-three to fourty-four events per hour. This study shows that while critically ill patients have a background level of sleep disturbance, secondary to factors such as pain, medications, staff interruptions, noise, and light, the mode of mechanical ventilation can further aggravate sleep disruption.
The difference between the average end-tidal CO2 (PETCO2) and the apnea threshold plotted against the number of central apneas per hour of pressure support alone (red symbols) and pressure support with added dead space (blue symbols) in six patients. The average end-tidal CO2 was measured during both sleep and wakefulness. The mean number of central apneas per hour was strongly correlated with the end-tidal CO2 during a mixture of both sleep and wakefulness (including the transitions between sleep and wakefulness) (r = −0.83, p < 0.001). (Used, with permission, from Parthasarathy et al.121)
The observation that ventilator mode can aggravate sleep disruption was confirmed by Bosma et al.122 In thirteen patients undergoing polysomnography, these investigators adjusted the levels of pressure support ventilation and proportional-assist ventilation to achieve a similar degree of inspiratory muscle unloading. The number of arousals and awakenings were significantly greater with pressure support than with proportional-assist ventilation. Moreover, the number of patient–ventilator asynchronies per hour correlated significantly with the number of arousals per hour (r 2 = 0.71).
Cabello et al123 compared sleep quality while patients were randomized between assist-control ventilation, pressure support set by the patient’s attending physician, and pressure support continuously adjusted by a closed-loop knowledge-based system. In contrast to the report by Parthasarathy and Tobin,121 sleep architecture, sleep quantity, and sleep fragmentation were equivalent with the three ventilator modes.
The different results in the two studies can be explained by the method of selecting tidal volume in the study of Cabello et al.123 A tidal volume of 8 mL/kg was employed during assist-control ventilation, whereas a tidal volume between 6 and 8 mL/kg was targeted during pressure support.123 The median (twenty-fifth to seventy-fifth percentile) tidal volumes were 500 mL (380 to 500 mL) during assist-control ventilation, 450 mL (357 to 521 mL) during clinician-adjusted pressure support, and 390 mL (330 to 492 mL) when a closed-loop knowledge-based system was used to continuously adjust pressure support. The findings of these three studies of mechanical ventilation during sleep in critically ill patients indicate that pressure support, when carefully adjusted to avoid hyperventilation, does not increase the amount of sleep fragmentation over that experienced with assist-control ventilation. If, however, pressure support is set in the manner of everyday clinical practice, it is likely to lead to sleep fragmentation.121,122
The alterations in breathing pattern and gas exchange induced by sleep have important implications for the selection of ventilator settings. During pressure support, sleep induced a 23% increase in inspiratory time and a 126% increase in expiratory time, as compared with wakefulness.121 Sleep caused the respiratory rate to decrease by 33% during pressure support and by 15% during assist-control ventilation (Fig. 53-23). The level of pressure support is commonly titrated to respiratory rate, which provides a reasonable guide to patient effort.115,124 Pressure support is commonly titrated during the daytime without the clinician being sure whether the patient is asleep or awake. If the patient is asleep at the time of the adjustment, a point at which respiratory rate will be relatively low, then, on awakening, the increase in a patient’s rate may cause a considerable increase in effort.
Respiratory rate during assist-control ventilation (AC) and pressure support (PS) in eleven critically ill patients. For each mode, the lines connect the mean value for each patient during wakefulness (W, left) and sleep (S, right). Compared with wakefulness, group mean respiratory rate was lower during sleep (blue symbols) than during wakefulness (red symbols). The difference between sleep and wakefulness was greater for pressure support than for assist-control ventilation. (Modified, with permission, from Parthasarathy and Tobin.121)