INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients in the intensive care unit are frequently unable to entrain their respiratory activity to a mechanical ventilator and fight the ventilator. These patients may require heavy sedation and occasionally even paralysis to achieve satisfactory patient-ventilator synchrony. Adverse patient-ventilator interactions occur not only as a result of the machine's failure to meet the patient's ventilatory demand, but also as a result of the patient's failure to modify his or her breathing in response to mechanical ventilation. To understand the problem of asynchrony between patient efforts and machine breaths requires an understanding of entrainment. Entrainment implies a resetting of the respiratory rhythma change in respiratory ratesuch that a fixed, repetitive, temporal relationship exists between the onset of neural inspiratory activity and a mechanical breath.

More is known about entrainment of the respiratory rhythm to periodic lung inflations delivered by a mechanical ventilator in experimental animals (1) than in humans. In anesthetized animals, the capacity of the respiratory control system to entrain neural respiratory activity to machine inflations differs depending on ventilator settings, i.e., machine frequency, tidal volume (VT), and inspiratory flow rate (2). The largest range of machine rates associated with 1:1 entrainment (one ventilator cycle to one inspiratory effort) during mechanical ventilation of anesthetized rabbits was obtained with the largest lung inflations and the lowest flow rates. Hering-Breuer reflexes are required for entrainment in anesthetized animals; entrainment to a ventilator was lost after bilateral vagotomy (3, 4). Furthermore, expiratory time (TE) prolongation consistent with the Hering-Breuer reflex was observed in anesthetized humans when ventilator inflations occurred during the inspiratory-expiratory transition, although shortening of inspiratory time (TI), indicative of the Hering-Breuer inspiratory inhibition reflex, was not observed in response to lung inflations during neural inspiration (5). The level of respiratory drive also modifies entrainment to mechanical ventilation (6). An increase in respiratory drive caused by CO₂ loading in anesthetized rabbits reduced the capacity of the respiratory control system to modify the respiratory rhythm and to maintain 1:1 entrainment when the ventilator rate was varied during mechanical ventilation (1). Entrainment responses of humans and animals to a mechanical ventilator are similar, but so far as we are aware, entrainment to mechanical ventilation has been studied only during anesthesia. Anesthesia, by virtue of its effects on the respiratory control system, may affect entrainment responses to a mechanical ventilator, and entrainment responses may differ among the anesthetized state, wakefulness, and non-rapid eye movement (NREM) sleep. Anesthesia is not a good model for sleep because respiratory drive differs between these two conditions, and anesthesia may reduce the sensitivity to, and integration of, afferent information in ways significantly different from sleep. Therefore, the first objective of this study was to examine the effects of wakefulness and NREM sleep on entrainment responses of normal humans to mechanical lung inflations when machine frequency was varied. The second objective was to evaluate the effect of increased respiratory drive induced by mild hypercapnia on entrainment responses to mechanical ventilation of normal, sleeping humans. On the basis of animal studies, we predicted that hypercapnia would reduce the capacity of the respiratory control system to entrain its rhythm to mechanical ventilation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

We recruited 53 normal adult volunteers without a history of cardiopulmonary disease. None of the subjects had a background in respiratory physiology, nor did they know the objectives of the study. The subjects were not trained for these studies. We report findings from 16 normal, healthy volunteers (10 women and 6 men), 18 to 35 yr of age, from whom data were obtained. Five subjects were studied during wakefulness. Twelve subjects underwent one of two protocols involving either isocapnia (n = 9) or hypercapnia (n = 6) during NREM sleep; 2 subjects participated in both NREM sleep protocols on separate nights, and 1 subject participated in all three protocols. The remaining 37 subjects were unable to sleep under the conditions of the experiment. The study was approved by the Institutional Review Board of the Mayo Clinic (Rochester, MN); informed consent was obtained from all subjects.

Measurements

Subjects were supine in bed during these studies. All subjects underwent volume-controlled mechanical ventilation through a nasal mask. Subjects were ventilated through a nasal continuous positive airway pressure (CPAP) mask attached to a Puritan-Bennett 7200 ventilator (Puritan-Bennett, Carlsbad, CA) modified for research purposes. First, auditory alarm functions were disabled to minimize subject arousal. Second, 12% CO₂ in oxygen was added to the inspired gas via the O₂ inlet, and the dial used to change the inspired O₂ fraction was adapted to permit adjustment of the inspired CO₂ fraction (FI_CO2) to the target end-tidal CO₂ (PET_CO2).

After calibration, measurements of airway pressure and flow were obtained from the analog output of the ventilator. Tidal volume was obtained by integrating flow. End-tidal gas was sampled from a port attached to the nasal CPAP mask. CO₂ concentration was measured with a calibrated Novametrix (Wallingford, CT) model 1260 infrared capnostat. Diaphragmatic and parasternal electromyographic (EMG) activity on the right side of the chest was monitored with surface electrodes (Red Dot; 3M, Medical Products Division, St. Paul, MN) placed in the anterior axillary line over the sixth and seventh intercostal spaces and in the first and second intercostal spaces, respectively. Inspiratory activity and respiratory timing were measured from diaphragm and parasternal EMG recordings. Electroencephalographic (EEG) activity, monitored from the C4-A1 and CZ-OZ leads, the submental EMG, and electro-oculographic (EOG) activity were used to document sleep stages. EMG and EEG activity was processed using a TECA 42 electromyographic instrument (Teca, Pleasantville, NY). All signals were displayed and recorded using an Astro-Med (West Warwick, RI) MT 8000 strip chart recorder and recorded on magnetic media using a computer acquisition program (LabVIEW; National Instruments, Austin, TX).

Experimental Protocols

Protocol during wakefulness. Subjects wore headphones with background music playing to distract them from the noise of the ventilator and possible clues that the ventilator settings were changing. The protocol began with a 5-min observation period, during which the subject breathed unassisted in the flow-by mode with the following ventilator settings: CPAP equal to 0 cm H₂O, baseline flow at 20 L/min, and flow sensitivity at 3 L/min. Eupneic VT and PET_CO2 were determined during this period. Next, volume-cycle ventilation set in the assist-control (AC) mode with constant inspiratory flow set at 40 L/min (square waveform) and with machine VT set equal to 150% of the spontaneous VT was initiated. The machine backup rate was set to 2 breaths/ min (bpm) with a flow-by threshold of 3 L/min to allow each subject to choose his or her own respiratory rate. The rate at which the subject triggered machine breaths (at a constant tidal volume and flow rate) was labeled the "spontaneous respiratory rate" and was determined over a 5-min period. In the final entrainment period, machine trigger mechanisms were disabled by decreasing the threshold to 20 cm H₂O, and machine rates were varied 2 bpm every 3 min above and below the spontaneous respiratory rate until either 1:1 entrainment (one ventilator cycle to one inspiratory effort) was lost or inspiratory activity was undetectable. These trials were performed without CO₂ supplementation.

Protocols during NREM sleep. Subjects participating in the protocols during sleep were asked to deprive themselves of sleep (< 2 h) the preceding night. Once the subject was in stable stage II or III-IV sleep, a protocol identical to the awake protocol was followed: a 5-min observation period during which the subject breathed unassisted with the ventilator set in the flow-by mode, CPAP equal to 0 cm H₂O, baseline flow at 20 L/min, flow sensitivity at 3 L/min to determine eupneic VT and PET_CO2, and a 5-min period of volume-cycle ventilation set in the AC mode (VT = 150% of spontaneous VT, inspiratory flow = 40 L/min) during which the spontaneous rate was determined. CO₂ was supplemented throughout to maintain isocapnia. Finally, machine trigger mechanisms were disabled and machine rates were started at or 1 bpm above the spontaneous respiratory rate and varied 1 bpm every 3 min below and above the spontaneous respiratory rate until entrainment was lost, inspiratory EMG activity was undetectable, or the subject aroused.

On a separate night, a protocol identical to the sleep protocol was performed under hypercapnic conditions. Once the subject was in stable stage II or III-IV sleep, CO₂ was added to the inspired gas to keep the PET_CO2 3 mm Hg above the eupneic PET_CO2 throughout the protocol.

Data Analysis and Statistics

The onset of a subject's spontaneous respiratory activity was determined by the onset of surface EMG activity of either the diaphragm or parasternal muscle, depending on which signal was of better quality. We defined entrainment as a pattern in which the inspiratory efforts of the subject occurred over a specific and repetitive phase of the ventilator cycle. The type of entrainment response was determined by the ratio of machine cycle to inspiratory efforts. For example, a 1:1 entrainment pattern was defined as one machine cycle associated with one respiratory cycle (Figure 1), whereas a 1:2 entrainment pattern was defined as one machine cycle associated with two respiratory cycles (cf. Figure 3).

View larger version (11K): [in this window] [in a new window]   Figure 1.   Stable temporal relationship between inspiratory efforts and machine inflations, with the diaphragm contractions occurring at a consistent time relative to the ventilator cycle. This example shows a 1:1 entrainment pattern, i.e., for each machine cycle there is one inspiratory effort.

View larger version (22K): [in this window] [in a new window]   Figure 3.   A phase scatter plot, in which the phase angles of all digitized breaths during a trial of mechanical ventilation are plotted as a function of sequential machine breaths. In most cases, there is a 1:1 ratio of subject efforts to machine breaths, but for breaths 3, 5, 7, 10, and 13 there were two subject efforts for each machine breath (a 1:2 ratio). The average of all the phase angles shown was 82 ± 76°, and there was a significant (p < 0.05) concentration of phase angles about this mean. Hence, a single, average phase angle sufficed to describe the entrainment response despite the presence of occasional 1:2 entrainment ratios.

We determined the phase relationships between the onset of surface EMG activity and the machine cycle from recordings like Figure 1 as illustrated in schematic form in Figure 2. The upper tracing in Figure 2 is a sketch of a volume tracing during controlled mechanical ventilation; the lower tracing represents diaphragm surface EMG activity. T_pump is the duration of the machine cycle. The phase delay is the time in seconds from the onset of a spontaneous inspiration to the onset of the machine inflation. The phase angle (), which describes the relationship between machine onset and surface EMG onset, was determined by calculating the phase delay, dividing by T_pump, and multiplying by 360°. Onset of machine inflation was assigned a  value of 0°, and the  value of inspiratory surface EMG activity ranged from 180° to +180°. A  value of 0° indicated that machine inflation and surface EMG onset occurred at the same time. When surface EMG activity preceded machine inflation,  was between 180° and 0°, and when surface EMG activity occurred during or after machine inflation,  was between 0° and +180°. Breaths were averaged over the last 3 min of the initial 5-min observation period before initiation of mechanical ventilation to determine spontaneous tidal volume and eupneic end-tidal CO₂. Before disabling the trigger mechanisms of the mechanical ventilator, breaths were averaged over the last 3 min of another 5-min period to determine the spontaneous respiratory rate during mechanical ventilation. The last 1.5 min of data acquired at each machine rate were digitized, and the phase angle was computed from the onset of the machine inflation and the onset of inspiratory EMG activity on a breath-by-breath basis as shown in Figure 2. Neuromechanical inhibition of surface EMG activity occurred at some ventilator frequency above the spontaneous rate in all subjects. Data from the ventilator frequency at which surface EMG activity first started to drop out in each subject and higher frequencies were not analyzed because the phase angle could not be known accurately. In addition, neural TI was obtained from the duration of inspiratory activity measured from surface EMG activity and compared at different phase angles in four subjects during NREM sleep under isocapnic conditions.

View larger version (14K): [in this window] [in a new window]   Figure 2.   The relationship between machineOnset and EMGOnset is described by calculating the phase angle () for each breath.

Traditionally, the analysis of entrainment has used phase plots in which each phase angle was plotted as a function of the breath sequence. This analysis allows one to see all the combinations of possible phase-locking ratios (1:2, 1:1, 2:1, etc.), but it does not provide a simple characterization of entrainment at each breath (cf. Figure 3). To remedy this, we have taken a statistical approach to entrainment and calculated the average and the standard deviation of  at each machine rate for each subject. We defined 1:1 entrainment as a significant (p < 0.05) concentration of phase angles about some mean  value. The standard deviation provided a measure of the tightness of the phase locking. If the distribution of phase angles was homogeneous from 180° and +180° (i.e., no significant single ), then we determined whether there were significant concentrations of phase angles from   180° to  < 0° and from   0° to < 180°. If significant concentrations of  were found in both of these ranges, then 1:2 entrainment existed. No other stable entrainment ratios were seen. The mean and standard deviation of phase angles at each particular machine rate for each subject were calculated using methods appropriate for angles (7). Furthermore, angles are not normally distributed; therefore, statistical inferences were made using the von Mises distribution, which is analogous to a normal distribution, but appropriate for periodic functions. Probability distribution curves for phase angles at particular breaths were calculated from the mean angle and a concentration parameter (K) as described by Mardia (7) and plotted as a function of machine rate expressed relative to the spontaneous rate using Matlab (Math Works, Natick, MA).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Respiratory Variables during Wakefulness and Sleep

The individual and average spontaneous tidal volumes and respiratory frequencies during mechanical ventilation and the range of frequencies studied in each subject under each condition are shown in Table 1. The tidal volumes during isocapnic and hypercapnic NREM sleep were significantly less than the tidal volume during wakefulness (p = 0.001 for both conditions versus wakefulness), but VT was not significantly different between isocapnic and hypercapnia during sleep. The spontaneous frequency during mechanical ventilation was highest during hypercapnic NREM sleep and significantly different from wakefulness (p = 0.04), but not significantly different from the spontaneous frequency during isocapnic NREM sleep. The lower limits of the entrainment ranges are discussed subsequently.

Effect of Wakefulness on Entrainment Responses

Figure 4 shows phase scatter plots and phase angle probability distribution functions for two representative subjects and a composite phase angle probability distribution plot for all five awake subjects. Examination of the machine rates at or greater than each subject's spontaneous respiratory rate associated with 1:1 entrainment was limited by loss of surface EMG signal in all five subjects. In two of the awake subjects, loss of surface EMG signal occurred at and intermittently below their spontaneous respiratory rate. The 1:1 entrainment patterns were lost at machine rates that varied between 1 and 9 bpm below the spontaneous respiratory rate (mean, 6 ± 3 bpm). Complex entrainment patterns, 1:2, 1:3, 1:4, and 1:5, intermittently replaced the 1:1 pattern in four of the five subjects, but the dominant, average response was 1:1 entrainment at frequencies well below the spontaneous rate. In all subjects, significant concentrations about one phase angle sufficed to describe the average entrainment response. Furthermore, the phase angle remained close to 0° with a small standard deviation (i.e., a high concentration parameter, K) until the slowest ventilator frequencies were reached in each subject, at which point the standard deviation increased without any consistent change in the mean phase angle. In only one subject (shown in Figure 4B) was entrainment lost at the lowest machine rate [i.e., no significant concentration(s) of phase angles]. The pattern of phase angles at this low frequency is best characterized as aperiodic. Aperiodic behavior was seen in anesthetized animals at frequencies and tidal volumes in the transitions between integral ratio entrainment relationships (8). Subjects maintained 1:1 entrainment with machine breaths when the ventilator rate was set at or above 41 ± 18% (mean ± SD) of their spontaneous respiratory rates, which defines the lower limit of 1:1 entrainment. We cannot determine the upper limit of 1:1 entrainment because there was neuromechanical inhibition of surface EMG activity before 1:1 entrainment was lost at ventilator frequencies above the spontaneous rate.

View larger version (34K): [in this window] [in a new window]   Figure 4.   (A and B) Phase scatter plots from two representative awake subjects. The dashed line indicates the spontaneous respiratory rate for each subject during volume preset ventilation at 150% of the spontaneous VT before the triggering mechanism of the ventilator was disabled. The individual phase angles are plotted to the left and the mean phase angle and standard deviation are plotted to the right at each ventilator frequency. The mean and standard deviation are plotted only if there was a significant (p  0.05) concentration of phase angles about some mean. The lack of a significant mean implies that the probability of any particular phase angle was homogeneous across the range of possible angles; a pattern that we believe implies aperiodic behavior. In (C ) and (D), the probability distribution of phase angles at each ventilatory frequency has been plotted for each subject in (A) and (B), respectively. The total area under the probability curve at each frequency is one. In (C ) and (D), ventilatory frequencies have been expressed relative to the spontaneous rate (set equal to 0 for each subject). Frequencies at which phase angles were densely packed about some mean angle have high probabilities at the mean phase angle, high concentration parameters (K), and probabilities that drop quickly to zero as one moves away from the mean phase angle. (E ) A composite probability distribution for all subjects. Note that we have no data from any subject at a frequency of 1 bpm below the spontaneous rate, and the deep valley in the probability distribution at this frequency does not reflect a marked change in mean phase angle. Furthermore, the probabilities between breaths moving along the respiratory frequency axis for which we do have data also dip, and these cols between integral breath numbers are an artifact of the surface-fitting routine in Matlab.

Effect of NREM Sleep on Entrainment Responses

Figure 5 shows phase scatter plots and phase angle probability distribution functions from two representative subjects and a composite phase angle probability distribution plot under isocapnic conditions (mean PET_CO2 was 42 ± 3 mm Hg) in nine sleeping subjects during mechanical ventilation. Machine rates greater than the subject's spontaneous respiratory rate associated with 1:1 entrainment were evaluated in six of the nine subjects. In three of these six subjects, the surface EMG signal was lost when machine rates were increased several breaths above the spontaneous respiratory rate. Compared with wakefulness, 1:1 entrainment patterns were lost at machine frequencies much closer to each subject's spontaneous respiratory rate during NREM sleep. When machine rates were decreased 2-4 bpm below the sleeping subject's spontaneous respiratory rate (mean, 3 ± 1 bpm), 1:2 entrainment patterns replaced 1:1 entrainment patterns. The lower limit of entrainment was significantly higher (p < 0.001) during isocapnic NREM sleep than during wakefulness. In some subjects, there was a machine rate associated with aperiodic respiratory frequencies at the point of bifurcation from 1:1 to 1:2 entrainment as the mechanical ventilator rate was reduced. Unlike wakefulness, no other complex entrainment patterns (i.e., 1:3, 2:3, etc.) emerged even briefly. During NREM sleep, the lower limit of 1:1 entrainment with machine breaths was at or above 84 ± 4% (mean ± SD) of the spontaneous respiratory rates.

View larger version (37K): [in this window] [in a new window]   Figure 5.   Phase scatter plots and probability distribution functions from two representative subjects and a composite probability distribution function for all subjects during isocapnic NREM sleep. The conventions in each graph are described in the caption to Figure 4.

There were also differences in the distribution of phase angles at different machine rates between wakefulness and NREM sleep (compare Figures 4 and 5). The average phase angles remained close to 0° across all frequencies during wakefulness, but the trajectory of phase angles progressed from positive phase angles at frequencies above the spontaneous rate to phase angles of approximately zero at the spontaneous rate to negative phase angles below the spontaneous rate, until a transition occurred at which a wider range of phase angles existed and below which the entrainment pattern bifurcated into 1:2 entrainment.

Effect of Increasing Respiratory Drive on Entrainment Responses

Figure 6 shows phase scatter plots and phase angle probability distribution functions for two representative subjects and a composite phase angle probability distribution plot of the entrainment responses under hypercapnic conditions (45 ± 3 mm Hg) (mean ± SD) from all six mechanically ventilated subjects during NREM sleep. In three of the six subjects 1:1 entrainment was lost and a 1:2 entrainment pattern emerged at machine rates that averaged 4 ± 2 bpm below the spontaneous respiratory rate of each subject (see Figure 6A for an example of this response pattern). In the remaining three subjects, the trial was ended because of arousal in two subjects (see Figure 6B for an example of this response pattern) and respiratory effort exceeding 20 cm H₂O in the remaining subject, which overrode the disabled triggering system of the ventilator. All of the subjects maintained a 1:1 entrainment pattern at least 1 bpm below their spontaneous respiratory rate. The average lower limit of 1:1 entrainment to machine frequencies was 86 ± 8% (mean ± SD) of each subject's spontaneous respiratory rate. The lower limit of entrainment was significantly higher (p < 0.001) during hypercapnic NREM sleep than during wakefulness, but not different from isocapnic NREM sleep. The trajectories of phase angles from positive to negative angles and the bifurcation from 1:1 to 1:2 entrainment were quite similar under hypercapnic and eucapnic conditions (compare the composite probability distributions in Figures 5 and 6). Hence, there were no apparent differences in the responses of the respiratory control system to a decrease in the machine rate during mechanical ventilation under hypercapnic compared with isocapnic conditions in any of the subjects.

View larger version (38K): [in this window] [in a new window]   Figure 6.   Phase scatter plots and probability distribution functions from two representative subjects and a composite probability distribution function for all subjects during hypercapnic NREM sleep. The PETCO2 increased 3 mm Hg above eupneic PETCO2 by adding inspired CO2. The conventions in each graph are described in the caption to Figure 4.

Effect of Timing between the Machine Cycle and Respiratory Cycle on Neural TI

In an assessment of the inspiratory inhibitory Hering-Breuer reflex, neural TI began to shorten in all four sleeping subjects examined during isocapnic NREM sleep as inspiratory efforts approached the onset of the machine inflation cycle. As inspiratory efforts coincided or extended into the machine inflation cycle, neural TI shortened further. This is depicted in Figure 7, in which inspiratory time has been plotted as a function of phase angles (relationship between machine_Onset and EMG_Onset) in one of the nine sleeping subjects.

View larger version (14K): [in this window] [in a new window]   Figure 7.   In normal subjects during isocapnic NREM sleep, neural TI is shorter for those inspiratory efforts that extend into the machine inflation cycle.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this article, we describe the entrainment responses of humans to a mechanical ventilator during wakefulness, during NREM sleep, and after CO₂-induced increases in respiratory drive. The capacity of the human respiratory control system to modify its respiratory rhythm in response to frequency changes in mechanical ventilator-delivered lung inflations was dependent on state. The range of ventilator frequencies over which 1:1 entrainment was maintained was narrower during NREM sleep compared with wakefulness. Furthermore, a mild increase in respiratory drive caused by CO₂ stimulation did not limit the capacity of the respiratory control system to entrain its respiratory rhythm to that of the ventilator cycle during NREM sleep. Neural TI changed during isocapnic NREM sleep depending on the relationship between inspiratory effort and the onset of machine inflation, a finding consistent with observations in animal studies in which vagal afferents were important in mediating entrainment of the respiratory rhythm to a mechanical ventilator.

Limitations

We chose to delineate a range of machine frequencies for a given tidal volume and inspiratory flow that resulted in 1:1 entrainment. Earlier entrainment studies (1) showed that the 1:1 entrainment pattern was both the most stable and the most frequent entrainment response observed. In our sleeping humans, the loss of 1:1 entrainment commonly occurred over a transition period usually confined to one machine rate at which 1:2 entrainment patterns sporadically interrupted 1:1 entrainment patterns; when the machine rate was lowered further, the 1:2 entrainment pattern became firmly established. The machine rate at which the transition period occurred defined the loss of 1:1 entrainment. Our statistical definition of entrainment was minimalistic in that if aperiodic behavior or brief complex entrainment patterns occurred, we still defined the pattern as 1:1 entrainment so long as a single, significant concentration of phase angles sufficed to describe the sample of phase angles at that machine rate. When more complex patterns occurred briefly, the standard deviation was increased by the diversity of entrainment patterns at the particular machine frequency.

Another limitation is that we did not specifically measure respiratory drive during CO₂ stimulation. There is little debate that CO₂ loading causes an increase in respiratory drive in normal humans reflected in an increase in respiratory motor output, and we did observe greater inspiratory effort in the airway pressure and expiratory flow waveforms and increased surface EMG activity. But VT and frequency did not increase significantly during sleep when the subjects were made hypercapnic, and we did not measure other indices of respiratory drive. Finally, our study is limited in that each subject did not act as his or her own control. It was simply too difficult to obtain adequate data during sleep in multiple subjects on multiple nights.

Vagal Influences on Entrainment

Previous work in anesthetized experimental animals showed loss of entrainment during vagotomy (3, 4). Studies in which the Hering-Breuer reflex(es) were used to explain entrainment (3, 4, 9) focused on vagal mechanisms to describe the temporal relationship between respiratory activity and the machine cycle. In previous entrainment studies in anesthetized animals, when the length of the mechanical ventilatory cycle was greater than the length of the respiratory cycle (machine frequency < spontaneous frequency), ventilator inflations commenced during the neural inspiratory-expiratory transition and increased lung volume during expiration; thereby prolonging the neural expiratory phase by a Hering-Breuer reflex mechanism and decreasing the respiratory rate. In this setting, neural inspiration led the onset of mechanical ventilation and the phase angles were negative. Graves and colleagues (5) observed TE prolongation in anesthetized humans when ventilator inflations commenced during the neural inspiratory-expiratory transition (machine frequency < spontaneous frequency), consistent with a Hering-Breuer reflex mechanism. When the ventilator cycle length was less than the respiratory cycle length (machine frequency > spontaneous frequency), the ventilator inflations commenced late in neural expiration and augmented lung inflation during neural inspiration; thereby shortening neural inspiration by a Hering- Breuer reflex mechanism and increasing the respiratory rate. When the ventilator cycle length was less than the respiratory cycle, neural inspiration lagged the onset of mechanical ventilation, and phase angles were positive. Contrary to the foregoing predictions, Graves and colleagues observed no consistent changes in neural TI (5).

We also found that inspiratory activity preceded the machine inflation cycle when machine frequency < spontaneous frequency, and inspiratory activity occurred during or after the machine inflation cycle when machine frequency > spontaneous frequency in our sleeping subjects. In contrast to the findings of Graves and coworkers, neural TI was shorter when inspiratory efforts extended into the machine inflation cycle. Shortening of neural TI under these circumstances implies an active Hering-Breuer inhibitory inspiratory reflex, and thus, vagal afferents probably play a role in entrainment to mechanical ventilation in sleeping humans. The reason for the lack of TI changes in the study by Graves and colleagues (5) is unclear; differences in both study design and the use of anesthesia are potential explanations for the differences between our studies.

Similar observations were not found in our awake subjects. The temporal relationship between inspiratory activity and the machine cycle did not change as machine frequency was varied: the smooth transition from positive to negative phase angles as ventilator frequency changed from rates above the spontaneous frequency to rates below the spontaneous frequency was not seen in waking subjects. Inspiratory activity occurred during or after the machine inflation cycle regardless of the relationship between machine frequency and spontaneous frequency, and phase angles were close to zero across all ventilator rates. Petrillo and Glass (8), using an "integrate and fire" model with time varying inspiratory off-switch and on-switch thresholds, were able to make excellent predictions of phase locking of respiration to mechanical ventilation over a large combination of tidal volumes and frequencies in anesthetized cats. We examined only a single tidal volume in each subject and a small frequency range (limited by arousal from sleep), but we saw many features in sleeping humans similar to the entrainment responses in anesthetized cats: 1:2 entrainment over a limited range of machine frequencies well below the spontaneous frequency, 1:1 entrainment over a wide range of machine frequencies bracketing the spontaneous frequency, and aperiodic behavior between the 1:2 and 1:1 entrainment patterns. We would probably have seen greater regions of aperiodic behavior had we varied tidal volume as well. However, we saw little or no evidence of the phase locking described by Petrillo and Glass in waking subjects. Vagal feedback is an essential element in the "integrate and fire" models of entrainment, but during wakefulness, it appeared that vagal feedback was ignored in resetting respiratory timing or more powerful entraining factors superceded vagal feedback and mediated entrainment over a wide range of respiratory frequencies. Cortical influences, discussed subsequently, probably contribute significantly to phase locking of the respiratory rhythm to maintain 1:1 entrainment in awake humans.

Effect of State on Entrainment Responses in Humans

Graves and colleagues (5) observed 1:1 entrainment in anesthetized subjects over a range of machine rates that were 40 to 140% of the subject's spontaneous breathing frequency. Outside the range of 1:1 entrainment, more complex entrainment patterns were seen. The entrainment responses that we observed in sleeping humans were quite different from those observed by Graves and colleagues in anesthetized humans (5). A 1:1 entrainment in sleeping humans was observed over a much narrower range of machine rates that was limited to ~ 85% of the spontaneous respiratory rate. Entrainment responses to lowering machine frequency in awake humans were similar to those of anesthetized humans; 1:1 entrainment was observed at or above 41% and at or above 40% of spontaneous respiratory rate in awake and anesthetized humans, respectively. Graves and co-workers (5) observed complex entrainment responses (e.g., 2:3 entrainment patterns) at machine rates above the anesthetized subjects' spontaneous rate, unlike our awake and sleeping humans, in whom only 1:1 entrainment was seen above the spontaneous rate. Data above the spontaneous respiratory rate both from sleeping subjects under isocapnic conditions and from awake subjects were limited because inspiratory activity ceased before 1:1 entrainment was lost when machine frequencies were increased 2-3 bpm above the spontaneous rate. Loss of the EMG signal was prevented in the study by Graves and colleagues (5) by adding low levels of CO₂ to the inspired air. During wakefulness, the reduction in respiratory motor output that occurred as the ventilator frequency exceeded the spontaneous frequency was not a simple consequence of hypocapnia and mechanosensory feedback, but may also be linked to behavioral feedback (10). In wakefulness, at least part of the effects of prolonged mechanical ventilation may be attributed simply to habituation of the subject to the ventilator and therefore relaxation. Respiratory drive increases during wakefulness compared with sleep. On this basis, one might have expected that 1:1 entrainment responses would have been restricted to a narrower range of machine frequencies during wakefulness when compared with 1:1 entrainment responses during sleep. However, our study showed that the wakefulness state facilitated 1:1 entrainment of the respiratory rhythm to the mechanical ventilator over a wider range of machine frequencies than during sleep.

We offer the following explanations. During wakefulness, forebrain influences on the respiratory control system can modulate the respiratory pattern in response to perceived respiratory sensations (11). For example, during wakefulness mechanical load compensation is achieved quickly; in contrast, load responses during NREM sleep are slow and incomplete (12). During wakefulness, subjects may have modified their respiratory rates to increase "comfort," i.e., to improve synchronization with machine inflation. This "learning" or "adaptation" could have occurred during the first few ventilator cycles. We suggest that respiratory sensations such as respiratory discomfort can result in conscious or subconscious forebrain influences on the respiratory control system that increase the respiratory control system's capacity to reset the respiratory rhythm to maintain 1:1 entrainment over a wider range of machine rates. Vagal afferents provide a powerful zeitgeber for entrainment, but respiratory vagal reflexes are weak during wakefulness, perhaps overridden by forebrain influences. Loss of these cortical influences during NREM sleep may restore the prominence of vagal reflex effects on respiration and limit the respiratory control system's resetting mechanisms such that respiratory rhythm cannot be slowed as effectively during sleep to maintain 1:1 entrainment once the machine rate is lowered several breaths below the spontaneous respiratory rate. Anesthesia also facilitates 1:1 entrainment over a much wider range of machine rates. Although forebrain influences and learning are probably absent during anesthesia, and anesthesia may reduce respiratory drive, it may also reduce mechanoreceptor activity from pulmonary stretch receptors (13), which may blunt the entraining power of vagal influences and expand the frequency range of 1:1 entrainment. In summary, different states (i.e., wakefulness, NREM sleep, and anesthesia) appear to have different effects on the respiratory control system's entrainment responses to mechanical ventilation. NREM sleep appears to restrict the capacity of the respiratory system to modify the respiratory rhythm and maintain 1:1 entrainment to machine inflations, whereas both wakefulness and anesthesia, by different mechanisms, appear to reduce the impact of the Hering-Breuer reflex and facilitate 1:1 entrainment over a wider range of machine frequencies.

Effect of Respiratory Drive on Entrainment Responses in Humans

The concept that the respiratory control system's capacity to modify its rate during mechanical ventilation is linked to the level of respiratory drive is supported by previous experiments in anesthetized animals that examined the effect of both single and multiple periodic perturbations on respiration timing (1, 6). Inspiratory neural activity is increased above baseline when Pa_CO2 is elevated above eupneic levels regardless of state. On the basis of these previous experiments, one might predict that the range of machine rates associated with 1:1 entrainment would be reduced under hypercapnic conditions because CO₂ would increase respiratory drive. However, we did not find differences in entrainment responses between isocapnic and mild hypercapnic conditions during NREM sleep, despite a twofold to threefold increase in inspiratory efforts. There are several possible explanations for the observed lack of effect. The magnitude of change in respiratory drive in our study was small compared with the changes achieved in anesthetized animals (1, 6). It is possible that the 1:1 entrainment response is not constrained by changes in respiratory drive until higher levels are reached. In normal spontaneously breathing humans, moderate levels of CO₂ stimulation primarily elicit a tidal volume response, and only at higher stimulus levels do changes in respiratory rate predominate. Therefore, it seems likely that our findings would have been different if we had chosen a higher CO₂ level; however, this was not feasible in our sleeping subjects because of problems with arousal. Alternatively, it is possible that we were unable to detect small differences in the range of machine rates associated with 1:1 entrainment between mild hypercapnia and isocapnia because the range of machine rates associated with 1:1 entrainment was quite narrow in sleeping humans. To the contrary, in the study by Benchetrit and coworkers (1), anesthetized rabbits maintained 1:1 entrainment over a much wider range of machine rates, and any differences between the ranges of machine rates associated with 1:1 entrainment under different CO₂ conditions may have been easier to detect.

Clinical Relevance

Lack of synchrony between patient efforts and machine breaths is a common problem in the management of mechanically ventilated patients with respiratory failure. Asynchrony can be viewed as a conflict between two competing oscillators: the ventilator and the patient's respiratory control system. Unavoidable limitations in sensing algorithms and response times prevent the ventilator from adjusting to patient demand. For example, the mode of ventilation may not be well suited to the patient's needs. Equally important is the failure of the respiratory control system to adjust to ventilator demands, which may reflect constraints in the resetting of the patient's respiratory rhythm in response to the mechanical ventilator. For example, mechanical impairment of the respiratory system may limit the flexibility of the respiratory controller as it responds to ventilator output. Patients with high ventilatory demands are particularly at risk for patient-ventilator asynchrony, and loss of entrainment in patients with high ventilatory demands (or "fighting the ventilator") can lead to lung injury. Heavy sedation or anesthesia is frequently used in these patients. We have shown that different arousal states have different effects on the respiratory control system's entrainment responses to mechanical ventilation. Anesthesia facilitates 1:1 entrainment, which may explain the observation that sedation improves patient-ventilator synchrony.

Cortical influences appear to facilitate resetting of the respiratory rhythm to maintain a 1:1 entrainment pattern when the ventilator frequency is changed; this is likely an adaptive strategy to avoid discomfort associated with patient-ventilator dysynchrony. NREM sleep appears to restrict the capacity of the respiratory control system to reset its rhythm and maintain 1:1 entrainment when the ventilator frequency is changed, and patients appear to be at greatest risk of patient-ventilator asynchrony during sleep. Understanding differences in entrainment responses to mechanical ventilation related to state are important when setting the ventilator, and attention should be paid to the appropriateness of ventilator settings, particularly the rate, when the arousal state of the patient changes from wakefulness to sleep.

The ability of patients to synchronize their efforts to machine output during nocturnal noninvasive positive pressure ventilation is likely an important determinant of the efficacy of this mode of treatment for patients with chronic respiratory failure. Failure of some patients (e.g., patients with chronic obstructive pulmonary disease) to benefit from nocturnal noninvasive mechanical ventilation may be related to an inability to entrain the respiratory rhythm to that of the ventilator, although this has not been proved.