INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Hypercapnia elicits a sense of respiratory discomfort and also evokes increases in breathing (1, 2). While the automatic respiratory centers in the brainstem are known to respond to this stimulus, it is possible that healthy, awake subjects increase breathing entirely using forebrain structures in order to reduce respiratory discomfort (see, for example, references 3 through 5). It is also possible that forebrain drive and automatic drive are summed, either at the level of the brainstem or at the level of the spinal neurons, to produce the observed ventilatory response in awake subjects. Forebrain control may also be important in patients having respiratory difficulty; these patients are aware of their breathing and may modify it using forebrain control to reduce respiratory discomfort (see, for example, reference 6). On the basis of transcranial magnetic stimulation, Murphy and colleagues (7) suggested that a behavioral component involving cortical innervation may contribute to the ventilatory response to hypercapnia, traditionally believed to be purely medullary. Since ventilation is ordinarily driven by the respiratory pump muscles which can receive drive from both the cortex and the brainstem, it is difficult to ascertain whether drive to the respiratory muscles in intact subjects originates in the forebrain, the medullary centers, or both. Sears (8) proposed that some breathing patterns that are at first learned and require willful control by the subject may subsequently become automatic, thereby making distinctions between volitional and automatic control of breathing even more difficult. Because of the difficulty in distinguishing the source of motor command to respiratory muscles, no study to date has demonstrated unequivocally a purely behavioral, cortical, ventilatory response to hypercapnia.

The present study examined a cortically generated single "ventilatory response" to hypercapnia in ventilator-dependent patients with paralyzed respiratory muscles. We tested whether these patients could be trained to control the tidal volume (VT) of their mechanical ventilator based on perceived air hunger alone. The motor act controlling ventilation in this circumstance (operation of a mouth switch) could not have originated in the brainstem respiratory centers because it did not involve the respiratory muscles. We adduce that the motor act controlling ventilation in this experiment originated in the cerebral cortex because operation of a mouth switch in response to respiratory discomfort is a novel learned act, not a reflex or habitual response.

Increased end-tidal partial pressure of CO₂ (PET_CO2) at constant ventilation produces the sensation of air hunger in subjects whose respiratory muscles are paralyzed (see, for example, references 9 and 10). At a given PET_CO2, respiratory discomfort is greater if ventilation is reduced; this has been shown in healthy subjects who voluntarily control their ventilation and in paralyzed patients whose ventilator VT is altered by an experimenter (11). We therefore hypothesized that, given the opportunity, paralyzed and ventilated patients could use respiratory sensations alone to consciously regulate their ventilator VT and increase ventilation to minimize air hunger induced by elevated PET_CO2.

We also investigated whether the sense of self-control itself influenced respiratory discomfort. Air hunger bears resemblance to pain in that it is an unpleasant sensory experience with an emotional component, and it is due to actual or potential physiologic impairment (16). Nonphysiologic mechanisms that typically influence perceptual responses to pain, therefore, are also likely to affect air hunger. For example, self-controlled analgesia tends to provide superior pain relief (as reflected in less medication used or less discomfort) compared with analgesics administered by a health care provider, presumably because subjects have a sense of control over their pain (17). Such findings led us to hypothesize that, for a given level of elevated PET_CO2, patients would report less air hunger when they were in control of their own ventilator compared with when an experimenter controlled the ventilator, even if ventilation were the same in both conditions. Surprisingly, we found that a sense of control of ventilation did not influence the sensation of air hunger over experimenter control.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was approved by internal review boards at the Harvard School of Public Health and at the performance sites: the Brockton/ West Roxbury Veterans Affairs Medical Center and the New England Sinai Hospital and Rehabilitation Center. Subjects provided informed consent.

Subjects

Two men and two women, tracheostomized and dependent on mechanical ventilation because of respiratory muscle paralysis, served as subjects; their characteristics are listed in Table 1. All subjects had participated in our earlier studies of respiratory sensation and are described in detail in previous publications using the same subject identity numbers (Subjects 1, 5, 10, 12; references, 9, 11, 20, 21).

Ventilation

Subjects were passively ventilated using a volume-cycled ventilator (Siemens 900C; Siemens-Elema AB, Solina, Sweden). The trigger pressure on the ventilator was set at 20 cm H₂O so that subjects could not generate a breath on their own. Inspired gas was heated and humidified with a sponge rehumidifier in the common ventilator line. Inspired CO₂ was mixed from one tank containing 10% CO₂, 50% O₂, and 40% N₂ and another tank of 50% O₂ and 50% N₂. PET_CO2 was held at the desired level by manipulating the inspired CO₂ fraction (Siemens mixing valve, Siemens O₂-Air Mixer 965). Inspiratory flow was delivered as a square wave occupying 33% of the respiratory cycle. End-inspiratory pause occupied 10% of the cycle. Expiration was passive; up to 2 cm H₂O positive end-expiratory pressure was applied (constant in each subject). To prevent air leaks, all subjects had their tracheostomy cuffs inflated during the testing periods.

Measurement and Recording

Tidal PCO₂ was sampled through a sidestream tube connected immediately distal to the tracheal cannula and recorded with an infrared analyzer (Datex Cardiocap II; Datex Medical Instruments, Helsinki, Finland). Gas analyzer responses were adequate to achieve end-expiratory and end-inspiratory plateaus. The analyzers were calibrated before each session with analyzed gas. Airway pressure and ventilatory flow were measured with transducers within the ventilator. Inspired and expired tidal volumes were compared to ensure that no significant gas leakage occurred. All physiologic data were recorded at a rate of 50 Hz per channel (CODAS acquisition hardware and software; Dataq Instruments, Akron, OH).

General Procedures

Each subject participated in four sessions which included training and experimental conditions, spanning an 8-d period. Sessions were performed at the same time of day to control for effects of routine medication and other diurnal rhythms (medication dose and time did not vary). In the experiment itself, we presented subjects with a very restricted task (described below) in which they were asked only to increase VT (and thereby ventilation) in response to air hunger induced by elevated PET_CO2. We used a simplified task because preliminary observations had shown that although subjects were able to control ventilation behaviorally, the task was not as easy for them as we initially expected. Specifically, preliminary data suggested that with no guidance subjects did not readily learn to increase ventilation with elevated PET_CO2; more important, they usually failed to decrease ventilation when PET_CO2 was reduced. Only when the additional training was provided (see VT ADJUSTMENT TRAINING) and the task was simplified to include only increases in PET_CO2 (thus requiring no decreases in ventilation) were the subjects able to consistently perform the task.

Preliminary Sessions

Pressure-volume curves. We determined the respiratory system pressure-volume characteristic of each subject by measuring end-inspiratory pressures at progressively increasing VT. We noted the pressure at which the respiratory system began to stiffen considerably, and to avoid barotrauma during testing we selected the maximal VT at 5 cm H₂O below that pressure (range: 1,250 to 2,000 ml).

VT detection. High-level quadriplegic, ventilator-dependent patients have no chest wall afferents yet they can detect VT changes in the range of 100 ml (20). The first training period was used to determine whether each subject could reliably detect the increment in VT, one-seventh of the subject's baseline VT (see Table 1), that would be used in the experimental conditions. All subjects were able to detect this change in VT.

Familiarization with mouth-operated switch. Subjects practiced using a mouth-operated switch, i.e., an apparatus attached to the ventilator which delivered increments in VT when the subject applied buccal pressure to a small tube (the switch was reversible; three subjects preferred to sip to increase VT, whereas Subject 5 opted to puff to increase VT). A green light flashed when subjects generated enough pressure to increase VT, and a red light flashed when subjects accidentally decreased VT.

Preliminary air hunger tests. This period was used to define air hunger to the subject and dissociate it from other respiratory sensations (e.g., pressure in the chest) and to determine the level of PET_CO2 necessary to evoke ratings of slight to moderate air hunger. At constant VT, PET_CO2 was adjusted in 3 to 4 mm Hg increments, at 3- to 5-min intervals. Subjects, uninformed of the timing and delivery of the PET_CO2 changes, were instructed "to rate how uncomfortable you are because of the air hunger you feel" and "not to rate the size of your breath." Subjects rated their air hunger on a seven-point ordinal scale throughout the test at 15-s intervals: zero (no air hunger), zero+, slight (clearly perceptible air hunger), slight+, moderate (uncomfortable but tolerable air hunger), moderate+, and extreme (intolerable air hunger; it was agreed that CO₂ would be decreased immediately if subjects rated their air hunger as extreme; for details of air hunger tests, see references 10 and 21).

VT adjustment training. Subjects were trained to adjust their VT in response to air hunger in the following way. The experimenters raised PET_CO2 and asked subjects to rate their air hunger. Upon reaching a rating of slight or moderate, subjects were instructed to use the mouth-operated switch to try to relieve the air hunger (PET_CO2 was held constant). All subjects reported relief with increased VT. Subjects were also instructed to decrease VT (while PET_CO2 was held constant). This last task was employed to confirm that decreases in VT made the subjects more uncomfortable.

Experimental Sessions

An important difference between the practice and experimental sessions was that during the practice sessions subjects were informed when the experimenter would increase PCO₂ and when they, in turn, should increase their VT. In contrast, during the experimental conditions, subjects were blind to the occurrence of the PCO₂ changes; furthermore, subjects were not told when to increase their VT. Because the trigger pressure on the ventilator was set so that the subjects could not generate a breath on their own, even if they had some ability to generate very small inspiratory pressures (for example, Subject 12; see Table 1), subjects were incapable of an automatic motor response (primary or accessory respiratory muscles) to alter ventilation in response to hypercapnia. Rather, subjects were required to pay close attention to their respiratory sensations and make a purely behavioral response to adjust VT by operating the mouth-operated switch using nonrespiratory, buccal muscles.

Procedures. Subjects participated in two experimental sessions and three test conditions: (1) Subject Control: the subject controlled VT while an experimenter adjusted PET_CO2; (2) Experimenter Control: the experimenter controlled VT while also manipulating PET_CO2; and (3) Constant Tidal Volume (Constant VT): VT was held constant while the experimenter adjusted PET_CO2. We controlled for order effects: Subject-Control periods were implemented at the beginning and repeated at the end of each session; the order of the Experimenter-Control and Constant-VT conditions was counterbalanced among subjects, and the order was reversed between the two sessions within each subject. (The order differed slightly in Subject 10.) Each condition followed a 5-min rest period.

Throughout all conditions, subjects rated their air hunger using the scale described above (see PRELIMINARY AIR HUNGER TESTS). Subjects rated every 30 s during the Subject-Control conditions and every 15 s during the Experimenter-Control and Constant-VT conditions. For each subject, PET_CO2 was increased to the same level in each of the conditions, and VT was the same in the first Subject-Control and the Experimenter-Control conditions (see below). At the end of each condition, subjects were asked to choose the best two descriptors of their respiratory sensations from a list of 10 phrases (i.e., "urge to breathe"; "breathing required work or effort"; "starved for air"; "tightness or constriction in the chest"; "air hunger"; "the size of the breaths"; "short of breath"; "a feeling of suffocation"; "did it feel like holding your breath?"; "did it feel like heavy exercise?").

Subject control. During the Subject-Control conditions, the subject was allowed to increase VT at any time. Subjects were instructed to increase VT using the mouth-operated switch as soon as they felt air hunger and to continue to increase VT until they no longer felt any sensation of air hunger. They were to "aim to be at a level of no air hungerso that you can spend an entire day at this level." This condition started at baseline VT and PET_CO2. In almost all instances, the subject was comfortable (i.e., air hunger ratings of zero and zero+) at these levels and made no signals for VT change. (The one exception was Subject 12, who, despite reporting zero and zero+ air hunger, increased VT during baseline [prior to change in PET_CO2] in the second Subject-Control condition of the second experimental session.)

Between 1 and 4 min after the start of the Subject-Control trial, PET_CO2 was increased surreptitiously from baseline to the level that had induced slight or moderate air hunger during the preliminary air hunger tests. While the experimenter maintained PET_CO2 at this elevated level (see VENTILATION), the subject increased VT as desired until he or she was comfortable. We terminated the trial 10 min following the initial increase in PET_CO2 or when the subject had stopped changing VT for 2 min. To allow for possible delays in perception of VT changes, subjects were instructed to wait 2 to 3 breaths between sips and to attend to whether they perceived a decrease in air hunger with each sip. Each PET_CO2 level was held for 4 min before data collection used for analysis started (5 times the half time for air hunger to reach a steady level after a step change in PET_CO2 [22]).

Experimenter control. During the Experimenter-Control condition, the experimenter matched the timing and delivery of PCO₂ and VT changes to that of the first Subject-Control condition.

Constant VT. During the Constant-VT condition, the experimenter matched the timing of PCO₂ changes to that of the first Subject-Control condition but VT was held at baseline level.

Data Analysis

Data recorded during the first 4 min following each step increase in PET_CO2 were not included in analysis; during this time, sensations were changing rapidly, and ventilator adjustments were frequent. Following this, the "steady-state" analysis period extended 3 to 6 min and included relatively few (0 to 3) changes in VT. (The first Subject-Control, Experimenter-Control, and Constant-VT conditions were always the same duration within each session, except for Subject 5 in which the Constant-VT conditions were terminated early because the subject reported extreme air hunger.) Mean VT and PET_CO2 were calculated separately for each condition, for each subject (CODAS analysis software). For all conditions, the mode was similar to the median air hunger ratings (in only two instances did they differ by as much as one scale increment); we present the median. Each subject participated in two experimental sessions; we discarded data from two experiments (one each for Subjects 1 and 10) because the previously determined level of PET_CO2 did not elicit air hunger on this particular experimental day, thus there was no relief to study. (This level of PET_CO2 had been chosen because it elicited slight or moderate air hunger during the preliminary sessions.) In four Subject-Control condition periods, the ventilator reached its safe VT limit before the subject stopped signaling for higher VT (i.e., Subject 5, the second Subject-Control condition in each session; Subject 12, both Subject-Control conditions in the second session); we included these data at the possible cost of slightly underestimating the VT change needed to complete relief. Data from each subject were weighted equally in group means.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Comparison of Reponses between Subject-Control, Experimeter-Control, and Constant-VT Conditions

An example of the time course of an experimental session for one subject is illustrated in Figure 1 (Subject 5, Session 1). Note that the time course of the manipulation of PCO₂ and VT during the Experimenter-Control condition closely followed that of the first Subject-Control condition. The increase in PET_CO2 during Constant VT elicited reports of extreme air hunger before steady state was reached in this case (this was uncommon). Despite this very strong discomfort at fixed VT, increased VT provided relief at elevated PET_CO2 during the Subject-Control and Experimenter-Control conditions (i.e., the subject reported minimal air hunger).

View larger version (21K): [in this window] [in a new window]   Figure 1.   An example of the time course of an experimental session in Subject 5. Experimental conditions: SC1 = first Subject-Control condition; EC = Experimenter-Control condition; CVT = Constant-VT condition; SC2 = second Subject-Control condition. Ratings: Z = zerono air hunger; S = slighta definite sensation of air hunger but one that is mild and easily tolerable; M = moderatea strong sensation but one that could be tolerated for a few minutes; E = extremean intolerable level of respiratory discomfort. Data are interrupted for rest periods between each condition. PETCO2 was increased to the same level in each of the conditions (note that for the first few breaths following an increase in PETCO2, PICO2 exceeds PETCO2). During the Constant-VT condition, elevated PETCO2 produced a rating of extreme air hunger (at which point we immediately reset PETCO2 to baseline and increased VT), whereas in the other conditions when VT was elevated in response to the increased PETCO2, either by the subject or the experimenter, the subject typically reported zero+ air hunger.

Data from each subject, along with the group means, are presented in Figure 2. When in control of the ventilator, all subjects increased VT in response to elevated PET_CO2; as a result, air hunger ratings were lower during the two Subject-Control conditions than during the Constant-VT condition. Two statistical tests in combination support this conclusion. First, we determined that the air hunger ratings differed significantly among these three conditions (Quade test; p = 0.0185). By inspection, it is evident that the two Subject-Control conditions produced nearly identical ratings and that air hunger was higher in the intervening Constant-VT condition. We confirmed this by testing whether the increase in air hunger going from the Subject-Control condition to the Constant-VT condition was equal to the decrease going from Constant VT to the second Subject-Control condition; the increase was not significantly different from the decrease (Wilcoxon sign test; p = 0.5). Within an experiment, each subject performed two Subject-Control conditions wherein they reliably increased VT to the same level (first Subject-Control condition, mean = 1,458 ml; second Subject-Control condition mean = 1,447 ml; Pearson's r = 0.93, p = 0.008) and achieved minimal air hunger (zero to zero+). We note here that this does include data from two subjects in whom the ventilator reached its limit before they stopped signaling for higher VT (see DATA ANALYSIS); yet, at this point, these subjects rated zero and zero+ air hunger.

View larger version (16K): [in this window] [in a new window]   Figure 2.   Individual and group average respiratory and perceptual responses for each condition. Experimental conditions: R = resting values before the start of the experimental condition; SC1 = first Subject-Control condition; CVT = Constant-VT condition; EC = Experimenter-Control condition; SC2 = second Subject-Control condition. Ratings of air hunger: Z = zero; S = slight; M = moderate; E = extreme. Within each condition, Subjects 1, 5, 10, and 12 are presented from left to right; triangle: first experimental session; inverted triangle: second experimental session (note: for subjects who have two sessions, the symbol looks like the Star of David when the two sessions overlap); horizontal line: group average; PETCO2 bars, ± SD (smaller than symbol size in many instances). Within subjects, PETCO2 is similar between conditions and VT is elevated to similar levels during the Subject-Control and Experimenter-Control conditions. Subjects report less air hunger during these two conditions than during Constant-VT, and there was no difference in air hunger ratings between Subject-Control and Experimenter-Control conditions. In addition, subjects increased VT to the same level during both Subject-Control conditions. (Note: the rating of extreme observed for Subject 5 during both Constant-VT conditions occurred before steady state and we report VT and PETCO2 levels accordingly; all other values were determined following steady state, as described in text.)

During the Experimenter-Control condition, the experimenter accurately copied the VT changes (timing and amplitude) made by the subject during the first Subject-Control condition (for example, see Figure 1; also note mean VT within subjects between these conditions, Figure 2). There were no differences in the air hunger reported between the Subject-Control condition and the Experimenter-Control condition (Figure 2). That is, out of six experiments, all but one air hunger rating point was exactly the same, and this rating was off by one increment on the scale (i.e., slight versus zero+).

Sense of Control

All subjects reported that they liked having control of the ventilator (e.g., Subject 12 reported that "I could control my breath size and make myself comfortable") and preferred participating in the Subject-Control conditions to the other two conditions. Three subjects reported that, given the opportunity, they would like to have control of their ventilation in everyday life (e.g., to be able to increase breath size during speech or when they were waiting to have their airways suctioned).

Quality of Respiratory Sensations

The quality of the respiratory sensation reported during elevated PET_CO2 in the Constant-VT condition was similar among subjects: the descriptors (see PROCEDURES) most often reported were "urge to breathe" (four of six Constant-VT experiments), "starved for air" (three of six Constant-VT experiments), and "short of breath" (three of six Constant-VT experiments). During Subject-Control and Experimenter-Control conditions, subjects reported minimal air hunger (median centers around zero+; see Figure 2), but, in those instances in which they did sense some discomfort, the quality of the sensation was similar to that reported in the Constant-VT condition.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study demonstrated that paralyzed, ventilator- dependent patients were able to relieve CO₂-induced air hunger ( 15 mm Hg above resting PET_CO2) by increasing VT using a mouth-operated switch (nonrespiratory cortical motor pathways). This is consistent with recent findings that showed that mechanically ventilated, healthy subjects could relieve CO₂-induced air hunger by requesting and receiving increased minute ventilation (23). Previous studies have shown that the perception of air hunger arises from chemoreceptive input, either via direct projections to the cortex, or via projections from the reflexively stimulated respiratory centers (9, 10, 24). Other studies have shown that air hunger is relieved by pulmonary vagal mechanoreceptor activity (11) and possibly chest wall mechanoreceptor activity (25, 26).

Heretofore, people with paralyzed respiratory muscles could not increase their breathing in response to hypercapnia. When PCO₂ is increased in intact individuals, it is not clear how much of the associated increase in ventilation originates in brainstem reflexes and how much originates in the cerebral cortex (i.e., a behavioral response to relieve air hunger). The mouth-operated switch used in the current experiment enabled the paralyzed subjects to change ventilation in response to hypercapnia, but that response could not have involved a reflex because the efferent limb of the reflex is absent in these patients. Moreover, the design of the experiment prohibited an automatic respiratory muscle response, since the trigger on the ventilator was set so that subjects could not generate a breath even if they had spared respiratory muscles (primary or accessory). Our data demonstrate that all of the air hunger could be relieved by a behavioral act; these subjects were able to consciously appreciate respiratory sensations (arising from chemoreceptors and lungs) and respond with willful (nonrespiratory) motor acts.

In addition to the neural input from mechanoreceptors, one might have expected additional relief from higher cortical processes, namely the sense of being in control of one's own ventilation. Subjects got the same relief regardless of whether they controlled VT or an experimenter controlled VT. Therefore, we conclude that the relief was solely a function of increased mechanoreceptor traffic (e.g., pulmonary stretch receptors) and was not due to having control of ventilation.

Implications

Our findings suggest that patients unable to trigger ventilator breaths because of inefficient respiratory muscles or severed motor pathways could effectively control their ventilator VT, and raise the novel idea of giving such patients a switch operated by nonrespiratory muscles to control their ventilator. Such control could provide some benefits; for example, we found that most subjects liked having control, and such control might enable patients to adjust ventilation for speech, or during airway leaks (see, for example, reference 27). On the other hand, within this small group of paralyzed patients, we found no difference in air hunger relief whether the experimenter or the subject controlled VT, indicating little reason to give patients control of their ventilation for that reason alone; such patients would likely be just as comfortable if their ventilatory parameters were properly adjusted by a health care provider.