INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A normal cough is considered to be of value as protection against infection of the respiratory tract. In patients unable to voluntarily activate the abdominal muscles it is possible to create an artificial cough by direct electrical stimulation of the abdominal muscles (1). These muscles may also be activated by percutaneous magnetic stimulation of the thoracic nerve roots (2, 3). We have previously demonstrated that an expiratory maneuver of a comparable magnitude to that achieved during a voluntary cough may be obtained using paired magnetic stimulation of the thoracic nerve roots (4). Recently, Lin and colleagues (5, 6) used a repetitive magnetic stimulator to simulate cough in both normal subjects (5) and patients with spinal cord injury (SCI) (6).

It therefore seems likely that repetitive magnetic stimulation of the thoracic nerve roots may have an increasing role to play in the prophylaxis and treatment of respiratory tract infection in selected patients. However, certain practical questions relating to the technique remain unanswered by current published work. Specifically, the optimal stimulation protocol in terms of train length, posture, and frequency is unknown. Secondly, the value of using a valve at the mouth to enhance glottic function in conjunction with repetitive magnetic stimulation of the thoracic nerve roots has not been evaluated. Thirdly, it is not clear whether the lung volume at the time of stimulation could importantly influence tension generation of the abdominal muscles. The aim of the present study was therefore to address these questions in healthy subjects in order to better define a stimulation protocol for clinical use. The factors governing cough efficacy (specifically is pressure or flow more useful) have not, to our knowledge, been validated from clinical studies and might in any case differ between patients with different diagnoses. Accordingly, rather than deciding a single most important outcome measure, the data are presented for the gastric and esophageal pressure generated as well as peak expiratory flow.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Nine healthy men 29 to 54 yr of age volunteered for this study; between five and seven completed each subsection. The subjects were experienced in physiologic studies and were free of neurologic and pulmonary disease. The protocol was approved by our ethical committee, and all subjects gave their informed consent.

Data Acquisition

Gastric, esophageal and transdiaphragmatic pressures (Pga, Pes, Pdi) were obtained using a pair of commercially available latex balloon catheters (PK Morgan, Rainham, Kent, UK) 110 cm in length placed in the stomach and esophagus in the conventional manner. The pressure-volume characteristics of the balloons were evaluated using the method of Mead and coworkers (7); because expiratory maneuvers were the subject of the investigation the esophageal balloon was filled with 2 ml of air instead of the 0.5 ml normally used. The catheters were connected to differential pressure transducers (Validyne MP45-1; Validyne, Northridge, CA), carrier amplifiers (PK Morgan), a 12-bit NB-MIO-16 analogue-digital board (National Instruments, Austin, TX) and a Macintosh Quadra Centris 650 personal computer (Apple Computer Inc., Cupertino, CA) running Labview software (National Instruments). Pdi was obtained on-line by subtraction of Pes from Pga. For some studies mouth pressure was also measured. For this purpose a mouthpiece was used containing a custom-built balloon occlusion valve driven by the wall supply of compressed air (8). The valve was used to close the mouthpiece by the operator at the appropriate time (usually resting end-expiration). This valve was controlled so that it automatically opened at a preset mouth pressure. For this purpose and for measurement, pressure proximal to the valve was sampled via a 70-cm tube of 1 mm internal diameter connected to a third transducer (Validyne MP45-1) and thence to the system detailed above. When necessary, expiratory flow was measured from a pneumotachograph head (Mercury CS5; Mercury, Kilwinning, Scotland) placed distal to the valve. Pressure and volume signals were sampled at a minimum of 100 Hz.

Diaphragm electromyographic (EMG) data were obtained from two subjects during thoracic nerve stimulation using an esophageal double-pair electrode constructed within our laboratory (9). The position of this electrode was confirmed at the start of the study by obtaining signals during a voluntary sniff and unilateral electrical phrenic nerve stimulation. The criteria used to define correct positioning was that unilateral stimuli produced action potentials in the upper and lower pairs of equal magnitude and opposing polarity; signals from the lower pair were used for analysis. These signals were passed via short leads to a Neurosign 100 amplifier (Magstim Co. Ltd., Whitland, Dyfed, Wales) and displayed via a combined pressure and EMG recording program also constructed using LabView software with a recording frequency of 2 kHz. The signals underwent bandpass filtering in the amplifier to exclude signals outside the range 10 Hz and 10 kHz but were not subsequently altered.

Stimulator: Description and Limitations

The stimulator used was a Magstim rapid (Magstim Co.) augmented by four booster packs. Construction of a repetitive magnetic stimulator presents more complex technical problems than are involved with a stimulator that gives only single stimuli. Essentially a conventional stimulator draws electrical energy (typically 500 J) from the mains supply, which is stored in a capacitor. This energy is then discharged into the stimulating coil in less than 100 µs in order to create the magnetic field, which in turn depolarizes the underlying nerve. Thus the main technical issues with a conventional stimulator relate to the safe transfer of electrical energy from capacitor to coil (for a more detailed consideration, see Reference 10). In the case of a repetitive stimulator there is an additional problem of how to charge the capacitor; for example, if the machine is operating at 20 Hz then the capacitor must be charged in less than 50 ms. Thus repetitive stimulators have a trade-off between the power that can be used and the frequency of stimulation. This problem can be alleviated by adding additional units to charge the capacitor more rapidly. For this reason we specifically modified our stimulator by adding four booster packs; even so, at frequencies greater than 25 Hz there was a progressive decline in the power of the stimulus that the machine could deliver such that at the maximal frequency (50 Hz) power was limited to 50% of maximal. A 19-pin 90-mm circular coil was used for all stimulation.

Stimulation Protocols

Determination of the optimal stimulation protocol. The aim of these studies was to determine the optimal stimulation train. Thus in each study all parameters except one were kept constant. Pga, Pes, mouth pressure (Pmo), and expiratory flow were measured throughout; between three and five stimulation trains were administered to each subject at each experimental condition. All stimuli were given at relaxed end-expiration (as judged by Pes) with the coil placed flat in the midline over the tenth intervertebral space with the handle at a right angle to the spine. This level was chosen for convenience since we have previously shown that positioning the coil in the intervertebral spaces between the seventh thoracic and first lumbar intervertebral space does not importantly influence the gastric pressure generated (3).

Train length. Six subjects were studied. Stimulation was performed with the subjects wearing noseclips and seated "cowboy style" such that their anterior chest wall rested against the back of a chair (4). Stimuli were given at 25 Hz with a valve opening pressure fixed for each subject; this pressure varied slightly between subjects but typically was 30 to 60 cm H₂O. The following train lengths were evaluated 100, 200, 300, 400, 500, and 600 ms.

Frequency. Six subjects were studied in the seated position as above. Stimuli were given in 600-ms trains with the valve opening pressure fixed as before. The following frequencies were evaluated 10, 15, 20, and 25 Hz.

Posture. Five subjects were studied. Stimuli were given at 25 Hz in 200-ms trains with the valve opening pressure fixed as before. Subjects were studied seated, prone, and supine.

Valve opening pressure. Five subjects were studied in the seated position as above. Stimuli were given at 25 Hz in 200-ms trains with the threshold opening pressure on the occlusion valve adjusted at increasing levels until the expiratory maneuver was no longer strong enough to open the valve.

Pressure-volume properties of the abdominal muscles. Seven subjects took part in this study. Pes and Pga were measured. Subjects were studied prone and stimuli were given over the tenth thoracic intervertebral space. Stimuli were given for 2 s at 50% of maximal power output. Stimuli were given at 1, 10, 20, 30, 40, and 50 Hz both at relaxed end-expiration (FRC) and also during relaxation after inspiration to TLC (Figure 1).

View larger version (14K): [in this window] [in a new window]   Figure 1.   Example of a trace obtained during the study of force-frequency and length. Pes and Pga are shown in the upper panels and transdiaphragmatic pressure (Pdi) is shown in the lower. Both traces show 2 s, 50% output in the prone position. The left panel was obtained at FRC and the right at TLC and illustrates how the subject achieved voluntary relaxation after inspiration.

At low frequencies, and particularly for the single twitches (1 Hz), the pressure generated can be influenced by prior contraction, a phenomenon termed potentiation (11). Therefore, the high frequency stimuli were given before the lower frequencies, so it may be assumed that the single twitches were potentiated.

Transmission of abdominal pressure to the thorax. During a voluntary cough (as for example in Figure 2) esophageal pressure closely matches gastric pressure and the Pdi is therefore numerically small. In contrast it became apparent that with repetitive thoracic nerve root stimulation the pattern of pressure generation resulted in an appreciable Pdi (as for example in Figures 3 and 4). To address this, additional studies were performed.

View larger version (13K): [in this window] [in a new window]   Figure 2.   Record obtained during a voluntary cough. Cough Pes is numerically similar to Cough Pga and, consequently, the Pdi (lower panel ) is small.

View larger version (13K): [in this window] [in a new window]   Figure 3.   Record obtained during a 600 ms, 25 Hz seated stimulation: upper trace, Pes and Pga; middle trace, mouth pressure; lower trace, expiratory flow. Note a single spike of expiratory flow coincident with the fall-off in mouth pressure as the valve opens.

View larger version (12K): [in this window] [in a new window]   Figure 4.   Record obtained during a 600 ms, 25 Hz seated stimulation: upper trace, Pes and Pga; middle trace, mouth pressure; lower trace, expiratory flow. Note that there are multiple spikes of expiratory flow, in contrast to Figure 3.

In two normal subjects recordings were made from an appropriately positioned diaphragm electrode during repetitive thoracic nerve stimulation as detailed above. In a third normal subject esophageal EMG recordings were obtained during supramaximal electrical phrenic nerve stimulation and during single thoracic nerve stimuli given from a Magstim DEM stimulator at 100% of maximal power output. In two normal subjects direct repetitive stimulation of the lower anterior abdominal muscle wall was performed by positioning the coil with its upper border at the level of the umbilicus and the handle perpendicular to the midline of the supine subject. In two subjects the effect of external compression was examined by applying a sudden pressure to the relaxed anterior abdominal wall of the supine subject.

Some observations are presented from clinical studies performed in patients with neurologic disease resulting in substantial diaphragm weakness with preservation of abdominal muscle strength. These experiments were performed to clarify the mechanisms responsible for the generation of transdiaphragmatic pressure during cough simulation in healthy subjects. Respiratory muscle strength was also measured in these patients using techniques described previously (12). Briefly, diaphragm strength was assessed as the Pdi elicited by bilateral anterior magnetic stimulation or cervical stimulation of the phrenic nerves and the maximal voluntary sniff. Expiratory muscle strength was measured by measuring the gastric pressure resulting from thoracic nerve root stimulation and a maximal voluntary cough (13). In one patient, EMG signals were also sampled from the esophagus during thoracic nerve stimulation using a multipair esophageal electrode; since this electrode could not be positioned with the aid of signals from the diaphragm we placed the proximal electrode of the pair 44 cm from the nose (14).

Conventions and Statistics

The gastric and transdiaphragmatic pressure elicited by a single stimulus is denoted by Tw Pga and Tw Pdi, respectively. The gastric and esophageal pressure changes elicited by repetitive thoracic nerve root stimulation is denoted by Pga_FMS and Pes_FMS, respectively; these changes were always positive in polarity. All pressures are measured baseline to peak. Statistics were analyzed using Statview 4.02 (Abacus Concepts, Berkeley, CA). We compared data for the same subjects for different experimental conditions using Wilcoxon's signed rank test. When appropriate, simple regression analysis was also used; a level of p < 0.05 was taken as significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Determination of the Optimal Stimulation Protocol

Stimulation was mildly uncomfortable but acceptable to all subjects. Normally, a single spike of expiratory flow was observed (Figure 3), but in some cases multiple spikes occurred (Figure 4), which were assumed to represent glottic incoordination. In some cases a fall off in Pga_FMS was observed; we suspect that this occurs because the coil is pushed away from the thoracic nerve roots by contraction of the paraspinal muscles.

Train length. Data from this study are shown in Table 1. In all subjects the rise in gastric pressure elicited by stimulation (Pga_FMS) increased progressively with train length up to 600 ms; however, the relationship between stimulus length and Pga_FMS was curvilinear (r = 0.96) such that increasing train duration from 300 to 600 ms resulted in a relatively small increase of mean Pga_FMS (from 148 to 164 cm H₂O: 10.8%). Similarly, the increase in Pes_FMS moving from 300 to 600 ms was 9 cm H₂O (86 to 95 cm H₂O). The median value of optimal stimulus interval was 350 ms. In contrast, as shown in Table 1, there was much greater variation for expiratory flow. The stimulation train length eliciting the greatest mean flow was 300 ms (351 L/min) and the mean greatest expiratory flow observed was 371 L/min (range, 195 to 665 L/min). These data suggest that a train length of approximately 300 ms is optimal for FMS.

Frequency. With increasing stimulation frequency Pga and Pes rose progressively (Figure 5). Both the mean Pga_FMS and Pes_FMS rose significantly between 10 and 15 Hz (both p < 0.03), and 15 and 20 Hz (both p < 0.03), but not 20 and 25 Hz (Pga_FMS, p = 0.14; Pes_FMS, p = 0.59). For all subjects Pes correlated with Pga (r ² = 0.78, p < 0.0001). The relationship between flow and stimulation frequency was less clear, with no significant difference between adjacent stimulation frequencies. There was no correlation between Pga_FMS or Pes_FMS and expiratory flow; nevertheless, mean expiratory flow elicited by 20 Hz stimulation (336 L/min) was significantly greater than that elicited by 10 Hz stimulation (279 L/min). We conclude that no additional advantage is secured by the use of 25 Hz stimulation over 20 Hz.

View larger version (13K): [in this window] [in a new window]   Figure 5.   Group data for the influence of stimulation frequency on gastric and esophageal pressure and expiratory flow. Error bars are the SEM.

Because the possibility of glottic narrowing had been suggested by the observation of multiple spikes of expiratory flow, we also analyzed the data from the frequency study in an alternative way, selecting for each subject the single largest flow from each experimental paradigm. The rationale for this analysis was that this trace was the one, for each individual, least likely to have glottic interference. Normalizing these data we found a relationship between driving pressure Pes_FMS and flow (r ² = 0.34, p = 0.01), suggesting the lack of relationship with the mean data did indeed result from glottic incoordination.

Posture. These data are presented in Figure 6. The seated posture resulted in a significantly greater expiratory flow as compared with both the supine and the prone postures (p < 0.05). There was no difference in the magnitude of the Pga_FMS between different postures, indicating a comparable intensity of stimulation.

View larger version (16K): [in this window] [in a new window]   Figure 6.   Group data for the influence of posture on gastric and esophageal pressure and expiratory flow. Error bars are the SEM.

Valve opening pressure. These data are presented in Figure 7. For each subject, Pmo at the time of valve opening showed a statistically significant relationship with peak expiratory flow. The mean magnitude of this increase was 2.8 L/min (range, 1.9 to 3.7 L/min) for each cm H₂O increase in Pmo.

View larger version (22K): [in this window] [in a new window]   Figure 7.   The effect of the occlusion valve is demonstrated by the relationship between mouth pressure at which the valve opens and expiratory flow. Individual data are shown. Where PmoFMS is very small this represents the pressure measured from the valve chamber when the valve is not used.

Pressure-Volume Properties of Human Abdominal Muscles

These data are presented in Table 2. Although Pga_FMS was significantly greater at TLC than at FRC for frequencies up to 30 Hz, the magnitude of the difference was small. Moreover, at 40 and 50 Hz, Pga_FMS was not significantly greater at TLC than at FRC. The ratio of Pga_FMS elicited by 10 Hz stimulation compared with that elicited by 50 Hz stimulation was significantly greater at TLC than at FRC (mean value, 0.41 at TLC compared with 0.32 at FRC; p < 0.02). The ratio PesF_MS:Pga_FMS was 0.57 at FRC and 0.72 at TLC, reflecting a smaller Pdi at TLC.

Transmission of Abdominal Pressure to the Thorax

In both normal subjects, signals were recorded from the esophageal electrode. However, these were small in amplitude (5 arbitrary units for Subject 8 and 10 arbitrary units for Subject 9) compared with signals elicited by phrenic nerve stimulation (120 arbitrary units for Subject 8 and 40 arbitrary units for Subject 9) and of different morphologic appearance.

Anterior stimulation of the lower abdominal wall resulted in a similar pattern of pressure generation to that elicited by thoracic nerve root stimulation; in particular, a substantial Pdi could be obtained. The maximal values obtained were: Subject 4, Pga_FMS was 74.8 cm H₂O with Pdi 47.8 cm H₂O; Subject 8, Pga_FMS was 76.2 cm H₂O with Pdi 44.5 cm H₂O. Sudden external compression of the abdomen also elicited a Pdi; an example is shown in Figure 8.

View larger version (12K): [in this window] [in a new window]   Figure 8.   The effect of the application of a sudden external pressure to the anterior abdominal wall. Pes is smaller than Pga, and a Pdi therefore exists.

Clinical data from the patients with diaphragm weakness are shown in Table 3. In these subjects thoracic nerve root stimulation failed to elicit a Pdi even when a substantial Tw Pga was obtained (Figure 9). The action potentials obtained during electrical phrenic nerve stimulation of a normal subject were narrow and of opposite polarity between upper and lower electrode pairs, confirming that the electrically active center of the diaphragm lay between the two pairs. In contrast, signals elicited by single T10 stimuli were of broad complex morphology and the polarity was similar between upper and lower electrode pairs, indicating a source either caudal or cranial to both electrode pairs (Figure 10A). In the context of T10 stimulation these signals were therefore assumed to originate from the abdominal muscles. In the patient in whom EMG recordings were made (Patient 4), no signals were obtained during electrical stimulation of the phrenic nerve, but low amplitude signals were obtained during thoracic nerve stimulation (Figure 10B), confirming that the esophageal electrode can record activity arising from muscles other than the diaphragm.

View larger version (15K): [in this window] [in a new window]   Figure 9.   The effect of a single thoracic nerve stimulation in a patient with diaphragm paralysis (data presented from Patient 1). No Pdi is observed.

View larger version (10K): [in this window] [in a new window]   Figure 10.   This figure shows EMG signals from an esophageal catheter containing two electrode pairs. In the right panel (A) data are shown from a normal subject. During electrical phrenic nerve stimulation ( panel A1) diaphragm action potentials are observed. Because the electrical activity of the diaphragm is centered between the upper and the lower pair, the signals are opposite in polarity. During single T10 stimulation ( panel A2) complex signals are obtained of similar polarity, suggesting a source or sources lying to the same side of both electrode pairs; probably the abdominal muscles. Similar records were obtained ( panel B) from an electrode placed in a patient (Patient 4) with bilateral diaphragm paralysis. No responses were obtained with phrenic nerve stimulation in this patient, demonstrating that potentials generated by the abdominal muscles can be recorded from an esophageal electrode.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study provides further evaluation, in healthy volunteers, of the technique of functional magnetic stimulation (FMS) of the abdominal muscles. In particular we report the influence of train length, posture, the use of an occlusion valve and lung volume that have not previously been examined. By measuring intra-abdominal and intrathoracic pressures we have been better able to understand the mechanism of FMS. Discussion of the significance of the findings follows a critique of the methods.

Critique of the Methods

Comparison with previous work. The aim of the present investigation was to better define a stimulation protocol for clinical use. The only other published work in this field is by Lin and coworkers (5, 6), and inevitably we compare our data with theirs. However, in making comparisons it should be noted that our stimulator model was different from theirs. Moreover, since our occlusion valve was constructed in our hospital (8), it is particular to our group.

The pressures and flows observed in the present study confirm previous data (4, 5) that repetitive magnetic stimulation of the thoracic nerve roots produces an expulsive maneuver comparable to a voluntary cough (for references see Reference 4).

What level to position the coil? It is acknowledged that the optimal level for coil positioning is not addressed by the present study. We investigated this question in our original study (3) and found that, over the lower thoracic intervertebral spaces, coil positioning did not importantly influence Tw Pga. We suspect therefore that positioning the coil at, for example, T9 or T11 would not materially influence our conclusions.

What parameter should be measured? Thoracic nerve root stimulation results in a positive gastric pressure, a positive intrathoracic pressure, a positive mouth pressure, and expiratory flow. It is unknown which parameter best correlates with cough efficacy in humans, although some data are available.

Bach and Saporito (15) studied 49 patients whose diagnoses were predominantly neuromuscular. The patients were weaned onto inspiratory pressure support, thus minimizing the importance of inspiratory muscle function. For these patients a peak expiratory flow during an assisted cough of > 160 L/min predicted eventual success. Thus (with a margin for safety), a simulated cough might be considered successful if it could generate a peak flow > 180 L/min. Similarly, we have previously shown, in patients with amyotrophic lateral sclerosis, that a cough Pga of greater than 50 cm H₂O is required to achieve intrathoracic airway compression (13). With appropriate adjustment of the stimulation protocol it proved possible to exceed these values during all experiment combinations, predicting that FMS could have a clinical role.

Previous studies (5, 6) have assessed outcome solely in terms of Pmo or expiratory flow. We observed a large variation in our subjects, with respect to the expiratory flow that could be obtained with FMS. This finding is well illustrated by Subjects 4 and 7 who had identical mean Pga_FMS at 500 ms, yet Subject 4 had an expiratory flow of 659 L/min against 175 L/min for Subject 7 (Table 1). Because our subjects were free of respiratory disease, we believe this is likely to be explained by the action of the glottis. This hypothesis is supported by the finding of multiple spikes of expiratory flow (Figure 4). During a normal cough, activity of the vocal cord muscles is closely coordinated with the activity of abdominal muscles and the diaphragm (16). This coordination is achieved centrally (17), and it is therefore not unexpected that glottic incoordination could be observed during an artificial cough. Therefore, we suggest that Pmo and expiratory flow could have potential limitations for the evaluation of an artificial cough.

Does glottic incoordination matter? Because it remains unclear whether the most important factor for expectoration is generation of flow or generation of intrathoracic pressure, it follows that it is unknown whether glottic incoordination might detract from the efficacy of the technique. However, for some potential clinical applications the patient would have a tracheostomy or endotracheal tube, thus eliminating considerations relating to glottic function.

Patient tolerance. Compared with functional electrical stimulation FMS has the advantage that cutaneous pain receptors are not activated. We have previously used single phrenic (18) and thoracic (13) nerve root stimuli in patients with severe medical illnesses and we predict that FMS would be acceptable to patients; this prediction is confirmed by the data of Lin and coworkers (6). Nevertheless the pressures produced can be substantial, and the technique would not be suitable in patients with unstable spinal cord injuries. As with all magnetic techniques, FMS is contraindicated in the presence of implanted metal objects.

Might the valve impair expectoration? Because our subjects were healthy, stimulation did not result in expectoration. It is conceivable that in patients with copious or viscous secretions that the use of an occlusion valve might be an obstacle to expectoration. If this proved to be so in clinical practice it would be necessary either to omit the valve (which would not preclude substantial pressures and flows; see Figure 7) or to develop protocols to remove secretions from the upper airway after stimulation. This might involve coming off the mouthpiece for conscious patients or poststimulation suctioning in intubated patients.

Electrical power. Repetitive magnetic stimulators require large current flows (up to 40 A). Before introducing the stimulator to service use it might be necessary to modify the electrical supply to the clinical area.

Significance of the Findings

Lung volume exerts a significant influence on the pressure- generating capacity of the inspiratory muscles (19, 20). In contrast, lung volume exerted a relatively modest influence on Pga_FMS, although, as with human limb (21) and diaphragm (20) muscle, the effect of length dependence of activation was demonstrable. It could be argued that the observation that lung volume has only a modest effect on abdominal muscle strength was obtained because subjects were studied prone, and in this position FRC is somewhat closer to TLC than when supine. However, the present data are consistent with data from our previous study (4) (which was performed seated) in which there was only a 17% increase in the Pga elicited by paired stimuli on moving from FRC to TLC. We therefore believe that the abdominal muscles are genuinely less influenced by lung volume change than are the inspiratory muscles. This may reflect the need for this muscle group to perform nonrespiratory tasks; for example, phonation (22) or posture control (23), which may be required at any lung volume. It is of interest that the Pdi observed at TLC was less than that at FRC. The origin of the Pdi is discussed below, but, for practical purposes, these data suggest that greater intrathoracic pressures could be achieved by stimulation at increased lung volume or, conversely, that lower stimulation intensities would be required to generate equivalent intrathoracic pressures.

Whether the efficacy of stimulated cough depends on patients moving to a high lung volume therefore depends on whether cough efficacy is dictated by flow or pressure. However, with the use of an interrupter valve it would be feasible to capture patients at high lung volume either after voluntary inspiration or (in the case of a ventilated patient) after an imposed breath. Indeed, FMS used in conjunction with a valve and positive pressure ventilation has similarities with the recognized technique of mechanical insufflation-exsufflation (24).

The present data (Figure 6) confirm our previous observation from single stimuli (3) that posture does not greatly influence Pga_FMS or Pes_FMS. In particular, the increase in expiratory flow observed between the prone or the supine position and the seated posture cannot be due to differences in stimulation technique since the Pga_FMS did not differ significantly between postures. The increase could, in part, be explained by increased end-expiratory lung volume since, when lying FRC decreases by between 0.5 to 1 L (25), with a consequent shift to the right along the expiratory flow-volume curve. A further contributing factor could be upper airway resistance, which is known to increase on adopting the supine posture (26). Whether adopting the upright posture is critical depends, like the conclusion with respect to strength, on whether flow or pressure is perceived as the most important outcome variable. Our data show, however, that substantial flows can be achieved even in the supine or prone positions and that posture does not importantly influence pressure generation.

The train length study was conducted at 25 Hz and we observed only a minimal increase in Pga_FMS after 300 ms; i.e., after 12 stimuli. The abdominal muscles have an approximately equal mix of Type I and Type II fibers (27), and our finding is therefore consistent with the known behavior of tension addition in mammalian skeletal muscle (for example, Reference 28). Lin and coworkers (5, 6) used a 2-s train, but a much shorter train may be equally effective. If so this would be more acceptable to patients.

The use of a valve during voluntary coughs and rapid expirations increases expired volume in a manner proportional to the pleural pressure generated (29). Our data confirm that peak expiratory flow is also increased using a valve with FMS, but the magnitude of increase that can be obtained using a valve (2.8 L/min for each cm H₂O of Pmo) has not been previously documented. It is, however, acknowledged that use of a valve may not be helpful in subjects who reach their maximal expiratory flow volume (MEFV) loop with FMS (30), but this does not exclude the possibility that a valve may benefit patients unable to do so. It may also be of practical benefit by arresting patients at a given (high) lung volume before performing FMS.

Because we chose to use our stimulator at 100% power, we could not evaluate the effect of stimulus frequencies greater than 25 Hz on expiratory flow. Lin and coworkers (5, 6) investigated the effect of stimulus frequency in normal subjects (5) and found a plateau in both mouth pressure and expired volume at 25 Hz. A similar phenomenon was found in their investigations of stimulus intensity in both normal subjects and patients with SCI (6). These observations could lead to the conclusion that further increases in stimulator power do not result in greater tension generation, but for this conclusion to be correct the force-frequency curve of the abdominal muscles would have to plateau at 25 Hz and the muscles within the field of stimulation would, electrophysiologically, have to be supramaximally stimulated. The abdominal muscles have a similar fiber type composition to the human quadriceps (27) that generates approximately 80% of its maximal tension in response to stimulation at 25 Hz and 100% at 50 Hz (28). In our study of lung volume we gave stimulations at frequencies as high as 50 Hz (albeit at 50% power output) and, as evidenced by data in Table 2, there was a mean increase in Pga_FMS of 34% between 30 and 50 Hz (p < 0.02). Our observations in this respect are similar to data presented by DiMarco and colleagues in the dog (31). Therefore, we suggest that the failure of Pmo and expiratory flow to increase beyond 25 Hz could be for reasons other than the magnitude of tension generation in the abdominal muscles.

To obtain a supramaximal response, stimulation should be supramaximal within the field of stimulation and, in addition, the edge of the field would have to have a clear edge such that increasing stimulus intensity did not increase field strength at the edge of the field (since the thoracic nerve roots extend beyond the limits of the coil). We did not investigate supramaximality in the present study but, in contrast to Lin and coworkers, previous investigators have been unable to demonstrate that thoracic nerve stimulation supramaximally activates the abdominal muscles either in terms of the action potential (2) or the pressure generation (3). Lin and coworkers reported in normal subjects that their stimulus was supramaximal in terms of the action potential recorded from the external oblique and rectus abdominis, although they did not examine transversus abdominis. Their data also show that, as expected from the known field pattern produced by circular coils (10), there is a gradual (rather than abrupt) fall off in the size of the action potential elicited if the coil is moved craniocaudally along the thoracic spine. Therefore, although the stimulus produced by FMS is very substantial, there presently exists insufficient proof that it is supramaximal. Nevertheless, at 25 Hz in the present study the mean Pga_FMS was 166 cm H₂O and the mean Pes_FMS was 108 cm H₂O compared with 103 cm H₂O (for Pmo) reported by Lin and coworkers in healthy subjects. Thus the maximal power of our machine appears comparable with that used by Lin and colleagues. However we caution that supramaximality cannot be satisfactorily addressed from Pmo since (as discussed below) Pmo is not a faithful reflection of Pga.

Our data show that Pes_FMS and Pmo_FMS are always numerically smaller than Pga_FMS (Figures 3, 4, and 5); by definition this creates a transdiaphragmatic pressure, the origin of which is of practical and physiologic interest. One possibility would be that thoracic nerve root stimulation directly stimulates the intramuscular branches of the phrenic nerve in the crural diaphragm. This hypothesis merits consideration since the crural diaphragm lies anterior to the lower thoracic spine and stimulation of crural diaphragm alone is known to result in a transdiaphragmatic pressure (32). This hypothesis is supported by the observation that patients with a paralyzed diaphragm but intact abdominal muscles do not generate a Pdi during thoracic nerve stimulation and also, potentially, by the observation that small amplitude signals were obtained from the esophageal electrode in the present study. However, even though these signals were recorded from the esophagus, they do not necessarily arise from the diaphragm; specifically they might represent far-field potentials from the abdominal muscles. In this connection we have demonstrated that our present electrode is sufficiently sensitive to detect far-field potentials presumed to arise from the muscles of the upper thorax (9) and from the abdominal muscles (Figure 10).

Moreover, other data exist which suggest that a Pdi can be obtained simply as a result of a rise in intra-abdominal pressure. First, we were able to generate a Pdi by stimulation of the anterior wall below the umbilicus; this site is distant from the diaphragm. It is of interest that a Pdi was also obtained by Mier and colleagues (33) when they performed direct electrical stimulation of the abdominal muscles. Second, sudden external pressure to the anterior abdominal wall also resulted in a Pdi (Figure 8). We therefore suspect that the Pdi does not, at least not wholly, arise because of exogenous stimulation of the crural diaphragm. An alternative mechanism could be that the diaphragm undergoes a reflex contraction in response to sudden stretch; such a mechanism has been demonstrated for opposing groups of insect flight muscles and in mammalian skeletal muscle (34). In support of this proposition are data from Muller and colleagues (35) showing both that the human diaphragm has spindles and that it does have a stretch reflex to an externally applied load to the abdomen. Clearly an intact contractile apparatus in the diaphragm is a prerequisite for such a reflex; thus our observations in patients with diaphragm weakness are also accommodated by this mechanism.

In conclusion we have confirmed that FMS of the abdominal muscles can generate substantial positive intra-abdominal and intrathoracic pressures. There were no adverse side effects of the technique and it was tolerated by normal subjects. With our apparatus pressure generation is maximized by having a train length of at least 300 ms and using a frequency of 25 Hz. Posture and valve use were not important determinants of Pga_FMS or Pes_FMS. Lung volume exerted only a minor influence on Pga_FMS, but the ratio Pes_FMS:Pga_FMS was increased at TLC compared with FRC. Expiratory flow can be increased by adopting a seated posture and using an occlusion valve with a threshold for opening close to the maximal Pes_FMS generated by the stimulus train; however, expiratory flow generation is susceptible to interference from glottic incoordination. The efficacy of FMS now requires testing in a functional model of cough (for example, clearance of inhaled radiolabeled particles); this approach could also be used to clarify whether pressure or flow is the most important determinant of cough efficacy.