INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It is now well established that an increase in ventilator flow rate (), at a constant tidal volume (VT), results in an increase in respiratory rate (RR) and minute ventilation (E), in both normal subjects (1) and in ventilator-dependent patients (5). We wished to determine whether this response is due to a reduction in neural TI (TIn), neural TE (TEn), or both. This information is relevant to understanding of the mechanism(s) by which the tachypneic effect of ventilator inspiratory flow is mediated. The relation between ventilator flow and TIn is additionally important for the following two reasons:

1. TIn determines to a considerable extent the average pressure output of inspiratory muscles during mechanical ventilation (see DISCUSSION). Defining the relation between ventilator flow and TIn would, accordingly, have implications relating to the work of breathing and inspiratory pressure output at different ventilator settings.
2. With conventional mechanical ventilation (volume cycled or pressure support ventilation [PSV]), the end of the ventilator cycle is not linked to the patient's neural inspiratory phase. The relation between flow rate and TIn may be expected to have important implications for synchrony between the patient's and the ventilator's inspiratory phases.

In the present study, we determined the response of neural TIn and TEn to systematic changes in inspiratory flow rate, at a constant VT, in normal subjects. We found that the increase in RR is produced principally by a reduction in TIn. The reduction in TIn with increasing flow was sufficiently pronounced that, above a certain flow, neural inspiration in many subjects terminated almost immediately after triggering. The results also indicate that the reduction in TIn at higher flow rates is not the result of more rapid attainment of the volume threshold for inspiratory termination (the classical Hering-Breuer [H-B] reflex), but is largely a specific effect of flow.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied 15 healthy nonsmoking volunteers (nine men and six women). Their average age was 28.0 yr (range: 20 to 39 yr), average height 170 cm (range: 152 to 188 cm), and average weight 70 kg (range: 45 to 97 kg). None of the subjects was aware of the purpose of the study. Subjects were simply told that they would be connected to a ventilator and that the ventilator settings would be changed from time to time. They were asked to relax and not to think about their breathing. They were instructed to signal or to disconnect themselves from the mouthpiece if they experienced any distress. The project was approved by the human experimentation committee of our institution, and a consent form was signed by each subject.

The subjects were studied while awake and in the sitting position in a comfortable chair. They were connected via a mouthpiece to a locally constructed (Winnipeg, Manitoba, Canada) ventilator. A pneumotachograph was inserted between the Y-piece of the ventilator and the mouthpiece to monitor and its integral, VT. Airway pressure (Paw) was monitored through an appropriate port near the mouthpiece. In five subjects, electrical activity of the diaphragm (Edi) was monitored by inserting an esophageal catheter equipped with two silver bands, 3 cm apart, near its tip. The catheter was first advanced to 35 cm from the nares. The final position of the catheter was associated with the most intense activity during spontaneous breathing.

The Winnipeg Ventilator is a flow triggered, piston-based ventilator. The flow level at which triggering occurred was set at the lowest level not associated with false triggering (usually 0.1 to 0.2 L/s). The flow pattern of this ventilator, in the volume-cycled mode, is shown in Figure 1 at different peak flow settings. Peak flow is reached at approximately 0.2 s after triggering, and then declines gently until the set volume is reached. For the purpose of our study, the backup rate of the ventilator was set at zero to insure that all breaths were triggered by the subject. End-expiratory pressure was zero (i.e., no positive end-expiratory pressure was applied).

View larger version (17K): [in this window] [in a new window]   Figure 1.   Tracings representing the typical flow pattern delivered by the ventilator at different peak flow settings.

Signals related to flow, volume, and Paw were stored at 50 Hz in a personal computer by means of commercial software (Windaq; DATAQ Instruments Inc., Akron, OH). When Edi was monitored, the electrical signal was bandpass filtered (30 to 1,000 Hz) and stored at 500 Hz (two subjects) or 1,000 Hz (three subjects). The signal was processed post hoc by computer as follows: (1) the signal was rectified; (2) a custom-written computer algorithm scanned the signal, identifying QRS complexes; (3) the raw electrical activity was deleted over the period corresponding to the QRS complex (usually 100 ms), and was replaced by artificial values representing a linear interpolation between average activities immediately before and immediately after the QRS complex (Figure 2, channel 5 [Processed Edi]); and (4) a moving average (100 ms) of the rectified signal was obtained (Figure 2, channel 6 [Edi MA]).

View larger version (73K): [in this window] [in a new window]   Figure 2.   Tracings of airway pressure (Paw); flow; volume; raw EMG of the diaphragm (raw Edi); Edi after removal of the QRS complex and rectification (processed Edi), and after averaging (Edi MA); and calculated Pmus. (A) Onset of neural inspiration (point at which Pmus begins rising and there is a change in trajectory of Paw and flow). (B) Ventilator triggering (point at which Paw begins rising). C1 = end of TIn defined from Pmus (point at which Pmus begins steady decline to baseline). C2 = end of TIn defined from Edi (point at which EdiMA begins a steady decline to baseline). (D) End of ventilator cycle (point at which Paw and flow begin a rapid decline). TIn is interval A-C. Triggering time (Ttr) is interval A-B. Tend is interval C-D. Vth is volume at the end of neural inspiration (point C1 or C2).

Protocol. Upon switching to the volume-cycled mode, we set the ventilator VT and peak flow settings to the lowest levels deemed comfortable by the subject. This was achieved by trying different settings while obtaining feedback from the subject. After these baseline settings were established ( = 0.81 ± 0.19 L/s, VT = 1.10 ± 0.4 L), ventilator flow was changed to different levels, with each level maintained for four breaths. From six to eight flow steps were tested. These spanned the range between baseline flow and a maximum peak flow setting of 2.5 L/s (Figure 1). VT was kept constant at its baseline value. Each flow setting was tested from four to eight times, which, after deletion of breaths having artifacts (swallowing, movement, etc.), resulted in 12 to 32 breaths per flow setting per subject. To evaluate the validity of the four breath tests, transitions from one flow rate to a higher flow rate were maintained for 1 to 2 min in 10 subjects. Three such longer transitions were made in each subject.

Throughout the foregoing protocol, silence was maintained in the laboratory and the subject was deprived of any clues that would alert him/her to an impending change in ventilator settings. The order of application of the different flow rates was randomized.

After completion of the flow-adjustment protocol, the ventilator backup rate and VT were gradually increased until ventilation was controlled without evidence of respiratory effort. This was determined from the absence of triggering efforts; from reproducible Paw waveforms during inspiration without evidence of scalloping; and from a smooth, exponential decline in flow during expiration (6). At this point the subject was told that exhalation would be blocked for a brief period, and was asked to relax during these maneuvers. Several such end-inspiratory occlusions were applied. Maneuvers resulting in an acceptably stable plateau pressure (Pplat) were used to calculate elastance (E) and resistance (R) according to standard formulae (E = Pplat/VT; R = [Ppeak  Pplat] / ).

Analysis. The pressure developed by respiratory muscles (Pmus) was continuously computed from the stored Paw, , and V data according to the equation of motion (7), modified for subjects on a ventilator (8):

(1)

We assumed that volume at the beginning of inspiration was passive FRC. We ignored inertia. Inertial pressure losses are usually very small during spontaneous breathing, since the acceleration and deceleration of flow during natural breathing are relatively very small (7). In the current study, particularly at the highest flow settings, the sudden reduction in flow at the end of the ventilator cycle was associated with deceleration values that were sufficiently high (up to 100 L/s²) to be associated with measurable inertial pressures. Because we did not include inertial forces in the equation of motion, a brief artifact occurred in the calculated Pmus in some subjects at the time of switching off of the ventilator. We elected not to include inertial losses in the computation of Pmus, since accurate values of inertia are not available for the conditions of our experiment, and because the artifactual nature of the pressure spike could be easily determined (its occurrence was limited to the phase of sharp decline in , it had a very brief duration, and its peak occurred at the point of fastest deceleration).

The onset of neural inspiration was defined as the point at which Pmus began rising from baseline (Point A, Figure 2). This point was invariably associated with a change in trajectory of Paw and flow in an inspiratory direction (in fact, it is this change in trajectory of Paw and flow that is responsible for the increase in calculated Pmus). The end of neural inspiration was defined as the point at which Pmus began a monotonic decline toward baseline (Point C₁, Figure 2). TIn was taken as the interval between these two points. We also determined the point at which the triggering of the ventilator occurred (the point at which Paw changed direction from falling to rising, and at which flow rate increased abruptly [Point B, Figure 2]). The time to triggering (Ttr) was the difference between the time of onset of inspiration and that of triggering (Points A and B, respectively, Figure 2). The end of the ventilator cycle was identified as the point at which Paw and flow began declining sharply (Point D, Figure 2). The interval between the end of the ventilator cycle and the end of neural inspiration (Tend) was calculated (Interval D-C₁, Figure 2). Total cycle duration (TTOT) was taken as the interval between the onset of inspiration and the onset of the next inspiration (Interval A-A, Figure 2). Neural expiratory duration was calculated from Ttot  TIn. The volume at which neural inspiration terminated (Vth) was noted (volume at point C₁, Vth, Figure 2).

In the five subjects in whom Edi was monitored, the same variables were measured from both the Pmus and averaged Edi signals. Scoring from the two signals was done independently (Edi was off the screen when scoring from Pmus was done, and vice versa). When scoring was done from Edi, the onset of inspiration was defined from the point of inspiratory inflection in Paw and flow. One or both of these signals was inspected at high gain to facilitate the detection of this event. We chose to identify inspiratory onset from Paw and (as opposed to using Edi) for two reasons. First, baseline artifacts in the moving average of Edi often made it difficult to precisely identify the point at which Edi began deviating from baseline. Second, we wanted to use the same point as the onset of inspiration for both Pmus and Edi (but without looking at Pmus), so that comparisons between TIn as measured with the two signals primarily reflected differences in the definition of the end of TIn. This is the point at which there is uncertainty about the validity of Pmus. The end of neural inspiration was defined as the point at which averaged Edi began a monotonic descent to baseline (Point C₂, Figure 2). Other variables were determined as with the Pmus analysis.

Results for each variable were segregated according to peak flow rate. Average values were obtained for each subject at each flow setting. The effect of flow on the variables of interest was determined for the 15 subjects through analysis of variance for repeated measures (ANOVA). A value of p < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Figures 3A to 3E shows the relation between TIn measured from Pmus (TI Pmus) and from Edi (TIEMG) in the five subjects in whom Edi was monitored. Each point in the figure represents a single breath. There was good agreement between the two measurements over the entire flow range in all subjects (r = 0.78 to 0.99, Figures 3A to 3E). The variability appeared to be mostly random, in that data points were scattered on both sides of the line of identity. To assess whether there were any systematic differences between the two measurements at the different flow rates, we averaged the values of TIPmus and TIEMG over the different flow ranges used in individual subjects. The results are shown in Figure 3F. As can be seen, although some differences remained in some cases after averaging, these were small. More importantly, these differences did not affect the estimated change in TIn as a result of changes in flow rate. Thus, the average (± SD) change in TIn between the highest and lowest flow rates in the five subjects was 0.612 ± 0.18 s when measured from Pmus, and 0.618 ± 0.24 s when measured from Edi. There was no significant difference between these means, whereas there was a significant correlation between the two measurements (r = 0.87).

View larger version (33K): [in this window] [in a new window]   Figure 3.   Relation between neural inspiratory duration determined from EMG of diaphragm (TIEMG) and from calculated Pmus in five subjects in whom both measurements were made. (A-E ) Relation in individual subjects. Each point represents a single breath. Data are from observations at all flow rates studied. (F ) Same results after averaging data points at different flow rates. Each symbol represents a different subject. In each subject, points with lower TI are associated with higher flow rates, and vice versa.

Figure 4 shows representative examples of the response of TIn at different flow settings in one subject. Note the progressive reduction in TIn and Ttot as flow increased. Note also that at the highest flow settings, TIn terminated very shortly (0.28 s) after triggering.

View larger version (68K): [in this window] [in a new window]   Figure 4.   Representative examples of responses observed at five different peak flow settings in one subject. Paw = airway pressure, Edi = diaphragmatic EMG, EdiMA = moving average of Edi after rectification and removal of QRS complex, Pmus = calculated pressure output of inspiratory muscles. Note the progressive reduction in TIn and Ttot as flow increases.

The individual changes in TIn and TEn are shown in Figure 5. With the exception of one subject, in whom the response lagged somewhat, TIn decreased progressively in all subjects as flow increased, beginning with the first step increase (Figure 5, left). The average response of TIn for the group is shown in Figure 6A. Highly significant reductions in TIn were observed beginning with the second step change in flow (p < 0.0001 or lower, ANOVA). TIn decreased to less than half its baseline value (0.62 ± 0.26 s versus 1.44 ± 0.34 s) between baseline flow ( = 0.82 ± 0.2 L/s) and the highest flow rate ( = 2.58 ± 0.2 L/s). The response appeared to level off at last step increase in flow. This, however, was largely due to the fact that in seven subjects, TIn had reached the minimum possible value that can be achieved in these experiments, this minimum value being dictated by the triggering delay (0.28 ± 0.05 s) and a finite response latency (Figure 5, left).

View larger version (23K): [in this window] [in a new window]   Figure 5.   Individual responses of neural inspiratory duration (left) and neural expiratory duration to changes in ventilator flow rate. Each line represents one subject. Solid lines are from five subjects in whom timing was assessed from Edi. Interrupted lines are from subjects in whom timing was assessed from Pmus. Note that individual responses of TIn are fairly uniform among subjects, whereas the responses of TEn are inconsistent.

View larger version (19K): [in this window] [in a new window]   Figure 6.   Average responses of various output variables at different flow rates. TIn = neural inspiratory duration, TEn = neural expiratory duration, Ttot = breath duration, TIn/Ttot = inspiratory duty cycle, Tend = time interval between the end of ventilator inflation phase and the end of neural inspiration. *Significant difference from value at lowest flow (ANOVA). Bars represent SEM.

The individual responses of TEn were inconsistent both within and among subjects (Figure 5, right). The average response showed a tendency (p = NS) for TEn to decrease over the first few flow steps (Figure 6B). Changes in Ttot (Figure 6C) reflected the combined effect of changes in TIn and TEn, showing a relatively steep decrease over the first three steps (both TIn and TEn decreasing), with a tendency for the response to level off in the high flow range as a result of reciprocal changes in TIn and TEn. TIn/Ttot decreased significantly as flow increased (Figure 6D). At baseline flow the ventilator cycle extended beyond TIn by an average (± SD) of 0.43 ± 0.28 s (Figure 6E). This difference (Tend) showed a biphasic response to increasing flow. Initially, TIn decreased less than ventilator TI, and the difference (Tend) narrowed to a minimum of 0.21 ± 0.19 (p < 0.003 by comparison with baseline). At still higher flows, TIn decreased more than TIM, and Tend increased again. There was no change in triggering delay as flow increased (Ttr = 0.32 ± 0.06 s at baseline versus 0.28 ± 0.05 s at the highest flow).

Figure 7 shows the average breath-by-breath change in TIn after flow transitions that were maintained for at least 12 breaths. The change in TIn was nearly complete within the first breath after transition, and there were no systematic time-dependent changes thereafter (ANOVA).

View larger version (15K): [in this window] [in a new window]   Figure 7.   Average time course of flow and neural inspiratory duration during flow transitions that lasted at least 12 breaths. The results are the average of data from 10 subjects in whom this was done. (B) Average value immediately before flow transition. *Significant differences from baseline value (ANOVA). There were no significant differences between any of the breaths following transition. Bars represent SEM.

Figure 8 shows the average relation between TIn and the volume delivered by the ventilator at the time of switching (Vth). As TIn decreased, Vth initially tended to increase, but subsequently declined progressively. Vth at the highest two flow rates (shortest TIn) was significantly lower than Vth at the third flow level (0.58 L and 0.59 L, respectively, versus 0.86 L; p < 0.01).

View larger version (13K): [in this window] [in a new window]   Figure 8.   Average relation between neural inspiratory duration and volume at point of termination of neural inspiration. Dotted line is the anticipated response based on the classical inspiratory terminating Herring-Breuer reflex. This line has no quantitative significance, but is placed here simply to show that according to this classical reflex, the volume threshold at inspiratory termination should be higher with shorter inspiratory times. Note that the response in the current experiments is qualitatively different from the expected response, in that the volume at switching tends to decrease as TIn decreases, indicating that the earlier switching is not related to earlier attainment of a volume threshold. Bars represent SEM (some bars were omitted for clarity). +Significantly higher than values at shortest TIn (p < 0.005).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main findings in the present study were that TIn is strongly influenced by ventilator flow rate, and that this response is not due to earlier attainment of the volume threshold for inspiratory termination (the classic H-B reflex). Rather, the response appears to be specifically related to flow.

Technical Considerations

In the majority of subjects (10 of 15), neural timing was inferred from a calculated mechanical output (Pmus). This approach may be criticized on the grounds that accurate Pmus determinations depend on accurate estimates of passive respiratory mechanics, which are difficult to obtain in awake subjects. Our faith in this approach is based on two considerations. First, although it has long been recognized that relaxation of respiratory muscles, on command, is difficult to achieve in awake, untrained subjects, it has recently been shown that loss of inspiratory muscle activity is readily attained in normal awake subjects when they are connected with a constant-volume ventilator and the mandatory ventilation rate is increased sufficiently (6). Furthermore, the criteria for relaxation under these conditions have been well established (6, 9). What was required of our subjects was simply not to voluntarily activate their expiratory muscles in response to the inspiratory hold maneuver. This is a much easier task than to voluntarily relax inspiratory muscles at different volumes. Furthermore, voluntary recruitment of expiratory muscles after the onset of end-expiratory occlusion would be easily recognized. Because the subjects did not anticipate the occlusions and because there is a minimum delay of 0.2 s for voluntary recruitment of respiratory muscles (11, 12), recruitment of expiratory muscles would appear as a secondary rise in Paw after an initial fall. Second, according to the equation of motion (Equation 1), errors in estimating elastance and resistance, in the face of a near constant flow, should result, respectively, in errors in estimating the rate of rise of Pmus or in a constant shift in the whole waveform without affecting the point in time at which Pmus changes direction from rising to falling. It was this inflection point in which we were interested. Although in theory the time of occurrence of the inflection point can be artifactually altered in the presence of nonlinearities in the passive pressure-flow and pressure-volume relations, we felt that these conditions would not apply in our normal, mouth-breathing subjects over the VT and flow ranges used (7, 10).

The validity of using calculated Pmus to estimate the end of TIn was determined independently from Edi and Pmus over a wide range of flows (Figure 3). Much of the difference between the two measurements appeared random. In retrospectively reviewing the reasons for the difference, we found that it was mostly due to uncertainty about when diaphragmatic electromyographic activity began declining. As can be seen from Figures 2 and 4, the Edi signal remained jagged despite elimination of the QRS complex and averaging at 100 ms. More severe filtering would have further smoothed the signal, but would also have altered the point of inflection. With the processing used in the study, there were often two peaks, and the decision about which constituted the end of TIn was subjective (e.g., see Figures 4B and 4C). We usually selected the last peak, but this could easily have been noise on the descending limb. Alternatively, a large spike occurring near the end of the rising phase might have led to false early identification of the end of TIn. After averaging to eliminate this random error, there were minor systematic differences (Figure 3F). In most of the cases in which differences existed, TIP_mus slightly but systematically exceeded TIEMG. This may have been related to pressure output essentially being a damped expression of muscle activation due to muscle mechanical delays. Furthermore, Pmus reflects the output of all muscles and not only the diaphragm, and it is possible that termination of activity in other muscles may have lagged somewhat. These systematic differences, however, were quite small (Figure 3F), and did not affect the magnitude of estimated change in TIn as a function of flow. In further support of the validity of Pmus in assessing TIn is that the changes observed in TIn with flow in five subjects in whom Edi was conducted did not differ from those observed in the other 10 subjects, in whom only Pmus was used (Figure 5, compare solid lines with broken lines). One exception, however, should be noted. In one subject, the data points were scattered around the line of identity at all flow rates except the highest flow (shortest TIn), where TIP_mus appeared to systematically underestimate TIEMG (Figure 3C). This was also apparent in the averaged data ("+" symbols, Figure 3F). An example of a breath in which TIP_mus underestimated TIE_di is shown in Figure 9A; peak Pmus preceded peak Edi by 0.2 s. We believe that this was due to a combination of underestimation of resistance, very high peak flow, and termination of neural inspiration soon after peak flow was reached. The underestimation of resistance would cause a stepwise reduction of estimated Pmus at triggering, which, in the face of a large flow, as in this case (~ 3.0 L/s) could cause a sharp, steep decline in estimated Pmus. Ordinarily, if inspiratory activity continued increasing well after peak flow was reached, the error would reach a maximum at peak flow, and Pmus would rise again to a new peak (e.g., Figure 9B). The artifactual nature of the first peak would be apparent. When inspiratory activity terminates soon after peak flow is reached, the early artifactual reduction in Pmus merges with the later reduction, owing to true termination of inspiration, and the error is not identified. This artifact was present in only one of the five subjects in whom Edi was measured, and only at the highest peak flow rate (3 L/s); however, it may have played a role in producing the very short TIP_mus points at the highest flow in some of the other 10 subjects. Nevertheless, it must be reiterated that occurrence of this artifact would require true TIn to terminate very early, soon after peak flow was reached. In other words, it can only make a short TIn appear somewhat shorter, but cannot cause TIn to appear short when it is not.

View larger version (61K): [in this window] [in a new window]   Figure 9.   Example of artifactual early termination of TIn Pmus in one subject. (A) Pmus begins declining before Edi. The decline coincides with ventilator triggering and is progressive beyond the point of peak flow. (B) Another breath soon after (A) in the same subject. TIn (by EMG) is somewhat longer. Pdi also begins declining at triggering. However, because TIn extends well beyond peak flow, Pmus begins rising again beyond peak flow, reaching a new peak. The artifactual nature of the first peak becomes apparent. See text for additional comments. The sharp positive spike in Pmus at the time of off-cycling of ventilator is related to inertial forces that are not taken into account in the Pmus calculation (see data analysis). This was particularly prominent in this subject (cf. subjects illustrated in Figures 2 and 4).

In summary, we believe that the changes in TIn, and hence in TEn, observed with Pmus adequately reflected the corresponding changes in neural timing in the entire group of our subjects.

Response of Neural Inspiratory Duration

The response of TIn to increasing flow rate was remarkably strong. Increasing flow rate resulted in an average reduction in TIn of 0.82 ± 0.30 s over the flow range used. When allowance was made for trigger delays (0.28 ± 0.05 s), TIn lasted an average of 0.34 ± 0.27 s beyond triggering at the highest flow rate. The response was graded and was fairly uniform among subjects (Figure 5, left).

The first issue to be discussed with respect to the mechanism of the TIn response is whether it is a deliberate behavioral response or is automatic (reflexive). Because the subjects were alert, and ventilator TI changed inversely with flow rate (constant VT), it may be argued that subjects deliberately reduced TIn as flow increased, in order to avoid continued effort after the end of the ventilator cycle. There are several arguments against this possibility. First, the response of TIn mirrors the response of Ttot and is in fact the main reason why Ttot decreases with increasing flow (Figure 6). The response of Ttot to ventilator flow has been shown to persist during sleep (2, 4), making it very likely that the response of TIn also persists during sleep, and is therefore not deliberate. Second, the considerable uniformity of response among subjects, and during several applications of the same flow at different times in the same subject, argues against volitional responses, which would be expected to show considerable interindividual and intraindividual (due to different levels of vigilance) variability. Third, changes in flow were applied in random order. At least for the first breath after a flow transition, there was no way for the subject to anticipate what ventilator TI would be. There was no significant difference between the response in the first breath and the response in subsequent breaths (Figure 7).

A host of reflexes are responsive to mechanical variables and are capable of altering respiratory timing. These originate from receptors in the upper airway (13), in the chest wall (14- 18), and in the lungs and bronchi (17). It is unlikely that the TIn response observed in our study was mediated primarily by upper-airway receptors. This assessment is based on our finding that the Ttot response was the result of shortening of TIn, coupled with recent observations that the Ttot response to increasing flow, at constant VT, is not altered by upper-airway anesthesia (3), and is intact in intubated patients in whom the upper airway is bypassed (5).

The TIn response is also unlikely to be mediated by chest wall receptors. This is because reflexes from the chest wall that have effects on TI are quite weak and tend to shorten TIn when the mechanical load is increased (21, 22), whereas what we observed was a reduction in TIn as the mechanical load was decreased through an increase in ventilator flow.

The reflex that would most obviously account for the current findings is the vagally mediated, inspiratory terminating (H-B) reflex (23, 24). In anesthetized animals, changes in VT, produced by changes in ventilator gain (23) or in mechanical load (24), result in reciprocal changes in TIn, as schematically represented by the dashed line in Figure 8. This reflexive activity is believed to be mediated by the slowly adapting vagal stretch receptors (19, 20), since it is these vagal receptors that are primarily sensitive to lung volume. Other, more rapidly adapting vagal receptors are not felt to be likely candidates for mediating such activity, since they respond primarily to flow with only modest volume sensitivity (25). One plausible explanation for the reduction in TIn with increasing flow rate in our experiments is, therefore, that a higher flow rate results in an earlier attainment of the volume threshold for inspiratory termination. To assess this possibility we determined the volume at which switching occurred (Vth), and plotted this against values of TIn observed at different flow rates (solid line, Figure 8). As can be seen in Figure 8, this possibility can be immediately discounted. Rather than being higher at low TIn values (as would be expected with the H-B reflex [Figure 8, dashed line]), the volume at switching was actually significantly lower at the highest flows rates relative to the volume values at lower flows, where TIn was longest. There are two possible explanations for this unexpected behavior as follows: (1). The response observed in the present study was not related to the volume-dependent H-B reflex, but represents a newly discovered, flow-specific response that is particularly prominent under the conditions of the present experiments. (2). A more provocative explanation is that the classical inspiratory inhibitory H-B reflex is mediated primarily, or to a considerable extent, by flow-sensitive receptors, with the apparent volume dependence being largely fortuitous. The following observations are noteworthy in this respect: First, in all experiments in which the "volume threshold" for inspiratory termination was determined, changes in VT were associated with, or produced by, concurrent changes in inspiratory flow (23, 24). Second, when the increase in volume is sustained over several breaths, the effect on TIn is lost (26, 27); the loss of this effect occurs very rapidly (usually by the second breath [personal observations]). This has been traditionally attributed to central adaptation (17, 26, 28). Although the development of some central adaptation has been unequivocally demonstrated (28), the rapid loss of effect of volume on TIn could, at least in part, be related to an important contribution from flow-sensitive receptors; inspirations occurring at an increased but constant volume are not subjected to the flow-related component of the reflex.

Our results do not permit a distinction between the two foregoing possibilities. Additional studies with animals are required to determine the specific contribution of flow to the inspiratory terminating response associated with tidal changes in volume and flow. One earlier study that addressed this issue, and which showed no independent flow effect (29), was limited by the small flow range tested.

Other Timing Responses

The response of TEn to changes in inspiratory flow was inconsistent and generally not significant (Figures 5 and 6B). It has been previously shown in anesthetized animals that isolated reduction in TIn, produced by vagal (28) or pontine (30) stimulation limited to inspiration, are associated with reductions in TEn. This indicates the presence of a central linkage between TIn and TEn. According to these findings in animals studies, we should have seen significant reductions in TEn as a result of the reduction in TIn. It is possible that our failure to observe such reductions was related to continued inflation during neural expiration, which, according to the H-B expiratory-prolonging reflex, would tend to lengthen TEn (17). The extent of continued inflation during neural expiration in our experiments, expressed by Tend, varied with inspiratory flow rate and showed an average biphasic relation (Figure 6E). There was a corresponding average biphasic TEn response (Figure 6B). To further investigate this possibility, we performed multiple linear regression analysis to determine the factors that correlated with changes in TEn (TEn). The independent variables selected for this analysis were Tend, TIn, and flow. The results for all subjects were pooled. Tend had a highly significant independent effect on TEn (p < 0.00001), with a coefficient of 1.2 indicating that an increase in T_end is associated with nearly the same increase (actually a slightly greater increase) in TEn. The change in TIn also had a significant independent effect (p = 0.01), but the coefficient was surprisingly negative (0.5). This suggests that rather than a reduction in TIn leading to a reduction TEn, the opposite occurs. It is possible that this negative correlation is related to errors in measurement of TIn; for a given Ttot, a given underestimation of TIn produces an overestimation of TEn, and vice versa. Additionally, flow had a significant effect (p = 0.004), independent of Tend and TIn. The coefficient was 0.34. The mechanism by which this effect may be exerted is not clear, and may be related to an increase in respiratory drive produced by increasing flow, as reported earlier (1).

That Tend exerts a strong influence on TEn is likely to explain the failure of the respiratory rate to increase when the flow rate is increased but VT is also increased to keep the ventilator TI constant (4, 31). Under these conditions, the reduction in TIn produced by the increase in flow results in an equal increase in Tend with a corresponding lengthening of TEn. This would offset the effect of inspiratory shortening on Ttot.

As a consequence of the continuous decrease in TIn with flow (Figure 6A) and the biphasic response of TEn (Figure 6B), TIn/Ttot was independent of flow up to a flow rate of 1.5 L/s (Figure 6D). At higher flow rates, TIn/Ttot decreased sharply.

The response of Ttot to an increasing flow rate (Figure 6C) was similar in magnitude to what was reported earlier over the comparable flow range (up to 1.5 L/s [1-3]). The present study shows that the response tends to saturate at a flow of approximately 1.5 L/s (Figure 6C) despite further decreases in TIn.

Clinical Implications

The flow range used in this study (0.8 to 2.5 L/s) covers the range of flow requirements encountered in critically ill patients receiving mechanical ventilation (32). Our results show that TIn and TIn/Ttot are quite sensitive to ventilator flow rate over this clinically relevant range. We are not aware of any studies in which the effect of ventilator flow on these variables was systematically assessed in mechanically ventilated patients. It is, however, almost certain that the responses observed here do occur in the intensive care unit (ICU) setting. Thus, Corne and associates recently reported that RR increases as a function of flow rate (at constant VT) in ventilator dependent patients (5). To the extent that the increase in RR is principally the result of a reduction in TIn (current findings), such patients are likely to display a dependence of TIn on flow. Furthermore, in several unrelated published studies, some of the illustrations clearly show instances in which inspiratory effort terminated virtually immediately after triggering (e.g., Figure 1, [33] and Figure 3D [34]). Because this immediate termination represents the extreme of the TIn response observed in the current study, it is almost certain that intermediate, less dramatic, responses also occur in the ICU setting.

Given that the responses demonstrated in our study are likely to occur in ventilator-dependent patients, the following clinical implications may be inferred from our findings, although these must await direct confirmation before being applied clinically.

1. Marini and colleagues (35) and Ward and coworkers (36) found that the work of breathing is inversely related to ventilator flow rate in the flow range below 1 L/s. The pressure-time product (time integral of Pmus per breath, PTP) was also found to decrease with increasing flow in the flow range from 0.4 to 1.1 L/s (36). The results presented by these authors (35, 36) do not permit an assessment of the mechanisms responsible for this response. Our results suggest that a flow-dependent reflex termination of TIn, with a consequent reduction in peak Pmus, may play an important role in this demonstrated unloading.
2. We have shown that neural TI/Ttot decreases significantly as flow rate increases (Figure 6D). TIn/Ttot is one of the two key variables determining inspiratory muscle energy consumption and endurance, the other variable being mean inspiratory muscle pressure during the inspiratory phase (37). Our findings therefore indicate that assessment of the impact of flow rate on inspiratory muscle energetics must include an evaluation of the effect of flow on TI/TTOT, since this effect can be substantial.
3. It is widely believed that flow rates in excess of 1 L/s do not result in further inspiratory muscle unloading. This is primarily based on the results of Marini and colleagues (35), who reported no reduction in the work of breathing (WOB) as flow was increased from 1.0 to 1.6 L/s in ventilator-dependent patients. These authors, however, reported WOB in joules per liter. This index is essentially a reflection of mean inspiratory muscle pressure over the duration of the ventilator's inflation phase (i.e., PTP/TI_vent). Because TI_vent decreases with increasing inspiratory flow, an unchanged WOB, expressed in joules/liter, signifies a commensurate reduction in PTP. Furthermore, the analysis done by Marini and colleagues did not consider the impact of flow on TIn/Ttot. It is possible, therefore, that the unloading effect of flow above 1 L/s was considerably underestimated in their study.
4. In some patients, WOB remains quite excessive, even at a ventilator flow of 100 L/min (see Table 1 in Ref. 35). We have shown that the effect of flow on TIn, and TIn/Ttot is continuous to at least a flow rate of 2.5 L/s (150 L/min). These findings suggest that substantial further unloading may be achieved in such patients at yet greater flow rates than are currently available on commercial ventilators. This prediction, if experimentally confirmed, may lessen the frequent need for sedation or paralysis in such patients.
5. In an effort to rapidly attain target pressure in the PSV mode, some commercial ventilators deliver a high initial flow rate (1.0 to 1.5 L/s) even at low pressure settings (38, 39). At low PSV levels, such as may be applied in preparation for weaning, TI_vent and VT are substantially dependent on the duration and amplitude of the patient's effort (40). Should the initial flow rate continue to be high at these low PSV levels, and should this result in brief inspiratory efforts (as per our current findings), the result would be brief, shallow inspirations at low PSV levels. This may be interpreted as "weaning failure." The impact of such brief, weak efforts, maintained over a long period (days), on muscle reconditioning in preparation for weaning, or on the ability of the respiratory control system to resume a normal breathing pattern during a trial of spontaneous breathing, would also have to be addressed.

In summary, we have demonstrated the existence of a powerful, graded, flow-related inspiratory terminating reflex in normal subjects. This reflex is sufficiently powerful as to be capable of reducing inspiratory effort to a mere triggering effort within the flow range encountered in critically ill patients. Accordingly, it may have implications for inspiratory muscle unloading and patient-ventilator synchrony during mechanical ventilation. The response is not related to earlier attainment of a volume-dependent threshold for inspiratory termination, but is specific to flow. Although there is indirect evidence that this reflex is operative in ventilator-dependent patients, further studies are clearly needed to establish its gain under different circumstances and its clinical significance.