Am J Respir Crit Care Med Vol 161. pp 819–826, 2000Internet address: www.atsjournals.org

Compensation for Increase in Respiratory Workload during Mechanical Ventilation

Pressure-Support versus Proportional-Assist Ventilation

SALVATORE GRASSO, FILOMENA PUNTILLO, LUCIANA MASCIA, GIOVANNI ANCONA, TOMMASO FIORE,FRANCESCO BRUNO, ARTHUR S. SLUTSKY, and V. MARCO RANIERI

Dipartimento di Emergenza e Trapianti d’Organo, sezione di Anestesiologia Rianimazione, Ospedale Policlinico, Università di Bari,and Ospedale DiVenere, Bari, Italy; and Department of Medicine, Samuel Lunenfeld Research Institute, Mount Sinai Hospital,University of Toronto, Toronto, Ontario, Canada

Variation in respiratory impedance may occur in mechanically ventilated patients. During pressure-targeted ventilatory support, this may lead to patient–ventilator asynchrony. We assessed the hy-pothesis that during pressure-support ventilation (PSV), preservation of minute ventilation (

E

) con-sequent to added mechanical loads would result in an increase in respiratory rate (RR) due to thelarge reduction in tidal volume (V

T

). W

ITH

proportional-assist ventilation (PAV), preservation of

E

would occur through the preservation of V

T

, with a smaller effect on RR. We anticipated that thiscompensatory strategy would result in greater patient comfort and a reduce work of breathing. Anincrease in respiratory impedance was obtained by chest and abdominal binding in 10 patients dur-ing weaning from mechanical ventilation.

E

remained constant in both ventilatory modes after chestand abdominal compression. During PSV, this maintenance of V

E

was obtained through a 58

6

3%increase in RR that compensated for a 29

6

2% reduction in V

T

. The magnitudes of the reduction inV

T

(10

6

3%) and of the increase in RR (14

6

2%) were smaller (p

,

0.001) during PAV. During bothPSV and PAV, chest and abdominal compression caused increases in both the pressure–time product(PTP) of the diaphragm per minute (142.9

6

26.9 cm H

2

O

?

s/min, PSV, and 117.6

6

16.4 cm H

2

O

?

s/min, PAV) and per liter (13.4

6

2.5 cm H

2

O

?

s/L, PSV, and 9.6

6

0.7 cm H

2

O

?

s/L, PAV). These in-crements were greater (p

,

0.001) during PSV than during PAV. The capability of keeping V

T

and

E

constant through increases in inspiratory effort after increases in mechanical loads is relatively pre-served only during PAV. The ventilatory response to an added respiratory load during PSV requiredgreater muscle effort than during PAV.

Grasso S, Puntillo F, Mascia L, Ancona G, Fiore T, BrunoF, Slutsky AS, Ranieri VM. Compensation for increase in respiratory workload during me-chanical ventilation: pressure-support versus proportional-assist ventilation.

AM J RESPIR CRIT CARE MED 2000;161:819–826.

.V

.V

.V

.V

Techniques for partial ventilatory support are intended for pa-tients who have normal respiratory drive but who have diffi-culty sustaining adequate spontaneous ventilation. The mostpopular mode of such assistance is pressure-support ventila-tion (PSV), in which the ventilator generates a constant pres-sure that acts in addition to the patient’s effort throughout theinspiratory period (1–4). With PSV, the pressure applied bythe ventilator (P

appl

) rises to a preset level that is held constantuntil a cycle-off criterion (a minimum inspiratory flow value)is reached. The inspiratory flow and tidal volume (V

T

) are thusrelated to the patient’s inspiratory effort, the level of P

appl

, and

the respiratory system impedance; breathing frequency is de-termined by the patient’s own respiratory drive (5).

Proportional-assist ventilation (PAV) is an alternativemode of partial ventilatory support in which the ventilatorgenerates pressure in proportion to the patient’s effort (5–7).Thus, during PAV, P

appl

is a function of patient effort: thegreater the inspiratory effort, the greater is the increase inP

appl

. Flow and V

T

will therefore be determined by the level ofproportionality between P

appl

and the patient’s effort, and theimpedance of the respiratory system. Ventilator assistance ter-minates with the end of the inspiratory effort, and RR is deter-mined by the patient’s own respiratory drive (5–7).

Spontaneous variations in impedance of the respiratorysystem commonly occur in mechanically ventilated patients(8–11), and may impair the matching between the ventilatoroutput and the patient’s ventilatory demand during partial ven-tilatory support (12), possibly leading to the development ofpatient–ventilator asynchrony (11). The different approachesused in PSV and PAV to pressurize the lung could theoreti-cally lead to marked differences in response to these varia-tions in respiratory system impedance (5).

(

Received in original form February 16, 1999 and in revised form September 3, 1999

)

Supported by grant 95. 00934. CT04 from the Consiglio Nazionale delle Ri-cherche, Italy.

Correspondence and requests for reprints should be addressed to V. MarcoRanieri, M.D., Università di Bari, Ospedale Policlinico, Anestesiologia e Rianimazi-one, Piazza Giulio Cesare 11, Bari 70100, Italy. E-mail: mranieri@teseo.it

820

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 161 2000

Respiratory loading has often been used to simulate changesin respiratory impedance (13, 14) and to evaluate the conse-quences of such changes on ventilatory patterns and respira-tory muscle performance (15). The aim of this study was to as-sess ventilatory responses to added mechanical loads duringPSV and PAV in patients during the weaning period. Our hy-pothesis was that during PSV, preservation of

E

consequentto respiratory loading would primarily be accomplished by anincrease in RR caused by a large reduction in

T

. By contrast,we hypothesized that preservation of

E

after respiratoryloading would occur through the relative preservation of V

T

,with a smaller effect on RR during ventilation with PAV. Weanticipated that this compensatory strategy would result ingreater patient comfort and less inspiratory muscle effort.

METHODS

Patient Selection

Patients were enrolled in the study after entering the weaning processas prescribed by the attending physician. Entry criteria also includedclinical and hemodynamic stability and a maximal inspiratory pressure(P

Imax

) of at least

2

20 cm H

2

O. Exclusion criteria included uncon-sciousness and/or presence of a chest tube. Ten patients admitted tothe intensive care units of the Policlinico and DiVenere Hospitals ofthe University of Bari were studied. They were intubated and me-chanically ventilated for the management of acute respiratory failure.All patients were receiving PSV (Siemens Servo Ventilator 300; Sie-mens Elema AB, Berlin, Germany) with pressures ranging between10 and 14 cm H

2

O (11.9

6

0.5 cm H

2

O [mean

6

SEM]). The local eth-ics committee approved the investigative protocol, and written in-formed consent was obtained from each patient. A physician not in-volved in the study was always present for patient care during theexperimental phase of the study.

Study Protocol

A key element of the study was to ensure that PSV and PAV providedan equivalent level of support. To ensure this, we provided an equaldegree of respiratory muscle unloading for both PSV and PAV, usingthe pressure–time product of the diaphragm per minute (PTP/min) asthe target variable, since it has been shown to correlate with the O

2

cost of breathing (16). A similar decrease in this variable with eachventilatory mode would therefore indicate equivalent levels of venti-latory support. To accomplish this, we disconnected patients from theventilator and allowed them to breathe spontaneously. Twenty to 30consecutive breaths were allowed over a period of 2 to 3 min. V

T

, RR,the ratio of inspiratory time to total breathing cycle time r (T

I

/Ttot),and PTP/min of the diaphragm during spontaneous breathing (SB)were then measured.

Respiratory muscle inactivity was than achieved by injecting ashort-acting hypnotic agent (propofol, 0.3 mg/kg/min for 5 min), andcontrolled mechanical ventilation was started (17). Ventilator param-eters were set in such a way as to match the breathing pattern re-corded during SB (18). Intrinsic positive end-expiratory pressure(PEEPi), static elastance (E

RSst

) and total resistance (R

RStot

) of therespiratory system were measured by applying end-expiratory andend-inspiratory airway occlusions as previously described (9).

Thirty minutes after these measurements, when respiratory mus-cle activity returned toward normal (judged to have occurred whennegative swings in esophageal pressure developed during inspiration)and the patients awakened, the Siemens ventilator was replaced by aWinnipeg ventilator (University of Manitoba, Winnipeg, MB, Canada).The design and operation of this unit are similar to those previouslydescribed (5, 6, 8, 12). The level of pressure (during PSV) and per-centage of unloading (during PAV) were set to obtain a 60 to 70% de-crease in PTP/min of the diaphragm relative to the values obtained dur-ing the spontaneous breathing (SB) trial. Values of E

RSst

and R

RStot

obtained during the trial of controlled mechanical ventilation wereused to set PAV. In patients with chronic obstructive pulmonary dis-ease, PEEP was set at 80% of the PEEP

i

measured during controlled

V·

V·

V·

mechanical ventilation (9). The set level of pressure (during PSV), thepercentage of assistance (during PAV), the fractional inspired oxygenconcentration (F

IO

2

), and the eventual PEEP level remained constantthroughout the different experimental conditions.

The increase in respiratory workload was obtained by strappingthe chest wall and the abdomen (19, 20). The chest wall was strappedwith a nearly inelastic cloth corset (25 cm long in the craniocaudalaxis) with adjustable straps, and a rectangular pneumatic cuff (20

3

30cm) was inserted under the corset and placed over the anterior chest.A similar corset (15 cm long in the craniocaudal axis) was wrappedaround the abdomen and a rectangular pneumatic cuff (20

3

30 cm)was placed between the corset and the ventral part of the abdomen(19, 20). The straps of the corsets were adjusted to that breathing wasnot hampered when the pneumatic cuffs were deflated. When all res-piratory variables were stable, the pneumatic cuffs were inflated to apressure of 20 mm Hg (19, 20).

Modes of ventilatory support were randomized through a con-cealed allocation approach, utilizing opaque, sealed envelopes con-taining the randomization schedule. Application of the respiratoryload was randomized once the ventilatory mode was selected. Mea-surements were obtained from 2 to 3 min of data recorded before(load off) and after 5 to 8 min of chest and abdominal compression(load on) for each mode of ventilation.

Measurements

Flow was measured with a heated pneumotachograph (Fleisch No. 2;Fleisch, Lausanne, Switzerland) connected to a differential pressuretransducer (Validyne MP 45

6

2 cm H

2

O; Validyne Co., Northridge,CA), which was inserted between the Y-piece of the ventilator circuitand the endotracheal tube. The pneumotachograph provided a linearrecording over the experimental range of flow. Equipment dead space(not including the endotracheal tube) was 70 ml. Airway openingpressure (Pao) was measured proximal to the endotracheal tube, witha pressure transducer (Validyne MP 45

6

100 cm H

2

O). Changes inintrathoracic and abdominal pressures were evaluated by assessingesophageal (Pes) and gastric (Pga) pressures. Pes and Pga were mea-sured with thin, latex balloon-tipped catheter systems connected bypolyethylene catheters to separate differential pressure transducers(Validyne MP 45

6

100 cm H

2

O). The esophageal balloon was cor-rectly positioned by means of an occlusion test (21). All of the vari-ables described here were displayed on an eight-channel strip-chartrecorder (Model 7718A; Hewlett-Packard Co., Cupertino, CA), andwere collected on a personal computer through a 12-bit analog-to-dig-ital converter at a sampling frequency of 100 Hz. Subsequent data anal-ysis was done with the ANADAT software package (RHT-InfoDat,Montreal, PQ, Canada). Patients were studied while in the semire-cumbent position.

Breathing pattern.

V

T

was computed by digital integration of theflow signal. P

Imax

was measured as previously described (9). Inspira-tory time (T

I

), expiratory time (T

E

), and total breathing cycle time(Ttot) were determined from the flow tracing.

Indexes of O

2

consumption of the diaphragm.

Tidal excursions ofPes (

D

Pes) and transdiaphragmatic pressure (

D

Pdi) were determined.Pdi was calculated as

D

Pga minus

D

Pes. All pressure swings are re-ported as changes from the end-expiratory value rather than from ab-solute zero pressure. This permits exclusion of the passive increase in-troduced by the application of corsets (and particularly the increase inPga), permitting comparison of the active pressures generated underthe different experimental conditions (19, 20). PTP per breath (PTP/b)was obtained by measuring the area under the Pdi signal from the be-ginning of the inspiratory deflection to the end of inspiratory flow(16). PTP/min was calculated as PTP/b multiplied by RR. PTP/L wascalculated as PTP/min divided by

E

.

Respiratory mechanics during PSV and PAV.

Respiratory mechan-ics during the different experimental conditions were assessed withthe Mead and Wittenberger technique (22). Briefly, inspiratory pul-monary resistance (R

L

) and elastance (E

L

) were calculated by fittingthe equation of motion of a single-compartment model using multilin-ear regression, as follows:

(1)

V·

∆PL RL flow EL VT×+×=

Grasso, Puntillo, Mascia,

et al.

: PSV versus PAV during Weaning 821

where

D

P

L

is inspiratory change in transpulmonary pressure (calcu-lated by subtracting Pes from Pao), V

T

is tidal volume, and flow ispeak inspiratory flow. The level of PEEP

i

during the different experi-mental conditions (PEEPi,

dyn

) was measured as the negative deflec-tion in Pes from the onset of inspiratory effort to the point of zeroflow. In the case of active recruitment of the abdominal muscles, thisvalue was corrected by subtracting the decrease in Pga, when present,from the decrease in Pes during the interval when PEEPi,

dyn

was mea-sured (10, 23).

Intensity of dyspnea.

The intensity of breathlessness was rated witha dyspnea visual analogue scale (VAS) at 10 to 15 min after the begin-ning of each experimental trial (24). Patients were asked to place avertical mark on a printed 100-mm horizontal scale in response to thequestion: “How short of breath are you right now?” The line had de-scriptors below its extreme ends. On the left was the word “none,” in-dicating no shortness of breath, and on the right was the oppositeresponse, “extremely severe.” For each experimental condition, pa-tients placed a vertical mark on the line at the point that best repre-sented the intensity of their dyspnea. Intensity was measured as thedistance in millimeters from the left end of the horizontal line (corre-sponding to no dyspnea) to the mark placed by the patient. A freshscale was presented on each occasion that these measurements ofbreathing comfort were made. Before the protocol began, directions forusing the scale were read aloud, and all patients then practiced mark-ing the scale.

Statistical Analysis

Results are expressed as mean

6

SEM. Values obtained during thedifferent experimental conditions were compared through repeated-measures two-way analysis of variance (ANOVA) and Bonferroni’stest. Regression analysis was done with the least-squares method, us-ing experimental data taken during the period from 2 to 3 min afterdata acquisition was begun (StatView, software package; Abacus Inc.,Berkeley, CA).

RESULTS

Causes of acute respiratory failure, gender, age, days of mechan-ical ventilation, static respiratory mechanical parameters, andblood gas values for the patients in the study are shown in Ta-ble 1. The PEEP level used in the study was 5.2

6

0.7 cm H

2

O.Application of ventilatory support decreased inspiratory

swings in Pdi (Figure 1), and PTP/b, PTP/min, and PTP/L(Figure 2) to a similar degree during PSV and PAV, relative totheir magnitude during spontaneous breathing. This was ob-tained through the application of 12

6

1 cm H

2

O pressure dur-

ing PSV; the percentages of elastic and resistive unloadingduring PAV were set to normalize patient resistive and elasticforces (9), and were 44

6

3% and 58

6

4%, respectively,which resulted in an applied pressure of 13

6

2 cm H

2

O. RRand levels of airway opening pressure (Pao) and V

T

were simi-lar in both ventilatory modes prior to chest and abdominalbinding (Table 2).

During SB, the dyspnea score was 46.0

6

1.2 mm. Whenasked to indicate changes in the degree of their breathlessnessduring PSV and PAV with respect to the preceding SB condi-tion, all patients reported similar reductions in dyspnea (38.9

6

4.3, 23.5

6

3.8, and 20.8

6

5.2 mm during SB, PSV, and PAV,respectively; p

,

0.001) (Figure 3).Chest and abdominal compression caused a similar in-

crease in

D

Pdi and PTP/b with both modes of ventilatory sup-port (Figure 2). During PAV, the increase in inspiratory mus-cle effort was followed by a concomitant increase in Pao,whereas during PSV, Pao remained constant after applicationof the respiratory load (Figure 1). Variables summarizing thepatients’ breathing patterns during the different experimentalconditions are given in Table 2. During PSV and PAV, chestand abdominal compression caused similar increases in E

L

(28

6

3% and 30

6

2%; p

5

NS) and R

L

(49

6

3% and 52

6

4%; p

5

NS), respectively. A 2- to 3-cm H

2

O increase in PEEPi wasalso observed in both modes of ventilatory support.

E

re-mained constant in both ventilatory modes after chest and ab-dominal compression. During PSV, this was achieved througha 58

6

3% increase in RR that compensated for a 29 6 2% re-duction in VT. The magnitudes of the reduction in VT (10 63%) and of the increase in RR (14 6 2%) were significantly(p , 0.001) smaller during PAV.

During both PSV and PAV, chest and abdominal compres-sion caused increases in PTP/min (142.9 6 26.9 cm H2O ? s/min, PSV, and 117.6 6 16.4 cm H2O ? s/min, PAV) and PTP/L(13.4 6 2.5 cm H2O ? s/L, PSV, and 9.6 6 0.7 cm H2O ? s/L,PAV) (Figure 2), as well as in the dyspnea score (53.4 6 4.0mm, PSV, and 33.8 6 5.9 mm, PAV) (Figure 3). During loadapplication, values of PTP/min, PTP/L, and dyspnea scorewere significantly (p , 0.001) greater during PSV than duringPAV (Figures 2 and 3).

Inspiratory muscle effort (estimated from PTP/b) was plot-ted against VT for 2 to 3 min of data acquisition during SB andduring load off and load on conditions with PSV and PAV(Figure 4). During SB, a significant correlation (p , 0.001) was

V·

TABLE 1

PATIENT CHARACTERISTICS

PatientNo. Gender

Age(yr) Cause of ARF

ERSst†

(cm H2O/L)RRSst

†

(cm H2O/L)PEEPi,stat

†

(cm H2O) FIO2*

PaO2*(mm Hg)

PaCO2*(mm Hg) pH*

PImax*(cm H2O) d

1 F 43 Pneumonia 19.4 11.4 1.6 0.3 97 43 7.42 230 42 F 32 Pneumonia 22.5 12.7 0.8 0.4 98 44 7.41 237 83 M 41 COPD 15.8 16.4 3.5 0.3 118 39 7.40 241 64 M 64 COPD 17.9 18.2 4.0 0.4 115 48 7.38 232 85 F 21 ARDS 23.5 12.0 1.3 0.3 109 42 7.40 241 146 M 58 COPD 16.5 17.2 3.9 0.3 110 46 7.37 238 97 M 67 Drug overdose 16.5 12.7 2.1 0.4 134 39 7.46 237 108 M 45 Pneumonia 23.2 9.2 4.4 0.4 105 38 7.41 240 89 F 52 ARDS 20.1 12.1 0.5 0.3 107 43 7.42 242 10

10 F 58 Pneumonia 18.6 15.3 0.2 0.4 106 38 7.43 244 5

Definition of abbreviations: ARDS 5 acute respiratory distress syndrome; ARF 5 acute respiratory failure; COPD 5 chronic obstructive pul-monary disease; FIO2

5 fractional inspired oxygen concentration; ERSst 5 static elastance of the respiratory system; PEEPi,stat 5 static intrinsicpositive end-expiratory pressure; PImax 5 maximum inspiratory pressure; PaCO2 5 arterial carbon dioxide tension; PaO2 5 arterial oxygentension; RRStot 5 total resistance of the respiratory system.

* Data obtained during pressure support ventilation.† Values obtained during controlled mechanical ventilation prior to randomization. Days on mechanical ventilation prior to study entry

are indicated.

822 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 161 2000

found between VT and PTP/b in all patients (slope 5 0.07 60.02). Relative to SB, application of PSV decreased and appli-cation of PAV increased the slope of this relationship (0.3 60.1 and 0.02 6 0.05, respectively; p , 0.001). During PSV,chest and abdominal compression displaced the correlationline downward without affecting its slope. During PAV, chestwall and abdominal compression decreased the slope of therelationship between PTP/b and VT in all patients to 0.1 60.05 (p , 0.0001).

DISCUSSION

This study demonstrates that in mechanically ventilated pa-tients who are being weaned from ventilation, the strategyused to compensate for an acute increase in respiratory im-pedance differs markedly with PSV and PAV. During PSV,despite a substantial increase in inspiratory effort, VT de-

creased after chest and abdominal binding, whereas E waskept constant by an increase in RR. During PAV, E wasmaintained by increasing Pao such that VT and RR remainedsubstantially unchanged. Larger values of PTP/min and PTP/Lwere required during PSV than during PAV to keep E con-stant in response to an increase in respiratory impedance. Fur-thermore, patients’ sensation of dyspnea during chest and ab-dominal binding was less intense during PAV than during PSV.

Before discussing the results of the present study, we willaddress a number of general considerations concerning its ex-perimental design, our assessment of inspiratory effort, andthe validity of measuring dyspnea after administration of ahypnotic agent.

First, both PSV and PAV are capable of unloading the re-spiratory muscles (5); indeed with either mode, the work per-formed by the ventilator can be varied from zero to nearly thetotal work of breathing required of the patient. A critical fea-

V·

V·

V·

Figure 1. Experimental recordsillustrating effects of chest andabdominal binding is a rep-resentative patient. From topto bottom: flow, Pao, volumechange (DV), and tidal ex-cursion of transdiaphragmaticpressure (DPdi). SB 5 spon-taneous breathing; PSV 5pressure-support ventilation;PAV 5 proportional-assist ven-tilation.

Figure 2. Effects of chest andabdominal binding on in-dexes of O2 of the diaphragm.DPdi 5 inspiratory tidal ex-cursion of transdiaphragmaticpressure; PTP/b 5 pressuretime product per breath;PTP/min 5 pressure timeproduct per breath; PTP/L 5pressure time product per li-ter; SB 5 spontaneous breath-ing; PSV 5 pressure-supportventilation; PAV 5 propor-tional-assist ventilation. Openbar: load off, closed bar: loadon.

Grasso, Puntillo, Mascia, et al.: PSV versus PAV during Weaning 823

ture of our experimental design was to provide equal supportwith both PSV and PAV before increasing respiratory systemimpedance. We chose to do this by setting the PSV and PAVsupport levels to provide a similar reduction in inspiratory ef-fort, relative to a trial of SB (Figures 1 and 2) before the im-pedance load was applied. We did this because McGregor andBecklake (25) have shown that the O2 cost of breathing isclosely related to the mean pressure developed by the dia-phragm, and assessed in the present study as the mean PTP/min (26, 27). Independent evidence suggesting that this ap-proach was successful was the observation that the dyspneascore decreased by an equivalent amount with PSV and PAV(Figure 3). In addition, the respiratory workload (i.e., PEEPi,RL, and EL for a given patient) was similar with the two modesof ventilatory support (Table 2). It is interesting to note thatunder these experimental conditions, PSV and PAV provideda similar breathing pattern (Table 2).

Second, the measurements in our study were obtained from2 to 3 min of data recorded before and after 5 to 8 min of chestand abdominal compression. Given this, the results of our studymay not reflect a steady-state condition.

Third, the time elapsed between the end of the the propo-fol infusion and the first assessment of dyspnea score was 68 63 min, and was therefore longer than the awakening time (14 613 min) reported for critically ill patients with a propofol infu-sion rate similar to the one used in our study (28).

Studies of the normal response to increases in chest andabdominal binding suggest that humans maintain nearly nor-mal levels of alveolar ventilation unless such binding createsmarked changes in impedance. The immediate response is areduced VT with minimal change in RR (13, 19, 20). Whenstrapping is maintained, VT progressively increases and E

returns to control values, with little change in RR and arte-rial blood gas values (13, 19, 20). Inspiratory muscle effort isincreased immediately after the chest wall and abdomenare strapped, and progressively increases during subsequentbreaths as VT approaches control values (13, 19, 20). This in-crease in inspiratory effort returns VT and E to control val-ues, with an increase in amplitude of respiratory muscle con-traction and changes in shape of the phases of inspiratorymuscle activity, without substantial changes in RR (13, 14).The steady-state response is due to the physiologic couplingbetween VT and inspiratory muscle effort (29). The magnitudeof effort and VT are linearly correlated, and in the presence ofan increase in respiratory impedance, the slope of the relation-ship between them will decrease; however, the greater outputwill restore VT to baseline levels (29–31). This physiologicability to integrate respiratory drive, inspiratory muscle effort,and VT on the basis of different ventilatory requirements hasbeen designated “neuroventilatory coupling” (29, 32).

Partial ventilatory support is principally indicated in pa-tients whose respiratory drive is normal or high but who havedifficulty in sustaining an adequate level of ventilation on theirown, and in whom an abnormal relationship between effortand ventilation is present (33). The great majority of these pa-tients have abnormal neuroventilatory coupling due to neuro-muscular weakness (which necessitates a greater effort to pro-duce a given pressure) and/or abnormal respiratory mechanics(which requires a greater pressure to generate a given level ofventilation). High ventilatory demand caused by sepsis, fever,or metabolic acidosis, for example, may further compound thissituation (34). During partial ventilatory support, the patient–ventilator interface can be described by using the equation ofmotion (22). At any instant during a breath, the total pressureapplied to the patient’s respiratory system includes the pres-

V·

V·

TABLE 2

EFFECTS OF INSPIRATORY LOAD ON BREATHING PATTERNUNDER DIFFERENT EXPERIMENTAL CONDITIONS

Parameter SB

PSV PAV

Load Off Load On Load Off Load On

VT, L 0.31 6 0.02 0.71 6 0.02 0.51 6 0.07† 0.68 6 0.13 0.61 6 0.14*TI, s 0.74 6 0.07 1.26 6 0.01 0.68 6 0.06† 1.05 6 0.10 1.02 6 0.11TE, s 1.07 6 0.10 3.18 6 0.09 2.11 6 0.05† 3.07 6 0.09 2.86 6 0.11Fr, s21 24.9 6 2.86 13.63 6 0.95 21.50 6 0.48† 16.41 6 1.96 18.57 6 1.44*

E, L/min 7.7 6 0.9 9.83 6 0.51 10.57 6 0.44 10.34 6 1.31 11.72 6 1.76Pmax, cm H2O 0.4 6 0.12 12.01 6 0.85 12.19 6 0.77 12.56 6 2.40 16.13 6 2.11†

EL, cm H2O/L 17.7 6 3.7 16.9 6 2.7 21.3 6 7.7† 17.3 6 1.3 22.8 6 3.4†

RL, cm H2O/L ? s 10.2 6 0.3 10.5 6 0.8 15.6 6 0.4† 9.8 6 1.4 15.1 6 0.6†

PEEPi,dyn, cm H2O 1.0 6 0.7 1.3 6 1.2 3.4 6 1.8† 1.2 6 0.7 3.8 6 1.1†

Definition of abbreviations: EL 5 lung elastance; Fr 5 respiratory frequency; PAV 5 proportional-assist ventilation; PEEPi,dyn 5 dynamicpositive end-expiratory pressure; Pmax 5 peak airway pressure; PSV 5 pressure-support ventilation; RL 5 lung resistance; SB 5 spontane-ous breathing; TE 5 expiratory time; TI 5 inspiratory time; E 5 minute ventilation; VT 5 tidal volume.

Data are mean 6 SEM.* p , 0.01, load off versus load on, ANOVA for repeated measures.† p , 0.001, load off versus load on, ANOVA for repeated measures.

V·

V·

Figure 3. Visual analog scale (VAS) of patient-perceived intensityof breathlessness under different experimental conditions. SB 5spontaneous breathing; PSV 5 pressure-support ventilation; PAV 5proportional-assist ventilation.

824 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 161 2000

sure generated by the respiratory muscles (Pmus) and thepressure applied by the ventilator (Pappl). This pressure is dis-sipated against: (1) PEEPi; (2) the patient’s resistance (Rtot);and (3) the patient’s static elastance (ELst). Under these cir-cumstances, the act of breathing in a mechanically ventilatedpatient can be described at any instant:

(2)

where Pres represents the resistive pressure and is a functionof inspiratory flow (Pres 5 inspiratory flow 3 Rtot) and Pelrepresents the elastic recoil pressure and is a function of VT

(PEL 5 VT 3 Est). Assuming that Rtot and Est are linear,Equation 2 becomes:

(3)

By providing a constant Pappl, PSV unloads the respiratorymuscles and improves the relation between patient effort andVT, in the sense that for a given inspiratory effort, the patientreceives a greater volume than would have been received dur-ing SB (33). However, since the ventilator provides the samePappl with every triggered breath, any increase in respiratorymuscle activity will not produce an increase in ventilator ap-plied pressure (5).

Pmus Pappl+ PEEPi Pres Pel+ +=

Pmus Pappl+ PEEPi inspiratory flow( Rtot )VT( Est )×

+×+=

With PAV the ventilator provides a pressure assist that isproportional to the instantaneous inspiratory effort. The pa-tient’s ability to alter VT through changes in effort is hencepreserved, and can be modulated by changing the level of pro-portionality (5–7, 9). With PAV, Pappl is a function of the flowand volume generated by patient effort. During PAV, Pappl 5kR 3 flow 1 kE 3 volume (5–7, 9), where the coefficients kR

and KE represent the proportionality between Pappl and flowand volume, respectively, generated by patient effort (5–7); inthe case of linearity of RL and EL, and according to Equation2, the use of PAV can be described by the following equation:

(4)

With PAV, what is preset is not the Pappl level (as in PSV), butthe proportion between inspiratory muscle effort and ventila-tor applied pressure (regulated by setting the KR and KE val-ues on the ventilator) (i.e., the extent to which Pappl will in-crease for a given increase in inspiratory muscle effort) (5, 6).Any increase in respiratory muscle activity will therefore befollowed by a concomitant increase in ventilator applied pres-sure (5).

To evaluate the physiologic implications of these theo-retical differences, we evaluated effects of PSV and PAV onneuroventilatory coupling at different levels of ventilatoryworkload (Figure 4) (26). The theoretical relation betweeninspiratory muscle effort (Pmus in Equation 1, quantified asPTP/b) and VT in a normal subject (slope 5 0.5 in Figure 4)was estimated with data from a previously described model (5,29). In our study, all patients had significant impairment oftheir neuroventilatory coupling during a trial of SB, as indi-cated by the values of the slope of the relationship betweenPTP/b and VT, which were below the normal predicted value(slope 5 0.07 6 0.02). By study design, application of PSV andPAV prior to chest and abdominal binding decreased PTP/band increased VT by a similar amount (2.2 6 0.8 cm H2O ? s and2.1 6 0.3 cm H2O ? s during PSV and PAV, respectively; and0.71 6 0.02 L and 0.68 6 0.13 L during PSV and PAV, respec-tively). However, during PSV, the improvement in breathingefficacy occurred concomitantly with the further reduction inthe slope of the relationship between PTP/b and VT (0.02 60.05). On the other hand, application of PAV decreased effortand increased ventilation with an increase in the slope of thisrelationship that virtually returned it to the normal theoreticalvalue (0.3 6 0.1).

Chest and abdominal binding increased PTP/b by a similaramount in both ventilatory modes (6.6 6 0.8 cm H2O ? s and6.4 6 0.6 cm H2O ? s during PSV and PAV, respectively; Figure2). However, during PSV, because of the parallel downwardshift of the PTP/b-versus-VT relationship, VT fell, since the in-crease in PTP/b did not offset the increase in impedance. E

remained constant only because of the increase in RR. On theother hand, during PAV, although chest wall and abdominalrestriction decreased the slope of the PTP/b-versus-VT rela-tionship (0.1 6 0.05), the increase in PTP/b was sufficient toincrease ventilator applied pressure to maintain T close tobaseline values. Under these circumstances, E remained con-stant without substantial changes in RR. These results there-fore explain why, despite a similar degree of inspiratory mus-cle effort, larger values of PTP/min and PTP/L during PSV(142.9 6 26.9 cm H2O ? s/min and 13.4 6 2.5 cm H2O ? s/L, re-spectively) than during PAV (117.6 6 16.4 cm H2O ? s/minand 9.6 6 0.7 cm H2O ? s/L, respectively) were required tokeep E constant in response to an increase in respiratory im-pedance. The decreased sense of dyspnea during chest walland abdominal binding observed during PAV is therefore

Pmus kR flow kE volume×+×( )+ PEEPiinspiratory flow Rtot×( ) Vt ELst×( )

++

=

V·

V·

V·

V·

Figure 4. Effects of increasing inspiratory effort on volume duringPSV (top panel) and PAV (bottom panel). PTP/b: pressure time prod-ucts of the diaphragm per breath; DV: volume change. Open squares:spontaneous breathing; open circles: load off; closed circles: loadon. Dotted line: theoretical relation between inspiratory muscle ef-fort and volume in a normal subject. See text for further details.

Grasso, Puntillo, Mascia, et al.: PSV versus PAV during Weaning 825

consistent with the lower O2 cost of breathing required tomaintain a constant E with this ventilatory mode (27). How-ever, before loading, RR tended to be lower during PSV thanduring PAV, although the difference was not significant; thiscould therefore have introduced a bias toward a larger in-crease in RR with load during PSV than during PAV.

A limitation of PSV may be the instability in VT of patientswith marked changes in respiratory impedance (5, 33, 34).This may be clinically important, since recent studies haveshown that spontaneous variations in total impedance of therespiratory system may occur in mechanically ventilated pa-tients (8–10). This has led to the suggestion that in addition tointact respiratory drive, stable impedance is also an importantrequirement for the use of PSV as a sole means of ventilatorysupport (5, 33, 35). Information about the magnitude and thefrequency of such changes in respiratory impedance in me-chanically ventilated patients is scanty. Despite this lack ofdata, we would suggest that the increases in impedance in-duced by the banding in our study are of a magnitude suffi-cient to be clinically relevant. The obtained increases in elas-tance and resistance (z 30% and 50%, respectively) were suchthat during PSV, they produced marked changes in bothbreathing pattern (58 6 3% increase in RR and 29 6 2 de-crease in VT) and inspiratory muscle effort (224 6 18% in-crease in PTP/b).

In a previous study, we showed that when ventilatory re-quirements were increased by acute hypercapnia, E in-creased mainly through changes in VT only during PAV; dur-ing PSV the increase in E was obtained through an increasein RR (26). This resulted in greater muscle effort and morepronounced dyspnea during PSV than during PAV (26). Re-sults of the present study confirm that in the presence of varia-tions in ventilatory requirements as a result of increases in me-chanical load, the use of PSV may also be limited by thereduction in VT. However, in contrast to our previous obser-vations (26), the present study suggests that impairment of thepatient–ventilator interaction after an increase in mechanicalload may also be observed during application of PAV. Al-though PAV was able to maintain a breathing pattern similarto that observed before loading conditions were applied, thisoccurred through a significant increase in patient workload.Although this preserved VT at a lower PTP/min and PTP/Lthan with PSV, the prolonged increase in inspiratory effortcould lead to muscle fatigue. Since Pappl is proportional to Pmusduring PAV, this may lead to an inadequate E, resulting inhypercapnia and acute respiratory acidosis (12). The full po-tential benefits of PAV can therefore be obtained only withcontinuous adaptation of the degree of ventilatory assistanceprovided by this modality to the changes in respiratory me-chanics that should be continuously monitored in the defini-tive technologic implementation of PAV (9).

In conclusion, our data show that in mechanically venti-lated patients in whom respiratory impedance is acutely in-creased by chest wall and abdominal binding, the physiologiccapability of keeping VT and E constant through increases ininspiratory effort was preserved only during PAV. DuringPSV, despite a similar increase in inspiratory effort to that inPAV, VT decreased, and the increase in RR preserved E. Theventilator response to an added respiratory load during PSVrequired greater muscle effort and caused more pronouncedpatient discomfort than during PAV. These data confirm thatwith PSV, although the patient receives a mandatory degreeof support, the patient’s ability to modulate the ventilatorypattern through changes in motor output remains impaired.Although with PAV the ability of the patient to unload thework of breathing in proportion to inspiratory effort is en-

V·

V·

V·

V·

V·

V·

hanced, changes in respiratory impedance are followed by in-creases in inspiratory effort. This may eventually lead to theimpairment of patient–ventilator interactions. This limitationof PAV may be related to the specific prototype used in ourstudy, and may be overcome by implementing positive feed-back to continuously adapt the level of assistance to changesin patient respiratory mechanics.

Acknowledgment : The authors thank the physicians and nursing staff of thePoliclinico and DiVenere hospitals for their valuable cooperation, and AnneMcClair-Turnbull for secretarial assistance. The Winnipeg ventilator was pro-vided by Mallinckrodt, Inc., and the Nellcor Puritan Bennett Corporation.

References1. Brochard, L., A. Harf, H. Lorino, and F. Lemaire. 1989. Inspiratory

pressure support prevents diaphragmatic fatigue during weaning frommechanical ventilation. Am. Rev. Respir. Dis. 139:513–521.

2. Brochard, L. 1994. Pressure support ventilation. In M. J. Tobin, editor.Principles and Practice of Mechanical Ventilation. McGraw-Hill, NewYork. 239–257.

3. MacIntyre, N. R. 1986. Respiratory function during pressure support ven-tilation. Chest 89:677–683.

4. MacIntyre, N. R., and N. E. Leatherman. 1990. Ventilatory muscle loadsand the frequency-tidal volume pattern during inspiratory pressure-assisted (pressure-supported) ventilation. Am. Rev. Respir. Dis. 141:327–331.

5. Younes, M. 1991. Proportional assist ventilation and pressure supportventilation: similarities and differences. In J. J. Marini and C. Roussos,editors. Ventilatory Failure. Springer-Verlag Berlin. 361–380.

6. Younes, M. 1992. Proportional assist ventilation, a new approach to ven-tilatory support. Am. Rev. Respir. Dis. 145:114–120.

7. Younes, M., A. Puddy, D. Roberts, R. B. Light, A. Quesada, K. Taylor,L. Oppenheimer, and H. Cramp. 1992. Proportional assist ventilation:results of an initial clinical trial. Am. Rev. Respir. Dis. 145:121–129.

8. Nava, S., C. Bruschi, C. Fracchia, A. Braschi, and F. Rubini. 1997. Pa-tient-ventilator interaction and inspiratory effort during pressure sup-port ventilation in patients with different pathologies. Eur. Respir. J.10:177–183.

9. Ranieri, V. M., S. Grasso, L. Mascia, S. Martino, T. Fiore, A. Brienza,and R. Giuliani. 1997. Effects of proportional assist ventilation on in-spiratory muscle effort in patients with chronic obstructive pulmonarydisease and acute respiratory failure. Anesthesiology 86:79–91.

10. Lessard, M. R., F. Lofaso, and L. Brochard. 1995. Expiratory muscle ac-tivity increases intrinsic positive end-expiratory pressure indepen-dently of dynamic hyperinflation in mechanically ventilated patients.Am. J. Respir. Crit. Care Med. 151:562–569.

11. Jubran, A., and M. J. Tobin. 1997. Pathophysiologic basis of acute respi-ratory distress in patients who fail a trial of weaning from mechanicalventilation. Am. J. Respir. Crit. Care Med. 155:906–915.

12. Bigatello, L. M., M. Nishimura, H. Imanaka, D. Hess, W. R. Kimbal, andR. M. Kacmareck. 1997. Unloading of the work of breathing by pro-portional assist ventilation in a lung model. Crit. Care Med. 25:267–272.

13. Milic-Emili, J., and W. A. Zin. 1986. Breathing responces to imposedmechanical loads. In N. S. Cherniack and J. G. Widdicombe, editors.Handbook of Physiology. The Respiratory System: Control of Breath-ing, Vol. 2. American Physiological Society, Bethesda, MD. 751–769.

14. Rebuck, A. S., and A. S. Slutsky. 1986. Control of breathing in diseasesof the respiratory tract and lungs. In N. S. Cherniack and J. G. Widdi-combe, editors. Handbook of Physiology. The Respiratory System:Control of Breathing, Vol. 2. American Physiological Society, Be-thesda, MD. 771–791.

15. Brack, T., A. Jubran, and M. J. Tobin. 1998. Effect of resistive loadingon variational activity of breathing. Am. J. Respir. Crit. Care Med. 157:1756–1763.

16. Sassoon, C. S. H., R. W. Light, R. Lodia, G. C. Sieck, and C. K. Mahutte.1991. Pressure-time product during continuous positive airway pres-sure, pressure support ventilation, and T-piece during weaning frommechanical ventilation. Am. Rev. Respir. Dis. 143:469–475.

17. Conti, G., V. Vilardi, M. Rocco, R. A. Deblasi, A. Lappa, M. Bufi, M.Antonelli, and A. Gasparetto. 1995. Paralysis has no effect on chestwall and respiratory system mechanics of mechanically ventilated, se-dated patients. Intensive Care Med. 10:808–812.

18. Zakynthinos, S. G., T. Vassilakopoulos, and C. Roussos. 1995. The loadof inspiratory muscles in patients needing mechanical ventilation. Am.J. Respir. Crit. Care Med. 152:1248–1255.

826 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 161 2000

19. Hussain S. N. A., B. Rabinovitch, P. T. Macklem, and R. L. Pardy. 1985.Effects of separate rib cage and abdominal restriction on exercise per-formance in normal humans. J. Appl. Physiol. 58:2020–2026.

20. Hussain, S. N. A., and R. L. Pardy. 1985. Inspiratory muscle functionwith restrictive chest wall loading during exercise in normal humans.J. Appl. Physiol. 58:2027–2032.

21. Baydur, A., P. K. Behrakis, W. A. Zin, M. I. Jaeger, and J. Milic-Emili.1982. A simple method for assessing the validity of the esophagealballoon technique. Am. Rev. Respir. Dis. 126:788–791.

22. Mead J., and J. L. Whittenberger. 1953. Physical properties of humanlungs measured during spontaneous respiration. J. Appl Physiol. 5:770–796.

23. Appendini, L., A. Patessio, S. Zanaboni, M. Carone, B. Gukov, C. F.Donner, and A. Rossi. 1994. Physiologic effects of positive end-expi-ratory pressure and mask pressure support during exacerbation ofchronic obstructive pulmonary disease. Am. J. Respir. Crit. Care Med.149:1069–1076.

24. Knebel, A. R., S. L. Janson-Bjerklie, J. D. Malley, A. G. Wilson, and J. J.Marini. 1994. Comparison of breathing comfort during weaning withtwo ventilatory modes. Am. J. Respir. Crit. Care Med. 149:14–18.

25. McGregor, M., and M. Becklake. 1961. The relationship of oxygen costof breathing to respiratory mechanical work and respiratory force. J.Clin. Invest. 40:971–980.

26. Ranieri, V. M., R. Giuliani, L. Mascia, S. Grasso, V. Petruzzelli, N. Pun-tillo, G. Perchiazzi, T. Fiore, and A. Brienza. 1996. Patient-ventilatorinteraction during acute hypercapnia: pressure support vs propor-tional assist ventilation. J. Appl. Physiol. 81:426–437.

27. Killian, K. J., and N. L. Jones. 1988. Respiratory muscles and dyspnea.Clin. Chest Med. 9:237–248.

28. Beauvoir, C., P. Chardon, F. d’Athis, J. C. Mathieu-Daude, and J. duCailar. 1992. Sedation with propofol and fentanyl in patients under in-tensive care. Ann. Fr. Anesth. Reanim. 11:27–34.

29. Younes, M., and W. Riddle. 1984. Relation between respiratory neuraloutput and tidal volume. J. Appl. Physiol. 56:1110–1119.

30. Otis, A. B. 1954. The work of breathing. Physiol. Rev. 34:449–458.31. Mead J. 1960. Control of respiratory frequency. J. Appl. Physiol. 15:

325–366.32. Younes, M. 1990. Load responses, dyspnea and respiratory failure. Chest

97:59S–68S.33. Hubmayer, R. D., M. D. Abel, and K. Rehder. 1990. Physiologic ap-

proach to mechanical ventilation. Crit. Care Med. 18:103–113.34. Hubmayr, R. D., and K. Rehder. 1992. Respiratory muscle failure in crit-

ically ill patients. Semin. Respir. Med. 13:14–21.35. Kacmarek, M. R. 1988. The role of pressure support ventilation in reduc-

ing work of breathing. Respir. Care 33:99–120.