Comparison of mathematical and mechanical models

of pressure-controlled ventilation

WILLIAM C. BURKE, PHILIP S. CROOKE III, THEODORE W. MARCY, ALEX B. ADAMS,

AND JOHN J. MARINI
Pulmonary/Critical Care Medicine, University of Minnesota, Minneapolis/St. Paul, Minnesota 55101-
2595; and Department of Mathematics, Vanderbilt University, Nashville, Tennessee37240

BURKE,WILLIAM C., PHILIP S. CROOKEIII, THEODOREW. way pressure limits alveolar distension and reduces the
MARCY,ALEX B. ADAMS,ANDJOHN J. MARINI. Comparisonof risk of barotrauma and hemodynamic compromise (1). In
mathematical and mechanical models of pressure-controlled ven- addition, the decelerating flow pattern, inherent to PCV,
tilation. J. Appl. Physiol. 74(Z): 922-933, 1993.-Recent evi- may facilitate gas distribution among lung units with dif-
dence that volume-cycled mechanical ventilation may itself ferent inspiratory time constants (2, 5). When used in
produce lung injury has focused clinical attention on the pres-
situations where the normal inspiratory-to-expiratory
sure waveform applied to the respiratory system. There has
been an increasing use of pressure-controlled ventilation ratio is inverted, prolonged application of high distend-
(PCV), because it limits peak cycling pressure and provides a ing pressures may recruit alveoli previously excluded
decelerating flow profile that may improve gas exchange. In from gas exchange (2).
this mode, however, the relationships are of machine adjust- An important goal of the clinician using PCV is opti-
ments to ventilation and alveolar pressure are not straightfor- mization of key “output” variables related to VE and in-
ward. Consequently, setting selection remains largely an empir- trathoracic pressure [peak (PAJ, mean (PA), and end-
ical process. In previous work, we developed a biexponential expiratory (Pex) alveolar pressures]. In a single-com-
model of PCV that provides a conceptual framework for under- partment respiratory system, PACT, achieved at end
standing these interactions (J. Appl. Physiol. 67: 1081-1092,
1989). We tested the validity of this mathematical model in a inspiration, is the sum of tidal elastic pressure [the quo-
single-compartment analogue of the respiratory system across tient of tidal volume (VT) and compliance] and Pex. In
wide ranges of clinician-set variables (frequency, duty cycle, turn, Pex is the sum of static and dynamic components,
applied pressure) and impedance conditions (inspiratory and the static component being the set level of end-expira-
expiratory resistance and system compliance). Our data con- tory pressure (PEEP) and the dynamic component being
firm the quantitative validity of the proposed model when ap- the flow-driving difference between Pex and PEEP.
proximately rectilinear waves of pressure are applied and ap- The ability of PCV to facilitate ventilation or to con-
propriate values for impedance are utilized. Despite a fixed-cir- trol intrathoracic pressure depends on the interaction
cuit configuration, however, resistance proved to be a function between the clinician-chosen ventilator settings and the
of each clinician-set variable, requiring remeasurement of sys-
tem impedance as adjustments in these variables were made. impedance presented by the lung and chest wall. During
With further modification, this model may provide a practical PCV, the clinician makes only three selections (f, Pset,
as well as a conceptual basis for understanding minute ventila- and TI/TT) while the patient’s mechanics [respiratory
tion and alveolar pressure fluctuations during PCV in the clini- system compliance (Crs), inspiratory resistance (RI), and
cal setting. expiratory resistance (RE)] determine the impedance to
ventilation. At present, there is no widely accepted
mathematical modeling; mechanical ventilation; dynamic hy- method for predicting the effect of changes in the clini-
perinflation; exponential kinetics cian-chosen ventilator settings on VE and intrathoracic
pressures. Therefore the selection of appropriate venti-
lator settings remains largely an empirical process.
IN RECENTYEARS, pressure-controlled ventilation (PCV) Numerous mathematical, electrical, and mechanical
has been used with increasing frequency to facilitate the models have been proposed to simulate respiratory sys-
ventilatory management of critically ill patients (9, 12, tem behavior (3, 4, 6, 7, 10, 11). In many, attention has
19). During PCV, approximately square waves of pres- concentrated on the inspiratory portion of the breathing
sure are applied and then withdrawn from the airway cycle while interactions between the inflation and defla-
opening at a pressure level (Pset), frequency (f), and in- tion half cycles have been neglected, despite their poten-
spiratory time fraction (TI/TT) selected by the clinician. tial importance. Moreover, the majority of these models
Fixing the airway pressure profile, f, and TI/TT prevents have assumed equivalence of RI and RE, an assumption
the machine from adjusting volume output in response to that seldom holds in the clinical setting (13, 18). Re-
changing inflation impedance. Therefore, minute venti- cently, we proposed a mathematical model that describes
lation (VE) may fluctuate with changes in respiratory important linkages between the critical inputs to PCV
mechanics. and the ventilatory and p lressure 0 utputs of clinical inter-
Despite this drawback, several potential benefits may est (14) (Table 1). In an attempt to incorp lorate charac-
justify the clinical use of PCV. Setting a maximum air- teristics of clinical relevance that have been neglected
922 0161-7567193 $2.00 Copyright 0 1993 the American Physiological Society
Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.
 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 923

TABLE 1. Modeling and limit equations METHODS

Mechanical Analogue
Modeling equations
VE = (f . (p& . Crs) . (1 - e-GOD’mlCrs) A manufactured “test lung” (model 1600, Michigan In-
. [ 1 - e-60(1-D)‘mECrs] }/
P
_ e-60(1-D)/fR~Crs . e-60D/fR&!rs
1 struments, Grand Rapids, MI) was used as our mechani-
PA= [f/(60. Cm)] [(V,. TI) - (V, - VEE) (RI Crs)
l l
cal analogue of the respiratory system. This mechanical
l .

(1 - e-Tr’RICrs ) + (VT + VEE) (RE Crs) (1 - e--TE’RECrS)] model mimics a single-compartment

l l

respiratory system
-60(1-D)/fRECrs
Pex = { Pset [e 1 . (1 _ e-60D/fRrCrs) } /
in that it allows independent settings of resistance and
l

_ e-60(1-D)/fRECrs. e-60D/fFtICrs
P I
compliance, the major components providing the imped-
Limit equations ance to ventilation, over the ranges typically observed in
vE= [60*Pset*D(l - D)]/[RE*D + R10(1 - D)] patients being treated for respiratory failure with me-
PA= (P~~~*D.RE)/[RE*D + RI+ - D)]
(P~~~*D.RE)/[RE*D chanical ventilation. In addition, bellows (simulated al-
Pex = + R10(1 - D)]
veolar) pressure is easily recorded. We used a mechanical
Modeling equations describe output variables of clinical interest as a ventilator (Servo 9OOC, Siemens-Elema, Solna, Sweden)
function of chest impedance and clinician-set variables. Limit equa- operating in the “pressure control” mode to approximate
tions describe theoretical bounding limits for output variables as f + a “square-wave” pressure generator. To enhance flow de-
co. [Adapted from Smith and Marini (14).] f, Frequency; Pset, level of
applied pressure; Crs, total respiratory system compliance; e, base of livery and thereby improve pressure-generating charac-
natural logarithm system (2.7183); D, duty cycle (TI/TT); RI, inspira- teristics, we used two such ventilators electrically cabled
tory resistance; RE, expiratory resistance; PA, mean alveolar pressure; together for cycle synchronization, and the internal reser-
V, = Pset Crs; TI, inspiratory time; VEE = Pex Crs, volume above
l

voir
l

of each ventilator was maintained at its maximum

functional residual capacity due to.Pex; TE, expiratory time; Pex, auto-
positive end-expiratory pressure; VE, minute ventilation. working pressure value (120 cmH,O) throughout all ex-
periments.

previously, our biexponential model addresses the inter- Variable Definitions

action between the inspiratory and expiratory half cycles In these experiments, output variables are the volume
and allows for different RI and RE. In addition to esti- (VE) and pressure (PA and Pex) variables predicted by
mating certain outputs of clinical interest, the -model the modeling equations. Input variables are of two types,
predicts distinct bounding limits for VE, Pex, and PA as f PI) and clinician set (CS). PD variables (RI, RE, and Crs)
is progressively increased at fixed values of Pset and describe the impedance to ventilation. CS variables (f,
TI/TT. TI/TT, and Pset) describe how the pressure wave is deliv-
As with all mathematical models, certain simplifying ered to the patient. PACT, another important variable of
assumptions are made. We assume that the respiratory clinical interest, was not specifically modeled, because
system behaves as a single compartment, with constant with VE, f, Crs, and Pex known, PACTis a dependent
“lumped” values for resistance and compliance. Further- function of these quantities: PACT= VT/h + Pex. To
more, it is assumed that perfectly rectilinear (“square”) stringently test the validity of the proposed mathemati-
waves of constant pressure are phasically applied to a cal model, we examined a wide range of inputs, as de-
passive system in which volume builds and decays as an tailed below.
exponential function of time and inflation/deflation im-
pedance. In addition, Crs, RI, and RE are assumed to lMeasurement Techniques
remain constant during the course of a tidal breath. The Output variables. VOLUME OUTPUT. All experiments
equations that define the maximum achievable values of were conducted using room air. Flow was measured in the
ventilation and alveolar pressure assume that Pset, TI/ common limb of the Y-piece of the ventilator circuit with
TT, RI, and RE remain constant as f is increased. We also a Fleisch no. 2 pneumotachograph, and VT was deter-
assume the effects of inertia to be negligible. mined by electrical integration of the inspiratory flow
The purpose of the present work was to determine how signal (HP 8815 respiratory integrator). The combined
well the simplified mathematical model we proposed pre- linearity of our pneumotachograph and differential pres-
diets the behavior of a passive mechanical analogue of sure transducer was measured to be t0.8% over the flow
the respiratory system over a spectrum of simulated clin- range O-60 l/min. VE was calculated as the product of VT
ical conditions. Both the mathematical model and me- and f. We confirmed the accuracy of this method in a
chanical analogue are designed to simulate a single-com- separate experiment by simultaneously measuring VE by
partment respiratory system. In clinical practice, the re- the timed displacement of a water-sealed spirometer at
spiratory system is actually composed of multiple the highest f studied [ 100 cycles/min (cpm)]. The VE
compartments, each with different impedance values. At determined from the pneumotachograph was within 5%
the bedside, however, it is possible only to measure single of the value obtained by volume displacement.
lumped impedance values that apply to the integrated PRESSURE OUTPUTS. All pressures were sensed by vari-
respiratory system. Because the patient-determined able-reluctance differential pressure transducers (model
(PO) impedance inputs to the mathematical model are MP45-871, Validyne Engineering, Northridge, CA) and
lumped values, the outputs of the model, being derived recorded simultaneously with flow and VT on a multi-
from the inputs, are also lumped values. Therefore, if channel strip chart (model 7754A, Hewlett-Packard,
valid, the model would only be useful in examining the Waltham, MA). The mean alveolar (bellows) pressure
overall response of the respiratory system and not the (PA) was determined by dividing the total area enclosed
responses of individual system compartments. by the alveolar pressure vs. time curve during both

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION

piratory valve as well as that imposed by the expiratory

resistor, the tubing, and the pneumotachograph. Compli-
ance was measured by dividing inspired VT by the differ-
ence between PACT and Pex. Because the volume com-
partment of our test lung has a maximum capacity of 2
liters, input combinations that result in end-inspiratory
e inspiration + - expiration ,-* volumes >2 liters could not be tested. Consequently, data
collection was initiated at f > 5 cpm for some conditions
(moderate Crs, high Pset).
CS-SET VARIABLES. Sufficient time (>2 min) was allot-
ted between each change of the CS input variables to
allow all output variables to achieve their final (“steady-
state”) values before data collection. TI was measured as
the time (seconds) between the beginning of inspiratory
flow and the initiation of expiratory flow; TT was mea-
sured from the flow tracings as the time between succes-
sive inspiration onsets; f was determined as 60/T~. To
partially compensate for the ventilators’ inability to
achieve the desired rectilinear pressure profile at rapid f
(Fig. 2), we used the airway pressure measured at end
inspiration as the Pset value entered in the modeling
equations. (No attempt was made to adjust for the shape
Time U
ii of the applied pressure waveform.)
FIG. 1. Relationship of airway pressure (fine line), alveolar pressure
(heavy line), and flow during pressure-controlled ventilation. Point
measures of resistance at 25, 50, and 75% of volume above functional Experimental Protocol
residual capacity were calculated by dividing driving pressure (P,) by
flow rate. Mean values for inspiratory resistance and expiratory resis- For two compliance values (moderate ~0.06 l/cmH,O,
tance were calculated by dividing respective resistive pressure-time low ~0.02 l/cmH,O), we tested three different combina-
product (hatched areas) by tidal volume. tions of resistances: RI > RE, RE > RI, and RI N RE. To
test the robustness of our linear model to obvious viola-
phases of the tidal cycle by total cycle time. The areas of tions of its assumption of flow-independent resistance,
recorded tracings were determined by computer-aided we varied resistance by placing combinations of para-
planimetry. Pex was measured as the end-expiratory bel- bolic resistive elements (7.7 and/or 5.6 mm ID, Michigan
lows pressure. No PEEP was applied in any experiment. Instruments) in the inspiratory and expiratory limbs of
Input uariables. PD VARIABLES. Although our mathe- the circuit. These resistors have a pressure-flow relation-
matical model requires single-valued inputs for RI and ship that is distinctly alinear (“parabolic”) in nature, re-
RE, to our knowledge there is no general consensus as sembling that described by others for endotracheal tubes
how best to quantify the average nonelastic impedance (17). We empirically determined which configurations of
under conditions of varying nonlaminar flow. We there- the parabolic resistors would give the desired combina-
fore examined four different methods of determining RI tions of RI and RE, as already defined. The system re-
and RE. Three resistance values were point measures of sponse to varying the CS inputs was examined in three
the ratio between driving pressure (P,) and the resulting experiments: the first defined the relationship off to out-
flow. The other was a more global measure. Point resis- put with all other inputs held constant, the second varied
tances were determined at 25,50, and 75% of inspired VT TI/TT in a similar fashion, and the third varied Pset. The
during both inspiration and expiration (RI,, . and RE,, , l

RI,, and RE,, , RI,, and RE,,) (Fig. 1). Duringinspiration, -30
P, for the RI calculation was considered to be the pres- =20
sure difference between the proximal airway (Pao) of the =10 =
external circuit and the bellows (PA) compartment of the cm -

test lung. Because PEEP was not applied, the pressure

driving flow during exhalation was the difference be-
tween PA and atmospheric pressure. Global resistance 1;;:: i \

(RI, i%E) was calculated by dividing the resistive pres-

sure-time product by VT (Fig. 1) t=30 t = 50 f=lOO

TI TE - Time -
P,dt P,dt 1 set
- s 0 - s TI I i
RI = RE =
FIG.2. Airway pressure profiles traced during pressure-controlled
VT VT ventilation at 4 different frequencies for a fixed-circuit configuration
(expt I). Note fall in end-inspiratory airway pressure and deformation
Each measure of RI included resistance imposed by the of pressure profile as frequency (f) increases. Applied pressure was
inspiratory resistor, co nnecting tubing, and pneumota- high, expiratory resistance exceeded inspiratory resistance, duty cycle
chograph; RE included resistance of the ven tilator’s ex- was 0.33, respiratory system compliance was 0.02.

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 925

TABLE 2. Ranges of clinician- and patient-set variables used to examine response of output variables
Clinician-Set Inputs Patient-Set Inputs

Expt. No. Pset, cmH,O f, wm TI/TT Crs, l/cmH,O Resistance class

RI > RE
Moderate (a 15) 5-100 0.33 Moderate (-0.06) RI = RE
RE > RI

RI > RE
I Low (aO.02) RI = RE
RE > RI

RI > RE
High (~30) 5-100 0.33 Moderate (~0.06) RI = RE
RE > RI

I
RI > RE
Low (zO.02) RI = RE
RE > RI
RI > RE
II High (~30) 10-50 0.2-0.8 Moderate (aO.06) RI = RE
RE > RI

III 10-80 20 0.5 0.04 RI = RE

For mean values for selected conditions, see Table 3.

measured PD and CS inputs were entered into the model-
ing equations to predict the output variables.
Experiment I (frequency varied). In experiment I, the
response of key output variables to increasing f was ex-
amined across a spectrum of conditions that varied im-
pedance and Pset while TIITT remained constant (Table
2). Data collection was performed by setting Pset and
Crs, placing the resistance elements, and then increasing
f from 5 to 100 cpm, with measurements at 5, 10, 15, 20,
30,40, 50,60,80, and 100 cpm. In addition to recording
the output variables of interest, all PD and CS input
variables were measured for one representative breath at
every condition tested.
Experiment II (TIITT varied). We examined each level
of TI/TT available on our ventilator (0.20, 0.25, 0.33,
0.50, 0.67, 0.80) at a single Pset for each of three imped-
ance classes(Table 2). This experiment was repeated for
each of five f (lo-50 cpm in steps of 10).
Experiment III (Pset varied). Apart from f and TI/TT,
Pset is the remaining CS input. When impedance, TI/TT,
least-squares linear regression of the form y = mx + b,
where y is the observed value, x is the predicted value,
and m and b represent the slope and y-intercept of the
expression, respectively (15).
RESULTS
Resistance
Although our intention was to observe the outputs re-
sulting from variation of only a single input variable, we
found resistance to be flow dependent and, therefore,
strongly influenced by Pset and TI/TT, two of the CS
inputs. As exemplified in Fig. 3, there was also within-cy-
cle variation in the values of resistance obtained by the
different methods. Single-factor analysis of variance sep-
arated methods of calculating RI into two groups:
RI,, and I&, resulted in higher values of resistance than
RI,, and RI (P 5 0.008). Within each group there was no
difference between methods. On average, the difference
between groups was only 3 cmH,O sol? For RE, the l

and fare held constant, the modeling equations (Table 1)
predict a linear relationship between Pset and each of the
output variables (VE, PA, and Pex). We tested these rela-
tionships in a separate experiment in which we varied
Pset (over lo-80 cmH,O) while keeping f constant at 20
four different methods resulted in three groups: RE,, and
RE resulted in the highest values of resistance and RE,,
the lowest. RE,, was intermediate to the other methods.
On average, the difference between the highest and low-
est groups was only 4 cmH,O s 1-l. Although there were
l l

cpm and TI/TT constant at 0.5. Here, the impedance
combination was Crs = 0.04 l/cmH,O with a single (5.6-
mm) parabolic resistor common to both inspiratory and
expiratory limbs of the circuit.
Statistical Analysis
Single-factor analysis of variance was used to deter-
mine statistical significance between the different meth-
ods of determining resistance. The Bonferroni multiple
comparison method was used to compare the means ob-
tained from the different resistance calculations to en-
sure an overall type I error risk 10.05. To assess the
overall accuracy of the mathematical model, we used
statistically significant differences in the values for resis-
tance determined by the various calculation methods,
such small disparities are unlikely to be of clinical signifi-
cance. Therefore the method of calculating resistance
was not considered critical to this analysis.
To characterize the flow dependence of resistance in
our circuit, we varied flow and used a logarithmic trans-
formation of the resulting data to solve for Ein the equa-
tion P, = KV, where P, is driving pressure, V is flow, K is
a proportionality constant that depends on geometry and
gas composition (ZO), and 6is a flow exponent that varies
from xl (under fully laminar conditions) to x2 (under
fully turbulent conditions). For the conditions RI > RE,
RI = RE, and RI < RE, the inspiratory flow exponents

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

926 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION

AknH20*s/l) Table 4 displays the regression information character-

izing the relationship between observed (y-axis) and pre-
dicted (x-axis) values for the three output variables stud-
20
1 ied over all conditions as well as the average difference
1 between the observed and predicted values. Results from
15. the three experiments will be categorized by their impli-
- RIRS m---m ((I#) - - RI7S *-1-s R-m cations for each output variable.
lo-
Output Variables
VE. When observed and predicted values for VE were
examined for all tested conditions, a good linear fit was
obtained, with some error occurring in data points scat-
tered to the right of and below the line of identity (Fig. 5).
40 60 80 100
FREQUENCY kpm)
The 95% confidence interval (generated from all obser-
vations) for the mean difference between actual and pre-
dicted VE was -2.7 to -1.8 I/min, indicating that the
6 mathematical model tended to overestimate VE.
kmH20*s/l)
Figures 6 and 7 demonstrate the agreement of pre-
dicted VE and observed TjE under conditions of increas-
ing f and TI/TT, respectively, for a fixed Pset and various
combinations of impedance. The model described the ob-
served behavior quite well at low and moderate f but de-
viated substantially from the observed behavior at high
cycling rates. Predictive accuracy was generally better at
i
6 high rather than low values of TI/TT.
5 - Rm --m-w km -- Rt7s .-.-.. ban
j

A
PRESSURE kfrd$O)
0 -Ir----Te-rp-I-- 7-----,17
c
0 20 40 60 80 100 50
1
FREQUENCY (cpm)
FIG. 3. Inspiratory resistance (RI, A) and expiratory resistance (RE, 40
i
B) plotted by frequency in cycles per minute (cpm) for duty cycle of 1
0.33, moderate respiratory system compliance, low applied pressure, 30 1
I
and RE > RI (eJcptr). Curves represent straight lines connecting dis- I
crete data points. 20
i

averaged 1.92, 2.06, and 1.66, respectively. The corre-

sponding expiratory flow exponents were 1.79,1.‘79, and
2.00, respectively. T ese data are consistent with the
0 30 40 50
characteristics of the esistors (isolated from the circuit) FLOW (liters/min)
exposed to a series of constant flows (Fig. 4). Such expo-
nents indicate that resistance in our system was highly
flow dependent and emphasize the need to remeasure RI B
PRESSURE (log base e)
and RE for each combination of CS and PD variables.
Resistance decreased as inspiration proceeded, reflect- 4’

ing the decelerating flow pattern. However, contrary to 3-

expectation, resistance tended to increase as expiration
2-
progressed. We attribute this behavior to the flow-re- +
tarding action of the ventilator’s scissorlike expiratory
valve, which seriously disrupts the exponential decay of
alveolar pressure assumed by the model. Because point
measures of resistance fluctuated throughout both half
cycles and because point measures of resistance during
exhalation may be seriously affected by valve action, we 1 2 3 4 5
reasoned that RI and i%E (which incorporate early, mid-, FLOW (log base e)
and late cycle changes in resistance) provide the most FIG. 4. A: pressure-flow profiles for the 2 parabolic resistors used in
representative overall measure of nonelastic impedance. these experiments [5.6 mm ID (0) and 7.7 mm ID (@I. Data were
Consequently, all equation-derived estimates were gen- obtained by measuring - pressure
- difference (vertical scale) across each
resistance exposed to a series of constant flows. B: logarithmic trans-
erated using the RI and i% values. Table 3 displays the formation ofthe same data with corresponding lines of best fit deter-
actual values of test lung impedance and Pset, averaged mined by least-squares linear regression. Slope of each best-fit line is
over f and TI/TT for all observations in the first two flow exponent (c) in expression P, = K” (C is similar for both resistors).
ext3eriments. Pressure intercept corresponds to In K (see text).
Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.
 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 927

TABLE 3. Input variables averaged over f and TIITT ,OBSERVED

for expts I and II 401

Resistance Class

Variable RI > RE RI = RE RE > RI

Moderate Pset, low Crs (n = 10) *

RI, cmH,O l s 1-l

l 12.1t2.2 4.7tl.4 2.9tl.l
RE, cmH,O l s 1-l l 6.3tl.O 5.8tl.4 13.421.3
Crs, ml/cmH,O 2O.OkO.7 20.6kO.3 20.9kO.9
Pset, cmH,O 17.2~11.6 17.M1.6 18.Ok1.7
Moderate Pset, moderate Crs (n = 1 O)*
0 10 20 30 40
PREDICTED
RI, cmH,O s 1-l 13.8+1.1 6.5kO.8 3.7kO.7 FIG. 5. Observed minute ventilation (VE) plotted against predicted
l l

RE, cmH,O l s 1-l l 5.8t1.3 5.9kl.O 13.6k0.7 VE for expts I-III (n = 215).0, f < 40 cpm and duty cycle (TI/TT) > 0.3
Crs, ml/cmH,O 57.7k5.8 62.4k2.4 63.2k2.7 (n = 122). +, f > 40 cpm and TI/TT < 0.3 (n = 93). With use of all
Pset, cmH,O 15.4k1.8 15.6t2.0 16.Ok1.8 observations, r = 0.92. With use of only observations in which f < 40
High Pset, low Crs (n = 10) * cpm and TI/TT > 0.3, r = 0.98. Dashed line, line of identity.
RI, cmH,O l s 1-l l 15.7k2.0 7.8t2.5 5.0t1.9
RE, cmH20 6 s 1-l l 7.3kl.6 8.6kl.O 20.1t1.3 values in the limit equations. We found that such calcu-
Crs, ml/cmH,O 19.5kO.5 19.2t0.4 18.9k0.4
Pset, cmH,O 27.7k2.8 28.6t3.0 29.2k2.8 lations resulted in theoretical limits for maximum \jE
that were consistently high in relation to those observed
High Pset, moderate Crs (n = 33) t
(Fig. 6).
RI, cmH,O s 1-l 18.2k2.9 8.9t2.0 6.3k2.0 P’A. Except for a slight offset, our mathematical model,
l l

RE, cmH,O s 1-l 7.7t2.5 9.1k2.1 22.5k5.0

adjusted for the value of peak pressure actually applied,
l l

Crs, ml/cmH,O 58.024.0 53.6t2.8 52.8t5.5

Pset, cmH,O 28.6t2.8 26.3k3.3 27.7k3.2 accurately predicted PA of the mechanical lung analogue
across all conditions (Fig. 9). With all observations con-
Values are means k SD. RE and RI, mean expiratory and inspiratory
resistance, respectively. * Results from expt I. t Combined results from sidered, the 95% confidence interval for the mean dif-
expts I and II. ference between measured PA and predicted PA was
-1.2 to -0.9 cmH,O, indicating that the prediction over-
At variance with the model’s prediction, we observed a estimated the PA actually observed.
curvilinear relationship of TjE to increasing Pset (Fig. 8). Figure 10 illustrates the ability of the model to describe
We attribute this behavior to the joint dependence of RI PA for various impedance combinations as f was varied.
and RE on the physical resistor used and Pset itself. Sub- The model described the observed behavior quite well at
stituting the RI and RE values actually measured (at each all f, with a near-constant positive offset of the predicted
Pset) into the model yielded predicted values almost in- values from those measured. With RI > RE, PA fell pro-
distinguishable from those observed (r = 0.99). Thus, for gressively as f was increased, approaching a lower
a specific set of impedance settings, the model accurately bounding limit under conditions of moderate Crs. Con-
described the behavior of the lung analogue in response versely, with RE > RI, PA tended to rise toward a final
to Pset changes, provided that accurate measurements of plateau in response to increasing f. As predicted by the
RI and RE were used as inputs. mathematical model, these f-dependent changes in PA
As previously noted, the measured inputs (RI, RE, and occurred primarily over the lower-to-moderate f range.
Pset) did not remain constant as f was increased toward Over the same f spectrum, the pressure waveform was
its extreme value. To exam.ine the usefulness of the limit quite rectilinear, and therefore mean airway pressure
equations with respect to VE (Table l), we averaged RI, (the product of Pset and TI/TT) remained essentially
RE, and Pset over all f and then used these f-averaged constant.
Despite the fact that RI and RE varied with Pset, we
TABLE 4. Values for m, b, and r and mean difference found a linear relationship between PA and Pset. The
between observed and predicted values linearity of this correlation was in contrast to the distinct
alinearity of the PS&-VE relationship (Fig. 8).
Observed
- Predicted
As f increases, the theoretical limit of PA is a compli-
Variable n m b r Mean ance-independent value that can be calculated from resis-
tance, TI/TT, and Pset alone (Table 1). To calculate
iJE 215 0.79kO.02 1.89t0.46 0.92 -2.23t0.22 these limits, we used RI, RE, and Pset averaged over all f
llmin
PA 215 0.97t0.01 -0.68kO.11 0.99 -1.06kO.07 for a given set of impedance conditions and constant TI/
cmH,O TT. In no instance did the measured PA achieve these
Auto-PEEP* 105 0.99t0.01 -0.5520.12 0.99 -0.64kO.07 theoretical limits (Fig. 10).
cmH,O Pex (auto-PEEP). With no PEEP applied, Pex = auto-
Ideal 1.00 0.00 1.00 0.00 PEEP. In our analysis of Pex, we only used observations
with measurable Pex (measured Pex > 0). When all pre-
Values are means k SE. m, slope; b, constant terms; r, correlation
coefficients. Data include all observations (expts I-III). * Data ana-
dicted and observed values (for measurable Pex) were
lyzed only for observations where auto-positive end-expiratory pres- plotted, a good linear fit was obtained, regardless of f or
sure (PEEP) was measurable. TIITT (Fig. 11). With all observations considered, the

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

928 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION

tB (liters/min) tB (liters/min) iB (literS/min)

FiwmcY kpd FfWUENCY kpr) FfEaENcY kprl

iB (h?r+in) iB (titers/min) irr (litm/min)

FIG. 6. Predicted VE (dashed line) and measured VE (solid line) as f is varied, for different impedance conditions.
Applied pressure (Pset) was low and TI/TT = 0.33 in all panels. Top: low compliance (Crs). Bottom: moderate Crs. Left,
middle, and right: RI > RE, RI = RE, and RE > RI, respectively. Note distinct f threshold, above which predicted values
progressively overestimate observed values. With use of f-averaged RI, RE, and Pset, calculated limits of ATE, proceed-
ing clockwise from top Left: 22.4, 46.6, 37.3, 30.5, 37.9, and 18.3 l/min, respectively. In no case did the measured values
exceed these theoretical limits. Curves represent straight lines connecting discrete data points.

95% confidence interval for the mean difference between
measured Pex and predicted Pex was -0.8 to -0.5
cmH,O. The response of measured and predicted Pex as f
was increased to 400 cpm is shown in Fig. 12. Contrary
to prediction, Pex was a curvilinear function of Pset, re-
flecting the pressure-dependent (flow-driven) rise in RE
(Fig. 8). When impedance inputs were measured at each
Pset, the mathematical model correlated very well with
the observed data (Fig. 8). As was true for VE and PA, the
observed f-driven limits for Pex were significantly below
those predicted (Fig. 12).
DISCUSSION
We found that our biexponential model accurately de-
scribed the behavior of the mechanical lung analogue
during PCV over a wide range of simulated clinical con-
ditions, provided that appropriate impedance variables
were input to the modeling equations and that approxi-
mately square waves of pressure were applied and re-
leased from the airway opening. To maintain accuracy,
however, it was necessary to input measurements of RI,
RE, and Crs that corresponded to the specific f, TI/TT,
and Pset combination of interest. RI and RE, the average
values of flow-resistive pressure loss per unit of flow,
varied with changes in Pset, TI/TT, and f, variables that
influenced the mean flow rate across the fixed parabolic
resistors we used in the system. Despite severalsimplifi-
cations and theoretical shortcomings, this model ex-
plained the great majority of the data we collected.
Our results confirm the general validity of several im-
portant implications of the mathematical model. First,
VE is a Curvilinear fu nction of f that demonstrates a dis-
Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.
tinct upper limit for each specified set of Pset, RI, RE,
and T~/T~.parameters (Fig. 6). Second, the rate of ap-
proach of VE to the bounding value (but not the limit
itself) is a function of system compliance (Fig. 6). Third,
unlike mean airway pressure, which is independent of f
and system impedance during PCV, PA proved to be a
dependent function of these variables that may exceed or
underestimate mean airway pressure, depending on the
relative magnitudes of RI and RE (Fig. 10).
Two primary difficulties prevented our mathematical
model from accurately predicting the output variables
across the entire spectrum of conditions tested. Resis-
tance, calculated as the quotient of flow-resistive pres-
sure and flow, was variable in this parabolic (nonlinear)
system, influenced by the specific flows generated. As a
result, changing Pset or TI/TT can cause RI and RE to
vary, even when all other inputs remain the same. In
addition, the profile of applied pressure often departed
from an ideal rectilinear configuration, especially under
conditions of high flow (low impedance) or high cycling f.
Qualitative assessment of model accuracy can be made
by graphing predicted and measured output variables
against one of the input variables. Figures 6-8,10, and 12
demonstrate the similarity of behavior for the mathemat-
ical model and mechanical analogue and illustrate the
conceptual validity of our equation system. A perfectly
accurate mathematical model would produce regression
coefficients between observed and predicted values of 0
and 1 for the constant term and slope, respectively, and
correlation coefficients across the range of data would
approach 1.0. When the average differ&c e between the
observed and predicted values is used as a. measure of
 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 929

14’

Measured h (liters/min)

c----+i
----X.,\
21:
if
4: ,‘%,
14i 14j 141
: i X -f=l() -c?>
----- fg)
71 74
’! 7;
--ff=Jo -- f=J(j --ff=Jo
‘1 I
OC - r-- ‘I’7 --- --.. r-- OC- - T----r- - T---T--V ok-, -r--- -r--f-.- -1

0.1 0.3 0.5 0.7 0.9 0.1 0.3 0.5 0.7 0.9 0.1 0.3 0.5 0.7 0.9
TI/lT TI/Tl TI/TT

FIG. 7. Relationship of predicted (top) and measured (bottom) values of VE to TI/TT for RI > RE (left), RI = RE
(middle), and RE > RI (right). Pset was high and compliance was moderate in all panels. Note tendency for predicted
values to overestimate observed values at-low TI/TT.

accuracy, the mean difference should approximate zero,
and twice the standard deviation of the difference should
be clinically insignificant. By these criteria, VE was the
only variable to show significant problems. Even here,
discrepancies were much smaller when the data for high f
(~40 cpm) and short TI/TT (~0.3) were deleted. The
correlation coefficient increased from 0.92 to 0.98 for the
subsetted data. We attribute the overestimation of our
model for VE at high f and brief TI/TT to the inability of
our pressure generator to maintain a square-wave profile
(Fig. 2).
MINUTE VENTILATION
(1 iters/min)
Flow-Dependent Variation in Resistance
As others have emphasized (17), the flow-resistive
properties of the intact respiratory system cannot be de-
scribed adequately by a single value. Yet two parameters
(K, C)and a single equation of the form P, = K’do surpris-
ingly well in fitting a variety of experimental data (16). t
varies with airflow turbulence and can theoretically
range from 1 (purely laminar) to 2 (fully turbulent), de-
pending on system impedance and flow range examined.
Our calculated exponents (1.7-2.0) indicated a high de-
Pa
kmH201

60 .50
1 1
t 401
FIG. 8. Relationship of Pset to mea-
sured resistance (bott0.m right) and to
each output variable [VE, auto-positive
end-expiratory pressure (auto-PEEP),
and mean alveolar pressure (PA)] for a
Ok-r-l--- 171--T OL- 7 --T--r-- r~--T--T---r-- -j----r.yr
fixed-circuit configuration, f and TI/TT.
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 0 and l , Predicted and measured val-
Pset (cmH20) Pset kmH20) ues, respectively, for output variables.
Bottom right: 0, RE; l , RI. Solid and
AUTO-PEEP RESISTANCE
kmH20) (cmH20*s/l)
dashed lines, linear regression relation-
ships. As Pset increased, RI and RE in-
creased. In graphs of output variables,
straight line was generated by model,
with assumption that RI and RE re-
mained constant at values measured at
Pset = 20 cmH,O. Crs = 0.04, TI/TT =
0.5, f = 20 cpm.

0 10 20 30
I ’ I f’r r1-’1-l7rr
40 50 60 70 80 90 0 10
7--I--l-- -l--‘--r----l-I
20 30 40 50 60 70 80 90
Pset (cmH20) Pset kmH20)
Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.
 930 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION

OBSERVED flow and volume exponents. In general, nonlinear prob-

501
1
0 10 20
PREDICTED
FIG. 9. Observed PA plotted against predicted
= 215). Dashed line, line of identity.
gree of turbulence, as expected when parabolic resistors
are employed. Such turbulence accounts for the range of
“point” resistances encountered within the span of a sin-
gle breath (Fig. 3). In general, resistances determined at
50% of VT approximated quite closely the “volume-aver-
aged” resistances used in our modeling calculations.
Nonetheless, we believe that the latter values more con-
sistently represent nonelastic impedance across a wide
range of conditions and flows.
In the current model, it is assumed that unique and
unchanging values can be specified for impedance param-
eters: RI, RE, and Crs. Such an assumption implies linear
relationships between dissipative (flow-resistive) pres-
sure losses and flow and between elastic pressure losses
and volume. Clearly, such assumptions of linearity are
routinely violated in the clinical setting, where flow is
turbulent and hysteresis is common. Not to make such
assumptions of linearity, however, renders the solution
of the resulting nonlinear differential equations, and
therefore the development of “closed-form” explicit alge-
braic solutions for the outputs of interest, exceedingly
difficult for all but a small subset of unusually tractable
lems yield only to iterative computer-assisted numerical
approximations of the exact solution.
Such nonalgebraic approaches often provide accurate
predictions of actual behavior. Yet, for many purposes it
may be neither desirable nor necessary to depart from a
linear model primed with appropriate empirical inputs.
As already noted, single lumped values for resistance and
compliance are all that can practically be determined in
clinical practice. Moreover, our results demonstrate that,
even in a highly nonlinear system, a linear model with
appropriate impedance inputs mimics the actual behav-
30 40 50 ior of certain output variables with considerable preci-
sion, provided that the assumed pressure waveform is
PA for expts I-III (n
actually applied.
Despite the obvious nonlinearity of the pressure-flow
relationship in our experiments, empirically determined
lumped estimates for RI and RE proved to be appropriate
inputs to the model, closely estimating the observed out-
puts produced by a truly rectilinear waveform of applied
pressure (Fig. 8). The mathematical model therefore ap-
pears robust to departures from ideal input characteris-
tics, so long as appropriate lumped values for impedance
are used.
To predict accurately across a range of conditions, our
mathematical model requires that all input variables re-
main independent of one another. Mutual independence
is characteristic of all CS variables (Pset, f, and TI/TT)
and between each of these and compliance. However, be-
cause calculated resistance is a function of flow rate in a
turbulent system, RI and RE were greatly affected by
adjustments of each CS variable. Average inspiratory
(VI) and expiratory (VE) flow rates can be computed as
the quotients of VE and TI/TT: 91 = ~EI(TIITT); VE =
iTE/[l - ( TI/TT)] . Consequently, variations in TI/TT
also affected phase-specific flow rates (and the corre-
sponding resistance value), even when they had little ef-
fect on VE. At the same f, reducing TI/TT tended to in-

PA (cmH20) FA (cmH20)
161RI GREATER 161
LOW Crs -----0
)e-- --0 -- ----a__
j ---- ---
12; & -
f
84
1 GREATER FIG. 10. Predicted PA (dashed line)
4i RE
LOW Crs
and measured PA (solid line) as f was
varied, for different conditions of simu-
o$- 7r- r---1-- --l - -r- --r. -T--‘-T ok-- T- -~~ r --7.- ~~l-T~-T----, , , r lated lung impedance. Pset was high and
0 20 40 60 80 100 0 20 40 60 90 100 TI/TT = 0.33 in all panels. Top: low Crs.
FREQUENCY (cpm) FREQUENCY kpm) Bottom: moderate Crs. Left: RI > RE.
Right: RE > RI. With use of f-averaged
FA (cmH20) FA (cmH20) RI, RE, and Pset, calculated limits of PA,
proceeding clockwise from top left, are
RI GREATER 5.2, 19.4, 15.8, and 3.3 cmH,O, respec-
16 16
MODERATE Crs tively. In no case did measured values
1 #@-------
exceed these theoretical limits. Curves
12 12
j represent straight lines connecting dis-
crete data points.

4 1 RE GREATER
MODERATE Crs
1
OC- r--- 1- .- ---I- _T_--- T--- - 0 c - -l--- -r-------- ---
0 20 40 60 80 100 0 20 40 60 80 100
FREQUENCY (cpm) FREQUENCY (cpm)
Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.
 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 931
OBSERVED after each change of machine settings. Nonetheless, with
30 prior assessment of the input variables under the condi-
tions of interest, the results obtained from our mathemat-
ical equations closely approximated those observed in
20 the mechanical analogue in most instances.

Distortion of the Applied Pressure Profile

10
When the applied pressure profile was of the desired
rectilinear (square) shape, the pressure outputs of inter-
est (PA and Pex) correlated well with their model-esti-
0 I---T- f- t-I‘-rT-t- r- T mated values. However, in circumstances where the pro-
0 10 20 30 file of the applied pressure was deformed (e.g., at high f
PREDICTED with low resistance and high compliance), the predictive
FIG. 11. Observed end-expiratory alveolar pressure (Pex) plotted equations consistently overestimated VE (Fig. 6). This
against predicted Pex for expts I-III with observed Pex > 0 (n = 105). tendency was especially evident in our calculations of
Dashed line, line of identity. “limit” values, due primarily to f-dependent distortion of
the airway pressure profile.
crease 91 and reduce VE. Therefore, in our turbulent sys- Compared with the simple relationship of applied
tem, shortening TI/TT increased RI while reducing its pressure to PA, the impact of variations in applied pres-
expiratory counterpart. Changing Pset or f to increase sure on ventilation appears less straightforward (Fig. 8).
VE also tended to increase both RI and RE for similar The impact of f on the interrelationship of Pset to VE
reasons. These changes in RI and RE occurred without bears especially close examination. Our data indicate
any adjustment to the geometry of the test lung. In this that at low f the predictions of the model were extremely
context, “resistance” calculations reflect structural accurate for the great majority of conditions examined.
changes (e.g., airway caliber) only to the extent that flow At higher f, however, the predictive accuracy of the
during the cycle phase under consideration does not model tended to deteriorate, with degrees of overestima-
change. tion quite dependent on the impedance condition under
Because resistance values were affected by changes in study (Fig. 6). As already noted, the clinical ventilators
the CS variables, our linear model cannot be considered we used fail as square-wave pressure generators when
“predictive” in the strict sense of the term; to maintain used at “nonconventional” rates (Fig. 2). Similar obser-
accuracy, the PD input variables must be remeasured vations were made a decade ago by Boros and colleagues

Pex (cmH20) Per (cmH20)

Pset-LOW 20 met-LOW

1
20
LOW Crs MODERATE Crs
16

8
16 1

0 20 60 80 100
I
4
n

0 20

FREQUENCY FREOUENCY

Pex (cm&O) Per (cmH20)

20 Pset-HIGH Pset-HIGH

1
20
LOU Crs MODERATE Crs
16 1 16

12
i

40 20 40 60 80 100
FFEEWENCY (cpa) FREQUENCY (cod
FIG. 12. Predicted Pex (dashed line) and measured Pex (solid line) as f is varied, for different combinations of Pset
and simulated lung impedance. RE > RI and TI/TT = 0.33 in all panels. Top: low Pset. Bottom: high Pset. Left: low Crs.
Right: moderate Crs. With use of f-averaged RI, RA, and Pset, calculated limits of Pex, proceeding clockwise from top
left are 12.5, 10.3, 15.8, and 19.5 cmH,O, respectively. In no case did measured values exceed these theoretical limits.
Curves represent straight lines connecting discrete data points.

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

932 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION

Predictive accuracy suffered primarily when equilib-

rium did not occur and the applied pressure waveform
was not square. Three circumstances (acting alone or
together) contributed to these adverse conditions: 1)
high f that shorten TI, 2) brief TI/TT that prevents end-
B- SQUARE WAVE inspiratory equilibration of Pao and PA, and 3) low inspi-
WITHOUT EQUILIBRATION ratory resistance-high compliance conditions that place
high flow demands on the pressure generator. Unlike the
situation for pressure outputs, there is no simple adjust-
ment to the square-wave model that will compensate for
c ,““-~
the influence of profile distortion on the VE estimates.
NON-SQUARE WAVE From a theoretical standpoint, precise compensation for
W EQUILIBRATION
profile distortion across the f range would require that
the pressure profile actually applied be expressed as an
integrable function of time. From a practical standpoint,
FIG. 13. Diagrammatic representation of 3 possible interactions be- distortion of the pressure profile remains a major prob-
tween actual airway pressure (fine line) and alveolar pressure (bold lem for our model but primarily at f values that are not
line). Theoretically “ideal” rectilinear waveform assumed by mathe- commonly applied during conventional ventilation.
matical model (dashed line) is shown for comparison. 0 and l , end-in-
spiratory airway and alveolar pressures, respectively. In cases where
alveolar and airway pressures reached end-inspiratory equilibration Summary
(A), perfectly “square” waves of applied airway pressure were not re-
quired for predictions of model to be accurate, provided that end-in- The behavior of a single-compartment lung analogue
spiratory airway pressure was used as pressure input. Under conditions
of moderate-to-high inspiratory impedance, applied pressure wave ap- with respect to certain variables of clinical interest
proximated a rectilinear configuration and also resulted in an accurate (VE, PA, and Pex) can be described quantitatively with a
prediction, even when equilibration of airway and alveolar pressure did biexponential model of PCV, provided that rectilinear
not occur (B). Model encountered predictive difficulties when a “non- pressure waves are applied or that equilibration between
square” wave was applied and equilibration was not achieved (C).
airway and alveolar pressures is achieved at end inspira-
tion. Our mathematical model is of limited use, however,
(8) in their study of infant ventilators. No mathematical when we are trying to predict what would happen to the
model (including our own) can be expected to predict output variables as the CS variables are changed without
accurately with inappropriate inputs. allowing for the dependence of RI and RE on changes in
A widening disparity between predicted and observed these parameters. Without further validation and refine-
values for VE tended to emerge above a distinct threshold ment, this simplified mathematical model cannot be con-
f, below which the predictive accuracy of the model for sidered to be a practical tool for the clinical setting. .We
ventilation was extremely good (Fig. 6). We noted that have yet to test its predictive ability in heterogenous bio-
this threshold point invariably occurred at or near the logical systems. However, we believe that our model does
value for f at which alveolar pressure at end inspiration provide a basis for understanding pulmonary dynamics
(the sum of VT/&s and Pex) equaled Pset (the “equilib- during PCV and, therefore, has considerable educational
rium pressure” point). Such thresholds occurred at rela- value. With further modifications, it may eventually
tively high f under low compliance conditions. We rea- prove a useful guide in clinical practice.
soned that, for VT, it does not matter whether a square-
wave or a nonsquare pressure profile is applied, so long The authors thank Lisa Scherer and Nancy Florian for expert assis-
as flow stops at end inspiration, exhausting the flow- tance in the preparation of the typescript.
driving pressure difference originally present between Address for reprint requests: J. J. Marini, Pulmonary/Critical Care
airway opening and alveolus (Fig. 13). In other words, for Medicine, University of Minnesota, St. Paul-Ramsey Medical Center,
640 Jackson St., St. Paul, MN 55101-2595.
all values of f less than the equilibrium pressure fre-
quency, sufficient time was allowed during inspiration to Received 18 March 1991; accepted in final form 14 August 1992.
permit the applied pressure waveform to “catch up” with
the idealized square-wave profile (which would have
caused alveolar pressure to build more rapidly but then REFERENCES
hold at the end-inspiratory value). 1. ABRAHAM, E., AND G. YOSHIHARA. Cardiorespiratory effects of
Airway-alveolar pressure equilibration proved to be a pressure controlled ventilation in severe respiratory failure. Chest
sufficient but not necessary condition for model accu- 98: 1445-1449, 1990.
racy. The model predicts ventilation extremely well 2. AL-SAADY, N., AND E. BENNETT. Decelerating inspiratory flow
waveform improves lung mechanics and gas exchange in patients
when the applied pressure waveform remains square and on intermittent positive pressure ventilation. Intensive Cure Med.
empirically measured frequency-specific resistances are 11: 68-75, 1985.
input. This was demonstrated by experiment 111, in which 3. BAKER, A. B. Analogs of respiratory mechanics. Int. AnaesthesioZ.
Pset was varied from 10 to 80 cmH,O at a moderate f (20 Clin. 9: 39-64, 1971.
cpm) and a high RI (Fig. 8). Although airway-to-alveolar 4. BAKER, A. B., AND C. E. W. HAHN. An analog study of controlled
ventilation. Respir. Physiol. 22: 227-239, 1974.
pressure equilibration was not achieved, a square-wave 5. BAUM, M., H. BENZER, N. MUTZ, G. PAUSER, AND L. TONCZAR.
profile of pressure was applied and the model proved ac- Inversed ratio ventilation (IRV). Anaesthesist 29: 592-596, 1980.
curate. 6. BERGMAN, N. A. Fourier analysis of effects of varying pressure

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.

 MODEL VALIDATION FOR PRESSURE-CONTROLLED VENTILATION 933

waveforms in electrical lung analogs. Acta Anczesthesiol. Scund. 28: chunics of Breathing. Bethesda, MD: Am. Physiol. Sot., 1979, sect.
178-181,1984. 3, vol. III, pt. 1, chapt. 18, p. 277-293.
7. BERGMAN, N. A. Intrapulmonary gas trapping during mechanical 14. MARINI, J. J., P. S. CROOKE, AND J. D. TRUWIT. Determinants and
ventilation at rapid frequencies. Anesthesiology 37: 626-633, 1972. limits of pressure-preset ventilation: a mathematical model of
8. BOROS, S. J., D. R. BING, M. C. MAMMEL, E. HAGEN, AND M. J. pressure control. J. Appl. Physiol. 67: 1081-1092, 1989.
GORDON. Using conventional infant ventilators at unconventional 15* NETER, J., W. WASSERMAN, AND M. H. KUTNER. Applied Stutisti-
rates. Pediatrics 74: 487-492, 1984. cul Models (2nd ed.). Irwin, IL: Homewood, 1985.
16. NUNN, J. F. Applied Respiratory Physiology (3rd ed.). Boston, MA:
9. BOYSEN, P. G., AND E. MCGOUGH. Pressure-control and pressure Butterworths, 1987, p. 50.
support ventilation: flow patterns, inspiratory time, and gas distri- 17 . ROSSI, A., S. B. GO?TFRIED, B. D. HIGGS, L. KOCCHI, A. GRASSINO,
bution. Respir. Cure 33: 126-134, 1988. AND J. MILIC-EMILI. Respiratory mechanics in mechanically venti-
10. CAMPBELL, D., AND J. BROWN. The electrical analog of the lung. lated patients with respiratory failure. J. Appl. Physiol. 58: 1849-
Br. J. Anuesth. 35: 684-693, 1963. 1858, 1985.
11. EPSTEIN, M. A., AND R. A. EPSTEIN. Airway flow patterns during 18. SMITH, T. C., AND J. J. MARINI. Impact of PEEP on lung mechan-
mechanical ventilation in infants: a mathematical model. IEEE ics and work of breathing in severe airflow obstruction. J. Appl.
Trans. Biomed. Eng. 26: 299-306, 1979. Physiol. 65: 1488-1499, 1988.
12. GUREVITCH, M. J., J. VANDYKE, E. YOUNG, AND K. JACKSON. Im- lg. THARRATT, R., R. ALLEN, AND T. ALBERTSON. Pressure controlled
proved oxygenation and lower peak airway pressure in severe adult inverse ratio ventilation in severe adult respiratory failure. Chest
respiratory distress syndrome. Treatment with inverse ratio venti- 94: 755-762,1988.
lation. Chest 89: 211-217, 1986. 20. WOOD, L. D. H., L. A. ENGEL, P. GRIFFIN, P. DESPAS, AND P. T.
13. INGRAM, R. H., AND T. J. PEDLEY. Pressure-flow relationships in MACKLEM. Effect of gas physical properties of flow on lower pulmo-
the lung. In: Handbook of Physiology. The Respiratory System. Me- nary resistance. J. Appl. Physiol. 41: 234-244, 1976.

Downloaded from journals.physiology.org/journal/jappl (123.116.066.080) on July 9, 2020.