Effects of Frequency, Tidal Volume, and Lung Volume

on CO2 Elimination in Dogs

by High Frequency(2-30 Hz), Low Tidal Volume VentilationARTHURS. SLUTSKY, ROGERD. KAMM,THOMASH. RosSING, STEPHENH. LORING,

JOHNLEHR, ASCHERH. SHAPIRO, ROLANDH. INGRAM, Jr., andJEFFREY M. DRAZEN, Department of Physiology, Harvard School of PublicHealth; Department of Mechanical Engineering, Massachusetts Institute ofTechnology; Departments of Medicine, Brigham and Women's Hospital, WestRoxbury Veterans Administration Hospital, Harvard Medical School, Boston,Massachusetts 02115

A B S T RA C T Recent studies have shown that effec-tive pulmonary ventilation is possible with tidalvolumes (VT) less than the anatomic dead-space if theoscillatory frequency (f) is sufficiently large. Wesystematically studied the effect on pulmonary CO2elimination (VCO2) of varying f (2-30 Hz) and VT(1-7 ml/kg) as well as lung volume (VL) in 13 anesthe-tized, paralyzed dogs in order to examine the contribu-tion of those variables that are thought to be im-portant in determining gas exchange by high fre-quency ventilation. All experiments were performedwhen the alveolar PCO2 was 40±+1.5 mmHg. In allstudies, VCO2 increased monotonically with f atconstant VT. Wequantitated the effects of f and VT onVCo2 by using the dimensionless equation VCO2/Vosc = a(VT/VTO)b(f/fo)c where: Vosc = f x VT, VTO= mean VT, fo = mean f and a, b, c, are constants ob-tained by multiple regression. The mean values of a, b,and c for all dogs were 2.12 x 10-3, 0.49, and 0.08,respectively. The most important variable in determin-ing VCO2was Vosc; however, there was considerablevariability among dogs in the independent effect of VTand f on VCO2, with a doubling of VT at a constantVosc causing changes in VCO2 ranging from -13 to+ 110% (mean = +35%). Increasing VL from functionalresidual capacity (FRC) to the lung volume at an airwayopening minus body surface pressure of 25 cmH20 hadno significant effect on VCO2.

Address correspondence to Dr. Arthur Slutsky, RespiratoryDivision. Brigham and Women's Hospital, Boston, Mass.02115.

Received for publication 21 January 1981 and in revisedform 19 August 1981.

INTRODUCTION

Traditional concepts of gas exchange during tidalventilation are based on the principle that there is aphysical distinction between the regions of the lungwhere gas transport is primarily convective (the dead-space volume) and regions of the lung where moleculardiffusion is the predominant gas transport mechanism(the alveolar region). This concept is supported by alarge body of experimental data as demonstrated by thesingle-breath nitrogen washout test. Following aninspiration of pure oxygen, if one plots the nitrogenconcentration of expired gas vs. volume expired, at leastthree distinct phases are evident. Initially, (phase I) theexpirate contains little detectable nitrogen and repre-sents exclusively gas expired from the dead-space. Asadditional volume is exhaled, the nitrogen concentra-tion of the expirate increases rapidly (phase II), repre-senting the interface between the dead-space volumeand the alveolar gas compartment. Finally, the endexpired gas contains the highest nitrogen concentra-tions (phase III), representing alveolar gas. If thesezones were distinct, one would predict that to achieveeffective gas exchange, the VT1 must be greater than thedead-space.

Recently, however, this traditional concept has beenchallenged by experimental observations demonstrat-

I Abbreviations used in this paper: f, oscillatory frequency;FACO2, fractional alveolar CO2 concentration; fo, mean f;HFV, high frequency ventilation; VDA, measured anatomicdead-space; VDE, measured equipment dead-space; VCO2, pul-monary CO2 elimination; VL, lung volume; VT, tidal volume;VTO, mean VT.

J. Clin. Invest. © The American Society for Clinical Investigation, Inc. 0021-9738181/12/1475/10 $1.00Volume 68 December 1981 1475-1484

1475

ing that effective alveolar ventilation can occur withVT considerably less than dead-space volume providedthat the ventilatory frequencies are sufficiently large(1-5). Although this has by now been well estab-lished by a number of studies, the physical mech-anisms that account for the observed gas exchange arenot known with certainty. Moreover, lacking from mostof these studies has been a comprehensive experi-mental data base and a clear interpretation of whatthe data signify with respect to possible physicalmechanisms that could account for gas exchange.

In a previous short publication (5) we presentedexperimental data showing that the most importantvariable determining the effectiveness of high fre-quency ventilation (HFV) with sinusoidal oscillationsof small tidal volumes is the flow amplitude of theoscillations imposed. We also presented a simpletheoretical model that could explain these results. Inthe present study we extend this previous work interms of a much larger data base and present results ofHFVperformed at different lung volumes.

METHODS

High frequency oscillatory circuit. Our experimentalsetup is shown in Fig. 1. It consists of a high frequencyventilator, a large animal ventilator (Harvard Apparatus,Ealing Corp., Natick, Mass.), a servo-controlled vacuumsource, a screen pneumotachograph connected to a pressuretransducer, a high impedance bias flow system, and a carbondioxide analyzer (LB-2 Beckman Instruments, Inc., Fullerton,Calif.). Weused two different methods of generating the highfrequency oscillations. The first was a unit consisting of four12-in. Diam, coated loudspeakers sealed in a special chamberand acoustically coupled in series. These speakers were drivenby an amplified signal from a sinewave generator. The system

allows the speakers to generate greater pressure than if we hadused a single speaker as the oscillating source. Whenall fourspeakers were driven in phase, it was possible to generatevolumes of - 100 ml at pressures of ±25 cm H2Oat a frequencyof 25 Hz. This setup allowed us to easily vary independentlythe frequency (f) and VT of oscillation for the experimentsin which (a) we maintained a constant frequency-tidalvolume product (Vosc = f x VT) at various VT (series A,below) or (b) we varied VT at constant frequency (series C,below). To achieve greater reproducibility of VT in those ex-periments in which we varied frequency at a constant VT(series B, below), we used a custom made volume displace-ment piston pump. This pump has been described in detailpreviously (6).

The air flow output of the speaker high frequency ven-tilator was measured by a 7 sq in. 400-mesh screen pneumo-tachograph connected to a pressure transducer (ValidyneDP-45 Validyne Engineering Corp. Northridge, Calif.:2 cm H2O). Since this combination is known to exhibit abroad resonance peak near 80 Hz (7) it was necessary tocalibrate the pneumotachograph-transducer combinationusing a motor driven piston pump. A calibration curve wasdetermined for frequencies ranging from 3 to 30 Hz. Weusedpump stroke volumes between 20 and 100 cm3 and foundthat for peak flows <10 liter/s, the calibration curve wassolely dependent on frequency and independent of VT. Thiscurve was unchanged by applying elastic loads (2- and 10-literbottles) to the distal end of the pneumotachograph or by apply-ing steady pressures in the bottles of up to 50 cm H20. For theexperiments in which we used the piston pump as the highfrequency ventilator, we did not use the pneumotachographtransducer combination.

Weprovided a steady supply of fresh air at 0.75 liter/s via aneedle valve, leading from a compressed air source to a tap inthe tracheal cannula. Another tap in the cannula connectedto a high vacuum source was located directly across from thisair source. During an experiment we could adjust the positiveand negative pressure needle valves to maintain the meanpressure at the airway opening to within 0.1 cm H2O ofatmospheric pressure. To ensure that the entire flow measuredby the pneumotachograph entered the dog, we checked it

COMPRESSEDAIRSOURCE

ROTAMETER

HIGH VACUUMSOURCE

FIGURE 1 Schematic diagram of experimental setup.

1476 Slutsky, Kamm, Rossing, Loring, Lehr, Shapiro, Ingram, and Drazen

initially with the bias flow on and subsequently with the biasflow taps closed off. No effect of the bias flow was foundon the measured oscillatory flow in the range of 2-30 Hz. As anadditional check, we placed an elastic load distal to the biasflow (in place of the dog in Fig. 1) and oscillated this loadwith a piston pump. The pressure oscillations within the elasticload were unaffected by the bias flow. Thus, the flow meas-ured by the pneumotachograph or generated by the pistonpump was the same as the flow entering the dog.

We used CO2 output as an index of the effectiveness ofpulmonary gas exchange. During a short burst of HFV of thedog we measured the percentage of CO2 in the gas in thetubing between the oscillator and the bias flow and foundthat the CO2concentration was <0.02%. Thus, small amountsof gas leakage into the speaker chamber could account for anerror of <2-4%. Therefore we assumed that all the CO2 fromthe dog was purged by the bias flow. Wecalculated CO2out-put by continuously drawing a small sample of gas from thebias flow, through the CO2analyzer, and electrically integratedthe CO2analyzer output neglecting an initial transient. Sincethe bias flow was constant, the integral of the CO2 con-centration in the bias flow over time was proportional to CO2output. Since we required precise measurements of CO2 atvery low CO2 concentrations (Cco,) we tested the accuracyand linearity of the system in the following manner. Aftercalibrating the CO2 analyzer, we measured the Cco2 using a0.80% CO2 test gas and electrically integrated this value overtime. We then performed serial dilutions of this test gas toobtain a CO2gas mixture with a concentration of 0.024% andmeasured Cco, and the integral of Cco2. We found that al-though the absolute value of Cco, varied due to analyzernoise, the time averaged value determined from the integral ofCco2 was within 4%of the true value. To confirm that duringan experimental run of HFV the oscillations did not intro-duce an artifact, we measured the integral of the CO2 con-centration from the bias flow using a constant CO2 sourcein place of the dog while oscillatory frequency and strokevolume were varied and found this integral to be constant. Indogs 3-6, whose data were presented in part in our previouspaper (5), and in dogs 8 and 9, we corrected the integral of CO2concentration for a small systematic error (7 ml/min) in our CO2analyzer calibration. This problem was eliminated for theexperiments in dogs 10-16.

Wefound that changes in the fractional alveolar CO2 con-centration (FACO2) resulted in a change in VCo2 at any Vosc.When the Vosc was such that the VCo2 was approximatelyequal to the metabolic CO2 production, the measured end-tidal PCO2(PetCO2) before and after HFVwere identical. Atthe higher values of Vosc, VCO2 was greater than themetabolic CO2 production and there was a fall in FACO2.Thus, as time progressed VCo2 continued to decrease. Thiswas readily apparent on the tracing of integrated CO2produc-tion, which exhibited a decreasing slope with time, and wasaccompanied by a fall in PetCO2 after the period of HFV. Tocorrectly estimate the effectiveness of HFVunder these condi-tions we examined the CO2 output until -10-20 ml of CO2was removed. This corresponded to a minimum integrationtime of -6-10 s depending on the value of VC02. At the verylow values of Vosc, the oscillations were ineffective at re-moving C02; thus the value of PetCO2 rose by an amountrelated to the CO2 production of the dog, or an average of-2-3 mmHg over a 30-s period. Weconfirmed this by meas-uring PetCO2 after each HFVrun and found that it changed by<3 mmHg when HFV was relatively ineffective. In thiscircumstance the errors introduced by assuming a constantFACO2instead of a changing FACO2, during the course of theoscillations would be small and would be in the direction ofoverestimating VCo2 at very low levels of Vosc. The

magnitude of this error would be roughly proportional to theaverage percent increase in PetCO2. If HFV was totallyineffective and PetCO2 increased by 3 mmHg in a linearfashion over the 30-s period of HFV, the average increase inPetCO2 would be 1.5 mmHg. This would thus introduce anerror of about 1.5 divided by 40 (=PetCO2) representing anerror of <5%.

Experimtiental protocolWestudied 13 mongrel dogs (body wt 9.8-22.5 kg) anesthe-

tizedl with intravenous sodium pentobarbital (initial dose 50mg/kg) and supplemented every hour with 10% of theinitial dose. A tracheostomy was performed and the tracheacannulated with a tight-fitting metal tube that contained apressure tap. The cervical vagi were isolated and cut tominimize variations in airway tone. Airway opening pressureminus body surface pressure was measured with a Hewlett-Packard 268B transducer (Hewlett-Packard Co., Palo Alto,Calif.). Catheters were placed in the femoral artery and veinfor sampling arterial blood and intravenous administration ofsolutions and drugs, respectively. Sodium bicarbonate wasgiven intravenously to maintain arterial pH between 7.25 and7.45. The dogs were ventilated with a conventional pistonventilator at frequencies from 10-15 breaths/min and thenparalyzed with a loading dose of succinylcholine (2 mg/kg)and a continuous infusion of 50 ,tg/kg per min. Under theseconditions respiratory CO2 output was determined asdescribed above.

In six dogs (3-6, 8, 9) the anatomic dead-space was meas-ured using a modification of the Fowler technique with CO2asthe test gas. In these measurements, CO2concentrations weredetermined just proximal to where the cannula entered thetrachea. This technique has been described in detail pre-viouslv (8). The mean of three or four determinations wasused as an estimate of the anatomic dead-space. The equip-ment dead-space, from CO2sampling site to the bias flow out-put, was determined by direct measurement in all ex-periments.

Measurements performned at FRC. After these deter-minations the ventilator was adjusted to maintain the end-tidal CO2between 38.5 and 41.5 Torr and base line measure-ments of CO2 output were made. Then, the animal wasventilated with the high frequency ventilator for 15-30 s,during which time measurements of VCO2were made. Aftereach measurement, the dog was returned to standard tidalventilation, and the end-tidal CO2 concentration was meas-ured. If the CO2 concentration had changed, the rate of theconventional ventilation was increased or decreased transientlyto allow the end-tidal CO2 concentration to return to its base-line value.

Using this protocol we examined the effectiveness of HFVat two to four fixed values of Vosc in dogs 3-6 over the rangeof frequencies from 4 to 28 Hz by decreasing VT as f increasedsuch that the product f x VT was held constant. Short burstsof HFVwere interspersed with normal tidal ventilation suchthat PetCO2 remained within its specified limits. This set ofexperiments will subsequently be referred to as series A. Inaddition, in these dogs (3-6) we performed further experi-ments in which we varied f and VT over a wide range of valuesbut in which Vosc and VT were not systematically heldconstant.

We examined the effectiveness of HFV in seven dogs(10-16) by using the piston pump with fixed stroke volumesranging from 3 to 5 ml/kg over the range of frequencies, 2-30Hz. These tests will be referred to as series B.

Measurements at VL above FRC(series C). Westudied theeffectiveness of HFV at different VL in two dogs (8, 9) by

C02 Elimination by High-Frequency Ventilation 1477

placing the animals in a rigid chamber, similar to an iron lung,in which the mean pressure was controlled by a servo-feed-back mechanism. By changing the pressure within thechamber we could change and control VL. Wemeasured CO2output at a number of different values of Vosc, by varying theVT of oscillation (dog 6: 8-30 ml; dog 9: 8-76 ml) but keepingthe frequency constant at 16 Hz. We made measurementsat four to six different VL: FRC and the VL correspondingto airway opening minus body surface pressures of 5, 10, 15,20,or 25 cm H20. A similar protocol was followed in that shortbursts of HFV were interspersed with conventional tidalventilation.

RESULTS

The weight, anatomic dead-space (VDA-where meas-ured), and equipment dead-space (VDE) for each of thedogs are given in Table I. Mean systemic arterialpressure was not significantly different for the experi-ments performed during the conventional tidal ventila-tion and the high frequency ventilation. However,there was a diminution in the respiratory variations inpressure as noted by others (4).

In Fig. 2, results obtained in series A for fixed Voscare presented for dogs 3 and 6. These two were chosensince the results for dog 3 did not show any substantialeffect on VCO2as f was increased and VT decreasedsuch that Vosc remained fixed, whereas dog 6 showedthe greatest such effect observed in series A.

In Fig. 3 results obtained in series B for fixed VT arepresented for dogs 12 and 15. These two were chosensince, as will be seen, dog 12 showed the smallest effectof VT at a fixed Vosc while dog 15 showed the largestsuch effect (as determined from values of b-c asdescribed below). The results for dogs 3, 6, 12, 15 arepresented in Fig. 4 as VCO2vs. Vosc. In this form, thedata fall in relatively narrow bands; however, there is atendency in all the dogs for the results obtained atthe larger values of VT to lie above the results for thelower VT. Although there does not appear to be much ofa difference between dogs 12 and 15, with regards tothe effect of VT, based on the results when plotted asVCo2 vs. f (Fig. 3), the plots of VCO2 vs. Vosc(Fig. 4) clearly show that dog 15 exhibits a greater VTdependence. Note that there are more data pointspresented for dogs 3 and 6 in Fig. 4 than in Fig. 2 be-cause in addition to the data obtained in series A, weperformed further experiments (as described pre-viously) in which we independently varied f and VTwithout aiming for a fixed Vosc.

The effects on VCo2 of varying VL (series C) areshown in Fig. 5. For both dogs at a fixed oscillationfrequency of 16 Hz, VCO2 increased as VT of oscilla-tion increased, but VCO2 remained relatively inde-pendent of VL.

Thus, although the product of f x VT is the main

TABLE IHFVData Summary

Dog VCOJ/Vosc ano. Weight VDA VDE VTO fo (X 10-3) (X 10-3) b c r n b -c

kg ml ml ml Hz

3 22.5 99 45 54 14.3 1.669 1.725 0.393 0.156 0.65 38 0.2374 12.2 43 39 25 18.0 1.878 2.278 1.022 1.231 0.75 12 -0.2095 11.2 45 39 41 14.0 1.210 1.219 0.513 0.060 0.67 51 0.4536 9.8 51 39 45 14.6 1.214 1.368 0.600 0.487 0.69 23 0.1138 15.0 62 45 18 16.0 2.486 2.575 0.496 0.98 519 14.8 86 45 35 16.0 2.975 2.953 0.039 0.94 65

10 13.2 39 37 13.5 1.603 1.485 - -0.236 0.92 1411 11.0 39 39 13.3 2.644 2.533 0.423 -0.154 0.93 26 0.57712 13.2 39 53 14.9 2.313 2.220 0.186 -0.139 0.79 47 0.32513 12.0 39 49 13.1 2.209 2.091 0.211 -0.277 0.93 20 0.48814 14.5 45 57 14.4 1.656 1.614 0.490 -0.090 0.97 24 0.58015 12.2 39 49 12.3 2.426 2.284 0.909 -0.158 0.91 39 1.06716 15.2 - 45 60 14.5 3.223 3.190 0.612 -0.046 0.90 47 0.658

Mean 13.6 64 41 43 14.5 2.116 2.118 0.491 0.076 0.85 35 0.429SD 23.2 3.0 12.4 1.5 0.638 0.610 0.270 0.439 0.12 16.5 0.344

Summary of data for all dogs based on equation 1:

Vco2 / VTb / f\ c

'Vosc VT. fo

where VT. and fo are the mean VT and frequency used for each dog, and the units of VCo2 and Vosc (= f x VT)are identical. VDA = measured anatomic dead-space; VDE = measured dead-space; r = correlation coefficient;n = number of data points; a, b, c = best fit numerical values for equation 1.

1478 Slutsky, Kamm, Rossing, Loring, Lehr, Shapiro, Ingram, and Drazen

determinant of C02 elimination, f and VT by themselvesare of some importance. To quantitate the effects ofVT and frequency individually, we fit the data obtainedfrom the experiments on each dog to a dimensionlessequation of the form:

VC02 (VT b f )-

Vosc VTO/ fo/(1)

where (in contrast to Figs. 2-5) VC02 and Vosc areexpressed in same units, VTO and fo are the mean VTand f used for each dog (Table I), and "a, b, c," areconstants determined by least squares regressionanalysis of each data set to this equation. Note thatsince the dimensions of VC02 and Vosc are identical inthis expression, the parameter "a" is dimensionless.This equation can be rewritten in alternative formsby substituting the quantity f = Vosc/VT into equation 1.

VC02 = a (-) (-) Voscl+cVTbc

- VTO)b()foc)-

A A

0

00

200T 12

ISO-

looy

50t

0

(2a)

(2b)

Vosc [liter/s]A @0.53

0 0.78A 1.05

00

AA

eo0A A0

A 0A 0

Aa 00

0A 0

A0 0

A0 00

A0 00 0

AO

o0io~~~~

15

A

A

A CA 0

A 0

A. 0.j000

0

0 52.9A 66.0

A

A

AA

o

A 0

0

0 VT IflI.J0 036.6

48.8A 61.0

0 15 20 25FREQUENCY[Hz]

FIGuRE 3 Plots of VCO2vs. fat constant values of VT for dogs12 (top) and 15 (bottom).

.50t 0

00

25tc

E

cu

501

60

0

0

0

25f

00

0 o

* 0

0

5 10 15 20FREQUENCY[Hz]

FIGuRE 2 Plots of CO2elimination (VCO2) vs. frfixed values of Vosc (=f x VT) for dogs 3 (top) a]

Table I shows for each dog the best-fit values of a,b and c. The mean correlation coefficient (r) for all thedata was 0.85 and the mean value of b - c was 0.429.Note that for any given value of Vosc the exponentb - c is an index of the independent effects of VT and fon VC02. For example, if b - c = 0, then VCO2is deter-mined only by Vosc and does not depend upon the

Vosc [liter/si individual values of either VT or f. If, however, b - c is00.46 non-zero, then VCO2will depend on both Vosc and VT,00.63 or Vosc and f. Suppose that b c = 0.5 and a given

tracheal flow, Vosc, is achieved with two VT, one twicethe other. The same VC02 would not be achieved eventhough the Vosc were equal. The VC02 with thehigher VT would be greater by a factor of 20°5, or -41%.

Using this type of analysis, we found that the effect ofdoubling VT while maintaining the same Vosc variedconsiderably from dog to dog. The minimum sucheffect on VC02 was in dog 4 (-13%), whereas the maxi-mumeffect was in dog 15 (110%). The mean effect for

25 30 all dogs of doubling VT at a fixed Vosc was -35%.To determine the effect of VL on VC02 we fit the data

from dogs 8 and 9 to the equation:equency ( t) atnd 6 (bottom). VCO2= ,BO + f81VOSC + 132VL + I83VOSCVL (3)

CO2Elimination by High-Frequency Ventilation

125f 3

loot

751

v-

75;

1479

3

0

0o0

c, *, c

.,*OA*

,. ..0

0 l00 6

75t

50

VT [ad]*s300 30-60A 60-75o 75-100

25

* 0

0.512

Ol

GAgo.IAS0.,

AS(,.A*

.,.A.

0.,.

.-

0.5

1.0 1.5A

0 AA0 0 A

0

A 250

200

1501

o00tVT[-'I]

036.1052.9*A66.0

5so

x L

xN

x

A0 K

x A&0

00

x

x 0ow

VT [mi]es 15olS-30&30-40040-6x 50-75

0.25 0.50 0.7515

0

0

A0A

A.O0

1.0 1.5 O.5Vosc [Iliter/s]

AaA

00

0

FiGuRE 4 Plots of VCO2vs. Vosc for dogs 3,6, 12, and 15. Each symbol represents a certain rangeof VT for dogs 3 and 6 and a specific VT for dogs 12 and 15.

where: VCO2is the observed CO2elimination for eachvalue of Vosc and VL and the i3i (i = 0, . . . 3) arecoefficients determined from least squares regression

analysis. The estimated values of /2 and /3 were notstatistically different from zero (P > 0.05) indicatingthat VL had no independent effect on VCO2.

'°°r 8 9 ;A

2001751

5010

S0. U

O 11

Ox

A

04m0

mxx

0

25k

0.1 0.2 0.3

1.501

* FRC0 FRC+ 5cm1420

AFRC+I Ocm H20*FRCl15cm H20xFRC+20cm H20

o FRC+25cm H20

0.4 0.5

1001

50F

0 0.25 0.50 0.75 1.00 1.25

Vosc (liter/s) Vosc (iter/s)FIGURE 5 Plots of VCO2vs. Vosc at different VL corresponding to FRCand the VL correspondingto FRCplus airway opening minus body surface pressures of 5, 10, 15, 20, and 25 cmH2Ofor dogs8 (left) and 9 (right). The frequency for all experiments was 16 Hz.

1480 Slutsky, Kamm, Rossing, Loring, Lehr, Shapiro, Ingram, and Drazen

200

150v

loot

E

0

u

501

v1

200

150

100[

50 v7(*iJ

o 36.6

A 61.0

1.0 1.5

c

E

N

0

~~~~~~~~~~~~~xil°~ ~~(f*x

0

1A6A 0

A0

o0

250 r

x

300r

E 200

E

*>10

2.0

Vosc (Iiter/s) Vosc (Iiter/s)FIGURE 6 Mean results for VCO2vs. Vosc at constant values of frequency (left) and stroke volume(right). These results were obtained using equation 1 and the mean values of the coefficients a, band c from Table I.

A plot of VCO2 vs. Vosc based on the mean regres-sion results for all dogs is presented in Fig. 6 for constantvalues of VT and f.

DISCUSSION

Previous studies of HFV. Although a number ofinvestigators have used rapid oscillations to achievegas exchange (9-11), only recently has it been demon-strated that rapid oscillations with volumes less thanthe anatomic dead-space volume could effect adequategas exchange (1-5, 12). Lunkenheimer and colleagues(1) while investigating the effect of high frequencyoscillation on cardiac performance, fortuitously foundthat they were able to maintain normal blood gases indogs without standard tidal ventilation by usingventilatory frequencies of up to 40 Hz and VT as low as10-20 cm3. Similar observations were also provided byBunnell et al. (2) and Custer et al. (3) who, using anasymmetric flow pulse at frequencies of 4-6 Hz, wereable to maintain normal blood gases before and afterinduction of diffuse lung injury with alloxan. Bohn etal. (4) also found that HFVwith VT equal to 25% of thedead-space could produce normal blood gases and sug-gested that there appeared to be a frequency optimum at-15 Hz that provided the most effective CO2elimina-

tion. In a recent report we also demonstrated that HFVcould support eucapnic gas exchange in anesthetizeddogs (5).

Although the observations mentioned clearly docu-ment the effectiveness of HFVwith small VT in achiev-ing gas exchange, a comprehensive experimental database and an understanding of the physical mechanismsresponsible for the gas exchange with this techniqueare necessary before the full potential of this mode of

ventilation can be realized. In this paper we present alarger data base exploring the effects of systematicvariations of VT and oscillation frequency on gas ex-change. Further, we explore the effects of changes in VLon gas exchange achieved by HFVin normal dogs. Withregard to all of these observations, we provide aunifying analysis of these results and highlight severalpotentially important mechanisms.

Comparison with previous results. The new andimportant feature of our experimental apparatus thatallowed us to obtain quantitative data relating VCO2to f, VT, and Vosc was the very high impedancebias flow circuit. This permitted large quantities offresh air to be delivered to the airway opening withoutloss of oscillation volume. In this regard, our experi-mental setup was substantially different from that ofBohn et al. (4) who found that at a fixed stroke volumeof the piston pump, arterial PCO2 passed through aminimum at 15 Hz, thus suggesting an optimum fre-quency for CO2elimination. To deliver fresh air to theanimal they used a flow of fresh gas delivered through along tube that acted as a low pass filter. Consequently,the stroke volume produced by their piston ventilatorwas not necessarily the same as the volume deliveredto the experimental subject since a variable fraction ofthe stroke volume would be lost to the bias flow, thefraction depending upon the ratio of the subject'simpedance to that of the bias flow. Both of theseimpedances are functions of the frequency of oscilla-tion. Thus, it is possible that as frequency increasedabove 15 Hz in the experiments of Bohn et al. (4), thefraction of stroke volume lost to the bias flow increasedsuch that the flows actually delivered to the dogs de-creased.

In a previous paper (5), we suggested that the effec-

CO2Elimination by High-Frequency Ventilation 1481

tive gas exchange during HFVmight be due to a num-ber of augmentive mechanisms including (a) Taylorlaminar (13) and turbulent dispersion (14), (b) mixingdue to asymmetrical velocity profiles as proposed byHaselton and Scherer (15) and (c) secondary flows atbifurcations. One of our predictions from this modeland that of Fredberg (16) was that VL should have rela-tively minor effects on VCo2 at any given trachealflow. The reason for this is that increasing VL producestwo competing phenomena: (a) the increasing cross-sectional area of the airways tends to increase VCO2but (b) at any given tracheal flow as cross-sectional areaincreases, the velocity in any airway decreases and thusmay lead to decreased enhancement of the mixingprocess. Our results provide evidence in support of thisprediction since we found no significant effect of VL(Fig. 5) on VCo2 as a function of Vosc. It should benoted that all the experiments were performed at VL ator greater than FRC, in which presumably all airwayswere open. It is possible that if experiments had beenperformed at VL below FRCand if airways were closedat these volumes, VCO2 might decrease. As well, atlower VL, flow limitation might occur and also furtherdecrease VCo2.

Our previous observations suggested that the mag-nitude of Vosc achieved during HFV was the criticalfactor in determining CO2 elimination. The currentexperimental data base indicates that although Voscis the most important factor, VT and f have independenteffects on VCO2 at any given Vosc. This can be ap-preciated by comparing the data presented in Figs. 3and 4. When the data are shown as VCO2 as a func-tion of frequency at a given VT the slope of the nearlylinear VCO2vs. f relationship obtained varies in mag-nitude as much as the VT varies, dog 12, or more thanVT varies, dog 15. Thus in Fig. 4 the VT dependenceis minimal when plotted as a function of Vosc in dog 12and is quite striking in dog 15. Using equation 1 wequantified the effects of VT and f and found that, on theaverage, a doubling in VT at a given Vosc resulted in anincrease in VCO2of -35% (=20.429; 0.429 is the meanvalue of b - c). There is a relatively large variability inthis effect, with some dogs showing a 13% decreasein VCO2with a doubling in VT (dog 4) and others show-ing - 100% increase in VCO2. In all but one of the dogs,the value of b - c was positive, indicating that at agiven Vosc, greater VCO2 is achieved with the com-bination of larger values of VT and lower values of f.Clearly, this must be the case when VT exceeds thedead-space volume. The values of the parametersestimated for equation 1 were obtained using datafrom experiments where f varied from 2 to 30 Hz andVT was always less than the estimated total (anatomicplus equipment) dead-space. As VT gets very small,approaching the volume of a single generation of thetracheobronchial tree, one might anticipate that gas

mixing efficiency might decrease even more. However,we do not have data to address this point becausewe did not systematically study VCO2 with verysmall VT.

This VT effect on CO2elimination was not predictedby our previous model (5) and most likely relates to thebulk transport of gas both at the airway opening and atthe junction of the dead-space and the alveolar region.This convective process would tend to purge the equip-ment dead-space of CO2 and enhance gas transport atthe alveolar-dead-space interface in a more efficientmanner than augmented transport, the only mechanismconsidered in our previous analysis. Thus, greater VTwould cause greater VCO2 at equivalent valuesof Vosc.

In an attempt to provide an estimate of the effect dueto convective purging, we modified our previoustheoretical analysis (5) by assuming that the diffusionalresistance is reduced to zero in two regions, oneadjacent to the bias flow and the other adjacent in thealveolar region. The volume of each of these two re-gions is taken to be one-half of the VT. Wethen appliedthe equations used previously (equations 1-5 of ref-erence [5]) to obtain estimates of VCO2. Theoreticalresults using this formulation are presented in Fig. 7.These theoretical results provide an approximateupper bound for the purging effect and show thatconvective purging could represent a reasonable mech-anism causing the VT effect. Further analyses will berequired to determine the exact nature of the effect.

In estimating VCO2 in our experiments we used arelatively short period of HFV(<30 s), so that we wouldbe able to obtain a large number of measurements forany given dog over a relatively short period of time.Our assumption was that data collected in this mannercould be extrapolated to predict the values of f and VTthat would be required to maintain eucapnia in thedogs during steady-state ventilation. To investigate thisquestion, in three of the dogs, we used the estimatedvalues of f and VT that would be required to producea VCO2 equal to the dogs' measured metabolic CO2production and ventilated the dogs for 1-2 h. Wemeas-ured arterial blood gases at the end of this period andfound arterial PCO2values were between 38 and 42.This not only confirms the previous findings of Bohnet al. (4), but also demonstrates that the experimentalvalues we determined for VCO2over 30 s reflect thesteady-state values that would exist over long periodsof HFV.

The experimental data presented in this paper, to-gether with our earlier experimental and theoreticalfindings (5) provide the rule that VCO2is approximatelyindependent of lung size and dependent mainly onVosc. If we use (Table I) the mean value of VCO2/Vosc for all 13 dogs, namely, 2.12 x 10-3, (where VCO2and Vosc have the same units), we can estimate the

1482 Slutsky, Kamm, Rossing, Loring, Lehr, Shapiro, Ingram, and Drazen

100 60

20

0 0.5 1.0 1.5 2.0

Vosc [liter/s]FIGURE 7 Theoretical results for VCO2vs. Vosc at various VT using our previous analysis (5) withthe addition of convective purging at the airway opening and alveolar regions (see text for details).

tracheal flows required to produce eucapnia in a givendog. For example, if the metabolic CO2production of a

dog is about 5 ml/min per kg (17), then one wouldpredict that a value of Vosc of -0.8 liter/s would berequired for eucapnia in a 20 kg dog. Applying the same

data to humans, where the metabolic CO2productionis about 3 ml/min per kg (18), one obtains a Vosc of

1.7 liter/s for eucapnia in a 70 kg man. This is, of course,a predicted value, to be assessed by clinical trials inhumans. Since these estimates are based on the averagevalue of VCO2/Vosc, it is expected that they would bemost accurate at the mean frequency used in our ex-

periment, 14.5 Hz. If the frequency of oscillation were

much different than 14.5 Hz, then more accurate esti-mates of Vosc would be obtained using equation 1 withthe appropriate mean values of the constants for thisequation from Table I. These estimates are probablysensitive to the specific geometry of the experimentalsetup. A further consideration in applying these datais the fact that the oscillatory peak flow may be near

or above flow limitation in many lungs. For example,the calculated Vosc of 1.7 liter/s corresponds to a

peak flow rate of 5.3 liter/s and not all patients wouldbe able to support this flow rate at FRC.

The potential benefits of HFV in the clinical settingare substantial. The mean tracheal pressure duringHFV may be maintained at zero and the peak trans-pulmonary pressure is low. Thus, the detrimental ef-fects of positive pressure ventilation, such as baro-trauma and impaired hemodynamics, may possibly bediminished or eliminated. In this paper we havepresented a relatively comprehensive experimental

data base relating CO2elimination to the various vari-ables thought to be important during HFV. Our dataclearly demonstrate that HFVis an effective method ofachieving pulmonary ventilation and that the magni-tude of the effective ventilation achieved varies withthe amplitude of tracheal volume flow rate. It is hopedthat these results will provide a framework on whichto base further work, with the aim of making this tech-nique clinically useful in humans.

ACKNOWLEDGMENTS

Wethank Dr. 0. McLetchie for statistical advice.This work was supported in part by Medical Research

Council (Canada), Parker B. Francis Foundation Fellowship(THR), grants from the National Heart, Lung, and BloodInstitute HL00549, HL26566, and HL22920 and the VeteransAdministration.

REFERENCES

1. Lunkenheimer, P. P., W. Rafflenbeul, H. Keller, I. Frank,H. H. Dickhut, and C. Fuhrmann. 1972. Application oftranstracheal pressure oscillations as a modification of"Diffusion Respiration". Br. J. Anaesth. 44: 627.

2. Bunnell, J., K. H. Karlson, and D. C. Shannon. 1978. Highfrequency positive pressure ventilation in dogs and rab-bits. Am. Rev. Respir. Dis. 117 (Suppl.): 289a.

3. Custer, J. R., and D. C. Shannon. 1979. High frequencyventilation enhances oxygenation and alveolar ventilationat reduced airway pressures in animal model of pul-monary edema. Am. Rev. Respir. Dis. 119 (Suppl.): 226a.

4. Bohn, D. J., K. Miyasaka, B. E. Marchak, W. K. Thompson,A. B. Froese, and A. C. Bryan. 1980. Ventilation byhigh frequency oscillation. J. Appl. Physiol. Respir.Environ. Exercise Physiol. 48: 710-716.

CO2Elimination by High-Frequency Ventilation

I

cJ.E"-E

0" 100

50

0

14843

5. Slutsky, A. S., J. M. Drazen, R. H. Ingram, Jr., R. D. Kamm,A. H. Shapiro, J. J. Fredberg, S. H. Loring, and J. Lehr.1980. Effective pulmonary ventilation with small volumeoscillations at high frequencies. Science (Wash. D.C.).209: 609-611.

6. Finucane, K. E., and J. Mead 1975. Estimation of al-veolar pressurp during forced oscillation of the respira-tory system. J. Appl. Physiol. 38(3): 531-537.

7. Jackson, A. C., and A. Vinegar, 1979. A technique formeasuring frequency response of pressure volume andflow transducers. J. Appl. Physiol. Respir. Environ.Exercise Physiol. 47(2): 462-467.

8. Drazen, J. M., S. Ii. Loring, A. C. Jackson, J. R. Snapper,and R. H. Ingram, Jr., 1979. Effects of volume historyon airway changes induced by histamine or vagalstimulation. J. Appl. Physiol. Respir. Environ. ExercisePhysiol. 47(4): 657-665.

9. Sjostrand, U. 1977. Review of the physiologic rationale forthe development of high frequency positive pressureventilation-HFPPV. Acta Anaesthesiol. Scand. Suppl. 64:7-27.

10. Sjostrand, U. 1977. Summary of experimental and clinicalfeatures of high-frequency positive-pressure ventilation-HFPPV. Acta Anaesthesiol. Scand. Suppl. 64: 165-178.

11. Venegas, J. G., and M. Venegas. 1979. Pulsos reversos depresion. Neuva alternativa para el tratamiento de las

enfermedades pulmonaras obstructivas cronicas. TribunaMedica 5: 33-42.

12. Butler, W. J., D. J. Bohn, A. C. Bryan, and A. B. Froese.1980. Ventilation by high frequency oscillation inhumans. Anaesth. Analg. 59: 577-584.

13. Taylor, G. I. 1953. Dispersion of soluble matter in solventflowing slowly through a tube. Proc. R. Soc. London Ser. A.219: 186-203.

14. Taylor, G. I. 1954. The dispersion of matter in turbulentflow through a pipe. Proc. R. Soc. London Ser. A. 223:446-468.

15. Haselton, F. R., and P. W. Scherer. 1980. Bronchicalbifurcations and respiratory mass transport. Science(Wash. D. C.). 208: 69-71.

16. Fredberg, J. J. 1980. Augmented diffusion in the airwayscan support pulmonary gas exchange. J. Appl. Physiol.Respir. Environ. Exercise Physiol. 49: 232-238.

17. Pappenheimer, J. R., A. P. Fishman, and L. M. Borrero.1952. New experimental methods for determination ofeffective alveolar gas composition and respiratory deadspace, in the anaesthetized dog and in man. J. Appl.Physiol. 4: 855-867.

18. Altman, P. L., and D. S. Dittmer, editors. 1971. Respirationand Circulation Biological Handbook. Federation ofAmerican Societies for Experimental Biology, Bethesda,Md. 78.

1484 Slutsky, Kamm, Rossing, Loring, Lehr, Shapiro, Ingram, and Drazen