migetprinciples and information content of the multiple inert gas elimination technique

Thorax 1993;49:815-824

Contribution of multiple inert gas elimination technique topulmonary medicine 1

Series editor: R Rodriguez-Roisin

Principles and information content of the multipleinert gas elimination technique

Josep Roca, Peter D Wagner

Servei dePneumologia iAl.lergia Respiratoria,Hospital Clinic,Villarroel 170,Barcelona 08036, SpainJ Roca

Department ofMedicine, Section ofPhysiology, Universityof California,San Diego (UCSD),La Jolla, California92093-0623, USAP D Wagner

Reprint requests to:Dr J Roca.

I 50-E V *E

0

C)

The chief function of the lung is pulmonary gasexchange which requires adequate levels ofventilation and perfusion of the alveoli. Thelung must match pulmonary oxygen uptake(Vo2) and elimination of carbon dioxide (Vco2)to the whole body metabolic oxygen consump-tion and carbon dioxide production whateverthe oxygen and carbon dioxide partial pressuresin the arterial blood.During the first half of this century'2 the

heterogeneity of the ventilation-perfusion(VA/Q) ratios within the lung was identified as afactor causing hypoxaemia (table). However,the importance ofVA/Q inequality in producinghypercapnia in patients with some forms oflung disease has been pointed out only morerecently.3 At the end of the second world war acrucial step towards a better understanding ofthe physiology of pulmonary gas exchange wasthe development of the three compartmentmodel of the lung almost simultaneously byFenn et alt and Riley and Cournand.5 Thegraphical analysis of this representation of thelung5 (fig 1) provided the conceptual basis forthe traditional interpretation of arterial bloodgas measurements in the clinical setting - thatis, venous admixture, physiological dead space.The progress in the mathematical description of

Po2 (mm Hg)

Figure 1 The three compartment lung: Riley's oxygen-carbon dioxide diagram.'The lung is represented by three units on the basis of the arterial and the expiredpartial pressures ox,ogen and carbon dioxide. One type of unit (dead space, I) isunperfused so its A/IQ ratio is infinity. The Po2 and Pco2 is then equivalent to those ofthe inspired gas mixture. A second type of unit (V) is unventilated so its (AI(Q ratio iszero and its Po2 and Pco2 are equivalent to those of mixed venous blood. Finally, a

third type of unit (ideal, A) presents balanced ventilation and perfusion rates (fPAI)ratzo= 1).

the behaviour of oxygen and carbon dioxide inthe blood"8 facilitated substantial contributionsto the numerical analysis of pulmonary gasexchange.9 These studies showed that tradi-tional variables derived from physiologicalgases (Pao2, Paco2, alveolar-arterial Po2 dif-ference (A-aPo2), venous admixture, andphysiological dead space) are sensitive to VA/Xmismatch, but they also vary with changes intotal lung indices such as minute ventilation,cardiac output, and inspired Po2. This be-haviour can lead to misinterpretations in theclinical setting. In the mid 1960s Farhi'°demonstrated the quantitative relation of theblood-gas partition coefficient (k) of a gas - theVA/Q ratio - and the capacity of an alveolar unitto exchange that gas. Using the principles laiddown by the work of Farhi,'° and extending theapplication of inert gas elimination by addingmore gases, the multiple inert gas eliminationtechnique (MIGET)"-13 was developed in themid 1970s. This technique provides more in-formation concerning the role of the VA/Qrelations on pulmonary gas exchange than pre-viously available. It overcomes some importantdrawbacks of other methods such as the topo-graphical radioactive tracer based techniques(ventilation-perfusion scans) which havelimited resolution and thus underestimate thedegree of VA/Q inequality. Approaches'4 sim-ilar to MIGET but using different inspiredoxygen fractions (Fio2) rather than differentinert gases as the forcing function erroneouslyassumed no effect of Fio2 on VA/Q inequality.The MIGET, in addition to its ability to

estimate VA/Q distributions in real lungs, pro-vides information on other pulmonary factors

Factors determining arterial hypoxaemia

Intrapulmonary Extrapulmonary

Main factors:VA, Q mismatching i Minute ventilationShunt ICardiac outputAlveolar-end capillary Inspired Po20, diffusion limitation T 02 uptake

Secondary factors:) P50Hb concentration

T pH

VAXQ=ventilation-perfusion; Hb=haemoglobin; P50 =Po2 thatcorresponds to 50% oxyhaemoglobin saturation.

815

Roca, Wagner

causing hypoxaemia (table) and also allows anumerical analysis of the influence of extra-pulmonary factors on arterial P02, as describedbelow.During the past 15 years a great deal of work

has been done on the theoretical basis of theMIGET (to define its limits), different modali-ties of the technique have been developed, and asubstantial amount of clinical research has beenproduced by several groups around the world.This introductory article to the series on thecontribution of MIGET to pulmonary medi-cine aims to report the essentials of the tech-nique, namely: (1) principles of measurement,(2) information content, (3) limitations, and (4)some practical details of the different modalitiesof the method. The mathematical complexitiesof the technique will be omitted as they areavailable elsewhere to the interested reader.'"-"3Finally, we will analyse the contribution of theMIGET to the understanding of the physi-ology of the normal lung and its behaviour instress situations such as exercise and altitude.

Physiological basis and assumptionsThe MIGET constitutes a conceptual andtechnical development of the historical workcarried out by Kety,'5 Farhi,'0 and others'6analysing the relations between inert gas ex-change in the lungs, the VA/4 ratio distribu-tion, and the solubilities of the gases used.These authors established the following basicequation for a single lung unit. This equationexpresses mass balance during steady stateelimination of the gas such as that which occursduring intravenous infusion of a dissolved gasin solution:

PC'PA = PV X/(X + VA/()

covering a broad spectrum of partition coeffi-cients from 0005 (SF6) to 300 (acetone) tocharacterise the distribution of the VA/Q ratioswithin the whole lung. Equation 3, describingthe events in a single lung unit, can be appliedto each lung unit and the results summed torepresent total lung gas exchange:

n =jR=Pa/Pvi= EZQJR[k/(k+VA/Q)]

n = ((equation 4)

n=jE = PE/PVT = E VAJ4X[/(X + VA/Q) ]

(equation 5)

Here, (j and VAJ represent, respectively, thefractional perfusion and ventilation of each ofthe N lung units present in the lungn=j n=j

(E i =E VAJ = 1). For each inert gas then=l n=l

retentions are calculated as the ratio betweenarterial partial pressure and mixed venous par-tial pressure (R= Pa/Pv) and the excretions asthe ratio between mixed expired partial pres-sure and mixed venous partial pressure (E= PE/Piv). By using a multicompartmental approachwith enforced smoothing, the retentions of thesix inert gases allow the estimation of a continu-ous distribution of the pulmonary blood flowagainst VA/( ratios on a logarithmic scale (fig2). Similarly, the excretions of the six inert

(equation 1)

This indicates that both end capillary (Pc') andalveolar (PA) partial pressures of an inert gas(assumed to be equal in a single lung unit'7)depend on the partial pressure of this gas in thevenous side of the capillary (Pv) and the expres-sion in the right hand side of the equation,which includes the solubility of the gas (k),expressed as the partition coefficient, and theVA/4 ratio of the lung unit. Equation 1 can berearranged as follows:

PC'/PV = PA/PV = X/(X + VA/Q)(equation 2)

The Pc'/Pv ratio indicates the r

of the gas in the blood while theexpress its excretion (E) to theThus, equation 2 can be expresselowing way:

R = E = X/(k + VA/Q)

According to this mathematicalthe retention and the excretion ofdepend only on the solubility of th4VA/4 ratio of the lung unit. The Asimultaneous venous infusion of siin trace concentrations (SF6, et]propane, enflurane, diethyl ether,

.etention (R)PA/Pv ratioambient air.

C

1.5

00 1.0

t0.

CL'a

05-

c

._4-O 30%deadspace

Shunt

1.0VA/Q ratio

d in the fol- Figure 2 Ventilation-perfusion distributions. Ventilation( 0 ) and perfusion ( * ) are plotted againstventilation-perfusion (tJAI/) ratio on a logarithmic scalein a resting young healthy subject breathing room air.

(equation 3) Both ventilation and blood flow curves are centered(equation X (first moment) around a VA/I) ratio of 1 and they are

narrow (second moment). No perfusion to low PA/dI expression, units (tPA/I ratios < 0 1 ) nor ventilation to high VAA/Qan inert gas areas (tA/Q ratios > 10) are observed. Note also the

absence of shunt. Each individual data point represents ae gas and the particular amount of bloodflow ( * ) or alveolarAIGET uses ventilation ( 0) to the corresponding pulmonaryix inert gases compartment (VA/a ratio). Total cardiac output

corresponds to the sum of the SO blood flow points andhiane, cyclo- total alveolar ventilation is the sum of the SO ventilationand acetone) points.

816

Principles and information content of the multiple inert gas elimination technique

gases provide an estimation of the distributionof the alveolar ventilation against VA/4 ratios.Because equations 4 and 5 have the term[X/(X+ VA/(Q)] in common, the blood flow andthe ventilation distributions are mathematicallyinterdependent, reflecting the same distributionof the VA/Q ratios seen from opposite sides ofthe blood-gas barrier (fig 3). The expressionenforced smoothing, in the mathematical sense,requires compartments with similar VA/4ratios to have similar perfusion. In summary,the key variables for recovering VA/Q distribu-tions from real lungs, either in normal subjectsor in patients with lung disease, are the PvT, Pa,and PE of the six inert gases, their correspond-ing partition coefficients, and VE. In the pres-ence of pulmonary and systemic arterial cath-eters, cardiac output is computed using steadystate mass balance equations as explainedbelow. The MIGET contains a number ofassumptions that can be grouped into threedifferent categories:

Healthy subject

1.0 Ether0.8 Enflurane ceto,0.6 K0.4 Cyclopropanef,1g0.2 -SF Ethane

X0.0X O-C 0.01 0.1 1 10 100 1000

C Bimodal VA/Q pattern0

._o, 1.0x(D0.8 7

([email protected]

C0.04 0.001 0.01 0.1 1 10 100 1000

Inm Shunta)

. 1.141.2 -

1.0108 ,8

0 6 0 --- - -I <

0.2 0,a0.00.001 0.01 0.1 1 10 100 1000

Partition coefficient

0 0.01 0.1 1 10 100C0

.0

35% xdead :space .2

Shunt t

0 0 01 0.1 1 10 100-C0)

15%

Iv~~~~~~~

~~~.ii9 oi:oi~~~~~~~~-

0 0.01 0.1 1

VA/Q ratio

Figure 3 Inert gas retention and excretion data versus ventilation-perfusi,distributions. Retentions (Pa/PvD) and excretions (PE/PV) for the six inertplotted against their blood-gas partition coefficient (solubility) (left panels)different types of ventilation-perfusion (tJA/0) distributions shown in the cright panels. From top to bottom: normal lung; bimodal both ventilation atdistributions (some patients with chronic obstructive pulmonary disease); ashunt (adult respiratory distress syndrome). In the retention-excretion plot,lines indicate the normal retention and excretion curves of the uppermost p,discontinuous lines represent the best fit to the measured data (dots). Notic,increased perfusion to units with low PA/l ratios (low A'A/Q areas) is esseindicated by the increase in the retentions of the more insoluble gases exceptthe decrease in the excretions of the more soluble gases except acetone indiccventilation to units with high VA/I ratios (high f7A/a areas). Retention oJexcretion of acetone are essentially the markers of the amount of shunt (%i

ratios < 0005) and dead space (% fA to tA/1 ratios > 100), respectivelj

STEADY STATE CONDITIONSThe technique requires the existence of steadystate conditions in all VA/(. units, as defined bymass balance between the amount of gasexchanged between the pulmonary capillaryblood and the alveoli, and that exchangedbetween the alveoli and the atmosphere. This isrequired for validity of equations 1-5. Thismeans that R and E must be constant during theperiod of measurement within a given experi-mental condition, though it is not necessary forabsolute partial inert gas pressures to be con-stant.

CHARACTERISTICS OF THE INERT GASESThe gases must show a linear relation betweenpartial pressure and concentration in blood(they must obey Henry's law). It is alsoassumed that diffusion equilibration occursbetween alveolar gas and end capillary blood foreach inert gas.

THE LUNG MODELThe lung is considered to consist of a number ofhomogeneous compartments arranged in paral-lel, each with constant and continuous ventila-tion and perfusion. The mathematical modelassumes that the distribution of ventilation andperfusion have regular contours and do notpresent sudden irregularities. Further import-ant assumptions are smoothing (described ear-lier) and non-negativity of perfusion.A substantial amount of research has been

done to test the principal assumptions andconstraints of the MIGET.'3 1-22 All are accept-able provided that the technique is used pro-perly, but must be kept in mind when resultsare interpreted.

Information content and limitationsThe principal objective in using the MIGET isto estimate the distribution of ventilation-per-fusion ratios. However, a number of additionalpieces of information become available frominert gas data, and all are now summarised.

dead VA/Q DISTRIBUTIONspace

This most fundamental outcome of theMIGET consists of a quantitative picture of thepresence or absence of functional lung units ofparticular VA/Q ratio. When such units are

10 100 identified as present, the amount of blood flowassociated with units of each identified VA/Q

'on are ratio is computed. The result is convenientlygases are presented graphically as in fig 2. The abscissa isfor three the VA/4 ratio, most conveniently plotted on a°orresponding logarithmic scale. The range of partition coeffi-id perfusionrnd increased cients of the six gases (from 0005 for SF6 to 300Fs continuous for acetone) allows for a similar range of VA/Qmel while the ratios to be identified. Thus, the lowest VA/Q,ntially resolved from zero is 0-005 and the highest (thatSF6,while cannot be differentiated from infinitely great) is

ite increased 100 0. Between these limits of resolution anyf SF6and arbitrary number of functional units can beL)T toeAa oy- ~~chosen, equally log spaced along the x axis.

817

Roca, Wagner

Traditionally 50 "compartments" are used inall. The ordinate consists of both ventilation(VA) and blood flow ((2) for each compartment,and the graph is then the frequency distributionof ventilation and blood flow as a function of theVA/Q ratio.This graphical representation is useful in

giving a general overview of the distribution. Itidentifies domains of interest along the VA/Qaxis (below normal, normal, and above normalVA/Q ratio), and suggests patterns of distribu-tion as unimodal, bimodal, or trimodal (fromsix gases it is mathematically impossible toresolve more than three modes).The graphical representation does not in

itself quantify the degree of inequality. To dothis we, and others, have developed a series ofconvenient parameters of the distributions ofboth ventilation and blood flow. One is the firstmoment of the distribution (on a log scale)which is simply the mean abscissa value, ormean VA/Q ratio, of each curve. More inform-ative are the second moments of the distribu-tions about their respective means on a logscale, and the square roots of these momentshave been called log SDQ and log SDv for bloodflow (() and ventilation (V) curves respectively.The letters "SD" refer to standard deviation,which is applicable only if the entire distribu-tion is logarithmically normal. In other cases(skewed or multimodal curves) there is nothingwrong with the parameter, but it is no longerstrictly reflective of the standard deviation.These parameters then quantify the amount ofVA/( inhomogeneity. This semantic issueshould not limit applicability of the parameterwhich is broad and well suited to comparingdistributions - for example, before and afterinterventions. Useful additional parameters arethe fractions of total ventilation and blood flowwithin arbitrarily defined ranges of VA/'Q ratio.Since the distribution in normal young subjectsis completely contained within the two decaderange between VA/Q=0-1 and VA/Q= 100, aconservative parameter of low VA/0 abnormal-ities is the total (fractional) blood flow in unitswhose VA/Q ratio is < 0-1. The correspondingparameter for abnormal high VA/Q regions istotal (fractional) ventilation of units having VA/0 ratios > 10-0. Still other parameters include(when more than one VA/Q mode is present) thefractional ventilation and blood flow of eachmode, shunt (blood flow in units of VA/Q ratio< 0.005), and dead space (ventilation in units ofVA/Q ratio > 100) (see below).

Parameters can also be derived directly fromthe retention and excretion data themselvesrather than from the associated computed dis-tribution. Hlastala et aP3 have used the areasbetween the retention and excretion curves asquantitative indices, and Gale et al'4 have usedrelated parameters such as the root mean squarevalue (over the six gases) of retention minusexcretion. This parameter is in effect an averagealveolar-arterial difference for the inert gases.

Rather than argue which set of parameters isbest, we prefer to use those parameters whichbest address the physiological question posed inthe particular case, and frequently many, if notall, of the above are used.

SHUNT AND DEAD SPACEThe principal strength of the MIGET is theability to identify the presence of lung unitsover a broad range of VA/Q ratios, and theabove describes information available over theVA/Q ratio range 0-005-100. Outside this rangeall perfusion in areas of very low (<0.005) orzero VA/Q (shunt or unventilated lung units) iscombined into a single parameter referred to asthe shunt. Similarly, all ventilation in units ofVA/Q > 100 including unperfused lung (VA/(infinitely high) is referred to as dead space.Shunt, when present, can be due to perfusion ofunventilated or extremely poorly ventilatedlung regions or to direct vascular communica-tions such as atrial septal defects with reverse(right to left) flow, or rarely to arteriovenouschannels in the lung. As one of the standard 50"compartments" used in the computed pro-gram of the MIGET shunt is directly esti-mated, as is flow to all other lung units. Thedistinguishing feature of shunt for the purposesof gas exchange is complete retention of all inertgases in the blood flowing through suchregions. Post-pulmonary shunt (through bron-chial or thebesian circulations) is not detectedby the MIGET.

In a corresponding manner the dead spacevalue returned by the computer algorithm re-flects anatomical (conducting airway) deadspace, together with any external instrumentaldead space and also that part of alveolar ventila-tion that reaches alveoli which are, for anyreason, completely unperfused (or which have aVA/( ratio of > 100).

It is not possible from MIGET data alone tofurther subdivide either shunt or dead spaceinto its potential components.

DIFFUSION LIMITATION OF GAS EXCHANGEThe VA/( distribution, shunt, and dead spaceare parameters based on convective gas andblood flow through the lung, and the MIGETalgorithm is based on a theory that expresslyassumes that all diffusive processes affectingpulmonary gas movement are sufficiently rapidnot to affect gas exchange. That the MIGETalgorithm can so often adequately fit experi-mental data in the face of these assumptionssuggests that under most conditions the as-sumptions are reasonable. Two forms of diffu-sion limitation are recognised as being theor-etically possible, however; (1) that associatedwith diffusion of gases between alveolar gas andcapillary blood, and (2) that associated withdiffusive mixing of inspired gas with residentalveolar gas.The former problem, when it occurs, impairs

oxygen exchange much earlier than inert gasexchange; the inert gases are in fact essentiallyinvulnerable to this form of diffusion limi-tation.2527 Consequently, under such con-ditions, application of MIGET will still estim-ate the VA/Q distribution despite oxygendiffusion limitation; however, oxygen exchangewill be affected not only by VA/( inequality butalso by diffusion limitation. When the MIGETalgorithm uses the derived VA/Q distribution toestimate the amount of hypoxaemia that VA/()

818


inequality should produce,25 27 the value of theexpected arterial Po2 will be too high (whencompared with the measured arterial Po2)because of the depression of actual arterial Po2by diffusion limitation. Thus, if the expectedarterial Po2 exceeds the measured value bymore than experimental error, the presence ofoxygen diffusion limitation is inferred. If oneassumes a uniform distribution of lung oxygendiffusing capacity to blood flow, it is thenpossible to compute the oxygen diffusing capa-city that would reconcile the measured andexpected arterial Po2 values.28 This is a usefultool when characterising impaired oxygen dif-fusive equilibration in the lung. Two otherphenomena may also cause a similar differencebetween expected and actual arterial Po2values.Firstly, significant pulmonary tissue oxygenutilisation which is hypothesised to occur, forexample, in pneumonia would result in a lowerarterial Po2 than the non-metabolised inertgases would predict. This has, however, notbeen found.29 Secondly, post-pulmonary shuntsvia the bronchial and/or thebesian circulationsmay depress actual arterial Po2 but not affectinert gas exchange, thus also resulting in ahigher expected Pao2. It can be difficult toexclude this phenomenon when expected Pao2exceeds the measured value, but the goodagreement between expected and measuredvalues almost always seen at rest suggests thatsuch shunts are generally undetectable. Duringnormoxic heavy exercise expected Pao2 oftenexceeds the measured value, suggesting diffu-sion limitation for oxygen, but the relativeimportance of diffusion limitation and post-pulmonary shunt are not definitively ascertain-able. However, logic favours diffusion limi-tation for two reasons: (1) the expected tomeasured difference becomes greater inhypoxia, and (2) the calculated post-pulmonaryshunt would (especially in hypoxia) have toexceed 10% of the cardiac output, an unreason-able value.

Impaired gas phase diffusive mixing can alsobe identified by MIGET. Here the factor thatseparates vulnerability of different gases to thisphenomenon is the gas molecular weight.Hence oxygen (MW = 32) is relatively invulner-able to this problem compared with the gasesSF6 (MW = 146) and enflurane (MW = 180)used in the MIGET. The other gases used inMIGET have molecular weights ranging from30 (ethane) to 74 (ether). Consequently, if mol-ecular weight is an important determinant ofinert gas exchange it would affect each of the sixgases differently and thus prevent the MIGETalgorithm from fitting the measured inert gasdata accurately. This is because the algorithmaccounts for the solubility of each gas butignores molecular weight. One would thereforeexpect a poor fit to the data by the algorithmand, in particular, directional errors in fittingthe data that reflected the molecular weight foreach gas. Thus high molecular weight gaseswould be retained in the blood to a greaterdegree, and low molecular weight gases to alesser degree, than would be expected fromsolubility alone. In normal animals and subjectssuch evidence for gas phase diffusive limitation

is difficult to obtain, but many years ago Truoget aPl saw this in the anaesthetised rat. Morerecently we have seen this to a minor degree inexercising pneumonectomised fox hounds." Itshould be kept in perspective, however. Suchgas phase diffusive limitation appears to ac-count for no more than a 1-2 mm Hg drop inarterial Po2, whereas VA/( mismatch and/oroxygen diffusion limitation across the blood gasbarrier often produces severe hypoxaemia.

MODEL ADEQUACYThe preceding argument indicated that gasphase diffusive limitation would produce a poorfit to the inert gas data by the MIGET algor-ithm, and this would be reflected by a highresidual sum of squares (between the datathemselves and the least squares best fit to thesedata produced by the algorithm). However, itwas pointed out that the errors for individualgases would systematically follow their molecu-lar weight.Another possibility exists to explain a poten-

tial poor fit of the model to experimental data -that is, a high residual sum of squares (RSS)due to random measurement errors. By estab-lishing the measurement accuracy of MIGETwe can estimate the variance of inert gas dataand thus compute the expected distribution ofRSS due to normal measurement errors. Datasets producing RSS values systematicallygreater than expected point to a discrepancybetween the basic physiological model de-scribed above and lungs under current study.Thus, if gas exchange is not adequately de-scribed by the conventional steady state equa-tions, either the assumptions have been violated- that is, the subject is not in a steady state asrequired - or the lung is for some reason notexchanging gas in concordance with the modelused. A good example of the latter is reportedby Powell,'233 who applied the MIGET to avianlungs. Using the standard MIGET algorithmthe RSS were clearly not compatible withexpectations; they were several fold higher.Measurements made with the same personneland equipment at the same time, but in caninelungs, did not produce abnormal RSS values.When the MIGET algorithm was changed toreflect the different gas exchange process ofavian versus mammalian lungs - that is, bymodelling cross-current rather than alveolar gasexchange" - the RSS reverted to normal for thesame data.

MEASUREMENT OF CARDIAC OUTPUTSteady state mass balance equations are thebasis of the MIGET. Thus, since the inertgases are infused intravenously rather thaninhaled they are absent from inspired gas, muchlike carbon dioxide. Consequently inert gasoutput measured by expired gas analysis (totalventilation times gas concentration) equals thatcomputed as the product of the differencebetween the pulmonary arterial and systemicarterial inert gas concentrations and cardiacoutput.

In the presence of pulmonary and systemic

819

820

arterial catheters it is thus a straightforwardmatter to compute cardiac output. In fact, eachgas gives an essentially independent estimate ofcardiac output, and thus a robust average valueof this variable can be determined. In comput-ing this average, however, it is important toproperly weigh the contribution of each gas inaccordance with conventional error propagationtheory."3

catheter in the pulmonary artery. When such acatheter is not used a modality of the MIGETwithout mixed venous sampling24 can be usedwith similar accuracy. In this instance cardiacoutput needs to be directly obtained (usingindocyanine green or perhaps one of the non-invasive techniques). Since all the necessaryvariables except Pv; are now measured, Piv canbe computed from mass balance:

QT.X.PV = QT.X.Pa + VE.PE (equation 6)INFLUENCE OF EXTRAPULMONARY FACTORS ONGAS EXCHANGEIt is well known that arterial Po2 is determinedby a combination of intrapulmonary and extra-pulmonary factors (table). The intrapulmonaryfactors include (1) the degree of VA/( mis-match, shunt and dead space, and (2) diffusionlimitation if present. The extrapulmonary fac-tors of primary importance are: (1) Fio2, (2)total ventilation, (3) cardiac output, and (4) Vo2(metabolic rate).As described above in the section on diffu-

sion limitation, it is possible to predict thearterial Po2 expected from the measured VA/(inequality and the particular combination ofextrapulmonary factors that existed at the timeofmeasurement. The MIGET algorithm, how-ever, allows the observer to change any or all ofthe extrapulmonary factors, and then to recom-pute the expected value of arterial Po2. Inaddition to the four primary extrapulmonaryvariables (table) secondary variables such ashaemoglobin concentration, haemoglobin P50,body temperature, and blood acid-base statuscan also be changed singly or in any com-bination.Such flexibility is useful in understanding not

only expected effects of therapeutic interven-tions, but also in separately determining thequantitative role of each extrapulmonary factorwhen they change between two conditions ofMIGET measurement. For example, if a vaso-dilator was given to a patient with high pulmon-ary artery pressure cardiac output may wellincrease, while at the same time the VA/Qdistribution may change for the worse. ArterialP02 will reflect the integrated effect of thesechanges. It is a simple matter to execute theMIGET algorithm (a) with the data obtainedbefore giving the drug, (b) using the post-druginert gas data with the predrug cardiac output,and (c) using all post-drug data to study theindividual influences on arterial Po2 of eachfactor. It is similarly possible to use the algor-ithm to separately assess, for example, the effectof shunt versus that of VA/Q inequality onarterial Po2.

Pv = Pa + (VE.PE/k.()T) (equation 7)This modality of the technique is also able to

AInert gasinfusion Mixed

venous (v)

(a)

B

C

Inert gasinfusion Expired (E)

/ Arterial (a)

Dye infusion(bolus) Dye dilution

measurement

Inert gas

!finfusion Expired (E)

Peripheral- svvenous(yen)

Practical considerationsThe original technique (fig 4) was describedusing inert gas levels determined from threedifferent sites: mixed venous blood (PvT), arter-ial blood (Pa), and mixed expired gas (PE). Thismodality uses direct measurements of all thevariables involved in the estimation of the VA/Qdistribution (equation 3) and provides ameasurement of cardiac output as describedearlier. However, it requires the placement of a

Figure 4 Different modalities of the MIGET. (A)Original description of the technique with arterial (Pa),mixed expired (PE) and mixed venous (P;v) samplings.In (B) only arterial (Pa) and mixed expired (PE)samplings are required. Cardiac output (OT) is thenmeasured by green dye and mixed venous partial pressures(Pv3) of the six inert gases are estimated by mass balance.It does not require the placement of a catheter in thepulmonary artery. In (C) only peripheral venous (Pven)and mixed expired (PE) samplings are done. Thismodality (see text) is useful when serial measurements inthe same subject are planned.

Roca, Wagner


estimate the predicted arterial Po2 according tothe VA/4 distributions and to test for alveolar-end capillary oxygen diffusion limitation27 ifmeasurements of Vo2 and Vco2 are obtainedsimultaneously since equations analogous to (6)and (7) can then be applied to calculate Pvio2and PiTco2.

Finally, a third modality of the MIGET34requiring only mixed expiratory and peripheralvenous sampling is also available. This isobtained from a distally oriented 20-gauge can-nula inserted into a peripheral vein (usually inthe forearm opposite to the side of the infusionof the inert gas solution). Because inert gasesare not metabolised in the tissues of the hand,and after some 90 minutes of inert gas infusionin a resting position, virtual equilibrationbetween blood and tissues is achieved and theperipheral venous blood reflects the inert gasconcentrations of the inflowing arterial blood.In this condition partial pressures of the inertgases in the peripheral vein (Pven) have beenshown to be 95% of Pa (Pa=.Pven/0 95) andreproducible enough to permit substitution ofperipheral venous for arterial sampling. How-ever, since (T is not directly measured there isstill an unknown in equation 7. Cardiac outputcan be estimated within a reasonable range onthe basis of relatively simple non-invasive pro-cedures such as Doppler echocardiography. Bynumerical analysis it has been shown that thelog SDQ and log SDv are mostly insensitive tothe uncertainty of the estimation of cardiacoutput. It must be pointed out, however, thatthis approach without arterial sampling is notuseful for estimating the first moment of thedistributions, the amount of shunt and deadspace, nor the percentage of perfusion or venti-lation at a given range of VA/Q ratios. Bycontrast, this less invasive approach of theMIGET is particularly useful in those situa-tions wherein the main focus of interest is theshape and the dispersion of the VA/(. distribu-tions, and arterial catheterisation needs to beavoided because repeated measurements overtime are planned. The results of this modalityhave been validated by two different studies inpatients with asthma using repeated measure-ments almost daily or weekly.3536

EXPERIMENTAL DETAILS AND DATA PROCESSINGThe ability to obtain accurate information usingthe MIGET requires careful and detailed tech-nical aspects. The different steps of the datacollection described below must be followedcarefully to maintain the experimental errorwithin the tolerable limits and to preserve thesteady state.

Preparation of the inert gas solution forsterile infusionA mixture of the six inert gases is equilibratedwith 1 litre of either normal saline or 5%dextrose. A tank of gas containing approxim-ately 20% SF6, 20% cyclopropane, and 60%ethane is connected to the plastic bag with thesolvent by using sterile intravenous tubing witha 0-22 gm Millipore filter inserted. Gas from the

tank is then pumped into the bag until the latteris somewhat distended. The bag is vigorouslyshaken for approximately one minute and thegas in the bag is then vented to the atmosphere.This cycle should be repeated once to reach fullequilibration between the saline or dextrose andthe tank gas mixture. Next, 0 7 ml liquiddiethyl ether/l saline is transferred into the bag,followed by the addition of 7 ml liquid acetone/l,and finally, 3 ml liquid enflurane/l. The bag isthen gently inverted several times to facilitatethe distribution of the last three gases into theliquid phase. Each of these steps must be doneusing sterile disposable material and followingthe necessary precautions to preserve sterility.

Infusion of the inert gas solutionThe solution is infused into a peripheral super-ficial vein of the forearm by means of a rollerpump directly from the bag in which it wasprepared. A fresh 0-22 jim Millipore filter isinserted into the line to preserve sterility. Aconstant and smooth rate of infusion of 3 ml/min is used in subjects studied at rest. Toachieve a steady state a period of about 30minutes with the infusion running is usuallyrequired before any sampling is carried out.

Sampling procedureAfter ensuring quiet breathing and stableminute ventilation of the subject (often tidalvolume, respiratory rate, end tidal Fo2 andFco2, ECG, and Vo2 are all monitored on line)arterial (4-8 ml) and mixed venous (4-8 ml)samples are drawn simultaneously using hepar-inised barrel matched ungreased glass syringes.The entire sample is collected by slow steadywithdrawal over several respiratory cycles last-ing about 30 seconds. For the collection of(duplicate) 15 ml samples of mixed expired gaswe use matched barrel/plunger glass syringes.A one way low resistance valve is connectedthrough a large bore heated plastic tube to amixing box containing a long coiled coppertube 10 cm in diameter and 70 cm in length,which is kept heated several degrees above bodytemperature. This allows adequate mixing ofthe expiratory gases without significant loss ofsoluble inert gases in condensate and produces alag time that must be allowed for in the sam-pling procedure. It is advantageous to takeduplicate sets of samples in each experimentalcondition. It is also important to take duplicatecontrol samples of the mixed expired gas andpulmonary and systemic arterial blood beforethe infusion of the inert gases to exclude spuri-ous signals that could confound the analysis oflater samples. The control blood samples arelater reused for (duplicate) measurement of thesolubilities of the inert gases.

Measurements of inert gas concentrationsBefore the study the syringes to be used forblood sampling must be weighed both beforeand after heparinisation. The heparin usedshould be tested in order to avoid those brandsthat contain acetone.'7 The mixed expired

821

Roca, Wagner

samples can be directly injected into the gaschromatograph to measure inert gas concentra-tions. For the blood samples we proceed in thefollowing way. Each syringe is weighed a thirdtime, after blood sampling, in order to measurethe volume of blood (calculated as weightdivided by density). Density is measured laterfrom pooled samples. Approximately 10 ml ofnitrogen is transferred into each syringe forequilibration of inert gases between blood andthe nitrogen phase. This procedure is carriedout in a shaking water bath at body temperaturefor approximately 40 minutes. The combinedvolume of the nitrogen and blood is thenmeasured at body temperature and the gasabove the blood is anaerobically transferred to adry syringe to be measured in the gas chromato-graph. The peak of the least soluble gas, SF6, ismeasured by an electron capture detector(ECD) while the remaining five gases (allhydrocarbons) are measured by a flame ionisa-tion detector (FID). The characteristics of theset-up of the gas chromatograph for theMIGET and the quality control procedureshave been detailed elsewhere."38 The chroma-tographic peaks so measured from the gas phaseafter the equilibration procedure are not equalto Pa and Pv of each inert gas. These partialpressures must therefore be calculated usingstandard mass balance formulae" that take intoaccount (1) height or area of the chromato-graphic peaks (fig 5), (2) partition coefficients ofeach inert gas, and (3) volumes of heparin,blood, and gas in the syringe during the equilib-ration procedure. Technicalities of the sampleprocessing and the measurement of solubilitiesare described elsewhere." 38 It is important tonote that the coefficient of variation of the entire

C L-

co a)QL CCO U)

0.

Q

a)0)o

0L) co

<

4-;T-c

(0

01)C

~~~~~~~~~~coC

0

LU ~~~~~~~~~~~co

FID ECD

Figure 5 Gas chromatogram obtained from a mixedvenous blood sample. Chromatographic measurements are

done after the equilibration of the blood sample (bloodphase) with a similar volume of nitrogen (gas phase), as

explained in the text. The peak of each inert gas is

identified in the figure. FID =flame ionisation detector;ECD= electron capture detector.

procedure for measuring gases in blood shouldnot exceed 3% for the five more soluble inertgases at physiological levels (generally deter-mined from 10 samples of a common pool). ForSF6 the coefficient of variation is invariablygreater and a value of 5-6% is reasonable.

Recovering the VA/Q distributionsThe data obtained during the inert gas elimina-tion studies - namely R (Pa/Pv ratio) and E (PE/Pv ratio) for each of the six inert gases and theirpartition coefficients - are fed into a computerprogram that estimates the distribution of VA/Qratios consistent with that quantitative patternof gas tension ratios. The program estimates thevariance of the retention of each inert gas andadjusts the numerical contribution (weight) ofthat gas according to the corresponding error ofmeasurement. Two complementary mathema-tical approaches can be used to examine the VA/( distributions associated with each set of sixinert gas data. One is the least square best fitregression analysis (LSBF) with enforcedsmoothing and the other is the linear program-ming method (LP). The purpose of the LP is toprovide an absolute upper bound of the familyof distributions compatible with the measureddata allowing for both the experimental errorsand the underdetermined nature of the problem(the gas exchange behaviour of approximately105 lung units is examined using only six sets ofdata). This approach has been useful for evalu-ating the limits of the MIGET'32l22 but it iscumbersome and time consuming. By contrast,the LSBF with enforced smoothing is a usefulway for day by day calculations in the clinicalsetting.39 The residual sum of squares (RSS) is,as mentioned earlier, a quantitative estimationof the overall experimental error in the proced-ure. The analysis of the frequency distributionof the RSS from many measurements carriedout in a particular laboratory constitutes a pre-cise way to evaluate adequacy of the data. TheX2 distribution for six degrees of freedom toassess the frequency occurrence of RSS gives avalue of 5-3 for p=05, 106 for p=01, and16 8 for p=001. In other words, 50% of datasets should have RSS < 5-3, 90% of RSSshould be less than 10 6, and 99% less than16 8. The range of the second moment (logSDQ and log SDv) is from 0 30 in healthyyoung subjects (see below) to about 2 5 inextreme lung disease. The overall intrasubjectcoefficient of variation for log SDQ and log SDvis 6% if the mean of duplicate measurements isused.40

Results in normal subjectsAT REST BREATHING ROOM AIRThe characteristic VA/(2 distribution in normalseated young subjects (<30 years) consists ofnarrow perfusion and ventilation curves cen-tered around a VA/Q ratio of one (fig 2). Meanvalues for the second moment of the distribu-tion (log SDQ and log SDv) range from 0-35 to0-43.26 The upper 95% confidence limit for logSDQ is 0 60, and for log SDv is 0.65.24284142 No(or virtually no) perfusion to VA/4 ratios

822


< 0 005 (shunt) is present.26 Likewise, theamount of ventilation to VA/( ratios > 100(including instrumental, anatomical, andphysiological dead space) is approximately30%. As fig 2 illustrates, no perfusion to lungunits with VA/(2 ratios <0 1 (low VA/(2) isobserved; similarly ventilation to lung unitswith VA/() ratios > 10 (high VA/Q) is not pre-sent. Age related changes in the second moment(increased heterogeneity with aging) isexpected,2643 similar to the decline in Pao2.However, the VA/() distributions of olderhealthy subjects remain to be determined.

AT REST BREATHING 100% OXYGENAfter 30 minutes of breathing 100% oxygen26no changes in VA/( mismatch have been shownin young subjects. The only variations observedin this condition were a moderate shift of theVA/( distributions to the right as a result of theincrease in the overall VA/Q ratio of the lungassociated with increased ventilation, and thepresence of a very small amount of shunt (0 5%of cardiac output). By contrast, overt increase inVA/( inequality was observed in a very smallstudy of older subjects (aged 39-60 years) dur-ing 100% oxygen breathing. Increase in bothlow VA/( areas (%( T to VA/( ratios < 0 1) andin the amount of shunt were detected. Thedevelopment of reabsorption atelectasis afterdenitrogenation of lung units with critically lowVA/( ratios, as predicted by Briscoe et al,44 hasbeen the mechanism invoked for the develop-ment of shunt during 100% oxygen breathing.Likewise, the release of hypoxic pulmonaryvasoconstriction in this condition would explainthe increase in the blood flow of lung units withlow VA/( ratios. The changes in the perfusiondistribution (log SDQ) during 100% oxygenbreathing have been used as a way of testing thereactivity of the pulmonary vascular tone indifferent situations.2545

EXERCISEDuring submaximal exercise the dispersion ofthe VA/Q distributions does not change24 2841 butthe VA/4 ratios at the mean of both ventilationand perfusion distributions substantially in-creases due to the higher overall VA/4 ratio.46The efficiency of the lung as an oxygen andcarbon dioxide exchanger therefore improves inthis condition. However, hypoxaemia (or in-crease in A-aPo2) may occur in athletes duringheavy exercise at sea level.47 The MIGET hasbeen used to analyse the physiological mechan-isms behind this phenomenon. It has beenshown that the increase in the A-aPo2 in exer-cise is due, in part, to VA/4 mismatching,242841but it is mostly explained by alveolar-end capil-lary oxygen diffusion limitation.272842 The tech-nique has indicated that pulmonary oxygendiffusing capacity during heavy exercise in manis high, but still somewhat lower than that pre-viously estimated by morphometric measure-ments.48

Abnormalities of pulmonary gas exchangeduring exercise in healthy subjects are clearlyaccentuated if exercise is carried out during

hypoxia produced either by breathing a lowinspired oxygen fraction42 or simulating altitudein a hypobaric chamber.24272849 The increase inlog SDQ during exercise is not associated withaltered spirometry, but shows a significant cor-relation with the increase in mean pulmonaryartery pressure. On the other hand, VA/( mis-matching in exercise is reversed during acutealtitude exposure by breathing 100% oxygen.24During chronic extreme altitude exposure thedevelopment of units with very low or zero VA/( ratios (shunt) is a common feature.49 Ventila-tion-perfusion mismatch occurring with exer-cise at altitude appears, in part, to be deter-mined by the rate of ascent as opposed to thealtitude reached.4950 It has been hypothesisedthat the development of some degree of pul-monary oedema2849 might explain the deteriora-tion in pulmonary gas exchange during heavyexercise, but no direct evidence has yet beenobtained. Alternative explanations for the VA/Qmismatching such as airway obstruction (due toexercise-induced or hypocapnia-induced bron-choconstriction) or non-uniform hypoxic vaso-constriction are not consistent with the de-scribed findings.

SummaryThis introductory review summarises four dif-ferent aspects of the multiple inert gas elimina-tion technique (MIGET). Firstly, the historicalbackground that facilitated, in the mid 1970s,the development of the MIGET as a tool toobtain more information about the entire spec-trum of VA/4 distribution in the lung by mea-suring the exchange of six gases of differentsolubility in trace concentrations. Its principleis based on the observation that the retention(or excretion) of any gas is dependent on thesolubility (X) of that gas and the VA/Q distribu-tion. A second major aspect is the analysis of theinformation content and limitations of the tech-nique. During the last 15 years a substantialamount of clinical research using the MIGEThas been generated by several groups aroundthe world. The technique has been shown to beadequate in understanding the mechanisms ofhypoxaemia in different forms of pulmonarydisease and the effects of therapeutic interven-tions, but also in separately determining thequantitative role of each extrapulmonary factoron systemic arterial Po2 when they changebetween two conditions of MIGET measure-ment. This information will be extensivelyreviewed in the forthcoming articles of thisseries. Next, the different modalities of theMIGET, practical considerations involved inthe measurements and the guidelines for qualitycontrol have been indicated. Finally, a sectionhas been devoted to the analysis of availabledata in healthy subjects under different con-ditions. The lack of systematic information onthe VA/4 distributions of older healthy subjectsis emphasised, since it will be required to fullyunderstand the changes brought about by dis-eases that affect the older population.

This study was supported in part by grants CICYT 90/0136 (JRoca) and NIH grant HL17731-07 (P D Wagner).

823

Roca, Wagner

1 Krogh A, Lindhard J. The volume of the dead space inbreathing and the mixing of gases in the lungs of man.Physiol (London) 1917;51:59-90.

2 Haldane JS. Respiration. New Haven: Yale University Press,1922.

3 West JB. Causes of carbon dioxide retention in lung disease.N Engl JfMed 1971;284:1232-6.

4 Fenn WO, Rahn H, Otis AB. A theoretical analysis of thecomposition of alveolar air at altitude. Am Physiol1946;146:637-53.

5 Riley RL, Cournand A. "Ideal" alveolar air and the analysisof ventilation-perfusion relationships in the lungs. ApplPhysiol 1949;1:825-49.

6 Kelman GR. Digital computer subroutine for the conversionof oxygen tension into saturation. Respir Physiol1966;1:1375-6.

7 Kelman GR. Digital computer procedure for the conversionof Pco2 into blood CO2 content. Respir Physiol 1967;3:111-6.

8 Olszowka AJ, Farhi LE. A system of digital computersubroutines for blood gas calculations. Respir Physiol1968;4:270-80.

9 West JB. Ventilation-perfusion inequality and overall gasexchange in computer models of the lung. Respir Physiol1969;7:88-1 10.

10 Farhi LE. Elimination of inert gases by the lung. RespirPhysiol 1967;3:1-1 1.

11 Wagner PD, Saltzman HA, West JB. Measurement of con-tinuous distributions of ventilation-perfusion ratios:theory. Appl Physiol 1974;36:588-99.

12 Wagner PD, Naumann PF, Laravuso R. Simultaneousmeasurement of eight foreign gases in blood by gas chro-matography. Appl Physiol 1974;36:600-5.

13 Evans JW, Wagner PD. Limits on VA/Q distributions fromanalysis of experimental inert gas elimination. App)Physiol 1977;42:889-98.

14 Lenfant C, Okubo T. Distribution function of pulmonaryblood flow and ventilation-perfusion ratio in man. ApplPhysiol 1964;19:683-92.

15 Kety SS. The theory and applications of the exchange ofinert gas at the lungs and tissues. Pharmacol Rev 1951;3:1-41.

16 Noehren TH. Pulmonary clearance of inert gases with par-ticular reference to ethyl ether. Appl Physiol1962;17:795-8.

17 Wagner PD. Diffusion and chemical reaction in pulmonarygas exchange. Physiol Rev 1977;57:257-312.

18 Wagner PD. Susceptibility of different gases to ventilation-perfusion inequality. Appl Physiol 1979;46:372-86.

19 Fortune JB, Wagner PD. Effects of common dead space on

inert gas exchange in mathematical models of the lung.Appl Physiol 1979;47:896-906.

20 Wagner PD, Evans JW. Conditions for equivalence of gasexchange in series and parallel models of the lung. RespirPhysiol 1977;31:1 17-38.

21 Kapitan KS, Wagner PD. Linear programming analysis ofVA/Q distributions: limits on central moments. App)Physiol 1986;60:1772-81.

22 Kapitan KS, Wagner PD. Linear programming analysis of

VA/4 distributions: average distribution. Appl Physiol1987;62: 1356-62.

23 Hlastala MP, Robertson HT. Inert gas elimination charac-

teristics of the normal and abnormal lung. Appl Physiol1978;44:258-66.

24 Gale GE, Torre-Bueno J, Moon RE, Saltzman HA, WagnerPD. Ventilation-perfusion inequality in normal humansduring exercise. App) Physiol 1985;58:978-88.

25 Agusti GN A, Roca J, Gea J, Wagner PD, Xaubet A,Rodriguez-Roisin R. Mechanisms of gas-exchange impair-ment in idiopathic pulmonary fibrosis. Am Rev Respir Dis

1991;143:219-25.26 Wagner PD, Laravuso RB, Uhl RR, West JB. Continuous

distributions of ventilation-perfusion ratios in normal sub-

jects breathing air and 100% 02- Clin Invest 1974;54:54-68.

27 Torre-Bueno J, Wagner PD, Saltzman HA, Gale GE, MoonRE. Diffusion limitation in normal humans during exercise

at sea level and simulated altitude. Appl Physiol1985;58:989-95.

28 Wagner PD, Gale GE, Moon RE, Torre-Bueno J, Saltzman

HA. Pulmonary gas exchange in humans exercising at sealevel and simulated altitude. Appl Physiol 1986;61:260-70.

29 Gea J, Roca J, Torres A, Agusti AGN, Wagner PD, Rodri-guez-Roisin R. Mechanisms of abnormal gas exchange inpatients with pneumonia. Anesthesiology 1991;75:782-9.1991;75:782-9.

30 Truog WE, Hlastala MP, Standaert TA, McKenna HP,Hodson WA. Oxygen-induced alteration of ventilation-perfusion relationships in rats. Appl Physiol1979;47:1112-7.

31 Longworth KE, Jones JH, Bicudo JEPW, Taylor CR, Wei-bel ER. High rate of 02 consumption in exercising foxes:large Po2 difference drives diffusion across the lung. RespirPhysiol 1989;77:263-76.

32 Powell FL, Wagner PD. Ventilation-perfusion inequality inavian lungs. Respir Physiol 1982;48:233-41.

33 Powell FL, Wagner PD. Measurement of continuous distri-butions of ventilation-perfusion in non-alveolar lungs.Respir Physiol 1982;48:219-32.

34 Wagner PD, Smith CM, Davies NJH, McEvoy RD, GaleGE. Estimation of ventilation-perfusion inequality by inertgas elimination without arterial sampling. Appl Physiol1985;59:376-83.

35 Wagner PD, Hedenstiema G, Bylin G. Ventilation-perfusioninequality in chronic asthma. Am Rev Respir Dis1987;136:605-12.

36 Roca J, Ramis Ll, Rodriguez-Roisin R, Ballester E, Mont-serrat JM, Wagner PD. Serial relationships between VA/Qinequality and spirometry in acute severe asthma requiringhospitalization. Am Rev Respir Dis 1988;137:605-12.

37 Powell FL, Lopez FA, Wagner PD. Effects of acetone inheparin on the multiple inert gas elimination technique.Appl Physiol 1985;58:143-7.

38 Wagner PD, Lopez FA. Gas chromatography techniques inrespiratory physiology. In: Otis AB, ed. Techniques in lifescience. Ireland: Elsevier, 1984:403/1-403/24.

39 Wagner PD. Calculating the distribution of ventilation-perfusion ratios from inert gas elimination data. Fed Proc1982;41: 136-9.

40 Wagner PD, Hedenstierna G, Bylin G, Lagerstrand L.Reproducibility of the multiple inert gas elimination tech-nique. Appl Physiol 1987;62: 1740-6.

41 Hammond MD, Gale GE, Kapitan KS, Ries A, Wagner PD.Pulmonary gas exchange in humans during exercise at sealevel. Appl Physiol 1986;60:1590-8.

42 Hammond MD, Gale GE, Kapitan KS, Ries A, Wagner PD.Pulmonary gas exchange in humans during normobarichypoxic exercise. Appl Physiol 1985;58:978-88.

43 Derks CM. Ventilation-perfusion distribution in young andold volunteers during mild exercise. Bull Eur PhysiopatholRespir 1980;16:145-54.

44 Briscoe WA, Cree EM, Filler J, Houssay HEJ, Cournand A.Lung volume, alveolar ventilation and perfusion interrela-tionships in chronic pulmonary emphysema. Y Appl Phy-siol 1960;15:785-95.

45 Rodriguez-Roisin R, Ballester E, Roca J, Torres A, WagnerPD. Mechanisms of hypoxaemia in patients with statusasthmaticus requiring mechanical ventilation. Am RevRespir Dis 1989;139:732-9.

46 Gledhill N, Froese AB, Dempsey JA. Ventilation to perfu-sion distribution during exercise in health. In: DempseyJA, Reed CE, eds. Muscular exercise and the lung. Madison,Wisconsin: University of Wisconsin Press, 1977:315-24.

47 Dempsey JA, Handson PG, Henderson KS. Exerciseinduced arterial hypoxaemia in healthy human subjects atsea level. Physiol (Lond) 1984;355:161-75.

48 Taylor CR, Karas RH, Weibel ER, Hoppeler H. Matchingstructures and functions in the respiratory system. In:Wood SC, ed. Comparative pulmonary physiology. NewYork: Marcel Dekker, 1989:27-65.

49 Wagner PD, Sutton JR, Reeves JT, Cymerman A, GrovesBM, Malconian MK. Operation Everest II: pulmonary gasexchange during a simulated ascent of Mt Everest. ApplPhysiol 1987;63:2348-59.

50 Bebout DE, Story D, Roca J, Hogan MC, Poole DC,Gonzalez-Camarena R, et al. Effects of altitude acclimat-ization on pulmonary gas exchange during exercise. ApplPhysiol 1989;67:2286-95.

824

migetprinciples and information content of the multiple inert gas elimination technique

Documents