A modeling-based evaluation of isothermal rebreathing for breath gas analyses of highly
soluble volatile organic compounds
Julian Kinga,b , Karl Unterkoflera,c, Gerald Teschlb , Susanne Teschld , Paweł Mochalskia,e, Helin Koçc , Hartmann Hinterhuberf,
Anton Amanna,g,∗
a Breath
Research Institute, Austrian Academy of Sciences, Rathausplatz 4, A-6850 Dornbirn, Austria
of Vienna, Faculty of Mathematics, Nordbergstr. 15, A-1090 Wien, Austria
c Vorarlberg University of Applied Sciences, Hochschulstr. 1, A-6850 Dornbirn, Austria
d University of Applied Sciences Technikum Wien, Höchstädtplatz 5, A-1200 Wien, Austria
e Institute of Nuclear Physics PAN, Radzikowskiego 152, PL-31342 Kraków, Poland
f Innsbruck Medical University, Department of Psychiatry, Anichstr. 35, A-6020 Innsbruck, Austria
g Innsbruck Medical University, Univ.-Clinic for Anesthesia, Anichstr. 35, A-6020 Innsbruck, Austria
arXiv:1105.0864v1 [q-bio.QM] 4 May 2011
b University
Abstract
Isothermal rebreathing has been proposed as an experimental technique for estimating the alveolar levels of hydrophilic volatile
organic compounds (VOCs) in exhaled breath. Using the prototypic test compound acetone we demonstrate that the end-tidal breath
profiles of such substances during isothermal rebreathing show characteristics that contradict the conventional pulmonary inert gas
elimination theory due to Farhi. On the other hand, these profiles can reliably be captured by virtue of a previously developed
mathematical model for the general exhalation kinetics of highly soluble, blood-borne VOCs, which explicitly takes into account
airway gas exchange as major determinant of the observable breath output.
This model allows for a mechanistic analysis of various rebreathing protocols suggested in the literature. In particular, it clarifies
the discrepancies between in vitro and in vivo blood-breath ratios of hydrophilic VOCs and yields further quantitative insights into
the physiological components of isothermal rebreathing.
Keywords: breath gas analysis, volatile organic compounds, highly soluble gas exchange, rebreathing, acetone, modeling
1. Introduction
Over the past decade, much advance has been made
in the attempt to extract the diagnostic and metabolic
information encapsulated in a number of endogenous
volatile organic compounds (VOCs) appearing in human
exhaled breath (Amann et al., 2007; Amann and Smith, 2005;
Miekisch et al., 2004). However, the quantitative understanding of the relationships between breath concentrations of such
trace gases and their underlying systemic levels clearly lags
behind the enormous analytical progress in breath gas analysis.
In particular, formal means for evaluating the predictive power
of various breath sampling regimes are still lacking.
Recently, several efforts have been undertaken to complement VOC measurements with adequate physical models
mapping substance-specific distribution mechanisms in the
pulmonary tract as well as in the body tissues. Some major
breath VOCs have already been investigated in this form,
e.g., during rest and exercise conditions or exposure scenarios (Tsoukias and George, 1998; Kumagai and Matsunaga,
2000; Anderson et al., 2003; Mörk and Johanson, 2006;
King et al., 2010a, 2011; Koc et al., 2011). Such mechanistic
descriptions of the observable exhalation kinetics give valuable
∗ Corresponding
author. Tel.: +43 676 5608520; fax: +43 512 504 6724636.
Email address: anton.amann@oeaw.ac.at (Anton Amann)
Preprint submitted to Respiratory Physiology & Neurobiology
insights into the relevance of the measured breath concentrations with respect to the endogenous situation and hence are
mandatory to fully exploit the diagnostic potential of breath
VOCs.
Within this context, the main focus of this article will be
on a modeling-based review of isothermal rebreathing, which
has been proposed as an experimental technique for estimating
the alveolar levels of hydrophilic exhaled trace gases (Jones,
1983; Ohlsson et al., 1990; O’Hara et al., 2008). This class
of compounds has been demonstrated to significantly interact
with the water-like mucus membrane lining the conductive airways, an effect which has become known as wash-in/wash-out
behavior. For further details we refer to Anderson et al. (2003);
Anderson and Hlastala (2007). As a phenomenological consequence, exhaled breath concentrations of such highly water
soluble substances tend to be diminished on their way up from
the deeper respiratory tract to the airway opening. The resulting
discrepancies between the “true” alveolar and the measured
breath concentration can be substantial (even if breath samples
are drawn in a strictly standardized manner employing, e.g.,
CO2 - or flow-controlled sampling) and will depend on a variety
of factors, including airway temperature profiles and airway
perfusion as well as breathing patterns (Tsu et al., 1991).
In particular, the above-mentioned effect considerably
departs from the classical Farhi description of pulmonary
May 5, 2011
inert gas exchange (Farhi, 1967), on the basis of which the
operational dogma has been established that end-tidal air will
reflect the alveolar level and that arterial concentrations can be
assessed by simply multiplying this value with the blood:gas
partition coefficient λb:air at body temperature. This “common
knowledge” has first been put into question in the field of
breath alcohol testing, revealing observable blood-breath
concentration ratios of ethanol during tidal breathing that
are unexpectedly high compared to the partition coefficient
derived in vitro (Jones, 1983). Similarly, excretion data (i.e.,
the ratios between steady state partial pressures in expired air
and mixed venous blood) of highly water soluble compounds
(including the MIGET test gas acetone) have been shown to
underestimate the values anticipated by treating the airways as
an inert tube (Zwart et al., 1986; Anderson and Hlastala, 2010).
at m/z = 59 (dwell time 200 ms). Additionally, we routinely
measure the mass-to-charge ratios m/z = 21 (isotopologue
of the primary hydronium ions used for normalization; dwell
time 500 ms), m/z = 37 (first monohydrate cluster H3 O+ (H2 O)
for estimating sample humidity; dwell time 2 ms), m/z = 69
(protonated isoprene; dwell time 200 ms), m/z = 33 (protonated methanol; dwell time 200 ms) as well as the parasitic
precursor ions NH+4 and O+2 at m/z = 18 and m/z = 32, respectively, with dwell times of 10 ms each. In particular, following O’Hara et al. (2008), the pseudo concentrations associated
with m/z = 32 determined according to standard PTR-MS practice (see, e.g., (Schwarz et al., 2009a, Eq. (1))) will be used as a
surrogate for the end-tidal oxygen partial pressure PO2 (relative
to an assumed nominal steady state level at rest of 100 mmHg).
Similarly, calibrated pseudo concentrations corresponding to
m/z = 37 are considered as an indicator for absolute sample
humidity Cwater , see (King et al., 2011) for further details. Partial pressures PCO2 of carbon dioxide are obtained via a separate sensor. Table 1 summarizes the measured quantities used
in this paper. In general, breath concentrations will always refer
to end-tidal levels.
In a previously published mathematical model for the breath
gas dynamics of highly soluble blood-borne trace gases, airway
gas exchange is taken into account by separating the lungs into a
bronchial and alveolar compartment, interacting via a diffusion
barrier mimicking pre- and post-alveolar uptake (King et al.,
2011). This formulation has proven its ability to reliably capture the end-tidal breath profiles as well as the systemic dynamics of acetone in a variety of experimental situations and
will be used here for illuminating the physiological processes
underlying isothermal rebreathing tests as carried out in the literature. For comparative reasons, the illustration in the sequel
will mainly be limited to acetone, with possible extrapolations
to other highly soluble VOCs indicated where appropriate.
Variable
Symbol
Nominal value (units)
Acetone concentration
Cmeasured
1 (µg/l) (Schwarz et al., 2009b)
CO2 partial pressure
PCO2
40 (mmHg) (Lumb, 2005)
Water content
Cwater
4.7 (%) (Hanna and Scherer, 1986)
O2 partial pressure
P O2
100 (mmHg) (Lumb, 2005)
Table 1: Summary of measured breath parameters together with some nominal
values in end-exhaled air during tidal breathing at rest.
2. Methods
2.1. Experiments
2.2. Physiological model
Extensive details regarding our experimental setup are given
elsewhere (King et al., 2009, 2010b, 2011). Here, we will only
briefly discuss the parts relevant in the context of isothermal
rebreathing. All phenomenological results were obtained in
conformity with the Declaration of Helsinki and with approval
by the Ethics Commission of Innsbruck Medical University.
A schematic sketch of the model structure is presented in
Fig. 1. For the associated compartmental mass balance equations we refer to Appendix A as well as to the original publication (King et al., 2011). Briefly, the body is divided into
four distinct functional units, for which the underlying concentration dynamics of the VOC under scrutiny will be taken into
account: bronchial/mucosal compartment (Cbro ; gas exchange),
alveolar/end-capillary compartment (CA ; gas exchange), liver
(Cliv ; production and metabolism) and tissue (Ctis ; storage).
The nomenclature is detailed in Table A.1.
The rebreathing system itself consists of a Tedlar bag (SKC,
Dorset, UK) with a volume of Ṽbag = 3 l, that can directly
be connected to a spirometer face mask covering mouth and
nose. From the latter, end-tidal exhalation segments are drawn
into a Proton Transfer Reaction Mass Spectrometer (PTR-MS;
Ionicon Analytik GmbH, Innsbruck, Austria), which allows
for VOC detection and quantification on a breath-by-breath
resolution as described in (King et al., 2009). The rebreathing bag is warmed to 37 ± 1◦ C by means of a specially designed outer heating bag (Infroheat Ltd., Wolverhampton, UK),
cf. (O’Hara et al., 2008). Heating is intended to assist the
thermal equilibration between the alveolar tract and the upper airways and prevents condensation and subsequent losses
of hydrophilic VOCs depositing onto water droplets forming
on the surface wall of the bag. End-tidal acetone concentrations are determined by monitoring the protonated compound
The measurement process is described by
Cmeasured = Cbro ,
(1)
i.e., we assume that the measured (end-tidal) VOC concentration reflects the bronchial level. Dashed boundaries indicate
a diffusion equilibrium, described by the appropriate partition
coefficients λ, e.g., λb:air . The nomenclature is detailed in Table A.1.
2
the model. More specifically, the factor z captures the change
of VOC solubility in the airway mucosa and bronchial blood
in response to fluctuations of the characteristic mean airway
and bronchial blood temperature T̄ (in ◦ C). In King et al. (2011,
Eqs. (3)-(5)) the latter has been shown to be estimable by virtue
of the measured sample humidity Cwater and typically ranges
around 34◦ C during tidal breathing of room air at rest. It is assumed that the temperature dependence of λb:air is proportional
to the temperature change of the mucosa:air partition coefficient
λmuc:air over the temperature range considered, i.e. z is given by
z(T̄ ) := λmuc:air (T̄ )/λmuc:air (37◦ C).
(4)
Remark 1. Note that z decreases with increasing temperature
and takes values greater than one for T̄ below body temperature.
A particular convenience of the above representation stems
from the fact that, as a first approximation, the mucosa layer
can be assumed to inherit the physico-chemical properties
of water. Correspondingly, λmuc:air can be estimated via the
respective water:air partition coefficient, which is usually
available from the literature (see, e.g., the extensive compendium in (Staudinger and Roberts, 2001)).
A central role is played by the gas exchange location parameter D, mimicking pre- and post-alveolar uptake in the mucosa. As has been discussed in (King et al., 2011), this quantity is a re-interpretation of stratified conductance according
to the series inhomogeneity model by Scheid et al. (1981) (see
also (Hlastala et al., 1981)) and will be close to zero for highly
water soluble substances during tidal breathing at rest. Moreover, an increase of D with ventilatory flow can be expected,
which will be modeled here as
Figure 1: Sketch of the model structure used for capturing dynamic VOC
concentrations (C). Subscripts connote as follows: bag–rebreathing bag;
I–inhaled; bro–bronchial; muc–mucosal; A–alveolar; c′ –end-capillary; liv–
liver; tis–tissue; b–blood.
In particular, for highly water- and blood-soluble compounds
such as acetone the present model replaces the familiar Farhi
equation describing the steady state relationship between inhaled (ambient) concentration CI , measured breath concentraFarhi
tions Cmeasured
, mixed venous concentrations Cv̄ and arterial
concentrations Ca during tidal breathing (Farhi, 1967),
Farhi
Cmeasured
= CA =
V̇A
C
Q̇c I
+ Cv̄
λb:air +
V̇A
Q̇c
=
Ca
,
λb:air
D := kdiff,1 max{0, VT − VTrest } + kdiff,2 max{0, V̇A − V̇Arest }, (5)
where V̇A and VT denote alveolar ventilation and tidal volume,
respectively. The positive coefficients kdiff,i can be estimated
from experimental data, cf. (King et al., 2011) and Table A.1.
(2)
3. Results and discussion
with the expression
Cmeasured = Cbro =
rbro CI + Ca
rbro CI + (1 − qbro )Cv̄
=
.
(1 − qbro )z(T̄ )λb:air + rbro z(T̄ )λb:air + rbro
3.1. Heuristic considerations
As has already been indicated in the introduction, Equation (2) is inappropriate for capturing experimentally obtained
arterial blood-breath concentration ratios (BBR) of highly water soluble trace gases during free breathing at rest (i.e., assuming CI = 0). For instance, in the specific case of acetone,
multiplying the proposed end-tidal population mean of approximately 1 µg/l (Schwarz et al., 2009b) with a blood:gas partition
coefficient of λb:air = 340 (Anderson et al., 2006) at body temperature appears to grossly underestimate arterial blood levels
spreading around 1 mg/l (Kalapos, 2003; Wigaeus et al., 1981).
In contrast, Equation (3) asserts that the observable arterial
blood-breath ratio in this case is
(3)
Here, qbro ≈ 0.01 is an estimate of the effective fractional
bronchial perfusion (which might be substance-specific), while
rbro =
V̇A
qbro Q̇c
denotes the associated bronchial ventilation-perfusion ratio.
The term z(T̄ )λb:air reflects an explicit temperature dependence of airway gas exchange that has been incorporated into
BBR =
3
Ca
Cmeasured
= z(T̄ )λb:air + rbro ,
(6)
and will thus depend on airway temperature and airway blood
flow as mentioned in the previous section. In particular, note
that the BBR will be greater than the in vitro blood:gas partition
coefficient λb:air for T̄ below body temperature, cf. Remark 1.
Formally, a model capturing the experimental situation during isothermal rebreathing can simply be derived by augmenting the model equations in Appendix A with an additional
compartment representing the rebreathing receptacle, i.e.,
Remark 2. From Equation (6) it is clear that the more soluble a VOC under scrutiny, the more drastically its observable BBR will be affected by the current airway temperature,
with an inverse relation between these two quantities. This
deduction is consistent with measurements by Ohlsson et al.
(1990) conducted in the field of alcohol breath tests, reporting a monotonous decrease of ethanol BBR values with increasing exhaled breath temperature. In the same contribution,
BBRs of ethanol during normal tidal breathing were shown to
typically exceed a value of 2500, which differs from the expected value λb:air = 1756 by more than 40%. For comparison, based on a mean characteristic airway temperature of
T̄ = 32◦ C and the water:air partition coefficient λmuc:air provided by Staudinger and Roberts (2001, p. 568), by substituting
the parameter values in Table A.1 we find that Equation (6) predicts an experimentally observable blood-breath ratio of ethanol
in the range of 2560. Equation (6) also suggests that in the case
of VOCs with an extremely high affinity for both blood and water (e.g., ethanol, methanol), the discrepancies between in vivo
and in vitro BBR values might largely be removed by warming
the airways to body temperature before the breath sampling procedure. Contrarily, blood-breath ratios of less soluble VOCs,
e.g., acetone, will additionally be affected by the comparatively
large value of rbro , which stems from the diffusion disequilibrium between the alveolar and bronchial space.
dCbag
Ṽbag = V̇A (Cbro − Cbag )
dt
(7)
and by setting CI = Cbag in Equation (A.1). Following the line
of argument in (King et al., 2011, Appendix B), it can easily
be checked that all fundamental system properties discussed
there remain valid. In particular, if we assume that the conducting airways are warmed to body temperature, i.e., if z(T̄ )
in Equation (4) approaches unity as temperature increases, we
may conclude that the compartmental concentrations will tend
to a globally asymptotically stable steady state obeying
rebr
rebr
rebr
Cmeasured
= Cbro
= Cbag
= CArebr =
Cv̄rebr Carebr
=
.
λb:air λb:air
(8)
This is a simple consequence of substituting CI with Cbro in
Equation (3) according to the steady state relation associated
with Equation (7). It is important to realize that if steady
state conditions hold they will depend solely on the blood:air
partition coefficient λb:air at 37◦ C, thus rendering isothermal
rebreathing as an extremely stable technique for providing
a reproducible coupling between VOC levels in breath and
blood. Particularly, it theoretically avoids the additional measurement of ventilation- and perfusion-related variables that
would otherwise affect this relationship, thereby significantly
simplifying the required technical setup for breath sampling.
However, as will be illustrated in the following, the major
practical obstacle is to guarantee that a steady state as in
Equation (8) is effectively attained.
Apart from providing some experimental evidence for the validity of Equation (6), these ad hoc calculations suggest that the
common practice of multiplying the measured breath concentration Cmeasured with λb:air to obtain arterial concentrations for
highly soluble trace gases will result in an estimation that might
drastically differ from the true blood level.
Isothermal rebreathing has been put forward as a valuable
method for removing the above-mentioned discrepancies. The
heuristic intention leading to isothermal rebreathing is to create
an experimental situation where the alveolar levels of highly
soluble VOCs are not altered during exhalation due to loss of
such substances to the cooler mucus layer of the airways. This
can be accomplished by “closing the respiratory loop”, i.e., by
continuous re-inspiration and -expiration of a fixed mass of air
from a rebreathing receptacle (e.g., a Tedlar bag), causing the
airstream to equilibrate with the mucosa linings over the entire
respiratory cycle (Ohlsson et al., 1990; Anderson et al., 2006;
O’Hara et al., 2008). Additionally, warming the rebreathing
volume to body temperature will ensure a similar solubility of
these VOCs in both regions, alveoli and airways.
For the purpose of comparing the qualitative implications of
Equation (2) and Equation (3), assume that Cv̄rebr ≈ Cv̄ , i.e.,
the mixed venous concentrations stay constant during rebreathing (which – at least in the first phase of rebreathing – can be
justified to some extent by reference to tissue-to-lung transport delays of the systemic circulation (Grodins et al., 1967)).
Then the ratios between measured rebreathing concentrations
and end-exhaled concentrations during free tidal breathing at
rest predicted by Equation (2) and Equation (3) are found to
follow an entirely different trend. To this end, note that while
in the first case we find that
Farhi,rebr
Cmeasured
Farhi
Cmeasured
λb:air +
→
λb:air
V̇A
Q̇c
,
(9)
the present model yields
rebr
Cmeasured
(1 − qbro )z(T̄ )λb:air + rbro
→
,
Cmeasured
(1 − qbro )λb:air
Remark 3. For the sake of completeness, it should be noted
that a small part of the air within the rebreathing circuit outlined in Section 2.1 is actually consumed by the PTR-MS device (roughly 40 ml/min), which, however, can safely be negelected in the present setting.
(10)
where z(T̄ ) ≥ 1, cf. Remark 1. For highly soluble trace gases,
this observation constitutes a simple test for assessing the adequacy of the Farhi formulation regarding its ability to describe
4
the corresponding exhalation kinetics. Indeed, for sufficiently
large λb:air , the right-hand side of Equation (9) will be close to
one, while the right-hand side of Equation (10) suggests that
rebreathing will increase the associated end-tidal breath concentrations. In other words, for this class of compounds a
markedly non-constant behavior during the initial isothermal
rebreathing period indicates that Equation (2) will fail to capture some fundamental characteristics of pulmonary excretion.
Such tests are of particular importance in the context of endogenous MIGET methodology (Multiple Inert Gas Elimination
Technique, based on endogenous rather than externally administered VOCs (Anderson and Hlastala, 2010)), as they might be
used for detecting deviations of the employed gases from the
underlying Farhi description (see also (King et al., 2010b)).
during free breathing, cf. Fig. 2. Here, for simplicity it is assumed that end-tidal PCO2 levels reflect those of the peripheral
and central chemoreceptor environment. The alveolar ventilation follows from
V̇A = f (VT − VD ),
(12)
using a deadspace volume of VD = 0.1 l.
Since cardiac output can be expected to stay relatively constant during the rebreathing phase (O’Hara et al., 2008), we fix
its value at a nominal level of 6 l/min. Using the parameter
values in Table A.1 for a male of height 180 cm and weight
70 kg this completes the necessary data for simulating the
above-mentioned rebreathing experiment. More specifically,
the model response in the first panel of Fig. 2 is obtained by integrating the differential equations (A.1)–(A.4) and (7), setting
CI = Cbag for t ∈ [tstart , tend ] and CI ≡ 0 otherwise. The initial
concentrations for each compartment as well as the underlying
endogenous acetone production rate kpr = 0.055 mg/min for
this specific experiment are derived from the algebraic steady
state conditions at t = 0.
3.2. Single cycle rebreathing
In this section we will discuss some simulations and preliminary experiments conducted in order to study the predictive
value of isothermal rebreathing within a realistic setting. For
this purpose we will first mimic isothermal rebreathing as it
has been carried out by various investigators (Jones, 1983;
Ohlsson et al., 1990). The volume of the rebreathing bag is
Ṽbag = 3 l according to Section 2.1.
[µg/l]
1
Typical profiles of end-tidal acetone, water, CO2 and oxygen content during normal breathing and isothermal rebreathing at rest are shown in Fig. 2. These representative data correspond to one single normal healthy male volunteer from the
study cohort in (King et al., 2009). As has been explained in
Section 2.1, PO2 is derived by scaling the end-tidal steady state
of the PTR-MS pseudo concentration signal at m/z = 32 to a
basal value of 100 mmHg during free tidal breathing. Rebreathing was instituted at tstart = 2 min by inhaling to total lung capacity and exhaling until the bag was filled, thereby providing
an initial bag concentration which can be assumed to resemble
the normal end-exhaled steady state, i.e.,
0.6
0
1
2
[µg/l]
C [µg/l]
3
4
5
6
0.3
1
A
C
bag
[µg/l]
Cliv [mg/l]
C
tis
[mg/l]
0.8
0.2
0.6
0.4
0
1
2
3
4
5
6
100
(11)
6
4
50
P
CO2
Rebreathing was then continued until either the individual
breathing limit was reached or the CO2 partial pressure increased above 55 mmHg (corresponding to tend ≈ 4.6 min for
the volunteer in Fig. 2).
Breathing frequency f and tidal volume VT were simulated on the basis of the monitored values for PCO2 and PO2 .
Under iso-oxic conditions, after a certain threshold value is
exceeded, both frequency and tidal volume are known to
increase linearly with alveolar PCO2 . Moreover, the corresponding slopes (reflecting the chemoreflex sensitivity of
breathing) are dependent on the current alveolar oxygen partial pressure (Mohan and Duffin, 1997; Lumb, 2005). More
specifically, hypoxia during rebreathing enhances chemoreflex
sensitivity, yielding a hyperbolic relation between the mentioned slopes and PO2 . These findings result in a simple
model capturing the chemoreflex control of breathing in humans (Duffin et al., 2000; Duffin, 2010), which has been reimplemented in order to compute f and VT from basal values
0
0.1
0
1
[mmHg]
2
P [mmHg]
H O content [%]
O2
3
2
2
4
5
6
3
150
tidal volume [l]
alveolar ventilation [l/min]
100
[l]
2
1
0
[%]
[µg/l]
bro
[mg/l]
C
50
0
1
2
3
[min]
4
5
6
[l/min]
0.4
[mmHg]
Cbag (0) = Cbro (0).
breath acetone
breath methanol [a.u.]
present model
Farhi model
0.8
0
Figure 2: Representative outcome of an isothermal rebreathing experiment during rest. Data correspond to one single normal healthy male volunteer from the
study cohort in (King et al., 2009). Isothermal rebreathing starts at tstart = 2 min
and ends at approximately tend = 4.6 min. Measured or derived quantities according to the experimental setup are shown in the first and third panel (red
tracings), while data in the second and fourth panel correspond to simulated
variables as described in the text (black tracings).
From Fig. 2, the proposed model is found to faithfully reproduce the observed data, which extends the range of exper5
{kpr , kmet , qbro , kdiff,1 , kdiff,2 }. Among these, the decisive quantities are the blood:gas partition coefficient (negative influence;
ς(λb:air ) = 2.07) as well as the initial concentrations and partition coefficients for the liver and tissue compartment (positive
influence; ς(Cliv (0)) = 0.06; ς(λb:liv ) = 0.06; ς(Ctis (0)) = 1.39;
ς(λb:tis ) = 1.39). In particular, the latter essentially define
the mixed venous blood acetone concentration Cv̄ according
to Equation (A.5). As this concentration stays almost constant
throughout the entire experiment, the above sensitivity analysis
results are in direct agreement with Equation (8), predicting a
rebreathing steady state that exclusively depends on λb:air and
Cv̄ . All other sensitivity indices take values below 0.015.
imental regimes for which the underlying formalism has been
validated. Further improvement of the goodness-of-fit might
be achieved by employing parameter estimation techniques as
described, e.g., in King et al. (2010a, 2011) which, however,
would be beyond the scope of this paper.
In contrast, as can be anticipated from Equation (9), the classical Farhi model fails to capture the given breath profile1 .
In particular, the presented data appear to consolidate the
heuristic considerations in Section 3.1 and confirm that the
alveolar concentration of acetone during free tidal breathing
can differ from the associated bronchial (i.e., measured endexhaled) level by a factor of more than 1.5. This is due to
an effective diffusion disequilibrium between the conducting
airways and the alveolar space. During isothermal rebreathing,
the diffusion barrier slowly vanishes and causes the measured
breath concentration to approach the underlying alveolar
concentration (which itself stays relatively constant during the
entire experiment). We stress the fact that in order to simulate
a similar response by using the conventional Farhi model, one
essentially would have to postulate a temporarily increased
endogenous acetone production during rebreathing, which,
however, lacks physiological plausibility.
3.3. Sequential rebreathing
On the basis of the results from the previous subsection, in
the following we will briefly discuss a sequential rebreathing
protocol developed by O’Hara et al. (2008). This regime aims
at improving the patient compliance of conventional rebreathing by repeatedly providing cycles of five rebreaths (postulated
to last approximately 0.5 min) with intermediate periods of
free tidal breathing lasting approximately 10 min. According
to above-mentioned protocol, isothermal rebreathing is again
instituted by inhaling to total lung capacity and exhaling to
residual volume into a Tedlar bag with a volume of Ṽbag = 3 l.
After each rebreathing cycle, the bag is closed, a small amount
of bag air (< 100 ml) is measured and the volunteer starts the
next cycle by exhaling to residual volume and inhaling from
the bag.
For comparative reasons, in Fig. 2 we also display the simultaneously obtained PTR-MS concentration profile of breath
methanol, scaled to match the initial level of breath acetone.
Taking into account a methanol blood:gas partition coefficient
of λb:air = 2590 at body temperature (Kumagai and Matsunaga,
2000), from Equation (10) it can be deduced that for this compound the differences between concentrations extracted during
free breathing and rebreathing primarily stem from the thermal
equilibration between airways and alveolar tract. The corresponding rise in temperature is mirrored by a steady increase
of sample water vapor Cwater , approaching an alveolar level of
about 6.2%. In particular, the presented profile for methanol
shows the necessity of including an explicit temperature dependence in models describing the rebreathing behavior of highly
soluble VOCs.
From the data in the last two panels of Fig. 2 it can be inferred that all physiological input variables will have returned
to pre-rebreathing values within the 10 min breaks. For simulation purposes, it will hence be assumed that their behavior
during repeated rebreathing segments is identical to the profiles
within the first 0.5 min of single cycle rebreathing. Values for
the initial compartment concentrations as well as for the additional parameters are adopted from the previous subsection.
These premises allow for simulating the evolution of breath
acetone within a repeated rebreathing regime as displayed in
Fig. 3. Here, the initial bag concentration at the onset of each
individual rebreathing cycle is determined by the final bag concentration after the preceding rebreathing cycle, i.e., no fresh
room air enters the bag.
The bag concentration profile in Fig. 3 qualitatively resembles the data presented by O’Hara et al. (2008). However,
what emerges from this modeling-based analysis is that in spite
of steadily increasing bag concentrations (finally reaching a
plateau level), the latter might not necessarily approach the underlying alveolar concentration as in the case of single cycle
rebreathing. The major reason for this is a lack of complete
thermal and diffusional equilibration between the airways and
the alveolar region within the individual rebreathing segments.
One potential way to circumvent this issue would be to reduce
the desaturation and cooling of the airway tissues between consecutive rebreathing segments by keeping the intermediate time
interval of free tidal breathing as short as possible (while si-
Remark 4. A ranking of specific model parameters and initial conditions pi with respect to their impact on the observable
breath acetone concentration during the isothermal rebreathing
period [tstart , tend ] can be obtained by numerically approximating the squared L2 -norm of the corresponding normalized sensitivities, viz.,
ς(pi ) :=
Ztend
tstart
pi
∂Cmeasured (t)
∂pi
max s |y(s)|
!2
dt.
(13)
Adopting the nomenclature in Appendix A, these indices
were calculated for all effective compartment volumes, partition coefficients, initial concentrations, and for pi ∈
1 Considering the fact that the Farhi formulation is included in the present
model as a limiting case for qbro = 0 and D → ∞ (King et al., 2011, Fig. 4), its
associated output can be computed in a similar manner as described above.
6
multaneously maintaining a regime allowing for comfortable
breathing).
The second panel in Fig. 3 displays the evolution of the predicted blood-bag concentration ratios during the course of experimentation. Note that the in vitro blood:gas partition coefficient λb:air = 340 is never attained. This observation can offer
some explanation for the discrepancies that continue to exist
with regard to theoretical and experimentally measured ratios
between blood and (rebreathed) breath levels (O’Hara et al.,
2009). Furthermore, the final plateau value and the observable
BBR of acetone will vary with temperature (cf. Equation (6)),
which is consistent with similar observations made in the case
of breath ethanol measurements (Ohlsson et al., 1990).
quantitative insights into the discrepancies between in vitro
and measured blood-breath ratios of such VOCs during steady
state. Dynamic data are presented for one single representative
subject only, inasmuch as our main emphasis was on describing
and clarifying some fundamental features and physiological
mechanisms related to the observable VOC behavior during the
above-mentioned experimental regime.
Several practical implications emerge from this modelingbased analysis. Firstly, it is demonstrated that the classical Farhi
setting will fail to reproduce the experimentally measured acetone exhalation data during isothermal rebreathing if a constant
endogenous production and metabolism rate is postulated. This
is due to the fact that airway gas exchange, being a major determinant affecting highly soluble gas exchange, is not taken
into account within this formalism. Furthermore, multiplying
end-tidal breath concentrations during free tidal breathing with
the substance-specific blood:gas partition coefficient λb:air will
generally underestimate the true arterial blood concentration for
highly blood and water soluble VOCs.
Excessive hypoxia and hypercapnia are the main factors limiting the duration of the rebreathing maneuver and preventing a
complete equilibration of VOC partial pressures in the alveoli
and the conducting airways. From an operational point of view,
our data indicate that even if isothermal rebreathing is continued until the individual breathing limit is reached, a steady state
according to Equation (8) might not necessarily be attained. As
can be deduced from Fig. 2 in the case of acetone, end-exhaled
breath (or bag) concentrations extracted after about 0.5 min of
rebreathing (corresponding to the common protocol of providing around five consecutive rebreaths) are still likely to underestimate the underlying alveolar level CA . Analogous conclusions
can be drawn for sequential rebreathing protocols designed to
allow for a recovery of the volunteer during the individual rebreathing segments.
A reliable extraction of meaningful breath levels for highly
soluble VOCs by virtue of isothermal rebreathing hence appears to require more sophisticated setups incorporating the
continuous removal of CO2 and replacement of metabolically
consumed oxygen. Provided that the influence of chemical
CO2 -absorption on the measured breath and bag concentrations
is negligible, such setups might for instance be adapted from
closed chamber techniques (Filser, 1992) or general closedcircuit anesthesia systems.
Alternatively, alveolar concentrations might be extrapolated
to some extent from partially equilibrated rebreathing samples,
using numerical parameter estimation schemes for reconstructing VOC exhalation kinetics according to a given model structure. While further validation and data gathering needs to be
carried out before such estimates can become practically relevant, the mechanistic descriptions discussed in this paper are
intended as a first step towards achieving this goal.
Figure 3: Simulation of a sequential rebreathing protocol according
to O’Hara et al. (2008) with intermediate pauses of 10 min characterized by
free tidal breathing. The underlying model parameters correspond to the individual fit in Fig. 2. Dash-dotted lines represent upper and lower bounds for the
bag concentration as well as for the observable blood-bag ratio with respect to
changes in airway temperature T̄ . These bounds were obtained by assuming
that the airway temperature either instantaneously rises to body core temperature during the individual rebreathing periods or remains constant (i.e., at its
initial level T̄ 0 ) throughout the entire experiment.
4. Conclusions
Here we have successfully applied a previously published
compartment model for the exhalation kinetics of highly
soluble, blood-borne VOCs to the experimental framework
of isothermal rebreathing. The proposed model has proven
sufficiently flexible for capturing the associated end-tidal breath
dynamics of acetone, which can be viewed as a prototypical
test compound within this context. Moreover, it gives new
Acknowledgements
The research leading to these results has received funding from the European Communitys Seventh Framework Pro7
respectively. Here,
gramme (FP7/2007-13) under grant agreement No. 217967.
We appreciate funding from the Austrian Federal Ministry
for Transport, Innovation and Technology (BMVIT/BMWA,
Project 818803, KIRAS). Gerald Teschl and Julian King acknowledge support from the Austrian Science Fund (FWF) under Grant No. Y330. We greatly appreciate the generous support of the government of Vorarlberg and its governor Landeshauptmann Dr. Herbert Sausgruber.
Cv̄ := qliv λb:livCliv + (1 − qliv )λb:tisCtis
(A.5)
Ca := (1 − qbro )λb:airCA + qbro z(T̄ )λb:airCbro
(A.6)
and
are the associated concentrations in mixed venous and arterial
blood, respectively.
Appendix A. Model equations and nomenclature
Parameter
This appendix serves to give a roughly self-contained outline
of the model structure sketched in Fig. 1 (King et al., 2011).
The time evolution of VOC concentrations is captured by taking into account standard conservation of mass laws for the individual compartments.
Local diffusion equilibria are assumed to hold at the airtissue, tissue-blood and air-blood interfaces, the ratio of the corresponding concentrations being described by the appropriate
partition coefficients λ, e.g., λb:air . Unlike for low blood soluble
compounds, the amount of highly soluble gas dissolved in the
local blood volume of perfused compartments cannot generally
be neglected, as it might significantly increase the corresponding compartmental capacities. This is particularly true for the
airspace compartments. We hence use the effective compartment volumes Ṽbro := Vbro + Vmuc λmuc:air , ṼA := VA + Vc′ λb:air ,
Ṽliv := Vliv + Vliv,b λb:liv as well as Ṽtis := Vtis and neglect blood
volumes only for the mucosal and tissue compartment.
Values for the individual compartment volumes,
temperature-dependent partition coefficients, and physiological variables such as cardiac output (Q̇c ) and alveolar
ventilation (V̇A ) are given in Table A.1. Fractional blood flows
to the bronchial tract (qbro ) and the liver (qliv ) as well as the
endogenous production (kpr ) and metabolic elimination rates
(kmet ) are treated as constants in the context of isothermal
rebreathing.
According to Fig. 1, the mass balance equation for the
bronchial compartment reads
dCbro
Ṽbro = V̇A (CI − Cbro ) + D(CA − Cbro )
dt
+ qbro Q̇c Ca − z(T̄ )λb:air Cbro ,
Bronchioles
Vbro
0.1 (l)a
Mucosa
Vmuc
0.005 (l)a
Alveoli
VA
4.1 (l)a
End-capillary
Vc′
0.15 (l)b
Liver
Vliv
0.0285 LBV (l)a
Blood liver
Vliv,b
1.1 (l)c
Tissue
Vtis
0.7036 LBV (l)a
Rebreathing bag
Ṽbag
3 (l)a
Respiratory parameters
Breathing frequency
f
12.5 (tides/min)d
Tidal volume
VT
0.593 (l)d
Alveolar ventilation
V̇A
6.2 (l/min)e
Cardiac output
Q̇c
6 (l/min) f
Fractional flow bronchioles
qbro
0.01g
Fractional flow liver
qliv
0.32a
λb:air
340h
Mucosa:air
λmuc:air
392i, j
Blood:liver
λb:liv
1.73 j
Blood:tissue
λb:tis
1.38h
Linear metabolic rate
kmet
0.0074 (l/kg0.75 /min)l
Endogenous production
kpr
0.19 (mg/min)l
Stratified conductance
D
0 (l/min)l
Blood flows
Partition coefficients (37◦ C)
Blood:air
Metabolic and diffusion constants
(A.1)
Constant Eq. (5)
kdiff,1
14.9 (min−1 )l
Constant Eq. (5)
kdiff,2
0.76l
Table A.1: Basic model parameters and nominal values during rest. LBV denotes the lean body volume in liters calculated according to LBV = −16.24 +
0.22 bh + 0.42 bw, with body height (bh) and weight (bw) given in cm and
kg, respectively (Mörk and Johanson, 2006); a (Mörk and Johanson, 2006);
b (Hughes and Morell, 2001); c (Ottesen et al., 2004); d (Duffin et al., 2000);
e cf. Eq. (12); f (Mohrman and Heller, 2006); g (Lumb, 2005); h (Anderson et al.,
2006); i (Staudinger and Roberts, 2001); j (Kumagai and Matsunaga, 1995);
l (King et al., 2011).
dCA
ṼA = D(Cbro − CA ) + (1 − qbro )Q̇c Cv̄ − λb:air CA , (A.2)
dt
and
(A.3)
References
and
dCtis
Ṽtis = (1 − qliv )(1 − qbro )Q̇c Ca − λb:tisCtis ,
dt
Nominal value (units)
Compartment volumes
where CI denotes the inhaled (ambient/bag) gas concentration.
Similarly, for the alveolar, liver and tissue compartment we find
that
dCliv
Ṽliv = kpr − kmet λb:livCliv
dt
+ qliv (1 − qbro )Q̇c Ca − λb:livCliv ,
Symbol
Amann, A., Smith, D. (Eds.), 2005. Breath Analysis for Clinical Diagnosis and
Therapeutic Monitoring. World Scientific, Singapore.
(A.4)
8
Amann, A., Spanel, P., Smith, D., 2007. Breath analysis: the approach towards
clinical applications. Mini reviews in Medicinal Chemistry 7, 115–129.
Anderson, J. C., Babb, A. L., Hlastala, M. P., 2003. Modeling soluble gas exchange in the airways and alveoli. Ann. Biomed. Eng. 31, 1402–22.
Anderson, J. C., Hlastala, M. P., 2007. Breath tests and airway gas exchange.
Pulm. Pharmacol. Ther. 20, 112–117.
Anderson, J. C., Hlastala, M. P., 2010. Impact of Airway Gas Exchange on the
Multiple Inert Gas Elimination Technique: Theory. Ann. Biomed. Eng. 38,
1017–1030.
Anderson, J. C., Lamm, W. J., Hlastala, M. P., 2006. Measuring airway exchange of endogenous acetone using a single-exhalation breathing maneuver. J. Appl. Physiol. 100, 880–9.
Duffin, J., 2010. The role of the central chemoreceptors: a modeling perspective. Respir. Physiol. Neurobiol. 173, 230–243.
Duffin, J., Mohan, R. M., Vasiliou, P., Stephenson, R., Mahamed, S., 2000. A
model of the chemoreflex control of breathing in humans: model parameters
measurement. Respir. Physiol. 120, 13–26.
Farhi, L. E., 1967. Elimination of inert gas by the lung. Respir. Physiol. 3, 1–11.
Filser, J. G., 1992. The closed chamber technique – uptake, endogenous production, excretion, steady-state kinetics and rates of metabolism of gases
and vapors. Arch. Toxicol. 66, 1–10.
Grodins, F. S., Buell, J., Bart, A. J., 1967. Mathematical analysis and digital
simulation of the respiratory control system. J. Appl. Physiol. 22, 260–76.
Hanna, L. M., Scherer, P. W., 1986. A theoretical model of localized heat and
water vapor transport in the human respiratory tract. J. Biomech. Eng. 108,
19–27.
Hlastala, M. P., Scheid, P., Piiper, J., 1981. Interpretation of inert gas retention
and excretion in the presence of stratified inhomogeneity. Respir. Physiol.
46, 247–259.
Hughes, J. M. B., Morell, N. W., 2001. Pulmonary Circulation. From basic
mechanisms to clinical practice. Imperial College Press, London.
Jones, A. W., 1983. Role of rebreathing in determination of the blood-breath
ratio of expired ethanol. J. Appl. Physiol. 55, 1237–1241.
Kalapos, M. P., 2003. On the mammalian acetone metabolism: from chemistry
to clinical implications. Biochim. Biophys. Acta 1621, 122–39.
King, J., Koc, H., Unterkofler, K., Mochalski, P., Kupferthaler, A., Teschl, G.,
Teschl, S., Hinterhuber, H., Amann, A., 2010a. Physiological modeling of
isoprene dynamics in exhaled breath. J. Theor. Biol. 267, 626–637.
King, J., Kupferthaler, A., Unterkofler, K., Koc, H., Teschl, S., Teschl, G.,
Miekisch, W., Schubert, J., Hinterhuber, H., Amann, A., 2009. Isoprene and
acetone concentration profiles during exercise on an ergometer. J. Breath
Res. 3, 027006 (16pp).
King, J., Mochalski, P., Kupferthaler, A., Unterkofler, K., Koc, H., Filipiak,
W., Teschl, S., Hinterhuber, H., Amann, A., 2010b. Dynamic profiles of
volatile organic compounds in exhaled breath as determined by a coupled
PTR-MS/GC-MS study. Physiol. Meas. 31, 1169–1184.
King, J., Unterkofler, K., Teschl, G., Teschl, S., Koc, H., Hinterhuber, H.,
Amann, A., 2011. A mathematical model for breath gas analysis of volatile
organic compounds with special emphasis on acetone. J. Math. Biol., DOI
10.1007/s00285–010–0398–9.
Koc, H., King, J., Teschl, G., Unterkofler, K., Teschl, S., Mochalski, P., Hinterhuber, H., Amann, A., 2011. The role of mathematical modeling in voc
analysis using isoprene as a prototypic example. J. Breath Res., accepted.
Kumagai, S., Matsunaga, I., 1995. Physiologically based pharmacokinetic
model for acetone. Occup. Environ. Med. 52, 344–352.
Kumagai, S., Matsunaga, I., 2000. A lung model describing uptake of organic
solvents and roles of mucosal blood flow and metabolism in the bronchioles.
Inhal. Toxicol. 12, 491–510.
Lumb, A. B., 2005. Nunn’s Applied Respiratory Physiology, 6th Edition.
Butterworth-Heinemann, Oxford.
Miekisch, W., Schubert, J. K., Noeldge-Schomburg, G. F. E., 2004. Diagnostic potential of breath analysis–focus on volatile organic compounds. Clin.
Chim. Acta 347, 25–39.
Mohan, R., Duffin, J., 1997. The effect of hypoxia on the ventilatory response
to carbon dioxide in man. Respir. Physiol. 108, 101–115.
Mohrman, D. E., Heller, L. J., 2006. Cardiovascular Physiology, 6th Edition.
Lange Medical Books/McGraw-Hill, New York.
Mörk, A. K., Johanson, G., 2006. A human physiological model describing acetone kinetics in blood and breath during various levels of physical exercise.
Toxicol. Lett. 164, 6–15.
O’Hara, M. E., Clutton-Brock, T. H., Green, S., Mayhew, C. A., 2009. Endoge-
nous volatile organic compounds in breath and blood of healthy volunteers:
examining breath analysis as a surrogate for blood measurements. J. Breath
Res. 3, 027005 (10pp).
O’Hara, M. E., O’Hehir, S., Green, S., Mayhew, C. A., 2008. Development of
a protocol to measure volatile organic compounds in human breath: a comparison of rebreathing and on-line single exhalations using proton transfer
reaction mass spectrometry. Physiol. Meas. 29, 309–30.
Ohlsson, J., Ralph, D. D., Mandelkorn, M. A., Babb, A. L., Hlastala, M. P.,
1990. Accurate measurement of blood alcohol concentration with isothermal
rebreathing. J. Stud. Alcohol 51, 6–13.
Ottesen, J. T., Olufsen, M. S., Larsen, J. K., 2004. Applied Mathematical Models in Human Physiology. SIAM, Philadelphia.
Scheid, P., Hlastala, M. P., Piiper, J., 1981. Inert gas elimination from lungs
with stratified inhomogeneity: Theory. Respir. Physiol. 44, 299–309.
Schwarz, K., Filipiak, W., Amann, A., 2009a. Determining concentration patterns of volatile compounds in exhaled breath by PTR-MS. J. Breath Res. 3,
027002 (15pp).
Schwarz, K., Pizzini, A., Arendacka, B., Zerlauth, K., Filipiak, W., Schmid,
A., Dzien, A., Neuner, S., Lechleitner, M., Scholl-Burgi, S., Miekisch, W.,
Schubert, J., Unterkofler, K., Witkovsky, V., Gastl, G., Amann, A., 2009b.
Breath acetone – aspects of normal physiology related to age and gender as
determined in a PTR-MS study. J. Breath Res. 3, 027003 (9pp).
Staudinger, J., Roberts, P. V., 2001. A critical compilation of Henry’s law
constant temperature dependence relations for organic compounds in dilute
aqueous solutions. Chemosphere 44, 561–576.
Tsoukias, N. M., George, S. C., 1998. A two-compartment model of pulmonary
nitric oxide exchange dynamics. J. Appl. Physiol. 85, 653–666.
Tsu, M. E., Babb, A. L., Sugiyama, E. M., Hlastala, M. P., 1991. Dynamics
of soluble gas exchange in the airways: 2. Effects of breathing conditions.
Respir. Physiol. 83, 261–276.
Wigaeus, E., Holm, S., Astrand, I., 1981. Exposure to acetone. Uptake and
elimination in man. Scand. J. Work Environ. Health 7, 84–94.
Zwart, A., Luijendijk, S. C., de Vries, W. R., 1986. Excretion-retention data of
steady state gas exchange in tidal breathing. I. Dependency on the blood-gas
partition coefficient. Pflügers Arch. 407, 204–210.
9