Intra-pulmonary shunt and pulmonary gas exchange during exercise in humans


Corresponding author M. K. Stickland: The John Rankin Laboratory of  Pulmonary Medicine, Department  of Population Health Sciences, University of  Wisconsin School of  Medicine, 1300 University Ave, Madison, WI 53706-1532, USA. Email:


In young, healthy people the alveolar–arterial Pmath image difference (A-aDO2) is small at rest, but frequently increases during exercise. Previously, investigators have focused on ventilation/perfusion mismatch and diffusion abnormalities to explain the impairment in gas exchange, as significant physiological intra-pulmonary shunt has not been found. The aim of this study was to use a non-gas exchange method to determine if anatomical intra-pulmonary (I-P) shunts develop during exercise, and, if so, whether there is a relationship between shunt and increased A-aDO2. Healthy male participants performed graded upright cycling to 90% inline image while pulmonary arterial (PAP) and pulmonary artery wedge pressures were measured. Blood samples were obtained from the radial artery, cardiac output inline image was calculated by the direct Fick method and I-P shunt was determined by administering agitated saline during continuous 2-D echocardiography. A-aDO2 progressively increased with exercise and was related to inline image (r= 0.86) and PAP (r= 0.75). No evidence of I-P shunt was found at rest in the upright position; however, 7 of 8 subjects developed I-P shunts during exercise. In these subjects, point bi-serial correlations indicated that I-P shunts were related to the increased A-aDO2 (r= 0.68), inline image (r= 0.76) and PAP (r= 0.73). During exercise, intra-pulmonary shunt always occurred when A-aDO2 exceeded 12 mmHg and inline image was greater than 24 l min−1. These results indicate that anatomical I-P shunts develop during exercise and we suggest that shunt recruitment may contribute to the widened A-aDO2 during exercise.

During aerobic exercise, there is typically an impairment in pulmonary gas exchange as demonstrated by an increase in the alveolar–arterial pressure difference for oxygen (A-aDO2). Depending on the ventilatory response and magnitude of A-aDO2, exercise-induced arterial hypoxemia can develop (Dempsey & Wagner, 1999). Abnormalities in pulmonary gas exchange could result from ventilation/perfusion inline image mismatch, diffusion impairment, right to left extra-pulmonary or intra-pulmonary (I-P) shunt. Current theories to explain the widened A-aDO2 during exercise include inadequate blood transit time in the lung (Dempsey et al. 1984), and a mismatch of inline image secondary to pulmonary hypertension (Wagner et al. 1986).

Intra-pulmonary shunt has been previously dismissed as an explanation for the increased A-aDO2 during exercise because oxygen breathing (Dempsey et al. 1984; Torre-Bueno et al. 1985; Hammond et al. 1986; Wagner et al. 1986) and the multiple inert gas elimination technique (MIGET; Hopkins et al. 1994; Dempsey & Wagner, 1999; Rice et al. 1999) consistently failed to detect significant right to left mixed-venous shunt. However, precapillary gas exchange has been documented in both humans (Jameson, 1963, 1964; Sobol et al. 1963) and cats (Conhaim & Staub, 1980). Conhaim & Staub (1980) reasoned that because of precapillary gas exchange, 100% O2 breathing underestimates shunt. Similarly, MIGET may underestimate I-P shunt during exercise if precapillary gas exchange occurs. Large arteriovenous vessels have been demonstrated in normal post-mortem human lungs (Tobin & Zariquiey, 1950; Tobin, 1966), and we previously questioned whether these arterial–venous anastamoses could act as shunt vessels during exercise (Stickland et al. 2002). Whyte et al. (1992) have previously documented an increase in shunt during exercise using technetium-99m labelled albumin microspheres in normal control subjects, and Eldridge et al. (2004) demonstrated I-P shunts during exercise with agitated saline contrast echocardiography. Accordingly, the purpose of this investigation was to confirm that I-P shunts occur during exercise and if so, determine the relationship to A-aDO2 and haemodynamic responses. We hypothesized that the recruitment of anatomical I-P shunts contributes to the widened A-aDO2 during exercise.


Research design

Institutional ethics review board approval was obtained and all participants provided written informed consent to participate. Three experimental sessions were completed during a 3 week period in the following order: a graded exercise test, a practice session, and the experimental day.


Nine healthy males (mean ±s.d., age: 29 ± 3.9 years, mass: 78.5 ± 6.0 kg) were initially recruited for participation in the study. Participants were free of exercise-induced bronchospasm, haematological abnormalities and ECG abnormalities. All subjects were physically active (mean inline image: 4.20 ± 0.6 l min−1, 53.7 ± 9.0 ml kg−1 min−1), and the sample included several recreational and competitive endurance athletes. During the study, one subject was found to develop a right to left intra-cardiac shunt with exercise. His data were removed, and therefore we report the results of eight subjects (mean age: 30 ± 3.9 years, inline image: 4.28 ± 0.6 l min−1, 54.7 ± 9.0 ml kg−1 min−1).

Day 1. Graded-exercise test

Subjects performed a graded-exercise test to determine ventilatory threshold (VT) and inline image. Exercise was performed on an electrically braked Seimens 740E ergometer (Malvern, PA, USA) with a custom-built seat and back rest to limit torso movement. Respiratory gas exchange data were collected continuously using a non-rebreathing valve (Hans-Rudolph, 2700, Kansas City MO, USA) and a metabolic measurement system (ParvoMedics, Truemax, Salt Lake City, UT, USA) while heart rate was recorded using a telemetry system (Polar, Kempele, Finland). The criteria for VT was a non-linear increase in the inline image (Wasserman, 1987) ratio. During the graded exercise test subjects were requested to self-select a consistent cadence between 70 and 90 r.p.m., while power output was increased 25 W every 2 min until exhaustion.

Day 2. Practice session

A practice session was conducted to familiarize each subject with the protocol. The set-up and exercise workloads were similar to those during the full experiment; however, blood sampling, echocardiography and invasive pressure measurements were not performed.

Day 3. Experimental trial

Subject preparation A radial artery catheter (20-gauge Angiocath; Becton-Dickson, Sandy, UT, USA) was inserted into the left radial artery using sterile techniques and local anaesthetic (Lidocaine HCl, 1%, Astra, Mississauga, ON, Canada). Thereafter, a Swan-Ganz catheter (Edwards Lifesciences; Irvine, CA, USA) was inserted through a standard Cordis sheath (7 French) via an antecubital vein and advanced under fluoroscopy to ensure proper placement. Patency of the catheters was maintained with a pressurized flush system of normal saline at a rate of 15 ml h−1. Following placement of the catheters, each subject rested quietly for 10 min before data collection.

Exercise protocol Data were collected both at rest and during graded exercise. At rest, supine data were collected first, each subject was then positioned on the cycle ergometer, pressure transducers were repositioned, and resting upright data were collected. Subsequently, the exercise protocol was conducted in the following order: (I) 75 W; (II) 150 W; (III) power output at VT; (IV) 25 W above VT; and (V) 90% of inline image. Five minute rest periods were given between workloads. Workloads III and IV were purposely selected to span VT as opposed to a percentage inline image because arterial pH is believed to affect pulmonary artery pressure, and intersubject variability in VT may have confounded the pulmonary pressure–cardiac output relationship if absolute workloads would have been selected (Schaffartzik et al. 1992). For workloads below 90% of inline image, data collection began after the first 2 min of each 5 min workload. At 90% inline image, data collection began once the target inline image was reached (typically 90 s), and the workload usually lasted about 3 min before the subject became fatigued. During the experimental session, subjects were encouraged to consume water, sports drinks (e.g. Gatorade) and food (e.g. Powerbars) and a fan was provided to avoid hyperthermia.

Respiratory gas-exchange measurement Respiratory gas exchange data were collected continuously during all conditions using the same system as for the graded-exercise test. Mean values from the final minute of sampling were used for subsequent analyses.

Blood-gas measurement Arterial blood samples (2–3 ml) drawn from the radial artery catheter and mixed venous samples drawn from the distal port of the Swan-Ganz catheter were immediately placed in ice water. Samples were later analysed for Pmath image, Pmath image, pH, haematocrit and haemoglobin (ABL 700 blood-gas analyser, Copenhagen, Denmark). Blood gases were corrected for pulmonary arterial temperature as measured by the Swan-Ganz catheter, with arterial and venous saturation (Smath image) corrected for temperature and pH.

Systemic and pulmonary pressures Systemic arterial blood pressure was measured from a pressure transducer attached to the radial arterial catheter, while mean pulmonary artery (PAP) and pulmonary artery wedge (PAWP) and right atrial pressures were obtained from the Swan-Ganz catheter. The pressure transducer was set at the level of the right atrium with the positioning monitored continuously. Pressure tracings were monitored constantly and recorded during the third and fourth minute of each workload. Mean pressures over at least three respiratory cycles are reported (Higginbotham et al. 1986; Wagner et al. 1986; Groves et al. 1987).

Contrast echocardiography Echocardiograms were performed by the same experienced sonographer using cardiac ultrasound (Sonos 5500, Hewlett Packard, Andover, MA, USA). The agitated saline contrast echocardiography technique was used to detect intra-cardiac and I-P shunt. Standard procedures were employed for injection of the solution (Weyman, 1994). Briefly, 10 ml of saline was combined with 0.5 ml of air and the solution forcefully agitated through a three-way stop-cock between two syringes to form fine suspended bubbles which are generally much larger than the pulmonary capillaries (Weyman, 1994). The solution was then injected through the proximal port of the Swan-Ganz catheter during the third minute of each 5 min workload. Concurrently, all four chambers of the heart were imaged and recorded onto a VHS tape. The presence of intra-cardiac shunt is determined by contrast appearance in the left ventricle in less than five heart beats, while with I-P shunt, contrast appearance in the left ventricle occurs after at least five heart beats (Weyman, 1994). This procedure has been performed repeatedly during exercise without any reported complication (Himelman et al. 1989).

At a later date, the images were reviewed by a cardiologist with substantial experience in echocardiography. The cardiologist was naïve as to the exact condition; however, he did have an indication of exercise intensity due to heart rate data. Each injection was categorized into one of (Weyman, 1994): (1) No shunt: visible contrast injection into the right ventricle, no contrast in the left ventricle; (2) I-P shunt: visible contrast injection into the right ventricle, and visible contrast in the left ventricle following at least five heart beats; (3) Intra-cardiac shunt: visible contrast injection into the right ventricle, and visible contrast in the left ventricle in less than five heart beats; (4) Inconclusive: no visible contrast injection into the right ventricle or no consistently visible left ventricle on echocardiogram. To evaluate intra-observer reliability, a total of 25 images were re-analysed from five randomly selected subjects by the same cardiologist after a time delay of at least 2 months. In all cases, the injections were classified into the same category as when first analysed (100% repeatability).

Calculations Cardiac output was calculated from the Fick equation. Systemic vascular resistance was determined as the difference between mean arterial pressure and mean right atrial pressure divided by inline image. Similarly, pulmonary vascular resistance was calculated as the difference between mean PAP and mean PAWP divided by inline image (West, 2000b).

Statistical analysis Group data for each dependent variable were analysed with a one-way ANOVA with repeated measures. To protect against violation of the sphericity assumption, the Geisser-Greenhouse conservative F test procedure was used (for details see Kirk, 1982). When the workload main effect was statistically significant, the Tukey honestly significant difference (HSD) procedure was utilized to compare workloads. The Tukey HSD procedure compared the means of each workload against the mean in the upright condition resulting in a total of six comparisons for each variable. For all inferential analyses, the probability of Type I error was set at 0.05.

Intra-individual Pearson-product-moment-correlation coefficients were calculated to describe the strength of relationships amongst the three continuous variables that were measured at least at the interval level (i.e. PAP, inline image, A-aDO2). To evaluate the strength of relationships between the three continuous variables and I-P shunt (a dichotomous variable), intra-individual point bi-serial correlation coefficients were calculated. The results from the correlation analysis are summarized in Tables 3 and 4. Of the 64 saline contrast echocardiography injections, two were non-diagnostic. These time points were excluded from correlational analyses.

Table 3.  Mean, standard deviation and range of within-subject correlation coefficients between alveolar-arterialPmath imagedifference and cardiac output and mean pulmonary artery pressure (n= 8)
  1. Highest and lowest values reported in parentheses.

Cardiac output0.860.100.25
Mean pulmonary artery pressure0.750.180.48
Table 4.  Mean, standard deviation and range of within-subject point bi-serial correlation coefficients between intra-pulmonary shunt as evaluated by agitated saline contrast echocardiography and alveolar-arterialPmath imagedifference (A-aDO2), cardiac output and mean pulmonary artery pressure (n= 7)
  1. Highest and lowest values reported in parentheses. The subject who did not develop intra-pulmonary shunt was excluded from analysis as no variance in shunt was present and therefore correlation coefficients between shunt and cardio-pulmonary variables could not be determined.

Cardiac output0.760.120.31
Mean pulmonary artery pressure0.730.190.50


Exercise haemodynamics

Group values for exercise haemodynamics are presented in Table 1. Stroke volume, PAP and PAWP decreased when the subjects went from the supine to the upright position (Table 1, Fig. 1). Exercise resulted in significant increases in inline image, heart rate, stroke volume, PAP and PAWP while pulmonary and systemic vascular resistance decreased.

Table 1.  Mean haemodynamic responses at rest (supine and upright) and during graded exercise (n= 8)
  1. Note: standard error values are in parentheses. SUP, supine; UP, upright; Level I–V, exercise load; inline image, oxygen consumption. *P < 0.05 versus upright at rest (UP)

Power output75150211235293
inline image0.360.411.59*2.49*3.22*3.69*4.08*
(l min−1)(0.01)(0.02)(0.05)(0.07)(0.21)(0.23)(0.21)
Cardiac output7.86.116.6*22.2*25.1*26.9*29.7*
(l min−1)(0.5)(0.4)(0.6)(0.7)(1.0)(1.6)(1.6)
Stroke volume124*94147*152*151*149*159*
Heart rate6465113*147*166*180*187*
(beats min−1)(3)(3)(3)(4)(5)(3)(2)
Mean arterial pressure99109120*126*130*132*136*
Systemic vascular resistance12.1*18.17.1*5.7*5.1*4.9*4.4*
(mmHg l−1 min−1)(0.8)(1.4)(0.4)(0.4)(0.3)(0.5)(0.4)
Pulmonary vascular resistance0.6220.8410.6190.495*0.442*0.378*0.390*
(mmHg l−1 min−1)(0.08)(0.17)(0.08)(0.06)(0.05)(0.05)(0.05)
Figure 1.

Mean (±s.e.m.) pulmonary artery pressure and pulmonary artery wedge pressure at rest (supine and upright) and during graded exercise (n= 8)
PAP, mean pulmonary artery pressure; PAWP, pulmonary artery wedge pressure; SUP, supine; UP, upright; Level I–V, exercise load; *P < 0.05 versus upright at rest.

Pulmonary gas exchange and I-P shunt

As expected, A-aDO2 increased and Smath image decreased with progressive exercise (Fig. 2). Mean within-subject correlations of all eight participants demonstrate that A-aDO2 was related to inline image (r= 0.86) and PAP (r= 0.75) (Table 3).

Figure 2.

Mean (±s.e.m.) pulmonary gas exchange at rest (supine and upright) and during graded exercise (n= 8)
Smath image, arterial saturation; Pmath image, arterial Pmath image; A-aDO2, alveolar-arterial Pmath image difference; SUP, supine; UP, upright; Level I–V, exercise load; *P < 0.05 versus upright at rest.

Two subjects had evidence of I-P shunt when supine at rest; however, none had shunt in the upright position (Table 2). With incremental exercise seven of eight subjects developed I-P shunt. In the seven subjects who developed shunt, mean within-subject point bi-serial correlations show that shunt development was related to A-aDO2 (r= 0.68), inline image (r= 0.76), and PAP (r= 0.73) (Table 4).

Table 2.  Frequency of shunt as assessed by agitated saline contrast echocardiography at rest (supine and upright) and during graded exercise (n= 8)
  1. SUP, supine; UP, upright; Level I–V, exercise load.

No shunt6843111
Intra-pulmonary shunt2035767


During exercise, seven of eight subjects developed positive agitated saline contrast echocardiograms as determined by the appearance of contrast bubbles in the left ventricle five or more heart beats after the arrival of contrast bubbles in the right ventricle. Notwithstanding the potential limitations of this technique during exercise (see below), these results indicate that healthy males develop I-P shunt with incremental exercise. I-P shunt and the increased A-aDO2 during exercise were strongly related to both inline image and to a lesser extent PAP (Tables 3 and 4). The occurrence of I-P shunt was related to A-aDO2 (mean intra-subject correlation, 0.68) but in some subjects A-aDO2 increased slightly without evidence for I-P shunt, while in others, I-P shunt occurred without an increase in A-aDO2 (see Figs 3 and 4). This variability could be due to the contribution of inline image matching and diffusion limitation to A-aDO2 (Wagner et al. 1986; Hopkins et al. 1994; Rice et al. 1999) and the inability of the agitated saline technique to quantify shunt. However it is important to note that during exercise, I-P shunt always occurred when A-aDO2 exceeded 12 mmHg and inline image was greater than 24 l min−1. Based on our results, we hypothesize that the recruitment of anatomical I-P shunts during exercise contributes to the impairment in pulmonary gas exchange.

Figure 3.

Individual relationships during upright rest and graded exercise between cardiac output and alveolar-arterialPmath imagedifference upright at rest and during graded exercise
A-aDO2, alveolar-arterial Pmath image difference. Arrow denotes peak work rate for the subject who did not develop I-P shunt during exercise.

Figure 4.

Individual relationships during upright rest and graded exercise between mean pulmonary artery pressure and alveolar-arterialPmath imagedifference
A-aDO2, alveolar-arterial Pmath image difference.

Anatomical shunt

The recruitment of I-P shunt during exercise is consistent with the documentation of I-P shunt in lungs of previously healthy human cadavers (Tobin & Zariquiey, 1950; Tobin, 1966) and isolated lung models (Rahn et al. 1952; Irwin et al. 1954). These anatomical shunt vessels can be in excess of 500 µm in diameter and they seem to predominate in the lung apices (Tobin & Zariquiey, 1950), which is congruent with the findings that shunt increases with an elevation in pulmonary artery pressure in dogs (Cheney et al. 1978) and with increases in inline image in humans (Muneyuki et al. 1971) and dogs (Berk et al. 1977; Bishop & Cheney, 1983). In addition, Sykes et al. (1970) demonstrated a widening A-aDO2 with hypervolaemia-induced increases in pulmonary artery pressure in resting dogs and ascribed this to right to left shunts in the lung. Our results are consistent with several previous investigations that showed I-P shunt during exercise (Whyte et al. 1992; Eldridge et al. 2004) and conditions of increased inline image and/or increased PAP (Sykes et al. 1970; Muneyuki et al. 1971; Berk et al. 1977; Cheney et al. 1978; Bishop & Cheney, 1983). In addition to the hypothesized contribution to impaired gas exchange, the evidence supporting I-P shunt raises questions about the effectiveness of the lungs as a biological filter during exercise.

According to classic theory on capillary recruitment (West et al. 1964), increasing pulmonary blood flow increases pulmonary microvascular pressure, and at some critical flow the resulting microvascular pressure leads to recruitment of additional pulmonary capillaries. We propose a similar explanation for the recruitment of I-P shunts during exercise; specifically that at some critical flow, arterial–venous vessels open and I-P shunt occurs. West et al. (1964) suggested that during exercise the additional kinetic energy of the blood, which is not reflected in the pressure measured lateral to flow, increases capillary recruitment and, as we suggest, I-P shunts. Whether shunts are recruited through flow-induced increases in microvascular pressure, or kinetic energy, the dominant factor for recruitment would be flow, thus explaining the stronger relationship between inline image with I-P shunt and A-aDO2 (see Figs 3 and 4). This hypothesis is supported by data from the subject who had the lowest inline image and A-aDO2 during exercise, as he did not develop I-P shunts despite very high pulmonary artery pressures.

Right ventricular afterload

Based on Poiseuille's law, an increase in vessel diameter would decrease the driving pressure needed to maintain flow. Berk et al. (1977) suggested that I-P shunts act as ‘pop-off valves’ in response to increases in flow and pulmonary vascular resistance. Exercise-induced pulmonary shunts may be an adaptive mechanism to reduce the potential damaging effects of high perfusion pressures during exercise (West, 2000a). Alternately, the development of I-P shunt may be a beneficial response to reduce right-ventricular afterload. Whyte et al. (1993) postulated that the higher inline image during exercise in patients with pulmonary arteriovenous malformations (direct right to left shunt vessels 23–45 µm in diameter) was the direct result of the lower pulmonary vascular resistance caused by these shunt vessels. We do not know the diameter or the length of the anatomical shunts demonstrated in this study, and are unsure if flow is laminar through these vessels, therefore it is impossible to calculate the impact of their recruitment on right ventricular afterload. Figure 5 illustrates that the single subject who did not develop shunt during exercise had, at a cardiac output of 20 l min−1, a PAP that was approximately 10 mmHg higher than the average PAP for the other seven subjects at the same inline image. This represents a substantial unloading of the right ventricle with I-P shunt recruitment. However, if we assume all anatomical shunt is also true physiological shunt (which may not be correct – see below), the estimated I-P shunt from the widened A-aDO2 is small and would have a correspondingly small affect on mean PAP. The impact of I-P shunt on right ventricular afterload is unclear, and it remains to be determined if the development of I-P shunt is a consequence of, or a requirement for, a high cardiac output.

Figure 5.

Individual relationships during upright rest and graded exercise between cardiac output and mean pulmonary artery pressure

Previous studies and gas exchange methods to measure shunt

Previous research has failed to detect significant right to left shunt during exercise with oxygen breathing (Dempsey et al. 1984; Torre-Bueno et al. 1985; Hammond et al. 1986; Wagner et al. 1986). However, Conhaim & Staub (1980) demonstrated that precapillary pulmonary arterial oxygenation occurs in the small arteries of isolated cat lungs. The small pulmonary arteries (100 µm) were found to take up oxygen from the alveoli during ventilation with room air, while blood in larger diameter arteries (400–500 µm) was completely oxygenated during 100% O2 ventilation. Similarly, rapid increases in oxygen and hydrogen in the pulmonary artery have been detected with increasing fractional inspired O2 and H2 in humans via the ports of a pulmonary arterial catheter (Sobol et al. 1963; Jameson, 1963, 1964). These changes occur rapidly (0.4–0.7 s) following a change in inhaled gas and precede the arrival of these gases in the descending aorta, arguing against the possibility of a bronchial arterial source for the increased pulmonary artery gas concentration (Sobol et al. 1963; Jameson, 1964). As pointed out by Conhaim & Staub (1980), in the presence of a large Pmath image gradient such as what would occur when breathing 100% O2, precapillary vessels up to 500 µm are fully oxygenated. However, with the removal of this large pressure gradient during normoxic breathing, these vessels may not take part in gas exchange. As a result, the ‘unphysiological’ state of 100% O2 breathing may underestimate arterial–venous shunts during normoxia. Of note, Genovesi et al. (1976) used similar logic to explain why I-P shunt calculated by oxygen breathing is −99m lower than anatomical shunt calculated using technicium-labelled albumin macroaggregates (Genovesi et al. 1976; Davis et al. 1978; Whyte et al. 1998). It is not unreasonable to expect that if small pulmonary arteries exchange oxygen then anatomical shunt vessels could do the same, and our speculation that I-P shunts act as lower resistance channels to unload the right ventricle could be more important than the modest increase in A-aDO2 would suggest.

Intra-pulmonary shunt during exercise has not been detected with MIGET (Hopkins et al. 1994; Dempsey & Wagner, 1999; Rice et al. 1999). However we hypothesize that precapillary gas exchange (Conhaim & Staub, 1980) may also impact MIGET. Specifically, the inert gases may be excreted from precapillary (or shunt) vessels and as a result, true physiological intra-pulmonary shunt may not be recorded with MIGET, although anatomical arterial–venous shunts which prevent full O2 diffusion during normoxic exercise may exist. Interestingly, diffusion limitation as measured by MIGET typically develops above an oxygen consumption of 2.5 l min−1 (Hammond et al. 1986; Wagner et al. 1986; Hopkins et al. 1994; Rice et al. 1999), and our results indicated that I-P shunt is also most common above this intensity. Clearly, our results documenting anatomical I-P shunts go against the current understanding of pulmonary gas exchange during exercise and require further research. Comparisons are needed between non-gas-dependent methods to quantify shunt, such as radio-labelled microspheres, and gas exchange methods such as MIGET to determine the relationship between anatomical shunt, inline image mismatch and diffusion limitation during exercise.

Resting data

Strauss et al. (1969) and Whyte et al. (1992) have documented a small amount of shunt with albumin microspheres in the supine position in normal resting humans. In the present study, two subjects had I-P shunt while supine, which disappeared once the subjects sat upright. Our haemodynamic data do not explain why these two subjects shunted when supine; however, this would be consistent with shunt vessels located predominately in the apex of the lung (Tobin & Zariquiey, 1950), which would be more likely recruited in the supine position.

Limitations of agitated saline contrast echocardiography

The conclusions that can be drawn from this paper are dependent on the predictive value of the agitated saline technique in detecting pulmonary arterial–venous shunts. Agitated saline contrast echocardiocraphy is a standard technique (Weyman, 1994) but it is not typically used during exercise. The technique produces stable air bubbles which are said to be much larger than the pulmonary capillaries (Weyman, 1994); however, a major limitation is that the exact size of the contrast bubbles produced are unknown. Furthermore, the agitated saline technique does not quantify shunt and the current method of gas exchange analysis could not assess inline image matching or diffusion limitation. The biological significance of anatomical I-P shunt is unknown, and it is possible that the majority of A-aDO2 during exercise is due to inline image mismatch and/or diffusion limitation (Wagner et al. 1986; Hopkins et al. 1994; Rice et al. 1999). In addition to I-P shunt, the appearance of saline contrast bubbles in the left ventricle could be due to other factors including: (1) small diameter bubbles which are able to pass through normal capillaries, (2) deformation of larger bubbles and their transit through the pulmonary capillaries, and (3) capillary distention. None of these possibilities can be conclusively excluded with the current methodology.

Bubbles less than 10 µm in diameter would transverse the pulmonary capillaries. Meltzer et al. (1980) estimated that the survival time for these bubbles is less than 200 ms, and mean whole-lung transit time in well-trained endurance athletes has been shown to exceed 2 s at intensities above 90% of inline image (Hopkins et al. 1996; Zavorsky et al. 2002). As well, bubble dissolution is greater with increasing fluid pressure (Tsujino & Shima, 1980) and increasing flow velocity (Yang et al. 1971), both of which occur during incremental exercise. Therefore, it is unlikely that contrasts entering the left ventricle during exercise would be the result of small diameter bubbles passing through the pulmonary capillaries.

Butler & Hills (1979) suggested that passage through the pulmonary circulation of a deformable gas may be different from that of solid particles. Indeed the potential that large diameter bubbles are able to deform and pass through pulmonary capillaries during exercise cannot be excluded. Meltzer et al. (1981) reported that agitated saline bubbles passed through the pulmonary circulation at rest when the distal port on a wedged Swan-Ganz catheter was used for injection. Roeland (1982) later found that an injection pressure of 300 mmHg through a firmly wedged catheter was needed to observe contrasts in the left ventricle, and suggested that extremely high injection pressures may cause deformation of the contrast bubbles resulting in their passage through the capillary bed. Our injections were made through the promixal (right atrial) port of an unwedged catheter, and mean individual PAP did not exceed 35 mmHg at peak exercise. Himelman et al. (1989) failed to document any positive saline contrast echocardiograms during exercise in pulmonary patients with peak pulmonary arterial systolic pressures of 80 mmHg, suggesting that deformed bubbles secondary to high pulmonary vascular pressures are unlikely to explain the positive echocardiographs in our study.

Increases in perfusion pressure could distend the pulmonary capillaries (Sobin et al. 1972), allowing for contrasts to pass through the pulmonary circulation. However, Glazier et al. (1969) have shown that the diameter of pulmonary capillaries does not increase above 13 µm, despite capillary pressures up to 100 cmH2O. Therefore, capillary distention is an unlikely explanation for the appearance of contrasts in the left ventricle.


Normal, healthy male subjects developed anatomical I-P shunts during exercise, as evaluated by the agitated saline contrast echocardiography technique. Both I-P shunt and the widened A-aDO2 with incremental exercise were related to inline image and, to a lesser extent, PAP. The occurrence of I-P shunts was associated with an elevated A-aDO2 and we hypothesize that these anatomical shunts contribute to the impairment in pulmonary gas exchange observed during exercise. Further research is needed as current methods do not allow for the quantification of shunt and our observations contradict a large body of gas exchange research which has not documented significant physiological shunt during exercise. It remains undetermined if these intra-pulmonary shunts are a consequence of, or a requirement for, a high cardiac output.



The authors wish to thank the subjects for their enthusiastic participation. The technical support of Allen McLean, Celeste Afonso, Rob Sawyer, Reginald Nugent and Matthew Koller are greatly appreciated. M.K. Stickland was supported by a scholarship from the Natural Sciences and Engineering Research Council of Canada.