Transpulmonary passage of 99mTc macroaggregated albumin in healthy humans at rest and during maximal exercise

Andrew T. Lovering, Hans C. Haverkamp, Lee M. Romer, John S. Hokanson, Marlowe W. Eldridge


We have demonstrated that 50-μm-diameter arteriovenous pathways exist in isolated, healthy human and baboon lungs, ventilated and perfused under physiological pressures. These findings have been confirmed and extended by demonstrating the passage of 25-μm microspheres through the lungs of exercising dogs, but not at rest. Determination of blood flow through these large-diameter intrapulmonary arteriovenous pathways would be an important first step to establish a physiological role for these vessels. Currently, we sought to estimate blood flow through these arteriovenous pathways using technetium-99m (99mTc)-labeled macroaggregated albumin (MAA) in healthy humans at rest and during maximal treadmill exercise. We hypothesized that the percentage of 99mTc MAA able to traverse the pulmonary circulation (%transpulmonary passage) would increase during exercise. Seven male subjects without patent foramen ovale were injected with 99mTc MAA at rest on 1 day and during maximal treadmill exercise on a separate day (>6 days). Within 5 min after injection, subjects began whole body imaging in the supine position. Six of the seven subjects showed an increase in transpulmonary passage of MAA with maximal exercise. Using two separate analysis methods, percent transpulmonary passage significantly increased with exercise from baseline to absolute values of 1.2 ± 0.8% (P = 0.008) and 1.3 ± 1.0% (P = 0.016), respectively (means ± SD; paired t-test). We conclude that MAA may be traversing the pulmonary circulation via large-diameter intrapulmonary arteriovenous conduits in healthy humans during exercise. Recruitment of these pathways may divert blood flow away from pulmonary capillaries during exercise and compromise the lung's function as a biological filter.

  • lung
  • contrast echocardiography
  • cryptogenic stroke
  • pulmonary gas exchange

arteriovenous anastomoses exist in almost every vascular bed in the human body, including the lung (10, 49). Although the existence of arteriovenous anastomoses in the lung of multiple mammalian species is well documented (31, 33, 45, 54), the conditions under which they have been documented have not been physiological. Furthermore, the physiological and/or pathophysiological roles for these vascular pathways remain controversial (17, 24, 38, 47).

Recently, our laboratory has demonstrated the transpulmonary passage (passage from pulmonary arterial circulation to pulmonary venous circulation) of large-diameter (50 μm) microspheres in isolated, ventilated, and perfused healthy human and baboon lungs under physiological conditions (25). We have further confirmed and expanded our laboratory's previous work by demonstrating the transpulmonary passage of large-diameter microspheres (25 μm) in healthy dogs during exercise, but not at rest (39). If pulmonary capillaries are 7–10 μm in diameter (50, 51), the only explanation for the 25- and 50-μm microspheres being able to travel through the pulmonary circulation would be through large-diameter arteriovenous anastomoses. These data, in combination with our laboratory's previous saline contrast echocardiography studies in healthy humans during exercise, provide good evidence for the existence of inducible intrapulmonary arteriovenous pathways (14, 40). However, it is unknown to what degree, if any, gas exchange occurs in these intrapulmonary arteriovenous pathways.

Although the idea of an inducible intrapulmonary arteriovenous pathway is an intriguing finding, the amount of blood flowing through these pathways during exercise needs to be determined to establish a physiological role for these vessels. Radiolabeled albumin macroaggregates and microspheres, in conjunction with gamma camera imaging, have been used for the detection and estimation of transpulmonary passage of solid large-diameter particles via large-diameter pathways (such as arteriovenous malformations) in patients (26, 41, 42, 53) and healthy humans during submaximal exercise (53). Accordingly, we used technetium-99m (99mTc)-labeled macroaggregated albumin (MAA) in conjunction with gamma camera imaging to detect and estimate the percentage of 99mTc MAA that are able to traverse the pulmonary circulation (%transpulmonary passage) in healthy humans at rest and during maximal treadmill exercise. We hypothesized that the percentage of transpulmonary passage of 99mTc MAA would be increased with maximal exercise.



This study received approval from the University of Wisconsin-Madison Human Subjects Committee, and each subject gave written, informed consent before participation. Eleven healthy, nonsmoking male volunteers, aged 24–49 yr, were recruited. A screening cardiopulmonary history and physical examination were performed, including resting saline contrast transthoracic echocardiography, as previously described (14). Four subjects (36%) had a patent foramen ovale and were excluded from further study. The remaining subjects (n = 7) completed the experimental studies outlined below. Four of the seven subjects (subjects 1–3 and 5) participating in this study were known to demonstrate transpulmonary passage of saline contrast bubbles during cycle exercise, as detected by saline contrast echocardiography.

Pulmonary Function and Lung Diffusion Capacity for Carbon Monoxide Testing

Baseline pulmonary function was determined using computerized spirometry (Pulmonizer model PFT 3000, Med Science, St. Louis, MO), according to American Thoracic Society standards (1). Lung diffusion capacity for carbon monoxide was determined by a single-breath, breath-holding method, according to American Thoracic Society standards (2). Values are reported in Table 1, percent predicted, as previously described (20).

View this table:
Table 1.

Anthropometric, exercise, and pulmonary function data

Maximal Exercise Test

To determine the subject's maximal oxygen consumption (V̇o2 max) in normoxia (21% O2, elevation 263 m), subjects completed a progressive incremental exercise test on a treadmill (4–6 mph, 0% grade, with treadmill speed increased by 2 mph/3 min, up to 6–10 mph, after which the grade was increased by 2%/3 min until volitional fatigue). The subject's maximal work rate was used during the treadmill run in the imaging studies (see Imaging Studies below). During the exercise testing, subjects breathed through a low-resistance, two-way non-rebreathing valve (model 2400, Hans Rudolph, Kansas City, MO) with expired gases sampled at the mouth and after a mixing chamber (8.6 liters) via a mass spectrometer (Perkin-Elmer model 110, Boston, MA). Inspiratory and expiratory flow rates were measured separately by two pneumotachographs. Mixed expired respiratory gas tensions and total ventilation were measured continuously to determine V̇o2 max. Peripheral arterial oxygen saturation (Sp Math) was estimated using a pulse oximeter, with optodes placed on the forehead (Nellcor OxiMax, Pleasanton, CA). Our laboratory has demonstrated a good correlation between SpMath and our directly measured HbO2 saturation (36). Heart rate was measured continuously with a three-lead ECG. Maximal exercise tests were not performed on the same days as imaging studies. Values for V̇o2 max are reported as percent predicted (6) in Table 1.

Imaging Studies

On imaging days, subjects were instrumented with a 20-gauge intravenous catheter placed in the median basilic vein. 99mTc-labeled MAA (Macrotec Kit, Bracco Diagnostics, Princeton, NJ) was injected via the intravenous catheter at rest while the subject was standing on the treadmill and during the final 30 s of a 3-min maximal exercise treadmill run, on 2 different days (separated by at least 6 days), one for each physiological state. All subjects had the resting study performed before the exercise study except subject 3. Of note, the results for subject 3 did not differ from the results of the other subjects. Once the MAA are injected, they become trapped within the pulmonary and/or systemic capillaries during the respective physiological state (rest or exercise), where they can subsequently be imaged in the supine position. The same lot of MAA was used within subjects to ensure identical particle size distributions at rest and during exercise. We used 99mTc MAA with a reported mean diameter of 20–40 μm [90% of MAA are between 10 and 90 μm in size; no particles are greater than 150 μm (5)]. Our analysis of the MAA using flow cytometry revealed a mean diameter of 45 μm, with 96% of the MAA between 15 and 105 μm. Each subject completed a 3-min warm-up period before performing the 3-min maximal exercise treadmill run. The warm-up consisted of a gradual increase in speed and grade so that the subjects were at their maximal speed and grade by the end of the warm-up. The dose (0.057 mCi/kg) of 99mTc MAA was given intravenously at rest and during the final 30 s of exercise, and, within 5 min after injection, subjects underwent quantitative whole body scanning while in the supine position. Whole body planar scanning consisted of a 20-min scan “step and shoot” protocol (5 sections, each 38 cm long for a total of 190 cm, within 25 min after exercise) that began image acquisition at the head. The “step and shoot” protocol obtains five separate anterior and five separate posterior images. The five posterior or anterior images are reconstructed by the acquisition software into a single whole body image. Anterior and posterior whole body images were obtained simultaneously in H mode (parallel heads) using a dual-head GE Millennium VG gamma camera (59 photomultiplier tubes/head; field of view 540 mm × 400 mm, 9.525-mm NaI crystal) with low-energy, high-resolution collimators [(VPC-45) General Electric Medical Systems, Waukesha, WI]. The camera was equipped with an Optitrack body contour sensing device that automatically placed the heads of the camera as close as possible to the subject to maximize counting efficiency and minimize scattering.

Imaging Data Analysis

The percentage of 99mTc MAA that was able to bypass the pulmonary capillary circulation was determined from geometric mean counts (GMT) of anterior and posterior whole body images, using two methods detailed below. We used the geometric mean because it is the more trusted measurement in radionuclide measurements of dosimetry (46), as it best reflects anterior and posterior whole body counts, and thus best corrects for attenuation, scatter, and variable detector response. All analyses were performed using GE eNTEGRA Workstation software (version 2.5202). GMT of anterior and posterior counts were obtained for all calculations and corrections. GMT were corrected for decay using standard decay tables for 99mTc (half-life 6.02 h). Labeling efficiency was determined for each batch of 99mTc MAA made using thin-layer chromatography. In this study, labeling efficiency was between 99.6 and 99.8%. The GMT of the peak pixel intensity (GMTpeak) in the lung during exercise (Lungex) was calculated. Then an automatic region of interest (ROI) was computer determined for the anterior and posterior images for both rest and exercise images to include all pixels with a minimum of 5% of the GMTpeak (±1 count) of the anterior and posterior images of the lungs at maximal exercise. That is, both rest and exercise ROIs for each subject had the same minimum pixel criteria based on the GMTpeak for the exercise images. The automated ROI was visually inspected for each subject in each condition to ensure only lung counts were included. We chose the exercise lung images to determine the 5% minimum pixel criteria, because pulmonary blood flow was more homogeneously distributed in the Lungex, and the peak pixel counts were always less than the peak pixel counts in the lung at rest (Lungrest). Thus the automated ROI would not fail to include the entire lung in either rest or exercise images.

We used two methods to estimate the transpulmonary passage via intrapulmonary arteriovenous conduits: a standard clinical method (lung-to-whole body ratio; method 1) and an adaptation of a method developed by Whyte et al. (53) for exercise studies (method 2). Because the standard clinical method has been suggested to be adequate only for detecting percentages >10% (55), we were originally concerned that this approach would not be sensitive enough to detect the small percentage of transpulmonary passage via arteriovenous conduits we expected (∼2–3%). Accordingly, we also used the adaptation of Whyte et al. (53).

Method 1: Lung-to-whole body method.

The lung-to-whole body ratio is the standard clinical method used to quantify transpulmonary passage via arteriovenous conduits in patients at our institution and is calculated as follows: Math At rest, there are possibly some MAA particles that traverse the pulmonary circulation. However, all of the resting saline contrast echocardiography studies showed no evidence of transpulmonary passage via arteriovenous conduits. Accordingly, the calculated transpulmonary passage via arteriovenous conduits at rest was set as the baseline (zero), and the calculated transpulmonary passage via arteriovenous conduits at maximal exercise was reported as the increase from the baseline.

Method 2: Adaptation of method by Whyte et al.

Whyte et al. (53) developed a method to calculate transpulmonary passage via arteriovenous conduits during submaximal exercise in patients with pulmonary arteriovenous malformations and healthy subjects using 99mTc-labeled albumin microspheres with a small size range (7–25 μm). To calculate the transpulmonary passage via arteriovenous conduits at rest, Whyte et al. determined the right kidney-to-lung ratio and estimated blood flow to the right kidney to be 10% of the total cardiac output (53). In our study, 96% of the MAA were between 15 and 105 μm (see above), and there was minimal, detectable transpulmonary passage of MAA at rest, resulting in no detectable activity in the kidneys. In addition, the proximity of the kidneys to the lung could potentially overestimate the counts in the kidney due to contamination from lung counts. Accordingly, we used head-to-lung ratio and assumed a 15% blood flow to the brain to calculate percent transpulmonary passage at rest (TPrest) (19, 21). Thus the percent TPrest was calculated as follows: Math where GMT was the corrected GMT of the head or lung as indicated; and Headrest is the head at rest. Since brain blood flow was assumed to be 15% of the cardiac output during resting conditions (19, 21), the GMT head was multiplied by 6.67 to equal 100% of cardiac output.

The equation used to calculate TPrest cannot be used for exercise, because of the redistribution of systemic blood flow during exercise. However, the change in percent transpulmonary passage from rest to exercise can be related to the change in lung activity from rest to exercise, since lung counts represent the percentage of MAA that does not traverse the pulmonary circulation. To calculate percent transpulmonary passage during exercise (TPex), a ratio of the normalized GMT Lungex to the normalized GMT Lungrest was expressed as a fraction (f). The GMT Lungrest was normalized for injected dose by dividing the GMT Lungrest by the GMT whole body at rest. This normalization was also done for exercise. Since lung counts represent the percentage of MAA that do not traverse the pulmonary circulation (transpulmonary passage via arteriovenous conduits), then Math If we rearrange this equation such that Math then, Math Thus, if the fraction is <1, it indicates that fewer counts were in the Lungex as a result of transpulmonary passage of MAA. If the fraction is >1, the subject had less transpulmonary passage of MAA during exercise.

Therefore, the percent TPex can be derived as: Math

Statistical Analysis

Comparisons of percent TPrest and TP during maximal exercise were made using Student's paired t-test. A Pearson correlation coefficient was calculated to describe the relationship between the increase in percent transpulmonary passage during maximal exercise and SpMath(average of last 30 s of maximal exercise). Results were considered significant when P < 0.05. All descriptive and physiological data are presented as means ± SD.


Anthropometric, pulmonary function, and exercise data are presented in Table 1. All subjects had values within the normal range.

Figure 1 shows the raw whole body images generated from subject 2 (Table 2) at rest and immediately after maximal treadmill exercise. The increased activity in the leg muscles with maximal exercise compared with rest is indicative of MAA particles passing through the lung and becoming subsequently trapped in the muscle capillaries. In six of seven (86%) subjects, the percentage of MAA able to get through the pulmonary circulation and into the systemic circulation (%transpulmonary passage) increased with exercise.

Fig. 1.

Anterior (Ant) and posterior (Post) planar whole body images obtained following injections with technetium-99m macroaggregated albumin at rest and during maximal treadmill exercise. The increased number of counts in the exercising muscles (legs) indicates the transpulmonary passage of technetium-99m macroaggregated albumin that has become trapped in systemic capillaries. The percent transpulmonary passage in this individual (subject 2, Table 2) at rest was 0.7%, which increased to 3.0% at maximal exercise. Color bar represents increasing count intensities with lighter colors.

View this table:
Table 2.

Change in percent shunt from resting baseline for methods 1 and 2 and peripheral estimate of arterial oxygen saturation at maximal treadmill exercise

The calculated percent transpulmonary passage with method 1 increased significantly from the resting baseline to 1.2 ± 0.8% at maximal exercise (P = 0.008; Table 2), and the increase in percent TPex was negatively correlated (r = −0.91) with SpMath. Using method 2, the calculated percent TPrest was 0.4 ± 0.1% (mean ± SD) and increased significantly to 1.7 ± 1.1% at maximal exercise (P = 0.016; Table 2). The absolute increase in percent transpulmonary passage from rest to exercise was 1.3 ± 1.0% and was negatively correlated (r = −0.92) with SpMath There was no statistically significant difference between the calculated values for method 1 (1.2 ± 0.8%) and method 2 (1.3 ± 1.0%).


We found that the percent transpulmonary passage of 99mTc MAA increased from rest to maximal exercise in 86% of the subjects studied. These data, along with our laboratory's previous isolated lung studies showing that large-diameter intrapulmonary arteriovenous pathways exist in healthy human lungs (25), further support our hypothesis that inducible intrapulmonary arteriovenous pathways open during exercise in healthy human subjects (14, 23, 40).

Methodological Considerations

Our laboratory's previous (14, 23) and present studies with saline contrast echocardiography did not identify any subjects without cardiac defects who demonstrated transpulmonary passage of saline contrast at rest, but the present study calculated a resting transpulmonary passage of MAA using methods 1 and 2. To account for the calculated transpulmonary passage via arteriovenous conduits at rest, we suspect that some smaller particles (<7–10 μm) were able to traverse the pulmonary circulation at rest that these particles would be able to traverse the pulmonary circulation during exercise as well and, therefore, would have the potential to overestimate the exercise percent transpulmonary passage. Accordingly, to avoid the potential to overestimate percent transpulmonary passage, for method 1 this resting transpulmonary passage of MAA was set as the baseline, and the exercise value was reported as the increase from baseline. For method 2, we calculated the resting values and the exercise value and reported the transpulmonary passage via arteriovenous conduits as the absolute increase from rest to maximal treadmill exercise. Furthermore, using the two different methods, the estimated percent transpulmonary passage of MAA at maximal exercise was virtually identical (1.2 and 1.3% for method 1 and method 2, respectively).

One of the assumptions for method 2 is that blood flow to the brain at rest was 15% of cardiac output. This value was based on the classic work of Kety and Schmidt (19) and Lassen (21) and, because it is an assumed value, it may not accurately represent the individual transpulmonary passage via arteriovenous conduits. However, altering the resting cerebral blood flow at rest from 15 to 10% of cardiac output changes the TPrest via arteriovenous conduit percentage from 0.4 to 0.6% and the TPex via arteriovenous conduit percentage from 1.7 to 1.9%, with the absolute increase from rest to maximal exercise remaining identical. Accordingly, the assumption that 15% of cardiac output was directed to cerebral blood flow at rest in our subjects resulted in a calculated (method 2) increase in percent transpulmonary passage of MAA that was essentially identical to that calculated using the lung-to-whole body ratio (method 1).

A possible source of error that cannot be corrected for is that the lung ROI encloses a large section of the thorax. Consequently, it is not possible to exclude, from the lung ROI, transpulmonary passage of macroaggregates via arteriovenous conduits (i.e., counts) that are distributed via the systemic circulation 1) to the myocardium by way of the coronary circulation, 2) back to the lung by way of the bronchial artery, and 3) to the respiratory muscles, such as the diaphragm and the intercostals. Therefore, it is likely that percent transpulmonary passage is underestimated (42) during maximal exercise when increased blood flow to the heart, lung, and respiratory muscles would be expected, and these counts would be included within the lung ROI.

Another potential confounder is the effect of lung density and changes in pulmonary blood flow postexercise on scatter and attenuation. The bones within the lung ROI would have been the major cause of scatter and attenuation. We would not expect that bone structure would be altered over the course of ∼6 days in the same subject, and, as such, only pulmonary blood flow and lung density are left as possible causes for attenuation and scatter. Notwithstanding the fact that the counts in the legs increase during exercise (See Fig. 1, which demonstrates not only a decrease in counts in the lung ROI, but a relocation of counts to the systemic circulation in the legs, which were not present at rest), increases in intra- and extravascular lung fluid could have theoretically attenuated lung counts postexercise. Some studies have demonstrated extravascular water accumulation 90–95 min postendurance exercise (45–120 min in duration) (7, 28). Combined, these studies suggest that 1.5 h after long-duration, high-intensity exercise, lung density increases. Interestingly, and to the contrary, in the same study by Caillaud and colleagues (7), the authors report no alteration in lung density after a 13-min incremental exhaustive exercise test using identical computed tomography methodology used to detect increased lung density 95 min after a triathlon. Similarly, others with study designs nearly identical to ours (i.e., brief maximal exercise and imaging immediately afterward) demonstrate that pulmonary blood volume rapidly returns to resting baseline levels, and that there is no evidence for significant residual extravascular lung water immediately after a maximal exercise test in healthy human subjects (3, 13). Our scans were completed within 25 min postexercise, with lung imaging complete within 15 min after 3 min of maximal treadmill exercise. Therefore, it is highly unlikely that increases in lung density due to persistent high pulmonary blood volume or residual extravascular lung water compromised these measurements.

Respiratory motion could have affected our measurements. However, respiratory motion artifact would primarily affect determination of the lung borders. To minimize the impact of this motion artifact, we used an objective automated determination of the lung borders to ensure that inclusion of appropriate lung counts were identical between conditions (see methods).

We are not able to determine the size of the particles able to traverse the pulmonary circulation. Even with particles of a known size distribution with a minimum particle size much larger in diameter than pulmonary capillaries, it would be difficult to know which particles are able to traverse the pulmonary circulation for several reasons. First of all, we would need to know the size distribution of all arteriovenous vessels throughout the healthy human lung. Currently, this is unknown; however, work from our laboratory suggests that some of these vessels are at least 50 μm in diameter (25). Second, a 50-μm vessel would allow for all particles <50 μm to bypass. Thus, even with a particle distribution of 25–50 μm, all particles 25–49 μm would be able to travel through a 50-μm vessel, which could be interpreted as multiple 20- to 25-μm pathways, when it may actually be a single 50-μm pathway. Alternatively, larger 50-μm particles may embolize arteriovenous pathways and would, therefore, prevent the potential passage of 20- and 25-μm particles.

Small particles up to 15 μm may have been able to traverse the pulmonary circulation during exercise. Although this could be a methodological limitation, the fact is that smaller particles bind significantly less radiation than larger particles and thus minimize the measured counts from the smaller particles. For example, since that radioactive labeling is a function of particle mass (not diameter) a 45-μm particle will contain 27 times more activity than a 15-μm particle [(45/15)3 = 27]. Let us assume that 90% of the particles were 45 μm, and 10% were 15 μm, and all of the 15-μm particles traversed the lung filter. The 15-μm particles would contain <0.4% of the label. In other words, for 100 total particles, the 45-μm (90%) particles would contribute 99.6% of the counts, and the remaining 15-μm (10%) particles would contribute 0.4% of the counts. In reality, 96% of the MAA particles we used were 15–105 μm, with a mean diameter of 45 μm. At most, 4% of the total particles are 15 μm (or less), and thus it is unlikely that the smaller particles contributed significantly to our counts. In fact, they would contribute significantly <0.4%, which is highlighted by the fact that the smaller particles, such as a 5-μm particle, would have 729 times less activity than a 45-μm particle [(45/5)3 = 729]. Likewise, 90-μm particles would have eight times as much label as a 45-μm particle, so using our average particle size of 45 μm for our calculations is a very conservative estimate. We cannot be certain of the particle sizes that traversed the lung filter; however, we do know that the larger particles contain the vast majority of the activity and, therefore, make the most significant contribution to our counts and calculations.

Similarly, particles could have potentially disintegrated and/or lost the radioactive label. If our findings were the result of larger particles disintegrating into smaller particles (e.g., 20 μm into two 10-μm particles), then we would have seen systematically that all subjects demonstrated transpulmonary passage via arteriovenous conduits, when, in fact, only six out of seven did so. Thus our results were not likely from particle disintegration. Likewise, when 99mTc undergoes radiolytic decomposition, it does so in the form of free pertechnetate ion, which has uptake specific to the bladder, stomach, thyroid, and salivary glands (9). Had this occurred, we would have seen it in our images, but our imaging results showed no such uptake pattern. Therefore, it is unlikely that this could explain our results.

Transpulmonary Passage of MAA via Arteriovenous Anastomoses or Capillaries?

One possible explanation for the results of our study is that significant capillary distention occurred during exercise, which would have allowed for the transpulmonary passage of MAA. Work by West and associates (4, 52) has demonstrated that average pulmonary capillary radii at high distending pressures are between 3.2 and 3.6 μm. Similarly, Glazier et al. (15) have demonstrated that the average capillary width under nonphysiological perfusion pressures of 50–100 cmH2O (37–74 mmHg) was 6.5 μm, with the largest measured capillary not exceeding 13 μm. Based on capillary distention calculations, the considerable cardiac outputs and pulmonary artery pressures that occur in thoroughbred horses during exercise should distend pulmonary capillaries up to and beyond 15 μm. Yet Manohar and Goetz (27) did not detect transpulmonary passage of a single 15-μm microsphere during exercise in the thoroughbred horse. Accordingly, significant capillary distention, if it occurs in the exercising human, is likely to be considerably less than that observed in the exercising thoroughbred horse, whose capillary distending pressures would be an estimated 95 mmHg (52) and apparently do not extend beyond 15 μm. In combination, these data provide solid evidence that capillary distention is limited even at very high, nonphysiological pressures, and, therefore, capillary distention is an unlikely explanation for our results in the present study. Furthermore, even if they did distend to 15 μm, this size of particle would likely have very little impact on our calculations, as discussed in detail above.

Another possibility would be transpulmonary passage via corner capillaries. Work by Conhaim and Rodenkirch (11, 12) report that corner vessels are 5–6 μm in diameter in rat under physiological conditions. In contrast, work by Rosenzweig and colleagues (37) reported a size range of 10–20 μm for corner capillaries in greyhound lungs. Although the size range is larger than the alveolar capillary, the authors report that the perfusion pressures were kept low (similar to resting conditions) (37). If the largest corner vessels (20 μm) allowed for the 1.2–1.3% transpulmonary passage of MAA during exercise in the present study, then these same vessels would also have allowed for TPrest, since the 10- to 20-μm range was determined under restlike conditions. However, this does not happen with either saline contrast bubbles (14, 23, 40) or microspheres (25, 39) and did not happen at rest in the present study.

In the absence of significant capillary distention or corner capillaries >20 μm, inducible intrapulmonary arteriovenous pathways remains the most likely explanation for our results. Morphological and anatomic-based studies have established the existence of arteriovenous anastomoses in normal human lungs (25, 44, 45, 54) and in the lungs of many mammalian species (8, 18, 3134, 39), with diameters of these vessel ranging from 25 to 420 μm. Despite the numerous reports demonstrating intrapulmonary arteriovenous pathways, there are also several reports that suggest otherwise (16, 22, 27, 30, 48). Although the reasons for the disparity in reports of intrapulmonary arteriovenous pathways are beyond the scope of this work, many possibilities exist. Some of the discrepancies are the result of the varying techniques (detecting microspheres in blood samples vs. tissue samples, etc.) used to study these vessels, while others are likely the varying conditions (perfusion pressure, oxygen tension, anesthesia, species, etc.) under which these vessels are studied.

Transpulmonary Passage of MAA and Microspheres via Large-Diameter Arteriovenous Pathways During Exercise

In the present study examining healthy humans, we found that the percent transpulmonary passage of MAA at rest was 0.4 ± 0.1% (mean ± SD), which increased to 1.7 ± 1.1% during maximal exercise, for an absolute increase of 1.3 ± 1.0%. The change in percent TPex was variable between subjects, with values ranging from a reduction to 0.3% to an increase to 2.7%. We have recently demonstrated that large-diameter (25 μm) polymer microspheres do not traverse the pulmonary circulation at rest, but are able to do so during exercise in dogs, and the transpulmonary passage via arteriovenous conduits was 1.4 ± 0.8% on average, with a range of 0.2–3.1% (39). The reason for the variability in the percent transpulmonary passage in both studies is not known. Clearly, the percentage of MAA and/or microspheres that is able to traverse the pulmonary circulation without being trapped will depend on the size and size distribution of the particles, as well as the variably unique pulmonary vascular morphometry of the individual organism being studied. With that said, a compelling observation is that, in both maximally exercising humans (99Tc MAA study) and dogs (25-μm neutron-activated microspheres) the estimated percent transpulmonary passage via arteriovenous conduits is essentially identical (1.3 and 1.4%, respectively).

Clinical and Physiological Relevance of Inducible Arteriovenous Conduits

Inducible large-diameter arteriovenous conduits may have a far-reaching impact on both physiological and pathophysiological conditions. With respect to pathophysiological roles, these dynamic intrapulmonary arteriovenous pathways may provide inducible conduits that allow for the passage of microembolic particles and subsequent neurological sequelae. Indeed, recent work has linked pulmonary arteriovenous malformations caused by hereditary hemorrhagic telangiectasia with migraine headaches (43). Furthermore, microemboli <100 μm have been shown to cause cerebral ischemia in rats (35). With respect to physiological roles, these vessels may be remnant fetal vessels that allow blood flow to bypass the capillary beds at a time when the fetal lung is not participating in gas exchange (54). Indeed, a recent study has demonstrated the existence of intrapulmonary arteriovenous pathways in the fetal lamb at rest (29).


Large-diameter intrapulmonary arteriovenous conduits exist in mammalian lungs. This study demonstrates that the percent transpulmonary passage of MAA significantly increases from rest to exercise in normal healthy humans. These data provide further evidence supporting the idea that large-diameter inducible intrapulmonary arteriovenous conduits open to allow blood, microbubbles, microspheres, and albumin particles to bypass the lung filter (i.e., pulmonary capillaries) during hyperdynamic conditions such as exercise. The origin and physiological relevance of these vessels is unknown. Nevertheless, the potential impact of an inducible pathway in the lung may be extensive and may contribute to a variety of physiological and pathophysiological conditions, ranging from pulmonary gas exchange inefficiency to microembolic ischemia.


Support for this project was provided by National Heart, Lung, and Blood Institute (NHLBI) Grant HL-15469, American Heart Association Grant-in-Aid 0550176Z, and the Department of Pediatrics, University of Wisconsin Medical School. H. C. Haverkamp and A. T. Lovering were supported by NHLBI Training Grant T32 HL07654.


We thank Dr. Jesus A. Bianco for help with imaging analysis, and Doug Fiers for excellent technical assistance with the gamma camera imaging. We gratefully acknowledge Jaime Bebee and David Pegelow for technical support.


View Abstract