Lung disease is responsible for one in seven deaths in the United States, totaling nearly 350,000 Americans per year.1 The most common form of acute lung disease is acute respiratory distress syndrome (ARDS), for which current therapy relies on mechanical ventilation. The ventilator settings required to achieve adequate gas exchange in some cases, however, can cause volutrauma and prevent recovery of the natural lungs (NLs).2 Mortality for ARDS is approximately 34%.3,4 Lung transplantation is the current treatment for end-stage chronic lung diseases such as primary pulmonary hypertension, idiopathic pulmonary fibrosis, cystic fibrosis, chronic obstructive pulmonary disease, and lung cancer. Donor lungs, however, are scarce and the average waiting period for a donor lung is 2 years.5 A thoracic artificial lung (TAL) attached to the pulmonary circulation could be used as a respiratory assist device in severe acute lung disease or as a bridge to transplant in chronic lung disease. A TAL is an oxygenator with a low resistance to blood flow.
Some pulmonary diseases primarily affect gas exchange, leading to hypoxemia and hypercapnia. A TAL attached in series with the NLs could provide supplemental gas transfer. Other pulmonary diseases are primarily the result of hemodynamic pathology, most notably an increase in pulmonary vascular resistance (PVR). The right ventricle (RV) is overloaded and cardiac output (CO) is reduced to problematic levels; the blood that passes through the lungs is oxygenated but total oxygen delivery is insufficient. Right ventricular failure is a major cause of death in these cases.6,7 A TAL attached in parallel with the NLs could act as an oxygenating shunt and unload the RV. In most pulmonary diseases, both gas exchange and hemodynamic disturbances are present to some degree. The appropriate mode of TAL attachment to the pulmonary circulation will depend on the specifics of the pathology and may change with time in a given TAL recipient.
An artificial lung must (1) oxygenate and decarbonate the blood; (2) operate without causing excessive blood trauma; (3) avoid causing excessive additional hemodynamic derangement or, better yet, alleviate any pathologic hemodynamic disturbance; and, (4) allow for the maintenance of significant blood flow through the NLs. The last requirement is important, because the NL vasculature provides significant compliance, metabolizes vasoactive compounds and multiple eicosanoids, and filters emboli by absorbing small gas bubbles and dissolving small clots.8–10 In acute lung disease, when recovery of the NLs is the clinical goal, sufficient NL perfusion is particularly important. These competing criteria must be considered in choosing the appropriate artificial lung attachment configuration.
The present study investigated the hemodynamic consequences of TAL attachment configuration. The TAL was assessed in a highly instrumented porcine model in parallel with the NLs, in series with the NLs, and in an intermediary hybrid of parallel and series. Parallel attachment diverted a portion of CO from the main pulmonary artery (PA) to the TAL and directed TAL outlet blood flow to the left atrium (LA). It advantageously unloaded the RV but made the least use of the nonrespiratory functions of the NLs. Series attachment diverted a portion or all of CO from the proximal main PA to the TAL and returned TAL outlet blood flow to the distal main PA. It advantageously directed total CO through the NLs but increased RV afterload. Hybrid attachment diverted a portion of CO from the proximal main PA to the TAL and split TAL outlet blood flow between the distal main PA and the LA. It allowed the TAL to oxygenate a large fraction of CO and the NLs to process a greater fraction of CO than in parallel, but did not force total CO through either the TAL or NLs. The RV drove TAL blood flow in all configurations without need for an additional pump. Pulmonary system pressures, flow rates, resistances, impedance, and compliance were measured to characterize pulmonary hemodynamic state. Right ventricular power, output, preload, afterload, contractility, oxygen consumption, and oxygen supply were measured to characterize RV function.
Pulmonary hemodynamics are affected not only by the choice of TAL attachment configuration, but also by the distribution of flow between the TAL and NLs for a given configuration. The protocol of the present study called for passage of two-thirds CO through the TAL and one-third CO through the NLs in parallel, two-thirds CO through both the TAL and NLs in hybrid, and total CO through both the TAL and NLs in series. Not all animals tolerated total diversion of CO to the TAL in series. A partial series configuration was substituted in those cases. The general strategy of this study was as follows. The TAL was attached with an inlet anastomosis on the proximal main PA and outlet anastomoses on the distal main PA and LA, as indicated in Figure 1. Bands on the PA and the two outlet grafts were adjusted to study the hemodynamics of the TAL attachment configurations.
Materials and Methods
Thoracic Artificial Lung
Each of the artificial lungs used in this study comprised a bundle of hollow, microporous, polypropylene fibers and a compliant polyurethane housing. Pure oxygen passed inside the fibers and blood flowed perpendicular to and around the outside of the fibers, resulting in efficient gas exchange across the fiber walls. A large blood-side void fraction of the gas transfer bundle and short path length through the bundle resulted in a low bundle resistance to blood flow, about 0.18 mm Hg/(L/min). The housing had a relatively large average static compliance, intended to lessen the mechanical impedance to RV output. Artificial lung design details are given in Cook et al.11
Thoracic artificial lungs were implanted in eight Yorkshire pigs weighing 71 ± 3.2 (SD) kg. All animals were treated humanely in accordance with a protocol approved by the Institutional Animal Care and Use Committee of Children's Memorial Hospital, Chicago, IL. Anesthesia was induced with an intramuscular injection of ketamine hydrochloride (20 mg/kg) and xylazine (2 mg/kg). Atropine sulfate (0.025 mg/kg) and pancuronium bromide (0.05-0.10 mg/kg) were administered; the pig was intubated; and, mechanical ventilation was initiated with a respiratory rate of 10–20 breaths/min and an end-inspiratory pressure less than 20 cm H2O. General anesthesia was maintained during surgery with isoflurane (1–5 vol%) vaporized in 60–100% oxygen.
Figure 2 depicts the experimental setup. Systemic arterial access and venous access were established by carotid artery and jugular vein catheterization, with the catheters connected to CDXIII pressure transducers (Argon, Athens, TX). A left thoracotomy was performed in the third intercostal space and an adjustable band placed around the main PA. A separate incision was made in the sixth intercostal space to place a band around the inferior vena cava (IVC). Two Gore-Tex (W.L. Gore & Associates, Newark, DE) vascular grafts, 18 mm in diameter, were attached by end-to-side anastomosis to the main PA, one proximal and one distal to the PA band. A third 18-mm graft was anastomosed to the LA. The grafts were epoxied, before surgery, to 0.5-inch inner diameter (ID) medical grade tubing for connection to the TAL inlet and outlet. The graft lengths were about 30 cm, and with the inclusion of pressure and flow sensors, the total length of the inlet line from the PA to the TAL was about 46 cm. A combination conductance catheter/pressure transducer (catheter and pressure console: Millar Instruments, Houston, TX; volume console: CD Leycom, Zoetermeer, The Netherlands) was inserted through PA wall just distal to the pulmonary valve and directed retrogradely through the pulmonary valve into the RV. A perivascular ultrasonic flow probe (Transonic Systems, Ithaca, NY) was placed around the PA distal to the conductance catheter insertion point and proximal to both graft anastomoses. A pressure transducer-tipped catheter (Camino Laboratories, San Diego, CA) was inserted into the main PA proximal to the flow probe. A 16G intravenous catheter introducer (Becton Dickinson, Rutherford, NJ) was inserted into the LA and connected to a CDXIII pressure transducer.
A 0.5-inch × 0.5-inch × 0.5-inch ID Y-connector was attached to the TAL blood outlet to split outflow between the distal PA and LA. The TAL was flushed with CO2 and primed with heparinized (1 U/ml) lactated Ringer's solution, an additional 1,000 U bolus of heparin, 5 g bovine serum albumin, and 1 g methylprednisolone. The sweep gas for the TAL was pure oxygen with vaporized isoflurane (1–5 vol%). The gas flow rate, measured with a Top Trak 821 mass flow meter (Sierra Instruments, Monterey, CA), was generally matched to the blood flow rate through the TAL and adjusted as necessary to maintain a normal arterial PCO2. A vacuum line attached to the TAL gas outlet maintained a negative pressure (–5 mm Hg) at the TAL gas inlet, as measured with a CDXIII pressure transducer. A capnograph (Datex, Puritan-Bennett, Carlsbad, CA) monitored the percentage CO2 in the TAL outlet gas line.
Before TAL attachment, a heparin bolus (100 U/kg) was administered, a heparin drip (360–1,000 U/h) was begun to maintain an activated clotting time above 200 seconds, and a steroid drip (1 g/h methylprednisolone) was begun in anticipation of an inflammatory reaction to blood flow through the unsterilized, prototype artificial lung. The proximal PA graft was attached to the TAL blood inlet. The distal PA graft and LA graft were attached to the branches of the Y-connector at the TAL blood outlet. CDXIII pressure transducers and sterile-tubing flow transducers (Transonic Systems) were attached to the TAL blood inlet and outlet, with the outlet flow transducer located on the LA outlet branch. Signals from all pressure, flow, and volume transducers were sampled at 250 Hz with Acquire Plus data acquisition software (Gould Instruments, Valley View, OH).
Initial hemodynamic data were acquired at steady-state for 15 seconds and during three transient occlusions of the IVC. The IVC was occluded to acquire RV contractility data with the conductance catheter. Cardiac output and systemic arterial pressure were allowed to recover between occlusions. The ventilator was temporarily suspended at end-expiration during all data acquisition periods. Arterial and venous blood samples were collected for blood gas analysis and measurement of specific conductivity of the venous blood, a quantity needed to determine RV volume with the conductance catheter.
A conditioning blood flow of 0.2 l/min was directed through the TAL in the parallel configuration. This flow rate was maintained during a 20-minute acclimation period. If PA pressure increased above 125% of the value before TAL flow, TAL flow was suspended and additional steroids were administered. When PA pressure returned to its pre-TAL level, 0.2 l/min were again directed through the TAL in the parallel configuration and the configuration was maintained for 20 minutes. After acclimation, flow to the TAL was suspended. Steady-state baseline hemodynamic recordings were acquired and blood samples were drawn.
Artificial lung assessment was initiated in a parallel configuration in which the distal PA graft was fully occluded, the LA graft was open, and the PA section between the two anastomoses was, if necessary, partially occluded to divert approximately two-thirds CO to the TAL and allow passage of the balance directly to the NLs. This configuration was maintained for 20 minutes and hemodynamic data were recorded at 0, 10, and 20 minutes. Blood samples were drawn from the TAL inlet, TAL outlet, and systemic arterial line between the first two hemodynamic data acquisition periods for blood gas analysis and specific conductivity measurement. At the end of the 20-minute period, hemodynamic data were recorded during three IVC occlusions. Two hybrid configurations, hybrid I and hybrid II, and a series configuration were also assessed. In hybrid I, the PA section between the two anastomoses was partially occluded, the distal PA graft was open, and the LA graft was partially occluded such that the TAL and NLs each received approximately two-thirds CO. In hybrid II, the PA section between the two anastomoses was fully occluded, the distal PA graft was open, and the LA graft was partially occluded such that the TAL received total CO and the NLs approximately two-thirds CO. In series, the PA section between the two anastomoses was fully occluded, the distal PA graft was open, and the LA graft was fully occluded such that the TAL and NLs each received the total CO. Attachment configurations were assessed in the order: baseline, parallel, hybrid I, hybrid II, series, repeat hybrid II, repeat hybrid I, repeat parallel, and repeat baseline. Data were collected in each configuration as described for parallel.
At the end of each experiment, the animal was euthanized with an intravenous injection of sodium pentobarbital (90–150 mg/kg). The RV was opened to observe conductance catheter placement.
Pulmonary system hemodynamics were assessed by tracking changes in pressures, flow rates, resistances, impedance, and compliance with change in attachment configuration. Average RV, PA, TAL inlet, TAL outlet, LA, systemic arterial, and systemic venous pressures and average PA, TAL inlet, and LA graft flow rates were calculated over a whole number of cardiac cycles.
The resistance of the combined TAL/NL system, the pulmonary system resistance, or PSR, was calculated as:
in which PPA and PLA are mean PA and LA pressures, respectively. Pulmonary input impedance is an alternative description of system resistance. Fast Fourier transforms were performed on PA pressure and flow time series. At a given frequency, ω, the ratio of the pressure modulus, |PPA(ω)|, to the flow modulus, |QPA(ω)|, yielded the impedance modulus |Z(ω)|= |PPA(ω)|/|QPA(ω)|. The impedance modulus is plotted as a function of frequency. The vascular compliance of the proximal portion of the combined TAL/NL pulmonary system, referred to as the pulmonary system compliance, or PSC, was assessed by applying a simple windkessel model to the system:
in which SV is stroke volume, tsys is duration of systole, and (PPA)max and (PPA)min are maximum and minimum PA pressures, respectively. The numerator of Equation 2 is the additional volume of fluid that was stored in the pulmonary system during systole, the total volume ejected by the right ventricle minus the volume that flowed out of the pulmonary system during systole. The denominator is the difference in pressure that resulted from storage of the additional volume.
The method for calculating NL vascular resistance, PVR, varied with attachment configuration. At baseline and in some parallel configurations, there was no PA band and the mean pressure at the inlet to the NLs was equal to PPA. In these cases, PVR was calculated as:
in which Q𝑄̅NL is the mean flow rate through the NLs, calculated as PA flow rate minus LA graft flow rate. In parallel configurations with a band on the PA, the NL inlet pressure was unknown and PVR could not be calculated. In hybrid and series, the NL inlet pressure was equal to the average TAL outlet pressure, PTALout, minus the average pressure drop across the distal PA anastomosis, δPA2. An in vitro study yielded data that were used to determine a relation for δPA2. In hybrid and series, therefore, PVR was calculated as:
Systemic vascular resistance, SVR, was calculated as:
in which MAP is mean systemic arterial pressure and CVP is central systemic venous pressure.
Right ventricular function was assessed by tracking changes in RV contractility, power, oxygen consumption, and oxygen supply, amongst other indices, with change in attachment configuration. Preload-recruitable stroke work, PRSW, was used as the index of cardiac contractility.12,13 Right ventricular stroke work was plotted versus end-diastolic RV volume over the initial cycles from each caval occlusion data set and a linear regression was fit to the data. The slope of this correlation was the PRSW index. The conductance catheter was calibrated against the PA flow probe by comparing conductance catheter-obtained stroke volume measurements to integrated systolic PA flow. The slope, α, of the correlation between the two measures of stroke volume was used to correct conductance catheter volume measurements.14
Instantaneous RV power was calculated as:
in which PPA is instantaneous PA pressure, in millimeters of mercury, QPA is instantaneous PA flow rate, in liters per minute, and 0.00222 converts the units to watts. A fast Fourier transform of the RV power time series yielded a frequency spectrum for the modulus of RV power.
The tension-time index, TTI, correlates with myocardial oxygen consumption15 and was calculated as:
in which PRVeject is average RV ejection pressure, teject is ejection duration, and HR is heart rate.
The pressure-time index, PTI, has been shown to correlate with myocardial oxygen supply in the left ventricle, LV.16 The LV is perfused primarily during diastole and the LV PTI is calculated as the integral over diastole of the driving pressure in the coronary arteries. The RV myocardium, however, receives significant perfusion throughout diastole and systole.17–19 A modified RV pressure-time index, RVPTI, was calculated for the RV free wall by integrating over the full cardiac cycle:
in which τ is cardiac period, SAP is instantaneous systemic arterial pressure, DSP is downstream pressure in the coronary vessels, and t is time. The DSP was determined according to the waterfall model.20 It was equal to the greater of central venous pressure or myocardial tissue pressure unless tissue pressure exceeded SAP, in which case DSP equaled SAP. Tissue pressure in the thin RV free wall is essentially constant across the wall and approximately equal to intraventricular pressure.20 Myocardial tissue pressure was assumed equal to intraventricular RV pressure.
The RVPTI correlates with the volume of blood entering the coronary vessels of the RV free wall in each cardiac period if the coronary resistance is constant. Oxygen supply depends on this volume, as well as blood oxygen content and cardiac frequency. An estimate of coronary oxygen supply to the RV free wall, O2RVPTI, was calculated as:
in which k is oxygen solubility in blood, PaO2 is arterial partial oxygen pressure, CHb is hemoglobin concentration, and SaO2 is arterial oxyhemoglobin saturation. The terms in the first set of parentheses on the right side of Equation 9 represent the oxygen content of the blood.
Differences between MAP values at three time points were compared by Student's t test. Differences in PSR, PSC, CO, PPA, end-diastolic RV pressure, PRSW, TTI, O2RVPTI, flow rate through the TAL (QTAL), flow rate from the RV through the PA to the NLs (QRV-NL), QNL, PVR, LA pressure, MAP, HR, SVR, and CVP between attachment configurations were assessed by single-factor analysis of variance and Tukey's post hoc multiple regression applied to data from initial and repeat assessments of all attachment configurations. Differences in these indices between the initial and repeat assessments of each specific attachment configuration were additionally assessed by Student's t test. Differences between two subsets of data distinguished by PVR-QNL and CO-PSR regression curves were assessed by grouping the data according to their abscissal values. Data groups for QNL were 0.5–1.5, 1.5–2.5,…, and 7.5–8.5 l/min. Data groups for PSR were 0.5–1.5, 1.5–2.5,…, and 15.5–16.5 mm Hg/(l/min). Within each data group, ordinal values of the two data subsets were compared by Student's t test. For all statistical tests, probability values less than 0.05 were considered significant.
Thoracic artificial lungs were implanted in eight animals and their effects assessed, as described above, at baseline and in parallel, hybrid, and series. The order of configuration testing was reversed in the second half of each experiment. Each animal thereby served as its own control. Results from two animals are excluded because of severe cardiac dysfunction observed during the experiment and, in one of the cases, a myocardial lesion found upon inspection after the experiment.
Blood flow through the prototype TALs induced an inflammatory response and a resultant increase in PVR. The TALs, thus, were tested in a disease model with a moderately elevated PVR of 3.9 ± 1.1 (SD) mm Hg/(l/min) at baseline. Mean systemic arterial pressure decreased significantly (p < 0.05) from an initial 85 ± 8.5 mm Hg at the start of the experiment to 64 ± 10 mm Hg after the completion of the surgery and the initiation of a steroid drip, but before the initiation of blood flow through the TAL. After acclimation to TAL flow, the resultant inflammatory reaction and the administration of additional steroids, MAP decreased further, although not significantly, to 60 ± 11 mm Hg. Postinflammation data are presented as the initial baseline state.
Variations in animal condition, prototype TAL construction and experimental conditions, as well as limited precision possible in adjusting the flow-controlling bands, contributed to variations between cases. With parallel attachment, for instance, the protocol specified diversion of 67% of CO to the TAL. Actual diversion in parallel ranged from 53–80%, but averaged 66% for the six cases. In two animals, the desired distribution of CO occurred spontaneously in parallel in the absence of a PA band. In four animals, the PA was constricted in parallel to force the desired fraction of CO through the TAL. In three animals, full constriction of the PA band in hybrid II and series induced a sudden and precipitous drop in CO. In these animals hybrid II was omitted and a partial series configuration, with an average diversion of 42% ± 5% of CO to the TAL and total CO through the NLs, was substituted for full series. Elevated PVR additionally complicated hybrid I in two of the three cases in which the PA could not be fully constricted. It was not always possible to constrict the LA outlet graft sufficiently, without fully occluding it, to make the resistance of that graft greater than that of the NLs and force some of the TAL outlet flow through the NLs. If the LA outlet graft was not fully occluded in these cases, a portion of CO shunted around both the TAL and NLs. Blood flowed from the distal PA to the TAL outlet and directly to the LA. Administration of additional steroids and/or constriction of the bands on both TAL outlet grafts resolved this condition. Constriction of both grafts increased the resistance of the path from the distal PA through the two grafts to the LA enough to prevent shunt flow. The additional constriction on the distal PA outlet graft, however, added to total PSR. Finally, in one animal the natural lungs were in such poor condition at the end of the experiment that the return from parallel to baseline caused right heart failure. The repeat baseline configuration was not maintained for the full 20-minute time period and that one data point is omitted.
Results for each configuration from the six cases were averaged, despite the variations described above, and are presented together. Steady-state was reached almost immediately upon switching to a new configuration and configurations were stable over the 20-minute assessment period. There was little difference between hybrid I, hybrid II, and series. As the results were the same in the initial and repeat assessments of both hybrid configurations and as hybrid II was achieved in only three cases, hybrid II results are omitted. The hybrid results presented, thus, are only for hybrid I. The presented series results are an aggregate of the three partial series and the three full series cases. There were not generally significant differences between hemodynamic indices in partial and full series. Due to varying NL conditions, each configuration placed maximal stress on the RV. Differences that did exist between partial and full series, nevertheless, are mentioned and a figure at the end of this section compares additional results between animals that could and could not tolerate full series.
Figures 3–10 present average and representative results from six animals. The bar graphs display mean values ± standard deviations of data obtained after 20 minutes in each configuration. A bar with a symbol at its base is significantly different (p < 0.05) than any other bar with the same symbol above its error limit, as assessed by analysis of variance and Tukey's multiple comparison. Alteration of attachment configuration caused systematic and significant changes in most hemodynamic parameters. If an index was significantly different between two attachment configurations in the first half of the experiment, then it was generally also significantly different between the repeat versions of the same configurations. If an index was significantly different in initial parallel than in series, for example, it was also generally significantly different in repeat parallel than in series. This symmetry suggests that the differences were due to attachment configuration and not other experimental variables, such as order of configuration testing. A link and symbol connecting the tops of two bars indicate a significant difference (p < 0.05) between the initial and repeat assessments of a given configuration, as assessed by Student's t test.
Figures 3a and 3b show that alteration of attachment configuration caused significant changes in pulmonary system resistance and compliance. Parallel TAL attachment reduced PSR relative to baseline, whereas hybrid and series attachments elevated PSR. The resistance in series was significantly greater than that at baseline or in parallel and the resistance in initial hybrid was significantly greater than that in parallel. The trend in pulmonary system compliance was generally the reverse of that in PSR. The system compliance with TAL attachment in any configuration was generally less than that of the baseline natural pulmonary system. The PSC in all attachment configurations was significantly less than that of initial baseline and the PSC in hybrid and series was significantly less than that of repeat baseline. With no access to the NL compliance, the PSC in the three animals that tolerated full series averaged only 0.17 ± 0.11 ml/mm Hg in that configuration.
Figure 3c shows representative impedance spectra for baseline and the three attachment configurations from one animal. A normal spectrum was observed at baseline. The spectrum dropped between the zero frequency and the first harmonic frequency (equal to heart rate), and leveled off above the first harmonic. Attachment of the TAL added a new conduit through the pulmonary circulation. That conduit, however, incorporated significant impedance mismatches at the graft anastomoses, the TAL inlet, and the TAL outlet. When the TAL was added in parallel in the absence of a PA band, the case shown in Figure 3c, the RV had full communication with the large compliance of the NLs and the impedance spectrum was hardly altered from baseline. The zero frequency modulus, which is similar to PSR when LA pressure is small and constant, dropped below baseline in parallel but the spectrum maintained a normal shape. When a band was required in parallel (data not shown), hybrid, or series, the impedance spectrum was markedly altered due to interruption of RV communication with the compliance of the NLs by a significant impedance mismatch at the PA band. Constriction of the PA disrupted the impedance spectrum more than did TAL attachment, resulting in increased impedance moduli at and above the first harmonic. Large impedance moduli at low frequency strain the RV. Elevation of the zero-frequency modulus in hybrid and series, due to elevated PSR, was particularly taxing. Large impedance moduli at high frequency are of relatively little importance because the modulus of RV power at high frequency is small (Figure 3d).
Figures 4a and 4b show how TAL attachment, with its effects on the mechanical properties of the pulmonary system, affected the ability of the RV to maintain flow and generate pressure. In parallel, with reduced PSR and despite reduced PSC, RV output increased and PA pressure decreased. In hybrid and series, with elevated PSR and reduced PSC, RV output decreased and PA pressure increased. Cardiac output and PA pressure were each significantly different in series than they were in parallel. Cardiac output, additionally, was significantly greater in partial series, 5.3 ± 0.86 l/min, than in full series, 3.5+ 1.3 l/min (p < 0.05, difference not shown in the figure). The inverse relation between CO and PA pressure was such that mean RV power (data not shown) did not change significantly with attachment configuration.
Cardiac performance on a beat-to-beat basis is traditionally described as governed by preload (end-diastolic RV volume), afterload (PA pressure), and contractility. Figures 4c and 4d show how TAL attachment affected end-diastolic RV pressure, an analogue for preload, and PRSW, a measure of contractility. Preload followed the same trend as afterload, although the differences between configurations were not significant. Mildly elevated preload at baseline was indicative of RV congestion, likely due to elevated PVR after the initial inflammatory response. The additional elevation of preload in hybrid and series was due, at least in part, to regurgitative pulmonary valve flow observed at end-systole in those configurations.21
The quality of correlation between change-in-volume measurements recorded by the conductance catheter and determined from integrated PA flow could not be assessed in real time during the experiments. Regression coefficients, thus, were poor (R2 < 0.73) in determining α for half of the PRSW data points. Those data were omitted in the analysis. The values of this index from the initial and repeat assessments of each attachment configuration, thus, were grouped together (Figure 4d). The initial baseline PRSW values reported, unlike the initial baseline values of all other indices, are from before the TAL acclimation period; postacclimation baseline caval occlusions were not performed. Contractility, as assessed by PRSW, was mildly elevated with all TAL attachment configurations, although not significantly so in any case due to a large standard deviation at baseline. The conductance catheter position varied between cases. It was located in the RV outflow tract in three cases and bent around the tricuspid valve in the other three cases. The conductance catheter data were calibrated against PA flow rate data in all cases and no difference in data quality was observed between the two positions.
Figures 5a and 5b further describe RV function. These figures show example RV pressure and PA flow rate traces at initial baseline and in series. The first of the two vertical lines marked in each figure indicates end-diastole as identified by a distinct change in slope of the RV pressure trace. Before this line RV pressure increased moderately due to RV filling. After this line RV pressure increased more sharply due to RV contraction. The second vertical line indicates the onset of RV ejection. In some cases, there was a second change in slope of the RV pressure trace that coincided with the onset of RV ejection, indicating additional RV work as the RV pumped against pulmonary impedance. High-impedance TAL attachment configurations such as hybrid and series tended to increase the time between end-diastole and commencement of RV ejection. Additionally, high-impedance TAL attachment configurations tended to smooth the shapes of the RV pressure and PA flow rate traces. The distinct change in slope of the RV pressure trace at end-diastole became less pronounced. The period of pulmonary valve closure during diastole decreased or disappeared and the magnitude of regurgitation through the pulmonary valve increased.21 It should be noted that the example shown in Figure 5b is an extreme one. Right ventricular pressure and PA flow rate traces did not always change so markedly in high impedance TAL attachment configurations. In all experiments, the traces had the same shapes at final baseline as they did at initial baseline (data not shown).
Figures 6a and 6b show the dependence on attachment configuration of the tension time index, an indicator of RV myocardial oxygen consumption, and the O2 RV pressure–time index, an indicator of oxygen supply to RV free wall tissue. The TTI was significantly elevated in hybrid and series compared with baseline. It was also elevated, although not significantly, in parallel compared with baseline. The trend for the O2RVPTI was opposite that for the TTI. The O2RVPTI index was significantly less in series than at the initial baseline.
The design protocol of these experiments was to divert two-thirds CO to the TAL in parallel, two-thirds CO in hybrid and total CO in series, and to pass one-third CO through the NLs in parallel, two-thirds CO in hybrid, and total CO in series. Because attachment configuration affects pulmonary system impedance and CO, however, the quantitative TAL and NL blood flow rates were not these fractions of baseline CO. Figures 7a and 7b show the blood flow rates through the TAL, Q𝑄̅TAL, and from the RV through the PA to the NLs, Q𝑄̅RV-NL, as functions of attachment configuration. The total amount of blood gases that can be delivered by the TAL is limited by Q𝑄̅TAL. The Q𝑄̅TAL was significantly greater in initial parallel, 4.5 ± 0.8 L/min, than in initial hybrid, 2.9 ± 0.6 L/min, or series, 2.8 ± 1.0 L/min. The Q𝑄̅RV-NL, which must be oxygenated by the NLs, did not differ significantly across attachment configurations. It was 2.3 ± 0.9 L/min in initial parallel, 2.2 ± 0.5 L/min in initial hybrid, and 1.5 ± 1.8 L/min in series. The Q𝑄̅RV-NL was, however, significantly greater in partial series, 3.1 ± 0.7 L/min, than in full series, 0 L/min (p < 0.05). Figures 8a and 8b show the blood flow rate through the NLs, Q𝑄̅NL, and PVR as functions of attachment configuration. The average blood flow rate through the NLs is the sum of Q𝑄̅RV-NL and the flow rate from the TAL outlet to the distal PA. It is blood that can be cleared of microemboli and metabolically processed by the NL tissue. The Q𝑄̅NLwas significantly greater in series than in parallel. The Q𝑄̅NLwas 4.4 ± 1.4 L/min in series compared with 2.3 ± 0.9 L/min in initial parallel, despite a significantly reduced CO in series (4.4 L/min) compared with that in initial parallel (6.8 L/min, Figure 4a). The Q𝑄̅NLwas, additionally, significantly greater in partial series, 5.4 ± 0.8 L/min, than in full series, 3.5 ± 1.3 L/min (p < 0.05). The NL resistance, PVR, depends on the blood flow rate through the NLs and it decreases, due to capillary recruitment, with increasing blood flow rate. Pulmonary vascular resistance was significantly greater in initial parallel than at baseline, in series or in repeat hybrid. Because PVR was significantly less in series, 2.8 ± 2.1 mm Hg/(L/min), than in initial parallel, 8.9 ± 2.8 mm Hg/(L/min), the increased PSR with series attachment, compared with that of parallel attachment, was somewhat less than might be expected. The PSR was, nevertheless, significantly greater in series, 7.1 mm Hg/(L/min), and in initial hybrid, 6.4 mm Hg/(L/min), than in initial parallel, 3.0 mm Hg/(L/min) (Figure 3a).
Figure 9 shows average LA pressure during the experiment. Unlike most other parameters, LA pressure did not exhibit the same correlation with attachment configuration in the second half of the experiment as in the first. It increased in the first half of the experiment and continued to increase in the second half. Left atrial pressure was significantly greater in the repeat baseline and hybrid configurations than in the initial baseline and hybrid configurations, respectively.
Figure 10 shows systemic data. Trends were evident but the differences were not significant; the systemic circulation is removed from the location of TAL attachment. Mean systemic arterial pressure (Figure 10a) was marginally elevated above baseline in parallel, due to reduced PSR and a resultant increase in CO, and marginally depressed below baseline in series, due to elevated PSR and a resultant decrease in CO. It was significantly lower in repeat parallel than in initial parallel, suggesting a slight drop in MAP over the course of the experiment. Heart rate (Figure 10b) and systemic vascular resistance (Figure 10c) rose marginally from baseline to series and decreased slightly with the return to baseline, indicating there was stimulation of the sympathetic nervous system in the configurations in which MAP was depressed below its baseline value. Central venous pressure (Figure 10d) rose marginally from baseline to series and decreased with the return to baseline. The trend in CVP, along with that in end-diastolic RV pressure (Figure 4c), indicates that elevated PSR in hybrid and series caused additional congestion in the systemic veins and right ventricle.
Figures 11a and 11b show inverse relations between PVR and NL blood flow rate, due to capillary recruitment, and between CO and PSR. The data are divided into two subsets: the animals that tolerated the full series configuration, called Group FS, and those that did not, called Group PS for the partial series configuration that was substituted. The PVR-Q𝑄̅NL curve for Group PS is shifted upwards and to the right of that for Group FS, indicating that PVR was greater at low flow rates in Group PS than in Group FS. Pulmonary vascular resistance was significantly greater in Group PS than in Group FS for natural lung flow rates less than 5.5 L/min. Cardiac output was higher at low PSR (typically parallel and baseline) but dropped off more sharply with increasing PSR (typically hybrid and series) in Group PS. Cardiac output was significantly greater in Group PS than in Group FS for PSR values less than 4.5 mm Hg/(L/min). Mild differences in both pulmonary vascular properties and cardiac function between the two groups, therefore, may have been responsible for the inability to fully occlude the PA in half of the animals.
The function of a thoracic artificial lung is to transfer blood gases and, in certain cases, to reduce right ventricular afterload. How well it accomplishes these goals depends on (1) TAL design; (2) blood flow rates through the components of the TAL/NL system, determined by attachment configuration and proportioning of CO between TAL and NL pathways, and their effect on gas transfer potential; and (3) underlying physiology/pathology. Attaching a TAL to the pulmonary circulation affects pulmonary hemodynamics and RV function, which in turn affect TAL/NL system performance. The present is a study of the effect on pulmonary hemodynamics of attaching a specific TAL in three configurations: parallel, hybrid, and series. Artificial lungs have been tested previously in parallel22,23 and in series.24–26 The hybrid configuration was tested here for the first time. The most appropriate attachment configuration and flow distribution depend on the specifics of the underlying pathology, and preimplant tests may be required.
The objective of the TAL/NL system is to provide adequate blood-gas transfer while minimally affecting right heart function. The TAL used in this study fully oxygenated the blood passing through it at all flow rates tested, but the TAL/NL system impedance was too large in the hybrid and series configurations. The high RV afterload of these configurations significantly reduced CO, which, in turn, limited the potential for blood-gas transfer. Right ventricular afterload depends on the resistance, compliance, and fluid inertia of the TAL/NL system distal to pulmonary valve. In the normal pulmonary circulation, the RV pumps directly into a large compliance proximal to the pulmonary vascular resistance. Perhaps the biggest shortcoming of the TAL system used in this study was that it presented the RV with a large resistance and inertia proximal to the TAL compliance. Constricting the PA to divert blood flow to the TAL reduced or eliminated RV access to the large vascular compliance of the pulmonary arterial system, as demonstrated by the PSC values in parallel. The average PSC in parallel in the four animals that required some PA constriction to achieve the designed parallel flows was 0.9 ± 0.4 ml/mm Hg, whereas the average PSC in the two animals that did not require any PA constriction in parallel was 1.5 ± 0.3 ml/mm Hg.
The TAL housing compliance, about 4.3 ml/mm Hg,11 was large compared to the average first baseline value of the natural system, 2.3 ± 1.0 ml/mm Hg. It was separated, however, from the RV by the (1) proximal PA anastomosis resistance, (2) inertia and resistance of the 46 cm long line from the anastomosis to the TAL inlet, and (3) TAL entrance resistance. The “minor loss” across the end-to-side proximal PA anastomosis was large. [Pressure losses due to locally abrupt changes in cross-sectional geometry (e.g., branches, sudden expansions or contractions of the cross section) are commonly referred to as “minor losses” by engineers. These mechanical energy losses are small relative to the losses in the long pipes of traditional engineering systems, hence the word minor. In this case, however, the “minor losses” across the end-to-side proximal PA anastomosis and the entrance to the TAL were the dominant pressure losses in the system.] The resistance of the Proximal anastomosis could be estimated using the mean PA and TAL inlet pressures andQ𝑄̅TAL. Because the flow was pulsatile and the proximal minor losses are flow dependent, the calculated anastomosis resistance showed considerable scatter. It was, nevertheless, about 3 mm Hg/(L/min) at the average TAL flow rate of 3.2 L/min. The TAL used in this study had a very low fiber-bundle resistance, about 0.2 mm Hg/(L/min), but relatively large entrance and exit resistances. The entrance resistance was about 1 mm Hg/(L/min) at 3.2 L/min.11 The approximate total resistance proximal to the TAL compliance, therefore, was about 4 mm Hg/(L/min). This resistance, coupled with the inertia of the blood in the long TAL inlet line, resulted in excessive impedance to RV ejection and prevented RV access to the TAL compliance. The major flow resistance in the TAL system, the anastomosis resistance, is proximal to the compliance, whereas the major resistance in the natural system, the PVR, is distal to the compliance. The effective measured PSC in full series was only 0.17 ml/mm Hg, indicating the TAL compliance did little to unload the RV. The long TAL inlet line, which added significant inertia, was required to accommodate the pressure and flow transducers for this detailed hemodynamic study. It would be significantly shorter in a clinical setting. Redesign of the TAL entrance,11 a shortened TAL inlet line, and, possibly, angled graft anastomoses to the PA should greatly increase RV access to the TAL compliance and reduce RV afterload much below the levels measured in these experiments.
Attachment Configuration, Blood Flow Distribution, and Gas Transfer Potential
Attachment configuration and blood flow distribution affect (a) pulmonary system impedance, which, in turn, affects CO, which, in turn, affects gas transfer potential; and (b) blood flow rate through the NLs for metabolic processing and potential removal of TAL-generated microemboli. The attachment configuration and fractional distribution of CO determine the quantitative blood flow rates (1) through the TAL, (2) from the RV directly to the NLs, (3) through the NLs, and (4) from the TAL to the LA. The first two flow rates are those of venous blood flow through the TAL and the NLs, respectively, which essentially determine the gas transfer potential of the system. The third is sum of venous blood from the RV and oxygenated blood from the TAL outlet that enters the NLs. It determines the degree of nonrespiratory blood processing by the NLs. The fourth is the shunt flow that bypasses the NLs.
The effect of attachment configuration and fractional distribution of CO on the quantitative blood flow rates was difficult to anticipate. “Minor loss” resistances at the PA anastomoses and TAL entrance and exit, all of which were nonlinear, increased with flow rate.11 Pulmonary vascular resistance, however, decreased with increasing flow rate (Figure 11a). The constrictions of the PA and the two TAL outlet grafts, used to control the distribution of blood flow in the system, also contributed to system resistance and system nonlinearity; these “minor loss” resistances should also have increased with increasing flow rate. Greater constriction of the PA increased the fractional diversion of CO to the TAL but with decreasing returns because of the increased resistance of the constriction itself and the nonlinear resistances of the pathway through the TAL. Greater constriction of the PA increased pulmonary system impedance and decreased CO. Greater constriction of the LA graft had a more ambiguous effect. The increased resistance of the constriction itself tended to increase pulmonary system impedance and decrease CO. With more blood flow to the NLs, however, PVR decreased. Overall, achieving the hybrid and series configurations called for in the present protocol with the current TAL significantly decreased CO. This decrease was such that the quantitative TAL flow rate was less with total CO to the TAL in full series, 3.5 L/min, than with two thirds CO to the TAL in parallel, 4.5 L/min. The TAL flow rate is particularly important when NL gas exchange capability is impaired.
Some differences in gas exchange between different attachment configurations can be anticipated from the recorded flow rates. The TAL fully oxygenated the blood passing through it11, and the NLs were ventilated with essentially 100% oxygen. No significant differences, therefore, in total combined TAL/NL gas transfer rate were evident across attachment configurations (data not shown). If the underlying pathology involved severely impaired NL oxygenation capability, however, then the blood flow rate through the TAL would determine total pulmonary system gas transfer capability. The blood flow rate through the TAL was markedly greater in parallel (4.5 ± 0.8 L/min) than in hybrid (2.9 ± 0.6 L/min), partial series (2.2 ± 0.1 L/min), or full series (3.5 ± 1.2 L/min) (Figure 7a). With residual or full NL gas transfer ability, however, the NLs would contribute to total oxygenation by oxygenating the venous blood passing directly from the RV to the NLs. The NLs would not transfer oxygen to blood in hybrid or series that had passed through the TAL before entering the NLs, because the TAL would have saturated that blood. The venous RV-to-NLs flow rate was 2.3 ± 0.9 L/min in parallel, 2.2 ± 0.5 L/min in hybrid, 3.1 ± 0.7 L/min in partial series, and 0 L/min in full series (Figure 7b). Parallel, due to a high CO, had the greatest TAL flow rate of any configuration in the present experiments and had a RV-to-NLs flow rate comparable to that in hybrid or series. Parallel, thus, provided the greatest gas exchange potential of any attachment configuration, regardless of NL oxygenation capability.
The above discussion pertains to results obtained with the specific TAL tested in the present study. Improvements to the anastomoses and TAL design should increase RV tolerance of hybrid or series configurations. Any improvements made to the design of the pathway through the TAL, however, would also improve the parallel configuration. With significant design improvement, parallel might unload the RV to a much greater degree than it did in the present study and might be especially beneficial in pathologies with elevated PVR. Given the many nonlinear components of the TAL/NL system, it is difficult to anticipate whether design improvements would cause changes in the relative benefits of different TAL attachment configurations. Parallel TAL attachment, however, is likely to provide the most favorable hemodynamics and significant gas exchange.
The major disadvantages of parallel TAL attachment are a reduction in NL metabolism of vasoactive molecules in the blood and the potential for passage of microemboli to the systemic circulation. Takewa et al.27 indicate passage of 35–50% of normal cardiac output through the NLs should be sufficient for metabolic blood processing. Flow through the NLs was generally at least 2.0 L/min in this study. The potential for passage of microemboli to the systemic circulation is the most significant drawback to parallel attachment.
Hybrid, like parallel, has the potential to improve pulmonary system hemodynamics over those at baseline. A significant redesign of the TAL used in the present study, however, would be required to achieve favorable hemodynamics in hybrid. The benefits of hybrid over parallel are the potential for greater NL blood processing and for a reduction, although not an elimination, of the possibility of embolic passage to the systemic circulation. The drawbacks of hybrid include a more difficult surgical procedure, less favorable hemodynamics than parallel and, with the present TAL, shunt flow around the NLs. Future inclusion of a one-way check valve in the distal PA graft would prevent shunt flow. Although mechanical valves can be problematic, all flow through this valve would subsequently pass through the NLs where any small emboli generated would likely be absorbed. Hybrid attachment with near-total or total CO to the TAL might be the best choice when the NLs have little or no gas transfer capability and virtually all gas transfer must be by the TAL. Only that portion of CO required for NL metabolic processing would have to be diverted to the NLs.
Both partial and full series provide the benefit of having total CO pass through the NL for embolic clearance. In the present study, partial series provided flow rates of 2.2 ± 0.1 L/min through the TAL and 3.1 ± 0.7 L/min directly from the RV to the NLs. If NL oxygenation capacity were only slightly impaired, partial series would have provided simple and effective blood gas transfer. If NL oxygenation capacity were significantly impaired, however, partial series would have failed to provide adequate gas exchange due to its low TAL flow rate. With significantly impaired NL oxygenation capacity, full series offered the greatest TAL flow rate in the present study, 3.5 ± 1.2 L/min, of any configuration other than parallel despite having the lowest CO of any configuration. Full occlusion of the PA may be a concern, however, even with an improved TAL design. A goal should be to improve TAL hemodynamics sufficiently such that CO and, as a result, TAL flow rate rise to acceptable levels in a partial or full series configuration.
The pathologic states of the lungs and heart are major factors in selecting the appropriate attachment configuration. In the present study, PVR was 3.9 ± 1.1 mm Hg/(L/min) at baseline and CO was 6.0 ± 1.3 L/min. This PVR value is about three times normal for both pigs and humans,28,29 but is typical for pulmonary diseases such as ARDS, end-stage pulmonary fibrosis, and severe chronic obstructive pulmonary disease.30–32 Parallel TAL attachment reduced PSR to 3.0 ± 0.76 mm Hg/(L/min) and increased CO to 6.8 ± 1.1 L/min. In a pathology such as primary pulmonary hypertension, in which PVR can be as high as 15 mm Hg/(L/min),33 parallel may be the only option and should markedly improve hemodynamics. Series TAL attachment in the present study increased PSR to 7.1 ± 2.9 mm Hg/(L/min) and decreased CO to 4.4 ± 1.4 L/min. Hybrid and series attachment of the TAL used in this study added to RV work load and should be considered only with minimally elevated PVR.
The calculated TTI and O2RVPTI indices correlate, although somewhat imperfectly, with RV oxygen consumption and supply, respectively. The indices have different units and direct comparison of their magnitudes is not meaningful. The TTI has been shown to correlate well with myocardial oxygen consumption at normal and moderately elevated inotropic stimulation levels but to correlate poorly at high inotropic stimulation levels, because it does not account for changes in contractility.34,35 The TTI, thus, underpredicts the increase in oxygen consumption with elevated contractility and is a conservative estimate of oxygen consumption. The O2RVPTI oxygen supply index used here (Equation 9) does not consider coronary vascular resistance, which is unknown for this study, and thus is not a calculation of the actual rate of coronary oxygen supply. Autoregulation is not accounted for, but once maximum vasodilation has occurred, coronary resistance should be constant and the O2RVPTI should accurately describe the proportional relation between coronary flow and the driving pressure in the coronary arteries.16 The significant trends in these two indices (Figures 6a and 6b) and the conservative prediction of the TTI, thus, indicate the RV may be at risk of ischemia with hybrid or series TAL attachment. Right ventricular ischemia could have an effect on systemic hemodynamics.
Any decrease in mean systemic arterial pressure should stimulate the sympathetic nervous system and increase cardiac contractility, heart rate, and systemic vascular resistance. Although MAP was depressed throughout these experiments, due in part to the inflammation-induced elevation in PVR, all three parameters increased slightly, albeit not significantly, from baseline to series (Figures 4d, 10b, and 10c). Heart rate and SVR each showed the same correlation with attachment configuration in the second half of the experiment as in the first. This symmetry suggests that the target pressure of the baroreceptors reset to the low MAP levels of the experiments. Increased RV afterload may also have contributed to increased RV contractility, through homeometric autoregulation.35–37 De Vroomen et al.37 observed a sudden and sustained increase in RV, but not LV, contractility after the generation of pulmonary hypertension by partial PA banding. The trend in RV contractility in the present study (Figure 4d), although weak because of large variation at baseline, is in agreement with their results. Coronary perfusion pressure38 and oxygen tension,39 additionally, affect contractility. Interplay between the multiple pathways that govern contractility may contribute to the large variability observed in this parameter.
Mean left atrial pressure rose over the time course of the experiment (Figure 9). It was higher in the repeat configurations of the second half of the experiment than in the initial configurations of the first half, with the highest mean LA pressure occurring at the repeat baseline at the end of the experiment. This progression suggests the effects of other experimental factors outweighed the effect of TAL attachment configuration on LA pressure. Graft attachment to the LA may have caused this trend by physically damaging LA myocardia or by altering chamber geometry and thus impairing LA function. Left atrial dysfunction, however, does not appear to have affected LV function.
Reduced MAP throughout this study should have simultaneously reduced both LV afterload and perfusion. This effect tends to keep oxygen supply to and consumption by the LV in balance. No direct measurements were made in the LV, but indirect indicators suggest normal function. The trend in MAP (Figure 10a) indicates the LV responded as expected to changes in attachment configuration. Reduced PSR in parallel increased RV output and should have led to an increase in LV preload, an increase in LV output, and the observed increase in MAP; increased PSR in hybrid and series should have caused the opposite response, resulting in the observed decrease in MAP. The above sequence of events depends on change of attachment configuration causing an expected change in end-diastolic LV volume, or LV preload. The unexpected trend in mean LA pressure does not appear to have altered the expected trend in LV preload.
Resistance-Flow Curves for Clinical Prediction
The present experiments were conducted in “normal” animals. The hemodynamic response to TAL attachment, nevertheless, varied somewhat from animal to animal. Only three of six animals tolerated full series TAL attachment, and two of six required no PA constriction with parallel TAL attachment. The hybrid and series configurations with the present TAL placed significant stress on the RV. To use a TAL clinically, it will be essential to predict ahead of time which attachment configurations will be tolerated in an individual case. Especially important are the strength and reserve of the RV and the dependence of PVR on blood flow rate. Although based on a small number of cases, the observed differences between the resistance-flow rate curves for the full series and partial series groups in this study may suggest a method for making such a prediction.
The PVR-Q𝑄̅NL and CO-PSR curves describe the resistance-flow rate relations for the NLs and TAL/NL system, respectively (Figures 11a and 11b). Without the TAL, these curves would represent the same information. With TAL attachment, these curves represent different, albeit related, information. The NLs are a component of the TAL/NL system. Thus, the PVR-Q𝑄̅NL curve could influence the CO-PSR curve; an elevated PVR-Q𝑄̅NL curve would be expected to cause a depressed CO-PSR curve. Despite the elevated PVR-Q𝑄̅NL curve in the PS group with respect to the FS group in this study, however, the CO-PSR curve was also elevated in the PS group. The CO-PSR curve, thus, primarily described how CO was affected by RV afterload. Elevated PVR and possibly altered RV function, evident as a greater percent decrease in CO at high PSR, appear to have contributed to the difficulty of supporting increased afterload in the PS group.
Generation of resistance-flow rate curves by right heart catheterization might enable prediction of RV capability to pump against increased afterload. Gradual inflation of a balloon catheter to partially occlude one main PA branch would increase PVR, decrease CO, and increase flow to the contralateral lung, inducing recruitment in that lung. A Swan-Ganz catheter could record PA and pulmonary capillary wedge pressures. Simultaneous use of either transthoracic Doppler echocardiography or a Doppler flow wire could enable recording of instantaneous flow rate in the main PA and, possibly, in the unobstructed PA branch. From the recorded data, resistance-flow rate curves could be composed. The response of CO to elevated PVR, although not independent of NL recruitment, would describe cardiac function. The PVR-Q𝑄̅NL curve specific to the unobstructed lung, which could be recorded while keeping LA pressure relatively constant,40 would describe NL capillary recruitment. These responses could be used to predict which TAL attachment configurations would be tolerated.