INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pulmonary capillary pressure (Ppc) is the most important factor influencing lung edema formation (1), suggesting the potential significance of its clinical measurement. Current bedside estimates of Ppc such as pulmonary arterial wedge pressure (Ppaw) or the Gaar equation (2) are inaccurate in the presence of significant alterations in pulmonary vascular resistance. The most likely candidate for ultimately determining Ppc under a broad range of conditions is the pulmonary arterial occlusion method. This method is based upon the assumption that the pattern of emptying of the pulmonary capillaries can be derived from the decaying pulmonary arterial occlusion pressure waveform. In theory, the pressure at the proximal end of the pulmonary capillaries is represented by the point at which net capillary emptying begins. Consequently, by accurately characterizing the discharge of pressure from the capillaries, the initial capillary pressure can be determined.

After mechanical inflow occlusion in an isolated perfused lung lobe, an immediate rapid fall in pressure is followed by a slower, exponential decrease (3). The rapid and slow pressure drops have been said to represent the emptying of the arteries and capillaries, respectively, with Ppc theoretically occurring at the inflection point between the two (3). As applied in intact animals and patients, however, occlusion is achieved by rapidly inflating the balloon of a pulmonary artery catheter. In this setting, two separate emptying phases cannot be reliably visualized, thus obscuring the point at which capillary pressure should be derived. In an effort to identify Ppc, the pulmonary circulation has been represented with an electrical circuit model (3), in which the capillaries are the primary capacitor. If the initial postocclusion pressure decrease is characterized as linear, then the simplest corresponding electrical circuit model includes an arterial resistance but no capacitance. Using this model, a monoexponential curve fit corresponding to the discharge of the capillary capacitance has been applied beginning 200 to 300 ms after the instant of occlusion (8). To derive Ppc, this curve fit has been extrapolated back to the moment of occlusion (8) as well as to later time points suggested to correspond to the "inflection point" seen with isolated perfused lung studies (3, 8, 14).

Because the initial pressure drop does not appear linear when pulmonary artery occlusion is initiated by balloon inflation, other investigators have added an arterial capacitance to the electrical circuit model. Accordingly, a biexponential curve fit was applied to the pressure trace beginning at the instant of occlusion (5, 11, 15). Using this model, capillary pressure has invariably been derived at time zero, by two general methods: first, by applying circuit theory to solve for the value of Ppc using the model fit parameters, or second, by simply taking the initial value of the slow component of the biexponential equation.

The use of a Swan-Ganz catheter in conjunction with the arterial occlusion method (4, 8, 10, 16) can induce error in measurement of Ppc. Occlusion via balloon inflation occurs relatively slowly because of the time required for the catheter balloon to inflate and subsequently seat itself against the arterial walls. Consequently, it is difficult to define the instant of occlusion, potentially reducing the accuracy and precision of mathematical estimations of Ppc. In addition, the use of a fluid-filled catheter frequently introduces oscillation artifact associated with balloon inflation, preventing the mathematical analysis of the first 200 to 300 ms postocclusion (8). Consequently a biexponential equation cannot always be applied accurately to the pressure trace following occlusion.

In order for the arterial occlusion method to gain widespread clinical acceptance as a tool for measuring Ppc, the numerous methods for estimating capillary pressure must be compared and validated, and the sources of error addressed. In addition, the validation of this technique must be performed using a preparation that approximates the clinical situation. Toward these ends, we used dogs with intact lungs, and closed the chest of each animal after sternotomy for necessary instrumentation. In addition, we designed and used a pulmonary artery balloon-flotation catheter that has catheter-mounted transducers located proximal and distal to the balloon. This high-fidelity catheter allowed us to more precisely determine the instant of occlusion, as well as consistently analyze the earliest portion of the pressure-decay waveform. The purpose of this study was to compare several algorithms, to determine which ones yielded the most accurate and precise calculations of Ppc under control conditions, and during increased pulmonary arterial or venous tone. We validated our estimations of Ppc by comparing them with double occlusion pressure (Pdo), which has been shown to be an accurate measure of Ppc (20, 21), but due to its invasive nature, cannot be applied in humans.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experimental Preparation

Thirteen heartworm-free mongrel dogs of either sex (16 to 28 kg) were anesthetized with sodium pentobarbital (30 mg/kg intravenously, supplemented hourly with 2 mg/kg intravenously), intubated, and placed on a water-perfused heating pad. Mechanical ventilation was begun with a respirator (Harvard Apparatus, Millis, MA) with 5 cm H₂O positive end-expiratory pressure. Tidal volume was set at approximately 15 ml/kg and the ventilatory rate was adjusted to maintain arterial PCO₂ between 35 and 37 mm Hg and arterial PO₂ at 80 to 105 mm Hg. Airway pressure was measured via a side-port connector placed in line with the inspiratory tubing. A saline-filled polyethylene catheter, to be used for fluid and drug administration, was placed in a femoral vein and advanced into the inferior vena cava. An identical catheter was inserted in a femoral artery and advanced into the thoracic aorta for arterial blood sampling and pressure measurement. A drip of dextran (1 ml/min) or 5% dextrose in water (3 ml/min) was begun and maintained throughout the experiment. Sodium bicarbonate was infused intravenously as necessary to keep arterial pH at 7.35 to 7.40. Arterial blood gas samples were withdrawn anaerobically into heparinized syringes. Blood gases were measured immediately with a blood gas analyzer (Instrumentation Laboratory, Lexington, MA). Systemic arterial and airway pressures were recorded using Statham P23 Db pressure transducers (Ventrex, Ventura, CA), and displayed continuously on a Beckman R612 eight-channel polygraph (Sensor Medics, Yorba Linda, CA). Zero pressure was considered to be at the level of the left atrium.

We performed a median sternotomy to open the chest. A pneumatic occluder (In Vivo Metric, Healdsburg, CA) was tied around the left lower lobe pulmonary vein, and a 2.5-French transducer-tipped catheter (Millar Instruments, Houston, TX) was advanced into the same vein, just past the occluder, via an incision in the left atrial appendage. A 7.8-French balloon flotation catheter (Millar) with two transducers, one located proximal, and the other one distal to the balloon, was inserted in the right external jugular vein and floated into the left lower lobe pulmonary artery. This catheter contained a thermodilution port for measurement of cardiac output (Baxter Edwards, Irvine, CA). Heparinized lactated Ringer's solution (1 ml/min) was dripped continuously through the catheter, except during protocol periods. Two chest tubes were inserted through bilateral incisions in the third intercostal space, and connected to a Pleur-evac (Deknatel, Inc., Fall River, MA) for blood drainage. The chest was then closed to reestablish negative intrathoracic pressure.

Pulmonary Arterial and Double Occlusions

Pulmonary arterial pressure (Ppa) measured by both the proximal and distal transducers, as well as pulmonary venous pressure (Ppv), were recorded at 200 Hz and displayed using a Power Macintosh 7100 personal computer (Apple Computer, Inc., Cupertino, CA) after analog-to-digital conversion of the transducer outputs. Arterial occlusions were achieved with rapid inflation of the balloon on the pulmonary artery catheter. We initiated double occlusions by simultaneously inflating the pulmonary artery catheter balloon and the pneumatic pulmonary venous occluder. All inflations were performed manually, or with the use of syringe drivers (Columbus Instruments, Columbus, OH) powered by compressed air. An electronic signal sent from the data-acquisition program triggered a solenoid valve within the syringe drivers, rapidly inflating the balloon. Variable programmed delays in the time at which balloon inflations began allowed better synchronization of the pulmonary arterial and venous occlusions, when necessary. Both arterial and double occlusions were recorded for a total of 12 s, typically for 2 s prior to occlusion and 10 s postocclusion.

Protocol

Before protocol initiation the animals were paralyzed with pancuronium bromide (0.1 mg/kg, intravenously), which was supplemented hourly (0.05 mg/kg, intravenously). Five to ten pairs of arterial and double occlusions were obtained under control conditions (n = 13), and during intravenous infusion of serotonin (30 µg/kg/min; n = 7) and/or histamine (19 µg/kg/min; n = 4). The order of arterial and double occlusions was randomized, and both occlusions in a pair were recorded within a span of 2 min. All occlusions were obtained at end- expiration, with the ventilator turned off. Cardiac outputs were measured in all but one of the dogs given serotonin, and in all dogs given histamine.

Analysis of Pressure Curves

Of the 5 to 10 occlusion pairs obtained during each experimental condition, three acceptable pairs were used for data analysis. An acceptable occlusion pair included a double occlusion in which arterial and venous occlusions occurred within 100 ms of each other. If more than three occlusion pairs met this criterion, then the three double occlusions whose two components occurred closest together were used.

All pulmonary arterial and venous signals were filtered (2-pole Bessel, cutoff at 18 Hz) prior to analysis. To estimate capillary pressure from arterial occlusion, it was first necessary to determine the instant of occlusion. This was done by analyzing the pressure signals arising from the proximal and distal pulmonary artery catheter pressure transducers. As the balloon inflated, the pressure recorded from the transducer proximal to the balloon increased sharply, whereas the pressure recorded from the distal transducer fell. Initially, the maximal difference between the two pressures during the first 2 s of the record was determined. This maximal difference was used as the threshold for detecting when the two traces truly diverged. The point at which the difference between the proximal- and distal-transducer pressures increased above the threshold value was defined as the instant of occlusion (Figure 1). We then fit the data from zero to 4.8 s postocclusion with a biexponential equation of the form

(1)

View larger version (20K): [in this window] [in a new window]   Figure 1.   Method of determination of instant of arterial occlusion. During the first 2 s of data recording, the maximal difference between the pressures measured by the proximal and distal pulmonary artery catheter transducers is measured. The instant of occlusion is subsequently identified as the point at which this maximal pressure difference is exceeded. PA proximal = pressure measured by transducer proximal to catheter balloon; PA distal = pressure measured by transducer distal to catheter balloon.

where Ppa(t) is the pulmonary arterial pressure as a function of time, t, after occlusion, and Ppaw is the pulmonary arterial wedge pressure. The gain constants (A and B) and rate constants ( and ) were determined using the Levenberg-Marquardt method (22). In addition, we fit the data between 0.2 and 3 s postocclusion with a monoexponential equation,

(2)

where H is the gain constant, and  is the rate constant. Ppaw was determined by averaging the pressures obtained from the distal pressure transducer between the 9th and 11th second of data recording.

Capillary Pressure Estimates

We derived several estimates of Ppc based upon methods previously used by other investigators. With the fit data beginning at the instant of occlusion, we employed the method of Collee and coworkers (5), which derives the value of the slow component of the biexponential equation at time zero, when  < . The resulting estimate of Ppc is described by the following equation:

(3)

Siegel and Pearl obtained an "exact" solution for Ppc by applying basic circuit theory to a two-capacitor model of the pulmonary circulation (7). We used a similar equation to calculate Ppc at time zero after fitting the biexponential equation shown above:

(4)

Two additional values of Ppc were calculated by interpolating the biexponential curve fit at 152 ms or by extrapolating the monoexponential curve to the same time point. This instant of interpolation/extrapolation was taken from work by Gilbert and Hakim (10), who determined that arteriolar flow stops at about 152 ms after the initial Ppa decay with occlusion. A final estimate of Ppc was made by extrapolating the monoexponential curve fit back to the instant of occlusion, as previously done by Hakim and coworkers (8).

As a means of comparing the curve fitting methods with the simple use of raw data, we examined the data values (spaced 5 ms apart) on each Ppa decay curve shortly after occlusion. This allowed us to determine the instant at which the difference between Ppa and Pdo was smallest. The average time point of all these determinations was then used to represent the instant on the raw data curve postocclusion at which Ppc should be estimated.

Determination of Pdo

We determined Pdo by computing a running variance (using 0.5 s of data) of the difference between the distal transducer pressure and Ppv over the entire recording. The instant at which this variance peaked was defined as the onset of occlusion. We considered occlusion to be complete by 0.5 s after the instant of peak variance. A line was then fit to the ensuing 3 s of points midway between the Ppa and Ppv signals. Pdo was defined as the value of that line fit at 0.5 s after the onset of occlusion. In some animals there was a slow linear drift upward or downward in the two pressures following their rapid convergence after occlusion. This method allowed us to determine the correct Pdo in such instances.

Statistics

For each animal, we obtained arterial and double occlusion pressures by averaging the results from the three acceptable occlusion pairs obtained under each condition. We then grouped arterial occlusions into systolic and diastolic events. By visually determining the point at which the pressures measured by the proximal and distal transducers began to separate, we were able to precisely locate the onset of occlusion in the cardiac cycle. Of the three occlusion pairs from which mean pressures were calculated for each condition, we separated arterial occlusions according to whether they occurred during systole or diastole. For each dog during each condition, multiple occlusions occurring in the same phase of the cardiac cycle were averaged. Double occlusions were not separated according to the time of occlusion, for more than 70% of these occlusions occurred during systole, and there was no consistent relationship between the time of occlusion and the resulting pressure. Because correlation coefficients and linear regression are poor indices of agreement between two measured variables (23), we chose to represent the agreement between arterial and double occlusion pressures with the bias, or the average difference between the two. In some instances, as a means of comparison with the previous results of other investigators, we determined correlation coefficients as well. The standard deviation of the bias was used as a measure of the precision, or repeatability of the capillary pressure estimates. The derived pulmonary arterial occlusion pressures, averaged from all occlusions, and sorted by cardiac cycle phase (systolic or diastolic), were compared with corresponding Pdo using paired t tests. T tests were also used to compare biases, as well as additional measured and calculated variables. For all analyses, p < 0.05 was considered significant. All data are expressed as mean ± SD.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pulmonary hemodynamic data for the dogs given serotonin and/or histamine are shown in Table 1. Serotonin significantly increased Ppa and maintained cardiac output, whereas histamine tended to decrease cardiac output and increase Ppa, although the changes were not significant. Figure 2 shows representative arterial occlusion curves from each experimental condition. Histamine caused a slow decay in Ppa after occlusion, which was associated with predominant venoconstriction, as evidenced by the close approximation of double occlusion and pulmonary arterial diastolic pressures (Table 1). By calculating arterial and venous resistances for each dog we determined that serotonin caused varying degrees of pulmonary arterial constriction and venoconstriction, the two extremes of which are shown in Figure 2. Predominant arterial constriction was associated with a rapid fall in Ppa after occlusion, whereas venoconstriction yielded a histamine-like pressure decay.

View larger version (20K): [in this window] [in a new window]   Figure 2.   Representative arterial occlusion pressure traces. Two traces are shown for serotonin infusion, indicating the extremes of varying degrees of arterial constriction and venoconstriction that were obtained.

The relationships between Pdo and the five pressures derived from arterial occlusion during all conditions are shown in Figure 3 and Table 2. Among the data fit from the instant of occlusion, both Ppc (exact) and Ppc (slow) markedly underestimated Pdo, whereas the three remaining algorithms agreed more closely with Pdo. Of these three, the biexponential curve fit interpolated at 152 ms yielded the closest approximation of Pdo, and was significantly different from the other two algorithms.

View larger version (18K): [in this window] [in a new window]   Figure 3.   Comparison of arterial and double occlusion pressures, all occlusions included. See Table 2 for abbreviations. Each symbol represents mean of three occlusions. (A) Ppc (exact): Bias (arterial occlusion pressure minus Pdo) = 5.5 ± 3.5 mm Hg; r = 0.64; Ppc (slow): Bias = 2.5 ± 2.7 mm Hg; r = 0.83. (B) Bias = 0.5 ± 2.0 mm Hg; r = 0.90. (C ) Bias = 1.3 ± 1.8 mm Hg; r = 0.91. (D) Bias = 0.2 ± 2.0 mm Hg; r = 0.89.

Because Ppc (exact) and Ppc (slow) were relatively inaccurate estimates of Pdo, we did not include them in further analysis. For the remaining three algorithms, we separated the occlusions according to the phase in the cardiac cycle in which they occurred, as shown in Table 3. Because in some animals all occlusions occurred during the same phase, the value of n is not constant from systole to diastole. Averaging occlusions for each pulmonary vascular state, the systolic arterial occlusion pressures obtained with the three Ppc-derivation methods ranged from 1.4 to 4.4 mm Hg higher than the pressures obtained from diastolic occlusions.

The difference between each arterial occlusion pressure and the corresponding Pdo is shown in parentheses in Table 3. With the exception of the monoexponential curve fit interpolated at 152 ms, systolic arterial occlusion pressures tended to overestimate Pdo, whereas diastolic arterial occlusion pressures underestimated Pdo. When all systolic occlusions were grouped together, derivations of capillary pressure at 152 ms were equally accurate when either a biexponential equation (bias = 0.7 ± 1.9 mm Hg) or a monoexponential equation (0.7 ± 1.6 mm Hg) was applied to the data. For diastolic balloon inflations, the monoexponential curve fit extrapolated to time zero yielded relatively accurate approximations of capillary pressure, yet appeared to exhibit poor precision during serotonin infusion, as did the same curve fit interpolated at 152 ms.

We further examined the systolic occlusion data to determine the optimal time points for derivation of Ppc. The optimal time point was defined as the instant of interpolation or extrapolation of the respective arterial occlusion curve fits that yielded the smallest average difference (< 0.1 mm Hg) from Pdo. Because the biexponential curve fits closely resembled the actual postocclusion pressure waveforms, we included the raw data in our analysis. The biexponential curve fit yielded optimal time points of 180, 199, and 145 ms postocclusion for the control, serotonin, and histamine protocol steps, respectively. The raw data yielded values of 198 ms for control, 183 ms for serotonin, and 142 ms for histamine, whereas the corresponding time points from the monoexponential curve fit were 109, 123, and 54 ms. We then averaged these three numbers and used the resulting time point to derive a value of Ppc for each algorithm. As shown in the top half of Figure 4, when occlusions from all protocol steps were grouped together, there were no significant differences in accuracy among the three algorithms. For each optimal time point, the respective biases for each protocol step are shown in the bottom half of Figure 4.

View larger version (16K): [in this window] [in a new window]   Figure 4.   Average difference between Ppao and Pdo for each optimal time point corresponding to each algorithm. Systolic occlusions only. (A) All occlusions are grouped together. (B) Occlusions are grouped according to protocol step. Raw data (175) = no curve fitting applied, interpolated at 175 ms; Bi (175) = biexponential curve fit from zero to 4.8 s postocclusion, interpolated at 175 ms; Mono (95) = monoexponential curve fit from 0.2 to 3 s postocclusion, extrapolated to 95 ms; Ppao = pulmonary arterial occlusion pressure; Pdo = double occlusion pressure.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

An approximation of Ppc is most commonly obtained clinically by measuring Ppaw. However, an increased pulmonary venous resistance will lead to a significant discrepancy between these two pressures, as has been shown in patients with adult respiratory distress syndrome (19). Consequently, a more accurate determination of Ppc may lead to more appropriate therapy in these patients. The arterial occlusion method is the most likely potential means of accurately determining Ppc at the bedside.

Methodology

The arterial occlusion method has been applied in isolated perfused lungs (3, 9, 21), in quasi-intact lung preparations (8, 12), in intact-lung animals (4, 6, 10, 17), and in patients (5, 14, 16). Both mechanical occluders and balloon inflation have been used to stop pulmonary arterial flow. The myriad experimental preparations and means of calculating Ppc have increased our knowledge of pulmonary vascular mechanics, but have not been reconciled in an attempt to develop a Ppc measurement tool for use at the bedside. In the current study we have used a clinically applicable form of the arterial occlusion method to determine the most accurate and precise algorithms for estimation of Ppc. In addition, we have used a preparation that, given experimental limitations, closely approximates the closed-chest patient. Because this is the first study to compare arterial and double occlusions in animals with completely intact lungs, it serves as a more advanced forum for determining the proper methodology for institution of the arterial occlusion method.

In theory, the application of curve fitting for estimation of Ppc should include the earliest portion of the pressure decay, which requires the ability to precisely determine the point of occlusion. Yamada and coworkers (16) introduced a pulmonary artery catheter that allows pressure measurement via fluid-filled ports proximal and distal to the balloon. Occlusion was identified as the moment at which the two pressure traces sharply diverged. Although this catheter increased the ease of identification of the time of occlusion, its low frequency response allows for occlusion artifact, thus frequently obscuring the first 200 to 300 ms of the postocclusion pressure traces (8). We have advanced the method of Yamada and coworkers (16) by replacing the fluid-filled measuring ports with catheter- embedded transducers. This high-fidelity catheter enabled the computer to consistently identify the point of occlusion, as well as characterize the initial portion of the decaying pressure trace.

To validate our estimations of Ppc, we compared all derived Ppc with measured Pdo. The double occlusion method for determining Ppc is based upon the same assumption as the arterial occlusion method: the pulmonary capillaries are the principal site of vascular capacitance. Pdo has been shown to correlate closely with capillary pressure measured by the isogravimetric technique (20), a method that determines the pressure at the major site of fluid filtration. Pdo was therefore the logical choice for a reference capillary pressure owing to its ease of measurement with little additional instrumentation. For Pdo to be considered a valid estimate of Ppc, arterial and venous occlusions must occur at the same time. We considered double occlusion to be acceptable if arterial and venous occlusions occurred within 100 ms of each other. Although this criterion is somewhat arbitrary, it resulted in an average coefficient of variation for Pdo of only 3.6% for sequential measurements.

Almost all studies comparing two methods of estimating Ppc have used correlation coefficients or linear regression. As pointed out by Altman and Bland (23), these are measures of association, rather than of agreement. We have therefore avoided using them as our primary method of comparison, in favor of the mean difference between two measurements, or the bias. However, in order to compare our results with those of other investigators, we have, when necessary, calculated correlation coefficients.

Of the five algorithms initially used to estimate Ppc, the poorest approximations of Pdo were Ppc (slow) and Ppc (exact), both of which are based upon a two-capacitor, two-resistor model of the pulmonary circulation. Ppc (slow) is based upon the assumption that net capillary emptying begins immediately after occlusion and can be described by the same exponential throughout the pressure decay to Ppaw. Siegel and Pearl have criticized Ppc (slow) as being conceptually inaccurate when used in association with a two-capacitor model of the pulmonary circulation (7). These investigators derived a solution for Ppc that depends upon the mathematical interactions of the model components, namely the arterial and capillary capacitances, as well as arterial and venous resistances. However, Ppc (exact), which utilizes Siegel and Pearl's application of basic circuit theory, was an even more inaccurate estimate of capillary pressure. Although the application of simple circuit theory may be the correct way to conceptualize capillary pressure decay after pulmonary arterial occlusion, it is apparent that its practical application did not work. This may be due to the limitations of the model of the pulmonary circulation, or because of the difficulty in accurately determining rate constants in the presence of cardiac pulsation and relatively slow occlusion.

The remaining three algorithms yielded better approximations of Pdo. However, upon separation of occlusions into systolic and diastolic events, we noticed a tendency for diastolic occlusions to be associated with poorer precision. The lesser ability of diastolic occlusions to consistently predict Pdo was a result of limitations in the method of identification of the instant of occlusion. Because in some dogs the peak systolic portion of the pulmonary arterial distal pressure trace was truncated relative to that of the proximal pressure, a gradient existed between the two pressures prior to occlusion. Consequently, a longer period of time was required to exceed this initial gradient during occlusion. This delay in identification of occlusion was significant only during diastole, when the flow and corresponding rate of pressure decay is much reduced. Therefore, diastolic occlusions tended to underestimate Pdo. This tendency was particularly evident during serotonin infusion, when pulse pressure, and therefore the baseline peak systolic pressure gradient between pulmonary arterial proximal- and distal-transducer pressures was greatest. Because of this potential error associated with diastolic occlusions, estimation of Ppc using curve fitting methods should probably only be applied to systolic occlusions.

Gilbert and Hakim (10) stated that Ppc should ideally be derived at the point at which arteriolar flow ceases, which they determined to be about 152 ms postocclusion when balloon inflation occurs during systole. Our data indicate that when Ppc is derived at 152 ms, application of a biexponential fit does not improve the estimation of Pdo over that obtained from a monoexponential fit. These results are in general accordance with the findings of Gilbert and Hakim regarding arteriolar flow. However, we found that the agreement between arterial occlusion pressure and Pdo varied during control, serotonin, and histamine conditions. It is not clear whether this disagreement is caused by changes in the time at which arteriolar flow ceases, or by alterations in microvascular resistance. According to the model upon which each is based, the Pdo represents the pressure at the midpoint of the capillary bed, whereas arterial occlusion pressure theoretically occurs at the proximal end of the capillaries. A significant increase in microvascular resistance will increase the discrepancy between arterial occlusion pressure and Pdo, as has been shown to occur during serotonin and histamine infusion (24).

By averaging the optimal time points for derivation of Ppc during control, serotonin, and histamine conditions, we attempted to utilize one time point at which Ppc could be calculated, regardless of the state of the pulmonary vasculature. By doing so, we were able to use a biexponential fit, a monoexponential fit, or even the raw data, with equal accuracy and similar precision, albeit with different optimal time points for each method. By taking into account the effects of serotonin and histamine, we are able to reduce some of the previously mentioned potential error associated with comparing arterial occlusion pressure and Pdos. It should be noted, however, that the inability to distinguish among the two curve fitting methods and the raw data in terms of accuracy may be due to the relatively limited number of animals, and therefore power to detect differences, involved in the study.

The fact that the raw data may be used in place of an exponential fit has potential clinical significance. By forgoing curve fitting, only about 200 ms of the pressure decay curve need be analyzed, assuming a sufficiently rapid occlusion. This would allow occlusions to be initiated between breaths, thereby obviating cessation of ventilation. However, use of the raw data is dependent upon occlusion being sufficient to prevent small pulsations in the initial pressure decay, equivalent to the smoothing effect of an exponential fit. In addition, it is not clear whether this method could be used in conjunction with a standard fluid-filled pulmonary artery catheter, because of the frequent presence of artifact associated with occlusion. These potential sources of error make it unlikely that exponential curve fitting could easily be replaced in clinical practice.

In summary, we have shown that the most accurate and precise means of determining Ppc is to estimate an optimal time point on the pulmonary arterial occlusion pressure curve. This time point occurred 95 ms after the instant of occlusion when a monoexponential curve was fit to the data, and at 175 ms when a biexponential curve fit was applied. A biexponential curve fit represents the pressure decay sufficiently well so that the raw data may be used in its place, with the same time point for derivation of Ppc, with little sacrifice in precision. Each optimal time point accurately represents Ppc despite changes in arterial and venous resistance and compliance. Although this study utilized transducer-tipped catheters, these methods could potentially be used with standard fluid-filled catheters, provided an instant of occlusion can be defined and occlusion-related artifact is minimal.