INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The beneficial effects of ventilatory strategies directed at minimizing cyclic parenchymal stretch during mechanical ventilation have been suggested (1). Using low distending pressures and permissive hypercapnia (PHY), Hickling and coworkers found evidences suggesting an improved survival in ARDS patients (2). In a recent prospective and randomized study on ARDS, the association of low distending pressures, PHY and ideal PEEP levels (a strategy from now on called lung protective approach [LPA] resulted in increased chances of early weaning and lung recovery (1) with a lower ultimate ICU mortality rate (5).

The massive presence of alveolar collapse during the ARDS processcyclic or persistenthas been associated with an exacerbation of ventilator-induced lung injury (3, 6, 7). Accordingly, the use of PEEP_IDEAL as part of a lung protective strategy has a good rationale (1, 3). By keeping end-expiratory pressures above the lower inflection point (PEEP_FLEX) of the inspiratory pressure-volume (P-V) curve, one can expect near-maximal alveolar recruitment (8), minimal alveolar reopening with better distribution of tidal ventilation (9), and a consequent reduction of shear forces with an attenuated lung injury (6).

But if PEEP has been consistently associated to a lung protective effect a major concern remains regarding its global hemodynamic impact. The well-known depressant effects of PEEP on venous return and systemic circulation (10, 11) is still considered as a potential source of problems in critically ill patients, specially when considering the great proportion of ARDS patients with hemodynamic instability.

On the other hand, PHY is also a strategy encompassing important hemodynamic consequences. Contrary to PEEP, however, its characteristic impact is a marked hyperdynamic response, usually the net result of a strong sympathetic stimulation overcoming the direct depressant effects of carbon dioxide on cardiac contractility (12). In animals and anesthetized human subjects submitted to acute hypercapnia, this response is very predictable, with an increase in cardiac output, heart rate and pulmonary arterial pressures, a maintained or slightly increased stroke volume, and an associated decrement in systemic vascular resistance (12). The same hyperdynamic response has also been observed in ARDS patients, as long as an acute and significant raise in arterial Pa_CO2 is allowed (15).

Considering then these opposing tendencies, what should we expect from the complex association between PHY and PEEP_IDEAL? Could the depressant effects of PEEP on systemic circulation be counterbalanced by the stimulatory effects of PHY? Should we expect a dangerous sum of effects of PHY and PEEP_IDEAL upon the pulmonary hypertension frequently found during the ARDS process? The answers to these questions are not obvious and, as far as we know, no consistent data has addressed this point in the medical literature.

In December 1990, we started a prospective randomized study on mechanical ventilation comparing a lung protective strategy with the conventional wisdom in ARDS management. Although the primary endpoint of this protocol was a survival analysis (5), the study was also designed to provide a parallel examination of the hemodynamic consequences of PHY associated with ideal PEEP levels (2 cm H₂O above the lower-P_FLEX).

The present study concerns this prospective collection of hemodynamic data. As we had a previous expectation about the fleeting nature of PHY effects, the primary goal of this study will be a descriptive analysis of its major hemodynamic consequences along the time. By using a multivariate analysis, we also tried to separate the respective contributions of PHY and PEEP_IDEAL to the observed hemodynamic effects.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Since December 1990, 48 patients with ARDS were enrolled in this study. The present report includes the first 28 patients already described in a previous publication focusing the pulmonary effects associated with the New Approach (1). Criteria for entry were an underlying disease process known to be associated with ARDS along with a Lung Injury Score (LIS)  2.5 (19) and a pulmonary arterial wedge pressure (PWEDGE) < 16 mm Hg. Exclusion criteria were previously defined (1). The study protocol was approved by the Hospital's Medical Ethics Committee and informed consent was obtained from the patient or next of kin.

Initial Homogenizing Procedures

After verifying a provisory LIS > 2.5, a Swan-Ganz no. 7F catheter (Baxter) was inserted in a central vein. Besides radiological control, the validation of zone-3 condition of the occluded pulmonary arterial catheter tip was assessed by two mechanical maneuvers (1, 20).

After confirming a PWEDGE < 16 mm Hg, all selected patients were pre-expanded until observing a PWEDGE > 8 mm Hg and a hemoglobin > 10 g/dl, receiving then similar infusions of dobutamine (target doses of 5 µg/kg/min) and nonadrenaline (minimal doses to maintain average systemic blood pressure > 60 mm Hg). They were subsequently sedated and paralyzed, and submitted to a fixed ventilatory support with : volume controlled ventilation, tidal volume (VT) 10 ml/ kg, respiratory rate 15 cycles/min, FI_O2 100%, and PEEP 5 cmH₂O or the minimum needed to maintain the arterial oxygen saturation > 85%.

Thirty minutes after the stabilization of these conditionswith stable and equivalent doses of vasoactive drugsa full set of respiratory, hemodynamic, and laboratory measurements was obtained. This period was called control-period and the baseline measures obtained at this time were used to calculate the predictor scores shown in Table 1. This initial but provisory infusion rate of catecholamines was strictly kept constant from this time until the next measurements corresponding to Time 1, a time period when both ventilatory strategies were implemented. The hemodynamic optimization (see below) was started only after collecting this second set of data (after Time 1).

Criteria for defining sepsis at entry and organ-failure were the same from the previous publication (1). Pneumonia at entry was objectively assessed by a preestablished algorithm for infectious care used in our unit.

A bedside P-V static curve was obtained for all patients, soon after the control period. Although P_FLEX determinations were useful only for the LPA group, the procedure was performed in all patients for homogenizing purposes. The overall procedure was done without disconnecting the patient from the ventilator, following all the steps described in a previous publication (1). A well-defined lower inflection point (P_FLEX) could be obtained in 44 patients.

Randomization

After obtaining the P-V curve, the patient was randomly assigned (sealed envelopes) to one of the two arms of the protocol: the conventional arm (C) and the lung-protective-approach arm (LPA), consisting of two different ventilatory strategies along with identical hemodynamic and general support. All the nutritional and dialyzing procedures, as well as the infectious care were kept the same for both groups, being described elsewhere (1).

General Ventilatory Support

The two different ventilatory approaches were immediately initiated and rigorously maintained until extubation or death. The P-V curve was performed only once, before the randomization. All patients were ventilated with an Inter-7 ventilator (INTERMED Equipamento Médico-Hospitalar LTDA, São Paulo, Brazil) or a Servo Siemens 900C (Siemens-Elema Sweden) connected to a closed system for aspirating tracheal secretions. In both groups, the target Pa_O2 was 80 mm Hg and the PEEP level was never set below 5 cm H₂O. The weaning protocol was the same for both arms (1).

The Conventional Approach (Control Group)

Arterial PaCO₂ levels were kept between 25 and 38 mm Hg (target = 35 mm Hg), independently of airway pressures. Priority was given to the maintenance of an FI_O2 < 60%, with adequate oxygen transport indexes (DO₂I). To reach these goals, the initial ventilatory parameters were: V_T = 12 ml/kg; volume assisted/controlled mode; flow-rates adjusted to avoid auto-PEEP and to keep inspiratory/expiratory ratio < 1:1; inspiratory pause = 0.4 s; respiratory rates between 10 and 24 cycles/min (depending on Pa_CO2); and minimum PEEP levels  5 cm H₂O, which were increased according to a stepwise algorithm (1) looking for a compromise between a safe FI_O2 and a safe hemodynamic profile.

The Lung-Protective-Approach (LPA)

Priority was given to the maintenance of an open lung, independent of hemodynamic concerns or considerations about the required FI_O2. Parenchymal overdistention was also aimed, disregarding Pa_CO2 levels. The following general strategies were adopted: (1) VT was kept < 6 ml/ kg and the respiratory (RR) < 30 cycles/min even during pressure support cycles; permissive hypercapnia and continuous infusions of fentanyl and diazepam were used as needed to avoid discomfort and tachypnea. Initial Pa_CO2 levels of up to 80 mm Hg were tolerated and slow sodium bicarbonate infusions were permitted if pHa < 7.2; (2) Inspiratory pressures above PEEP (termed driving pressures) were kept below 20 cm H₂O and peak airway pressures were kept below 40 cm H₂O; (3) The initial PEEP/CPAP level was preset at 2 cm H₂O above P_FLEX; independent of hemodynamic considerations or FI_O2 levels (this level was called PEEP_IDEAL). When auto-PEEP was present, the total PEEP (external PEEP plus auto-PEEP) was considered and adjusted in order to match this PEEP_IDEAL. Whenever a sharp P_FLEX could not be determined on the P-V static curve, a total PEEP level of 16 cm H₂O was initiated, corresponding to the mean value of PEEP_IDEAL found in a previous unpublished study on ARDS at our institution; (4) PC-IRV was the ventilatory mode of choice during critical periods of severe lung injury (signalized by FI_O2 requirements above 50%). Once the patient presented signals of lung recovery (FI_O2 < 50%) he was stepwise placed under VAPSV (21) or PSV.

A complete description of all the stepwise maneuvers employed in each arm was presented elsewhere (1).

Hemodynamic Measurements and Support

Mean pulmonary artery pressure (PAP_M) and other intravascular pressures including PWEDGE were measured at end-expiration, maintaining the patient connected to the ventilator. Cardiac output was measured by the thermodilution technique, using 5-6 injections of 10 ml cold saline (< 4° C) started at random relative to respiratory cycles. All measurements were registered in an HP-78339A monitor (Hewlett-Packard, USA), used to calculate vascular resistance indexes, left and right ventricle stroke work indexes, the oxygen delivery (DO₂I), the oxygen consumption (O₂I), and the extraction ratio, according to standard formulas. Systemic and pulmonary arterial blood samples were collected soon after cardiac-output determinations.

A full set of measurements, including plasma lactate, blood-gases and hemoglobin determinations was performed one hour after the initiation of the randomized strategy (Time 1), and again after precisely 8, 16, and 24 h. Subsequently, at least one set of measurements was performed each day until extubation or death. When frequent measures were required during the day (for hemodynamic optimization), the average value was used as the representative value for that day.

Soon after collecting data regarding Time 1, patients in both arms of the protocol were submitted to a stepwise algorithm of hemodynamic optimization (1) strictly followed till extubation, which encompassed: (1) target hemoglobin concentrations  12 g/dl; (2) a minimal dobutamine infusion-rate of 5 µg/kg/min for all; (3) maintenance of PWEDGE < 16 mm Hg; and (4) dopamine infusions at doses < 3 µg/kg/ min whenever oliguria was present. Rigid guidelines for packed red blood cells and colloid infusions, daily negative fluid balances, avoidance of signs of systemic oxygen debt; and progressive doses of dobutamine (as needed, up to 30 µg/kg/min) or norepinephrine were applied (1).

Sedation

Only fentanyl (up to 8 mg/d) and diazepam (up to 250 mg/d) were used in this study, sometimes associated with pancuronium administered intermittently. The protocol for sedation was the same for both arms, but patients in LPA certainly received higher doses since generous infusions of sedatives were employed at any signal of increased respiratory drive caused by hypercapnia. Fentanyl (continuous infusion) was used for all patients, even during weaning. During daily measurements of compliance in both arms, all patients received transient infusions of succinylcholine (1 mg/kg).

Statistics

Whenever possible, statistical tests used were shown in figures and tables. All reported p values are two-tailed and calculated with the BMDP statistical software.

The comparative analysis of respiratory and hemodynamics parameters along the time (Tables 3 and 4 and Figures 1-5) was performed by considering the incremental areas under the individual curves (AUC) which represented the average tendency of a variable along some specific time-interval. The area was calculated for each patient after considering the measurements during the control-period as the relative origin. Thus, negative areas indicated a progressive decrement in the respective variable from the control period up to the considered time-interval. Three time-intervals were considered: the immediate-changes interval (the area under the short line linking the control measure to the measure at Time 1), the early-changes interval (the area under the line joining the five measurements performed from Time 1 up to the first 36 h) and the late-changes interval (the area under the line joining the six measurements collected from the second up to the seventh day). To avoid bias that might occur in the analysis if a patient did not complete a whole interval (producing an artifactual decrement in the calculated area or trend), these areas were divided by the time span of recordings for each patient (obtaining so an average trend parameter). Differences in these trend parameters between both groups (LPA versus C) were analyzed by independent-sample t tests with appropriate transformations.

View larger version (43K): [in this window] [in a new window]   Figure 1.   Evolution of pH and PaCO2 along the first week of mechanical ventilation. Bars represent 1 SEM. Early changes (shaded ) are artificially split from late changes. The corresponding significance level (analyzing AUC) expresses the different evolution between both groups during the interval. *p < 0.00001 for the immediate differences from control-period up to Time 1. NS = not significant; AUC = incremental area under the curve.

View larger version (42K): [in this window] [in a new window]   Figure 2.   Evolution of cardiac index and heart rate along the first week of mechanical ventilation. Bars represent 1 SEM. Early changes (shaded ) are artificially split from late changes. The corresponding significance level (analyzing AUC) expresses the different evolution between both groups during the interval. *p < 0.0002 for the immediate differences from control-period up to Time 1. NS = not significant; AUC = incremental area under the curve.

View larger version (46K): [in this window] [in a new window]   Figure 3.   Evolution of vascular resistances along the first week of mechanical ventilation. Bars represent 1 SEM. Early changes (shaded) are artificially split from late changes. The corresponding significance level (analyzing AUC) expresses the different evolution between both groups during the interval. *p = 0.00005 for the immediate differences from control-period up to Time 1. NS = not significant; AUC = incremental area under the curve.

View larger version (47K): [in this window] [in a new window]   Figure 4.   Evolution of pulmonary arterial pressure (mean) and PWEDGE along the first week of mechanical ventilation. Bars represent 1 SEM. Early changes (shaded) are artificially split from late changes. The corresponding significance level (analyzing AUC) expresses the different evolution between both groups during the interval. *p < 0.0005 for the immediate differences from control- period up to Time 1. NS = not significant; AUC = incremental area under the curve.

View larger version (46K): [in this window] [in a new window]   Figure 5.   Evolution of oxygen delivery and plasma lactate along the first week of mechanical ventilation. Bars represent 1 SEM. Early changes (shaded ) are artificially split from late changes. The corresponding significance level (analyzing AUC) expresses the different evolution between both groups during the interval. *p < 0.0005 for the immediate differences from control-period up to Time 1. NS = not significant; AUC = incremental area under the curve.

The absolute values of AUC so obtained estimating the changes or "" in each parameter were also submitted to a stepwise multiple linear regression aimed at scrutinizing the relationship between the changes in respiratory parameters (the AUCs obtained for PEEP, Pa_CO2, airway pressures, compliance measurements, etc.) and the changes in hemodynamic parameters (the AUCs calculated for the PAP_M recordings, PWEDGE, cardiac index, etc.) during the same time interval. We decided to focus the analysis on the immediate changes only (from control to Time 1 period), avoiding complex interactions along the time. The normality assumption was always checked for the dependent variables.

As there was a high degree of multicollinearity among the variables studied, the reported p values were not adjusted for multiple comparisons, except for those corrections intrinsically performed during the multivariate analysis.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The patient's characteristics and the primary diagnosis at entry are summarized on Tables 1 and 2. No statistically significant differences could be detected between both arms.

As established by the protocol, the installation of permissive hypercapnia was abrupt. Starting with a Pa_CO2 close to 35 mm Hg at the control period, the patients in LPA experienced an instantaneous increment of 23 ± 11 mm Hg (mean ± SD) in the Pa_CO2 levels at Time 1. The maximal Pa_CO2 observed was 114 mm Hg. The average maximal value reached during the first week was 70 ± 18 mm Hg. The minimal pHa observed was 6.88 and the average minimal pHa during the first week was 7.13 ± 0.14. Three patients in LPA never exhibited a pHa below 7.30.

As shown on Figure 1, the average Pa_CO2 in LPA remained significantly elevated ( 20 mm Hg above the values found in C) along the whole week (p < 0.00001). In contrast, the arterial pH exhibited a progressive recovery to baseline conditions soon after the first 36 h. Eleven patients received slow bicarbonate infusions (< 50 mEq/h with a mean total dose of 380 ± 260 mEq) and, among them, seven were submitted to dialysis.

The evolution of hemodynamic variables were analyzed all along the first week of mechanical ventilation (Figures 2-5). In spite of the equivalent infusions of fluids and vasoactive drugs accomplished in both arms (Table 3), the hemodynamic profiles significantly differed between both groups and according to the interval studied:

Baseline Measurement

As expected, both groups presented a very similar hemodynamic condition at the control-period, receiving a rather similar hemodynamic support (METHODS). At this time, both arms presented a global hemodynamic profile compatible with a hyperdynamic state and high plasma lactate levels (Figures 2-5, see control measures)

Early Changes

The major hemodynamic effects associated with LPA were transitory, lasting for only 36-48 h. Most effects were maximal at time 1 (a time-period during which the catecholamine infusions were kept as during the control-period), decreasing progressively after that. Some few effects (for instance, lactate changes and the increments in PWEDGE, Figures 4, 5) achieved their maximal expressions some hours later.

Figures 2-5 and Table 3 demonstrate that there were three major blocks of early hemodynamic effects associated with LPA: (1) the hyperdynamic changes in the systemic circulation (increase in cardiac output and heart-rate, associated with a fall in systemic vascular resistance); (2) a large increment in pulmonary vascular pressures (increase in PAP_M, right atrial pressure and PWEDGE, resulting in an increment of the right ventricular work of more than 40%); and (3) the alterations in the systemic oxygenation status (increase in DO₂I and Pv_O2, associated with a latter fall in plasma lactate levels).

Late Changes

As shown on Figures 2-5 and Table 3, the major hemodynamic effects associated with the LPA are no longer observed after the first 36 h. This return to baseline conditions coincided with the progressive recovery of arterial pH observed afterwards (Figure 1). In order to point out this transient aspect, Figures 1-5 were split into the early-changes-interval (up to first 36 h) and the late-changes-interval, with separate statistical analysis.

In contrast, Table 4 demonstrates the persistence of some beneficial respiratory effects in LPA. The static compliance improved progressively (p < 0.00001) and the Pa_O2/FI_O2 ratio remained well above that of C along the whole week (p < 0.00001), despite progressively lower mean airway pressures.

At the end of the first week of mechanical ventilation, the patient in the LPA group had received a mean of 200 ± 114 ml/day of packed red blood cells, whereas the patient in C received 314 ± 290 ml/day (p = 0.09). The mean cumulative fluid balance was: 5.2 ± 3.7 l for LPA: and 6.0 ± 3.8 l for C group; (p = 0.70)

Multivariate Analysis

Considering the major hemodynamic alterations observed in LPA at Time 1 (immediate changes), we performed a stepwise multiple regression analysis focusing the following questions:

1. Were the changes in cardiac output related to changes in airway pressures or to PHY? (dependent variable =  cardiac output; independent variables examined: Pa_CO2, Pa_O2, pH, PEEP, plateau pressures, peak pressures, mean airway pressures, VT, RR, minute ventilation, compliance, Pv_O2). The immediate increments in cardiac index (CI) were not related to PEEP changes (p = 0.48) or to the changes in mean airway pressure (p = 0.45), either in a univariate or in a multivariate model. In contrast, a strong and positive relationship was found between CI and Pa_CO2 changes (r = 0.47; p = 0.002, best multivariate model), and a negative association between CI and changes in plateau pressures (p < 0.0001; r = 0.57; best multivariate model encompassing Pa_CO2 also as an independent variable).

2. Were the immediate changes in pulmonary vascular pressures related to changes in airway pressures or to PHY? (dependent variable = PAP_M; independent variables = the same scrutinized in the analysis above). The correlation between pulmonary hypertension (PAPM) and Pa_CO2 changes was very strong in a univariate model (r = 0.68; p < 0.00001) and this positive association was maintained even after forcing the inclusion of PEEP, P_PLAT or  mean-airway-pressures in a multivariate model (r = 0.49; p = 0.001). In the same way, the variations in filling pressures (dependent variables = -right-atrial-pressure or PWEDGE) exhibited significant and positive relationships with the increments in Pa_CO2 (r = 0.55; p = 0.0002, considering PWEDGE versus Pa_CO2; and r = 0.61; p = 0.0001, considering RAP versus Pa_CO2) which were maintained even after the forced inclusion of PEEP in a multivariate model (p < 0.01, for both). The high degree of collinearity between PEEP and Pa_CO2 preclude us from discarding some PEEP effect upon the raised vascular pressures. However, the analysis above suggested that a great part of the increment in vascular pressures was associated to Pa_CO2- effects independently of PEEP influences.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study showed that permissive hypercapnia associated with ideal PEEP levels can be acutely induced without any obvious deleterious influence on hemodynamics. Even in the context of a great proportion of patients with sepsis and hemodynamic instability (Table 1) with many patients needing nonadrenaline infusions to keep arterial blood pressures  60 mm Hg this combined strategy was well tolerated, producing a consistent hyperdynamic response.

Although this study was not blinded to the investigators, the equivalent hemodynamic support strictly accomplished in both groups (Table 3), allowed us to interpret the different hemodynamic profiles as an almost exclusive consequence of the different ventilatory strategies.

The immediate hyperdynamic response observed in LPA, characterized by an enhanced cardiac output due to an increased heart rate, a maintained stroke volume and a decreased systemic vascular resistance, is consistent with the reported effects of acute hypercapnia in intact animals and anesthetized humans (12). This similarity suggests that the observed response in LPA was predominantly a PHY response. Despite the marked increment of PEEP and mean airway pressure, we did not find any indication of cardiovascular depression in these patients. Accordingly, either because these patients had a very low lung compliance with a damped transmission of end-expiratory pressures to the cardiovascular system or because PHY could obscure any possible depressant effect of high PEEP levels, the combined use of PEEP_IDEAL was extremely well tolerated in our patients. The net result was almost a pure PHY effect. Of note, there was a spontaneous attenuation of this hyperdynamic response along the first 36 h, despite persisting hypercapnia. This fading effect coincided with the progressive return of arterial pH to baseline levels.

The absence of depressant PEEP effects in our work must be considered into the particular context of this study design: although using PEEP levels as high as 24 cm H₂O in the LPA group, the distending pressures above PEEP never exceeded 20 cm H₂O (and the absolute P_PLAT never exceeded 40 cm H₂O). Had the strict control of P_PLAT in this group been less rigid, the results might be rather different.

Some recent studies have observed a marked depression of the cardiovascular system associated with the use of high VT and consequently of high P_PLAT (22, 23). This observation allow us to speculate that part of the depressant effects classically attributed to PEEP was, in fact, the result of associated high inspiratory pressures and not a direct PEEP-effect per se. The result of our multivariate analysis, indicating a strong and negative correlation between changes in cardiac output and changes in P_PLAT is in consonance with this reasoning. Of note, we did not find any correlation between changes in PEEP and changes in cardiac output (p = 0.48).

Our findings differ from the results of McIntyre (16) and Simon (17) who did not find significant changes in hemodynamic variables during PHY. In our opinion, the gradual and modest increment in Pa_CO2 obtained in both studies can easily explain these differences. Actually, despite the combined use of PEEP_IDEAL, our data are in close agreement with the short term hyperdynamic response observed by Puybasset (15) and Thorens (18).

The most striking effect associated with LPA was the immediate pulmonary hypertension. The mean increment of PAP_M was about 9 mm Hg and, as suggested by the multivariate analysis, this increment was closely related to the raise in Pa_CO2 (each 10 mm Hg increment in Pa_CO2 was associated to a 3 mm Hg increment in PAP_M: 95% confidence interval: 1.6- 4.4 mm Hg). In association with pulmonary hypertension, two other hemodynamic perturbations were also observed: an overflow condition in the pulmonary vascular bed created by an increased cardiac output and an increased effective downstream pressure observed as acute elevations of PWEDGE in this group. Interestingly, we did not observe any increment in pulmonary vascular resistance associated to PHY (p > 0.20). Considering all these findings together, the logical conclusion is that PHY was related to pulmonary hypertension through indirect mechanisms: by its stimulating effects on cardiac output (as discussed previously) and by increasing the cardiac filling pressures (elevating all the pressures in the pulmonary vascular systemdiscussed later). Apparently, the pulmonary vascular tone was not directly affected by LPA and this result differs from previous studies on PHY effects, where the authors showed a consistent vasoconstrictor response of the pulmonary bed to hypercapnia (12, 15, 24, 25).

The data above must be carefully analyzed, since some controversy stands on the exact meaning of the raw calculations of pulmonary vascular resistance extracted from single-point measures (26, 27). As already demonstrated, in face of an acutely increased cardiac output and/or an associated raised PWEDGE, both conditions occurring in LPA, the assumption that PWEDGE represents the effective vascular outflow pressure might lead to a fake reduction in the calculated value of pulmonary vascular resistance, despite a well-preserved vascular tone (26, 27).

Another word of caution should be dedicated to the fact that this lack of vasoconstrictor response was obtained during a marked and concomitant change in airway pressures. Considering the J-shaped function of pulmonary vascular resistance according to progressive PEEP levels (28), one could speculate that our patients were below to a reasonable lung volume during the control period and that the use of PEEP_IDEAL brought these patients closer to their optimal lung volume where the vascular bed is maximally recruited and the vascular resistance reaches its minimal.

Therefore, both mechanisms above could have obscured some underlying increment in pulmonary vascular tone associated to PHY. Interestingly, some recent reports showed a controversial and direct vasodilator effect of CO₂ on pulmonary vessels (12, 29) and, in a recent study in adult patients with ARDS, Thorens et al also observed a preserved "single-point" pulmonary vascular resistance during PHY (18).

A marked and immediate increment in filling pressures, RAP and PWEDGE, was observed in the LPA group. We speculate that this phenomenon was the combined result of two conditions: the use of PHY and the use of high PEEP levels.

According to some experimental data, a marked loading effect may be produced by PHY (12, 15, 30). Acute raises in Pa_CO2 seem to elicit constriction of the vascular capacitance system (sympathetically mediated), increasing the mean circulatory filling pressure and, consequently, the atrial filling pressures. Although this effect has not been consistently described in humans, it is interesting to notice that some trend in this direction could be observed in two recent clinical studies (15, 18). Despite all the limitations of the post-hoc multiple regression analysis performed in our study, the results also indicated some active role of the acute raise in Pa_CO2 on systemic filling pressures independently of any PEEP-mediated effect.

On the other hand, PEEP increments usually result in increased intra-atrial pressures (10, 11), although through a different and complex mechanism. According to recent experimental data, whenever PEEP increases the mean circulatory filling pressure (by increasing the effective systemic volemia), it also externally constrains the cardiac chambers and, accordingly, the transmural atrial pressures are maintained or even diminished (31). The net result is usually a maintained or decreased pre-load and venous return. One could say that whereas PHY has the potential for increasing the true filling pressure with an increased venous return (30), the use of high PEEP would have the potential for increasing intra-atrial pressures but not the true pre-load. As we did not measure juxtacardiac or esophageal pressures, it is difficult to make some definite statement about the effective pre-load in our patients. However, as the patients in LPA did not present any change in the systolic volume along the time, marked changes in the effective pre-load were very unlikely.

At this point we must consider some potential adverse effects of our LPA: Intracranial pressure. Besides the well-known vasodilator effects of PHY on the brain (12), the marked increment in atrial pressures observed in LPA might adversely affect patients with documented intracranial hypertension. In our personal experience, however, we have submitted head-trauma patients to LPA (after installing intracranial pressure monitoring) and found that the strategy can be well tolerated in many cases.

Right ventricle. Despite marked pulmonary hypertension, we did not observe any clinical signal of right ventricle dysfunction or arrhythmia in our study. The systolic volume was preserved during LPA (data not shown) indicating that those patients made use of their cardiac reserve, keeping the right stroke volume despite an increased afterload. These findings are in accordance with recent evidences suggesting that right ventricular failure is relatively rare in ARDS (32). The best conduct in the presence of a documented cardiac or coronary disease, however, is an open question. The increased heart rate and right ventricular overload elicited by LPA might have harmful consequences to an ischemic myocardium.

Pulmonary edema. It is reasonable to suppose that, not only the increments in pulmonary arterial pressure and PWEDGE, but also the pulmonary overflow elicited by LPA might alter the longitudinal distribution of pulmonary resistance, increasing the effective pressures at the filtration site in the lung (33) and worsening lung edema. Although we have already demonstrated improved lung function associated with LPA (1), the situation could be even better if the intravascular pressures were minimized (34). In this context, the associated use of nitric oxide (15, 34) or alkalinizing agents (25) to attenuate pulmonary hypertension might be good theoretical alternatives.

Finally, notwithstanding the major hemodynamic alterations observed in the LPA patients, no organ dysfunction appeared during the 7-d period that could be related to this strategy. Moreover, we observed a significant decrease in the levels of lactate in the first 36 h of the protocol (Figure 5). As long as differences in global oxygen debt cannot account for these findings (O₂I was the same in both groups), there are three possible explanations for this decrease: one is the blockade of the glycolytic pathway caused by respiratory acidosis (35), the other is the local increase of oxygen availability in small territories provided by higher Pv_O2 (18) and/or splanchnic vasodilatation associated to high CO₂ tensions, and finally, a fairly attractive hypothesis is the reduction of systemic effects or cytokine release triggered by persisting lung injury imposed by our conventional ventilatory support (36). Whether the observed reduction is due to one or the other of the above is beyond the scope of the present study.

In conclusion, the data above are demonstrating that the cardiovascular tolerance to the immediate installation of PHYfrom a Pa_CO2 close to 35 mm Hg up to 60-80 mm Hg plus high PEEP levels (PEEP_IDEAL) was very adequate in our population (mean age = 34 yr). So, except for situations of previous heart and/or neurological disease, we believe that the installation of PHY in a progressive and gradual manner is not justifiable and may consist in a risky waste of time.