Pulse Pressure Variation Is Comparable With Central Venous Pressure to Guide Fluid Resuscitation in Experimental Hemorrhagic Shock With Endotoxemia

Abstract

Introduction: Pulse pressure variation (PPV) has been proposed as a promising resuscitation goal, but its ability to predict fluid responsiveness has been questioned in various conditions. The purpose of this study was to assess the performance of PPV in predicting fluid responsiveness in experimental hemorrhagic shock with endotoxemia, while comparing it with goals determined by a conventional set of guidelines. Methods: Twenty-seven pigs were submitted to acute hemorrhagic shock with intravenous infusion of endotoxin and randomized to three groups: (i) control; (ii) conventional treatment with crystalloids to achieve and maintain central venous pressure (CVP) 12 to 15 mmHg, mean arterial pressure of 65 mmHg or greater, and SvO₂ (mixed venous oxygen saturation) of 65% or greater; (iii) treatment to achieve and maintain PPV of 13% or less. Parametric data were analyzed by two-way analysis of variance and Tukey test and differences in crystalloid volumes by t test. Predictive values of variables regarding fluid responsiveness were evaluated by receiver operating characteristic curves and multiple logistic regression. Results: Both treatments produced satisfactory hemodynamic recovery, without statistical differences in fluid administration (P = 0.066), but conventional treatment induced higher CVP (P = 0.001). Areas under receiver operating characteristic curves were larger for CVP (0.77; 95% confidence interval, 0.68–0.86) and PPV (0.74; 95% confidence interval, 0.65–0.83), and these variables were further selected by multiple logistic regression as independent predictors of responsiveness. Optimal PPV cutoff was 15%, with false-positive results involving mean pulmonary arterial pressure of 27 mmHg or greater. Conclusions: Acute resuscitation guided by PPV was comparable with the strategy guided by CVP, mean arterial pressure, and SvO₂. Central venous pressure and PPV were individually limited but independently predictive of fluid responsiveness.

INTRODUCTION

Fluid therapy is considered first-line treatment for patients in hemorrhagic or septic shock, for the restoration of circulating volume and tissue perfusion, but several issues remain under debate, particularly regarding resuscitation goals represented by physiological variables to be achieved (^1–5).

The clinical decision for fluid administration is often based on cardiac filling pressures, but these static variables have been unable to predict fluid responsiveness (i.e., hemodynamic improvement from fluid administration) accurately (^6–9). Guidelines for the treatment of trauma and sepsis have been developed, which propose the achievement of particular goals within determined time frames, especially during the initial phase of resuscitation (^{2, 3, 10}). However, despite improvements in outcomes, morbidity and mortality remain elevated, so that not only do these strategies remain under debate, but also there is ongoing research involving diverse interventions and technologies to improve standard of care (^{3, 11}).

Pulse pressure variation (PPV) has been demonstrated as one of the most effective and accurate indices to predict fluid responsiveness in numerous clinical instances and thus has been proposed as a promising resuscitation goal (^{6, 7, 12–15}). As one of many dynamic variables, its use is based on the observation that intermittent positive-pressure ventilation induces cyclic changes in left ventricular stroke volume (LVSV) that are measurable in the peripheral circulation and that the intensity of such changes is related with preload dependence (^{6, 12, 14}). It is an index with the potential to locate each patient on their individual Frank-Starling curve dynamically, using minimally invasive and often already available arterial access, and, in specific multiparameter monitors, it is calculated automatically at the bedside and updated in real time, at each mechanical breath (^{7, 12, 14}).

Nevertheless, recent studies have shown that several pathophysiologic conditions that reflect on heart-lung interactions or on the impact of LVSV on the peripheral circulation may interfere with PPV and its predictive value for fluid responsiveness, so that further evaluation for its use in diverse clinical settings is warranted (^{9, 14, 16–23}).

The main purpose of this study was to compare initial fluid resuscitation guided by PPV with another strategy based on a set of goals derived from conventional guidelines, comprising central venous pressure (CVP), mean arterial pressure (MAP), and mixed venous oxygen saturation (SvO₂), in an experimental model of combined hemorrhagic shock with endotoxemia. The model was designed to mimic clinical hemorrhagic shock with contamination, requiring damage control surgery. The individual performance of PPV regarding prediction of fluid responsiveness was assessed in this setting and further compared with those of the other variables used as goals.

MATERIALS AND METHODS

Animal preparation

The study was approved previously by the institution’s ethics committee and performed in accordance with the National Institutes of Health guidelines. Twenty-seven large white and landrace cross-bred female pigs (28.4 ± 2.4 kg) were fasted overnight, with free access to water. They were premedicated with an intramuscular mixture of midazolam (0.25 mg/kg) and ketamine (5 mg/kg), and a marginal ear vein was catheterized for the administration of drugs and maintenance fluids (5 mL/kg per hour of lactated Ringer’s [LR], by ANNE anesthesia infuser; Abbott Laboratories, Chicago, Ill.). Anesthesia was induced with propofol (5 mg/kg intravenously [i.v.]), and, after orotracheal intubation, isoflurane (1.5%) was used for maintenance (Primus; Dräger Medical GmbH, Lübeck, Germany). Mechanical ventilation was performed by the anesthesia machine, with tidal volume (Vt) 8 mL/kg, positive end-expiratory pressure (PEEP) 5 cmH₂O, I:E ratio 1:2, FIO₂ 0.5, and respiratory rate to maintain end-tidal carbon dioxide of 40 to 45 mmHg. Pancuronium bromide (0.1 mg/kg bolus, followed by 5 μg/kg per minute i.v. by the infusion pump) was administered to ensure muscular relaxation. Pigs were placed in supine position, and, before surgical preparation, 2% lidocaine was injected in planned incision sites, and fentanyl (5 μg/kg i.v.) was administered. At the end of the experimental protocol, animals were killed with anesthetic overdose followed by central line injection of potassium chloride 19.1%.

A pulmonary artery catheter (Swan-Ganz CCO/SvO₂/CEDV 774HF75; Edwards Lifesciences, Irvine, Calif.) was inserted in the right jugular vein. Femoral arteries and veins were catheterized for hemodynamic monitoring, blood withdrawal, and fluid administration. A tonometric catheter (TRIP 16F; Tonometrics Division, Helsinki, Finland) was introduced in the midjejunum by laparotomy, and the incision closed. A transesophageal echocardiography (TEE) probe was positioned at the level of the midpapillary muscle for the short-axis, cross-sectional view of the left heart chambers (Philips Omniplano 7.5/5.0 MHz, EnVisor; Philips Ultrasound, Inc, Bothell, Wash). Urethral catheterization was performed for measurement of urinary output. Heart rate and rhythm, MAP, CVP, mean pulmonary artery pressure (mPAP), and pulmonary artery occlusion pressure (PAOP) were measured by a multiparameter monitor (Philips IntelliVue MP50; Philips Healthcare, Boedlingen, Germany). Pulse pressure variation (%) was defined as 100 × {(PPmax − PPmin) / [(PPmax + PPmin) / 2]} (²⁴), measured automatically and displayed in real time by a separate monitor (DX 2020, Dixtal, Manaus, Brazil), as described previously by Lopes et al. (¹²). After preparation, three consecutive recruitment maneuvers were performed (20 cmH₂O for 30 s), followed by 30 s of regular ventilation. Finally, a period of 30 min was allowed for stabilization.

Experimental protocol

All pigs were submitted to acute hemorrhagic shock of 50% the estimated blood volume (estimated as 76 mL/kg (²⁵)) in 20 min. The first 500 mL were stored in a conventional blood bag containing citrate. Simultaneously, i.v. infusion of lipopolysaccharide from Escherichia coli O111:B4 (Sigma-Aldrich Co, St Louis, Mo), diluted in 0.9% saline solution at 10 μg/mL, was started (Samtronic infusion pump 680, Socorro, Brazil), following a protocol adapted from Lipcsey et al. (²⁶), in which pigs received 4.0 μg/kg per hour for 80 min, 0.5 μg/kg per hour for 60 min, and 0.1 μg/kg per hour for the remainder of the experiment. After hemorrhage, shock was maintained for 60 min during which, if MAP decreased below 30 mmHg, autologous blood was administered to achieve MAP 40 mmHg.

Pigs were randomly assigned to three groups of nine animals. In the control group (CTL), they were monitored until the end of protocol. In the conventional group (CNV) and pulse pressure variation group (dPP), all pigs were submitted initially to a standard resuscitation protocol consisting of three consecutive boluses of 500 mL LR, each administered in 10 min, with 5-min intervals. In the CNV group, additional 500 mL fluid boluses were administered, at the same rate and interval, to achieve and maintain CVP 12 to 15 mmHg, MAP of 65 mmHg or greater, and SvO₂ of 65% or greater. At every 1,000 mL of LR, hemodynamic improvement was assessed, defined by increase in CVP by 1 mmHg or greater or increase by 10% or greater in either MAP or SvO₂ (^{2, 27}). If there was a positive response and endpoints had not been met, additional fluid boluses were given as before. If hemodynamic improvement was not observed and endpoints had not been met, pigs were checked for need of transfusion and, if not required, norepinephrine (NE) was initiated (0.1–1.0 μg/kg per minute i.v.). In the dPP group, additional 500 mL fluid boluses were administered, at the described rate and interval, until PPV was 13% or less. If PPV was 13% or less but MAP remained below 65 mmHg, the need for transfusion was verified and, if not required, NE was initiated. The trigger for blood transfusion was defined as hematocrit of less than 15%, based on observations of another study (²⁸). Treatments lasted for 3 h.

Complete sets of data were recorded at baseline (TB), after 1 h of sustained hemorrhagic shock (TS; 80 min after TB), and at hourly intervals during the treatment period (T1-T3). These included ventilator and hemodynamic data from monitors and results from bolus thermodilution, which was performed in triplicate with 10-mL injections of 5% dextrose (Vigilance Continuous Cardiac Output/Oximetry Monitors; Baxter, Irvine, Calif). Measurements by TEE were performed in triplicate and averaged at each timepoint, and in addition, qualitative analyses of interventricular septal configuration and movement were made. Analyses of arterial and mixed venous blood gases were performed immediately at each timepoint (ABL 555; Radiometer Medical, Brønshøj, Denmark), and results inserted into the tonometry monitor (Tonocap; Datex-Engstrom Division, Helsinki, Finland), which displayed intramucosal pH (pHi), regional CO₂ partial pressure, and the regional-to-arterial CO₂ gradient (PrPa). Hematocrit was obtained by centrifugation (Centrimicro 211; Fanem, São Paulo, Brazil), and oxygenation variables were calculated with standard formulae.

In both treated groups, PPV, CVP, MAP, mPAP, PAOP, continuous SvO₂, and cardiac index (CI) by bolus thermodilution were recorded immediately before and 5 min after each fluid challenge. Responsiveness to fluid challenge was defined as increase in CI of 15% or greater.

Statistical analyses

Parametric data were evaluated by two-way repeated-measures analysis of variance followed by Tukey test, when appropriate. Normality was assessed by Kolmogorov-Smirnov test. In group CTL, missing values on account of mortality were filled with the last available record. Urinary output was evaluated by Kruskal-Wallis test followed by Tukey test. Survival was assessed by Kaplan-Meier method and curves compared by log-rank test.

Unpaired t test was used to compare amounts of LR to reach goals and total amounts infused. Differences in need for NE use were assessed by Fisher exact test. Unpaired t test was used to evaluate differences in amounts of NE considering only the pigs that received it, and Mann-Whitney rank sum test was used to compare total amounts of NE required in treated groups. The predictive value of each variable used as resuscitation goal was evaluated by Student t test and further assessed by receiver operating characteristic (ROC) curves, to verify and compare individual predictive performances, optimal cutoff points as defined by Youden index, and sensitivities and specificities of the optimal cutoff values versus cutoff values used experimentally (^{29, 30}). Finally, forward stepwise multiple logistic regression was used to identify independent variables capable of predicting nonresponsiveness to fluid challenge. Statistical programs were used for calculations (SigmaPlot v.11, Systat Software, Inc., San Jose, Calif.; SPSS v.18, SPSS Inc., Chicago, Ill.; Microsoft Excel 2010, Microsoft Corp., Redmond, Wash. P < 0.05 was considered statistically significant, and unless otherwise stated, results are represented as means ± SD.

RESULTS

The combination of hemorrhage and endotoxemia produced hemodynamic impairment, respiratory compromise, and global and splanchnic hypoperfusion (Tables 1 and 2). At TS, all groups presented proportional, significant decreases in MAP, CVP, SvO₂, CI, left ventricular end-diastolic area index (LVEDAI), arterial pH, base excess (BE), oxygen delivery index and pHi and significant increases in PPV, mPAP, pulmonary vascular resistance index (PVRI), lactate, oxygen extraction ratio, and PrPa (all, P < 0.001 vs. TB). Decreases in static compliance accompanied increases in peak inspiratory pressure (PIP) (all, P < 0.001 vs. TB). In group CTL, these changes persisted or worsened over time, and five pigs died before completing the protocol. All treated animals survived (Kaplan-Meier survival analysis: log rank, 9.70; P = 0.002 vs. CTL).

Table 1: Hemodynamic and ventilation variables assessed at established timepoints

Table 2: Arterial blood gases, oxygenation, and global and splanchnic perfusion obtained at timepoints

Over T1-T3, treated groups presented normalization of several hemodynamic and perfusion variables. At T3, however, group CNV still presented statistical differences from TB in arterial pH (P = 0.007), lactate, and BE (both, P = 0.004). Between treated groups, CVP was significantly higher in group CNV than in group dPP at T2 (P = 0.009) and T3 (P = 0.001), as well as to its own TB and T1 at T3 (both, P = 0.006). Also at T3, PIP in group CNV was significantly higher than that in groups CTL (P = 0.001) and dPP (P = 0.042).

In all groups, mPAP reached peak values at T1 and decreased at T3; however, remaining above TB (T3, P < 0.001 vs. TB, T1 and T2; P = 0.029 vs. TS) and mean PAOP remained below 15 mmHg. Transpulmonary gradients (TPGs, mPAP-PAOP) were similar in all groups, and, from TS to T3, mean values approximated or exceeded 15 mmHg. Pulmonary vascular resistance index decreased in treated groups, but remained greater than 3 Wood units. By TEE, there were no relevant alterations in ejection fraction and no signs of septal dislocation or dyskinesia.

To achieve endpoint, group CNV received 99.8 ± 29.8 mL/kg LR, and group dPP, 89.1 ± 31.6 mL/kg (P = 0.471). Total volumes administered to achieve and maintain endpoints over the 3-h treatment period were 138.3 ± 18.6 and 119.0 ± 22.6 mL/kg, respectively (P = 0.066). Norepinephrine was required in two pigs from CNV and four from dPP, without statistically significant differences in frequency (P = 0.620) or in doses contemplating only pigs receiving NE (CNV, 33 ± 23; dPP, 10± 12 μg/kg; P = 0.166) or all pigs within treated groups (median [interquartile range], CNV, 0.0 [0.0–4.2]; dPP, 0.0 [0.0–3.8] μg/kg; P = 0.563). No transfusions were needed. Urinary output (median) was 8.7 mL/kg (interquartile range, 4.7–12.3 mL/kg) in group CNV and 7.2 mL/kg (interquartile range, 5.8–14.1 mL/kg) in group dPP, which were significantly greater than group CTL (1.1 mL/kg [interquartile range, 0.70–1.5 mL/kg]).

During treatments, 118 fluid challenges were registered. All variables used as goals were univariately significant for fluid responsiveness (CVP and PPV, P < 0.0001; MAP, P = 0.005; SvO₂, P = 0.021). Areas under the ROC curves (AUCs) were larger for CVP and PPV, which approximated 0.75 (Fig. 1). Stepwise multiple logistic regression retained CVP (odds ratio, 1.70; 95% confidence interval, 1.25–2.32; P = 0.001) and PPV (odds ratio, 0.91; 95% confidence interval, 0.84–0.98; P = 0.010) in the model, as independent predictors of nonresponsiveness to fluid challenge (i.e., lack of increase in CI by at least 15% in response to fluid bolus). Their interaction was represented in Figure 2, based on the model’s equation: ln[p / (1 − p)] = −3.583 + (0.529 × CVP) + (−0.098 × PPV), where ln is natural logarithm, and p, probability of nonresponsiveness to fluid challenge.

Fig. 1: Receiver operating characteristic curves of variables used as resuscitation goals. Area under the ROC curve for each variable used as goal (95% confidence interval) regarding prediction of fluid responsiveness: PPV, 0.74 (0.65–0.83), P < 0.0001; CVP, 0.77 (0.68–0.86), P < 0.0001; MAP, 0.65 (0.55–0.75), P = 0.006; SvO₂, 0.59 (0.49–0.69), P = 0.103. Pulse pressure variation (solid line), CVP (dashed line), MAP (dotted line); SvO₂ (dash-dot line).

Fig. 2: Logistic curves representing prediction of fluid responsiveness based on simultaneous use of CVP and PPV. Based on data obtained at each fluid challenge, stepwise multiple logistic regression retained CVP and PPV in the model as independent predictors of nonresponsiveness to fluid challenge (i.e., fluid bolus not resulting in increase in CI by at least 15%). Ultimately the model proposes that, among all of the variables used as goals in the present conditions, fluid responsiveness would be better predicted by the simultaneous assessment of CVP and PPV. This graph, derived from the equation presented in the text, illustrates the concurrent use of both variables before the fluid challenge. Pulse pressure variation is plotted on the x axis, and the probability of the event (i.e., nonresponsiveness to fluid challenge) is indicated in the y axis. Each curve represents a different value of CVP: dotted line, 6 mmHg; short-dashed line, 7 mmHg; long-dashed line, 8 mmHg; light solid line, 9 mmHg; medium solid line, 10 mmHg; dark solid line, 11 mmHg. Additional curves may be added by solving the equation. Within this model, while PPV 20% associated with CVP 10 mmHg would yield an estimated 45% chance of nonresponsiveness (i.e., 55% chance of responsiveness), the association of the same value of PPV with CVP 8 mmHg would yield an estimated 20% probability of nonresponsiveness (i.e., 80% probability of responsiveness). Further testing of this model is required.

Differences between optimal cutoff values and those used in the study were larger for PPV and CVP (Table 3). The change in the cutoff value of PPV from 13% to 15% improved its predictive performance, and false-positive classification rate was reduced by 44% (Table 4). Analyses considering the individual performance of optimal PPV in predicting fluid responsiveness were as follows: all false-positive results involved mPAP of 27 mmHg or greater, however irregularly, because all four response types were observed above this threshold. Higher mPAP 40 to 59 mmHg was detected in eight pigs, three from CNV, and three from dPP. Of the 22 fluid challenges performed under these conditions, four yielded false-positive results, and four false-negative results. Similarly, of the 81 challenges in which PAOP was acquired, all false-positive results involved TPG of 14 mmHg or greater. Two animals in group dPP and one in CNV presented transient increases in PAOP 16 to 20 mmHg, but none of the seven fluid challenges performed in such conditions were false-positive. Central venous pressure greater than PAOP was not detected. In general, false-positive results were more common during earlier fluid challenges, and false-negatives were obtained throughout, although approximately 30% of these responses occurred at the third fluid challenge (i.e., immediately preceding T1). Assuming no other hemodynamic changes, if in group dPP cutoff for PPV had been 15% and in group CNV responsiveness to fluid challenge had been evaluated at each 500 mL and required compulsory increase in CVP by 1 mmHg or greater, then 15 fluid challenges would have been considered in excess in either group (means, 29.4 and 29.1 mL/kg, respectively).

Table 3: Sensitivities and specificities of variables used as goals according to optimal and used cutoff values

Table 4: Proportion of correct and incorrect response classifications by PPV according to cutoff values (decimal)

DISCUSSION

In the present experiment, PPV-guided resuscitation was comparable with the strategy guided by CVP, MAP, and SvO₂. In both treated groups, hemodynamic recovery was satisfactory, without actual statistical differences in fluid balance or in NE use, but with resulting statistically higher CVP and PIP in group CNV. Regarding prediction of fluid responsiveness, PPV and CVP were superior to other variables used as goals and, even though their individual performances were limited, multiple logistic regression selected them as independently predictive of nonresponsiveness, indicating that their combined use may increase their predictive power. All of these findings must be interpreted in light of the experimental model.

The concurrent induction of hemorrhage and endotoxemia sought to induce a state of severe shock, conceivably simulating a condition consistent with penetrating trauma, associated with acute hemorrhagic shock and contamination, requiring surgical intervention for damage control (^{10, 31}). Indeed, the present model produced a state of severe hemodynamic shock, with a definite decrease in cardiac preload, accompanied by metabolic and respiratory dysfunctions and ultimately a high mortality rate in the absence of treatment. The induction of mPAP of 25 mmHg or greater, PAOP of 15 mmHg or lower, and normal to reduced CI were consistent with precapillary pulmonary hypertension (^{32, 33}), and deteriorating lung mechanics were likely associated with lung inflammation (³⁴).

The resulting AUC for PPV was lower than reported in some studies (^{6, 7}), but similar to others (^{19, 20, 22}), its value bordering the acceptable limit of a biomarker (^{20, 29}). The cutoff value chosen for PPV (13%) was based on previous findings (⁶), but the adjustment of PPV to the ideal cutoff value of 15%, in a posterior analysis, improved specificity and resulted in considerably lower incidence of false-positive results.

Incidence of elevated PPV not related with fluid responsiveness (i.e., false-positive) has been verified in previous reports, having been attributed to right ventricular failure (RVF) or increased RV afterload and, possibly, with the added effects of mechanical ventilation (^{9, 16–19, 35}). In such conditions, if the RV is incapable of delivering flow, LVSV will not increase even if it is preload-dependent (^{9, 36}). Clinically, a normal RV may be incapable of overcoming acute increases in pressure greater than 40 to 60 mmHg (³³). The hearts of the pigs in the present study were originally healthy but, even though mean mPAP was mostly below this interval, false-positive results were detected as of mPAP of 27 mmHg and TPG of 14 mmHg. Of note, persistent increases in TPG above 15 mmHg and/or in PVR above 2.5 Wood units, as detected, have been associated with increased risk of RVF in heart transplant patients (³²). Nevertheless, the occurrence of acute cor pulmonale was not verified by TEE, although such evaluation was valid mainly upon LVEDAI recovery (³⁷). Also, identification of RVF has been proposed by the association of mPAP greater than 25 mmHg, CVP greater than PAOP, and stroke volume index less than 30 mL/m² (³⁸). In the present experiment, mean stroke volume index below 30 mL/m² was not observed beyond T1 in treated groups, and because at no point CVP greater than PAOP was noted, RVF was not supported by these standards. Still, episodic RV inability to deliver volume across the resistance offered by elevated mPAP, PVRI, and concurrent mechanical ventilation should not be rejected, particularly during severe hypovolemia. Occurrence of “deltaPP-up effect” (^{9, 39}) could be considered, but TEE did not reveal left ventricular insufficiency, and transient increases in PAOP were not associated with false-positive results.

A recent study also involving endotoxemia and hemorrhage had a different design and protocol (¹⁹). In this study, fluid challenges with a colloid (hydroxyethyl starch) were performed in five pigs at two moments: after endotoxemia and again after hemorrhage of 20% the estimated blood volume and reinfusion of all shed blood. Volume challenges were performed at 10% the estimated blood volume, for as long as cardiac output increased by more than 10%. Pulse pressure variation was not used as an endpoint, but was calculated offline, and responsiveness to fluid challenge during analysis was defined as increase in stroke volume of 10% or more. In this experiment, a higher mPAP (45 ± 6 mmHg) was achieved, and, after volume expansions, results comprised higher CVP, PAOP, and MAP values than those targeted in the present study. From the presented data, it is possible to infer that CVP greater than PAOP may have occurred, but as highlighted by the authors, echocardiography was not performed to evaluate cardiac status, and volume overload was not considered as a possibility. In this setting, AUC for PPV was 0.64 (95% confidence interval, 0.51–0.78; P = 0.045), and the best cutoff value was 12%. Pulse pressure variation was hence considered an inadequate predictor of fluid responsiveness during pulmonary hypertension of this magnitude. A similar conclusion was drawn in a clinical study by the same group, with postsurgical patients with RVF or pulmonary hypertension associated or not with sepsis (¹⁸).

A previous clinical study also found false-positive results in 12 of 35 ICU patients, with PPV greater than 12% (¹⁷). These patients composed a somewhat heterogeneous population of medical and surgical patients, all with either sepsis or systemic inflammatory response syndrome, in ongoing treatment for more than 24 h, some receiving vasopressor or inotropic support. Volume challenge consisted of 500 mL of gelofusine 4% over 30 min, and responsiveness was defined as increase in stroke volume greater than 15% by transthoracic echocardiography. Tissue Doppler echocardiography was further used to assess peak systolic velocity of tricuspid annular motion (Sta), a parameter or RV function. Mean pulmonary artery pressure was not reported. Between true- and false-positive results, there were no significant differences in Vt, PEEP, or baseline PPV. A greater plateau pressure was associated with false results, but values overlapped considerably with those of true results. No signs of cor pulmonale were observed, but false-positive results were associated with a more marked dilatation of the right ventricle (not significant), and both patients with acute respiratory distress syndrome were nonresponders, suggesting increased afterload in these patients. The authors found that an Sta of less than 0.15 m/s anticipated false-positive results and concluded that this parameter should be associated with PPV measurements before performing fluid challenges.

At optimal cutoff, the rate of false-negative results was proportional to false-positive results, but their widespread incidence implied they were due to an ongoing process. False-negative results have been reported in critical patients ventilated with Vt less than 8 mL/kg (^{8, 20}), and it has been argued that decreased lung compliance also diminishes transmission of transpulmonary pressure, thereby decreasing PPV, even if the patient is fluid-responsive (^{40, 41}). By these standards, decrease in lung compliance secondary to lung inflammation could be the cause of false-negative results, even if Vt 8 mL/kg and PEEP 5 cmH₂O were maintained. In addition, changes in vascular tone must also be considered (^{4, 42}). The observation that PPV may fail to reflect LVSV variations at increased vascular compliance (⁶) was verified recently (²²). In the present study, significant decrease in systemic vascular resistance index at T1 coincided with a higher incidence of false-negative results.

The use of CVP for the assessment of preload is limited (⁴). Although it has been considered the standard for the determination of RV pressure and volume, cardiac pressures suffer the interference of several factors, such as of changes in vascular tone, which have occurred, and of ventricular function, which was likely (^{4, 42}). In group dPP, CVP was rapidly restored to baseline levels and remained stable, whereas, in group CNV, CVP 12 to 15 mmHg was one of the endpoints, so that it became significantly higher than that in group dPP, as well as its own baseline and T1. Whereas the actual consequences of a greater CVP and PIP were not verified during the experiment, a clinical trial in patients with acute lung injury showed that fluid therapy targeting lower CVP or PAOP resulted in improved lung function and reduction in duration of mechanical ventilation and of intensive care, without increasing incidence of nonpulmonary organ failure (⁴³). In this regard, the slightly greater (but statistically significant) elevation in PIP in group CNV was likely related to the greater CVP in the same group. For the prediction of fluid responsiveness, CVP of 12 mmHg or greater presented very low sensitivity, whereas CVP 9 to 10 mmHg was associated with a greater rate of correct classifications, so that the proposed goal for CVP could be questioned, at least in these conditions. Furthermore, it could be considered that, in the event of acute cor pulmonale, CVP could increase by backpressure (³³), thereby producing the false assumption of fluid responsiveness if only CVP were used for this evaluation.

All of the variables used as goals were univariately predictive of fluid responsiveness. Areas under the receiver operating characteristic curves for PPV and CVP were larger and equivalent to each other, but, as stated for PPV, their values rendered them unsuitable for their individual use to this end. Also, optimal cutoff values were different from those used experimentally for CVP and PPV, but not for MAP and SvO₂. Of note, optimal cutoff values were determined by Youden index, seeking to maximize the correct classifications of fluid responsiveness; however, clinically, the choice of a cutoff value for a biomarker should encompass additional considerations, which include benefits and consequences of increasing or reducing the incidence of false-positive or false-negative results, among others (^{29, 30}).

Because there was no clear superiority of one treatment strategy over the other, and given that neither CVP nor PPV was considered an adequate predictor of fluid responsiveness per se, stepwise multiple logistic regression was used with the purpose of seeking whether one of the variables used as goals, or a distinct variable combination that had not been contemplated, could be more effective in predicting fluid responsiveness in the present conditions. The retention of CVP and PPV in the multiple logistic regression model, as independent variables associated with the prediction of nonresponsiveness to fluid challenge, indicated that their combined use could enhance their predictive power. Still, it should be noted that this analysis was based on data from the fluid challenges performed in the study, and, even though they were recorded prospectively at each bolus and about half of the fluid challenges were performed regardless of treatment strategies; additional challenges within each group involved either the algorithm which included CVP, or the use of PPV, and therefore potential overestimation of their predictive values should be considered, so that this model should be investigated further in new studies before it is accepted (⁴⁴). The combination of these parameters for the prediction of fluid responsiveness may be reasonable, given their different physiological bases and methods for measurements. In the present model, the simultaneous use of CVP and PPV could be achieved by solving the equation and was further represented by the logistic curves in Figure 2. One interesting aspect involving CVP and PPV monitoring is that invasive arterial pressure monitoring and placement of a central line are often necessary in the critical care setting (²).

Regarding study limitations, it should be noted that changes in PPV secondary to possible abdominal hypertension could not be determined, as abdominal pressure was not measured (^{21, 45}). In addition, a longer treatment period might have added to the results, perhaps even revealing a significant difference in total fluids administered between treated groups, but the study sought to evaluate the initial, acute phase of resuscitation. Finally, results must be interpreted in the context of their experimental nature (⁴⁵), and the inherent limitations for use of PPV must always be taken into account (^{8, 14, 24}).

This study reinforces the concepts that decreased preload is not equivalent to fluid responsiveness and that PPV is an index that should not be followed blindly (^{27, 35, 42}). While evaluating PPV, all possible factors that may interfere with cardiopulmonary interactions must be considered (^{8, 14, 18, 19, 21, 24}). In clinical practice, there should be full awareness of the mechanisms and limitations involving PPV, and more importantly, the patient’s needs should be pondered (^{27, 35}). Particularly in situations in which PPV may perform poorly, it should be combined or even replaced with other means for the decision about fluid administration (e.g., TEE), and of course, subsequent confirmation of responsiveness and patient re-evaluation is advised (^{7, 17, 22, 27, 40, 42}).

In summary, the results of initial fluid resuscitation guided by PPV were comparable with those obtained with the strategy guided by CVP, MAP, and SvO₂ in the present model. The individual performances of PPV and CVP for the prediction of fluid responsiveness were similar to each other and superior to those of MAP and SvO₂, but were considered limited. Still, results of the logistic regression indicated that the combined use of CVP and PPV may enhance their predictive power, which is a hypothesis to be investigated further. Pulse-pressure variation is proposed as an additional variable to aid in patient monitoring, but awareness of its limitations is indispensable.

ACKNOWLEDGMENTS

The authors thank Mr Gilberto de Mello Nascimento for valuable technical assistance.

ABBREVIATIONS

AUC: areas under the receiver operating characteristic curves

BE: base excess

CI: cardiac index

CNV: conventional group

CTL: control group

CVP: central venous pressure

dPP: pulse pressure variation group

LVEDAI: left ventricular end-diastolic area index

LVSV: left ventricular stroke volume

MAP: mean arterial pressure

mPAP: mean pulmonary artery pressure

NE: norepinephrine

PAOP: pulmonary artery occlusion pressure

pHi: intramucosal pH

PIP: peak inspiratory pressure

PPV: pulse pressure variation

PrPa: regional-to-arterial CO2 gradient

PVRI: pulmonary vascular resistance index

ROC: receiver operating characteristic

Svo2: mixed venous oxygen saturation

T1: after 1-h treatment

T2: after 2-h treatment

T3: after 3-h treatment

TB: baseline

TEE: transesophageal echocardiography

TPG: transpulmonary gradient

TS: 1 h after sustained hemorrhagic shock

Vt: tidal volume