Fluid management in critically ill patients: the role of extravascular lung water, abdominal hypertension, capillary leak, and fluid balance

Introduction Capillary leak in critically ill patients leads to interstitial edema. Fluid overload is independently associated with poor prognosis. Bedside measurement of intra-abdominal pressure (IAP), extravascular lung water index (EVLWI), fluid balance, and capillary leak index (CLI) may provide a valuable prognostic tool in mechanically ventilated patients. Methods We performed an observational study of 123 mechanically ventilated patients with extended hemodynamic monitoring, analyzing process-of-care variables for the first week of ICU admission. The primary outcome parameter was 28-day mortality. ΔmaxEVLWI indicated the maximum difference between EVLWI measurements during ICU stay. Patients with a ΔmaxEVLWI <−2 mL/kg were called 'responders'. CLI was defined as C-reactive protein (milligrams per deciliter) over albumin (grams per liter) ratio and conservative late fluid management (CLFM) as even-to-negative fluid balance on at least two consecutive days. Results CLI had a biphasic course. ΔmaxEVLWI was lower if CLFM was achieved and in survivors (−2.4 ± 4.8 vs 1.0 ± 5.5 mL/kg, p = 0.001; −3.3 ± 3.8 vs 2.5 ± 5.3 mL/kg, p = 0.001, respectively). No CLFM achievement was associated with increased CLI and IAPmean on day 3 and higher risk to be nonresponder (odds ratio (OR) 2.76, p = 0.046; OR 1.28, p = 0.011; OR 5.52, p = 0.001, respectively). Responders had more ventilator-free days during the first week (2.5 ± 2.3 vs 1.5 ± 2.3, p = 0.023). Not achieving CLFM and being nonresponder were strong independent predictors of mortality (OR 9.34, p = 0.001 and OR 7.14, p = 0.001, respectively). Conclusion There seems to be an important correlation between CLI, EVLWI kinetics, IAP, and fluid balance in mechanically ventilated patients, associated with organ dysfunction and poor prognosis. In this context, we introduce the global increased permeability syndrome.

As the lungs are maximally exposed to the proinflammatory cascade, receiving the entire cardiac output, they provide valuable insight into dynamic microcirculatory changes during systemic inflammation [17]. Consequently, bedside measurement of extravascular lung water index (EVLWI) performed by single transpulmonary thermodilution allows the estimation of the extent of capillary leak and fluid overload [11,[18][19][20][21][22][23]. In this study, we investigated the prognostic value of EVLWI, capillary leak parameters, IAH, and fluid balance in critically ill patients.

Patients
We collected data from March 2004 to August 2007 in 123 patients treated in two ICU's in Ziekenhuis Netwerk Antwerpen (ZNA) Campus ZNA Stuivenberg, Antwerp, Belgium. Critically ill patients requiring mechanical ventilation (MV) and, according to clinical appraisal, extended hemodynamic monitoring by single transpulmonary thermodilution technique were consecutively included. Internal review board approval was obtained, and due to the non-interventional and retrospective nature of the study, the need for informed consent was waived (EC approval number 3765).

Definitions
Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) were diagnosed according to international criteria [24].
EVLWI was recorded as the mean of two daily EVLWI measurements. EVLWI min,max,mean were the minimal, maximal, and mean EVLWI during ICU stay, respectively. Maximum EVLWI was measured on Day max . Δ max EVLWI indicated the maximum difference between all EVLWI measurements during ICU stay and was computed in accordance with overall EVLWI trend (ΔEVLWI or the difference between the first and the last recorded EVLWI). If during ICU stay an increase of EVLWI was recorded followed by an equal EVLWI drop, Δ max EVLWI was given the sign of ΔEVLWI. Patients with an EVLWI decrease of >2 mL/kg (Δ max-EVLWI <−2 mL/kg) and an overall drop in EVLWI during the first week of ICU admission (negative ΔEVLWI) were called 'responders'.
Intra-abdominal pressure (IAP) was the mean of two daily IAP measurements. IAP max,mean were the maximum and the mean IAP during ICU stay. IAH was defined as IAP mean ≥ 12 mmHg and abdominal perfusion pressure (APP) as mean arterial pressure (MAP) minus IAP according to consensus definitions [15].
Daily fluid balance was calculated by subtracting the urinary output from the fluid intake (including both IV and enteral fluid administration); each day cumulative fluid balance was computed by the addition of daily fluid balances.
Conservative late fluid management (CLFM) was determined as even-to-negative fluid balance on at least two consecutive days during the first week of ICU stay [12]. In this study, CLFM was used as a descriptive term and did not signify any study intervention.

Data collection and methods
For the entire duration of the ICU stay, relevant demographic, clinical, and laboratory data along with daily assessment of fluid balance, sequential organ failure assessment (SOFA) score [26], IAP, MV settings, and extended hemodynamic monitoring variables were registered in an electronic database, supplemented by mortality on day 28. Severity of illness on ICU admission was described by an averaged simplified acute physiology score [27] and acute physiology and chronic health evaluation score [28].
IAP was measured via a Foley bladder catheter as described previously [29], in the complete supine position and in stable conditions twice daily. In patients with IAH, the IAP was also continuously monitored via a balloon-tipped catheter placed in the stomach connected to the CiMON monitor (Pulsion Medical Systems, Munich, Germany).
A central venous catheter and a thermistor-tipped arterial thermodilution catheter (Pulsiocath 5F) inserted into the femoral artery and attached to a PiCCOplus ® system (Pulsion Medical Systems, Munich, Germany) were already in place for each patient. Transpulmonary thermodilution measurements were obtained by central venous injection of three 20-mL boluses of cooled saline (<8°C). For each set of thermodilution determinations, the mean values were used for statistical analysis. Cardiac output (CO), global end diastolic volume (GEDV), extravascular lung water (EVLW), global ejection fraction (GEF), pulmonary vascular permeability index (PVPI), stroke volume variation, and pulse pressure variation were calculated using the PiCCOplus ® [18]. EVLW was indexed to body weight (EVLWI) and CO and GEDV to body surface area (cardiac index, GEDV index).

Study design
In this observational study, no protocol-directed intervention was performed; treatment was based on recent ICU guidelines. We analyzed process-of-care variables for the first 7 days of ICU admission. The primary outcome parameter was 28-day mortality. Secondary outcome parameters were organ dysfunction, duration of MV, and achievement of CLFM.

Statistical analysis
The primary data analysis compared survivors to nonsurvivors according to 28-day mortality. Subsequently, patients were stratified by occurrence of IAH, achievement of CLFM, and responders vs nonresponders. Continuous data were expressed by mean ± SD, and intergroup differences were determined by one-way analysis of variance (ANOVA) analyses day by day for 1 week. Categorical data were expressed as frequency distributions and/or percentages, and the c 2 test was used to determine intergroup differences. Two-sided p values <0.05 were considered to indicate statistical significance.
Time course of CLI, total SOFA score, EVLWI, APP, daily, and cumulative fluid balance was described by clustered error bar graphs representing mean ± SE. Receiver-operating characteristic (ROC) curves were determined and optimal cutoffs for CLI, EVLWI, and Δ max EVLWI were derived, creating categorical data.
Stepwise multivariate logistic regression was performed to determine the independent risk factors for 28-day mortality and for not achieving CLFM. Risk factors significant at the 0.1 level in univariate analysis were included in the models. The Hosmer-Lemeshow test was used assessing the goodness of fit.
The Kaplan-Meier method was used to analyze differences in cumulative survival and duration of MV; distribution was compared using the log-rank test. We used SPSS software package (version 17.0.1; SPSS, Chicago, IL, USA) for data analysis.

Patients
We included 123 predominantly medical (n = 109) patients on MV, of whom 65 (53%) died after 28 days. At baseline, no significant differences were found between groups, as shown in Table 1, except for lower MAP and GEF in nonsurvivors.

CLI
CLI had a biphasic course with a maximum on day 3, which was significantly higher in patients not achieving CLFM (76.1 ± 49.6 vs 53.2 ± 45.6, p = 0.017). ROC statistics for CLI on day 3 to predict no CLFM achievement revealed an area under the curve (AUC) of 0.658 and a derived cutoff point of >61 (sensitivity 62%, specificity 68%, and positive predictive value (PPV) 80%).

Total SOFA score
Total SOFA score remained significantly lower on each day from day 2 in survivors, patients achieving CLFM, and responders (p < 0.001).

Clinical outcomes
Outcomes concerning organ function were described by the course of total SOFA score as above. Other major outcomes are shown in Table 4 and Kaplan-Meier plots are shown in Figure 4.
Mortality and duration of MV were lower in patients achieving CLFM and in responders. Responders had fewer days with cardiovascular, respiratory, liver, and coagulation failure during the first week of ICU admission.
Multivariate analysis identified that increasing IAP mean and CLI on day 3 and being a nonresponder were independent risk factors for not achieving CLFM (p = 0.919 Hosmer-Lemeshow test) ( Table 5). Increasing baseline creatinine and EVLWI max , decreasing APP on day 3, not achieving CLFM, and being a nonresponder were independent risk factors for 28-day mortality (p = 0.808 Hosmer-Lemeshow test) ( Table 6).

Discussion
Our study demonstrated that a persistent increase in CLI, EVLWI, and fluid balance in critically ill patients is associated with poor outcome. We investigated the Cordemans et al. Annals of Intensive Care 2012, 2(Suppl 1):S1 http://www.annalsofintensivecare.com/content/2/S1/S1 precise prognostic value of these parameters and were able to formulate a unifying hypothesis implementing concepts of earlier studies ( Figure 5). As early as 1942, Cuthbertson introduced the concept of a dual metabolic response to bodily injury [30]. In direct response to initial proinflammatory cytokines and stress hormones, the ebb phase represents a distributive shock characterized by arterial vasodilatation and transcapillary albumin leak [31] abating plasma oncotic pressure. Arterial underfilling, microcirculatory dysfunction, and secondary interstitial edema lead to systemic hypoperfusion and impaired regional tissue oxygenation [2]. In this early stage of shock, adequate fluid therapy comprises of goal-directed filling [3] to prevent evolution to multiple organ dysfunction syndrome (MODS). As compensatory neuroendocrine reflexes and potential renal dysfunction result in sodium and water retention [32], positive fluid balances are inherent in the ebb phase. Patients with higher severity of illness need more fluids to achieve cardiovascular optimization. Therefore, at this point, fluid balance may be considered a biomarker of critical illness [33].
Patients overcoming shock attain homeostasis inflammatory mediators within 3 days [1]. Subsequent hemodynamic stabilization and restoration of plasma oncotic pressure set off the flow phase with resumption of diuresis and mobilization of extravascular fluid resulting in negative fluid balances. In line with Murphy et al. [12], we found CLFM achievement to be a strong and independent predictor of survival. In contrast, patients with persistent systemic inflammation maintain capillary leak and do not reach the flow phase, accumulating further positive fluid balances. In this context, we introduce the global increased permeability syndrome (GIPS), characterized by nonresponders with increased CLI, no CLFM achievement, and progressing organ failure. GIPS represents a 'third hit' of shock following acute injury and MODS.
We defined CLI as a parameter of capillary leak, assuming that increased vascular permeability caused by systemic inflammation is associated with high CRP levels [34] and hypoalbuminemia [31]. CLI had a biphasic course and the maximum reached on the third day of shock was an independent predictor of CLFM achievement. Previously, a negative cumulative balance [8,12,13,35,36] and lower PVPI [22] on day 3 were correlated with better survival. The third day of shock seems to be a crucial turning point [37] at which homeostasis of cytokines is accompanied by the healing of microcirculatory disruptions and 'closure' of the capillary leak. This interpretation is supported by Boerma et al. who demonstrated normalization of the microcirculatory blood flow on day 3 in septic patients [38].
As a result of capillary leak and an impaired flow phase, overzealous administration of fluids in GIPS will lead to gross fluid overload and tissue edema [14]. Interstitial edema raises the pressure in all four major body compartments: head, chest, abdomen, and extremities. Consequently, venous resistance of organs within compartments increases and perfusion pressure decreases contributing to progression of organ failure. As different compartments interact and reciprocally transmit compartment pressures, the concept of polycompartment syndrome is suggested [39].
The abdomen plays a central role in GIPS and polycompartment syndrome. Positive cumulative fluid balance is a known risk factor for secondary IAH [40] which in turn is associated with renal dysfunction [41]. Therefore, fluid overload leading to IAH and renal dysfunction may counteract its own resolution. Data from our study support these ideas, demonstrating higher average positive cumulative fluid balance and renal SOFA score after 1 week in patients developing IAH. Moreover, we determined increased IAP mean as an independent risk factor for no CLFM achievement and decreased APP as risk factor for 28-day mortality.
As the adverse effects of fluid overload in states of capillary leak are particularly pronounced in the lungs [17], monitoring EVLWI may offer a valuable tool to guide fluid management in the critically ill. In line with previous reports, we established a correlation between EVLWI max during admission and poor outcome [42]. An increased EVLWI max may indicate a state of capillary leak, associated with a higher severity of illness and mortality [11,22,23,42]. In this context, data from Sturm et al. are particularly of interest, correlating EVLWI with albumin extravasation in patients after multiple trauma [43]. BMI, body mass index; MV, mechanical ventilation; ARDS, acute respiratory distress syndrome; COPD, chronic obstructive pulmonary disease; SAPS, simplified acute physiology score; APACHE, acute physiology and chronic health evaluation; SOFA, sequential organ failure assessment; HR, heart rate; CI, cardiac index; SVV, stroke volume variation; GEF, global ejection fraction; GEDVI, global end-diastolic volume index; EVLWI, extravascular lung water index; PVPI, pulmonary vascular permeability index; PBW, predicted body weight; PEEP, positive end-expiratory pressure; CRP, C-reactive protein.
The course of EVLWI during the first week of admission may even be a better outcome predictor. Responders, defined as patients with an EVLWI decrease of >2 mL/kg, were more likely to achieve CLFM, had more organ-failure-free and ventilator-free days, and a better 28-day outcome. These data suggest that responders overcome the distributive shock and make a transition to the flow phase. Nonresponders on the other hand stay in the grip of the ebb phase and progress to GIPS associated with interstitial fluid accumulation, organ failure, and death. In this hypothesis, (the change in) EVLWI has a prognostic value as a reflection of the extent of capillary leak rather than as a quantification of lung function impairment. Indeed, the degree of hypoxemia in ARDS is an inferior prognostic factor, as extrapulmonary organ failure mostly determines outcome [44]. Accordingly, in a subgroup analysis of patients with ARDS, Sakka et al. found no higher maximum EVLWI in nonsurvivors [42]. Therefore, in an established state of capillary leak, time-dependent changes in EVLWI appear to be of superior value.
Our observations may have direct consequences on fluid management in the critically ill. Patients at risk for GIPS require restrictive fluid strategies and even fluid removal to avoid interstitial edema formation.
Our study has several important limitations. First, the observational nature of this study does not allow discrimination between a primary and secondary effect of fluid balance on outcome; prospective trials are warranted to determine if fluid overload is cause or      Cordemans et al. Annals of Intensive Care 2012, 2(Suppl 1):S1 http://www.annalsofintensivecare.com/content/2/S1/S1   Cordemans et al. Annals of Intensive Care 2012, 2(Suppl 1):S1 http://www.annalsofintensivecare.com/content/2/S1/S1 consequence of worse outcome. Second, inclusion of patients was based on clinical appraisal of the need of MV and thermodilution catheter monitoring. Therefore, the studied population was a specific case mix of seriously ill patients selected without well-defined objective rules making simple extrapolation of our results to a general ICU population impossible. However, albeit in this particular population, our observations contributed to some basic ideas regarding fluid management in patients with capillary leak as proposed in earlier reports [1,14,16,37,40] and raised questions that should be addressed in future prospective investigations. Third, our database did not supply detailed information on the amounts of fluids administrated specified for the first 6 h. Early fluid resuscitation has an important impact on outcome [12,45]. There were no data on the type of fluids and infusion rates used during ICU stay either. Fourth, differences in MAP and GEF at baseline may be important confounding factors as they may reflect different hemodynamic states dictating whether a patient can mobilize fluids in the flow phase.

Conclusions
We identified a subgroup of mechanically ventilated patients with persistent capillary leak failing to reach the flow phase. In these patients, GIPS may reflect a 'third hit' and superfluous fluid administration may be considered toxic. Future prospective clinical trials evaluating any therapy aimed at a reduction of EVLWI are warranted.