Journal of Critical Care 29 (2014) 1011–1015

Contents lists available at ScienceDirect

Journal of Critical Care

j ourna l homepage: www. jcc journa l .org

Volume of fluids administered during resuscitation for severe sepsis

and septic shock and the development of the acute respiratorydistress syndrome☆

Dong W. Chang, MD a,⁎, Richard Huynh, MD a, Eric Sandoval, MD a, Neung Han, MD a,Clinton J. Coil, MD, MPH c, Brad J. Spellberg, MD b

a Divisions of Respiratory and Critical Care Physiology and Medicine, Los Angeles County Department of Health Services and Los Angeles Biomedical Research Institute at Harbor-Universityof California Los Angeles Medical Center, Torrance, CA, USAb General Internal Medicine, Los Angeles County Department of Health Services and Los Angeles Biomedical Research Institute at Harbor-University of California Los Angeles Medical Center,Torrance, CA, USAc Department of Medicine and the Department of Emergency Medicine, Los Angeles County Department of Health Services and Los Angeles Biomedical Research Institute at Harbor-University ofCalifornia Los Angeles Medical Center, Torrance, CA, USA

a b s t r a c ta r t i c l e i n f o

☆ Research support: none.⁎ Corresponding author. Department of Medicine, Har

405, 1000W. Carson St, Torrance, CA90509. Tel.:+1 3102E-mail address: dchang@labiomed.org (D.W. Chang)

http://dx.doi.org/10.1016/j.jcrc.2014.06.0050883-9441/© 2014 Elsevier Inc. All rights reserved.

Keywords:

SepsisAcute respiratory distress syndromeResuscitationFluid balance

Purpose: The purpose of this study was to examine the association between the volume of intravenous (IV)fluids administered in the resuscitative phase of severe sepsis and septic shock and the development of theacute respiratory distress syndrome (ARDS).Materials and methods: This was a retrospective cohort study of adult patients admitted with severe sepsis andseptic shock at a large academic public hospital. The relationship between the volume of IV fluids

administered and the development of ARDS was examined using multivariable logistic regression analysis.Results: Among 296 patients hospitalized for severe sepsis and septic shock, 75 (25.3%) developed ARDS. Aftercontrolling for confounding variables, there was no significant association between the volume of IV fluidsadministered in the first 24 hours of hospitalization and the development of ARDS (odds ratio [OR], 1.05; 95%confidence interval [CI], 0.95-1.18). Serum albumin (OR, 0.52; 95% CI, 0.31-0.87) and Acute Physiology andChronic Health Evaluation II score (OR, 1.08; 95% CI, 1.04-1.13) on admission were the most informativecovariates for the development of ARDS in the regression model.Conclusions: For patients hospitalized for severe sepsis and septic shock, fluid administration to improve end-organ perfusion should remain the top priority in early resuscitation despite the potential risk of inducing ARDS.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Severe sepsis and septic shock are the most severe manifestationsof the sepsis syndrome and characterized by end-organ hypoperfu-sion and hypotension due to infection [1]. Previous studies haveshown that early goal-directed resuscitation of patients with severesepsis and septic shock improves mortality [1,2]. One of the keyinterventions in early goal-directed therapy is aggressive administra-tion of intravenous (IV) fluids using physiologic targets to assess forimprovements in end-organ perfusion [2]. For patients with severesepsis or septic shock, the Surviving Sepsis Guidelines recommend aninitial bolus of 30 mL/kg of fluid followed by repeated fluidadministration as long as there are continued responses in hemody-namic parameters [1]. The rationale for these recommendations is

bor-UCLA Medical Center, Box22 3803; fax:+1310 328 9849..

that, in the resuscitative phase of severe sepsis and septic shock,restoring intravascular volume and maintaining end-organ perfusionare the top priorities. However, one concern regarding aggressivevolume resuscitation is that it may increase the risk for complicationsdue to volume overload, such as the acute respiratory distresssyndrome (ARDS) [3–5].

Acute respiratory distress syndrome is a devastating complicationof sepsis that affects its clinical management and outcomes [6–8].Previous studies have shown that interventions that minimize theadministration of IV fluids in hemodynamically stable patients withARDS decrease the duration of mechanical ventilation and intensivecare unit (ICU) stay without compromising end-organ perfusion [9].As such, the fluid management strategies for ARDS can becomediscordant with those of sepsis when pulmonary edema complicatesthe early resuscitative phase of the sepsis syndrome.

Given the increased propensity of the lungs to develop pulmonaryedema during sepsis, it is possible that the effects of positive fluidbalance during fluid resuscitation in patients admitted with severesepsis and septic shock may increase the risk for developing ARDS.

1012 D.W. Chang et al. / Journal of Critical Care 29 (2014) 1011–1015

The objective of this study was explore this risk by examining theassociation between the volume of IV fluids administered in the first24 hours of hospitalization for severe sepsis or septic shock and thedevelopment of ARDS. We hypothesized that increased IV fluidadministration is associated with a greater incidence of ARDS despitecontrolling for known predisposing factors.

2. Materials and methods

2.1. Study design and patients

This was a retrospective observational study of patients withsevere sepsis or septic shock admitted to the emergency departmentof a large academic county hospital (Harbor-UCLA Medical Center,Torrance, CA) between December 2011 and January 2013. The cohortof patients with severe sepsis and septic shock was identified byretrospective chart review of all patients seen in the emergencydepartment using the clinical definition from the Surviving SepsisGuidelines and the Society of Critical Care Medicine/European Societyof Intensive Care Medicine/American College of Chest PhysiciansInternational Sepsis Definitions Conference [1,10]. Specifically, sepsiswas defined as a probable and suspected infection as documented by aphysician in the medical record with at least 2 manifestations of asystemic inflammatory response (temperature N 38.3°C or b35.6°C,heart rate N 90 beats per minute, respiratory rate N 20/min, or whiteblood cell count N 12.0 × 103 or b4.0 × 103, or normal white blood cellcount with N 10% immature band forms). Severe sepsis was defined assepsis with organ dysfunction or tissue hypoperfusion (systolic bloodpressure b 90mmHg,mean arterial pressure b 70mmHg, lactate abovethe upper limit of normal, urine output b 30 mL/h despite initial fluidresuscitation, creatinine N 2.0 mg/dL, bilirubin N 2 mg/dL, plateletcount b 100,000, international normalized ratio N 1.5). Septic shockwasdefined as sepsis-induced hypotension resulting in a systolic bloodpressure less than 90mmHg ormean arterial pressure less than 70mmHg requiring vasopressor support despite IV fluid administration [1,10].

We examined the association between the total volume of IV fluidsadministered in the first 24 hours of hospitalization and thedevelopment of ARDS within 72 hours of hospital admission inpatients admitted with severe sepsis and septic shock. The totalvolume of IV fluids administered was determined from nursing flowsheets that documented the total fluid intake and output for eachpatient. The development of ARDS was identified by chart reviewusing the Berlin definition of ARDS [11,12]. Specifically, patients wereidentified as having ARDS if they had bilateral opacities on chestradiograph, acute onset of respiratory failure not fully explained bycardiac failure, and PaO2/FIO2 ratio less than 300 mm Hg within72 hours of admission for severe sepsis or septic shock.

2.2. Risk factors for the development of ARDS

The primary predictor variable for the development of ARDS wasthe volume of IV fluids (liters) administered in the first 24 hours ofhospitalization. Secondary variables included the administration of aninitial fluid bolus of greater than 20 mL/kg and the volume of IV fluidsgiven in the first 6 hours of hospitalization. Additional risk factors thatwere examined included patient demographics (age, sex, race/ethnicity), medical comorbidities (Charlson Comorbidity Index),presenting vital signs and laboratory data (serum sodium, bicarbon-ate, glomerular filtration rate, albumin, lactate), medications, use ofblood products, composite scores for severity of illness (AcutePhysiology and Chronic Health Evaluation [APACHE] II, SequentialOrgan Failure Assessment [SOFA]), and suspected source of sepsis.Additional clinical outcomes that were examined include hospital andICU mortality. Laboratory studies that were not normally distributedwere log transformed for statistical analysis. The APACHE II and SOFAscores were calculated from the clinical data available on admission.

All data were collected by retrospective chart review by 2 of theinvestigators (RH and NH) using a structured data abstraction form.Patient identifiers were removed from aggregated data and werecoded using a numeric identifier. The list of numeric codes andcorresponding patient identifiers was maintained in a password-protected computer by one of the investigators (DWC). The study wasapproved as an exempt protocol by the John G Wolf InstitutionalReview Board at the Los Angeles Biomedical Research Institute. Theneed for informed consent was waived.

2.3. Statistical analysis

The primary analysis compared the volume of IV fluids adminis-tered in the first 24 hours of hospitalization in patients who developedARDS vs those who did not. The Student t test was used to comparethe volume of fluids administered to each group at 6 and 24 hours. Thesecondary predictor variables, volume of fluids administered in thefirst 6 hours of hospitalization and the administration of an initial fluidbolus, were analyzed using a t test or χ2 test, as appropriate. A P valueb .05 was used to define statistical significance. To control for factorsthat may confound the relationship between the volume of fluidsadministered in the first 24 hours (primary predictor variable) andthe development of ARDS (primary outcome variable), multivariablelogistic regression analysis was performed [13,14]. Prior to multivar-iable analysis, univariate analysis was performed on all covariates toidentify those with the strongest association with the development ofARDS. All covariates that were significant at a level of P b .10 wereconsidered for multivariable analysis [15]. For multivariable regres-sion analysis, 2 models were generated. First, all covariates of interestwere included in a comprehensive model that contained any factorsthat showed an association in the univariate analysis. Next, the samecovariates were included in stepwise logistic regression analysis togenerate a parsimonious model that contained only the mostinformative covariates. These 2 models were used as complementaryapproaches to control for the influence of confounding factors on therelationship between primary and secondary predictor variables andthe development of ARDS. The volume of IV fluids administered in thefirst 24 hours (primary variable of interest) and the administration ofan initial fluid bolus (secondary variable) were included in all models.The administration of an initial fluid bolus was included in themultivariable models despite the lack of association in univariateanalysis because this variable was of clinical interest as a potential riskfactor for ARDS. The volume of IV fluids administered in the first6 hours of hospitalization (secondary variable) was not included inthe multivariable models because it was collinear with the volume ofIV fluids administered in the first 24 hours. We evaluated the modelsfor multicollinearity using a correlation coefficient matrix of thecovariates [15]. All covariates included in the models showed acorrelation of less than 0.60. The values of the variables in the modelsare reported as adjusted odds ratios (ORs) with 95% confidenceintervals (CIs). A P value b .05 was used for statistical significance inthe model variables. The data analysis was performed using JMPversion 11.0 (SAS Institute, Cary, NC).

3. Results

3.1. Baseline characteristics between the ARDS and control groups

The study cohort consisted of 296 patients who were admittedwith severe sepsis or septic shock. Of these, 75 patients (25.3%)developed ARDS within 72 hours of hospital admission. There weremultiple differences in the baseline characteristics between patientsthat developed ARDS and those that did not develop ARDS (Table 1).The ARDS groupwas older (mean age, 71.5 vs 62.9 years; P= .004). Inaddition, the ARDS group had a lower proportion of Hispanic patientscompared to the no-ARDS group (17.3% vs 38.9%, P b .001). The serum

Table 1Baseline characteristics

Characteristic ARDS (n= 75) No ARDS (n=221) P-value

Age (SD) 71.5 (18.7) 62.9 (18.2) .004Female (%) 35 (46.7) 109 (49.3) .691Race/ethnicity (%)

White 21 (28.0) 62 (28.1) .993African American 22 (29.3) 45 (20.4) .116Hispanic 13 (17.3) 86 (38.9) b .001Asian 16 (21.3) 27 (12.2) .061Other 3 (4.0) 1 (0.5) .287

Body mass index (SD) 25.6 (7.3) 26.4 (6.7) .375Serum albumin (SD) 2.4 (0.7) 2.7 (0.7) .001Lactate (SD) 3.8 (3.6) 2.5 (2.3) .002APACHE II (SD) 24.1 (9.0) 18.8 (8.1) b .001SOFA (SD) 9.5 (4.1) 5.1 (3.6) b .001Serum sodium (SD) 137 (7.7) 134 (7.0) .009Serum creatinine (SD) 2.3 (2.2) 2.0 (2.4) .356Source of sepsis (%)

Pneumonia 44 (58.7) 87 (39.4) .004GI tract 4 (5.3) 23 (10.4) .322GU tract 14 (18.7) 56 (25.3) .29Skin/soft tissue 2 (2.7) 20 (9.1) .035Other 11 (14.7) 35 (15.8) .347

Comorbidities (%)Congestive heart failure 14 (18.7) 55 (24.9) .263Diabetes mellitus 29 (38.7) 82 (37.1) .809Chronic kidney disease (GFR b 60) 16 (21.3) 38 (17.2) .429

Tobacco use (%) 15 (20.0) 37 (16.7) .526Medications (%)

Statin 16 (21.3) 42 (19.0) .663ACE inhibitor 11 (14.7) 51 (23.1) .111Aspirin 18 (24.0) 47 (21.3) .624β-Blockers 23 (30.7) 67 (30.3) .955

PRBC transfusion in 1st 24 h (yes, %) 13 (17.3) 40 (18.1) .881FFP transfusion in 1st 24 h (yes, %) 4 (5.3) 14 (6.3) .751ICU mortality (%) 19 (25.3) 15 (6.8) b .001Hospital mortality (%) 29 (38.7) 30 (13.5) b .001

Abbreviations: GI- gastrointestinal, GU- genitourinary, GFR- glomerular filtration rate,ACE- angiotensin converting enzyme, PRBC- packed red blood cells, FFP- fresh frozenplasma, ICU- intensive care unit.

Figure. The mean (±SE) volume of fluids administered during the first 6 and 24 hoursof hospitalization to patients who developed ARDS compared to those who did nodevelop ARDS.

Table 2Univariate analysis examining the association between demographic/clinical factorsand the development of ARDS

Variable OR 95% CI P-value

Volume of IV fluids in first 24 h, 1-L increments 1.11 (1.01-1.22) .029Initial IV fluid bolus N20 mL/kg 0.98 (0.49-1.88) .948Age 1.03 (1.01-1.04) .004Female sex 0.90 (0.53-1.52) .691Race/ethnicity

White 1.00African American 1.44 (0.71-2.95) .311Hispanic 0.45 (0.20-0.95) .036Asian 1.75 (0.79-3.87) .168Other 8.86 (1.07-184.44) .043

Temperature 0.92 (0.86-0.99) .032Heart rate 1.00 (0.99-1.01) .872Systolic blood pressure 1.00 (0.99-1.01) .938Body mass index 0.98 (0.94-1.02) .375Central venous pressure 1.02 (0.94-1.10) .675Serum sodium 1.05 (1.01-1.09) .009Serum bicarbonate 0.98 (0.94-1.02) .255Log glomerular filtration rate 0.71 (0.52-0.95) .021Serum albumin 0.49 (0.32-0.73) .001Log BNP 1.08 (0.68-1.75) .738White blood cell count 1.00 (0.98-1.02) .996Log lactate 1.83 (1.26-2.71) .002Use of statin 1.16 (0.59-2.17) .663Use of β-blocker 1.02 (0.57-1.78) .955APACHE II 1.08 (1.04-1.12) b .001History of congestive heart failure 0.69 (0.35-1.31) .263History of diabetes mellitus 1.07 (0.62-1.82) .809History of chronic renal disease 1.31 (0.67-2.48) .429Charlson Comorbidity Index 1.09 (0.99-1.20) .064Current tobacco use 1.24 (0.62-2.39) .526Source of sepsis

Pneumonia 1.00GI 0.34 (0.10-0.96) .041GU 0.49 (0.24-0.97) .039Soft tissue/skin 0.20 (0.03-0.72) .011Other 0.62 (0.28-1.31) .215

Transfusion of blood products in 1st 24 h 1.03 (0.52-1.96) .919

1013D.W. Chang et al. / Journal of Critical Care 29 (2014) 1011–1015

lactate and the APACHE II and SOFA scores were higher in the ARDSgroup, suggesting a higher severity of illness and end-organ injury(P b .001). Although pneumonia was the most common source of sepsisinbothgroups, itwasmore frequent in theARDSgroup (58.7%vs39.4,P=.004). The distribution of comorbidities between the ARDS and non-ARDSgroups is shown in Table 1. The frequency of blood transfusions betweenthe groups was similar. The ICU mortality (25.3% vs 6.8%, P b .001) andhospital mortality (38.7% vs 13.8%, P b .001) were higher in patients whodeveloped ARDS than in those that did not develop ARDS.

3.2. Volume of fluids administered in the first 24 hours of hospitalization

Patients who developed ARDS received more IV fluids than non-ARDS patients (Figure). There was a nonsignificant trend of increasedfluid administration after 6 hours of hospitalization (mean difference,280 mL; P = .07) in patients who developed ARDS. After 24 hours ofhospitalization, patients who developed ARDS received an average of830 mL more IV fluids compared to the control group (P b .05). Allpatients received crystalloid fluids during the initial volume resusci-tation. The urine output in the first 24 hours of hospitalization wassimilar between the groups (1.30 L in ARDS patients vs 1.37 L in non-ARDS patients, P = .68). To control for factors that can confound theassociation between IV fluid administration and the development ofARDS, logistic regression analysis was performed. By univariateanalysis age, temperature, serum sodium, glomerular filtration rate(log), serum albumin, serum lactate (log), APACHE II score, CharlsonComorbidity Index, and source of sepsis showed an association (P b

.10) with the presence of ARDS (Table 2). These variables wereincluded in a comprehensive multivariable logistic model (Table 3).This model showed that, after controlling for confounding variables,

t

the volume of fluid administered in the first 24 hours of hospitaliza-tion was not significantly associated with the presence of ARDS (OR,1.07; 95% CI, 0.95-1.21). The association between the volume of fluidadministered and the presence of ARDS was also examined using aparsimonious model that included only the most significantlyassociated covariates as identified by stepwise logistic regression(Table 4). This model contained serum albumin (OR, 0.52; 95% CI,0.31-0.87) and the APACHE II score (OR, 1.08; 95% CI, 1.04-1.13). Like

Table 3Multivariable logistic regression analysis showing the relationship between the volumeof IV fluids in the first 24 hours of hospitalization and the development of ARDS aftercontrolling for confounding covariates identified in univariate analysis

Variable AdjustedOR

95% CI P-value

Volume of IV fluids in first 24 h, 1-L increments 1.07 (0.95-1.21) .27Initial IV fluid bolus of N20 mL/kg 0.65 (0.28-1.45) .30Serum albumin 0.50 (0.28-0.84) .01APACHE II score, 1-point increments 1.09 (1.03-1.15) b .01Age 0.99 (0.97-1.02) .70Temperature 1.02 (0.93-1.02) .72Serum sodium 1.02 (0.97-1.06) .52Glomerular filtration rate (log) 0.94 (0.60-1.48) .78Lactate (log) 0.76 (0.45-1.26) .28Charlson Comorbidity Index, 1-point increments 1.00 (0.86-1.16) .99Source of sepsisPneumonia 1.00GI tract 0.71 (0.16-2.81) .64GU tract 0.80 (0.31-1.98) .63Skin/soft tissue 0.37 (0.05-1.93) .25Other 0.85 (0.32-2.17) .73

An adjusted OR of greater than 1 indicates that the variable is associated with greaterodds of developing ARDS.

1014 D.W. Chang et al. / Journal of Critical Care 29 (2014) 1011–1015

the comprehensive model, this parsimonious model also showed thatthe association between the volume of fluids given in the first 24hours of hospitalization and the presence of ARDS was not significant(OR, 1.05; 95% CI, 0.95-1.18). The secondary predictor variable,whether a fluid bolus of greater than 20 mL/kg was initiallyadministered, was also not significantly associated with the develop-ment of ARDS in both models (Tables 3 and 4).

4. Discussion

Fluid management in severe sepsis and septic shock is challenging,and the optimal strategy is unknown. Although there is generalconsensus that resuscitation with IV fluids should be a first-lineintervention in patients with severe sepsis and septic shock, recentstudies have shown that excessive fluid administration is associatedwithprolongedend-organdysfunction and increasedmortality [16–18].These studies support the concern among clinicians that aggressivevolume resuscitation in severe sepsis and septic shock, while initiallyimproving end organ perfusion, may inadvertently expose patients tocomplications from excessive fluid administration.

Previous studies have shown that the volume of fluids administeredduring a hospitalization increases the likelihood of developing ARDSamong patients at risk. [19–21] Jia et al [20] retrospectively examined amixed cohort ofmechanically ventilated patients in a large ICUdatabase(Multi Parameter IntelligentMonitoringof Intensive CareDatabase) andshowed that net fluid balance was significantly associated with thedevelopment of ARDS (OR, 1.30; 95% CI, 1.09-1.56). Plurad andcolleagues [21] found that, among patients admitted after trauma tosurgical ICUs, fluid balance of greater than 2 L in the first 48 hours ofhospitalization was independently associated with ARDS.

Table 4Stepwise multivariable logistic regression analysis showing the relationship betweenthe volume of IV fluids in the first 24 hours of hospitalization and the development ofARDS after controlling for the most informative confounding covariates

Variable Adjusted OR 95% CI P-value

Volume of IV fluids in first 24 h,1-L increments

1.05 (0.95-1.18) .35

Initial IV fluid bolus of N20 mL/kg 0.70 (0.31-1.52) .38Serum albumin 0.52 (0.31-0.87) .01APACHE II score, 1-point increments 1.08 (1.04-1.13) b .0001

An adjusted OR of greater than 1 indicates that the variable is associated with greaterodds of developing ARDS.

Although these studies show the overall association betweenpositive fluid balance and the development of ARDS, they do notaddress how aggressive volume resuscitation during the early phasesof sepsis affects this risk. Theoretically, early goal-directed volumeresuscitation can decrease the risk of ARDS by improving end-organperfusion, thereby reducing the physiologic derangements associatedwith the systemic inflammatory response syndrome. Murphy et al[18] examined a retrospective cohort of 212 patients hospitalizedwith ARDS secondary to septic shock and found that patients who hadvolume resuscitation according to the Surviving Sepsis Guidelines hadlower hospital mortality. In that study, the survivors had a greatervolume of fluids in the first 6 and 24 hours of hospitalization but hadsignificantly less total volume by day 7 and at hospital discharge.These results are consistent with the aggressive fluid managementstrategy in early sepsis followed by a conservative fluid managementstrategy in ARDS. Although those findings demonstrated thatadequate volume resuscitation is associated with improved clinicaloutcomes in patients with established sepsis-related ARDS, they didnot address whether patients who are treated for sepsis develop ARDSmore readily as a result of aggressive volume resuscitation. Our studyaddresses this question and found no association between the volumeof fluids administered within the first 6 or 24 hours and thedevelopment of ARDS within 72 hours of hospitalization, irrespectiveof comorbidities or other factors. Although the absence of astatistically significant association between the volume of fluidsadministered in the first 24 hours and the development of ARDS doesnot completely exclude the possibility of an increased risk, theadjusted OR of 1.07 and the narrow 95% CI suggest that, if there is anincreased risk, the magnitude is small. Thus, our results, takentogether with the findings by Murphy et al, add support to thefavorable risk-benefit profile of aggressive volume resuscitation in theearly phases of severe sepsis and septic shock. Our results also indicatethat concern regarding an increased propensity to develop ARDS dueto underlying comorbidities, such as congestive heart failure, shouldnot limit the use of adequate fluid for resuscitation in the early stagesof sepsis.

In the multivariable logistic regression models, the serum albuminand the APACHE II score were the most informative covariates for thedevelopment of ARDS. The associations between these variables andthe development of ARDS have been previously reported in themedical literature [7,22–24]. Notably, a recent study by Mikkelsen etal [7] identified several risk factors for the development of ARDS inpatients presenting to the emergency department with severe sepsisincluding baseline APACHE II score. The adjusted OR for APACHE IIscore in that studywas 1.06 (95% CI, 1.01-1.11), which is similar to ourresults. The consistency of our findings with the medical literaturewhere overlap exists supports the validity of our models. In addition,the associations between serum albumin and the baseline APACHE IIscore to the development of ARDS are biologically plausible. Serumalbumin levels influence vascular oncotic pressure and are abiomarker of the burden of acute and chronic illness. The APACHE IIscore is a multidimensional clinical marker that has been shown tocorrelate with the extent of systemic inflammation, clinical outcomes,and the severity of acute illness in sepsis and ARDS [7,22,23,25–27].From these findings, we hypothesize that patients with severe sepsisand septic shock do not develop ARDS because they receive more IVfluids, but instead receive more IV fluids in the first 24 hours ofhospitalization because they have greater burden of acute illnessrequiringmoreaggressive resuscitation andadiminishedhemodynamicresponse to IV fluids due to lower intravascular oncotic pressure.

There are several limitations to our study. First, the retrospective,observational design limits our ability to identify causal relationshipsbetween the fluid management strategies and the development ofARDS. Second, because our study was performed at a single, urban,public hospital, the results may not generalize to other clinicalenvironments. However, the consistency of our findings with other

1015D.W. Chang et al. / Journal of Critical Care 29 (2014) 1011–1015

studies that examined the epidemiology and risk factors for sepsis-induced ARDS suggests that our findings are broadly applicable. Third,the resuscitation fluid that was used in the study was exclusivelycrystalloids. However, a recent study by Caironi et al [28] found thatthere were no differences in mortality, respiratory failure, andduration of mechanical failure in patients with severe sepsis whowere resuscitated with crystalloids vs crystalloids and colloids. Thesefindings support the interpretation that differences in the type ofresuscitation fluids do not limit the external validity of our studyresults. Finally, because this was not a randomized study, it is possiblethat unmeasured or unknown confounding variables that were notaccounted for in the multivariable models could bias the results.

5. Conclusions

The volume of fluid administered in the first 24 hours ofhospitalization in patients with severe sepsis and septic shock is notsignificantly associated with the development of ARDS after controllingfor confounding factors, including underlying comorbidities. Althoughthis does not preclude the possibility of a small increased risk ofdeveloping ARDS due to aggressive fluid resuscitation, the findings ofour study support the overall favorable risk-to-benefit profile ofadequate volume resuscitation in patients admitted with severe sepsisand septic shock and indicate that fluid resuscitation should not belimited by clinical concerns regarding proclivity to induce ARDS.

Acknowledgments

We thank Chi-hong Tseng, PhD, for his critical review of thestatistical analysis and manuscript.

References

[1] Dellinger RP, Levy MM, Rhodes A, et al. Surviving sepsis campaign: internationalguidelines for management of severe sepsis and septic shock: 2012. Crit Care Med2013;41:580–637.

[2] Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment ofsevere sepsis and septic shock. N Engl J Med 2001;345:1368–77.

[3] Groeneveld AB, Polderman KH. Acute lung injury, overhydration or both? CritCare 2005;9:136–7.

[4] Rivers EP, Jaehne AK, Eichhorn-Wharry L, et al. Fluid therapy in septic shock. CurrOpin Crit Care 2010;16:297–308.

[5] Vincent JL, Gerlach H. Fluid resuscitation in severe sepsis and septic shock: anevidence-based review. Crit Care Med 2004;32:S451–4.

[6] BrowerRG,WareLB, BerthiaumeY, et al. TreatmentofARDS.Chest 2001;120:1347–67.[7] Mikkelsen ME, Shah CV, Meyer NJ, et al. The epidemiology of acute respiratory

distress syndrome in patients presenting to the emergency department withsevere sepsis. Shock 2013;40:375–81.

[8] Ware LB, Matthay MA. The acute respiratory distress syndrome. N Engl J Med2000;342:1334–49.

[9] Wiedemann HP, Wheeler AP, Bernard GR, et al. Comparison of two fluid-management strategies in acute lung injury. N Engl J Med 2006;354:2564–75.

[10] Levy MM, Fink MP, Marshall JC, et al. 2001 SCCM/ESICM/ACCP/ATS/SISinternational sepsis definitions conference. Crit Care Med 2003;31:1250–6.

[11] Bernard GR, Artigas A, Brigham KL, et al. The American-European ConsensusConference on ARDS. Definitions, mechanisms, relevant outcomes, and clinicaltrial coordination. Am J Respir Crit Care Med 1994;149:818–24.

[12] Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distresssyndrome: the Berlin definition. JAMA 2012;307:2526–33.

[13] Hosmer DW, Hosmer T, Le Cessie S, et al. A comparison of goodness-of-fit tests forthe logistic regression model. Stat Med 1997;16:965–80.

[14] Lemeshow S, Hosmer Jr DW. Estimating odds ratios with categorically scaledcovariates inmultiple logistic regression analysis. Am J Epidemiol 1984;119:147–51.

[15] Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley; 2000.[16] Boyd JH, Forbes J, Nakada TA, et al. Fluid resuscitation in septic shock: a positive

fluid balance and elevated central venous pressure are associated with increasedmortality. Crit Care Med 2011;39:259–65.

[17] Micek ST, McEvoy C, McKenzie M, et al. Fluid balance and cardiac function in septicshock as predictors of hospital mortality. Crit Care 2013;17:R246.

[18] Murphy CV, Schramm GE, Doherty JA, et al. The importance of fluid managementin acute lung injury secondary to septic shock. Chest 2009;136:102–9.

[19] Hughes CG, Weavind L, Banerjee A, et al. Intraoperative Risk Factors for AcuteRespiratoryDistress Syndrome inCritically Ill Patients. AnesthAnalg 2010;111:464–7.http://dx.doi.org/10.1213/ANE.1210b1013e3181d1218a1216a.

[20] Jia X, Malhotra A, Saeed M, et al. Risk factors for ARDS in patients receivingmechanical ventilation for N 48 h. Chest 2008;133:853–61.

[21] Plurad D, Martin M, Green D, et al. The decreasing incidence of late posttraumaticacute respiratory distress syndrome: the potential role of lung protectiveventilation and conservative transfusion practice. J Trauma Acute Care Surg2007;63:1–8. http://dx.doi.org/10.1097/TA.1090b1013e318068b318061ed.

[22] Fein AM, LippmannM, Holtzman H, et al. The risk factors, incidence, and prognosisof ARDS following septicemia. Chest 1983;83:40–2.

[23] Iscimen R, Cartin-Ceba R, Yilmaz M, et al. Risk factors for the development of acutelung injury in patients with septic shock: an observational cohort study. Crit CareMed 2008;36:1518–22.

[24] Mangialardi RJ, Martin GS, Bernard GR, et al. Hypoproteinemia predicts acuterespiratory distress syndrome development, weight gain, and death in patientswith sepsis. Ibuprofen in Sepsis Study Group. Crit Care Med 2000;28:3137–45.

[25] Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of diseaseclassification system. Crit Care Med 1985;13:818–29.

[26] Moran JL, Solomon PJ, Fox V, et al. Modelling thirty-day mortality in the AcuteRespiratory Distress Syndrome (ARDS) in an adult ICU. Anaesth Intensive Care2004;32:317–29.

[27] Shim HJ, Jang JY, Lee SH, et al. The effect of positive balance on the outcomes ofcritically ill noncardiac postsurgical patients: a retrospective cohort study. J CritCare 2014;29:43–8.

[28] Caironi P, Tognoni G, Masson S, et al. Albumin replacement in patients with severesepsis or septic shock. N Engl J Med 2014;370:1412–21.