Infections and the inflammatory response in acute respiratory distress syndrome
A. Stacey Headley
Study objective: Systemic inflammatory response syndrome (SIRS) and infections are frequently associated with the development and progression of acute respiratory distress syndrome (ARDS) and multiple organ dysfunction syndrome (MODS). We investigated, at onset and during the progression of ARDS, the relationships among (1) clinical variables and biological markers of SIRS, (2) infections defined by strict criteria, and (3) patient outcome. Biological markers of SIRS included serial measurements of inflammatory cytokines (IC)–tumor necrosis factor-[Alpha] (TNF-[Alpha]) and interleukins (IL) 1[Beta], 2, 4, 6, and 8–in plasma and BAL fluid.
Methods: We prospectively studied two groups of ARDS patients: 34 patients treated conventionally (group 1) and nine patients who received glucocorticoid rescue treatment for unresolving ARDS (group 2). Individual SIRS criteria and SIRS composite score were recorded daily for all patients. Plasma IC levels were measured by enzyme-linked immunosorbent assay on days 1, 2, 3, 5, 7, 10, and 12 of ARDS and every third day thereafter while patients received mechanical ventilation. Unless contraindicated, bilateral BAL was performed on day 1, weekly, and when ventilator-associated pneumonia was suspected. Patients were closely monitored for the development of nosocomial infections (NIs).
Results: ICU mortality was similar among patients with and without sepsis on admission (54% vs 40%; p[is less than]0.45). Among patients with sepsis-induced ARDS, mortality was higher in those who subsequently developed NIs (71% vs 18%; p[is less than]0.05). At the onset of ARDS, plasma TNF-[Alpha], IL-1[Beta], IL-6, and IL-8 levels were significantly higher (p[is less than]0.0001) in nonsurvivors (NS) and in those with sepsis (p[is less than]0.0001). The NS group, contrary to survivors (S), had persistency elevated plasma IC levels over time. In 17 patients, 36 definitive NIs (17 in group 1 and 19 m group 2) were diagnosed by strict criteria. No definitive or presumed NIs caused an increase m plasma IC levels above patients’ preinfection baseline. Daily SIRS components and SIRS composite scores were similar among S and NS and among patients with and without sepsis-induced ARDS, were unaffected by the development of NI, and did not correlate with plasma IC levels.
Conclusions: Sepsis as a precipitating cause of ARDS was associated with higher plasma IC levels. However, NIs were not associated with an increase m SIRS composite scores, individual SIRS criteria, or plasma IC levels above patients’ preinfection baseline. SIRS composite scores over time were similar in S and NS. SIRS criteria, including fever, were found to be nonspecific for NI. Irrespective of etiology of ARDS, plasma IC levels, but not clinical criteria, correlated with patient outcome. These findings suggest that final outcome m patients with ARDS is related to the magnitude and duration of the host inflammatory response and is independent of the precipitating cause of ARDS or the development of intercurrent NIs. (CHEST 1997;111:1306-21)
Abbreviations: ARDS=acute respiratory distress syndrome, ANOVA=analysis of variance, APACHE=acute physiology and chronic health evaluation; CAP = community-acquired pneumonia; CRI = catheter-related infection GCRT= glucocorticoid rescue treatment; IC =inflammatory cytokine; IL=interleukin; LIS = lung injury score; LPS= lipopolysaccharide; MODS = multiple organ dysfunction syndrome; MV= mechanical ventilation; NI = nosocomial infection; SIRS =systemic inflammatory response syndrome; TN-[Alpha]=tumor necrosis factor-[Alpha]; UTI=urinary tract infection; VAP=ventilator-associated pneumonia
Acute respiratory distress syndrome (ARDS) represents an intense host inflammatory response of the lung to an infectious or noninfectious pulmonary or extrapulmonary insult. In a nonsurgical population, sepsis is the most frequent condition precipitating ARDS. Despite aggressive supportive treatment, mortality in ARDS still remains [is greater than]60% and may account for more than 100,000 deaths annually in the United States.[3,4] Early death in ARDS, defined as within 3 days of onset, is directly related to the condition precipitating acute respiratory failure. In mortality data, after day 3 of ARDS, most patients die following a prolonged period of ventilatory support, during which they often develop fever, systemic inflammatory response syndrome (SIRS), clinical manifestations of sepsis,[5,7,8] and multiple organ dysfunction syndrome (MODS).[9,10] In the medical literature, sepsis is associated with fatality in 36 to 90% of patients with ARDS.[5,7,9,10] At necropsy, 69% of ARDS nonsurvivors show histologic evidence of pneumonia.
Based on these observations, it has been hypothesized that in ARDS a direct correlation exists among nosocomial infection (NI), amplification of the systemic inflammatory response, and higher mortality. Support for this hypothesis, however, relied only on clinical studies that did not use strict criteria for diagnosing NI. Often positive blood cultures are required to diagnose sepsis, but this criterion is too exclusive because only a minority (20 to 40%) of patients with NI manifest bacteremia. Recent investigative developments provide an opportunity to clarify the complex interaction between infections and the inflammatory response in patients with ARDS: recognizing the key role of cytokines as proximal mediators of inflammation, applying bronchoscopic techniques to improve diagnostic accuracy for pneumonia, and standardizing definitions of SIRS and sepsis.
We conducted a prospective study to investigate, at the onset and during the progression of ARDS, the relationship among infection, clinical parameters and biological markers of SIRS, and patient outcome. Potential sites of infection were sought by using a comprehensive investigative protocol with final diagnosis requiring strict diagnostic criteria. Biological markers of SIRS included serial measurements of inflammatory cytokines (ICs): tumor necrosis factor-[Alpha] (TNF-[Alpha]) and interleukins (IL) 1[Beta], 2, 4, 6, and S. The inflammatory response was monitored at both the systemic (SIRS composite score and plasma IC levels) and pulmonary levels (BAL IC levels). Our results support the hypothesis that patient outcome in ARDS is related to the magnitude and duration of the host inflammatory response and is independent of the precipitating cause of ARDS and the development of intercurrent NIs.
Materials and Methods
The study was conducted between January 1992 and May 1993 at the Medical ICU of the Regional Medical Center and the University of Tennessee Medical Center, Memphis. The protocol was approved by the University of Tennessee institutional review board, and informed consent was obtained before entrance into the study. We prospectively studied 43 consecutive patients who developed medical ARDS. Thirty-eight patients were admitted to the ICU directly from the emergency department, and five patients were admitted from the general medicine ward. Patients entered the study within 24 h of developing ARDS. Study patients were divided into two groups–34 patients (group 1) treated conventionally and nine patients (group 2) who received glucocorticoid rescue treatment (GCRT) for unresolving ARDS. We previously described 21 of 34 patients in group 1 and the nine patients in group 2.
The following data were compiled for each patient: (1) demographic information (sex, age, race); (2) APACHE II (acute physiologic and chronic health evaluation) score and predicted mortality on ICU admission and APACHE II score on ICU day 3; (3) cause of ARDS and lung injury score (LIS); (4) presence or absence of SIRS, sepsis, shock, and MODS and MODS score at the time ARDS developed; (5) cardiopulmonary physiologic parameters; and (6) outcome.
We investigated, at the onset and during the progression of ARDS, the relationship among (1) infections defined by strict criteria, (2) SIRS, and (3) patient outcome. Infections were classified as (a) precipitating ARDS, and (b) NI developing after day 3 of mechanical ventilation (MV). SIRS individual clinical variables and SIRS composite score (see “Definitions” section) were monitored daily. Biological markers of SIRS included plasma IC levels obtained on days 1, 2, 3, 5, 7, 10, and 12 of ARDS and every third day thereafter while patients received MV. ICU outcome was classified as early death ([less than or equal to] 3 days) or late death ([greater than] 3 days).
ARDS was defined by the presence of the following criteria: (1) acute respiratory failure requiring MV; (2) bilateral pulmonary infiltrates on chest radiograph; (3) pulmonary artery occlusion pressure [is less than]18 mm Hg (available in 36 patients with hemodynamic monitoring); (4) static pulmonary compliance [is less than]50 mL/cm [H.sub.2]O; (5) ratio of arteriolar oxygen tension to inspired oxygen concentrations [is less than]200; and (6) appropriate etiology for the development of ARDS. The precipitating cause of ARDS was classified as either direct or indirect lung injury. MODS criteria and MODS score were previously described. Shock was defined as the presence of a mean arterial BP [less than or equal to]60 mm Hg or systolic BP [less than or equal to]90 mm Hg that did not resolve following a 500-mL bolus infusion of IV fluids and required vasopressors. Survival was analyzed at day 3 and 7 of ARDS and at the time of ICU and hospital discharge. Patients were classified as survivors if they were discharged alive from the ICU without requiring MV. Causes of death were classified as being due to MODS, shock, respiratory failure, or other.
SIRS was defined using a modification of the guidelines developed by the American College of Chest Physicians and Society of Critical Care Medicine Consensus Conference: (1) temperature [is greater than] 38 [degrees] C or [is less than] 36 [degrees] C; (2) lowest heart rate per day [is greater than or equal to] 290 beats/mini (3) respiratory rate [is greater than] 20 breaths/min or ventilatory requirements in excess of 10 L/min; (4) WBC count [is greater than]12,000/[mm.sup.3] or [is less than] 4,000/[mm.sup.3]; and (5) [greater than or equal to] 10% immature (band) forms on the peripheral blood smear. The SIRS composite score (SIRS score) was calculated daily by assigning one point to each one of the five SIRS criteria. In each patient, maximum score was defined as the highest SIRS score attained during the course of ARDS, and mean SIRS score over time was defined as the sum of the daily SIRS score divided by the number of measured days.
The diagnosis of sepsis was defined using recently developed guidelines and required at least two SIRS criteria and microbiologic documentation of infection using strict diagnostic criteria. All microbiologic cultures not meeting strict diagnostic criteria for infection were recorded and analyzed separately. When infections were suspected, a previously described systematic protocol was used with careful search for ventilator-associated pneumonia (VAP), sinusitis, catheter-related infection (CRI), urinary tract infection (UTI), and abdominal abnormality. The following diagnostic procedures were used and the results were carefully reviewed retrospectively to determine the source(s) of infection: 92 bronchoscopies with BAL (62 with bilateral BAL); 30 CT scans of the chest; 13 thoracenteses; 44 CT scans of the sinuses; nine bilateral maxillary sinus aspirates; 92 semiquantitative cultures of central line catheters; 127 sets of blood cultures; 94 urinalyses and cultures; 26 CT scans of the abdomen and pelvis; 11 abdominal ultrasounds; five paracenteses; three exploratory laparotomies; six lumbar punctures; nine gallium-67 scintigraphies; one indium scan; one arthrocentesis; two autopsies; and 14 others. Infections were classified as present on admission if diagnosed within 48 h of ARDS development or NI if suspected and diagnosed after day 3 of ARDS onset. Chemical aspiration was defined as witnessed aspiration with negative bronchoscopic findings for pneumonia. Primary bacteremia was defined as a positive blood culture (coagulase-negative staphylococci were excluded) without an identified primary source of infection. Secondary bacteremia was defined as a positive blood culture with recovery of an identical microorganism from a culture of the suspected primary source of infection.
Blood and BAL Collection and Laboratory Processing
Blood samples were obtained on days 1, 2, 3, 5, 7, 10, and 12 of ARDS and every third day thereafter until extubation. All blood samples were obtained from a central venous line or an antecubital venipuncture, placed in a specimen (Vacutainer; Becton Dickinson; Franklin Lakes, NJ) tube containing edetic acid, and transported to the laboratory for immediate processing. Blood samples were centrifuged at 1,500 g for 10 min, and plasma was aspirated and stored at -70 [degrees] C. BAL specimens were collected, following a previously reported method, on day 1 of ARDS and every 7 days (unless contraindicated) or when the patient developed clinical manifestations suggesting VAP, defined as the development of fever, purulent tracheal secretions, and new or progressive densities visible on chest radiograph. Right and left BAL samples were kept separate and processed immediately for quantitative bacterial culture, total and differential cell count, and cytologic study following a previously described method. After obtaining the first BAL sample, the bronchoscope was removed from the patient and the suction channel was flushed with 60 to 100 mL of sterile saline solution, and the distal 6 to 8 cm of the bronchoscope was swabbed with sterile alcohol preparations prior to completing the bilateral BAL procedure to minimize potential cross-contamination between left and right BAL samples. Diagnostic threshold for positive quantitative bacterial culture was a growth [greater than or equal to] [10.sup.4] cfu/mL. The remainder of each BAL specimen was centrifuged at 3,500 rpm for 10 min in a centrifuge (Beckman TJ6; Fullerton, Calif), and the supernatants were collected and stored at -70[degrees] C. Prior to assay, each BAL supernatant was concentrated 20-fold using a concentrator (Savant Speed Vac; Farmingdale, NY) and then centrifuged in a microfuge at 15,000 rpm for 1 min. The method for measuring BAL albumin and protein levels was reported previously.
Cytokine levels were determined by a solid-phase enzyme-linked immunosorbent assay method based on the quantitative immunometric sandwich enzyme immunoassay technique. Reagents for the various cytokines were obtained from several sources (IL-2, IL-4, IL-6, and TNF-[Alpha] from Genzyme, Cam bridge, Mass; IL-[Beta] from Endogen, Boston; and IL-8 from R & D Systems, Minneapolis). The laboratory method was previously described.[1,15]
Several statistical methods were used for describing and comparing specified groups and subgroups of patients. Comparisons between groups 1 and 2 for demographic and clinical variables at the onset of ARDS were made with t test, [chi square] tests, and Fishers Exact two-tailed tests. Likewise, comparisons between survivors and nonsurvivors in group 1 for clinical variables at the onset of ARDS were made with similar tests. For group 1 patients, survival analysis, based on Kaplan-Meier methods, was used to depict and compare the distributions of survival estimates for (1) all survivors and nonsurvivors, and (2) survivors and nonsurvivors with and without NIs, given that they had survived for at least 3 days after the onset of ARDS. Actual and predicted mortalities were compared with z tests. Although power for tests was low, t tests were used to assess whether, on the day of onset of ARDS, group 2 patients were comparable to nonsurvivors in group 1.
To make inferences about relationships among variables, several statistical methods were used. For group 1 patients, repeated measures of analysis of variance (ANOVA) were used to assess the relationships between clinical variables (de, survival, sepsis on admission, and SIRS scores on admission) and IC values measured on the day of onset of ARDS and thereafter. Because the normality assumptions for ANOVA were not met, IC values were log transformed; antilogs of the means and SEs are presented in the text and tables. A separate analysis was conducted to compare each of the following over time: (1) survivors and nonsurvivors; (2) patients with and without sepsis; (3) patients with composite SIRS scores [greater than or equal to] 3 on admission and those with lower scores; and (4) patients with composite SIRS scores [greater than or equal to] 4 on admission and those with lower scores. For each ANOVA, patients were classified into two groups according to the particular clinical variable of interest. Then, these groups and the patients tested within those groups were cross-classified with days. For each day, least squares means for the two specified groups were compared with preplanned orthogonal contrasts (ie, F tests, but equivalent to t tests). In addition, for each day Pearson product moment correlation coefficients were estimated. Because the shared variance would have accounted for about 50% of the total variance of either variable, an estimated correlation coefficient of 0.7 was arbitrarily defined as clinically useful.
Other statistical methods were also used for making inferences. For patients with NIs, we hypothesized that on the day of infection, plasma and BAL IC, BAL albumin, BAL total protein, BAL neutrophil levels, composite SIRS scores, and individual SIRS components would exhibit change relative to values obtained 3 days before and after diagnosis. Therefore, paired t tests, Wilcoxon signed rank tests, and sign tests were used to compare values obtained on the day in which NI was diagnosed with the mean of values obtained 3 days before and after. Although power was low, the relationship between presence of NI and survival was assessed with Fisher’s Exact two-tailed test. Finally, for patients with pneumonia, the ANOVA method (ie, one-way ANOVA with a completely random model) was used to estimate intraclass correlation coefficients for quantifying the agreement among IC, total protein, albumin, and neutrophil values in right and left BAL fluid.
Clinical Variables at the Onset of ARDS for Groups 1 and 2
Demographics and clinical variables at the onset of ARDS for groups 1 and 2 are shown in Table 1. No significant difference was found between the two groups except for sex, LIS, cardiovascular dysfunction, and the presence of shock. Causes of direct lung injury for group 1 included 17 pneumonia (7 with bacteremia), two chemical aspiration, and one bleomycin-oxygen toxic reaction. Causes of direct lung injury for group 2 included two pneumonia and three chemical aspiration. Causes of indirect lung injury for group 1 included four extrapulmonary infections, three primary bacteremia, one transfusion reaction, one pancreatitis, one sickle cell chest syndrome, one hemorrhagic shock, one toxemia of pregnancy, one intracranial bleed with aspirin toxic reaction, and one extrathoracic trauma. Causes of indirect lung injury for group 2 included one intra-abdominal infection, one sickle cell chest syndrome, and two UTIs. In group 1, prevalence of specific organ dysfunction in addition to the lung at the onset of ARDS included 17 cardiovascular (50%), 13 renal (38%), 10 GI (29%), 10 hematologic (29%), seven neurologic (21%), and six hepatic (18%).
Table 1-Demographics and Clinical Variables at Onset
of ARDS for Groups 1 and 2
Group 1 Group 2
No. of patients 34 9
Age,([dagger]) yr 44[+ or -]2 37[+ or -]4
Sex, M:F 22:12 2:7
APACHE II scores([dagger])
Day 1 23[+ or -]1 22[+ or -]3
Day 3 25[+ or -]1 20[+ or -]3
LIS day 1([dagger]) 2.72[+ or -]0.12 3.25[+ or -]0.13
Direct lung injury 21 5
SIRS scores([dagger]) 3.37[+ or -].33 3.66
[+ or -]
Sepsis 24 5
Septic shock 16 1
Shock, any etiology 20 1
MODS score 0([sections]) 8 6
MODS score 1-2([sections]) 15 3
[is greater than]2([sections]) 11 0
ICU mortality, % 50 44
No. of patients
Age,([dagger]) yr 0.18
Sex, M:F 0.03
APACHE II scores([dagger])
Day 1 0.55
Day 3 0.07
LIS day 1([dagger]) 0.007
Direct lung injury 0.9
SIRS scores([dagger]) 0.6
Septic shock 0.07
Shock, any etiology 0.02
MODS score 0[sections] –
MODS score 1-2[sections] –
MODS score [is greater than]2([sections]) 0.2
ICU mortality, %
(*) Probability value associated with the test statistic given
that the null hypothesis (that the two groups are the same) is true.
([dagger]) Mean [+ or -] SE.
([double daggers]) n=6.
([sections]) Number of organ dysfunctions in addition to
the lung. MODS score 3 in four patients, 4 in five patients,
and 5 in two patients.
Day 1 of ARDS: Relationship Between Infections Precipitating ARDS and Outcome in Group 1
Overall ICU mortality was similar in patients with and without sepsis on admission (54% vs 40%; p=0.45). However, most (six of seven) patients who died early had sepsis-induced ARDS. Causes of sepsis inducing ARDS and outcome are shown in Table 4. Seventeen patients had community-acquired pneumonia (CAP) that was diagnosed by bronchoscopy in 16 and by recovering an identical microorganism in the endotracheal aspirate and in blood cultures in one. Pneumonia was caused by Streptococcus pneumoniae in five, Gram-negative organisms in five (Acinetobacter cloacae, Enterobacter aerogenes, Haemophilus influenzae, Klebsiella pneumoniae [two]), polymicrobial (mixed Gram-positive and Gram-negative) in three, and other pathogens in four. Among patients with CAP, ICU mortality was higher in those with secondary bacteremia (71% vs 30%; p=0.16). ICU mortality in the three patients with primary bacteremia on admission was 100%.
Table 4-Outcome Among Group 1 Patient# With
Sepsis on Admission
Sepsis on admission 11 13
Pneumonia 9(2)[*] 8(5)[*]
Primary bacteremia 0 3
Sinusitis 2 1
Abdominal infection 0 1
Cause of Death
Sepsis on admission
Primary bacteremia MODS-1; shock-2
Abdominal infection MODS-1
[*] Number with secondary bacteremia in parentheses.
Day 1 of ARDS: Relationships Between Plasma Cytokine Levels and Outcome in Groups 1 and 2
Group 1: Patients dying early ([less than or equal] 3 days) had significantly higher mean ([+ or -]SE) plasma cytokine levels for TNF-[Alpha] (463[+ or -]19 vs 322[+ or -]25; p=0.002), L-1[Beta] (484[+ or -]12 vs 347[+ or -]16; p=0.0001), IL-6 (642[+ or -]13 vs 615[+ or -]10; p=0.02), and IL-8 (996[+ or -]27 vs 793[+ or -]20; p=0.0001). As shown in Table 3, nonsurvivors of ICU admission and patients with sepsis-induced ARDS had significantly higher plasma TNF-[Alpha], IL-1[Beta], IL-6, and IL-8 levels. These findings are similar to previously reported data incorporated into this report for 21 patients. In that study, we found IL-1[Beta] to be a consistent and efficient predictor of outcome on admission and over time. In the current study, 13 of the 15 patients with day 1 plasma IL-1[Beta]400 pg/mL died in the ICU, and 15 of 19 patients with IL-[Beta][is less than]400 pg/mL survived.
Group 2: At the onset of ARDS, mean ([+ or -]SE) plasma cytokine levels (pg/mL) in group 2 patients were similar to nonsurvivors in group 1: TNF-[Alpha] (370[+ or -]24 vs 403[+ or -]70; p=0.68), IL-1[Beta] (509[+ or -]40 vs 518[+ or -]35; p=0.97), IL-2 (466[+ or -]143 vs 303[+ or -]29; p=0.55), IL-4 (222[+ or -]58 vs 188[+ or -]19; p=0.88), IL-6 (1,023[+ or -]291 vs 654[+ or -]92; p=0.15), and IL-8 (554[+ or -]55 vs 762[+ or -]154; p=0.93).
Day 1 of ARDS: Relationships Between Plasma Cytokine Levels and SIRS Criteria in Group 1
The relationship between mean ([+ or -]SE) plasma cytokine levels and SIRS score on day 1 of ARDS is shown in Table 3. There were weak correlations (not clinically useful; eg, r[is less than]0.7) between SIRS score and plasma levels of TNF-[Alpha] (r=0.2; p=0.25), IL-1[Beta] (r=0.3; p=0.1), IL-2 (r=0.33; p=0.06), IL-4 (r=0.35; p=0.04), IL-6 (r=0.33; p=0.06), and IL-8 (r=0.33; p=0.06). When the presence and absence of individual SIRS components were analyzed independently and mean ([+ or -]SE) plasma IC levels (pg/mL) were compared, the only significant correlation (not clinically useful; eg, r[is less than]0.7) was found for tachycardia (present in 30 of 34 patients) and TNF-[Alpha] (present 327[+ or -]45 vs absent 83[+ or -]11; r=0.53; p=0.001), IL-1[Beta] (present 425[+ or -]31 vs absent 218[+ or -]31; r=0.41; p=0.02), IL-6 (present 578[+ or -]62 vs absent 183[+ or -]20; r=0.57; p=0.0005), and IL-8 (present 652[+ or -]110 vs absent 113[+ or -]15, r=0.49; p=0.003); leukocytosis (present in 20 of 34 patients) and IL-4 (present 195[+ or -]91 vs absent 133[+ or -]70; r=0.35; p=0.04); and presence of band forms (present in 7 of 34 patients) and IL-6 (present 785[+ or -]149 vs absent 465[+ or -]58; r=0.40, p=0.02). Cytokine levels were similar in patients with (n=26) and without fever on admission.
ARDS Over Time: Relationship Between NIs and SIRS in Groups 1 and 2
Group 1: Irrespective of sepsis, fever was a frequent finding. After day 3 of ARDS, fever was present in all but four patients (89% with NI and 79% without NJ). Fever was persistent (maximum temperature [is greater than]38[degrees]C on each day of MV) throughout the course of MV in two of nine patients (22%) with NI and in 9 of 14 patients (64%) without NI, while a new episode of fever ([is greater than]24 h duration following a 48-h interval without fever) developed in 6 of 9 (67%) patients with NI and in 2 of 14 (14%) patients without NI. Within [+ or -] 3 days of the 17 documented NIs, core body temperature (Tmax) increased by [is greater than]1[degrees]F (above 100.4[degrees]F [38[degrees]C] or above baseline Tmax in patients with preexisting fever) in 8 (47%) episodes of NI, and leukocyte count increased by [is greater than]3,000/[mm.sup.3] in 12 (71%) episodes of NI. Among the 14 patients receiving MV [is greater than]3 days without NI, core body temperature increased by [is greater than]1[degrees]F in six patients (43%) and leukocyte count increased by [is greater than]3,000/ [mm.sup.3] in eight patients (57%). At any time of ARDS, no differences in individual SIRS components were found between patients with infections caused by Gram-positive and those with infections caused by Gram-negative microorganisms.
For each of the 36 NIs (groups 1 and 2 combined), we compared SIRS score and presence of each of the SIRS components on the day the NI was diagnosed with the average of values obtained 3 days before and after. None of the NIs was associated with a significant increase or decrease in individual components or the SIRS score. The highest level of significance (lowest p) was found for fever in group 1 (p=0.06).
ARDS Over Time: Relationship Between NIs and Plasma Cytokine Levels in Groups 1 and 2
This analysis includes patients from both groups 1 and 2. For each of 36 NIs, we compared plasma cytokine levels obtained on the day of a positive culture with (1) plasma cytokine levels obtained 3 days before development of NI and (2) the mean of plasma cytokine levels obtained 3 days before and 3 days after development of NI. None of the NI was associated with any increase or decrease in plasma TNF-[Alpha], IL-1[Beta], IL-2, IL-4, IL-6, or IL-8 levels. Furthermore, similar findings were also obtained for each of the 58 positive cultures not meeting strict diagnostic criteria for infection. Figure 2 shows plasma IL-1[Beta] and IL-6 levels of individual patients over time and the timing of NI in nonsurvivors of group 1. A similar finding was observed for TNF-[Alpha] and IL-8 in nonsurvivors and for TNF-[Alpha], IL-1[Beta], IL-6, and IL-8 levels in survivors (not shown). Figure 3 shows plasma IL-1[Beta] and IL-6 levels of individual patients over time during GCRT and timing of NI in survivors and nonsurvivors of group 2. Similar findings were also observed for TNF-[Alpha] and IL-8 (not shown).
[Figure 2 and 3 ILLUSTRATION OMITTED]
Relationships Between Pneumonia and BAL Inflammatory Cytokine Levels in Groups 1 and 2
This analysis includes patients from both groups 1 and 2, and is limited to the 30 patients who were subjected to 61 bronchoscopies with bilateral BAL sampling. Only 5 of the 21 cases of pneumonia (24%) had significant growth on both right and left BAL. There was excellent agreement between right and left BAL cytokine levels (TNF-[Alpha], IL-1[Beta], IL-2, IL-4, IL-6, and IL-8), albumin, and total protein irrespective of the presence of significant unilateral or bilateral bacterial growth (Table 5). There was poor agreement (eg, r[is less than]0.7) between right and left BAL for total inflammatory cell count and neutrophil percentage irrespective of unilateral or bilateral significant bacterial growth. In patients with unilateral pneumonia, total inflammatory cell count and neutrophil percentage were not higher in the BAL with significant bacterial growth compared to the BAL without significant growth and, therefore, were not useful in predicting the presence or absence of pneumonia. Correlations between right and left BAL cytokine levels, albumin, and total protein for patients without pneumonia were reported previously. ARDS Over Time: Relationship Between NIs and Outcome in Group 1
Twenty-three patients remained ventilator dependent after 3 days of ARDS, and nine (39%) developed NI. The relationship between developing a specific NI and outcome is shown in Table 6. Among patients (n = 18) with sepsis-induced ARDS, mortality was higher in the seven patients with NI (71% vs 18%; p = 0.05). Among patients without sepsis-induced ARDS, mortality was similar in patients with (n = 2) and without (n = 8) NI (50% vs 38%; p = 0.99). Although overall mortality was higher in patients with NI (67% vs 29%), it did not reach statistical significance (p =0.1). Six episodes of VAP developed in five patients, and four had CAP-induced ARDS. Mortality rate in patients with VAP was 80%. The bacterial etiology of VAP included one Escherichia coli, one Staphylococcus aureus, and four polymicrobial pneumonia (E coli, Enterobacter cloacae, Proteus mirabilis, Pseudomonas aeruginosa, and S aureus). Three patients with bacteremic VAP had bacteremia at the onset of ARDS caused by a different pathogen (two secondary to CAP and one primary). NI developed more frequently in patients receiving prolonged MV. In group 1, 9 of 17 NIs (53%) developed after day 10 of MV. The mean ([+ or -] SE) duration of MV was significantly longer in patients developing NI (27 [+ or -] 6 vs 10 [+ or -] 1 days; p = 0.003). A Kaplan-Meier curve with the duration of MV in patients with and without NI is shown in Figure 1, bottom. The overall rate of NI per day of MV was 5%-1% in survivors and 8% in nonsurvivors.
ARDS Over Time: Relationship Between SIRS and Outcome in Groups 1 and 2
Group 1: For any day of MV, 97% of patients met two SIRS criteria, and 78% met three SIRS criteria. As shown in Figure 4, top, on any day of ARDS, no differences in mean SIRS scores were found between survivors and nonsurvivors (p = 0.12). With rare exceptions, incidences of individual SIRS components were similar (lowest p = 0.73) between survivors and nonsurvivors on any day of ARDS. The only significant differences between survivors and nonsurvivors were found for fever on days 2 (88% vs 53%; p = 0.02) and 3 (88% vs 53%; p = 0.02) of ARDS and for tachycardia on days 2 (94% vs 71%; p = 0.03) and 12 (94% vs 71%; p = 0.03) of ARDS. As shown in Figure 4, bottom, on any day of MV, no differences in mean ([+ or -] SE) SIRS scores were found between patients with and without sepsis-induced ARDS (p = 0.15) with the single exception of day 3 of MV, when patients with sepsis had a lower SIRS score (3.0 [+ or -] 0.1 vs 3.6 [+ or -] 0.2; p = 0.04).
[Figure 4 ILLUSTRATION OMITTED]
Mortality did not increase with sequentially higher values of the SIRS composite score or maximum score (Fig 5, top). Among patients with similar SIRS composite scores, mortality increased in the presence of sepsis, although it did not reach statistical significance (lowest p value 0.7); the reason for this result may have been attributable to lack of power. The relationship between infection and mortality is shown in Figure 5, center. The relationship among mean SIRS composite score over time, infections, and mortality is shown in Figure 5, bottom. Eight patients who never developed sepsis all had a SIRS score [is greater than or equal to] 3 over time and a mortality rate of 40%.
[Figure 5 ILLUSTRATION OMITTED]
Group 2: In group 2, 19 NIs occurred among eight patients, and the core body temperature (Tmax) increased (within [+ or -] 3 days) by [is greater than] 1 [degrees] F in association with 10 (53%) episodes of NI and the leukocyte count increased ([is greater than or equal to] 3,000/[mm.sup.3] within [+ or -] 3 days) in 10 (53%) episodes of NI. Within [+ or -] 3 days of the 13 documented NIs occurring during GCRT, temperature increased in nine (69%), and the leukocyte count increased in seven (54%) episodes.
ARDS Over Time: Relationships Between Plasma Cytokine Levels and Outcome in Group 1
Plasma IL-1[Beta] and IL-6 levels over time in survivors and nonsurvivors are shown in Figure 6, top. A similar pattern was observed for TNF-[Alpha] and IL-8 (not shown). Plasma IL-1[Beta] and IL-6 levels over time in patients with and without sepsis-induced ARDS are shown in Figure 6, bottom. During the first week of ARDS, plasma IL-1[Beta] level declined in all survivors but remained persistently elevated (within 10% of the initial day 1 value) in all but two (88%) nonsurvivors. The latter two patients, however, had plasma IL-1[Beta] levels [is greater than] 400 pg/mL on day 1 and 7 of ARDS and an average decline of 20% over 7 days. Plasma IL-1[Beta] level on day 1 of ARDS and over time was a consistent and efficient predictor of outcome. At the onset of ARDS, a plasma IL-1[Beta] level [is greater than] 400 pg/mL had a sensitivity of 77% and a specificity of 82% (positive predictive value, 81%; negative predictive value, 78%) for predicting death. By day 7 of ARDS, either a plasma IL-1[Bet] level [is greater than] 400 pg/mL or a persistently elevated plasma IL-1[Beta] level (within 10% of the initial day 1 value) had a sensitivity of 100% and a specificity of 100% for predicting death.
[Figure 6 ILLUSTRATION OMITTED]
ARDS Over Time: Relationships Between Plasma Cytokine Levels and SIRS Criteria in Group 1
On any day of ARDS, there were weak correlations (not clinically useful; eg, r [is less than] 0.7) between SIRS score and plasma levels of TNF-[Alpha] (r = 0.22; p = 0.02) and IL-8 (r = 0.1; p = 0.3). There was no correlation between SIRS score and plasma levels of IL-1[Beta] (r = 0.02; p = 0.8), IL-2 (r = 0.01; p = 0.9), IL-4 (r = 0.01; p = 0.9), and IL-6 (r = 0.02; p = 0.9). When individual SIRS components were analyzed independently, weak (not clinically useful; eg, r [is less than] 0.7) correlations were found among plasma TNF-[Alpha], IL-1[Beta], IL-6, and IL-8 levels (lowest p value reported in parentheses) and tachycardia (r = 0.18; p = 0.08), leukocytosis (r = 0.25; p = 0.008), band forms (r = 0.23; p = 0.01), and fever (r = 0.13; p = 0.2).
ARDS During Glucocorticoid Treatment: Relationships Between NIs and Plasma Cytokine Levels in Group 2
Five of nine patients in group 2 had sepsis-induced ARDS. All but one patient had a day 1 and day 7 plasma IL-1[Beta] level [is greater than] 400 pg/mL. Before receiving GCRT, patients met both physiologic (LIS) and biological (plasma IC levels) criteria predicting an excessive ([is greater than] 80%) mortality rate. Five patients had six episodes of NI, which were treated prior to initiating GCRT and included three VAP (A calcoaceticus, S aureus, P aeruginosa), one abdominal abscess (Candida albicans), one UTI (K pneumoniae and E cloacae), and one primary bacteremia (P aeruginosa). During GCRT, six of nine patients developed one or more episodes of VAP and six additional episodes of NI (Table 7). The bacterial etiology of VAP during GCRT included three P aeruginosa, one A calcoaceticus, one E cloacae, one Streptococcus viridans, and one Xanthomonas maltophilia. Mortality rate in the six patients with NI during GCRT was 33%. Both patients with bacteremia (one primary and one secondary to VAP) died. The overall rate of NI per day of MV was 8%-9% for survivors and 5% for nonsurvivors. Fifteen of 19 NIs (79%) developed after day 10 of MV. Survivors of GCRT had a longer duration of MV (36 [+ or -] 4 days vs 24 [+ or -] 4 days; p = 0.07) and a higher number of NIs per patient (2 vs 0.8). Figure 3 shows plasma IL-1[Beta] and IL-6 levels during GCRT and the timing of NI in survivors and nonsurvivors. A similar finding was seen for TNF-[Alpha] and IL-8. Survivors, but not nonsurvivors, had a significant reduction in plasma IC levels during GCRT.
Table 7-Relationship Between NIs and Outcome for Group 2 Patients
Sepsis on admission 2 3
NIs before HDCS(*)
(n = 5 patients)
Pneumonia 2 1
Abdomen 1 0
Bacteremia 0 1
UTI 1 0
NIs during HDCS
(n = 6 patients)
Pneumonia 5 2 (2)[dagger]
CRI 2 1
Sinusitis 1 0
UTI 2 0
(*) HDCS = high-dose corticosteroids. [dagger] Number of episodes of secondary bacteremia associated with NI.
The host defense response to an insult is similar, regardless of the tissue involved, and consists of an interactive network of simultaneously activated pathways that act in synergy to increase the chance of survival. The host defense is essentially a protective response of tissues and serves to destroy, dilute, or wall off injurious agents and to repair any tissue damage. This repair consists of replacing injured tissue by regenerating parenchymal cells and filling defects with fibroblastic tissue (fibroproliferation). Cellular responses are regulated by a complex interaction among cytokines with final effects on the microenvironment not directly induced by the initiating insult. In this regard, cytokines have concentration-dependent biological effects. At low concentration, they regulate homeostasis, and at progressively higher concentrations, they mediate proportionately stronger local and finally systemic responses.
Among a broad spectrum of proximal mediators, cytokines of the IL-1 and TNF family appear uniquely important in initiating all key aspects of the host defense response.[23,24] TNF-[Alpha] and IL-1[Beta] stimulate their own and each other’s secretion. and both promote the release of IL-6. The biological actions of cytokines during the host defense response are numerous and redundant–each mediator exerts multiple effects on different target cells, and different mediators act on the same cell to induce similar effects. Once released, TNF-[Alpha] and IL-1[Beta] act on epithelial cells, stromal cells (fibroblasts and endothelial), the extracellular matrix, and recruited circulating cells (neutrophils, platelets, lymphocytes) resulting in secondary waves of cytokine release with subsequent amplification of the host defense response. When homeostasis cannot be restored, a massive systemic reaction follows. At this stage, the predominant effects of cytokines become destructive rather than protective. Cytokines spill into the circulation, reach end organs, and produce additional sites of damage. The clinical expression of this host response has been termed the “systemic inflammatory response syndrome” (SIRS), and this definition has recently been standardized. Although, the close association among infection, SIRS, and MODS in patients with ARDS is well established, it is often unclear as to whether these are epiphenomena or are causally related. Our study was designed to prospectively investigate the relationship among infections diagnosed by strict criteria, clinical variables, biological markers of SIRS, and patient outcome.
Several of our findings are in agreement with previous reports and underscore the similarity between our patients and those described in previous ARDS studies: (1) infections are a common precipitating cause of ARDS and a frequent complication of MV; (2) NIs are more frequent in nonsurvivors of ARDS; (3) MODS score on admission is higher in nonsurvivors, and this syndrome is the most frequent cause of death; (4) early death ([is less than or equal to] 3 days) is related to the condition precipitating ARDS; and (5) failure to improve lung function (LIS) in the first week of ARDS is associated with a poor outcome. We provide additional evidence indicating that clinical criteria for recognizing infection are nonspecific. Throughout the course of ARDS, we found no differences among individual SIRS components or composite score (Fig 4, bottom) in patients with or without infection, yet irrespective of sepsis, fever and leukocytosis were frequent findings. This lack of specificity is probably attributable to the underlying inflammatory response of ARDS and should be taken into account when evaluating ARDS patients with possible sepsis. Because clinical signs of infection are unreliable prognosticators in critically ill patients, it is essential that clinical investigations define sepsis by strict diagnostic criteria and that a comprehensive protocol be used systematically to identify all possible sites of infection.
In analyzing the relationships among infection, SIRS, and patient outcome, we found the following: (1) patients with sepsis-induced ARDS had higher initial IC levels with greater early mortality; (2) degree and duration of the host inflammatory response, as measured by plasma IL-1[Beta] levels during the first week of ARDS, outperformed all clinical and physiologic parameters in predicting outcome for individual patients; (3) initially and over time, there was weak correlation between the magnitude of clinical inflammation (SIRS composite score) and the magnitude of biological inflammation (plasma IC levels); (4) NI did not amplify clinical nor laboratory variables of the host systemic inflammatory response; (5) during ARDS, nosocomial pneumonia was not associated with a localized increase in the pulmonary inflammatory response; and (6) containment of the inflammatory response with GCRT improved lung function and patient outcome despite frequent development of potentially lethal NI. Our findings indicate that the degree of initial host inflammatory response, rather than any infectious or noninfectious precipitating etiology, is the determinant of ARDS progression. Although NIs (correctly diagnosed and treated) are a frequent complication (60%), they do not amplify SIRS and may not be directly responsible for patient demise. Selected aspects of our investigation are reviewed in the context of recent medical literature.
Host Inflammatory Response and Outcome in ARDS
In this investigation, we evaluated, over time, clinical variables and biological markers of SIRS. To our knowledge, this is the first study longitudinally evaluating the relationships among SIRS score, IC levels, and intercurrent infections during ARDS. We found that daily individual SIRS components and composite score (Fig 4, top) and average and maximum SIRS score over time (Fig 5, top) were similar in both survivors and nonsurvivors. However, among patients with identical average SIRS score over time, mortality was higher in those with infection and reached 80% when bacteremia was present (Fig 5, bottom).
Systemic cytokine release is a common pathway underlying SIRS and MODS. On day 1 of ARDS and over time, nonsurvivors had significantly (p [is less than] 0.0001) higher plasma TNF-[Alpha], IL-1[Beta], IL-6, and IL-8 levels (Fig 6). Furthermore, among nonsurvivors, plasma IC levels were significantly higher in patients dying within 3 days of ARDS onset. In a previous publication from our group, a plasma IL-1[Beta] level [is greater than or equal to] 400 pg/mL was found to be a consistent predictor of patient outcome over time. In the study reported herein, we found that plasma IL-1[Beta] levels declined in all survivors during the first week of ARDS but remained persistently elevated (within 10% of the initial day 1 value) in all except two of our nonsurvivors (88%), both of whom had plasma IL-1[Beta] levels [is greater than] 400 pg/mL on day 7. These observations support the view that an overaggressive and protracted host inflammatory response, rather than the condition precipitating respiratory failure, is the major influence on patient outcome in ARDS.
Similar conclusions were reported previously by groups investigating MODS in surgical patients suffering either major trauma or intra-abdominal sepsis.[25-30] In those studies, patient survival was related to the magnitude of the host response as measured by a sepsis score and/or MODS score and was independent of the presence, bacteriologic characteristics, or treatment of the source of infection.[25,26,28-30]
Infections and Host Systemic Inflammatory Response
Patients with sepsis-induced ARDS had a higher early mortality (25% vs 10%), but later in the course ([is greater than] 3 days), patient outcome was similar to those without sepsis-induced ARDS. In agreement with prior studies,[8,10] severe CAP was the most common precipitating cause of ARDS, and S pneumoniae was the most frequently isolated microorganism (30%), with solitary isolates of Gram-negative pathogens accounting for another 30% of cases. Irrespective of time, there was no difference in individual SIRS components or composite score among patients with or without infection. Both fever and leukocytosis were actually more common in patients without sepsis-induced ARDS.
Table 5–Estimation of Agreement in Levels of Cytokines, Albumin, Total Protein, and Neutrophils Between the Right and Left BAL Using Intraclass Correlation Coefficients in Patients With Pneumonia(*)
Unilateral Pneumonia Bilateral
(n = 9) (n = 7) (n=5)
Group 1 Group 2 Groups 1 and 2
TNF-[Alpha] 0.99 0.70 0.95
IL-1[Beta] 0.99 0.75 0.99
IL-2 0.93 0.79 0.98
IL-4 0.96 0.85 0.95
IL-6 0.97 0.85 0.99
IL-8 0.97 0.92 0.99
Albumin 0.99 0.99 0.99
TP 0.91 0.99 0.99
Total PMN count 0.69 0.11 0.00
PMN% 0.42 0.17 0.60
(*) TP = total protein; PMN = neutrophils.
Table 6–Relationship Between Specific NIs and Outcome for Group 1 Patients With NI
Sepsis on admission 2
Pneumonia 1 (1)([double dagger])
Sepsis on admission 5([dagger])
Pneumonia 5 (4)([double dagger])
(*) One patient died with refractory hypoxia.
([dagger]) All patients died with MODS.
([double dagger]) Episodes of bacteremia associated with infection in parentheses.
Also irrespective of time, patients with sepsis-induced ARDS had significantly (p400 pg/mL at onset of ARDS (67% vs 29%), a persistent cytokine elevation (67% vs 29%), a longer duration of MV (24 vs 8 days; p=0.0003), and were more likely to die (67% vs 29%). No differences in clinical, physiologic, or plasma IC levels were found between patients with Gram-positive or Gram-negative pathogens; this observation is in agreement with prior reports.[27-29,31-33] In nonsurvivors, the infections were frequently at multiple sites (86%) and caused by different pathogens (57%).
Our observations confirm earlier work reporting the frequent association of NI, with the development and progression of MODS, and decreased survival.[5,9,10] Faist and coworkers proposed a two-hit hypothesis of MODS in which NI represents a second insult to a previously injured and primed host, converting a low-grade or stable host response (stable SIRS) into an accelerated or severe host response (accelerated SIRS), triggering the development and progression of MODS. Contrary to this model, our findings indicate that NIs are intervening variables relative to mortality and may not be the direct cause of patient death. We found that none of the proved (n=36) or suspected (n=55) NIs caused transient or sustained amplification of either clinical or biochemical parameters of systemic inflammation (Figs 2 and 3). The inflammatory response in ARDS appears to be autonomous, and the immunologic response to secondary infections is either blunted or downregulated. We documented a lack of increase in all measured IC levels with NI at both the pulmonary and systemic levels. We cannot exclude that undefined mechanisms operating independently of the host inflammatory response might impact on the outcome of patients with NI.
Recent experimental and clinical studies (Table 8) have shown that production. of IC in response to a bacterial challenge is suppressed by prior activation of the host inflammatory response. In agreement with the findings of animal studies,[34,35] human studies using circulating monocytes and alveolar macrophages of patients with sepsis have shown downregulation of IC release in response to lipopolysaccharide (LPS).[33-38] Among these patients with sepsis, decreased production of IC in response to LPS persisted for up to 10 days.[36,37] Potential mediators of IC downregulation include prostaglandin [E.sub.2], glucocorticoids, IL-4, IL-10, transforming growth factor-[Beta], and the induction of heat-shock proteins. Because we found that the degree and the duration of the host inflammatory response is the major determinant of patient outcome, downregulation in the presence of NI most likely represents cytoprotection from overwhelming cytokinemia and not a detrimental form of immunosuppression. Postmortem examinations of ARDS patients with nosocomial pneumonia have shown no histologic manifestation of immunosuppression. Furthermore, 67% of the NI in our study occurred after the 10th day of ARDS; the ability of plasma IL-1[Beta] levels during the first week of ARDS to accurately predict outcome (sensitivity, 100%; specificity, 100%) prior to the development of most NI strongly argues against NI playing a significant role in the decreased survival of patients with ARDS.
Table 8–Human Ex Vivo Studies Demonstrating the Phenomenon of
Downregulation of IC Production in Subjects
With Experimental Sepsis
Donor(*) Cell([dagger]) Stimulate
Cytokine Population Type ([double dagger])
TNF-[Alpha] 1,2 WB LPS
IL-1[Beta] 1,2 WB LPS
IL-2 2,3 PBM PHA
IL-6 1,2 WB LPS
IL-8 1,2 WB LPS
Cytokine Production Study
TNF-[Alpha] [down arrow] Septic>control Ertel et al[36,37]
Simpson et al
IL-1[Beta] [down arrow] Septic>control Ertel et al[36,37]
Lugar et al
IL-2 [down arrow] Burn>control Grbic et al
IL-6 [down arrow] Septic>control Ertel et al[36,37]
IL-8 [down arrow] Septic >control Ertel et al
(*) Donor Population: 1=patients with sepsis;
2=healthy control subjects; 3=burn patients.
([dagger]) Cell type: WB=whole blood; AM=alveolar macrophage;
PBM=peripheral blood mononuclear cells.
([double dagger]) Stimulate: PHA = phytohemagglutinin.
 Calandra T, Baumgartner JD, Grau GE, et al. Prognostic values of tumor necrosis factor/cachectin, interleukin-1, interferon-[Alpha], and interferon-[Gamma] in the serum of patients with septic shock. J Infect Dis 1990; 161:982-87
 Pinski MR, Vincent JL, Deviere J, et al. Serum cytokine levels in human septic shock: relation to multiple-system organ failure and mortality. Chest 1993; 103:565-75
 Nelson S, Mason C, Bagby GJ, et al. Lipopolysaccharide-induced inhibition of intrapulmonary tumor necrosis factor and lung antibacterial defenses [abstract]. Am Rev Respir Dis 1990; 141:A512
 Wilson JD Jr, Stewart RM, Fabian TC, et al. Plasma tumor necrosis factor and post-traumatic hyperdynamic sepsis evoked by endotoxin. Shock 1994; 1:176-83
 Ertel W, Kremer JP, Kenney J, et al. Downregulation of proinflammatory cytokine release in whole blood from septic patients. Blood 1995; 85:1341-47
 Ertel W, Jarrar D, Jochum M, et al. Enhanced release of elastase is not concomitant with increased secretion of gran ulocyte-activating cytokines in whole blood from patients with sepsis. Arch Surg 1994; 129:90-8
 Simpson SQ, Modi HN, Balk RA, et al. Reduced alveolar macrophage production of tumor necrosis factor during sepsis in mice and men. Crit Care Med 1991; 19:1060-66
 Ribeiro SP, Villar J, Slutsky AS. Induction of the stress response to prevent organ injury. New Horizons 1995; 3:301-11
 Rouby JJ, deLassale EM, Poete P, et al. Nosocomial bronchopneumonia in the critically ill: histologic and bacteriologic aspects. Am Rev Respir Dis 1992; 146:1059-66
 Lugar H, Graf H, Schwarz HP, et al. The serum interleukin 1 activity and monocyte interleukin 1 production in patients with fatal sepsis. Crit Care Med 1986; 14:458-61
 Grbic JT, Mannick JA, Gough DB, et al. The role of prostaglandin [E.sub.2] in immune suppression following injury. Ann Surg 1991; 214:253-63
(*) From the Department of Medicine, Pulmonary and Critical Care Division (Drs. Headley and Meduri), and the Department of Preventive Medicine, Division of Biostatistics and Epidemiology (Dr. Tolley), University of Tennessee Medical Center, Baptist Memorial Hospital Medical Center, Regional Medical Center, and Veteran Affairs Medical Center, Memphis. This study was supported by Clinical Research Center grant 5M01RR00211 and Baptist Health Care Foundation grant 9506. This study won a 1994 DuPont Pharmaceuticals American College of Chest Physicians Young Investigator Award (Dr. Headley).
Manuscript received August 27, 1996; revision accepted November 22.
Reprint requests: Dr. Headley, The University of Tennessee, Memphis, 956 Court Avenue, Room H314, Memphis, TN 38163
COPYRIGHT 1997 American College of Chest Physicians
COPYRIGHT 2004 Gale Group