Abstract
Background
Death following trauma, infection, or other critical illness has been attributed to unbalanced inflammation, where dysregulation of cytokines leads to multiple organ dysfunction and death. We hypothesized that admission cytokine profiles associated with death would differ based on admitting diagnosis.
Study Design
This five-year study included patients admitted for trauma or surgical intensive care for more than 48 hours at two academic, tertiary care hospitals between 10/01 and 05/06. Cytokine analysis for IL-1, 2, 4, 6, 8, 10, 12, IFN-gamma, and TNF alpha was performed using ELISA on specimens drawn within 72 hours of admission. Mann-Whitney U test was used to compare median admission cytokines levels between alive and deceased patients. Relative risks and odds of death associated with admission cytokines were generated using univariate analysis and multivariate logistic regression models, respectively.
Results
1655 patients had complete cytokine data: 290 infected, non-trauma, 343 non-infected, non-trauma, and 1022 trauma. Among infected patients, non-survivors had higher median admission levels of IL-2, -8, -10 and GMCSF; non-infected, non-trauma patients IL-6, -8 and IL-10; and non-surviving trauma patients had higher IL-4, -6, -8 and TNF-α. Interleukin-4 was the most significant predictor of death and carried the highest relative risk of dying in trauma patients, and IL-8 in non-trauma, non-infected patients. In infected patients, no cytokine independently predicted death.
Conclusions
Cytokine profiles of certain disease states may identify persons at risk of dying and allow for selective targeting of multiple cytokines to prevent organ dysfunction and death.
Cytokine release is a normal, highly complex and tightly modulated response to traumatic insult or infection, capable of producing different effects depending on the body's regional composition.1 Infection or trauma induce an immediate, system-wide pro-inflammatory release, unleashing cytokines that help to recruit neutrophils, B cells and T cells, platelets and coagulation factors to the site of damage.1-3 This leads to destruction of the already wounded tissue, healthy tissue growth promotion and an attempt at eradication of pathogenic microorganisms or foreign antigens.1-3 Compensatory anti-inflammatory responses soon follow and are believed to exist to attenuate the pro-inflammatory state by decreasing expression of monocytic major histocompatibility complex (MHC) class II, imparing antigen presenting activity and reducing cells' ability to produce pro-inflammatory cytokines.4-9
Disequilibrium between pro- and anti-inflammatory cytokines is now believed to initiate a physiologic state that can ultimately lead to patient demise. The disequilibrium may cause the Systemic Inflammatory Response Syndrome (SIRS) – a generalized, cytokine response in organs distant from the original site of injury or infection.1 SIRS is characterized by progressive microvascular permeability,10-14 organ ischemia due to microcirculation plugging,15 activation of the coagulation system,16 and vasodilation with fluid transudation and global tissue hypoxia.17-19 SIRS can in turn progress to a multiple organ dysfunction syndrome (MODS), which may cause death in up to 30% of critically ill patients unless cytokine homeostasis is restored.20
Although great clinical advances have been made regarding mortality among injured or septic patients through early goal-directed therapy and resuscitation, the frequency of organ dysfunction and its ensuing death have remained largely unchanged.21,25 There is an increased recognition of the role of cytokines in inflammatory dysregulation, with recent evidence suggesting that elevated cytokine levels correlate with poor patient outcomes.23,25 Furthermore, the highest cytokine concentrations have been obtained early in the posttraumatic and infectious periods.39 Nonetheless, specific cytokine patterns, truly predictive of outcomes, are yet to be established; with further uncertainty stemming from the fact that the release of systemic cytokines can occur in a variety of diseases without leading to organ dysfunction.1,22 For the purpose of this study, we hypothesized that admission cytokine levels and patterns would predict mortality in patients admitted for intensive care and would differ based on admission diagnosis.
Methods
Study design/Data origin
Patients, 18 years of age or older, admitted to the surgical or trauma intensive care units (ICU) of two institutions - Vanderbilt University Medical Center and the University of Virginia Medical Center – were enrolled in this prospective, multicenter cohort study from October 2001 to May 2006. The study was approved by the local Institutional Review Board at each medical center, with the University of Virginia having approval for waiver of consent, while Vanderbilt required an assent from a surrogate prior to data collection and informed consent from the patient in case of resolution of the critical illness.
Critically injured or infected patients with a minimum of 48-hour ICU stay were eligible for enrollment in this study. This time point was created to ensure that patients with overwhelming events (e.g., non-survivable hemorrhage, traumatic brain injury, etc.) and those with less severe illnesses or at lower risk for mortality would be excluded from enrollment. Individuals primarily admitted for burn injuries and/or without complete cytokine data were not included in this report. Subjects were divided into three separate groups based on the reason for ICU admission: (1) trauma patients, (2) non-trauma patients admitted for the treatment of infection and (3) non-infected, non-trauma patients.
Data collection/Variables
Full-time research nurses collected clinical data at each facility by interviewing patients, their families and healthcare providers, or reviewing electronic and paper medical records. Baseline demographics as well as date of hospital/ICU admission and discharge, in-hospital death, co-morbidities (diabetes mellitus, hypertension, cardiovascular disease, malignancy etc), and in-hospital complications were entered into a secure, password-protected computer database. Data including Acute Physiology and Chronic Health Evaluation (APACHE) II score,26 the Marshall or Multiple Organ Dysfunction score (MODS),27 and the McCabe (underlying disease) score28 were calculated for all patients at the time of ICU admission. The Injury Severity Score (ISS)29 and the Trauma-Score Injury Severity Score (TRISS),30 evaluating the probability of survival in injured patients, were calculated for trauma patients at the time of admission. Glasgow Coma Scale (GCS) score was calculated for all patients. For patients who were iatrogenically sedated, the GCS score was estimated as if the patient was not sedated, and that value was recorded and used in the calculation of APACHE II score and MODS. Mortality included deaths from any cause occurring prior to hospital discharge.
Infections
Definitions for infections with the exception of catheter related infections followed the United States Centers for Disease Control and Prevention guidelines.31 Diagnosis of pneumonia was based on systemic evidence of infection, production of sputum, isolation of a predominant organism, development of a new or changing infiltrate or effusion on chest radiograph or growth of >100,000 colony-forming units (CFU) on quantitative culture via endotracheal aspiration or >10,000 CFU via broncheoalveolar lavage. Bloodstream infections were diagnosed by isolating organisms from any single blood culture using aseptic technique, except for coagulase-negative Staphylococcus (i.e. Staphylococcus Epidermidis). Diagnosis of coagulase-negative Staphylococcus blood stream infections required growth from two separate positive blood cultures. Urinary tract infections required growth of either >100,000 organisms/ml of urine or >10,000 organisms/ml with symptoms. Catheter-related infections were defined as isolation of >15 colonies with the semi-quantitative roll plate technique in the setting of clinical infection with possible but not necessary positive blood culture with the same organism. Catheter tips were cultured only when removed from patients with a temperature of >38.5°C or a persistently rising or elevated white blood cell count.32 For infections of the skin, soft tissue, wound and peritoneal infections, culture-positive evidence was helpful, but not required, since most of these infections were diagnosed clinically.
Cytokine assay
At the time of study entry (within 72 hours of admission), a 10-ml blood sample was collected from each patient for assessment of cytokine levels including pro-inflammatory cytokines (IL-1, -2,-6, -12, interferon –γ (IFN), tumor necrosis factor-α (TNFα), granulocyte-macrophage colony-stimulating factor (GM-CSF)), anti-inflammatory cytokines (IL-4 and -10), the chemokine IL-8 and the miscellaneous cytokine IL-5. The blood sample was centrifuged and plasma separated and stored at -70°C prior to evaluation. To determine cytokine concentrations, plasma was assayed with an ELISA-based technology, the Luminex100 system® (Miraibio, Inc., Alameda, CA), and all data recorded into our database. Minimal level of cytokine detectability was defined as 2.7 pg/mL.
Statistical analysis
Univariate analysis
Descriptive analyses comparing demographic data between survivors and non-survivors for the three patient groups (trauma, infected non-trauma, and non-infected, non-trauma) were performed using univariate analysis with a statistical significance set at <0.05. Continuous variables are presented as a mean ± 95% confidence interval (CI) and compared using a two-sample t test for independent samples. Categorical variables were analyzed using a Chi-square or Fisher's exact test. Cytokine levels had a marked rightward skew and are also presented as a median and interquartile range (IQR) with p-value generated through a group comparison using the Wilcoxon rank sum test. Cytokine concentrations were subsequently divided into quartiles with all further comparisons carried out between the highest cytokine concentrations (top quartile) and all other quartiles combined. Relative risk (RR) of death imposed by or associated with each cytokine, found on univariate analysis to be significant, was then calculated. To describe the relationship between the cytokines that were found to be statistically significant on univariate analysis and the severity of illness (TRISS in trauma patients and APACHE II score in non-trauma patients), Spearman correlation coefficient was calculated.
Multivariate analysis
Cytokines that were statistically significantly different by univariate analysis were used to build three multivariable logistic regression models for trauma, infected non-trauma, and non-infected non-trauma patients. In order to predict the odds of death with respect to cytokine levels, admission cytokine concentrations were divided into quartiles and entered into the models as categorical variables; the patients with the highest quartile of cytokine expression were compared to those in the lowest three quartiles, ultimately generating odds ratios for death. For each model, maximum adjusted R2 value and c statistics were calculated to ensure that the models performed well and were able to adequately discriminate between survivors and non-survivors, respectively. To confirm the results generated by the logistic regression analysis, our model was internally validated via bootstrapping method (80:20).
Additional variables found in previous publications to be important predictors of mortality in critically ill patients were included in the logistic regression model as follows. The trauma model utilized age, gender, temperature,32 serum creatinine,32 GCS, TRISS, MODS and ISS score in addition to significant cytokines. Only patients with ISS ≥ 16 were included in the model in order to ensure severe injuries requiring surgical critical care.25 In addition to significant cytokines, the model predicting survival in infected patients included age, gender, GCS score, APACHE II score, MODS, McCabe admit score and the model for non-injured non-infected patients included age, gender, GCS, APACHE II score, MODS, McCabe admit score.
In previous publications,33 cytokine concentrations have been log-transformed with models created using a natural log scale. However, our model performance was unaltered using log-transformed cytokine concentrations and thus, results using only the original form of data are included in this paper. Additionally, no imputation techniques were required in our study, since only patients with 100% of cytokine data were included. All statistical analyses were performed using SAS software, version 9.1.3 (SAS Institute, Cary, NC).
Results
Baseline demographics
A total of 2291 patients were enrolled in this study, with 1655 (72.2%) having complete admission cytokine data. One-thousand-twenty-two (61.8%) patients were included in the trauma group and 633 (38.2%) in the non-trauma group. The non-trauma patients were further subdivided into infected (290) and non-infected (343) patients. One-hundred-nine (10.7%) injured patients, 87 (30%) infected, non-trauma patients, and 54 (15.7%) non-infected, non-trauma patients died during the study period.
Population factors associated with survival
Univariate analysis of individual variables within each study group revealed several differences between alive and deceased patients. For all cohorts, patients who survived were younger and had lower APACHE II scores. Alive trauma patients tended to be white, have longer hospital and ICU length of stay, higher GCS score, lower MODS score and increased TRISS score and comorbidities including cardiovascular, cerebrovascular and renal disease (table 1). Infected, non-trauma patients who survived also had longer hospital stay, higher McCabe scores, and a greater incidence of malignancy (table 2). Non-infected, non-trauma patients who lived had higher GCS score and incidence of McCabe score of 3, but lower white blood cell (WBC) count (table 3).
Table 1.
Frequency and Percentage Distribution of Trauma Population Characteristics, by Discharge Status
| Discharged alive | Discharged deceased | p Value | |||
|---|---|---|---|---|---|
| n | %* | n | %* | ||
| n | 913 | 89.3 | 109 | 10.7 | - |
| Age, y† | 41.9 ± 0.6 | - | 51.5 ± 1.9 | - | <0.01 |
| Hospital length of stay† | 23.4 ± 1.0 | - | 13.6 ± 1.4 | - | <0.01 |
| ICU length of stay† | 12.0 ± 0.3 | - | 10.8 ± 1.2 | - | <0.01 |
| White blood cell count† | 16.1 ± 0.3 | - | 16.3 ± 0.8 | - | 0.37 |
| Race/Ethnicity | |||||
| White | 743 | 81.4 | 98 | 89.9 | 0.03 |
| African American | 119 | 13.0 | 8 | 7.3 | 0.09 |
| Hispanic | 41 | 4.5 | 3 | 2.8 | 0.40 |
| Native American | 2 | 0.2 | 0 | 0 | 0.62 |
| Other | 8 | 0.9 | 0 | 0 | 0.44 |
| Comorbidities | |||||
| Coronary artery disease | 98 | 10.7 | 30 | 27.5 | <0.01 |
| Cerebrovascular disease | 21 | 2.3 | 8 | 7.3 | <0.01 |
| Chronic renal disease | 5 | 0.5 | 4 | 3.7 | <0.01 |
| Corticosteroids | 8 | 0.9 | 3 | 2.8 | 0.07 |
| Liver disease | 23 | 2.5 | 9 | 8.3 | 0.62 |
| Malignancy | 17 | 1.9 | 4 | 3.7 | 0.21 |
| Peripheral vascular disease | 16 | 1.8 | 2 | 1.8 | 0.95 |
| Pulmonary disease | 83 | 9.1 | 15 | 13.8 | 0.12 |
| Autoimmune disorders | 10 | 1.1 | 3 | 2.8 | 0.15 |
| Diabetes mellitus | 64 | 7.0 | 10 | 9.2 | 0.41 |
| Other | 7 | 0.8 | 3 | 2.8 | 0.08 |
| Interventions | |||||
| Mechanical ventilation | 845 | 92.6 | 103 | 94.5 | 0.48 |
| Total days intubated† | 9.0 ± 0.3 | - | 8.6 ± 0.6 | - | 0.75 |
| Tracheostomy | 90 | 9.9 | 9 | 8.3 | 0.56 |
| Transfusions | 638 | 69.9 | 79 | 72.5 | 0.72 |
| Scoring† | |||||
| ISS | 28.6 ± 0.4 | - | 29.8 ± 1.4 | - | 0.48 |
| TRISS | 0.70 ± 0.01 | - | 0.54 ± 0.04 | - | <0.01 |
| GCS | 12.9 ± 0.1 | - | 10.1 ± 0.4 | - | <0.01 |
| APACHE II | 15.4 ± 0.2 | - | 21.8 ± 0.6 | - | <0.01 |
| MODS | 7.2 ± 0.1 | - | 8.8 ± 0.3 | - | <0.01 |
Percent based on total trauma population.
Expressed as mean ± Standard error of the mean.
Table 2.
Frequency and Percentage Distribution of Infected Non-Trauma Population Characteristics, by Discharge Status
| Discharged alive | Discharged deceased | p Value | |||
|---|---|---|---|---|---|
| n | %* | n | %* | ||
| n | 203 | 70 | 87 | 30 | - |
| Age, y† | 54.8 ± 1.0 | - | 63.0 ± 1.6 | - | <0.01 |
| Hospital length of stay† | 41.2 ± 2.4 | - | 31.2 ± 3.0 | - | <0.01 |
| ICU length of stay† | 20.1 ± 1.0 | - | 21.3 ± 2.2 | - | 0.90 |
| White blood cell count† | 15.9 ± 0.9 | - | 16.2 ± 1.3 | - | 0.73 |
| Race/Ethnicity | |||||
| White | 176 | 86.7 | 75 | 86.2 | 0.91 |
| African American | 22 | 10.8 | 7 | 8.0 | 0.47 |
| Hispanic | 1 | 0.5 | 2 | 2.3 | 0.16 |
| Other | 4 | 2.0 | 3 | 3.4 | 0.03 |
| No. of infections treated during ICU admission | |||||
| 1 | 124 | 61.1 | 48 | 55.2 | 0.35 |
| 2 | 51 | 25.1 | 15 | 17.2 | 0.14 |
| 3 or more | 16 | 7.9 | 13 | 14.5 | <0.01 |
| Total infections during ICU admission | 310 | - | 150 | - | - |
| Comorbidities | |||||
| Coronary artery disease | 66 | 32.5 | 33 | 37.9 | 0.39 |
| Cerebrovascular disease | 17 | 8.4 | 10 | 11.5 | 0.41 |
| Chronic renal disease | 21 | 10.3 | 7 | 8.0 | 0.54 |
| Corticosteroids | 20 | 9.9 | 14 | 16.1 | 0.13 |
| Liver disease | 29 | 14.3 | 18 | 20.7 | 0.18 |
| Malignancy | 48 | 23.6 | 35 | 40.2 | <0.01 |
| Peripheral vascular disease | 20 | 9.9 | 8 | 9.2 | 0.85 |
| Pulmonary disease | 50 | 24.6 | 20 | 23.0 | 0.75 |
| Autoimmune disorders | 6 | 3.0 | 3 | 3.4 | 0.83 |
| Diabetes mellitus | 68 | 33.5 | 20 | 23.0 | 0.07 |
| Other | 42 | 20.7 | 17 | 19.5 | 0.82 |
| Interventions | |||||
| Mechanical ventilation | 175 | 86.2 | 74 | 85.1 | 0.80 |
| Total days intubated† | 14.9 ± 0.8 | - | 15.9 ± 1.0 | - | 0.35 |
| Tracheostomy | 15 | 7.4 | 5 | 5.7 | 0.60 |
| Transfusions | 142 | 70.0 | 68 | 78.2 | 0.07 |
| Scoring† | |||||
| GCS | 14.2 ± 0.2 | - | 13.8 ± 0.2 | - | 0.08 |
| APACHE II | 18.5 ± 0.5 | - | 22.2 ± 0.7 | - | <0.01 |
| MODS | 8 ± 0.3 | - | 8.3 ± 0.5 | - | 0.60 |
| McCabe score | |||||
| 0 | 37 | 18.2 | 13 | 14.9 | 0.50 |
| 1 | 133 | 65.5 | 45 | 51.7 | 0.03 |
| 2 | 29 | 14.3 | 24 | 27.6 | <0.01 |
| 3 | 3 | 1.5 | 5 | 5.7 | 0.04 |
Percent based on total infected non-trauma population.
Expressed as mean ± Standard error of the mean.
Table 3.
Frequency and Percentage Distribution of Non-Infected Non-Trauma Population Characteristics, by Discharge Status and Overall
| Discharged alive | Discharged deceased | p Value | |||
|---|---|---|---|---|---|
| n | %* | n | n | ||
| n | 289 | 84.3 | 54 | 15.7 | - |
| Age, y† | 59.2 ± 0.9 | - | 65.6 ± 2.22 | - | <0.01 |
| Hospital length of stay, d† | 22.7 ± 1.1 | - | 40.1 ±14.4 | - | 0.84 |
| ICU length of stay, d† | 8.9 ± 0.4 | - | 22.9 ± 13.5 | - | 0.49 |
| White blood cell count† | 13.0 ± 0.6 | - | 17.4 ± 2.0 | - | <0.01 |
| Race/Ethnicity | |||||
| White | 249 | 86.2 | 45 | 83.3 | 0.59 |
| African American | 35 | 12.1 | 9 | 16.7 | 0.36 |
| Other | 5 | 1.7 | 0 | 0 | 0.33 |
| Comorbidities | |||||
| Coronary artery disease | 93 | 32.2 | 24 | 44.4 | 0.08 |
| Cerebrovascular disease | 28 | 9.7 | 5 | 9.3 | 0.92 |
| Chronic renal disease | 34 | 11.8 | 8 | 14.8 | 0.53 |
| Corticosteroids | 18 | 6.2 | 5 | 9.3 | 0.41 |
| Crohn's/Ulcerative colitis | 17 | 5.9 | 0 | 0 | 0.07 |
| Hemodialysis | 13 | 4.5 | 3 | 5.6 | 0.74 |
| Liver disease | 55 | 19.0 | 11 | 20.4 | 0.82 |
| Malignancy | 64 | 22.1 | 17 | 31.5 | 0.14 |
| Peripheral vascular disease | 32 | 11.1 | 4 | 7.4 | 0.42 |
| Pulmonary disease | 79 | 27.3 | 11 | 20.4 | 0.29 |
| Inflammatory disorders | 8 | 2.8 | 1 | 1.9 | 0.70 |
| Diabetes mellitus | 78 | 27.0 | 12 | 22.2 | 0.46 |
| Transplant | 45 | 15.6 | 6 | 11.1 | 0.40 |
| Interventions | |||||
| Mechanical ventilation | 246 | 85.1 | 51 | 94.4 | 0.07 |
| Total days intubated | 7.1 ± 0.5 | - | 8.2 ± 1.2 | - | 0.26 |
| Tracheostomy | 17 | 5.9 | 4 | 7.4 | 0.59 |
| Transfusions | 200 | 69.2 | 40 | 74.1 | 0.24 |
| Scoring† | |||||
| GCS | 14.2 ±0.1 | - | 13.4 ± 0.3 | - | <0.01 |
| APACHE II | 17.6 ± 0.4 | - | 22.2 ± 0.9 | - | <0.01 |
| MODS | 7.6 ± ± 0.3 | - | 7.1 ± 0.8 | - | 0.47 |
| McCabe score | |||||
| 0 | 60 | 20.8 | 9 | 16.7 | 0.49 |
| 1 | 163 | 56.4 | 29 | 53.7 | 0.71 |
| 2 | 52 | 18.0 | 8 | 14.8 | 0.57 |
| 3 | 14 | 4.8 | 7 | 13.0 | 0.02 |
Percent based on total non-infected non-trauma population.
Expressed as mean ± Standard error of the mean.
Mean and median cytokine levels of the three respective study groups were listed in table 4 with medians compared using the Mann-Whitney U test due to a non-Gaussian distribution of values. Significant differences were noted between the following patient categories in terms of admission cytokine levels: IL-1, IFN, GMCSF, TNFα, IL-10, IL-8 and IL-5.
Table 4.
Mean and Median Levels of Pro-Inflammatory, Anti-Inflammatory and Miscellaneous Cytokines and Chemokines, within 48 h of ICU Admission
| p Values | ||||||||
|---|---|---|---|---|---|---|---|---|
| Trauma | Infected, non-trauma | Non-infected, non-trauma | All groups | Infected versus non-infected | Infected versus trauma | Non-infected versus trauma | ||
| Pro-inflammatory cytokines | ||||||||
| IL-1 | 8.8 ± 1.2 2.7 (2.7, 5.2) |
7.3 ± 0.7 2.7 (2.7, 4.4) |
8.5 ± 1.2 2.7 (2.7,2.9) |
0.0314 | 0.3977 | 0.1361 | 0.0111 | |
| IL-2 | 14.3 ± 1.8 3.2 (2.7, 12.7) |
21.7 ± 6.8 3.7 (2.7, 14.5) |
13.3 ± 2.5 2.7 (2.7, 10.9) |
0.1642 | ||||
| IL-6 | 363.2 ± 36.5 144.1 (57.3, 320.6) |
529.0 ± 63.1 170.9 (45.4, 497.3) |
408.2 ± 46.5 120.0 (43.0, 393.7) |
0.2041 | ||||
| IL-12 | 16.5 ± 3.0 2.7 (2.7, 9.8) |
45.2 ± 17.2 2.7 (2.7, 12.8) |
58.0 ± 26.1 2.7 (2.7, 9.1) |
0.102 | ||||
| IFN | 15.0 ± 1.7 3.2 (2.7, 10.7) |
17.8 ± 6.4 2.7 (2.7, 8.3) |
43.9 ± 19.2 2.7 (2.7, 8.4) |
0.0063 | 0.4728 | 0.0314 | 0.0035 | |
| GMCSF | 11.0 ± 1.7 4.4 (2.7, 9.6) |
8.8 ± 1.2 2.9 (2.7,7.9) |
13.4 ± 2.6 2.8 (2.7, 7.3) |
0.0011 | 0.8837 | 0.0041 | 0.0014 | |
| TNFα | 19.6 ± 11.1 4.9 (2.7, 9.2) |
15.7 ± 1.9 9.2 (2.7, 17.6) |
16.9 ± 2.3 8.8 (3.5, 16.2) |
<0.0001 | 0.7887 | <0.0001 | <0.0001 | |
| Anti-inflammatory cytokines | ||||||||
| IL-4 | 243.9 ± 23.8 31.6 (6.1, 236.4) |
221.6 ± 27.2 36.9 (7.1, 223.4) |
215.8 ± 27.6 31.9 (4.4, 174.5) |
0.6352 | ||||
| IL-10 | 187.8 ± 25.3 49.9 (20.4, 123.0) |
352.6 ± 53.1 104.9 (37.8, 275.0) |
339.6 ± 54.0 60.5 (24.5, 225.3) |
<0.0001 | 0.0005 | <0.0001 | 0.0067 | |
| Chemokines | ||||||||
| IL-8 | 87.7 ± 16.3 28.0 (12.6, 60.4) |
102.9 ± 10.7 52.1 (14.8, 121.8) |
140.1 ± 29.4 46.6 (16.3, 100.1) |
<0.0001 | 0.6378 | <0.0001 | <0.0001 | |
| Miscellaneous | ||||||||
| IL-5 | 6.9 ± 0.5 2.7 (2.7, 4.7) |
5.7 ± 0.5 2.7 (2.7, 5.3) |
6.8 ± 0.8 2.7 (2.7,4.0) |
0.088 | ||||
Levels of cytokines are expressed as mean ± SE, median (interquartile range).
Divided according to their respective groups: trauma; infected, non-trauma; and non-infected non-trauma. Because of the non-Gaussian distribution of the data, medians were compared using the Mann-Whitney U test.
Cytokine differences between survivors and non-survivors
Analysis of the entire study population revealed that median cytokine differences between survivors and non-survivors were most significant for IL-4, -6, -8, -10, and TNFα. In trauma patients, IL-4, -6, -8, and TNFα were found to be significantly higher in non-survivors. Median levels of IL-2, -8, -10 and GMCSF in infected, non-trauma patients and IL-6, -8, and -10 in non-infected, non-trauma patients were higher in patients who eventually died (table 5 and 6). A considerable number of patients had cytokine levels that were below the minimal level of detectability for the assay used, ranging from 3.1% of IL-10 to 67.0% of IL-1 assays. The most likely detected cytokines/chemokines were IL-6,-8, and IL-10, while the least likely detected cytokines were IL-1,-5 and IL-12.
Table 5.
Mean and Median Levels of the Highest Concentration of Cytokines (Top Quartile), Compared between the Living and Deceased Patients Using Mann-Whitney U Test
| All pts. | Trauma pts. | Non-trauma patients, infected at ICU admission |
Non-trauma patients, not infected at ICU admission |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Pro- inflammatory cytokines |
Lived (1,405) |
Died (250) | p Value |
Lived (913) | Died (109) | p Value |
Lived (203) |
Died (87) | p Value |
Lived (289) |
Died (54) | p Value |
| IL-1 | 8.4±0.4 2.7 (2.7,5.0) |
10.9±2.4 2.7 (2.7,2.6) |
0.82 | 8.5±0.5 2.7 (2.7,5.5) |
14.9±5.3 2.7 (2.7,6.6) |
0.55 | 7.1±0.8 2.7 (2.7,3.8) |
7.9±1.3 2.7 (2.7,5.2) |
0.48 | 8.7±1.4 2.7 (2.7,2.9) |
7.8±1.9 2.7 (2.7,2.7) |
0.78 |
| IL-2 | 16.4±1.8 3.1 (2.7,12.5) |
19.6±4.5 3.3 (2.7,13.3) |
0.67 | 15.7±1.6 3.2 (2.7,13.4) |
25.4±9.8 3.1 (2.7,10.2) |
0.56 | 24.5±9.6 2.7 (2.7,13.8) |
15.4±2.3 5.9 (2.7,18.7) |
0.04 | 13.1±2.9 2.8 (2.7,10.9) |
14.6±4.7 2.7 (2.7,8.4) |
0.25 |
| IL-6 | 374.3±21.5 137.4 (49.7,359.6) |
702.3±83.8 218 (71.2,629.5) |
<0.01 | 355.1±25.3 144.4 (54.9,349.8) |
769.5±139.5 192.7 (99.4,668.4) |
<0.01 | 522.1±78.6 154.6 (40.1,424.4) |
545.1±104 215.1 (65.2,602.0) |
0.10 | 331.3±37.8 113.3 (43.0,301.4) |
820.3±207.4 261.9 (46.3,670.8) |
0.02 |
| IL-12 | 32.2±7.4 2.7 (2.7,10.3) |
13.9±1.9 2.7 (2.7,10.7) |
0.53 | 15.8±2.0 2.7 (2.7,9.9) |
12.6±2.8 2.7 (2.7,18.7) |
0.98 | 56.9±24.5 2.7 (2.7,13.1) |
17.8±4 3.4 (2.7,12.8) |
0.37 | 66.8±31.0 2.7 (2.7,9.2) |
10.5±2.25 2.7 (2.7,7.5) |
0.71 |
| IFN | 19.2±2.6 2.7 (2.7,10.1) |
53.1±25.3 2.7 (2.7,12.0) |
0.70 | 19.5±2.8 3.4 (2.7,11.0) |
20.9±5.5 2.7 (2.7,13.2) |
0.90 | 11.5±3.8 2.7 (2.7,7.3) |
32.4±19.2 2.8 (2.7,9.5) |
0.15 | 23.9±9.0 2.7 (2.7,8.1) |
151.4±112.1 2.7 (2.7,11.5) |
0.54 |
| TNFα | 15.2±4.2 5.8 (2.7,11.6) |
18.1±3.1 8.8 (3.0,16.2) |
<0.01 | 15.8±6.5 4.9 (2.7,9.5) |
10.1±0.9 6.1 (3.0,13.3) |
0.01 | 13.4±1.4 8.7 (2.7,16.6) |
21.0±5.3 11.5 (2.8,21.8) |
0.12 | 14.6±1.7 8.7 (3.5,16.1) |
29.4±11.4 10.3 (4.2,23.1) |
0.31 |
| GMCSF | 11.1±1.0 3.3 (2.7,8.6) |
14.8±2.8 4.2 (2.7,9.8) |
0.20 | 12.0±1.3 4.2 (2.7,9.6) |
10.8±1.9 4.9 (2.7,11.7) |
0.34 | 7.5±1.2 2.7 (2.7,7.1) |
11.9±3 4.4 (2.7,9.5) |
0.01 | 10.8±2.2 2.8 (2.7,7.3) |
27.6±11.1 2.7 (2.7,7.6) |
0.85 |
Table 6.
Mean and Median Levels of the Highest Concentration of Anti-Inflammatory Cytokines, Chemokines, and Miscellaneous (Top Quartile), Compared between the Living and Deceased Patients Using Mann-Whitney U Test
| Anti-inflammatory cytokines | Chemokines | Miscellaneous | ||
|---|---|---|---|---|
| IL-4 | IL-10 | IL-8 | IL-5 | |
| All patients | ||||
| Lived (1405) | 215.8±14.0 30.9 (5.8,175.8) |
240.7±19.9 54.4 (22.7,146.8) |
83.0±7.7 34.3 (13.6,73.0) |
7.1±0.5 2.7 (2.7,4.6) |
| Died (250) | 324.0±38.3 86.7 (9.8,355.2) |
279.2±44.6 102.4 (36.7,240.0) |
187.0±37.6 61 (15.0,143.2) |
6.5±0.7 2.7 (2.7,5.3) |
| p Value | <0.01 | <0.01 | <0.01 | 0.38 |
| Trauma patients | ||||
| Lived (913) | 228.7±18.7 31.1 (5.9,177.1) |
183.3±18.3 51.9 (21.3,126.6) |
77.9±10.0 30.0 (12.8,64.0) |
7.6±0.7 2.7 (2.7,4.7) |
| Died (109) | 340.4±55.1 108.1 (17.3,424.0) |
200.5±59.3 65.3 (24.9,143.3) |
132.1±30.2 42.5 (16.1,102.9) |
6.4±0.9 2.7 (2.7,5.9) |
| p Value | <0.01 | =0.24 | <0.01 | 0.35 |
| Non-trauma patients, infected at ICU admission | ||||
| Lived (203) | 193.6±29.6 30.5 (6.7,194.2) |
35.7±68.6 75.2 (30.0,264.6) |
85.7±9.1 44.0 (14.4,99.9) |
5.4±0.5 2.7 (2.7,4.7) |
| Died (87) | 287.0±58.7 64.2 (8.8,249.9) |
341.1±76.4 167.8 (64.2,343.1) |
143.0±28.2 72.9 (14.8,169.8) |
6.8±0.8 2.7 (2.7,5.9) |
| p Value | 0.20 | <0.01 | 0.04 | 0.31 |
| Non-trauma patients, not infected at ICU admission | ||||
| Lived (289) | 190.6±26.5 29.1 (4.6,155.7) |
339.0±60.5 55.4 (22.3,201.8) |
97.3±18.7 43.8 (16.6,93.3) |
6.8±0.8 2.7 (2.7,4.0) |
| Died (54) | 350.0±102 67.9 (4.2,281.6) |
338.0±114.3 148.2 (31.7,320.6) |
368.8±155.2 84.0 (11.4,207.3) |
7.2±2.5 2.7 (2.7,2.7) |
| p Value | 0.16 | 0.03 | 0.02 | 0.45 |
Relative risk associated with elevated cytokine concentrations
To further assess the importance of elevated admission cytokine concentrations, the relative risk of death was estimated by comparing patients with the highest cytokine concentrations (top quartile) to the other three quartiles within their respective study groups. Severely injured patients were found to have the highest relative risk of dying with the highest concentrations of IL-4 (relative risk = 2.18, p<0.01). Interleukin-6, -8 and TNFα, individually also carried an increased risk of death in trauma patients as follows: 60%, 73% and 48%, respectively. IL-8 was associated with a 56% increased risk of death in infected, non-trauma patients. An increased risk of death in non-infected, non-trauma patients was associated with the highest levels of IL-8 with a relative risk of 2.22 (p<0.01). These results are shown in table 7.
Table 7.
Relative Risk of Death Estimated by Comparing the Highest Concentration of Cytokines (Top Quartile) to All Other Quartiles Combined
| Patients | Cytokines | |||
|---|---|---|---|---|
| Trauma patients | IL4 | IL6 | IL8 | TNF |
| Relative risk | 2.18 | 1.60 | 1.73 | 1.48 |
| 95% CI | (1.535, 3.109) | (1.109, 2.313) | (1.207, 2.494) | (1.016, 2.143) |
| p Value | <0.01 | 0.01 | <0.01 | 0.04 |
| Infected, non-trauma patients | IL2 | IL8 | IL10 | GMCSF |
| Relative risk | 1.20 | 1.56 | 1.27 | 1.34 |
| 95% CI | (0.819, 1.755) | (1.098, 2.228) | (0.871, 1.843) | (0.925, 1.934) |
| p Value | 0.36 | 0.02 | 0.23 | 0.13 |
| Non-infected, non-trauma patients | IL6 | IL8 | IL10 | |
| Relative risk | 1.62 | 2.22 | 1.49 | |
| 95% CI | (0.982, 2.681) | (1.371, 3.587) | (0.897, 2.489) | |
| p Value | 0.06 | <0.01 | 0.13 | |
Correlation between cytokines and severity of illness
In trauma patients, IL-4 and IL-8 significantly and inversely correlated with TRISS (σIL-4= -0.12, p<0.01; σIL-8= -0.09, p=0.02). APACHE II score correlated significantly with IL-8 and IL-10 in infected (σIL-8= 0.19, p<0.01; σIL-10= 0.20, p<0.01), and non-infected non-trauma (σIL-8= 0.13, p=0.02; σIL-10= 0.17, p<0.01), patients.
Multivariate models
A multivariable logistic regression model was estimated to predict patient death as a function of cytokine concentrations identified by univariate analysis to differ significantly between survivors and non-survivors. All cytokine data were adjusted for the previously established risk factors (see methods section). Results of the multivariate logistic regression model for cytokines only are shown in table 8, while table 9 displays the most statistically significant independent predictors of mortality.
Table 8.
Adjusted Odds of Death with their Respective 95% Confidence Intervals for Patients with Admission Cytokine Levels in the Top Quartile as Generated by the Logistic Regression Model
| Cytokines | Odds ratio | 95% CI |
|---|---|---|
| Trauma | ||
| IL4 | 2.538 | 1.204 - 5.351 |
| IL6 | 2.333 | 0.898 - 6.059 |
| IL8 | 0.989 | 0.348 - 2.809 |
| TNF | 0.943 | 0.391 - 2.274 |
| Infected non-trauma | ||
| IL2 | 1.313 | 0.652 - 2.646 |
| IL8 | 1.594 | 0.783 - 3.242 |
| IL10 | 1.238 | 0.615 - 2.494 |
| GMCSF | 1.704 | 0.867 - 3.349 |
| Non-infected non-trauma | ||
| IL6 | 1.197 | 0.503 - 2.853 |
| IL8 | 2.543 | 1.049 - 6.165 |
| IL10 | 0.836 | 0.352 - 1.990 |
Table 9.
Statistically Significant Independent Predictors of Mortality Arranged by their Wald Chi-Square Values
| Effects | Odds ratio | 95% CI | Wald Chi-square | p Value |
|---|---|---|---|---|
| Trauma | ||||
| GCS score | 0.642 | 0.569 - 0.725 | 51.6 | <0.0001 |
| Age | 1.032 | 1.010 - 1.054 | 8.10 | 0.0044 |
| Creatinine | 2.483 | 1.233 - 4.999 | 6.49 | 0.0108 |
| IL4 | 2.538 | 1.204 - 5.351 | 5.99 | 0.0144 |
| ISS | 2.643 | 1.126 - 2.643 | 4.99 | 0.0256 |
| Infected non-trauma | ||||
| Age | 1.035 | 1.012 - 1.059 | 9.18 | 0.0025 |
| McCabe (1 versus 0) | 2.430 | 1.230 - 4.801 | 6.54 | 0.0106 |
| APACHE II score | 1.098 | 1.004 - 1.117 | 4.45 | 0.0350 |
| Non-infected non-trauma | ||||
| APACHE II score | 1.098 | 1.029 - 1.171 | 7.93 | 0.0049 |
| Age | 1.035 | 1.008 - 1.064 | 6.24 | 0.0125 |
| McCabe score (3 vs 0) | 5.380 | 1.133 - 25.554 | 6.20 | 0.0128 |
| GCS | 0.843 | 0.717 - 0.991 | 4.28 | 0.0385 |
| IL8 | 2.543 | 1.049 - 6.165 | 4.27 | 0.0389 |
Cytokine values are calculated based on the comparison of the highest quartile concentration of admission cytokines to all other quartiles combined. Also shown, adjusted odds ratios for each independent mortality predictor with their respective 95% confidence intervals (95% CI) as generated by the multivariate logistic regression analysis.
In trauma patients, cytokine results were adjusted for presenting temperature, GCS, TRISS, age, creatinine, MODS, ISS, and gender. Only patients with ISS≥16 were included, a total of 811 patients. IL-4 was shown to be an independent predictor of mortality with a 2.5 times adjusted increased odds of death (95%CI 1.2 – 5.4). The logistic regression model was an accurate predictor of outcome in trauma patients with a c-statistic of 0.88, demonstrating an adequate discrimination between survivors and decedents, and had a maximum adjusted R2 value of 0.40, designating the model as a good performer. The most predictive information for the model was provided by the GCS score, age and creatinine, followed closely by IL-4 and ISS. Predicting death in infected, non-trauma patients was more difficult with a c-statistic of 0.74 and maximum adjusted R2 value of 0.20. No cytokines were found to be independent predictors of death, but age, McCabe score and APACHE II score provided the most predictive information. Admission IL-8 concentration was independently associated with a 2.5 fold (95%CI 1.0 – 6.2) increased risk of death when adjusted for GCS score, APACHE II score, age, McCabe score, MODS, gender and cytokines in non-infected, non-trauma patients. The model's c-statistic was 0.81 and maximum adjusted R2 0.26. The most predictive information in the model was provided by the APACHE II score, age, McCabe score, GCS score and IL-8.
Discussion
Although the term Systemic Inflammatory Response Syndrome (SIRS) generically refers to a common pathologic event that ensues as a result of an injury or an infectious process, the body's response is a highly complex and heterogeneous sequence of events.36 It is dependent on a number of factors, including the nature of the inciting event, patient's genetic background, clinical interventions, etc.36 Patient management to date has been limited to antimicrobial administration, surgical or radiologic interventions and a multitude of supportive therapies including fluids, vasoactive drugs and other ICU interventions with the intent of preventing progression of SIRS to MODS and death.20,36 Animal and human experiments have suggested the possibility of modifying the pathophysiologic processes that lead to SIRS by modulating the host inflammatory response. Clinical trials directed at altering the activity of these modulators, however, have been almost uniformly unsuccessful. Perhaps the most important difficulty has been in differentiating actual mediators of inflammation from inactive markers of inflammation. Marshall et al36 described mediators as ‘rational targets for therapeutic intervention,’ while markers merely represent the sequelae of the inflammatory processes. We believe, however, that as patterns of cytokines emerge and allow us to discriminate consistently between survivors and non-survivors (in combination with more sophisticated models), that beneficial mediator immunomodulation may yet be achieved.
In this study, we have analyzed prospectively collected admission cytokine data in patients who were admitted to the intensive care units of two medical centers. Demographic data suggested that the patients had a considerable severity of illness, whether based on scoring systems (mean ISS of 29 for trauma patients and mean APACHE II of 19 for non-trauma patients), or overall mortality (11 to 30% based on the group). The cohort was further used to create models using both clinical and cytokine data that could predict patient death after a severe illness. To our knowledge, this is the largest dataset examining the relationship between admission serum cytokine levels and outcomes among surgical patients. By solely analyzing admission concentrations, our initial hypothesis that the early cytokine expression and the relationship between cytokine expression and death are dependent on the nature of the critical illness leading to ICU admission appears valid. Compared to the other groups, trauma was associated with high levels of GMCSF and IFNγ, and lower levels of TNF- α, IL-8 and IL-10. Infected patients had the highest levels of IL-10. Death following trauma was associated with increased levels of IL-4, IL-6, IL-8, and TNF-α; death following an admission for infection was associated with increased concentrations of IL-2, IL-8, IL-10, and GM-CSF; while death following admissions without trauma or infection were associated with increased expression of IL-6, IL-8, and IL-10.
Although some animal models suggest a relatively stereotypic proinflammatory response to most forms of acute stress, the populations described here demonstrated significant intergroup differences and we are left to speculate on the causes of these findings. Obviously, the nature of the inciting event is different between traumatized patients and patients with other forms of critical illness. For example, traumatically injured patients are exhibiting a host response to pain, injury, and potentially hemorrhage and reperfusion, while infected patients are displaying an immune response to live pathogens. Also, the timing of cytokine assay related to the beginning of the patient's illness is undoubtedly different between trauma and non-trauma patients. The exact time of a traumatic event is generally easy to ascertain and treatment is ideally rapid, while patients with other diseases may only seek treatment after many hours and days of illness.
Finally, the cytokine levels measured even at admission must reflect to some degree differences in early therapies between groups, such as pre-hospital resuscitation for trauma patients or early antibiotics for infected patients. Whatever the reasons for these observations, it is clear that different subgroups of critically ill surgical patients cannot be considered similarly under a generic category of “hyperinflamed.”
Examining the individual cytokines, it appears that the association between elevated IL-6 and IL-8 levels and increasing risk of mortality is fairly uniform across patients. On the other hand, among counter inflammatory mediators, IL-4 seems to be most closely associated with death among trauma patients while IL-10 is expressed more among non-trauma patients who die. Although IL-4 inhibits cytokine production to some degree, it more profoundly acts to polarize lymphocyte toward a less inflammatory Th2 phenotype, more closely associated with B cell activity, immunoglobulin production, and a more chronic response to infection. IL-10, on the other hand, significantly and rapidly downregulates the production and secretion of multiple pro-inflammatory mediators. These differences might imply that traumatized patients expressing IL-4 would be deficient in Th1 response and at risk for death from cell-mediated subacute infections, while non-traumatized patients may be less able to mount an appropriate innate immune response and be more likely to die from acute infections, though data to support this theory will be hard to obtain.
Given previous data generated in animals and smaller human studies, the finding that death was not associated with either a clear predominance of either pro-inflammatory or anti-inflammatory mediators, even after analyzing different etiologies of disease, was somewhat surprising. These results, however, corroborate the work of Kellum et al34 who studied 1886 patients admitted for the treatment of community-acquired pneumonia. They found that the risk of death was highest among patients expressing high levels of both IL-6 and IL-10, and the best survival was among patients with low levels of both cytokines. Discerning which patients would be the best candidates for mediator targeted therapy, therefore, will be more difficult than detecting patients with merely an overly pro- or anti-inflammatory response.
Several weaknesses in our study ought to be highlighted. First, due to the study design we analyzed cytokine levels drawn within 72 hours of admission, and ideally we would have considered values only at the actual time of admission. Although some changes undoubtedly occurred over the first three days of illness, around 40% of the values analyzed actually were from the first day and no notable differences could be discerned between cytokines from day one and those drawn on days two or three after admission (data not shown). Second, although the increased expression of several cytokines were associated with mortality by univariate analysis, only IL-4 for trauma patients and IL-8 for non-infected, non-trauma patients were independent predictors of death in our final models. Part of the explanation for this, however, might lie in the close relationship between severity of illness and cytokine expression. It is possible that the inclusion of severity of illness scores in our predictive models nullified the effects of cytokine levels. Finally, even though the object of the current study was to analyze the effects of cytokine expression on different patient subgroups, many more factors must be important determinants of mediator release. Among the infected patients, for example, there may be a different relationship between cytokines and survival conditioned on whether the infection is Gram negative, Gram positive, fungal, or even viral in origin. Thus, it is clear that even among the subgroups studied that a varied response to mediators will occur.
In conclusion, among surgical patients admitted for critical care, there are significant differences in cytokine levels conditioned on the admitting diagnosis. In addition, for the subgroups studied, the relationship between the different cytokines and survival are also dependent on the admitting diagnosis. These data can be used as an early step to designing trials intended to improve outcomes through the modulation of these important mediators.
Acknowledgments
We are indebted to the patients, their families as well as the following healthcare professionals who supported the trial: Sera L Downing, Michelle Pomphrey, Jeffrey S Young, MD; Stewart M Lowson, MD; Judy L Jenkins, Donna L Hollis
This work was supported by the National Institute of Health grant: R01 AI49989-01 (ClinicalTrials.gov identifier NCT00170560), and the University of Virginia General Clinical Research Center. Clinicaltrials.gov Identifier: NCT00170560.
Footnotes
Disclosure Information: Nothing to disclose.
Presented at Southern Surgical Association 121st Annual Meeting, Hot Springs, VA, December 2009.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Bone RC, Grodzin CJ, Balk RA. Sepsis: a new hypothesis for pathogenesis of the disease process. Chest. 1997;112:235–243. doi: 10.1378/chest.112.1.235. [DOI] [PubMed] [Google Scholar]
- 2.Zuckerman SH, Bendele AM. Regulation of serum tumor necrosis factor in glucocorticoid-sensitive and -resistant rodent endotoxin shock models. Infect Immun. 1989;57:3009–3013. doi: 10.1128/iai.57.10.3009-3013.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bone RC. The pathogenesis of sepsis. Ann Intern Med. 1991;115:457–469. doi: 10.7326/0003-4819-115-6-457. [DOI] [PubMed] [Google Scholar]
- 4.Fukushima R, Alexander JW, Gianotti L, et al. Isolated pulmonary infection acts as a source of systemic tumor necrosis factor. Crit Care Med. 1994;22:114–120. doi: 10.1097/00003246-199401000-00022. [DOI] [PubMed] [Google Scholar]
- 5.Ford HR, Hoffman RA, Wing EJ, et al. Characterization of wound cytokines in the sponge matrix model. Arch Surg. 1989;124:1422–1428. doi: 10.1001/archsurg.1989.01410120068014. [DOI] [PubMed] [Google Scholar]
- 6.Meduri GU, Kohler G, Headley S, et al. Inflammatory cytokines in the BAL of patients with ARDS: persistent elevation over time predicts a poor outcome. Chest. 1995;108:1303–1314. doi: 10.1378/chest.108.5.1303. [DOI] [PubMed] [Google Scholar]
- 7.Sauder DN, Semple J, Truscott D, et al. Stimulation of muscle protein degradation by murine and human epidermal cytokines: relationship to thermal injury. J Invest Dermatol. 1986;87:711–714. doi: 10.1111/1523-1747.ep12456681. [DOI] [PubMed] [Google Scholar]
- 8.Kupper TS, Deitch EA, Baker CC, et al. The human burn wound as a primary source of interleukin-1 activity. Surgery. 1986;100:409–414. [PubMed] [Google Scholar]
- 9.Puren AJ, Feldman C, Savage N, et al. Patterns of cytokine expression in community-acquired pneumonia. Chest. 1995;107:1342–1349. doi: 10.1378/chest.107.5.1342. [DOI] [PubMed] [Google Scholar]
- 10.Tracey KJ, Lowry SF, Ceremi A. Cachectin/TNF-alpha in septic shock and adult respiratory distress syndrome. Am Rev Respir Dis. 1988;138:1377–1379. doi: 10.1164/ajrccm/138.6.1377. editorial. [DOI] [PubMed] [Google Scholar]
- 11.Stephens KE, Ishikaza A, Larrick JW, et al. Tumor necrosis factor causes increased pulmonary permeability and edema: comparison to septic acute lung injury. Am Rev Respir Dis. 1988;137:1364–1370. doi: 10.1164/ajrccm/137.6.1364. [DOI] [PubMed] [Google Scholar]
- 12.Ibbotson GC, Wallace JL. Beneficial effects of prostaglandin E2 in endotoxin shock are unrelated to effects of PAF-acether synthesis. Prostaglandins. 1989;37:237–250. doi: 10.1016/0090-6980(89)90060-9. [DOI] [PubMed] [Google Scholar]
- 13.Lewis RA, Austen KF, Soberman RJ. Leukotrienes and other products of the 5-lipoxygenase pathway: biochemistry and relation to pathobiology in human diseases. N Engl J Med. 1990;323:645–655. doi: 10.1056/NEJM199009063231006. [DOI] [PubMed] [Google Scholar]
- 14.Petrak RA, Balk RA, Bone RC. Prostaglandins, cyclo-oxygenase inhibitors, and thromboxane synthesis inhibitors in the pathogenesis of multiple organ failure. Crit Care Clin. 1989;35:303–314. [PubMed] [Google Scholar]
- 15.Sigurdsson GH, Christenson JT, Bader el-Rakshy M, et al. Intestinal platelet trapping after traumatic and septic shock: an early sign of sepsis and multiorgan failure in critically ill patients? JAMA. 1992;20:458–467. doi: 10.1097/00003246-199204000-00005. [DOI] [PubMed] [Google Scholar]
- 16.Levi M, ten Cate H, van der Poll T, et al. Pathogenesis of disseminated intravascular coagulation in sepsis. JAMA. 1993;270:975–979. [PubMed] [Google Scholar]
- 17.Gomez-Jimenez J, Salgado A, Mourelle M, et al. L-arginine: nitric oxide pathway in endotoxemia and human septic shock. Crit Care Med. 1995;23:253–258. doi: 10.1097/00003246-199502000-00009. [DOI] [PubMed] [Google Scholar]
- 18.Miyauchi T, Tomobe Y, Shiba R, et al. Involvement of endothelin in the regulation of human vascular tonus. Circulation. 1990;81:1874–1880. doi: 10.1161/01.cir.81.6.1874. [DOI] [PubMed] [Google Scholar]
- 19.Astiz ME, Rackow EC, Falk JL, et al. Oxygen delivery and consumption in patients with hyperdynamic septic shock. Crit Care Med. 1987;15:26–28. doi: 10.1097/00003246-198701000-00005. [DOI] [PubMed] [Google Scholar]
- 20.Goris JA, te Boekhorst TPA, Nuytinck JKS, et al. Multipleorgan failure: generalized autodestructive inflammation? Arch Surg. 1985;120:1109–1115. doi: 10.1001/archsurg.1985.01390340007001. [DOI] [PubMed] [Google Scholar]
- 21.Jastrow KM, Gonzales EA, McGuire MF, et al. Early Cytokine Production Risk Stratifies Trauma Patients for Multiple Organ Failure. J Am Coll Surg. 2009;209:320–331. doi: 10.1016/j.jamcollsurg.2009.05.002. [DOI] [PubMed] [Google Scholar]
- 22.Bozza FA, Salluh JI, Japiassu AM, et al. Cytokine profiles as markers of isease severity in sepsis: a multiplex analysis. Crit Care. 2007;11:R49. doi: 10.1186/cc5783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rivers EP, Kruse JA, Jacobsen G, et al. The influence of early hemodynamic optimization on biomarker patterns of severe sepsis and septic shock. Crit Care Med. 2007;35:2016–2024. doi: 10.1097/01.ccm.0000281637.08984.6e. [DOI] [PubMed] [Google Scholar]
- 24.Maier B, Lefering R, Lehnert M, et al. Early versus late onset of multiple organ failure is associated with differing patterns of plasma cytokine biomarker expression and outcome after severe trauma. Shock. 2007;28:668–674. [PubMed] [Google Scholar]
- 25.Frink M, van Griensven M, Kobbe P, et al. IL-6 predicts organ dysfunction and mortality in patients with multiple injuries. Scand J Trauma Resusc Emerg Med. 2009;17:49. doi: 10.1186/1757-7241-17-49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818–829. [PubMed] [Google Scholar]
- 27.Marshall JC, Cook DJ, Christou NV, et al. Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med. 1995;23:1638–1652. doi: 10.1097/00003246-199510000-00007. [DOI] [PubMed] [Google Scholar]
- 28.Gross PA, Stein MR, van Antwerpen C, et al. Comparison of severity of illness indicators in an intensive care unit. Arch Intern Med. 1991;151:2201–2205. [PubMed] [Google Scholar]
- 29.Baker SP, O'Neill B, Haddon W, Jr, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974;14:187–196. [PubMed] [Google Scholar]
- 30.Boyd CR, Tolson MA, Copes WS. Evaluating trauma care: the TRISS method. Trauma Score and the Injury Severity Score. J Trauma. 1987;27:370–378. [PubMed] [Google Scholar]
- 31.Delgado-Rodriguez M, Gómez-Ortega A, Llorca J, et al. Nosocomial infection, indices of intrinsic infection risk, and inhospital mortality in general surgery. J Hosp Infect. 1999;41:203–211. doi: 10.1016/s0195-6701(99)90017-8. [DOI] [PubMed] [Google Scholar]
- 32.Smith RL, Meixler SM, Simberkoff MS. Excess mortality in critically ill patients with nosocomial blood stream infections. Chest. 1991;100:164–167. doi: 10.1378/chest.100.1.164. [DOI] [PubMed] [Google Scholar]
- 33.Dossett LA, Redhage LA, Sawyer RG, May AK. Revisiting the validity of APACHE II in the trauma ICU: improved risk stratification in critically injured adults. Injury. 2009;40:993–998. doi: 10.1016/j.injury.2009.03.004. Epub 2009 Jun 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Kellum JA, Kong L, Fink MP, Weissfeld LA, et al. Understanding the inflammatory cytokine response in pneumonia and sepsis: results of the Genetic and Inflammatory Markers of Sepsis (GenIMS) Study. Arch Intern Med. 2007;167:1655–1663. doi: 10.1001/archinte.167.15.1655. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Macias WL, Nelson DR, Williams M, et al. Lack of evidence for qualitative treatment by disease severity interactions in clinical studies of severe sepsis. Crit Care. 2005;9:R607–622. doi: 10.1186/cc3795. Epub 2005 Sep 22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Marshall JC, Vincent JL, Fink MP, Cook DJ, et al. Measures, markers, and mediators: toward a staging system for clinical sepsis. A report of the Fifth Toronto Sepsis Roundtable, Toronto, Ontario, Canada, October 25-26, 2000. Crit Care Med. 2003;31:1560–1567. doi: 10.1097/01.CCM.0000065186.67848.3A. [DOI] [PubMed] [Google Scholar]
- 37.Bozza FA, Salluh JI, Japiassu AM, et al. Cytokine profiles as markers of disease severity in sepsis: a multiplex analysis. Crit Care. 2007;11:R49. doi: 10.1186/cc5783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Maier B, Lefering R, Lehnert M, et al. Early versus late onset of multiple organ failure is associated with differing patterns of plasma cytokine biomarker expression and outcome after severe trauma. Shock. 2007;28:668–674. [PubMed] [Google Scholar]
- 39.Roumen RM, Hendriks T, van der Ven-Jongekrijg J, et al. Cytokine patterns in patients after major vascular surgery, hemorrhagic shock, and severe blunt trauma. Relation with subsequent adult respiratory distress syndrome and multiple organ failure. Ann Surg. 1993;218:769–776. doi: 10.1097/00000658-199312000-00011. [DOI] [PMC free article] [PubMed] [Google Scholar]