Multiple organ failure (MOF) remains the major cause of late postinjury death and consumes inordinate resources in the intensive care unit (ICU) (1-3). Yet, the epidemiological descriptions of this condition vary widely. Indeed, the reported incidence of postinjury MOF varies from as low as single-digit numbers to values approaching 40%, with case-fatality rates also varying within similarly large ranges (1, 4-8). Studies had suggested that MOF is disappearing owing to advances in trauma and critical care, whereas other studies have not found a consistent change in either the incidence or the mortality rate associated with postinjury MOF (1, 9-11). These disparities are in part due to the difficulty in defining and measuring MOF. The syndrome seems to fit well with Justice Stewart's famous quote: "I may not be able to define it, but I know it when I see it."
Recently, Baue (12), in a thorough review of the MOF literature, highlighted the plethora of definitions and scoring systems. Most definitions were developed over a decade ago, yet little has been done in terms of validating them or, moreover, to understand how and why they differ. Indeed, Marshall (13), who led the construction of the multiple organ dysfunction score (MODS), noted in a recent editorial that we need to better understand the differences between scales to determine whether one approach is superior or, alternatively, whether two different but informative facets of a question are being addressed.
In the absence of a gold standard, validation must be done examining the association of different scores with objective, adverse outcomes, reflecting clinical status and resource utilization. In addition, we must fully understand how patients classified as "MOF" or "no MOF" by disparate scores differ in terms of their initial insult, risk factors, and outcomes. This process would be greatly enhanced if conducted on a homogenous population observed for an extended period. Therefore, we propose to use our long-term trauma database to validate two scores (Marshall MODS [Marshall] and Denver MOF score [Denver]). These two scores have been used successfully in multiple trials, which underscores the importance of validating them and understanding how their classifications of patients differ, so we can better appraise the body of literature using them as end points (6, 14-22). Our database, started in 1992, contains prospectively collected clinical data on patients at risk for postinjury MOF for the first 28 postinjury days (1, 8, 23-27). To our knowledge, this is one of the longest sustained databases on postinjury MOF using standardized data collection methods. As part of the validation process, we also examined the scores' potential as continuous scales that could be used to monitor patients' response to treatment and prognosis, a critical step in monitoring the dynamic process of organ dysfunction.
MATERIAL AND METHODS
Study population and data collection
Acutely injured patients admitted to the Rocky Mountain Regional Trauma Center surgical ICU (SICU) at Denver Health Medical Center were studied prospectively from 1992 through 2004. Denver Health Medical Center is a state-designated level I trauma center verified by the American College of Surgeons Committee on Trauma. Inclusion criteria were Injury Severity Score (ISS) greater than 15, survival longer than 48 h from injury, admission to the SICU within 24 h of injury, and age older than 15 years. Patients with isolated head injuries (head injuries with an extracranial abbreviated injury score less than 2) were excluded. Burn patients and hanging injuries were not included.
Patient characteristics were recorded at the time of hospital admission. Daily physiological and laboratory data were collected through SICU day 28, and clinical events were recorded on all patients thereafter until death or hospital discharge. The data collection and storage processes are in compliance with Health Insurance Portability and Accountability Act regulations and have been approved by our institutional review board. The data collection and storage processes are in compliance with Health Insurance Portability and Accountability Act regulations and have been approved by our institutional review board and informed consent was obtained.
Multiple organ dysfunction scores
Data were collected daily to calculate the Denver MOF score (Denver score) and the Marshall MODS (Marshall score) (27-30). In brief, the Denver score rates the dysfunction of four organ systems (pulmonary, renal, hepatic, and cardiac), which are evaluated daily throughout the patient's ICU stay and graded on a scale from 0 to 3 (Table 1), with the total score ranging from 0 to 12. Table 2 outlines the Marshall score, which evaluates the same four organ dysfunctions as the Denver score, using slightly different criteria, plus two additional organ dysfunctions, that is, hematologic and neurological. The Marshall score grades dysfunction from 0 to 4, and the total score ranges from 0 to 24. The values for respiratory dysfunction have been adjusted for altitude by multiplication of the value by the ratio of atmospheric pressure in Denver to that at sea level (630/760 mmHg). Both MOF scores were calculated as the sum of the simultaneously obtained individual organ scores on each hospital day. Daily laboratorial and physiological values were used for calculation of both scores.
Multiple organ failure status by the Denver score was defined as recommended by the authors as well as others investigators, that is, score of more than 3 occurring any day after 48 h after injury (31, 32). The authors of the Marshall score have not recommended a specific cutoff, rather they established score ranges for evaluation of mortality, as follows: 9 to 12, 13 to 16, 17 to 20, and greater than 20 with associated mortalities of 25%, 50%, 75%, and 100%, respectively (28). Other studies, including those published by the Glue Grant group (www.gluegrant.org), have used a Denver score of more than 3 for at least 2 consecutive days and a Marshall score of more than 5 for at least 2 consecutive days. In our study, we revisited all these cutoffs using receiver operating characteristic (ROC) curves, which plot the sensitivity in the y axis and 1-specificity in the x axis, with each point representing the sensitivity and 1-specificity of all potential cutoffs of the measurement in question. The best cutoff for a measurement, that is, the cutoff that has the best combination of sensitivity and specificity, is the data point located at the left-uppermost position in the ROC curve.
Validation of the two MOF scores and patient adverse outcomes
Validation was performed by evaluating the scores' association with objective adverse outcomes occurring within 28 days after injury. We used two hospital resource utilization outcomes-length of ICU stay (LOS) and mechanical ventilation (MV) days-as well as two patient-specific outcomes-ventilator-free days (VFDs) and death. Ventilator-free days have been recently proposed as an alternative outcome for ICU patients that accounts for patients who died early and consequently had shorter MV times. It was calculated as recommended by Schoenfeld and Bernard (33). Because LOS, MV days, and VFDs were not normally distributed, these outcomes were categorized in LOS of longer than 14 days, MV time of longer than 7 days, and VFDs of less than 21 days, respectively. These cutoffs were chosen based on a combination of clinical relevance and the data distribution in tertiles. The prediction of these outcomes was evaluated by building logistic regression models in which the above described adverse outcomes were the dependent variables and the MOF status as defined by the scores was the independent variable or predictor. The performance of each MOF score was then evaluated using the area under the ROC curves, sensitivity, specificity, and measures of goodness-of-fit of logistic regression models: (a) Akaike Information Criterion, for which lower values reflect better goodness-of-fit; and (b) −2 log-likelihood reduction, for which larger reductions reflect better goodness-of-fit. We also examined the correlation with the above-mentioned score ranges for mortality evaluation as proposed by the Marshall score authors (28).
Describing the patients classified by each MOF score
Patients could fall into one of four potential scenarios, as follows:
- classified as MOF by both scores,
- classified as MOF by the Denver score and as no MOF by the Marshall score,
- classified as MOF by the Marshall score and as no MOF by the Denver score, and
- classified as no MOF by both scores.
For each one of these four groups, we evaluated MOF risk factors, clinical outcomes (death, VFDs), and utilization outcomes (MV time, LOS).
Evaluation of scores as continuous scales
We evaluated the performance of each score as continuous scales by examining (a) the correlation between mortality rates and score values over time, and (b) the association of mortality with score changes from day 2 (baseline) to days 4 and 5. The first was used to evaluate the scores as prognostic indices, whereas the second was used to indicate if improvement/deterioration of the score could be used to monitor patient's response to treatment. Improvement/deterioration of the scores over time (delta) was calculated as (day 4 score−day 2 score) and (day 5 score−day 2 score).
All analyses were performed using SAS for Windows, version 9.0 (SAS Institute, Cary, NC), ROC curves were constructed, and the areas under the ROC curve were compared using StatsDirect (1990-2007 StatsDirect Limited, UK). Categorical variables were analyzed using χ2 test or the Fisher exact test when expected cell values were less than 5. For continuous variables with normal distribution, ANOVA or Student t tests (with the appropriate Satterwhaite modification when the assumption of equal variances did not hold) were used. P < 0.05 was considered significant. Logistic regression analysis was used to adjust for multiple confounders. Data were expressed as mean, SD, SEM, or, when variables were not normally distributed, as median and interquartile range (lower and upper quartiles).
Overall, 1,389 consecutive patients met the inclusion criteria and were prospectively followed up during the 12-year study period. Table 3 depicts the characteristics of the patient population. In brief, our population consisted of mostly young, healthy males, with moderate to severe injuries, mostly due to blunt force in motor vehicle accidents. Mortality and number of VFDs were low, but health care resources utilization was high, with more than one third of these patients requiring longer than 14 admission days in the SICU (LOS) and/or more than 7 days of MV time.
Table 4 shows the distribution of risk factors stratified by MOF classification, whereas Table 5 depicts the outcomes. Patients could fall into one of four potential scenarios: (a) classified as MOF by both scores, (b) classified as MOF by the Denver score but as no MOF by the Marshall score, (c) classified as MOF by the Marshall score but as no MOF by the Denver score, and (d) classified as no MOF by both scores. As shown in Tables 4 and 5, patients in scenarios 1 and 2 presented very similar risk factor and outcome rates (χ2 test, P > 0.05 for all comparisons) so they were collapsed into one group. Therefore, as illustrated in Figure 1, three major groups emerged: (a) patients classified by both scores as MOF and also those patients classified as MOF by the Denver score but as no MOF by the Marshall score (or in other words, patients classified as MOF by the Denver score regardless of their classification by the Marshall score) who comprised a severe injury group of patients with high rates of risk factors, mortality, and utilization, (b) patients classified as MOF by the Marshall score but as no MOF by the Denver score represented a moderate injury group with medium risk factor rates, medium utilization, and low mortality, and (c) patients classified as no MOF by both scores constituted a mild injury group for whom risk factor rates, utilization, and mortality were all low.
Validation of the scores using ROC curves
The ROC curves for the prediction of outcomes (death, VFDs <21 days, LOS >14 days, and MV time >7 days) by the Denver and Marshall MOF scores using the 1-day duration requirement are shown in Figures 2 to 5. Overall, both scores were associated with areas 80 or greater (ideal value = 100). The area under the ROC curve for the Denver score was slightly greater than for the Marshall score regarding prediction of death, VFDs, and MV time, although these differences were not significant, as demonstrated by the overlap in the areas' 95% confidence intervals. The ROC curves suggested that the best cutoffs for the Denver and Marshall scores were more than 3 and more than 8, respectively. The traditional cutoff of more than 5 for the Marshall score was associated with a worse combination of sensitivity and specificity. Requiring scores higher than the cutoffs for at least two consecutive days (as opposed to scores higher than cutoff for just any 1 day) resulted in lower prediction power, as indicated by a decreased area under the ROC curve for both the Denver and Marshall scores prediction of death (Fig. 2, A and B), as well as for the other outcomes.
Validation of the scores using goodness-of-fit measures
Goodness-of-fit of prediction models for death was better for the Denver MOF score of more than 3 for 1 day compared with the Marshall score of more than 5 for 1 day (−2 log likelihood reduction: 192 vs. 106; Akaike Information Criterion 610 vs. 696). Changing the Marshall score to cutoff of more than 8 (best cutoff by the ROC curve) for death improved the goodness-of-fit of the Marshall score (−2 log likelihood reduction: 162; Akaike Information Criterion: 640), but its predictive performance remained inferior to the Denver MOF score's goodness-of-fit. When the requirement for a 2-day minimum MOF duration was applied, the Denver score continued to perform better than the Marshall score, yet for both models, the goodness-of-fit was worse than for the models using the 1-day requirement (−2 log likelihood reduction: 157 vs. 98.7; Akaike Information Criterion: 645.1 vs. 703.3; C-statistic: 0.76 vs.0.74), confirming the findings with the ROC curve analyses that requiring that the score was higher than a cutoff for at least 2 days did not improve the performance of the scores.
Tables 6 to 9 summarize other measures of goodness-of-fit for the association of the scores with death, VFDs of less than 21 days, MV time of more than 7 days, and LOS of more than 14 days for the total scores. Both the Denver and Marshall scores were associated with values of sensitivity and specificity of more than 70% for death and VFDs of less than 21 days, but either the sensitivity or the specificity was less than 70% for the MV days and LOS, suggesting that these scores are appropriately biased toward clinical outcomes as opposed to resource utilization.
Tables 6 to 9 also depict the performance of each individual organ dysfunction in predicting the outcomes. Heart dysfunction, graded by either score, had the best predictive power for death. Kidney dysfunction when graded by the Marshall score correlated well with death, suggesting that the Denver kidney dysfunction score could be adjusted to values similar to those used by the Marshall score to increase sensitivity. The association of kidney dysfunction with other adverse outcomes was generally low. Not surprisingly, lung dysfunction, coded by either score, was predictive of both VFDs and MV days, but did not perform well with the other two outcomes. Hematologic and central nervous system (CNS) dysfunctions were relatively poor predictors of all the adverse outcomes.
Use of scores as continuous scales
Figure 6, A and B, depicts the correlation between mortality and MOF scores obtained in days 2 (the first 24 hours was defined as day 0), 3, and 7 after injury. Multiple organ failure scores obtained on days 4 to 6 were also examined, but the results were similar and are not shown here. The Denver score (Fig. 6A) had a consistent, uniform increase in mortality risk associated with each one-point increase in the total score, particularly up to a total score of 4 (range 0-12). Scores of greater than 4 exhibited a larger variability, possibly a function of the smaller number of patients in these ranks. The Marshall score (Fig. 6B) showed a consistent, yet relatively small increase in mortality risk as the scores increased, up to a score of 4 (range 0-24), above which the correlation of scores and mortality risk was inconsistent.
We also examined the association of the Marshall score with mortality as suggested by the creators of the score, that is, using ranges for mortality. We observed that levels between 9 to 12 and 13 to 16 were indeed associated with mortality rates of approximately 25% and 50%, respectively (Fig. 7), as reported by the score's authors. Marshall scores of 17 to 20 and more than 20 were uncommon and did not differ in mortality (both associated with a 100% mortality). Therefore, changes of less than four points in the Marshall score were not associated with variation in the risk of death.
Improvement/deterioration of the scores over time was explored next. Figure 8 illustrates the association of the delta (day 5 score−day 2 score) with mortality. Remarkably similar numbers were observed for the delta (day 4 score−day 2 score) (data not shown). Patients who increased their Denver MOF score by one to three points showed a consistent increase in mortality risk from 10% to about 33%. The Marshall score showed a less consistent, yet similar pattern. Of note, patients with large improvements from days 2 to 5 (delta = −3) were more likely to die than patients with no improvement or with smaller improvements. This suggested a that the initial, baseline score was important and not only the delta. Indeed, stratifying patients by the value of their baseline Denver MOF score (day 2 score ≤2 and day 2 score >2) minimized this effect, as shown in Figure 9.
The Denver MOF score was first proposed in 1991 as part of a National Institutes of Health Center Project (34), then modified in 1993 (35) by a panel of expert trauma surgeons and trauma investigators. The Marshall score was developed in 1994 through a systematic review of the literature to define organ dysfunction, with specific cutoffs derived and validated based on probability of subsequent mortality in a sample of 692 patients from a Canadian ICU (13, 28, 30). Major differences between the two scoring systems include (a) the Denver score considers only four organ systems (heart, kidney, liver, and lungs), whereas the Marshall score also includes the CNS and the hematologic system; (b) scoring of heart dysfunction is based on physiological variables for the Marshall score (1994 version = heart rate and lactate; 1995 = pressure-adjusted heart rate [PAR]), whereas the Denver scores uses inotrope support requirement; and (c) cutoffs for organ dysfunction in the Denver score are, in general, more specific and less sensitive than the cutoffs used by the Marshall score. This difference in sensitivity explains, at least in part, why the incidence of MOF in our study according to the Denver definition was 22% with a case-fatality rate of 30%, whereas according to the Marshall score (using the traditional >5 cutoff), the incidence was much higher at 50% with a lower case-fatality rate of 15%. Using the cutoff of more than 8 (defined as optimum by the ROC curve analysis) improved the performance of the Marshall score but still to a level inferior to the Denver score regarding association with adverse outcomes. A similar pattern emerged when we looked at known risk factors for MOF, such as age, ISS, and red blood cells (RBCs) (8, 29, 36, 37). Incidentally, once more RBCs transfused in the first 6 h emerged as a major risk factor increasing the risk of MOF by at least 3-fold, regardless of definition confirming previous findings in this arena (26, 38, 39).
Collectively, our findings suggest that both scores performed reasonably well, with the Denver score having a slightly better performance than the Marshall score in the discrimination of adverse clinical and utilization outcomes because of greater specificity. A similar finding was reported by Grotz et al. (31) in a recent study comparing three MOF scores (Denver, Marshall, and Goris) on 301 severely injured patients who were graded daily into a group with and a group without MOF by an experienced intensive care physician. The Denver score's sensitivity was 81%, and its specificity was 88%, superior to the other scores. The authors attributed the lower specificity of the Goris and Marshall scores to liver and cardiovascular dysfunction grading (31). Also similar to our findings, Grotz et al. found that grading the heart, lungs, kidney, and liver seemed sufficient and that other organs dysfunctions did not add to the performance of the scores (31).
Our study found that both scores had areas under the ROC curve equal or superior to 80, a value much larger than the values recently found by Zygun et al. (40) who compared the Sepsis-related Organ Failure Assessment score, developed by a consensus conference at a meeting of the European Society of Intensive Care Medicine, and the Marshall score. In their study, all areas under the ROC curves were less than 63. It is important to note that their sample was composed mostly of general ICU patients (excluding patients admitted to a general coronary unit and elective surgery with shorter stay), with trauma patients representing less than 12% of their study group. We found that limiting our sample to trauma patients and further excluding isolated head injuries, burns, and hanging injuries strengthen our study design by providing a homogeneous, relatively young population among whom previous medical conditions affecting major organ systems are less frequent than in the general ICU population. In addition, the time of the initial insult is often known among trauma patients, which is usually not possible in other critically ill patients.
We observed that both scores were appropriately biased toward clinical outcomes (death and VFDs) as opposed to resource utilization, as measured by MV days and LOS. This is not surprising as resource utilization may be influenced by factors not directly related to the patient condition, such as ICU and hospital management, ICU and hospital staff issues, protocols, availability of finite resources, and so on.
In previous studies, we have avoided using the Denver MOF score as a continuous measure, fearing that a scale constructed as ordinal would not serve as continuous. In this study, we examined more closely the performance of the Denver MOF score as a continuous scale, which could be used to monitor patients over time, deriving prognosis and evaluating response to treatment based on temporal changes in the score. This preliminary evaluation seems to support the use of the Denver MOF score in this fashion. Denver MOF scores obtained in different days were consistently and uniformly associated with a similar mortality risk. Increases (or decreases) of a single point from days 2 to 4 and from days 2 to 5 were associated with measurable, consistent changes in prognosis. The Denver score seems to perform better within the range of 0 to 4, suggesting that we may need to continue to refine the scale and ranges with regard to higher scores. Nevertheless, our findings indicate that the parsimonious Denver score can be a useful device to monitor patient status in the clinical environment.
Overall, individual organ dysfunctions functioned inconsistently as predictors of adverse outcomes, with few exceptions. In brief, kidney dysfunction, as graded by the Marshall score, seems to correlate well with mortality, suggesting that their ranges have better discriminatory performance than the Denver MOF score. We are now working on modifications of the renal dysfunction cutoffs of the Denver MOF score to improve its performance. Not surprisingly, lung dysfunction scores were associated with MV outcomes (MV time >7 days and VFDs <21 days). Liver dysfunction, measured by either score, had a poor predictive performance in general, possibly because bilirubin is potentially affected by factors unconnected to the liver function (e.g., hemolysis). This finding underscores the importance of assessing other measures of liver function. One of the goals of our clinical research now is on evaluating the role of new markers of individual organ dysfunction that can detect deterioration of function earlier and more precisely.
Admittedly, we did not compare grading of the cardiac dysfunction by the two versions of the Marshall score: (a) the treatment-dependent version, in which heart rate, use of inotropes, and lactate levels are used to grade the dysfunction; and (b) the treatment-independent version, which uses a composite measure named pressure-adjusted heart rate (PAR = heart rate × central venous pressure [CVP] / MAP) (28, 30). The authors' rationale was that this composite measure corrected cardiovascular function for physiological support and showed the desired incremental correlation with ICU mortality rate. Values for each of the three component variables had to be measured simultaneously. According to Marshall et al. (28), the PAR satisfied most of the characteristics of an ideal variable, except that it is potentially altered as a result of transient changes associated with resuscitation, and may be affected by such therapeutic interventions as the use of β-adrenergic receptor blockers or vasopressors.
Although we agree that the treatment independence (although this independence was relative because PAR is directly affected by treatment) is an attractive feature, the component variable CVP is often not available; in fact, it was not available in half of the patients in the original authors' data set (28). We decided that imputation of zero values for the score [or a normal value of 8 for the CVP as more recently proposed by Marshall (13)] in a large proportion of patients (because of missing CVP data) introduced more uncertainty than using the earlier version, which had a limited use of a treatment-dependent measurement (use of inotropes) as seen in Table 2. In addition, Marshall's group reverted to the 1994 simpler version of their cardiovascular dysfunction score in a subsequent study published in 2001 (41). Thus, we opted to use the 1994 version using heart rate, inotrope use, and lactate levels (41). Existing alternatives to the use of treatment-dependent variables are either missing in a large proportion of patients or under investigation (42, 43). Interestingly, both the Marshall and Denver cardiac dysfunction scores functioned well as predictors of death with large areas under the ROC curves as well as sensitivity and specificity values greater than 80%.
Another limitation of our study is the inclusion of patients of a single ICU, which restricts our ability to evaluate the influence of different protocols on the scoring of MOF. The Glue grant, which uses the Denver MOF score as one of their outcome measures and includes patients from several centers, should provide additional data to verify the applicability of our score in other trauma facilities. Finally, we use worst daily values for both scores, yet the Marshall score authors recommend the use of physiological values measured at the same point in time every day (first morning values) to avoid capturing momentary physiological changes unrelated to changes in the patient's underlying physiological status. We made a decision to obtain the same values (worst values) for both scores, which may have an effect on the Marshall score.
In conclusion, (a) both scores perform reasonably well as indicators of adverse outcomes in critically ill patients, with the Denver MOF score performing slightly better due to greater specificity; (b) the Denver MOF score performs better as a continuous scale to monitor individual patient's response to treatment; and (c) the analysis of individual organ dysfunction scores suggests that concepts of the two scores can be combined to develop a superior score. The Denver MOF score provides an appealing combination of good performance with simplicity, making it a more attractive tool to be used in clinical research both as an outcome in trials and in risk adjustment as well as a monitoring device at the bedside.
The authors thank the program manager, Ms Victoria Bress, for her assistance and the medical staff at the Denver Health Medical Center for their invaluable contribution.
1. Ciesla DJ, Moore EE, Johnson JL, Burch JM, Cothren CC, Sauaia A: A 12-year prospective study of postinjury multiple organ failure: has anything changed? Arch Surg
2. Teixeira PGRM, Inaba KM, Hadjizacharia PM, Brown CM, Salim AM, Rhee PM, Browder T, Noguchi TTM, Demetriades D: Preventable or potentially preventable mortality
at a mature trauma center [article]. J Trauma Inj Infect Crit Care
3. Sauaia A, Moore FA, Moore EE, Moser KS, Brennan R, Read RA, Pons P: Epidemiology of trauma deaths: a reassessment. J Trauma
4. Goris RJ: Prevention of ARDS and MOF by prophylactic mechanical ventilation and early fracture stabilisation. Prog Clin Biol Res
5. Laudi S, Donaubauer B, Busch T, Kerner T, Bercker S, Bail H, Feldheiser A, Haas N, Kaisers U: Low incidence of multiple organ failure after major trauma. Injury
6. Sperry JL, Frankel HL, Vanek SL, Nathens AB, Moore EE, Maier RV, Minei JP: Early hyperglycemia predicts multiple organ failure and mortality
but not infection. J Trauma
7. Nast-Kolb D, Aufmkolk M, Rucholtz S, Obertacke U, Waydhas C: Multiple organ failure still a major cause of morbidity but not mortality
in blunt multiple trauma. J Trauma
8. Sauaia A, Moore FA, Moore EE, Norris JM, Lezotte DC, Hamman RF: Multiple organ failure can be predicted as early as 12 hours after injury. J Trauma
9. Levine JH, Durham RM, Moran J, Baue A: Multiple organ failure: is it disappearing? World J Surg
10. Durham RM, Moran JJ, Mazuski JE, Shapiro MJ, Baue AE, Flint LM: Multiple organ failure in trauma patients. J Trauma
11. Regel G, Grotz M, Weltner T, Sturm JA, Tscherne H: Pattern of organ failure following severe trauma. World J Surg
12. Baue AE: MOF, MODS, and SIRS: what is in a name or an acronym? Shock
13. Marshall JC: Measuring organ dysfunction in the intensive care unit: why and how? Can J Anaesth
14. Klein DJ, Derzko A, Foster D, Seely AJ, Brunet F, Romaschin AD, Marshall JC: Daily variation in endotoxin levels is associated with increased organ failure in critically ill patients. Shock
15. Frink M, Pape HC, van Griensven M, Krettek C, Chaudry IH, Hildebrand F: Influence of sex and age on MODS and cytokines after multiple injuries. Shock
16. Nathens AB, Nester TA, Rubenfeld GD, Nirula R, Gernsheimer TB: The effects of leukoreduced blood transfusion
on infection risk following injury: a randomized controlled trial. Shock
17. Bulger EM, Jurkovich GJ, Farver CL, Klotz P, Maier RV: Oxandrolone does not improve outcome of ventilator dependent surgical patients. Ann Surg
18. Cohn SM, Nathens AB, Moore FA, Rhee P, Puyana JC, Moore EE, Beilman GJ: Tissue oxygen saturation predicts the development of organ dysfunction during traumatic shock resuscitation. J Trauma
19. Ikossi DG, Knudson MM, Morabito DJ, Cohen MJ, Wan JJ, Khaw L, Stewart CJ, Hemphill C, Manley GT: Continuous muscle tissue oxygenation in critically injured patients: a prospective observational study. J Trauma
20. Johnson JL, Moore EE, Gonzalez RJ, Fedel N, Partrick DA, Silliman CC: Alteration of the postinjury hyperinflammatory response by means of resuscitation with a red cell substitute. J Trauma
21. Bulger EM, Jurkovich GJ, Nathens AB, Copass MK, Hanson S, Cooper C, Liu PY, Neff M, Awan AB, Warner K, et al: Hypertonic resuscitation of hypovolemic shock after blunt trauma: a randomized controlled trial. Arch Surg
22. Sperry JL, Friese RS, Frankel HL, West MA, Cuschieri J, Moore EE, Harbrecht BG, Peitzman AB, Billiar TR, Maier RV, et al: Male gender is associated with excessive IL-6 expression following severe injury. J Trauma
23. Ciesla DJ, Moore EE, Johnson JL, Cothren CC, Banerjee A, Burch JM, Sauaia A: Decreased progression of postinjury lung dysfunction to the acute respiratory distress syndrome and multiple organ failure. Surgery
24. Ciesla DJ, Moore EE, Johnson JL, Burch JM, Cothren CC, Sauaia A: Obesity increases risk of organ failure after severe trauma. J Am Coll Surg
25. Ciesla DJ, Moore EE, Johnson JL, Burch JM, Cothren CC, Sauaia A: A 12-year prospective study of postinjury multiple organ failure: has anything changed? Arch Surg
26. Moore FA, Moore EE, Sauaia A: Blood transfusion
. An independent risk factor for postinjury multiple organ failure. Arch Surg
27. Moore FA, Sauaia A, Moore EE, Haenel JB, Burch JM, Lezotte DC: Postinjury multiple organ failure: a bimodal phenomenon. J Trauma
28. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL, Sibbald WJ: Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med
29. Sauaia A, Moore FA, Moore EE, Lezotte DC: Early risk factors for postinjury multiple organ failure. World J Surg
30. Marshall JC: A scoring system for the multiple organ dysfunction syndrome (MODS). In: Reinhart K, Eyrich K, Sprung C, eds.: Sepsis: Current Perspectives in Pathophysiology and Therapy
. Berlin: Springer-Verlag, 38-49, 1994.
31. Grotz M, von Griensven M, Stalp M, Kaufmann U, Hildebrand F, Pape HC: Scoring multiple organ failure after severe trauma. Comparison of the Goris, Marshall and Moore scores. Chirurg
32. Sauaia A, Moore FA, Moore EE, Haenel JB, Read RA, Lezotte DC: Early predictors of postinjury multiple organ failure. Arch Surg
33. Schoenfeld DA, Bernard GR: Statistical evaluation of ventilator-free days as an efficacy measure in clinical trials of treatments for acute respiratory distress syndrome. Crit Care Med
34. Moore FA, Moore EE, Poggetti R, McAnena OJ, Peterson VM, Abernathy CM, Parsons PE: Gut bacterial translocation via the portal vein: a clinical perspective with major torso trauma. J Trauma
35. Sauaia A, Moore FA, Moore EE, Haenel JB, Read RA: Pneumonia: cause or symptom of postinjury multiple organ failure? Am J Surg
36. Hensler T, Heinemann B, Sauerland S, Lefering R, Bouillon B, Andermahr J, Neugebauer EA: Immunologic alterations associated with high blood transfusion
volume after multiple injury: effects on plasmatic cytokine and cytokine receptor concentrations. Shock
37. Hensler T, Sauerland S, Lefering R, Nagelschmidt M, Bouillon B, Andermahr J, Neugebauer EA: The clinical value of procalcitonin and neopterin in predicting sepsis and organ failure after major trauma. Shock
38. Silliman CC, Moore EE, Johnson JL, Gonzalez RJ, Biffl WL: Transfusion of the injured patient: proceed with caution. Shock
39. Aiboshi J, Moore EE, Ciesla DJ, Silliman CC: Blood transfusion
and the two-insult model of post-injury multiple organ failure. Shock
40. Zygun DA, Laupland KB, Fick GH, Sandham JD, Doig CJ: Limited ability of SOFA and MOD scores to discriminate outcome: a prospective evaluation in 1,436 patients. Can J Anaesth
41. Cook R, Cook D, Tilley J, Lee K, Marshall J: Multiple organ dysfunction: baseline and serial component scores. Crit Care Med
42. Wu TT, Yuan A, Chen CY, Chen WJ, Luh KT, Kuo SH, Lin FY, Yang PC: Cardiac troponin I levels are a risk factor for mortality
and multiple organ failure in noncardiac critically ill patients and have an additive effect to the APACHE II score in outcome prediction. Shock
43. Edouard AR, Benoist JF, Cosson C, Mimoz O, Legrand A, Samii K: Circulating cardiac troponin I in trauma patients without cardiac contusion. Intensive Care Med