INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Since its identification in 1967 by Ashbaugh and colleagues, many investigators have examined the features that affect the outcome of acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) (1). Univariate analyses have identified multiple factors that are associated with increased mortality rates for ALI and ARDS, including sepsis syndrome preceding (2, 3) or following ARDS (4, 5); Acute Physiology and Chronic Health Evaluation II (APACHE II) score (5, 6); preceding (2, 7, 8) or subsequent organ failure (5, 7, 9); older age (5, 8, 10); or the presence of comorbid or chronic conditions, such as cirrhosis (2, 14); human immunodeficiency virus (HIV) infection (5, 15); organ transplantation (2, 16); and active malignancy (2, 5). In contrast, few multivariate analyses of factors associated with increased mortality in ALI or ARDS are available, and these have included either a mixed patient population (e.g., combining medical and surgical patients) or have analyzed only a portion of the important univariate predictors previously identified (2, 3, 6, 17). A priori identification of populations at high risk for in-hospital mortality will have important implications for interpreting mortality statistics provided by studies of new therapeutic strategies for ALI and ARDS. In addition, all factors that affect mortality must be considered in evaluating trends in ALI and ARDS outcome (11, 18, 19). To fully interpret analyses of such data the prevalence of important comorbid conditions and how they change over time must be known. To identify conditions present at the onset of respiratory failure that contribute independently to mortality, we conducted an exclusive study of medical-intensive-care-unit (MICU) patients with ALI and ARDS (20).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Over a 36-mo period, 107 consecutive patients meeting criteria for ALI (see the subsequent discussion) were recruited from the MICU of the New England Medical Center, the principal teaching hospital of Tufts University School of Medicine and a major tertiary-care referral center in Boston, Massachusetts. These patients represented 22% of all patients mechanically ventilated within this period. All of the patients were exclusively medical; with the exception of a single patient with ARDS secondary to fat emboli following an orthopedic procedure, there were no cases of major trauma or postoperative acute lung injury. In addition, no patient received nitric oxide (NO), surfactant, or other experimental pharmacologic intervention for either ALI and ARDS or for sepsis. Patients were admitted to the MICU from three sources: transfer from a floor bed, direct admission from the emergency department (ED), and transfer from an outside hospital. Candidates for the study were prospectively identified by a daily review of all ventilated patients in the 10-bed MICU.

Criteria for Diagnosis of ALI and ARDS

The criteria for diagnosis of ALI and ARDS as set by the North American/European Consensus Conference were used and were as follows: acute onset of lung injury, diffuse bilateral infiltrates seen upon chest radiography, Pa_O2/FI_O2 < 200 mm Hg for ARDS and Pa_O2/ FI_O2 < 300 mm Hg for ALI, pulmonary artery occlusion pressure (PAOP) < 19 mm Hg, or no clinical evidence of congestive heart failure (CHF) (20).

Data Collection

At the onset of mechanical ventilation, the history and physical, laboratory, and roentgenographic data and hemodynamic and ventilatory variables were recorded, and the presence of comorbid conditions (see the subsequent discussion) was noted. The APACHE II score and acute physiology score (APS) were calculated within 6 h of initiation of mechanical ventilation (21). On the basis of the data available in the first 24 h of mechanical ventilation, an LIS was calculated (22). Physiologic and laboratory characteristics prior to the initiation of mechanical ventilation were measured to determine the development of nonpulmonary organ-system dysfunction (ODS) between the time of hospital admission and the onset of mechanical ventilation. A cause for ALI or ARDS was assigned, using prospectively defined criteria (see the subsequent discussion). All patient screening, data collection, and chest-radiograph interpretation were performed by the investigators.

Definitions

On the basis of a review of previously published univariate analyses, comorbid conditions most likely to affect mortality were prospectively selected and defined as "primary" chronic diseases or comorbidities as follows:

Malignancy: Active, untreated, or undergoing current treatment (patients having received a bone-marrow transplant for their malignancy were excluded from this category).

Cirrhosis: Biopsy-proven or with evidence of portal hypertension, such as variceal bleeding, ascites, or encephalopathy in the appropriate clinical setting.

HIV infection: Serologic evidence of infection with or without previous acquired immune deficiency syndrome (AIDS)-defining illness.

Organ transplantation: A history of bone marrow, liver, or kidney transplantation.

We also identified the presence of other potentially important chronic conditions, as follows:

Chronic obstructive pulmonary disease (COPD): A known history of obstructive lung disease or obstructive defect on pulmonary function studies.

Alcohol abuse: Current active use of alcohol.

Intravenous drug use (IVDU): Current active use of intravenous drugs.

Chronic steroid use: Current use of corticosteroids for  2 weeks and dose  20 mg/d.

Diabetes mellitus: A history of diabetes requiring chronic therapy with insulin or an oral hypoglycemic agent.

The impact of age on outcome was determined by categorizing patients by age, using a threshold age of 65 yr. Preceding nonpulmonary OSD was defined as such dysfunction developing after admission to the hospital and persisting through the time of initiation of mechanical ventilation, in a specific organ system, as follows:

Hepatic dysfunction: Total serum bilirubin > 2.0 mg/dl, along with a prothrombin time > 3 s longer than the control value.

Renal dysfunction: Serum creatinine > 2.0 mg/dl.

Hematologic dysfunction: A platelet count of < 75,000/mm³, total white blood cell (WBC) count < 1,000 cells/mm³, or evidence of disseminated intravascular coagulation (DIC).

Neurologic dysfunction: Obtundation not attributable to a sedative drug effect or a primary central nervous system (CNS) event, such as seizures, hemorrhage, or acute cerebrovascular accident.

Gastrointestinal dysfunction: Ileus lasting > 24 h or hemorrhage requiring transfusions (23).

Risk Factors for ALI/ARDS

For each case, an etiology of ALI and ARDS was prospectively defined. Sepsis was defined as present if at least three of the following criteria were satisfied: rectal temperature > 39° C or < 35° C; unexplained hypotension with a systolic blood pressure < 90 mm Hg for at least 2 h in the absence of dehydration; hemorrhage or cardiac failure; leukocytosis manifested by an increase of at least 3,000/mm³ in one day or > 20% bands; positive blood culture; and systemic vascular resistance < 800 dynes/s/cm⁵ in the absence of chronic liver disease (7). Aspiration of gastric contents was said to be present if it was documented by direct observation by medical personnel or if gastric contents were suctioned from the endotracheal tube (9). The presence of pneumonia was defined from the radiographic presence of new infiltrate(s), coupled with identification of at least one pathogen in a sputum specimen or bronchoalveolar lavage fluid (BALF) culture and clinical evidence of infection (7). Patients meeting the criteria for both sepsis and pneumonia were classified as having sepsis and were excluded from the pneumonia category. Patients who met the criteria for ALI or ARDS in the setting of other clinical conditions, including drug overdose (24), hypertransfusion (17), pancreatitis (7), and fat embolism (24), or when no definite etiology could be identified, were classified as having another condition.

Statistical Analysis

The principal outcome studied was hospital survival. Survivors and nonsurvivors were compared through Student's t test for continuous variables and a chi-square test with two-tailed Fisher's exact test for dichotomous variables. Univariate analysis was done by comparing the mortality rates in the presence and absence of a comorbid condition, previous nonpulmonary OSD, APACHE II score > 20, APS > 15, LIS > 2.5, and the individual risk factors for ALI and ARDS. Relative risk (RR) and 95% confidence intervals (CIs) were determined. Our goal was to ascertain whether factors previously found to be associated with increased mortality in lung injury in univariate analyses would continue to have an independent effect after controlling for other important variables. Therefore, to ascertain which factors contributed independently to mortality, we constructed a multiple-logistic-regression model. This was done in a forward, stepwise manner, with hospital death as the dependent variable and the prospectively identified independent variables including older age (> 65 yr), individual primary comorbidities, septic and gastric aspiration-related etiologies of lung injury, APACHE II score or APS, LIS, and development of nonpulmonary OSD prior to the onset of mechanical ventilation. Another analysis was then done, using the other ("nonprimary") comorbidities as additional independent variables. These analyses were then repeated in the cohort of patients meeting criteria for ARDS. All statistical analysis was done with SPSS v.6.1 software (SPSS, Inc., Chicago, IL).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Of the 107 patients enrolled in the study, there were 65 men and 42 women, aged 51 ± 2 yr (mean ± SEM; range: 20 to 73 yr) (Table 1). A prospectively defined primary comorbid condition was often present, including cirrhosis (n = 21; 20%), HIV infection (n = 11; 10%), active malignancy (n = 26; 24%), and organ transplantation (n = 10; 9%). Of the 10 transplant patients, nine had received a bone-marrow and one a kidney transplant. The most prevalent predisposing conditions for ALI and ARDS were pneumonia, sepsis, and aspiration of gastric contents (Figure 1). A similar distribution of comorbid conditions and etiologies was found in the 81 (76%) patients meeting the criteria for ARDS (Table 2).

View larger version (26K): [in this window] [in a new window]   Figure 1.   Clinical conditions associated with acute lung injury in 107 medical-intensive-care-unit patients.

Sixty-two (58%) of the 107 patients, including 47 (58%) of the 81 patients with ARDS, died in the hospital. Mortality was greater among patients with sepsis or gastric aspiration than among those with pneumonia or other causes of ARDS (p < 0.05) (Figure 2, Table 2). When survivors were compared with nonsurvivors, the latter were slightly older and had a higher APACHE II score, whereas there was no difference in gender, APS, Pa_O2/FI_O2, or LIS (Table 1). A comorbid condition or age > 65 yr was twice as likely to be present in the nonsurvivors as in the survivors.

View larger version (21K): [in this window] [in a new window]   Figure 2.   Mortality based on the etiology of acute lung injury in 107 medical intensive care unit patients.

In the univariate analysis, all of the primary chronic conditions under investigation, except cirrhosis, were associated with at least a 1.2-fold greater RR of dying in the hospital, although statistical significance at the p < 0.05 level was achieved only for organ transplantation (Table 2). The presence of any one of these underlying comorbidities was associated with an increased mortality (Table 1). For the other chronic conditions, only steroid use approached statistical significance for the ARDS cohort (Table 2). Patients over the age of 65 yr were 1.45 times (1.56 times for the ARDS group) as likely to die as those 65 yr of age or younger. The presence of both older age and a primary comorbidity increased the RR of dying among all patients to 5.0 (95% CI: 2.0 to 12.5; p < 0.0001), and in the ARDS group to 7.1 (95% CI: 1.9 to 25.0; p < 0.0001). Among patients  65 yr old who did not have a primary comorbidity, 84% and 90%, respectively, of those with ALI and ARDS survived, as compared with only 20% and 25% of those with ALI and ARDS, respectively, who were older than 65 yr and who had a primary comorbidity. Both septic and aspiration-related etiologies were associated with an increased RR of death during hospitalization, with only the former in the ARDS cohort reaching statistical significance (Table 2). The RR of dying in the hospital in the presence of any preceding nonpulmonary OSD, APACHE II score > 20, APS > 15, or LIS > 2.5 was not increased.

With multiple logistic regression, cirrhosis, HIV infection, active malignancy, organ transplantation, age > 65 yr, and sepsis were all independent predictors of hospital death in patients with ALI, whereas in the ARDS cohort, cirrhosis and active malignancy failed to achieve statistical significance (Table 3). Nonpulmonary OSD developing between hospital admission and the onset of mechanical ventilation, aspiration of gastric contents as an etiology, APACHE II score or APS, and LIS measured during the first 24 h of ALI/ARDS were not found to be independently associated with increased mortality in either group. This was the case whether the APS and LIS were analyzed as continuous variables or whether threshold values were used. When the regression analysis was reconfigured to exclude age and comorbid conditions, the APACHE II score became an independent predictor of mortality. In contrast (under the same conditions), when the APS was substituted for the APACHE II score, it was not an independent predictor of outcome.

When other widely prevalent comorbidities, including COPD, steroid use, alcohol or intravenous drug abuse, and diabetes mellitus were analyzed with respect to their impact on mortality, none reached statistical significance in either the univariate or the multivariate analysis (Tables 2 and 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The principal findings of this study were that comorbid conditions, age > 65 yr, and sepsis are important independent predictors of hospital death in MICU patients with ALI, 75% of whom satisfied the criteria for ARDS. In contrast, after controlling for these independent predictors, severity of illness, severity of lung injury, gastric aspiration as an etiology of ALI, and preceding nonpulmonary OSD were not independently predictive of outcome.

A number of studies, using univariate analyses, have identified conditions associated with increased mortality in ALI and ARDS. These conditions include older age (3, 5, 10, 17); cirrhosis (2, 14); HIV infection (15); bone-marrow and solid-organ transplantation (2, 16); active malignancy (2, 5); and etiology of lung injury (6, 10, 11, 17, 25, 26). By finding mortality rates of  67% for cirrhosis, active malignancy, HIV infection, organ transplantation, sepsis, and gastric aspiration, our univariate analysis done exclusively with MICU patients is in general agreement with the majority of the studies cited earlier of mixed-patient (i.e., combined medical and surgical) or single-disease populations.

Previous investigators have demonstrated that nonpulmonary OSD following the onset of ARDS has a profound negative impact on survival (5, 7). Recently, Doyle and coworkers found preceding organ dysfunction in 21% of medical and surgical ALI patients, and its presence was associated with an 88% mortality rate (2). Although our criteria for OSD were similar, we used the interval from hospital admission to the onset of mechanical ventilation, rather than to ICU admission, as the period for occurrence of such dysfunction. Although nonpulmonary OSD was identified in 46% of our MICU patients, the associated mortality rate, in contrast to that in previous studies, was not increased with univariate analysis. There are several reasons why preceding nonpulmonary OSD may not be associated with increased mortality. First, although our criteria were similar to those of other investigators, the low threshold values (i.e., creatinine > 2.0 mg/dl, bilirubin > 2.0 mg/dl, etc.) set in our study raise the possibility that although organ dysfunction was present, it was milder than in previous studies. The potential impact of this is supported by studies showing that a multiorgan-system-dysfunction score, with increasing points given for increasing dysfunction, more accurately predicts outcome than does a single threshold value (27). Second, because mortality rates differ for individual organ failures, different distributions of OSDs in different studies may affect their combined influence on mortality. Moreover, preceding OSD may be more important in surgical ICU patients, a group excluded from our analysis.

Measures of lung injury severity and of severity of illness have variable success in predicting ARDS-associated outcome. A number of reports have found a direct association between poor oxygenation and mortality (3, 10, 12, 25), whereas other studies have failed to identify a correlation (2, 5). When tested prospectively, increasing severity of lung injury (using LIS score), measured at 24 h, 48 h, and 72 h, was not associated with increased mortality (2). Our findings confirm that survival in MICU patients is independent of the LIS determined within the first 24 h of mechanical ventilation. Similarly, although some have found that APACHE scores predict outcome for ARDS (4, 6), this has not been confirmed by other investigators (28). APACHE II score was only an independent predictor when age > 65 yr and comorbid conditions were excluded from our multivariate model. In contrast, when the acute physiology component is substituted for the APACHE II score, it is not an independent predictor, irrespective of whether age and comorbidities are included in the model. In other words, it is the contribution of age and certain comorbidities, rather than the acute physiology component, that make the APACHE II score a predictor of outcome. One potential limitation of our APACHE II-score analysis is that we used data for the first 6 h of mechanical ventilation rather than for the first 24 h of ICU stay. We believe the impact of this to be minor, because 75% of our patients were intubated on the first ICU day (91% within the first 3 d), and the APACHE II score measured after the first day in mechanically ventilated patients appears to be an accurate predictor of outcome (31). It is also possible that our study population size did not provide sufficient statistical power to show an independent effect of acute physiologic derangement, given the very strong effect of comorbidity.

Although numerous univariate analyses of factors associated with increased mortality rates for ALI and ARDS have been published, few studies have used multivariate techniques to identify truly independent predictors of mortality. Moreover, for the most part, the investigations using univariate analyses have focused exclusively on physiologic variables (8) or have used parameters present several days after the onset of ARDS (3, 6). When comorbidities have been included, the roles of individual conditions have not always been specified (6). The most complete multivariate analysis to date is that done by Doyle and colleagues, who examined 123 patients with ALI (2). Their use of multiple logistic regression showed cirrhosis, sepsis, and preceding OSD, but not LIS (determined on Days 1, 2, and 3) to be independently associated with increased hospital mortality. Our findings confirm both the independent effects of cirrhosis and sepsis on ALI outcome, and the failure of the 24-h LIS to predict hospital mortality. In contrast, we identified additional independent factors governing ALI outcome, including organ transplantation (primarily bone-marrow transplantation), HIV infection, and the presence of active malignancy. In further distinction to Doyle and colleagues, we did not find that preceding OSD achieved statistical significance in the multivariate model. Although similarities between the studies exist, including overall mortality rate, average age, and the percentage of patients with sepsis as the cause for ALI and ARDS, important differences are present. Most significantly, we studied only MICU patients, rather than a mixed medical-surgical ICU population. We also controlled for age and generalized severity of illness (APS), finding the former, but not the latter, to be an important independent predictor of hospital outcome. Interestingly, Doyle and colleagues found no difference in mortality in comparing patients on the basis of a Pa_O2/FI_O2 threshold value of 150. Similarly, we found the same 58% mortality among patients with ARDS and those with ALI but without ARDS with the use of a Pa_O2/FI_O2 threshold of 200. Unfortunately, the small number of patients in the latter group and large number of independent variables studied precluded analysis with multiple logistic regression.

The findings of the current study have several important clinical implications. First, we have identified factors present at the onset of respiratory failure, including age > 65 yr and a number of comorbid conditions, that should be considered to ensure balanced assignment of patients in investigations of new therapies for lung injury. Failure to account for these factors may lead to erroneous conclusions about the efficacy or lack of efficacy of the therapeutic modality under investigation. For example, a number of recent studies, although noting exclusion criteria, have not specified the proportions or allocation of patients with cirrhosis, active malignancy, HIV infection and organ transplantation (19, 32). Given the high mortality associated with these conditions, failure to account for their presence may be important when their prevalence is high and/or study size is small (35). Similar considerations exist in interpreting studies that analyze time trends in mortality. Although such analyses show reductions in mortality after controlling for age and etiology, they do not consider possible changes in the prevalence of comorbid conditions (11). On the basis of earlier published univariate analyses showing poor outcome for ALI and ARDS, one might anticipate an increase in the proportion of patients with cirrhosis, HIV infection, organ transplantation, and active malignancy opting to forego mechanical ventilation. If this proves true, then recent improvement in mortality rates may in part reflect decreases in the prevalence of these conditions among patients with ALI and ARDS. As in recent studies of new therapies for ALI and ARDS (19, 32) and those showing a decrease in mortality over time (11), our mortality rate, in the absence of a chronic condition, was only 35% (only 16% when age  65 yr). Conversely, recent advances in the management of patients with ALI and ARDS would appear even more impressive if decreased mortality trends occurred in the setting of increasing prevalence of these conditions. Additionally, it has been suggested that gender affects the outcome of mechanical ventilation, with women having a higher mortality (36). In contrast, among MICU patients with acute lung injury, we found mortality in men and women to be similar.

In conclusion, using multivariate techniques, we found that independent predictors of hospital mortality in exclusively MICU patients with ALI (75% of whom satisfied the criteria for ARDS) include cirrhosis, the presence of HIV, active malignancy, organ transplantation, age > 65 yr, and sepsis as a cause of ALI. Preceding nonpulmonary OSD, APS, LIS, aspiration as the etiology, and other chronic diseases were not found to be independent predictors of hospital mortality. These factors must be considered when analyzing results of trials of new treatment modalities for lung injury. Additionally, these factors, and their changing prevalence, must be taken into account when interpreting trends in ALI and ARDS mortality.