Published ahead of print on February 13, 2003, doi:10.1164/rccm.200210-1196OC
© 2003 American Thoracic Society Acute Respiratory Distress Syndrome after Bacteremic Sepsis Does Not Increase MortalityDepartment of Medicine, Infection Control Program; Department of Anesthesiology, Pharmacology, and Surgical Intensive Care, Division of Surgical Intensive Care, University of Geneva Hospitals, Geneva, Switzerland Correspondence and requests for reprints should be addressed to Correspondence and requests for reprints should be addressed to Didier Pittet, M.D., M.S., Infection Control Program, University of Geneva Hospitals, 1211 Geneva 14, Switzerland. E-mail: didier.pittet{at}hcuge.ch
To determine whether acute respiratory distress syndrome (ARDS) complicating bacteremic sepsis independently affects mortality in critically ill patients, we conducted a 3-year retrospective cohort study in a surgical intensive care unit. We included all consecutive patients with blood culture–positive sepsis and measured organ dysfunctions and mortality. Among 4,530 admissions, 196 cases of bacteremic sepsis were recorded. ARDS occurred in 31 (16%) of these patients. The case fatality rate was 58% in patients with ARDS compared with 31% in patients without ARDS. Using Cox proportional hazards regression with time-dependent variables, the unadjusted hazard ratio for death was 1.8 (95% confidence interval [CI], 1.0–3.2). After adjusting for comorbid factors that were present before the onset of sepsis, the hazard ratio was 2.2 (95% CI, 1.2–3.9). After further adjustment was made for nonpulmonary organ dysfunctions and microbiologic factors that were independently associated with mortality, the adjusted hazard ratio for ARDS was 0.6 (95% CI, 0.3–1.2). Among critically ill surgical patients, ARDS complicating bacteremic sepsis remains common, but it is not independently associated with short-term mortality, after adjusting for severity of illness and nonpulmonary organ dysfunctions evolving after the onset of sepsis.
Key Words: acute lung injury • critical care • prognosis • respiratory failure • sepsis Sepsis remains an important cause of morbidity and mortality among hospitalized patients. Acute respiratory failure after the onset of sepsis is common and can be associated with adverse outcomes (1–4). To explain the high case fatality rate among critically ill patients with sepsis and acute respiratory distress syndrome (ARDS), investigators have hypothesized that the prognosis may be determined by the severity of underlying conditions that also predispose patients to ARDS and by simultaneous, nonpulmonary complications, usually related to multiple organ failure (4, 5). Russell and coworkers (6) have suggested in a prospective, observational study that although pulmonary dysfunction is a frequent complication among septic patients, the severity of lung injury and hypoxemia may not be correlated with an increased risk of mortality. In contrast, Martin and Bernard stated in a recently published textbook chapter that respiratory dysfunction related to sepsis is associated with significant independent mortality (7). To test the hypothesis that severe pulmonary dysfunction complicating sepsis has no independent impact on mortality in critically ill patients, we conducted a 3-year retrospective cohort study using survival regression methods and examined the independent effect of ARDS on mortality in 196 critically ill patients with blood culture-positive sepsis.
Study Cohort This study was conducted at the surgical intensive care unit of the University of Geneva Hospitals, which is a 22-bed referral unit, admitting over 1,500 patients per year (8). The source population consisted of all patients admitted between June 1, 1994, and May 31, 1997. Patients were included if they had microbiologically confirmed bacteremia or fungemia with signs of sepsis, that is, "bacteremic sepsis" (9).
Data Collection and Definitions The time of onset of bacteremic sepsis was defined as the time when the first positive blood culture was drawn. Bacteremic sepsis was defined as a systemic inflammatory response syndrome (see online supplement) associated with positive blood cultures, the latter being characterized by the presence of viable microorganisms (e.g., bacteria or fungi) in the blood (9). Pneumonia required the presence of new radiographic infiltrates, coupled with identification of a significant amount (more than 103 cfu/ml) of at least one pathogen in a culture of bronchoalveolar lavage fluid and clinical evidence of infection (14). Beginning at the first day of blood culture-positive sepsis, we recorded organ dysfunctions according to the type and number of days of dysfunction, based on pre-established definitions (15–17). ARDS was defined according to established criteria as an acute onset respiratory failure with presence of bilateral diffuse chest infiltrates, a PaO2/FIO2 of 200 mm Hg or less with a pulmonary wedge pressure of 18 mm Hg or less, and the presence of a recognized predisposing factor for ARDS (16, 18). Definitions of other organ failures can be found in the online supplement.
Statistical Analysis
Among 4,530 admissions, 196 cases of blood culture-positive sepsis were recorded (incidence, 4.3%). The causative pathogens are shown in Table E1 of the online supplement. The mean Acute Physiology and Chronic Health Evaluation II score at admission to the unit was 21.4 ± 8.4; the mean age of study subjects was 56.9 ± 17.1 years. Forty-eight patients (24.5%) were women. The median length of stay in the intensive care unit was 11 days (interquartile range, 5 to 22). The majority of the patients (n = 153, 78%) underwent surgery before admission to the intensive care unit. Cardiovascular disease (n = 48, 24%), gastrointestinal disease (n = 42, 21%), and multiple trauma (n = 38, 19%) were the most frequent reasons for admission. ARDS occurred in 31 bacteremic patients (16%). As expected, bacteremic patients with ARDS had greater comorbidity and severity of illness (Table 1) . The case fatality rate was 58% (18 of 31) in patients with ARDS compared with 31% (51 of 165) in patients without. Figure 1 illustrates survival curves for both cohorts. The unadjusted hazard ratio for death was 1.8 (95% CI, 1.0–3.2; p = 0.03 by the log-rank test).
The follow-up period for the 196 patients represented 2,087 patient days. During that time interval, at least one organ dysfunction was present during 890 patient days. The timing and cumulative prevalence of the different organ dysfunctions are illustrated in Figure E1 in the online supplement. The most frequent organ dysfunctions in terms of organ-failure days were neurologic dysfunction (460 days, 22%), ARDS (327 days, 16%), renal failure (211 days, 10%), and septic shock (182 days, 9%). Among the 31 patients with ARDS, respiratory failure was associated with no other organ dysfunction in 2 patients, with one additional, nonpulmonary organ dysfunction in 8, with two or three nonpulmonary organ dysfunctions in 12, and with four or more nonpulmonary organ dysfunctions in 9 patients. After adjusting for baseline factors and comorbidities (Table 2 , model 1), ARDS was still associated with an increased mortality rate (adjusted hazard ratio, 2.2; 95% CI, 1.2–3.9). We then included microbiologic variables and nonpulmonary organ dysfunctions that were associated with mortality on bivariate analysis (Table 2, model 2). After including these variables in a multivariable survival model, the risk of death was significantly correlated with neurologic dysfunction (hazard ratio, 6.7; 95% CI, 3.4–13.0), shock (hazard ratio, 4.2; 95% CI, 2.4–7.5), cardiac dysfunction (hazard ratio, 2.2; 95% CI, 1.2–4.3), and the score of the Simplified Acute Physiology Scoring II system at onset of sepsis (hazard ratio per one-point increase, 1.03; 95% CI, 1.01–1.05). Appropriate antimicrobial therapy was independently associated with decreased mortality (hazard ratio, 0.5; 95% CI, 0.3–0.8). After adjusting for the previously mentioned independent risk factors for mortality, ARDS was not associated with an increased risk of death (adjusted hazard ratio, 0.6; 95% CI, 0.3–1.2). In a separate analysis, we confirmed these results for the lack of association between ARDS and mortality after excluding 29 patients (15%) who were censored or died within 2 days after onset of bacteremic sepsis (adjusted hazard ratio, 0.7; 95% CI, 0.3–1.3).
To assess whether the effect of ARDS on survival was affected by possible interactions with other nonpulmonary organ dysfunctions, we added several interaction terms to the Cox regression model. For instance, for the interaction of ARDS with neurologic failure, the p value of the interaction term was 0.65. The multiplicative terms for other interactions were also not statistically significant (e.g., interaction of ARDS with cardiac failure, p = 0.23; interaction of ARDS with liver failure, p = 0.74; interaction of ARDS with shock, p = 0.33).
Despite a large body of research about ARDS, no study has specifically assessed, after adjustment for evolving organ dysfunctions and severity of illness, the independent short-term mortality of patients with ARDS complicating bacteremic sepsis. Our results confirm the hypothesis that ARDS, although it may be a good marker of the severity of the underlying disease, is rarely the direct cause of death in patients with sepsis. These data suggest that patients with sepsis who subsequently develop ARDS do not die from respiratory failure but rather from nonpulmonary organ dysfunctions, such as cardiac or neurologic failure leading to refractory shock and multiorgan failure (4, 6, 21). Thus, improvements in outcome of patients with sepsis and with ARDS may depend more on treatment of sepsis and multiorgan failure than on oxygenation measures and ventilation strategies (22). The association between acute respiratory failure and mortality has been extensively studied in patients with mechanical ventilation and ARDS, showing contradictory results (4, 23–25). Several authors found that the presence of ARDS with sepsis leading to mechanical ventilation was independently associated with mortality (24, 26–28). Others showed that death of septic patients with ARDS is more frequently associated with refractory sepsis and multiorgan failure, suggesting that the degree of respiratory system failure and severity of lung injury are not independent predictors of outcome (1, 4, 5, 21, 22, 29, 30). For instance, Ferring and Vincent (22) looked at 129 patients who developed ARDS with a case fatality rate of 52% and found that sepsis and multiorgan failure was the primary cause of death (49%), followed by respiratory failure (16%). Our results suggest that death in septic patients cannot be attributed to the development of respiratory failure but to the severity of underlying illness and to the nonpulmonary organ failures developing after the onset of sepsis. It is to note, however, that the 95% CIs surrounding the effect of ARDS on the risk of death in our study overlap that reported by other studies (24, 27, 28). For instance, in the study performed by Esteban and colleagues (24), the odds ratio of mortality after ARDS was 1.4 (95% CI, 1.0–2.0). According to the limited number of observations in this article, we cannot completely refute the results obtained previously. Long-term evaluation of critically ill patients with ARDS complicating sepsis has been performed in at least two studies (27, 31), showing again incoherent results. Perl and colleagues (27) found that development of ARDS after suspected Gram-negative sepsis independently increased the long-term likelihood of death (hazard ratio, 2.3; 95% CI, 1.1–4.8). In contrast, Davidson and colleagues (31) assessed in a prospective, matched cohort study the effect of ARDS on long-term survival and concluded that if sepsis or trauma patients survive to hospital discharge, ARDS does not independently increase their risk of subsequent death (hazard ratio, 1.0; 95% CI, 0.5–2.1). Our study results complement the findings by Davidson and colleagues (31), as in the latter study no information was given about in-hospital mortality and nonpulmonary organ dysfunctions evolving after the onset of sepsis. In further distinction to Davidson and colleagues, we included detailed patient-level data about microbiologic factors and appropriateness of antimicrobial therapy in the final Cox regression model. Hitherto, reported biologic data focusing on ARDS after bacteremic sepsis have been limited (3, 32). Meduri and colleagues showed that for patients with ARDS secondary to sepsis, the likelihood of death was correlated with the degree of pulmonary and extrapulmonary inflammation, estimated by measuring levels of different cytokines in bronchoalveolar lavage fluids and plasma (33, 34). Thus, elevated systemic inflammatory reaction may be causally related to outcome in septic patients with ARDS (3). Conversely, clinical studies confirmed that nonpulmonary organ dysfunction due to systemic inflammation markedly decreases survival in patients with ARDS (23). Several features may have contributed to the low adjusted risk of death in patients with ARDS evolving after onset of bacteremic sepsis. Attending physicians in charge of the unit have a large experience with ARDS management (35–37). Furthermore, new strategies were progressively introduced in the management of these patients during the last 10 years, such as protective lung ventilation (36), less sedation, and early detection of ventilator-associated pneumonia (38). Moreover, the institution is very active in trying to improve the quality of critical care (39, 40). There are several limitations to this study, most importantly, the relatively small number of patients with ARDS included. However, this report represents one of the largest reported studies of ARDS complicating blood culture-positive sepsis. Second, this experience may not be representative for patients with culture-negative sepsis or patients in other settings such as medical or pediatric intensive care units. Nevertheless, the comorbidities were diverse and likely represent a patient population encountered in other surgical intensive care units. Third, we cannot exclude that nonpulmonary organ dysfunctions related to systemic infection started already more than 24 hours before blood culture positivity. Future studies may use better indicators of early systemic infection to assess the interacting effects of different organ dysfunctions at the earliest possible moment (41). Finally, our study design did not allow us to examine in detail the hypothesis that lung inflammation due to ARDS may contribute to nonpulmonary organ dysfunctions. However, we looked at the interactions between ARDS and the development of important organ dysfunctions and did not retrieve statistically significant interaction terms. Despite these limitations, our study presents a new perspective on the effect of ARDS on risk of death in septic patients and provides a strong quantitative basis on this subject. Our results refute the hypothesis that severe lung injury after bacteremic sepsis is associated with significant independent mortality. Important strengths of our study were the availability of detailed data on evolving organ dysfunctions and the use of advanced regression methods with time-dependent covariables, which allowed adjustment for the time interval between onset of bacteremic sepsis and discharge or death. Overall, our results suggest that although ARDS after bacteremic sepsis remains a common complication, it may not be independently associated with short-term mortality.
The authors thank Peter Rohner for assistance in retrieving microbiologic data, Nadia Colaizzi for help with data management, and Rosemary Sudan for editorial assistance.
Supported by an educational grant from the Department of Anesthesiology, Pharmacology and Surgical Intensive Care, University of Geneva Hospitals. This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Received in original form October 18, 2002; accepted in final form February 11, 2003
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
