© 2009 American Thoracic Society doi: 10.1164/rccm.200811-1712ED
Changing Mortality in Acute Respiratory Distress Syndrome? Yes, We Can!
Réanimation Médicale
Réanimation Polyvalente In an article that appears in this issue of the AJRCCM (pp. 220–227), Phua and coworkers (1) performed an extensive search of the literature to assess whether acute respiratory distress syndrome (ARDS) mortality has decreased over time. Their review suggests that mortality has not changed since the 1994 definition of ARDS adopted by the American-European Consensus Conference (AECC). Such disappointing results contradict commonly held views that are based on the following. (1) Experimentally, high volume ventilation produces lung injury (2, 3). (2) Clinically, reduction of tidal volume reduces ARDS mortality (4). (3) Several single-center studies have reported reductions in ARDS mortality since 1993 (5–8). (4) A recent large review of the literature did find a mortality reduction of 1.1% per year between 1994 and 2006 (9). To understand why the conclusions of the present review go in the opposite direction, several hypotheses can be raised. The methodology used by Phua and colleagues (1) has obvious limitations. When mortality ranges between 30 and 85% in observational and randomized controlled studies published after 1994, one wonders whether pooling such heterogeneous studies makes sense. For instance, a study by Bauer and coworkers had a mortality rate of 80%, but required that patients meet two different definitions for ARDS and be hospitalized for at least 48 hours before enrollment (10). Lu and colleagues reported a 68% hospital mortality in patients with ARDS from Shanghai with a mean age of 60 (11), whereas the mean age was 45 in a study from San Francisco by Kallet and coworkers that reported a 32% hospital mortality in patients with either acute lung injury (ALI) or ARDS (7). Gattinoni and colleagues (12) studied patients with either ALI or ARDS, with a mean SAPS II of 37 and a mortality of 28%. Ferguson and coworkers (13) described a series of patients with ARDS with a much higher SAPS II (47) and, not surprisingly, a higher mortality rate (60.2%). Therefore, major heterogeneity exists among the selected studies with respect to severity of illness, health system organization, and criteria for enrollment. A multivariate analysis was used to adjust for such variables. However, mean values instead of individual data were used and missing severity scores were extrapolated in 26% of the studies, using multiple imputations. A second confounding factor is the absence of description of lung morphology in most studies. The initial distribution of aeration loss critically impacts late prognosis: patients with a diffuse and homogeneously distributed loss of lung aeration have a mortality greater than or equal to 70%, whereas patients with focal loss of aeration have a mortality rate less than or equal to 40% (14). The respective (and unknown) proportion of "focal" and "diffuse" ARDS in each study may explain the wide variability in mortality rate. A third source of confusion concerns the definition of ARDS, which does not take into account positive end-expiratory pressure and FIO2 at which PaO2/FIO2 is measured, and causes large variations in ARDS incidence among centers. Sensitivity of the AECC definition for predicting pathologic and computerized tomography patterns characteristic of ARDS is around 75%. This implies that at least 25% of patients in each study did not suffer from significant lung injury and that the final outcome in this subgroup had nothing to do with ARDS. Last but not least, the variable number of available studies each year may affect the evolution of mortality rate over time. As shown in Figure 4 of the article (1), mortality peaked in 1998 and 2001, two years in which a single randomized controlled trial with mortality rate greater than or equal to 50% and two observational studies with unusually high mortality were published. (In 1998, one of the four observational studies, described above, reported a mortality rate > 80% [10]; in 2001, another study mentioned among the five observational studies reported approximately 70% mortality, a result attributed by the authors themselves to inappropriate respiratory care [11].) Inclusion of these results may have overestimated mortality in 1998 and 2001, thereby concealing the decrease in mortality reported in Zambon's analysis which, interestingly, did not include the two observational studies with such high mortality (9). If one admits, however, that there is no strong signal indicating a decrease in mortality from ARDS, then different hypotheses must be considered. A major factor that should have reduced mortality is pressure and volume reduction, raising the difficult question as to when changes in practice might have taken place. Very few clinicians took heed and changed clinical practice on the basis of the experimental study of Webb and Tierney in 1974 (2). Likely, more clinicians became aware of the risks of high tidal volume ventilation after the impressive experimental studies by Dreyfuss and coworkers (3), the landmark paper of Hickling and colleagues in patients (15), and the concept of biotrauma and atelectrauma developed by Slutsky (16) during the decade 1985–1995. This recognition of risk is shown by a consensus conference held in 1993 that recommended limiting plateau pressure in clinical practice (17). Many clinicians started changing their practice after the 2000 NIH ARDS Network study (4), whereas others did not change anything. It is therefore difficult to determine a precise cutoff time for changes in clinical practice. Because of missing information on ventilator settings in many published studies, it is also impossible to conclude whether the lack of change in mortality since 1994 could reflect delayed modification of clinical practice to reduce tidal volume or, possibly, changes in management of ventilators taking place before 1994. Mortality may also have remained unchanged because of an increased incidence of sepsis-induced ARDS associated with too-liberal fluid administration. The results of this study may be also related to our current practice in treating critically ill patients. The heroic times in which intensivists could only do one thing—save life—are over. Recent improvements regarding processes of critical care have mainly concerned better recognition of potential deleterious effects of our therapies, strategies, and attitudes. Using overly aggressive ventilatory modalities, sedation, antibiotics, colloids, or insulin therapy may be detrimental. Taking an opposite attitude may not be better, suggesting that titration of therapy for individual patients is necessary. We may improve outcome with one therapeutic intervention but worsen it with other excessive or inappropriate treatments. This could be reflected in the lack of change in mortality. Caring for the critically ill is a complex and difficult task. The clinician must be aware that what matters lie in the details, God as well as the devil. For the years to come, we can remain optimistic if we are more and more patient-centered, pragmatic rather than dogmatic, less invasive, and concerned about safety. FOOTNOTES
Conflict of Interest Statement: L.B.'s research laboratory has received research grants during 2006, 2007, and 2008 from Drager ( REFERENCES
Related articles in AJRCCM:
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
20,000), General Electric (
