To the Editor:

Results of the meta-analysis by Dr. Phua and colleagues, as reported in their article (1), contrast with our analysis (2), which showed a decrease in mortality from ALI/ARDS over time between 1994 and 2006. There are indeed some methodological differences between the analyses. First, distinction between interventional and epidemiological studies is crucial. Studies in which patients are carefully selected cannot be included in the same analysis as those where no selection is performed. For example, three observational studies (3–5) had exclusion criteria that resulted in a study population with lower mortality rates; these studies were conducted in the 1990s, so may have biased the analysis of change in mortality over time. Second, in controlled trials the mortality rate of the control group, which receives the current standard of care, is more representative of general outcome at that time than that of the intervention group. Phua and coworkers used pooled mortalities (control + treatment), justifying this decision with the fact that in most studies there was no significant difference between groups. We disagree. For example, pooling mortality rates from the control and treatment groups of the ARDS Network trial comparing low and higher tidal-volume ventilation (6) would give a rate not representative of that period. Third, Phua and coworkers did not analyze the “slope” of mortality rates over time, which may more correctly assess any potential trend. In addition, it may be preferable to evaluate variations in mortality rates grouped, for example, in blocks of two years (as in our review) to avoid excessive heterogeneity.

Phua and coworkers observed that we included studies with overlapping populations. We excluded articles which stated explicitly that the study population had been used in another publication, but acknowledge that in three cases populations may have overlapped. They also suggested that some of the differences in our results could be explained by our inclusion of three studies conducted mostly before 1994. However, these studies used inclusion criteria that were identical to the AECC definitions; we excluded older studies that used other criteria.

Phua and coworkers used a broader selection process, and did not limit their search to English-language publications. However, their analysis had methodological limitations. We stand by our results and believe there has been a gradual improvement in outcomes for patients with ARDS. However, we agree with Phua and coworkers that mortality rates remain high and there is still “a great need for future effective therapeutic interventions for this still highly lethal syndrome” (1).

References

Section:

ChooseTop of pageReferences <<CITING ARTICLES