Lieuwe D. Bos Olaf L. Cremer David S. Y. Ong Eliana B. Caser Carmen S. V. Barbas Jesus Villar Robert M. Kacmarek Marcus J. Schultz On behalf of the MARS consortium External validation confirms the legitimacy of a new clinical classification of ARDS for predicting outcome Accepted: 6 July 2015 Ó Springer-Verlag Berlin Heidelberg and ESICM 2015 A complete list of members of the MARS consortium is given in the ‘‘Acknowledgments’’ section. Electronic supplementary material The online version of this article (doi:10.1007/s00134-015-3992-x) contains supplementary material, which is available to authorized users. Dear Editor, The diagnosis of acute respiratory distress syndrome (ARDS) is largely pragmatic and based on the presence of hypoxia and bilateral infiltrates on the chest X-ray that are not fully explained by cardiac failure or vol- ume overload within 1 week of a known clinical insult [1]. Risk strati- fication by categorization into mild, moderate, and severe ARDS was an important aspect of the Berlin defini- tion for ARDS [1]. However, robustness of the PaO 2 /FiO 2 thresh- olds as suggested in the Berlin definition was not confirmed by external validation [2]. It has been proposed to include additional infor- mation, collected later in the course of ARDS, to improve mortality pre- diction [3]. Recently, the ALIEN network investigators reported on a categorization based on both PaO 2 / FiO 2 and positive end-expiratory pressure (PEEP) thresholds, measured 24 h after the initial ARDS diagnosis, for improved hospital mortality pre- diction [3]. We assessed the external validity of this new classification system. Consecutive adult patients admit- ted to the mixed intensive care units (ICU) of two university hospitals in the Netherlands and 14 hospitals in Brazil were prospectively screened for the incidence of moderate and severe ARDS [1, 4, 5]. In short, patients were classified into four groups based on PaO 2 /FiO 2 and PEEP measured at onset of clinical symp- toms and 24 h later (Table 1). Primary outcome was all-cause hos- pital mortality; secondary outcomes were 30-day mortality (for method- ological details, see the online supplementary material). A total of 519 subjects had at least a single episode of moderate or severe ARDS while in the ICU with a length of stay greater than 24 h (Fig. E1). Overall hospital mortality was 42.5 % in the Netherlands (Table E1) and 57.5 % in Brazil (Table E2). Using the thresholds at onset of ARDS for categorization, there were no differ- ences in mortality between the four groups (Table 1). Reclassification with data collected 24 h after onset of ARDS showed clear differences between the groups (Table 1; Fig. E2), albeit that hospital and 30-day mortality rates in group I were higher than in group II. A post hoc analysis with an adapted PaO 2 /FiO 2 cutoff of 100 mmHg, in line with the Berlin definition threshold, showed a similar signal only for 30-day mor- tality (Table E3). The results of our analysis confirm and extend the recent findings by the ALIEN network investigators who showed that reclassification by PEEP and PaO 2 /FiO 2 24 h after the diag- nosis of ARDS could be used to stratify for hospital mortality. Indeed, our findings also suggest that ARDS is not a homogeneous complication that can be simply categorized at its Table 1 Distribution and outcomes of each subset of patients with the acute respiratory distress syndrome Group I Group II Group III Group IV P for trend Odds ratio group IV vs. group I PaO 2 /FiO 2 C150 and PEEP \ 10 PaO 2 /FiO 2 C150 and PEEP C10 PaO 2 /FiO 2 \ 150 and PEEP \ 10 PaO 2 /FiO 2 \ 150 and PEEP C10 At onset of ARDS (N) 90 91 109 227 All-cause hospital mortality 48.9 % 44.0 % 41.2 % 45.4 % 0.68 OR 0.95 [0.6–1.6] 30-day mortality 34.4 % 34.1 % 33.0 % 37.4 % 0.54 OR 1.30 [0.8–2.3] After 24 h (N) 128 177 57 155 All-cause hospital mortality 43.8 % 35.0 % 42.1 % 58.1 % 0.002 OR 1.33 [0.8–2.1] 30-day mortality 30.5 % 24.9 % 38.6 % 50.3 % 0.00002 OR 1.75 [1.1–2.9] Intensive Care Med DOI 10.1007/s00134-015-3992-x LETTER