In this multicentre survey, we found that ALI/ARDS occurred in approximately 19% of all ICU admissions. Although the incidence is high, it is in keeping with recent well designed epidemiological studies that have estimated the incidence of ALI/ARDS to be 15% to 26% in patients mechanically ventilated for more than 24 hours [9, 10]. Because mechanical ventilation was included in our definition of ALI/ARDS, which is not a requirement in the AECC criteria, it is likely that this is a conservative estimate of the incidence of ALI/ARDS.
The crude ICU mortality for ALI/ARDS was lower compared with recent European findings, which estimate the mortality of ARDS at 49% to 53% [9, 11]. The mortality rate in this study is in keeping with mortality reported in ARDSnet studies that have utilized protective lung ventilation [12–14]. The lower mortality in this study may reflect the use of a lower Vt strategy. The mean Vt in this study was 7.0 ml/kg (approximately 8.4 ml/kg predicted body weight), as compared with a mean Vt in the European ALIVE study of 8.3 ml/kg (approximately 10.0 ml/kg predicted body weight)  and 9.2 ml/kg (approximately 11.0 ml/kg predicted body weight) in a Scottish study . These data suggest advances in the ventilatory management of ALI/ARDS have translated into practice, which is contrast to evidence that many centres still do not apply protective lung ventilation strategies [15, 16].
Consistent with previous studies, organ dysfunction predicts mortality in ALI/ARDS [9, 17, 18]. We found the PaO2/FiO2 ratio at diagnosis of ALI/ARDS to be associated with mortality. This is contrary to most [19–21] but not all  previous reports in ALI/ARDS. The reports demonstrating no association with PaO2/FiO2 ratio are older and predate the era of protective lung ventilation. With low Vt ventilation the aim of mechanical ventilation is limitation of ventilator-associated lung injury rather than correction of hypoxaemia. As a result, oxygenation may be more likely to reflect severity of pulmonary dysfunction and outcome. This is supported by the finding there is no correlation between Vt/kg and PaO2/FiO2 ratio at diagnosis of ALI/ARDS, which suggests that clinicians do not react to worse oxygenation by using higher Vts. It is also important to consider that the PaO2/FiO2 ratio is highly variable, depending on the ventilatory strategy employed as well as the distribution of ventilation-perfusion, which is influenced by the pattern of mechanical ventilation (including PEEP and I:E ratio) . It is relevant that the initial PaO2/FiO2 ratio was predictive of outcome in several large clinical trials that implemented a protocolized ventilatory strategy [12, 23]. In the present study, although ventilatory strategy was not protocolized, and in particular PEEP/FiO2 combinations were not mandated, one possible explanation for the finding that the initial PaO2/FiO2 ratio was associated with mortality is that ventilatory strategy was sufficiently standardized, which is supported by the close adherence to protective lung ventilation strategy. However, it should also be recognized that physiological indices such as cardiac output may also affect the PaO2/FiO2 ratio and may have influenced the relationship between PaO2/FiO2 ratio and mortality. Finally, it is acknowledged that these data relate to PaO2/FiO2 ratio and Vt on the day of diagnosis of ALI/ARDS, and the influence of any subsequent alteration in Vt on the relationship between P/F ratio and mortality is unknown.
Vt was not associated with mortality in this study, which is not surprising given the relatively good adherence to lung protective ventilation, with only 5% of patients of patients receiving a tidal volume above 10 ml/kg. This is in accordance with previous trials of protective lung ventilation, in which benefit was only seen where there was the largest difference between the lower and higher Vt strategies [12, 24].
Plateau pressure was not related to mortality in this study, in contrast to data describing a direct relationship between plateau pressure and mortality [25, 26]. Similarly to the contention that there is no relationship with Vt, this may reflect adherence to lung protective ventilation, with only 12% of patients having a plateau pressure above 30 cmH2O. Furthermore, because plateau pressure is determined by multiple factors, including Vt (with possible harmful effects as Vt increases) and PEEP (with possible beneficial effects as PEEP increases ), it is possible that this finding reflects opposing influences of Vt and PEEP in lung injury. We did not define how plateau pressure was to be measured. In addition, 75% of patients received pressure-controlled ventilation, in which inspiratory pressures may overestimate plateau pressure. It is possible that these factors influenced the relationship between plateau pressure and mortality.
Contrary to our findings, age [9, 11, 17] and duration of ventilation pre-ALI/ARDS  have been associated with increased mortality in ALI/ARDS. It is possible that as a result of selection bias elderly patients who were admitted to ICU were less severely ill and had less co-morbidity, which confounded the correlation of age with mortality in our study. Although in some studies risk factors for ALI were found to be associated with mortality [19, 27], this has not been a universal finding [9, 28], and no such relationship existed in our study. It is more likely that the severity of physiological derangement and organ dysfunction, as in our findings, are more important determinants of outcome.
Accurate epidemiological data are essential to provide pilot data as well as to inform the design and feasibility assessment of clinical trials. These prospective data from an unselected cohort of critically ill patients with ALI/ARDS provide important information to inform phase III clinical trials.
In a study in which mortality is the primary outcome, overestimating the mortality rate in the power calculation will result in the study being under-powered. If the mortality rate of 49% from European epidemiological data  were used to determine the required sample size for an ALI/ARDS study in Ireland, where the mortality is lower, this would result in a large and expensive phase III study being significantly under-powered. Another important consideration that may result in overestimation of the current mortality rate for ALI/ARDS is that the European ALIVE study  was undertaken in 1999, before the publication of definitive evidence regarding the role of low Vt ventilation, and subsequent studies showing low adherence to protective lung ventilation [11, 15, 16] were undertaken relatively soon after publication of the ARDSnet study findings that confirmed benefit. With more widespread adherence to protective lung ventilation, as suggested in this current study, it might be expected that mortality is lower. This has important implications for multicentre studies being performed across Europe.
Furthermore, if the expected mortality benefit from pilot data is small, then powering a study with mortality as the primary outcome may not be possible, because the sample size would be too large to allow recruitment within an acceptable time period. On the basis of such factors, it may be necessary to consider other important clinical end-points such as ventilator-free days in order to design an feasible trial. In addition, these data will inform the decision regarding whether a clinical study could successfully be undertaken nationally or would require multinational trial group co-operation in order to recruit the numbers required.
In undertaking multicentre trials it is essential to demonstrate standardized care between centres. It is reassuring that in this unselected population the standardization of ventilatory strategy was already apparent, with use of lower Vts, and that outcomes are in keeping with international standards [13, 14].
There is experimental and preclinical evidence that statins  may be beneficial in ALI/ARDS. The present study supports an association with improved outcome in patients who received statin therapy. The patients who received statins reflected the general population included in this study. These data are interesting and valuable in terms of generating hypotheses. However, it must be emphasized that the numbers of patients receiving statins were small and that subsequent randomised clinical trials examining statins in ALI/ARDS will be required. A major limitation is that no information was collected on the specific statin or dose administered, which may be important in determining possible benefit. In addition, no information is available on how many patients not receiving statins during their ICU admission had received statins before admission. A further confounding issue is that sicker patients would perhaps be less likely to receive their statin because of potential concerns about increased toxicity or failure of the enteral route in this group.
There are a number of limitations to our study. With the diagnosis of ALI/ARDS based on AECC criteria, it is recognized that misclassification may occur because of misinterpretation of chest radiographs and exclusion of left atrial hypertension on clinical assessment only. However, misclassification is no more likely in this study than in similar epidemiological studies using these current definitions. Data on co-morbidities were not collected. The influence of co-morbidities on overall mortality as well as the potential effect of statins may be important, but this cannot be quantified from the data presented here. Additionally, height was not recorded, and therefore it is not possible to calculate predicted body weight accurately to adjust Vt. For comparison with the ARDSnet protective lung ventilation study actual body weight has been estimated at 20% greater than predicted body weight, and approximate values for Vt/predicted body weight have been presented . Data were collected over a 10-week period and seasonal variation in the incidence of ALI/ARDS, although unlikely, cannot be excluded.