Results

There were no significant differences between the two groups in terms of age (40.6 (3.3) versus 45.5 (2.2) years, p = 0.22), APACHE score (14.5 (0.72) versus 13.6 (0.1), p = 0.2), Lung Injury Score (2.95 (0.07) versus 2.8 (0.1), p = 0.22), Pao ₂/Fio ₂ ratio (11.8 (0.67) versus 12.0 (0.6), p = 0.13), incidence of multi-system organ failure on admission (29% (12/41) versus 32% (25/78), p = 0.5), or incidence of sepsis on admission (31% versus 39%, p = 0.40). There was a significantly lower proportion of men in group 1 (51% versus 74%, p = 0.01; table 1)

The case mix of the two groups was closely matched (postoperative 48% versus 48%, trauma 17% versus 16%, primary lung injury 12% versus 24% for groups 1 and 2, respectively, p = 0.23). Patients in group 1 were ventilated using several different modes (volume preset, non-inverse ratio, n = 15; pressure controlled, PC-IRV, n = 11; ultra high frequency jet ventilation (UHFJV), n = 13; extracorporeal or intracorporeal gas exchange (ECGE), n = 2). All the patients in group 2 received PC-IRV but, in addition, some received other support techniques (UHFJV, n = 4; ECGE, n = 2).

There was a highly significant reduction in mortality between patients in group 1 (66%) and group 2 (34%; p = 0.0037; relative risk, 1.77; CI 1.23 to 2.55).

In group 1 the presence of sepsis on admission, seen in 21% of survivors and 56% of non-survivors, was the sole predictor of death (p = 0.05). No significant difference in mortality was observed between patients receiving volume-controlled ventilation and the remainder. In group 2 age (41.2 (2.6) versus 52.6 (3.5), p = 0.01), APACHE II score (12.2 (0.6) versus 15.8 (0.9), p = 0.001), and Pao ₂/Fio ₂ (12.8 (0.86) versus 10.5 (0.72), p = 0.04) for survivors and non-survivors, respectively, were significant predictors of survival. Sepsis (survivors 47% versus non-survivors 31%, p = 0.2) and multi-organ failure (survivors 36% versus non-survivors 41%, p = 0.2) at the time of admission were not predictive of survival.

Discussion

Recent publications3 have suggested that a dramatic reduction in the mortality associated with ARDS may have occurred in recent years and the results of this study seem to confirm this impression. The authors acknowledge that there was a slight loosening of the oxygenation entry criteria in 1994 following publication of the consensus guidelines. However, the two cohorts were not significantly different with regard to lung injury score, APACHE II score, and incidence of multi-organ failure or sepsis on admission, and the case mix was almost identical in terms of precipitating condition which is known to influence outcome in ARDS. In particular, the mean Pao ₂/Fio ₂ ratio of the two groups was almost identical. Thus, the change in entry Pao ₂/Fio ₂ ratio cannot explain the reduction in mortality.

Examination of possible predictive factors for mortality was revealing. Age, APACHE II scores, and Pao ₂/Fio ₂ ratios which predicted survival in group 2 were not significant in group 1. Moreover, the presence of sepsis at admission was a significant predictor of death in group 1, yet was not in group 2. Increased awareness of the dangers of sepsis in recent years has led to advances in the prevention and diagnosis of nosocomial infection10which suggests that strategies to prevent this common complication of critical illness have improved in recent years. However, within group 2 47% of survivors displayed evidence of sepsis at the time of admission compared with 31% of non-survivors (p>0.05), which suggests that better management of sepsis may also have been an important factor in improving outcome.

The predictive power of indices of the severity of illness (APACHE II) and respiratory failure (LIS, Pao ₂/Fio ₂) was surprising in that neither have had clear prognostic significance in other published series.2 Moreover, the majority of patients with ARDS do not die of respiratory insufficiency but rather of multiple organ failure,11 the incidence of which was not predictive of mortality in this study.

The optimal mode of ventilatory support for patients with ARDS has been the subject of considerable debate in recent years. Specifically, a recently published randomised controlled study has shown improved outcome in patients with ARDS using a low volume, pressure-limited strategy aimed at alveolar recruitment compared with a traditional preset volume-controlled ventilatory approach.4 Our ventilatory strategy has certainly changed in recent years. All patients in group 2 were supported using PC-IRV with only six receiving other support techniques. In contrast, patients in group 1 received a variety of ventilatory and other support techniques. However, within group 1 there were no differences in survival between those who received a traditional volume controlled strategy and those who received other support techniques. Unfortunately, as the various modes of support were not allocated as part of a structured protocol over the eight year period, it is not possible to draw reliable conclusions about the efficacy of any particular mode in this study. Other therapeutic support systems such as inhaled nitric oxide, turning patients prone, and (late) administration of corticosteroids have also been widely used on an individual patient basis. Their potential contribution to the improved overall mortality figures in patients in group 2 cannot therefore be meaningfully assessed.

This study has clearly shown that, within our unit, there has been an improvement in outcome associated with ARDS over an eight year period in patients well matched for disease severity and case mix. Unfortunately no clear answers as to why this should be so have emerged from our data or, indeed, that published by others. We suggest that the improvement is most likely to be multifactorial in origin, attributable both to better general patient management strategies—particularly those of ventilation and sepsis—and to the use of newer therapeutic strategies such as inhaled nitric oxide, prone positioning, and late administration of corticosteroids.