High Positive End-expiratory Pressure and Mortality in Acute Respiratory Distress Syndrome

Abstract

Thank you for giving us the opportunity to respond to the communication by Dr. David Pestaña. The author is correct in his assertion that most of the studies do not define the terms high positive end-expiratory pressure (PEEP) and low PEEP and that the appropriate level of PEEP has been selected on the basis of oxygenation and peak/plateau airway pressures.1–3This reflects the practical difficulties inherent in recruiting large numbers of patients into clinical trials using highly individualized criteria. We are therefore left with the mean values recoded in each of the groups to infer the threshold values for “low” and “high” PEEP. Nevertheless, the underlying physiologic principles are clear and suggest that there are several biologic benefits associated with selecting PEEP levels between 10 and 15 cm H2O in patients with severe acute lung injury. This is particularly so during the early stages of the illness, when lung edema is maximal and therefore the tendency for cyclical opening and collapse of alveolar units is maximal.1–5The author is also correct in stating that the most appropriate level of PEEP in a given patient can only be determined through an individualized titration protocol. He raises a pertinent point in his final statement that “it is unlikely that any attempt to demonstrate the superiority of a ventilatory strategy will be conclusive.” We agree entirely and would like to pose the question of whether the current emphasis on the need to demonstrate significant improvements in final outcome–based endpoints (mortality, duration of stay, duration of mechanical ventilation, and so forth) is appropriate for evaluating new ventilation strategies.Ventilation is a supportive measure needed in the management of other systemic illnesses such as sepsis, acute lung injury/acute respiratory distress syndrome, systemic inflammatory response syndrome, and heart failure. Clinical outcome in such patients is usually a manifestation of the underlying disease process itself or the “mediator variables.” Ventilation, in this context, is best seen as a “moderator variable” that alters the quantitative relation between disease severity and its consequence (mortality and or morbidity). Improvements in ventilation strategies can, therefore, have only a modest impact on disease specific mortality. As iatrogenic contributions to mortality (such as excess sedation, barotrauma/volutrauma, and ventilator-induced lung injury) are recognized and rectified, it becomes inevitable that further improvements in ventilator technology will require an unrealistic sample size to demonstrate mortality/morbidity benefits based on the basic principles of diminishing returns.5Such large numbers cannot be recruited within a geographically, culturally, and economically homogeneous area or during a reasonable time period during which clinical practices remain comparable across several other domains. More importantly, it is well recognized that interactions between organ systems in humans are nonlinear, and the importance of such nonlinearity in critical illness was highlighted elegantly by Buchman6and Rixen et al. 7If we agree on the most fundamental premise that the initial manifestations and subsequent development of a disease state are governed by nonlinear interactions between the severity of the initial insult (the mediator variables), host's physiologic responses, and other moderator variables (such as ventilation, secondary infections, iatrogenic complications, and nutritional status), it follows that each patient would follow a unique trajectory as dictated by nonlinear dynamics. In such nonlinear systems, the final clinical outcome (survival, death, or prolonged morbidity) is unpredictable and is sensitively dependant on the initial conditions (the mediator variables) and subsequent modulator variables. It does not follow simple rules based on linear assumptions. That is, a “small change” in one of the moderator variable does not always lead to a “small change” in the final outcome. Such “unpredictable” events occurring (in the control or treatment arms) in clinical trials involving moderator variables, with a relatively modest influence on the overall disease process, will necessarily lead to conclusions that are difficult to reproduce and at times erroneous. Therefore, the current emphasis placed on clinical outcome alone reflects a mind-set (promoted by the business world) that is rooted in cost–benefit analysis and aims to identify and support only those interventions with a relatively large effect size. This approach, if adopted blindly and dogmatically, is likely to lead to the abandonment of several interventions that may be beneficial to individual patients.Estimating the qualitative and quantitative improvements to patient care that can be achieved by refining moderator variables (such as ventilation), in our view, requires the adoption of more dynamic models as suggested by Dr. Pestaña, rather than the final clinical outcome alone.*Manchester Royal Infirmary and Manchester Academic Health Sciences Centre, Manchester, United Kingdom. m.nirmalan@manchester.ac.uk