The trusted source for
healthcare information and
In a multicenter randomized trial, 79 patients with the acute respiratory distress syndrome (ARDS) received either pressure-controlled ventilation (PCV) or volume-controlled ventilation (VCV), using the same plateau pressure goals. The patients received ventilation with nearly identical plateau pressure, tidal volume, and total positive end-expiratory pressure (PEEP). Hospital mortality was significantly higher (78% vs 51%) in the VCV group, and there were also more extra-pulmonary organ failures (average 3.7 vs 2.6, due to an increase in renal failure) in the VCV patients. A multivariate analysis showed that the mode of ventilation was not associated with mortality, and Esteban and colleagues conclude that the mode of ventilation was not the cause of the increased mortality (Esteban A, et al. Chest 2000;117:1690-1696).
Regardless of how little evidence there is to address many clinical questions in critical care, it is difficult to imagine an intensivist without strong opinions. We all bring clinical experience and preconceptions based on our reading of the existing literature to every new study. As Thomas Kuhn pointed out in The Structure of Scientific Revolutions, scientists sometimes go to great lengths to make their data fit their paradigms.
One has the sense that Esteban et al struggled with this problem in trying to interpret the results of their novel study. Esteban et al accepted the "low stretch" ventilator strategy by targeting a plateau pressure of 35 cm H2O in all patients with ARDS and tried to identify the ideal mode of mechanical ventilation by randomizing patients to PCV and VCV. When total PEEP and plateau pressure are kept constant, the only real difference between the modes is the inspiratory flow pattern. Could this subtle difference in mechanical ventilation really account for the 35% reduction in mortality and the marked reduction in organ failure in the PCV group?
One of the first places we look in randomized trials when the results seem a bit strange is Table 1. In this article, Table 1 provides some disquieting information. Although the patients were balanced with regard to severity of illness score and organ failure, there were almost twice as many patients in the VCV group with renal failure at baseline. Perhaps the study turned out the way it did because of an "unlucky" randomization? This is a common concern in clinical trials and can be addressed simply by performing an "adjusted" analysis that incorporates baseline severity of illness. Esteban et al tried to provide this analysis which, they argue, showed that the increased mortality in the VCV group was due to differences in organ failure. Unfortunately, they performed the wrong analysis and, therefore, readers can conclude relatively little from the study results.
When Esteban et al are concerned about an unlucky randomization, they can address this concern fairly simply with an "adjusted" analysis that controls for baseline differences in the patients. Differences that occur after randomization should not be included in the adjusted analysis because they may be caused by the therapy. For example, imagine a study of patients with septic shock that showed a higher mortality in patients treated with corticosteroids. Controlling for differences in baseline severity of illness in this study would be fine. However, an analysis that controlled for infection at any time over the study period would obscure any effect that corticosteroids exerted by increasing infections. Just like in this hypothetical example, Esteban et al relied on an analysis that controlled for "the presence of two or more extrapulmonary organ failures over the study period" to argue that VCV did not increase mortality. The correct way to present the results of this study would be to present the treatment effect adjusting only for baseline differences in organ failure. Their analysis, which adjusts for organ failure after randomization, obscures any influence VCV might exert through its effect on organ failure during the course of ARDS.
The truth is, I share Esteban et al’s skepticism that VCV caused the observed increase in mortality and organ failure in this study. My paradigm is that the mode of ventilation has little or no effect on outcome in ARDS if plateau pressure, tidal volume, PEEP, and inspired oxygen concentration are all the same. Since the article comes to the same conclusion, maybe I should stop there? Unfortunately, Esteban et al arrive at this conclusion through a flawed analysis that might obscure either a benefit or a harm of PCV. This article won’t change my practice, but it challenges the way I have thought about the effects of mechanical ventilation.