Alternative approaches to analyzing ventilator-free days, mortality and duration of ventilation in critical care research.

Abstract

To discuss the strengths and limitations of ventilator-free days and to provide a comprehensive discussion of the different analytic methods for analyzing and interpreting this outcome. Using simulations, the power of different analytical methods was assessed, namely: quantile (median) regression, cumulative logistic regression, generalized pairwise comparison, conditional approach and truncated approach. Overall, 3,000 simulations of a two-arm trial with n = 300 per arm were computed using a two-sided alternative hypothesis and a type I error rate of α = 0.05. When considering power, median regression did not perform well in studies where the treatment effect was mainly driven by mortality. Median regression performed better in situations with a weak effect on mortality but a strong effect on duration, duration only, and moderate mortality and duration. Cumulative logistic regression was found to produce similar power to the Wilcoxon rank-sum test across all scenarios, being the best strategy for the scenarios of moderate mortality and duration, weak mortality and strong duration, and duration only. In this study, we describe the relative power of new methods for analyzing ventilator-free days in critical care research. Our data provide validation and guidance for the use of the cumulative logistic model, median regression, generalized pairwise comparisons, and the conditional and truncated approach in specific scenarios.