With a maximum enrollment of 1000 patients and 4 planned interim analyses, the study had statistical power of 90.7% to detect a 2.25-day increase in VFDs, assuming a mean of 14 VFDs and standard deviation of 10.5. The study was monitored using a group sequential design with asymmetric stopping boundaries for efficacy and futility designed using alpha and beta spending boundaries (eMethods ).19 An independent data and safety monitoring board (DSMB) conducted an analysis of serum n-3 levels after enrollment of the first 60 patients, a safety analysis after enrollment of 100 patients, and an interim analysis after enrollment of 272 patients. Although VFDs was the primary end point, the DSMB was advised to consider mortality in their decision to stop the trial for either efficacy or futility.
Means and standard deviations are reported for baseline continuous variables, and counts and percentages are reported for baseline categorical variables, with differences assessed using t tests and χ2 (link) tests, respectively. Plasma levels of IL-6, IL-8, leukotrienes, and urinary isoprostanes were log transformed and compared using analysis of variance with baseline levels as covariates. Categorical outcome variables are reported as percentages with 95% confidence intervals. The continuous outcome variables (VFDs, ICU-free days, and organ failure–free days) are reported as means and standard deviations, with differences assessed using analysis of variance controlling for baseline shock and enrollment group of the EDEN study.
Logistic regression controlling for baseline shock and randomization group of the EDEN study was used to analyze mortality. Adjusted mortality rates were calculated using 7 baseline mortality-predicting covariates derived from a previous study of similar populations20 (link): age, Acute Physiology and Chronic Health Evaluation III (APACHE III) score, plateau pressure, missing plateau pressure, number of organ failures, and the alveolar-arterial difference in PaO2 value. Proportion curves over time were plotted for survival and unassisted breathing.
All analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) on an intention-to-treat basis, with 2-sided P≤.05 considered significant. P values were not corrected for multiple comparisons or early stopping.
Means and standard deviations are reported for baseline continuous variables, and counts and percentages are reported for baseline categorical variables, with differences assessed using t tests and χ2 (link) tests, respectively. Plasma levels of IL-6, IL-8, leukotrienes, and urinary isoprostanes were log transformed and compared using analysis of variance with baseline levels as covariates. Categorical outcome variables are reported as percentages with 95% confidence intervals. The continuous outcome variables (VFDs, ICU-free days, and organ failure–free days) are reported as means and standard deviations, with differences assessed using analysis of variance controlling for baseline shock and enrollment group of the EDEN study.
Logistic regression controlling for baseline shock and randomization group of the EDEN study was used to analyze mortality. Adjusted mortality rates were calculated using 7 baseline mortality-predicting covariates derived from a previous study of similar populations20 (link): age, Acute Physiology and Chronic Health Evaluation III (APACHE III) score, plateau pressure, missing plateau pressure, number of organ failures, and the alveolar-arterial difference in PaO2 value. Proportion curves over time were plotted for survival and unassisted breathing.
All analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) on an intention-to-treat basis, with 2-sided P≤.05 considered significant. P values were not corrected for multiple comparisons or early stopping.
