The key challenge faced in the analysis of these data was the application of statistical procedures to demonstrate the independent prognostic information provided by the measurement of pulmonary artery pressure. Given the high degree of colinearity observed among the various physiologic measures of cardiac structure and function performed as part of this study, multivariate techniques were required. As others have noted, however, statistical “control” in multivariate models often does not provide adequate tests of independence, especially when the degree of precision of measurement of the variables in the model is unequal or unknown. buy flovent inhaler
To address this problem, two basic analytic strategies were applied. First, the standard multivariate regression models were employed and interaction terms were generated based on the outcomes of the analyses that used first-order terms. Second, the sample was stratified on one variable (eg, ejection fraction) while outcome analysis was performed with a second variable (eg, pulmonary artery pressure). Only when the outcome was concordant for all analyses were the results presented. While use of these multiple analytic approaches add complexity to the presentation of the results, this procedure increases the degree of confidence that can be ascribed to the conclusions.