Page 47 - Assessing right ventricular function and the pulmonary circulation in pulmonary hypertension Onno Anthonius Spruijt

Statistical analysis
Continuous data are presented as mean ± SD if the Kolmogorov-Smirnov test showed a normal distribution; otherwise, data are presented as median ± interquartile range (IQR). Categorical variables are expressed as frequency and percentage. Comparisons between groups were performed using 2-sided Student t tests with adjustment for unequal variance as needed. For non- normally distributed variables such as NT-pro-BNP level, transformation to the common logarithm was performed before analysis. Linear regression analysis was used to determine independent associations between hemodynamic and structural or functional right heart parameters. The association between clinical and echocardiographic parameters and outcome was analyzed using Cox proportional hazards models. The assumption of proportional hazards was assessed by plotting the scaled Schoenfeld residuals for each independent variable against time; these correlations were found to be nonsignificant for all variables included in the multivariable model. We used hierarchical modeling to determine factors independently associated with outcome and chose to include at maximum 1 covariate per 10 events to minimize overfitting of the model. We avoided including in the model variables that were collinearly related to each other. We used bootstrapping with 5000 iterations to estimate hazard ratios and bias-corrected 95% confidence interval (CI) for the multivariable models. For building the predictive score, the smallest absolute beta coefficient was assigned a value of 0 and values for subsequent variables were assigned on the basis of multiples of their respective beta coefficients to the nearest 0.5 approximation for categories with significantly different beta coefficients [16]. The survival c-statistic was calculated to show the discriminatory ability of the models and was used to compare the predictive score with the validated REVEAL score and National Institutes of Health (NIH) survival equation. Intraobserver variability was assessed using the average difference in absolute measurement and the intraclass correlation coefficient (ICC). Statistical analysis was done using PASW statistical program (version 18.0, SPSS Inc., Chicago, Illinois).
Results
Study population
Of 128 patients with idiopathic and drug- and toxin-associated PAH who were seen during the study period, 106 were enrolled in the prospective registry. Eleven patients were excluded from the study for the following reasons: unavailable echocardiogram (n=2), atrial fibrillation (n=1), lost to follow- up (n=5), left heart failure (n=2), and restrictive lung disease (n=1).
Chapter 3
45
3