Analyses were performed with SAS version 9.1 (SAS Institute, Inc., Cary, North Carolina). In preliminary analyses, all variables were examined for missing values and outliers with cross (e.g., scattergrams) and range checks. Continuous variables were tested for normality with the Kolmogorov-Smirnov D statistic as well as visual inspection of the univariate distributions. Chi-square statistics were used to determine difference in proportions between categorical variables. Descriptive statistics (mean, SD) were generated for the demographic characteristics. To compare differences before and after surgery for normally distributed continuous variables, Student t tests for paired data were performed. The Wilcoxon signed-rank test, the nonparametric equivalent, was used to assess changes in the following variables: height, age, heart rate, Ea ratio, and Ea-sept. Statistical significance was determined at a p value of <0.05. Spearman’s and Pearson’s correlations were calculated to further determine which factors were associated with measures of change in cardiac structure and change in cardiac function after weight loss. Specifically, the dependent variables were change in: indexed LVM, E/A, Ea/Aa-lat, E/Ea-lat, LA size, and indexed LA size. The variables included in correlation analysis as potential independent risk factors were: gender, race, and change in: weight, BMI z-score, BMI, kilograms over ideal body weight (IBW), percent over IBW, heart rate, systolic blood pressure z score (SBP-z), diastolic blood pressure z score (DBP-z), indexed LVM, IVSd, LVPWd, LVED, RWT, SF, and VCF difference. Multivariate regression models were built with the results from Pearson’s analysis to explore potential determinants of change in indexed LVM. These models were built by selecting either clinically or statistically significant independent variables from the univariate correlation analyses. The following independent variables were included in the model: gender, race, change in percent over IBW, change in SBP-z, and change in DBP-z. As an additional analysis, the pre-operative values for indexed LV mass and LV geometry were added to the aforementioned independent variables. Analysis of covariance was performed to further elaborate upon the results of the regression analyses, where appropriate. No regression models were built for changes in diastolic function, because there were no significant correlations in univariate analyses.