Statistical analyses
Normality of the data was assessed by the Kolmogoroff–Smirnoff test and data were expressed as mean (standard deviation) and median (interquartile range) for normally and non-normally distributed data, respectively. Categorical variables were expressed as number (percentage). The groups were compared using the unpaired Student t-test/Mann-Whitney or chi-square (χ2) for numerical and categorical data, respectively.
Logarithmic transformation was performed for non-parametric data and multilevel linear mixed-effects models were used to compare the groups. The fixed effect component included time (the three study visits), study group, maternal age, race, smoking, gestational age at the time of echocardiography, development of GDM and first-order interaction between time and study group. A higher numbers of cases compared to controls were chosen in order to produce more precise effect measures, also considering that echocardiographic data of pregnant women with normal BMI has been previously reported; 8 our sample size reflected a paucity of data in obese pregnant women. Considering a 4:1 ratio for cases and controls, our study had >80% power (at alpha=0.05) to detect an 1829ml/min difference in CO, 0.12 difference in E/A ratio and 65g difference in LV mass between groups.14 Statistical analyses were performed using International Business Machines (IBM) Statistical Package for the Social Sciences (SPSS) for Windows 2019, version 26.0, IBM Corp., Armonk, New York, USA. Differences were considered statistically significant at p<0.05.