Statistical analysis
Continuous variables and categorical variables were summarized as mean ± standard deviation (SD) and count (percent) respectively. The Society of Thoracic Surgery (STS) Risk Score for each patient was calculated using the online STS calculator [29]. Propensity score matching techniques were used to control for the difference in the baseline covariates. The propensity score was estimated using a multivariable logistic regression model with biological sex as the dependent variable and the baseline characteristics as covariates including age, BMI, pulmonary disease, cerebrovascular disease, renal disease, current smoker, past smoker, hypertension, dyslipidemia, liver disease, gastrointestinal disease, malignancy, peripheral vascular disease, diabetes mellitus, congestive heart failure, prior MI, prior percutaneous coronary intervention (PCI), prior coronary artery bypass grafting (CABG), prior AF/flutter, implant type, pump time, cross-clamp time, STS score and LVEF. Greedy matching techniques without replacement and a caliper width equal to 0.2 of the standard deviation of the logit of the propensity score were applied to match male patients 1:1 to female patients. Standardized mean difference was used to evaluate the balance before and after matching. A standardized difference of 0.1 or less was deemed to be the ideal balance.
The paired t-test was used to compare the pre- and post-operative echocardiographic measurements and the absolute changes in LV size, LA size, and LA volume index between male and female patients. Cox proportional hazards regression models and the Fine & Gray model [30] were implemented to determine the hazard ratios of sex difference on the primary and non-fatal secondary outcomes. McNemar test was performed to compare the postoperative complications by sex. The survival curve was plotted for all-cause mortality at longest follow-up using Kaplan-Meier methods. The interaction between sex and implant type for all-cause mortality at longest follow-up was tested by adding implant type and an interaction term of sex and implant type into the Cox model. Reverse Kaplan-Meier methods were used to estimate the median follow-up time. Statistical analyses were executed using the SAS 9.4. (SAS Institute, Cary NC). A p-value <0.05 was deemed of statistical significance. All statistical tests were two-sided.