Statistical analysis
Data distributions were checked for normality before further analysis with the Shapiro–Wilk test. Continuous data are presented as the mean and standard deviation (SD) or median and interquartile range (IQR). Unpaired t-tests or Wilcoxon rank sum tests were used for statistical comparisons. Categorical data are presented as proportions and were compared using the chi-squared tests.
A generalized linear model (logit model) and linear model was used to assess the effect of multiple variables on in hospital mortality and post-operative LOS. Candidate explanatory variables (n=10) included were: MetS/no-MetS, age, sex, left ventricle ejection fraction (LVEF), renal failure, chronic obstructive pulmonary disease (COPD), redo surgery, atrial fibrillation (AF), peripheral vascular disease and previously treated coronary artery disease. In order to examine the effect of MetS stratified for each of the cohort of treatments (MVS, SAVR and TAVR), interaction term was added to the model (MetS*Treatment), keeping mitral intervention as reference level.
Linearity assumptions were checked (cran.r-project.org/web/packages=sjPlot).
As sensitivity analysis, regression model was also built including the risk factors incorporated in the MetS definition such as: BMI>30 kg/m2 and as continuous variable, systemic hypertension, atherogenic dyslipidaemia and insulin resistance.
The models were also tested for multicollinearity with variance inflation factor (VIF), and explanatory variables with VIF higher than 5 were excluded since considered poor regression estimates (cran.r-project.org/web/packages=VIF). Results were presented as adjusted Odds Ratio (adj. OR) and beta-coefficients and 95% confidence intervals (Cis) (models diagnostic in Supplementary Material ).
30-, 60- , 90- and 120-days in hospital-survival probability was also analysed using the Kaplan-Meier and corresponding survival curves were built by plotting all observations. Comparisons of survival estimates for two different patient strata (overall MetS vs. no-MetS) were performed with the log-rank statistic
In between centres mortality variability was also evaluated and pooled mortality proportion (overall MetS vs. no-MetS) plotted along with the prediction interval using a random effects model (cran.r-project.org/web/packages=meta).
All statistical analyses were performed with RStudio Team (2020) (RStudio: Integrated Development for R. RStudio, PBC, Boston, MA, USA.