Statistical analysis

Normality was assessed using the Shapiro-Wilk statistical test. Continuous variables with normal distribution were expressed as means with standard deviations, and these variables were compared using the t-test. The chi-square test or Fisher’s exact test was used to assess categorical variables. The Mann-Whitney test was used to compare nonnormally distributed continuous data, which were expressed as medians with interquartile ranges (IQRs).

Given that patients with severe symptoms are more susceptible to underlying comorbidities, propensity score (PS) matching and overlap PS weighting were performed to address potential confounding factors. A PS between New York Heart Association (NYHA) classes I/II and III/IV was estimated using a multivariable logistic regression model incorporating patient age, sex, body mass index (BMI), hypertension, diabetes mellitus, atrial fibrillation (AF) history, lung disease, stroke history, hepatitis C, functional TR, Ebstein’s Anomaly, cardiac surgery history, ejection fraction (EF), and isolated TVR. PS matching was performed using the nearest neighbor matching algorithm with a caliper size of 0.2 logit PS standard deviation units and a 1:1 ratio, after which the overlap PS weighting method was performed, with the weight of each patient corresponding to the probability of that patient being assigned to the opposite group in this method (8). Overlap PS-weighted logistic regression models were used to analyze the association between symptoms severity and the probability of long-term survival, as well as other clinical outcomes. After PS matching and overlap PS weighing, the covariate balance was assessed using standardized mean differences, with values less than 0.2 reflecting adequate balance. We used the Kaplan-Meier method and the log-rank test to compare long-term follow-up survival rates between the NYHA I/II group and the NYHA III/IV group. Overall survival among NYHA classes was compared using unweighted and weighted log-rank tests. The adjustment covariates were those used to create the PS models, which were modeled using unadjusted and covariate-adjusted Cox regression models.

All tests were two-tailed, and a P value <0.05 was considered significant. Analyses were performed using R, version 4.0.4 (https://www.r-project.org).