Impact of MetS on Long-Term Prognosis Among STEMI Patients Treated with pPCI—Ten-Year Follow-Up Study

Figure 1. Kaplan–Meier cumulative event rate curves for the three primary outcomes. Cumulative event rates [1 − S(t)] for all-cause mortality (A), MACE (B), and hospitalization for heart failure (C), stratified by metabolic syndrome status. Shaded bands show 95% pointwise confidence intervals; the vertical dotted line marks the 60-month timepoint. p-values from the log-rank test are shown in each panel. Number-at-risk tables are shown beneath each panel for both groups. Abbreviations: MetS—metabolic syndrome; MACE—major adverse cardiovascular event; HF—heart failure.

Figure 2. Kaplan–Meier curves stratified by the number of metabolic syndrome criteria met. Cumulative event rate curves for all-cause mortality (A), MACE (B), and heart failure hospitalization (C), stratified by the number of MetS criteria met (3, 4, or 5) and compared with the reference group of patients with no criteria met (dashed gray line, n = 12). Log-rank p-values for each subgroup versus the reference group are shown in each panel. Number-at-risk tables for all four subgroups are shown beneath each panel. Abbreviations as in Figure 1.

Figure 3. Aalen–Johansen cumulative incidence function (CIF) for heart failure hospitalization under the competing risk of death: (A) MetS-positive (n = 216); (B) MetS-negative (n = 290). Solid lines, Aalen–Johansen CIF; dashed gray lines, naive Kaplan–Meier estimator (1 − S(t)) for comparison. Error bars at 10 years represent 95% bootstrap confidence intervals (1000 resamples). Number-at-risk tables are shown beneath each panel.

Figure 4. SHAP attribution analysis of the XGBoost equivalent classifier for the three primary ten-year endpoints. Top 15 features are shown for each endpoint, ranked by mean absolute SHAP value computed on a stratified random subsample of 200 patients (random seed 42). (A) Beeswarm plots display the distribution of patient-level SHAP values; each dot represents one patient, with horizontal position indicating the magnitude and direction of the feature’s contribution to the model prediction. Color encodes the original feature value (red, high; blue, low; and gray, intermediate or missing), revealing the direction of the relationship between each feature and the predicted probability. (B) Mean absolute SHAP values with 95% bootstrap confidence intervals (1000 resamples), quantifying the overall importance of each feature and providing a measure of statistical uncertainty. MetS-related features are shown in bold.

Figure 5. Random Survival Forest permutation feature importance. Permutation importance for MACE (A), all-cause mortality (B), and heart failure hospitalization (C), reported as the drop in the concordance index when each feature is individually permuted in a Random Survival Forest fitted to the full analytical sample. Metabolic syndrome features are shown in orange. The number of MetS criteria ranks first for heart failure prediction (Δc-index = +0.035), providing convergent evidence with the SHAP analysis (Figure 4B, where MetS criteria ranked second by mean |SHAP|). The cross-validated Harrell’s c-index for each outcome is shown in the panel header. Abbreviations as in Figure 4.