Construction and validation of a prediction model for in-stent restenosis following coronary stent implantation during dual antiplatelet therapy

Figures

• 

  Figure 1. The study design and the selection procession of patients with CAD.

• 

  Figure 2. Risk factors selected using the LASSO model. Optimal parameter (lambda) selection for the LASSO model was cross validated using the minimum criterion. Partial likelihood deviation (binomial deviation) curves vs. log (lambda). Dotted vertical lines are drawn at the best values of 1SE (1-SE criterion) using the minimum criterion and the maximum criterion.

• 

  Figure 3. LASSO coefficient profiles for the 34 candidate clinical features. As the regularization parameter increases, the coefficients are progressively shrunk toward zero. At the optimal lambda value selected via 10-fold cross-validation, exactly six predictors retain non-zero coefficients, while the others are eliminated. This penalization process effectively simplifies the model and ensures the identification of the most robust clinical predictors.

• 

  Figure 4. A nomogram to predict the probability of ISR in patients with stent implantation after a PCI. The nomogram included the factors of lower ADP inhibition rate, lower AA inhibition rate, age, the reaction time of thromboelastograph, N-terminal pro-brain natriuretic peptide, and the coagulation index of thromboelastography.

• 

  Figure 5. ROC curves for validating the discrimination power of a nomogram.

• 

  Figure 6. A histogram of the area under the receiver operating characteristic curve (AUC) values derived from 1,000 bootstrap internal validation replicates of the ISR prediction nomogram. This histogram displays the distribution of AUC values across 1,000 bootstrap resamples performed for internal validation. The x-axis represents the AUC value obtained in each bootstrap replicate; the y-axis indicates the frequency (number of replicates). The distribution approximates a normal distribution with a peak (mode) at approximately 0.75 and a mean of 0.758 (95% confidence interval 0.709–0.807). Importantly, 94.6% of the bootstrap AUC values fell within the narrow range of 0.70–0.80, demonstrating excellent stability and robustness of the model's discriminative performance.

• 

  Figure 7. Calibration plots of the nomogram for the probability of PCI patients with ISR.

• 

  Figure 8. Decision curve analysis for the ISR prediction nomogram