Incremental prognostic and risk-stratification value of ultrasound-based Carotid Plaque Reporting and Data System (Plaque-RADS) classification for recurrent ischemic stroke in patients with low-to-moderate-grade carotid stenosis

Background: The Carotid Plaque Reporting and Data System (Plaque-RADS) is a system for stroke risk assessment. However, the clinical utility of ultrasound (US)-based Plaque-RADS remains unclear. This study aimed to assess the incremental prognostic value and risk stratification utility of US-based Plaque-RADS in patients with prior stroke and low-to-moderate-grade carotid stenosis. Methods: Between January 2020 and December 2022, 231 patients with carotid plaque ipsilateral to the prior stroke and low-to-moderate-grade carotid stenosis were retrospectively screened, of whom 158 were included in the final analysis. US-based Plaque-RADS classification was assigned based on duplex US assessment of plaque features. Cox regression and subgroup analyses were performed to evaluate the association between high-risk US-based Plaque-RADS categories and recurrent ischemic stroke. Model discrimination was assessed using receiver operating characteristic curves and area under the curve (AUC), and clinical utility was evaluated using decision curve analysis. The association between increasing US-based Plaque-RADS categories and the risk of recurrent ischemic stroke was assessed by trend analysis. The incremental risk-stratification value of refined high-risk Plaque-RADS thresholds was quantified with net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Internal validation was performed using bootstrap resampling. Results: During the 3-year follow-up period, recurrent ischemic stroke occurred in 33 patients. High-risk US-based Plaque-RADS categories were significantly associated with recurrent ischemic stroke in patients with prior stroke and low-to-moderate-grade carotid stenosis, after adjusting for smoking, homocysteine, and intraplaque neovascularization [adjusted hazard ratio: 3.56; 95% confidence interval (CI): 1.07–11.83, P=0.038]. No statistically significant interactions were observed across the subgroups (all P for interaction >0.05). The Plaque-RADS ≥3 model demonstrated a higher AUC than the conventional US model [0.816 (95% CI: 0.731–0.901) vs. 0.810 (95% CI: 0.720–0.899)] and provided a slight net benefit across selected threshold probabilities. Additionally, the risk of recurrent ischemic stroke increased with higher Plaque-RADS categories (P for trend <0.001). Among the models using different high-risk Plaque-RADS thresholds, the Plaque-RADS ≥3b model showed the best predictive performance (0.865, 95% CI: 0.799–0.930), compared with the Plaque-RADS ≥3 model and the Plaque-RADS ≥3c model [0.824 (95% CI: 0.740–0.906)]. The Plaque-RADS ≥3b model also showed moderate improvements in NRI [0.250; (95% CI: 0.088–0.557), P=0.017] and IDI [0.090; (95% CI: 0.000–0.185), P=0.049], compared with the Plaque-RADS ≥3 model. Bootstrap internal validation indicated limited optimism across all models. Conclusions: In patients with prior stroke and low-to-moderate-grade carotid stenosis, high-risk US-based Plaque-RADS categories were significantly associated with recurrent ischemic stroke. The US-based Plaque-RADS may help identify plaque phenotypes associated with an increased risk of recurrent ipsilateral ischemic stroke, while refined high-risk thresholds, particularly Plaque-RADS ≥3b, may provide additional value for risk stratification.