Multimodal CT-echocardiographic score for preprocedural prediction of permanent pacemaker implantation after transcatheter aortic valve replacement
X J C Chen, Y Li, W Zhu, D Wu, H Sun, R Wang, M Hong, D Xu, Z Su, H Xu, Z Xiao, M Fu, J Hong, M C Fang, Y GuoAbstract
Background
Permanent pacemaker implantation (PPI) occurs in 10–25% of patients after transcatheter aortic valve replacement (TAVR) and adversely affects long-term outcomes. As TAVR expands to younger, lower-risk populations, accurate preprocedural risk stratification becomes critical. Current prediction models rely on limited clinical or single-modality imaging variables and demonstrate modest discriminatory performance (c-statistics 0.65–0.75). No preprocedural score integrating multimodal CT geometry, aortic valve calcification burden, echocardiographic diastolic function, and baseline conduction status has been developed.
Purpose
To derive and temporally validate a preprocedural multimodal imaging-based Conduction Risk Index (CRI) for predicting 30-day PPI after TAVR and to compare its performance against existing risk scores and machine learning algorithms.
Methods
This single-centre retrospective study analysed 1,226 consecutive patients who underwent TAVR between June 2019 and December 2024. We derived the CRI, a dimensionless composite index incorporating standardised CT-derived annular geometry, device-landing-zone calcium volume, and echocardiographic diastolic function parameters. Internal-external cross-validation was conducted across three non-overlapping temporal blocks (2019–2020, 2021–2022, 2023–2024). Model discrimination, calibration, decision curve analysis, and continuous net reclassification improvement (NRI) were compared against a preprocedural modified Maeno score, clinical variables alone, and a gradient boosting machine. Interobserver reproducibility was assessed in a stratified random subsample.
Results
Among 1,226 patients (mean age 71 ± 8 years; 47.8% male), 151 (12.3%) required PPI within 30 days. The final model combining CRI, age, and baseline right bundle branch block achieved a pooled area under the curve (AUC) of 0.89 (95% CI 0.86–0.91) with excellent calibration (slope 1.02; Brier score 0.09). It outperformed the modified Maeno score (AUC 0.71; ΔAUC 0.18, p < 0.001), clinical variables alone (ΔAUC 0.21, p < 0.001), and performed comparably to a gradient boosting machine (AUC 0.87). Continuous NRI versus the modified Maeno score was 0.58 (95% CI 0.44–0.72, p < 0.001). Performance remained stable across all pre-specified subgroups (all interaction p > 0.20). Interobserver agreement for CT measurements was excellent (intraclass correlation coefficient > 0.94).
Conclusions
A dimensionless multimodal imaging index integrating CT-derived geometric and calcific parameters with echocardiographic diastolic function and baseline conduction status provides robust preprocedural PPI risk stratification after TAVR. External validation in independent multicentre cohorts is warranted before clinical implementation.Multimodal Imaging Score for PredictingFor image description, please refer to the figure legend and surrounding text.