Artificial intelligence-enhanced strategies for reducing mortality in transcatheter aortic valve replacement: improving outcomes and minimizing risks

Dear Editor, In the UK, the prevalence of aortic stenosis (AS) is estimated at 1.48% of the population. Among those over 55 years, around 300 000 people are estimated to have severe AS. Unfortunately, a recent study by Johann Auer et al. [1] has revealed that transcatheter aortic valve replacement (TAVR) is associated with higher mortality rates compared to surgical aortic valve replacement (sB-AVR). Interventions in aortic valve replacement techniques are urgently needed, as TAVR demonstrates a higher hazard ratio of 1.552 (95% confidence interval), in comparison to sB-AVR. Moreover, the estimated median survival was significantly longer with sB-AVR at 8.8 years, compared to just 5 years with TAVR. Notably, TAVR is also associated with a higher risk of permanent pacemaker implantation compared to sB-AVR. These findings emphasize the urgent need for interventions aimed at reducing the high mortality rates observed in patients undergoing TAVR. As a prevalent valvular disorder, AS particularly affects the elderly, leading to left ventricular outflow obstruction. With the growing incidence of AS globally, various treatment options are under consideration. While sB-AVR has long been the standard treatment, TAVR has gained popularity over the past decade due to its less invasive nature, shorter recovery time and reduced complication rates [2]. However, TAVR’s higher mortality rates remain a concern, necessitating new approaches to improve outcomes. To address these concerns associated with TAVR and improve patient outcomes, novel methods, including artificial intelligence (AI)-powered software, are being investigated. A recent study found that AI-generated annular measurements closely matched manual expert measurements, with correlation coefficients of 0.97 for both perimeter and area. Moreover, the preprocedural process for TAVR typically involves screening for coronary artery disease, often using coronary CT angiography. The growing volume of TAVR procedures has strained hospital resources. AI-based software can streamline the process by uploading multiple CTs to a cloud platform, saving time and improving efficiency [3]. Moreover, AI models demonstrate greater predictive accuracy compared to traditional risk scores in forecasting post-TAVI mortality. AI models show a significantly higher mean Area Under Curve (AUC) for mortality predictions compared to traditional risk scores (P < 0.00001), with notable improvements for both 30-day (P = 0.001) and 1-year mortality (P < 0.0001) [4]. Integrating AI-based models could enhance patient selection, leading to better outcomes and reduced complications in severe AS cases. Additionally, we commend the author for their dedication to this research. However, we must note several limitations within the study. The reliance on retrospective data could introduce biases that may impact the accuracy of the findings. Moreover, the generalizability of the results is limited, as the data were exclusively sourced from the Austrian Health Insurance funds. Therefore, larger and more diverse prospective studies are needed to validate the findings and enhance cardiothoracic patient care. In conclusion, the integration of AI in TAVR enhances the precision of annular measurements, streamlines CT processing and surpasses traditional methods in predicting mortality. This technology has the potential to optimize patient selection and resource allocation, ultimately leading to improved outcomes and fewer complications. However, further research is necessary to address existing limitations and refine TAVR treatment. No funding source to declare. Conflict of interest: none declared.