Radial wall strain for residual risk stratification after percutaneous coronary intervention

The study developed an artificial intelligence algorithm for subpixel-level lumen delineation that enables real-time automated assessment of radial wall strain (RWS) from routine coronary angiography, whereas plaque strain assessment traditionally required complex modelling and expensive intracoronary imaging. Relative to traditional plaque strain assessment requiring complex modelling and expensive intracoronary imaging, this algorithm bases RWS assessment on routine angiography, lowering the technical barrier. The algorithm is described by the study as newly developed; the abstract does not provide independent validation data on algorithm performance.

In 1,384 non-target vessels from 802 patients, baseline RWSmax independently predicted the 5-year NT-VOCE, with an adjusted HR of 4.82 (95% CI 3.14-7.40, p<0.0001) and an adjusted AUC of 0.73 (95% CI 0.69-0.78, p<0.0001). Previously, the prognostic value of plaque strain as a marker of plaque vulnerability largely depended on complex modelling and intracoronary imaging; this study defines increased plaque vulnerability using a prespecified cutoff of RWSmax ≥13% on routine angiography and validates its long-term prognostic association in a randomised trial population. Based on all 1,551 enrolled patients in the TARGET All Comers randomised trial, using blinded analysis, a prespecified cutoff, and prospectively adjudicated endpoints, with RWS successfully evaluated in 1,384 non-target vessels; the effect size is large and the confidence interval does not cross the null.

RWS showed the highest predictive accuracy for non-target vessel revascularisation, with an adjusted AUC of 0.92 (95% CI 0.88-0.95, p<0.0001). Among the components of NT-VOCE, the prognostic value of RWS was most prominent for subsequent non-target vessel revascularisation, suggesting a closer association with lesion progression requiring repeat intervention. Based on vessel-level endpoint associations within the same blinded analysis framework, with an AUC approaching 0.9, though the abstract does not provide component event counts or competing-risk handling details.

In the subgroup with high-quality angiographic images, RWS demonstrated even greater predictive performance for the NT-VOCE, with an adjusted HR of 6.89 (95% CI 3.15-15.07, p<0.0001). Angiographic image quality was independently assessed prior to blinded RWS computation, and the effect size was larger in the high-quality image subgroup, suggesting image quality is an important condition for RWS assessment. Subgroup analysis based on the high-quality angiographic image subset, with a larger effect size than the overall population but a smaller subgroup sample size and wider confidence interval.

This study is a post hoc analysis based on the TARGET All Comers randomised trial, applicable to patients who have undergone percutaneous coronary intervention and have untreated non-target vessels, with conclusions conditioned on the availability and assessability of routine coronary angiography; RWS assessment was successfully completed in 1,384 non-target vessels, and predictive performance was stronger in the high-quality angiographic image subgroup, indicating that the method's applicability is closely tied to image quality. The study concludes that ongoing and planned prospective randomised trials are evaluating the role of RWS-guided risk stratification, so the current results primarily serve to generate hypotheses and guide subsequent validation rather than directly change clinical practice.

As a post hoc analysis, the association between RWS and NT-VOCE, although adjusted and with a large effect size, remains an observational association and cannot directly support causal inference; RWS assessment was successfully completed in only 1,384 non-target vessels from 802 patients, and the characteristics of vessels not successfully assessed and their impact on the representativeness of results remain unclear; the high-quality image subgroup showed a larger effect size but a wider confidence interval, and the degree to which image quality affects RWS assessability and predictive performance still needs further quantification; the event counts for each NT-VOCE component, competing-risk handling, and the derivation and external validation of the RWSmax ≥13% cutoff are not elaborated in the abstract, so readers needing to assess robustness should consult the full text; the clinical utility of RWS-guided risk stratification awaits answers from ongoing and planned prospective randomised trials.