From Measurement to Prediction: What Peak Areal Stretch Can and Cannot Tell Us About Aortic Root Rupture in Transcatheter Aortic Valve Replacement

Aortic root rupture is among the most feared complications of transcatheter aortic valve replacement (TAVR), with a reported mortality of 41%-50%.^1–3 Contemporary multicenter data from 10,541 patients across 18 centers (2014-2024) confirm that cardiac structural complications occur in approximately 2% of TAVR recipients, with annular rupture comprising 41% of these events and carrying the highest mortality rate; notably, the incidence has not decreased over the past decade despite advances in device technology and patient selection. ³ The incidence of annular rupture is approximately 1% in tricuspid anatomy and rises to 4.5% in highest-risk bicuspid aortic valve (BAV) subgroups.^1,4,5 Current pre-procedural planning relies on anatomical surrogates such as calcium burden, valve oversizing, and BAV morphology.^1,4,6 However, such metrics characterize the patient's anatomy prior to TAVR and do not capture the biomechanical interaction between the deployed device and the aortic root . Their ability to predict aortic root rupture is therefore inherently constrained.

In this issue of The Journal of the Heart Valve Society, Becker et al ⁷ report on the use of patient-specific 3D computational modeling to predict the risk of aortic root rupture following TAVR. Using the FDA-cleared PrecisionTAVI platform (DASI Simulations LLC; FDA clearance K223809), the authors simulated balloon-expandable (BE) transcatheter heart valve deployment from pre-procedural CT angiography and extracted peak areal stretch as a candidate biomarker for aortic root rupture. The study was conducted in two phases: a derivation cohort of 26 patients (12 rupture-positive, 14 rupture-negative) to establish a stretch threshold of 1.6, followed by a prospectively blinded validation cohort of 40 patients in which clinical outcomes were withheld from the modeling team throughout the analysis. In the blinded validation cohort, peak areal stretch achieved an AUC of 0.810 at the pre-specified threshold, identifying all five rupture events (100% sensitivity, approximately 51% specificity). These results exceeded those of all conventional comparators tested in the same sample: calcium score (AUC 0.643), calcium volume (AUC 0.634), valve oversizing (AUC 0.697), and age (AUC 0.477).

Despite these promising results, it is important to assess what the data can and cannot support. The central question is whether the measurement supports the claim. The paper's title and central message frame the findings as a prediction of aortic root rupture in TAVR. However, the model measures calcium-induced sinus wall deformation during BE valve deployment in a retrospective, pre-selected high-risk cohort. Several factors limit the extent to which this measurement can be equated with clinical prediction of rupture.

It should also be noted that the study has important methodological strengths. The blinded validation design represents a rigorous approach that is not always applied in computational modeling studies. Furthermore, peak areal stretch is mechanistically coupled to the causal pathway of rupture rather than being an anatomical proxy: it quantifies the consequence of calcific nodule displacement under valve deployment loading rather than the nodule burden itself. Interestingly, maximum stretch in the left coronary sinus (LCS) was significantly higher in rupture-positive cases (P = .050), which is clinically relevant given that LCS wall disruption may lead to rapid pericardial hemorrhage and acute cardiac tamponade. ¹ The computational efficiency reported, under 2 hours per patient, is also compatible with pre-procedural planning workflows in clinical practice. The study represents an encouraging step toward predicting the risks of aortic root rupture following TAVR, but several important limitations must be considered: (i) The derivation cohort carries a 46% rupture event rate, which is approximately 46 times the population-level clinical incidence of approximately 1%. ² A threshold derived at this enrichment level cannot be directly generalize to an unselected TAVR population. Moreover, the 40-patient validation cohort was not a consecutive or unselected sample. All patients had been designated high-risk for rupture by the heart team using conventional pre-procedural planning (the same metrics the model proposes to supplement). The AUC of 0.810 was therefore computed within a pre-enriched stratum. Whether peak areal stretch retains discriminative performance in the broader TAVR population remains to be demonstrated. (ii) Five positive events underpin the validation performance estimates. The optimal threshold shifted from 1.60 in the derivation cohort to 1.69 in the validation cohort, a drift of 5.6%. The authors attribute this to small sample size and morphological variability, which is plausible. However, a threshold that changes by 5.6% on a validation sample of 40 patients has not yet demonstrated the reproducibility required for a clinical decision rule. These results suggest that a larger prospective validation cohort is still needed before a specific threshold can be recommended in clinical practice. (iii) The mechanism scope of the model is narrower than the title implies. Aortic root rupture encompasses multiple biomechanical mechanisms: calcium-induced sinus wall stretch, LVOT calcium impingement, direct stent-frame injury, post-dilation trauma, and raphe calcification in BAV anatomy.^1,6 The current model addresses only calcium-induced sinus stretch during BE valve deployment. Mechanisms involving self-expanding valves, LVOT calcium, and stent-frame injury are not considered. Early patient-specific finite-element studies by Wang et al not only demonstrated that simulated rupture location could match clinical observations in individual cases, but also highlighted that the mechanism in each case was driven by the specific interaction between calcific deposits and the stent frame, underscoring the heterogeneity of rupture pathways that a single biomarker must capture. ⁸ (iv) The FDA clearance cited for PrecisionTAVI pertains to stent deformation prediction accuracy - specifically, that predicted stent shape matches post-TAVR CT to within 10% error. This is a different endpoint from rupture prediction and should not be conflated in clinical communication. Independent reviews have confirmed that DASI Simulations holds FDA clearance as a software for pre-procedural TAVR planning, but the cleared indication relates to device-anatomy interaction modeling, not to clinical outcome prediction.^9,10 (v) In daily clinical practice, the performance characteristics require careful interpretation. At the reported 100% sensitivity and 51% specificity, applying the threshold of 1.6 to an unselected TAVR population with a 1% rupture incidence would yield a positive predictive value of approximately 2%, meaning that for every true positive, there would be approximately 49 false positives. ¹¹ The clinical consequence of a false positive (whether procedural modification, alternative device selection, or deferral) is not negligible and has not been quantified in a population-representative dataset. The authors argue that prioritizing sensitivity over specificity is acceptable given the lethality of rupture. This position is reasonable, but it requires prospective evaluation in an unselected cohort before it can inform clinical guidance.

Whether computational modeling of peak areal stretch will become a useful tool in TAVR pre-procedural planning will depend on validation in a prospective, consecutive, unselected patient cohort with systematic documentation of rupture occurrence and location. Extension of the modeling framework to self-expanding valves and additional rupture mechanisms would further strengthen the clinical applicability of the approach. There is already evidence that patient-specific computational simulations can improve procedural decision-making in TAVR: the prospective TAVIguide study demonstrated that simulation affected implantation depth selection and procedural execution in 38% of cases, and the randomized GUIDE-TAVI trial is currently evaluating whether simulation-guided planning improves clinical outcomes.^12,13 These results suggest that peak areal stretch may represent a good step toward more precise pre-procedural risk stratification, provided that the necessary prospective validation is undertaken.

In summary, the present study reports an original and methodologically rigorous approach to a clinical problem that existing pre-procedural metrics still do not fully capture. The use of patient-specific computational modeling to quantify calcium-induced tissue deformation during valve deployment is a conceptually sound advance. The biomarker is mechanistically grounded, the blinded validation design is commendable, and the signal is consistent across both cohorts. Peak areal stretch represents a genuine step forward in pre-procedural risk stratification for TAVR, and the framework presented here provides a solid foundation for future work (including larger prospective studies and extension to self-expanding valve platforms) that will determine its place in clinical practice.