Our analysis of younger patients (60–75 years) reveals a complex trade-off between clinical outcomes and costs. Clinically, TF-TAVR was associated with significantly lower rates of bleeding, postoperative delirium, and prolonged mechanical ventilation compared to SAVR, alongside a risk-adjusted in-hospital mortality benefit. However, these clinical advantages come at a higher financial cost to the healthcare system. The key question is whether these additional costs are justified by the observed clinical benefits. Overall, the risk-adjusted mortality benefit of TF-TAVR over SAVR is significant but relatively small, while the additional reimbursement is still eminent. If we consider the life expectancy of the general population, the life expectancy of the patients studied is between 25 years (60-year-old female patient) and 11 years (75-year-old male patient), and the additional costs per life saved due to a hypothetical shift from SAVR toward TF-TAVR may be considered acceptable under certain willingness-to-pay assumptions but should be interpreted cautiously in the analyzed set of patients [32]. Clinically, an ICER of €857,413 from the in-hospital perspective reflects a small absolute mortality difference combined with a substantial reimbursement difference. This does not imply that TF-TAVR is or is not “worth it” per se; rather, it quantifies the economic trade-off under the study’s perspective and assumptions. Nevertheless, the individual situation of the patient always must be taken into account for decision-making. Of course, one cannot use the life expectancy of the general population to estimate the remaining life expectancy after successful TAVR implantation. Instead, the median survival time of a comparable population should be used. However, this information is not available for a post-TAVR population of younger age. For SAVR, these data are available and indicate a residual life expectancy of 13 years for patients aged 65 to 69 years [33,34,35]. Again, the additional costs per life saved due to a hypothetical shift from SAVR toward TF-TAVR appear to be well justified given a life expectancy of 13 years. However, it is important to note that these 13 years had come from patients whose SAVR procedures were largely performed in the last millennium, which again limits comparability regarding today’s general cardiac treatment options. Furthermore, it is important to note that the greater baseline risk profile of the TAVI patients suggests that the observed mortality benefit in favor of TAVI may even be an underestimation.

TAVR is still a relatively new technique, especially compared to its direct alternative, SAVR, and the evolution of outcomes since TAVR was introduced reflects a significant learning curve [8, 36, 37]. This means that any treatment effect may vary depending on the exact year of the intervention [38]. As a result, studies or trials conducted over a relatively short period of time provide a snapshot that may be outdated a short time later [39]. In our data, the share treated with TF-TAVR increased from 34% (2018) to 49% (2022), reflecting broader adoption and potentially modest shifts in baseline risk over time (Table 1). Accordingly, the DML/AIPW framework (Fig. 1) remains interpretable despite temporal shifts, as adjustment included year-of-procedure alongside clinical covariates.

For the period between 2018 and 2022, our study shows that TF-TAVR is more expensive than SAVR. Nonetheless, it is accompanied by a notable benefit in terms of reduced mortality. Considering the (hypothetical) remaining life expectancy of the patients under consideration, the cost–benefit ratio appears to be quite justifiable. However, the considerable differences in the costs of treatment also show that the economic aspect of extending the indication should not be neglected. Furthermore, temporal trends in reimbursement play a crucial role in the economic evaluation [10]. As shown in Table 3, the unadjusted reimbursement for TF-TAVR in our cohort decreased notably from €30,986 in 2018 to €27,177 in 2022, likely reflecting the devaluation of DRG rates and reduced device costs. In contrast, SAVR reimbursement remained relatively flatter (ranging from €21,356 to €20,253). This narrowing gap indicates that TF-TAVR is steadily becoming more cost-effective relative to surgery over time. Although we could not perform a risk-adjusted yearly cost-effectiveness analysis due to the limited number of endpoints per year, this descriptive trend suggests that future ICERs may prove even more favorable for TAVR.

Our research has numerous limitations that extend beyond the typical constraints associated with retrospective analyses [10, 20]. Firstly, it is based on administrative data. Consequently, coding errors are inevitable. Secondly, our model’s risk adjustment may not encompass all necessary parameters, as some included parameters might lack full reliability. Moreover, our administrative dataset does not contain crucial clinical data, like echocardiographic results or anatomical details, hindering the assessment of operative risks and a deeper understanding of valvular pathomechanisms. Consequently, we apply only an approximate version of the logistic EuroSCORE, essentially a conservative or “best-case” estimate. In summary, analyses rely on administrative data with limited clinical granularity and a best-case imputation for missing EuroSCORE components; residual misclassification and unmeasured confounding cannot be excluded; both may bias effect estimates and warrant cautious interpretation. Importantly, the administrative dataset does not contain echocardiographic or anatomical information (including valve morphology). We therefore cannot differentiate tricuspid from bicuspid aortic stenosis. As evidence of non-inferiority in younger/low-risk populations in randomized trials (DEDICATE-DZHK6, NOTION-2) has largely focused on tricuspid anatomy [31, 40], the extent to which our findings generalize to bicuspid disease remains uncertain. The NOTION-2 trial enrolled 100 bicuspid patients with less favorable outcomes [40]. Consequently, our results should be interpreted as descriptive and hypothesis-generating and should not be used to replace individualized clinical decision-making. It is important to emphasize that this 1-year calculation relies on external mortality data from the DEDICATE-DZHK6 trial [31]. This introduces specific limitations, as the trial population primarily consisted of tricuspid aortic stenosis patients and excluded certain high-risk anatomies, which may not perfectly match our real-world administrative cohort. Moreover, our cost perspective is in-hospital only; linkage to post-discharge events and resource use is not available in DESTATIS, so downstream costs of complications or follow-up care could not be incorporated, and the 1-year ICER relies on external mortality data without post-discharge costs. In addition, device-level analyses (balloon-expandable vs. self-expanding, or by generation) were not feasible because the dataset does not capture valve generation, and finer stratification by device and year would markedly reduce power; we therefore refrained from device-type subgrouping. Finally, the dataset excludes patients with a baseline diagnosis of aortic regurgitation and those undergoing TF-TAVR or SAVR with other concomitant cardiac procedures. From a clinical standpoint, this approach is logical. However, it poses challenges for making direct comparisons with other administrative datasets. An additional limitation is the absence of specific data in our dataset regarding the exact model of the device used, beyond its type. For instance, in the case of self-expanding (SE) valves, the ACURATE neo [41] and the CoreValve [42] were in use in Germany [43] at that time, but this level of detail is not captured in our data.