Time- and dose-dependent effect of high-sensitivity cardiac Troponin T compared to VARC-3 defined myocardial injury for predicting outcomes after Transcatheter Aortic Valve Implantation

Background Pre- and post-procedural high-sensitivity cardiac troponin T (hs-cTnT) predict periprocedural myocardial injury (PPMI) and adverse events following transcatheter aortic valve implantation (TAVI). However, current studies have not incorporated time- and dose-dependent effects of hs-cTnT, limiting the precision of risk stratification and subsequent patient management. Purpose To investigate the non-linear and time-dependent effects of pre- and post-procedural hs-cTnT on outcomes after TAVI and compare these findings with the valve academic research consortium-3 (VARC-3) dichotomized definition of PPMI. Methods Consecutive patients undergoing TAVI at two tertiary university hospitals between 2011 and 2025 with available hs-cTnT concentrations measured pre- and post-procedurally were enrolled. Primary outcome was all-cause mortality at 1 year. A multivariable Cox proportional hazards model was fitted. To relax the proportional hazards assumption, allowing for hazard ratios (HRs) to vary over time, multivariable penalised spline survival models were fitted. In both models, hs-cTnT was modelled with splines allowing for non-linear effects. HRs were calculated for each timepoint & hs-cTnT-value relative to day 0 & a hs-cTnT of 14 ng/L. Time-varying HR plots, 3D mesh plots & contour plots were constructed. Results Among 6178 patients, 5823 had both pre- and post-hs-cTnT measurements available and were included. The HR for patients with a hs-cTnT > 35 or 70 times the upper reference limit of 14 ng/L (VARC-3 criterion) was not prognostic for all-cause death (1.65, 95%CI: 0.98,2.76, p=0.06). Cox proportional hazard models handling hs-cTnT as a continuous variable showed that both higher pre- and post-procedural hs-cTnT levels were associated with increased hazard for all-cause mortality at 1 year (Figure 1A,B). Models allowing for time-varying HRs revealed that the hazard was highest early post-procedure and diminished over time (Figure 1 C-P), indicating a non-constant hazard. The duration of being at risk increased proportionally for higher pre- and post-procedural hs-cTnT and was longest for post-procedural hs-cTnT. For example, a patient with a post-procedural-hs-cTnT of 600 ng/mL remained at increased risk for mortality during the first 3 months, versus only 1 month with a post-procedural hs-cTnT of 50 ng/mL. Conclusions hs-cTnT as a binary variable, either through VARC-3 defined PPMI or direct use of hs-cTnT was not prognostic. The dynamic nature of risk post-TAVI could only be modelled when using hs-cTnT as a continuum. Higher levels of both pre- and post-procedural hs-cTnT are associated with a prolonged period of increased mortality risk, with duration ranging from 1 to 3 months when using a hs-cTnT of 14ng/L as reference. Incorporating time-varying and non-linear effects of hs-cTnT into risk prediction models could enhance clinical decision-making by tailoring post-procedural surveillance to individual patient risk profiles.