The Jenaer Aortenklappenregister (JAKR), established in 2016, prospectively includes all patients that undergo transcatheter aortic valve implantation (TAVI) at the University Hospital Jena. In this study, we investigated two patient cohorts with the aim to, first, identify novel markers for mortality risk prediction and define a simple sum-score system (identification cohort, n = 122) and second, validate it in a larger study population (validation cohort, n = 295). All patients included gave written informed consent for participation and the local ethics committee of the University Hospital Jena has approved the study (registration number: 4815-06/16). Detailed clinical, laboratory, functional and imaging analysis according to local standard operating procedures were performed in adherence to the principles of the current version of the Declaration of Helsinki and good clinical practice guidelines. Structured patient follow-up was carried out at 6 weeks, 6 months, 12 months and 24 months according to the JAKR protocol. In principle, all patients undergoing TAVI and agreed to participate were consecutively included in the study. Due to potential extra-cardiovascular sources of the serum biomarkers determined by Enzyme linked Immunosorbent Assay (ELISA), the following conditions were excluded: active malignant or autoimmune disease, hyperthyroidism, infections or systemic intake of corticosteroids.

For the assessment of patients’ health status, the EuroQol visual analogue scale (EQ-VAS; between 0 and 100 points) was used as a component of the EuroQol questionnaire (EQ-5D-5L), which is an accepted tool for the standardized simple generic Quality of Life (QoL) assessment already used in TAVI patients [28, 29, 30]. The Clinical Frailty Scale (CFS), known to predict death or the need for institutional care, was applied to assess clinical frailty of the patients by assigning the following categories: very fit (1), well (2), managing well (3), vulnerable (4), mildly frail (5), moderately frail (6), severely frail (7), very severely frail (8) and terminally ill (9) [31].

In the frame of the JAKR study protocol, blood samples were taken from all study participants by standard venous puncture. The collection tubes were centrifuged within 20 min after withdrawal and serum was transferred into special low binding tubes (Protein LoBind, Eppendorf AG, Hamburg, Germany), snap frozen in liquid nitrogen to reduce artificial protein degradation and finally stored at –80 ${}^{\circ }$C until further analysis. Routine laboratory parameters, e.g., brain natriuretic peptide (BNP) levels or creatinine, were measured according to local standard operating procedures. The circulating levels of preselected biomarkers of cardiovascular tissue remodelling were quantified by ELISA technique. For the following biomarkers, commercially available assays were used according to the instructions of the manufacturer: matrix metalloproteinase 9 (MMP-9; assay: Human MMP-9 Immunoassay, R&D Systems GmbH, Wiesbaden, Germany), tissue inhibitor of metalloproteinase 1 (TIMP-1, assay: Human TIMP-1 Immunoassay, R&D Systems GmbH, Wiesbaden, Germany), B domain containing tenascin-C (B${}^{+}$ Tn-C, assay: Tenascin-B Large (FNIII-B) ELISA, IBL International GmbH, Hamburg, Germany), endothelin 1 (ET-1, assay: Endothelin-1 Immunoassay, R&D Systems GmbH, Wiesbaden, Germany) and neutrophil gelatinase-associated lipocalin (NGAL, assay: Human Lipocalin-2/NGAL Immunoassay, R&D Systems, Wiesbaden, Germany). For the serum quantification of extra domain A containing fibronectin (ED-A${}^{+}$ Fn) and extra domain B containing fibronectin and (ED-B${}^{+}$ Fn), we applied protocols, which were recently established and validated in our group [22, 32, 33] because commercial ELISA assays are not available at present.

Statistical analyses were performed using IBM SPSS statistical software, version 25.0 (IBM SPSS Statistics for Windows. Armonk, NY, USA). Data are expressed as mean/median $\pm$ standard deviation, as appropriate. Mann–Whitney-U test was used to test for significant differences between the groups with respect to clinical, laboratory, functional and imaging parameters as well as biomarker levels.

For the identification of independent predictors for one- and two-year mortality, multivariate regression analysis was performed by using a binary logistic model (backward elimination method: Wald). One- or two-year mortality was defined as the dependent variable.

In the identification cohort, (n = 122), a wide range of baseline characteristics were assigned to three clusters with 4 to 10 items each. For each of the three clusters, a multivariate analysis was performed by using one-year mortality as dependent variable and the following items as cofounders. For cluster 1 (classical clinical parameters/comorbidities), New York Heart Association (NYHA) functional class, coronary artery disease (CAD), peripheral artery disease (PAD), diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD), atrial fibrillation (Afib), presence of pacemaker (PM) and glomerular filtration rate (GFR) were included as cofounders. The items for cluster 2 (risk assessment scores) were EQ-VAS (0 to 100), CFS (1 to 9), STS score (%) and staging classification of extra-valvular cardiac damage (stage 0 to 4) and for cluster 3 (novel biomarkers of cardiovascular remodeling) MMP-9, TIMP-1, B${}^{+}$ Tn-C, ET-1, NGAL, ED-A${}^{+}$ Fn, ED-B${}^{+}$ Fn and BNP serum levels.

In each cluster, one parameter could be identified to significantly increase one-year mortality risk in the identification cohort (Afib, CFS and TIMP-1). Basing on these findings, a simplified sum-score system was defined as follows: presence of Afib (no = 0, yes = 1); dichotomized CFS (1–4 = 0; 5–9 = 1); TIMP-1 range (cut-off value: 187.2 ng/mL; below = 0, above = 1). The resulting sum-score, denominated as Jena Mortality Score (JMS), ranged from level 0 to 3. The one-year mortality rates per JMS level were calculated in the identification cohort and a multivariate analysis was performed as described including all parameters showing a p-value 0.1 in group comparison (survivors versus non-survivors after one year) except the 3 parameters contributing to JMS.

In a next step, to confirm our findings in a larger patients’ collective, JMS was tested with respect to its predictive value for one- and two-year mortality in the validation cohort (n = 295). Therefore, survivors and non-survivors after one and two years were compared with respect to a wide range of clinical, laboratory, functional and imaging parameters including TIMP-1 serum levels using the Mann–Whitney-U test. Then, multivariate regression analysis was performed, again by using a binary logistic model (backward elimination method: Wald), with one- or two-year mortality as dependent variable and all parameters showing a p-value 0.1 in group comparison (and not contributing to the included scores or classifications) as cofounders to identify independent predictors of long-term mortality.

Finally, Kaplan-Meier survival analysis including log-rank test as well as receiver operating characteristic (ROC) analysis was performed to verify differences between high and low-risk patients according to JMS compared with STS score with respect to both, one- and two-year mortality. Comparison of area under the curve (AUC) values was performed by DeLong method using MedCalc statistical software (Version 20.023; MedCalc Software Ltd, Ostend, Belgium).

Baseline characteristics of the identification cohort, which represented a typical TAVI collective with a mean STS score of 4.8%, are given in Table 1. The one-year mortality rate was 14%. By performing binary logistic regression analysis with one-year mortality as dependent variable and the items per cluster as cofounders as described in the material and methods section, Afib (cluster 1; OR 7.583, 95% CI: 2.051–28.040, p = 0.002), CFS (cluster 2; OR 2.258, 95% CI: 1.262–4.039, p = 0.006) and TIMP-1 (cluster 3; OR 1.006, 95% CI: 1.001–1.011, p = 0.019) could be identified to significantly increase long-term mortality risk. These 3 parameters were integrated into a simplified sum-score (JMS) as follows as described. As cut-off value for TIMP-1, the median of the 122 patients, which was 187.2 ng/mL, was chosen. The resulting JMS value ranged from 0 to 3. The mortality rates per score level in the identification group were 0% for JMS 0, 8.8% for JMS 1, 22.2% for JMS 2 and 33.3% for JMS 3. When performing binary logistic regression analysis including all parameters showing a p 0.1 in group comparison (survivors versus non-survivors), except the 3 parameters contributing to the score, only JMS was independently associated with an increased risk for long-term mortality after TAVI (OR 2.720, 95% CI: 1.467–5.042, p = 0.001).

With respect to the biomarkers of cardiovascular remodeling measured by ELISA and included in cluster 3 as described in Material and Methods, the median values $\pm$ standard deviation (SD) both, in the whole identification cohort (n = 122) as well as in survivors (n = 105) and non-survivors (n = 17) are given in Table 2.

Baseline characteristics of the validation cohort, which again represented a typical TAVI collective with a mean STS score of 4.8%, are given in Table 1. The mortality rate was 18% after one and 23% after two years.

Binary logistic regression analysis with one- and two-year mortality as dependent variable and STS score (STS), staging of extra-valvular cardiac damage (stage), presence of high gradient aortic stenosis (HGAS), brain natriuretic peptide (BNP), EQ visual analogue scale (EQ-VAS) and JMS as cofounders was performed. The cofounders were defined as all parameters not contributing to the implemented scores and showing a p-value 0.1 in group comparison between survivors and non-survivors after one year.

As a result, besides STS score, only JMS could be proven to serve as independent predictor of both, one-year (JMS: OR 1.626, 95% CI: 1.053–2.509, p = 0.028; STS score: OR 1.180, 95% CI: 1.052–1.325, p = 0.005) and two-year (JMS: OR 1.649, 95% CI: 1.093–2.488, p = 0.017; STS score: OR 1.253, 95% CI: 1.106–1.419, p 0.001) mortality.

The one-/two-year mortality rates per score level in the validation cohort were 1.75/2.62% for JMS 0, 3.93/4.80% for JMS 1, 6.11/7.86% for JMS 2 and 6.55/7.86% for JMS 3.

After dichotomization of patients into a low- and high-risk group according to JMS and STS score (JMS: 0 or 1 = low-risk, 2 or 3 = high-risk; STS score: 4% = low-risk, $>$4% = high-risk) Kaplan-Meier survival analysis was performed. The results are shown in Fig. 1a (JMS) and Fig. 1b (STS score). In brief, Kaplan-Meier survival analysis displayed a significant survival benefit for the low-risk groups according to both, JMS and STS score, after one and two years (p 0.05 for all). Of note, survival rate in the JMS high risk group is lower (66.0%) compared to those in the STS score high risk group (71.4%). ROC analysis for the discrimination of survivors from non-survivors revealed an AUC for JMS of 0.676 and for the STS score of 0.710 after one year (Fig. 2a,b, p = 0.534) and an AUC for JMS of 0.686 and for the STS score of 0.725 after two years (Fig. 2c,d, p = 0.437).

Fig. 1.

Kaplan-Meier survival analysis for JMS and STS scoredisplaying a significant survival benefit for the low-risk groups according after one and two years. (a) JMS. (b) STS score.

Fig. 2.

ROC analysis for the discrimination of survivors from non-survivors revealing an AUC for the JMS of 0.676 and for the STS score of 0.710 after one year and an AUC for the JMS of 0.686 and for the STS score of 0.725 after two years. (a) and (b) results after one year.

With explicit focus on the serum levels of TIMP-1, which have been already reported above as part of JMS, the following observations are of value to be reported separately. When comparing survivors and non-survivors after one and two years, there are significantly increased TIMP-1 concentrations in the non-survivors (p 0.001 for both timepoints) (Fig. 3a,b). ROC analysis revealed an AUC of 0.678 for the discrimination of survivors and non-survivors after one and an AUC of 0.670 after two years, which are superior compared to, e.g., baseline BNP levels (AUC 0.642 after one and 0.639 after two years) or baseline GFR (AUC 0.587 after one and 0.586 after two years).

Fig. 3.

Baseline TIMP-1 concentrations in survivors and non-survivors after one and two years showing significantly increased levels in non-survivors for both time-points. (a) Results after one year. (b) After two years.