Extracellular Matrix Proteins Improve Risk Prediction in Patients Undergoing Transcatheter Aortic Valve Replacement

Abstract

Background: Cardiac fibrosis is common in patients with severe aortic stenosis and an independent predictor of death. Therefore, we examined the additional value of circulating fibrosis markers as a putative biomarker platform to identify patients with aortic stenosis undergoing transcatheter aortic valve replacement (TAVR) who are at a higher risk of death.

Methods: In this study, 2-year survival analyses were conducted in 378 consecutive patients undergoing TAVR to evaluate the association between fibrosis marker and risk of adverse long-term outcome. Implementation of fibrosis marker into TAVR risk stratification was tested by a machine-learning algorithm.

Results: Among 20 circulating fibrosis markers involved in pathological extracellular matrix remodeling, high tissue inhibitor of metalloproteinase-1 (TIMP-1) levels independently predicted risk of death in univariable (hazard ratio, 5.0 [95% CI, 2.6-9.7]; P<0.001) and multivariable (adjusted hazard ratio, 2.2 [95% CI, 1.0-4.7]; P=0.046) Cox regression analyses. Consequently, higher TIMP-1 levels offered a significantly higher overall prediction of reduced survival compared with the conventional Society of Thoracic Surgeons Predicted Risk of Mortality score (area under the curve, 0.753 [95% CI, 0.682-0.824] versus area under the curve, 0.656 [95% CI, 0.578-0.734]; P<0.05). Applying an independent machine-learning algorithm allowed identification of a simple combination of 2 biomarkers (TIMP-1 and high-sensitivity cardiac troponin T) with superior prognostic value compared with Society of Thoracic Surgeons Predicted Risk of Mortality alone (area under the curve, 0.757 [95% CI, 0.686-0.828] versus 0.656 [95% CI, 0.578-0.34]; P<0.05).

Conclusions: Circulating TIMP-1 is an independent predictor of reduced 2-year overall survival in patients undergoing TAVR. Combined with high-sensitivity cardiac troponin T, circulating TIMP-1 should be incorporated into risk stratification to identify patients undergoing TAVR who are at a higher risk of death.

Keywords: ECM; TAVR, TAVI, aortic stenosis; TIMP‐1; biomarker; hs‐cTnT.

Figure 1. TIMP‐1 predicts outcomes in patients undergoing TAVR.

A, Distribution of TIMP‐1 levels within the total study cohort in survivors and nonsurvivors. B and C, Landmark analysis with survival curves based on circulating TIMP‐1 (B) and ROC analysis comparing the prognostic value of TIMP‐1 and STS‐PROM (C). AUC indicates area under the curve; ROC, receiver operating characteristic [curve]; STS‐PROM, Society of Thoracic Surgeons Score Predicted Risk of Mortality; TAVR, transcatheter aortic valve replacement; and TIMP‐1, tissue inhibitor of metalloproteinase‐1.

Figure 2. Incremental prognostic value of machine learning–derived blood biomarkers to predict outcomes of patients undergoing TAVR.

A, Recursive partitioning survival trees on the selected biomarkers in the total study cohort. B and C, Landmark analysis with Kaplan–Meier plots for patients with TAVR stratified by risk groups defined by XGB‐based machine‐learning model (B) and ROC analysis assessing the prognostic value of a simple 2‐biomarker combination (TIMP‐1 and hs‐cTnT) compared with the conventional risk score STS‐PROM (C). AUC indicates area under the curve; hs‐cTnT high‐sensitivity cardiac troponin T; ROC, receiver operating characteristic [curve]; STS‐PROM, Society of Thoracic Surgeons Score Predicted Risk of Mortality; TIMP‐1, tissue inhibitor of metalloproteinase‐1; TnT, troponin T; and XGB, extreme gradient‐boosting.