This study included 439 consecutive patients with symptomatic severe aortic stenosis undergoing transfemoral TAVI at the Department of Cardiovascular Disease, Deutsches Herzzentrum München, Munich, Germany between April 2015 and December 2016. The multidisciplinary heart team discussed all patients and consensus was reached on the therapeutic strategy in each patient. Biomarkers were determined from routine blood samples at least 24hours before TAVI. Gal-3 was measured using an immunoassay (Quantikine Human Galectin-3 Immunoassay; R&D Systems, Minneapolis, MN, USA) and CA125 was determined using an electrochemiluminescence immunoassay (Cobas e411 analyzer, Roche Diagnostics GmbH, Mannheim, Germany). All patients provided written informed consent.

The primary endpoint of this study was a composite of all-cause mortality or unplanned readmission for worsening CHF during follow-up. In addition, all-cause mortality and readmission for CHF were analyzed separately. All clinical endpoints, procedural data and in-hospital complications were categorized according to the updated Valve Academic Research Consortium-2 criteria.15 All data were prospectively collected during routine visits at the outpatient clinic, by referring to hospital documentation, contacting the primary care physician, or by direct contact with patients or their relatives.

Data distribution was tested for normality using the Shapiro-Wilks test. Categorical variables are expressed as frequencies and proportions and were compared using the chi-square or Fisher exact test, as appropriate. Continuous variables are presented as mean with standard deviation (SD) or median with interquartile range [IQR] and compared using the Student t-test or Mann-Whitney U-test, respectively.

For dichotomous analysis, 2 groups of patients with elevated and not elevated Gal-3 levels were identified by maximally selected rank statistics (R-package “survminer”, version 0.4.0). CA125 was dichotomized according to a previously published cutoff (≥ 18.4 U/mL) into elevated and not elevated.3 Event rates of the primary endpoint and separately all-cause mortality and readmission for CHF were calculated as crude rates. Cumulative events during the first 12 months after TAVI were visualized using the Kaplan-Meier method with differences tested using the log-rank test.

The independent association of elevated Gal-3 and CA125 values, modeled separately and together, with time to the primary endpoint, as well as all-cause mortality and CHF separately, was assessed using multivariable Cox proportional hazards regression analyses. The hazard ratios (HR) with their 95% confidence intervals (95%CI) were computed. For CHF, the subdistribution hazard ratios (Fine and Gray model)16 with death as a competing event were also calculated (R-package “cmprsk”, version 2.2-7). Covariates in the multivariable model were selected using the LASSO (Least Absolute Shrinkage and Selection Operator) regression method entering all baseline variables as candidates (R-package “glmnet”, version 2.0-13). The resulting variables were age, logistic EuroSCORE I, previous myocardial infarction, previous cancer, atrial fibrillation, mitral regurgitation grade III/IV, pulmonary hypertension (defined as pulmonary artery pressure > 60 mmHg), hemoglobin values, creatinine clearance, N-terminal pro-B-type natriuretic peptide, and mean transvalvular gradient. Additionally, left ventricular ejection fraction<35% and C-reactive protein values were included due to their known prognostic value and the findings in the univariate analyses. Missing baseline data (0.14%) were imputed by predictive mean matching (R-package “mice”, version 2.46).

The proportional hazard assumption of multivariable models for the primary endpoint containing Gal-3 (dichotomic), CA125 (dichotomic) and both variables (dichotomic) was checked using scaled Schoenfeld residuals. The assumption of proportional hazards was met, expressed by a nonsignificant relationship between global residual and time (global P values 0.121, 0.290 and 0.151, respectively).

When modeled together, the interaction of continuous and dichotomous Gal-3 values within CA125 strata (elevated vs not elevated CA125) was tested. The differential prognostic value of elevated Gal-3 regarding the primary endpoint according to CA125 status (elevated vs not elevated CA125) was further analyzed. The incremental prognostic usefulness of elevated Gal-3 to a model containing elevated CA125 and baseline variables, as well as within in each CA125 stratum, was evaluated by calculating the integrated discrimination improvement and the net reclassification improvement with the corresponding 95%CIs (Stata package “incrisk” version 1.0.5). These statistical indices have been developed to evaluate the added predictive value of a new marker to a base model and have been described in detail previously.17

A 2-sided P value < .05 was considered statistically significant. Statistical analyses were performed using R (version 3.3.2, R Foundation for Statistical Computing, Vienna, Austria) and Stata 14 (Stata Corp, TX, USA).

Baseline and procedural characteristics of the entire patient population and according to the primary endpoint are displayed in Table 1 and Table 2. During a median follow-up of 371 days [219-402], the primary endpoint occurred in 16.4% (72 of 439) of the patients. Separately, all-cause mortality was 8.4% (37 of 439) and readmission for CHF occurred in 9.8% (43 of 439) of the patients. Patients experiencing the primary endpoint were older, had a higher predicted operative risk according to logistic EuroSCORE I, and were more symptomatic according to the New York Heart Association functional class. Procedural and in-hospital course was more complicated in these patients (Table 2).

Median Gal-3 and CA125 levels were 7.2 ng/mL [5.2-9.3 ng/mL] and 18.7 U/mL [11.1-38.9 U/mL] with higher values in those experiencing the primary endpoint (8.6 ng/mL [5.1-10.6] vs 7.0 [5.2-9.1];P=.043) and (39.8 U/mL [20.7-82.7] vs 17.1 [10.5-32.0]; P=.001), respectively.

The ideal cutoff to predict the primary endpoint with Gal-3 was 8.71 ng/mL (Figure 1). Gal-3 values were elevated (≥ 8.71 ng/mL) in 31.9% (140/439) of the patients. Clinical and procedural characteristics and in-hospital outcomes of patients with elevated Gal-3 values are displayed in Table 1 of the supplementary data.

Figure 1.

Distribution and rank statistics of galectin-3 values.

(0.15MB).

The cumulative event rate according to elevated Gal-3 and results from the univariate and multivariable analyses for the primary endpoint, as well as death and CHF separately, are provided in Table 3A. Elevated Gal-3 values were associated with a higher crude rate and risk of the primary endpoint (25% [35 of 140] vs 12.4% [37 of 299], HR, 2.26; 95%CI, 1.42-3.59; P=.001) in univariate analysis (Figure 2). After multivariable adjustment for baseline variables, the association of elevated Gal-3 with the primary endpoint was borderline significant (HR, 1.59; 95%CI, 0.97-2.62;P=.068) (Table 3A and Table 4).

Figure 2.

Kaplan-Meier event rate according to galectin-3. Kaplan-Meier estimates for the primary endpoint with differences tested using the log-rank test. +, elevated, –, not elevated; TAVI, transcatheter aortic valve implantation.

(0.1MB).

Univariate and multivariable analyses for prediction of death and CHF are displayed in Table 3A. Elevated Gal-3 was associated with a significantly higher risk of death and CHF in univariate analysis, which, however, was lost after multivariable adjustment.

CA125 values were elevated (≥ 18.4 U/mL) in 51.9% (228 of 439) of the patients. The cumulative event rate according to elevated CA125 and results from the univariate and multivariable analyses for the primary endpoint, as well as death and CHF separately, are provided in Table 3B. Elevated CA125 values were associated with a higher crude rate and risk of the primary endpoint (25.4% [58of 228] vs 6.6% [14 of 211], HR, 4.20; 95%CI, 2.34-7.53; P=.001) in the univariate analysis (Figure 1 of the supplementary data). After multivariable adjustment, elevated CA125 (HR, 2.83; 95%CI, 1.52-5.29; P=.001) remained independently associated with the primary endpoint (Table 3B and Table 4). Uni- and multivariable analysis for separate prediction of death and CHF are displayed in Table 3B. Elevated CA125 independently predicted readmission for CHF but not all-cause mortality after multivariable adjustment.

Information on both biomarkers was modeled together in a multivariable model for predicting time to the primary endpoint adjusted for baseline variables. Elevated Gal-3 (HR, 1.67; 95%CI, 1.02-2.75; P=.043) and elevated CA125 values (HR, 2.91; 95%CI, 1.56-5.43; P=.001) were both independently associated with the primary endpoint (Table 4).

The interaction between Gal-3 values and CA125 stratae (elevated vs not elevated) was significant when Gal-3 was modeled as a continuous (0.049) and borderline significant as a dichotomous variable (0.115). This potential differential prognostic value of Gal-3 depending upon the state of CA125 was further analyzed.

When CA125 was not elevated, elevated Gal-3 was not associated with an increased rate and adjusted risk of the primary endpoint (6.6% [10 of 151] vs 6.7% [4 of 60], HR, 1.16; 95%CI, 0.28-3.74; P=0.981). However, when CA125 was elevated, elevated Gal-3 was associated with an increased rate and adjusted risk of the primary endpoint (38.8% [31 of 80] vs 18.2% [27 of 148], HR, 2.02; 95%CI, 1.15-3.55; P=.015], Figure 3 and Table 5). For death and readmission for CHF, there was also no difference in crude event rates for elevated Gal-3 when CA125 was not elevated, but crude event rates were almost twice as high when CA125 was elevated. After multivariable adjustment, this significant difference disappeared for mortality and was borderline significant for CHF (HR, 1.83; 95%CI, 0.89-3.77; P=.099, Table 5).

Figure 3.

Kaplan-Meier event rate according to galectin-3 within CA125 stratae. Kaplan-Meier estimates for the primary endpoint with differences tested using log-rank test. +, elevated, –, not elevated; CA125, carbohydrate antigen 125; TAVI, transcatheter aortic valve implantation.

(0.13MB).

When Gal-3 elevation was added to a model containing unelevated CA125 values and baseline variables, no improvement was found in the area under the curve of the model or in net reclassification improvement, or the integrated discrimination improvement. Only when CA125 was elevated, the addition of Gal-3 elevation provided incremental prognostic value (Table 2 of the supplementary data).

TAVI is increasingly performed in patients with severe symptomatic aortic stenosis18 and its efficacy and safety in patients with high- and intermediate-risk for conventional aortic valve replacement have been demonstrated in large randomized controlled trials.19,20 Risk stratification in patients referred for TAVI is challenging and available risk scores, such as the EuroSCORE, have limited value and do not provide TAVI-specific variables.21,22 Several biomarkers have been tested for prognostic purposes2,13,23–25 and offer incremental value over and above established risk scores.4 Because several processes, such as inflammation, calcification, fibrosis/remodeling, and ventricular stress are involved in the pathophysiology of aortic valve stenosis, the incorporation of different biomarkers into clinical practice could potentially optimize risk stratification and the timing of the procedure.

Gal-3 is an emerging biomarker in cardiovascular research. It is expressed in a variety of cells, including fibroblasts and leukocytes, and is a central mediator of several pathophysiological processes, including inflammation and fibrogenesis,26–28 which are major pathophysiological mechanisms operating in the failing heart.27 These effects are supported by clinical findings showing that higher levels of Gal-3 are related to severity of heart failure and adverse clinical outcomes, in both acute and chronic CHF.7,11,12,29

Clinical studies on Gal-3 in aortic stenosis are sparse.29,30 In these studies, Gal-3 levels did not correlate well with the severity of aortic valve stenosis and did not provide further prognostic information on the occurrence of adverse events.30 In TAVI, its prognostic value remains to be determined because the results from the available studies are conflicting.13,14 While Gal-3 independently predicted all-cause mortality in 1 study,13 its significance disappeared after multivariable adjustment in another.14

The conflicting results from available studies and the fact that baseline Gal-3 loses its significance after multivariable adjustment suggests an essential role of unidentified factors that warrants further investigations.

Elevated CA125 levels have been linked to disease severity in CHF due to aortic stenosis31–33 and are independently associated with adverse prognosis after TAVI.2 Recently, it has been shown that CA125 offers independent prognostic value beyond that of the EuroSCORE4 and is superior to natriuretic peptides,3 which are the only biomarkers that are acknowledged in current guidelines.34

Recent experimental and clinical studies have suggested a potential molecular interaction between Gal-3 and CA125. Gal-3 contains a carbohydrate-recognition domain enabling specific interactions with glycosylated proteins, such as CA125.5,6 Both biomarkers seem to be involved in similar molecular signaling pathways, especially inflammatory responses,26,35 although their precise roles in cardiac inflammation have not been elucidated. Inflammation and fibrosis are key mechanisms operating in the failing heart.36

In acute heart failure, it has already been shown that the prognostic effect of Gal-3 depends on CA125 levels. Besides a mechanistic, molecular interaction, a possible heart failure-related inflammatory response for this association has been postulated.37 The biological activity of galectins is strongly dependent on the number of glycoconjugates.37 In prior experimental studies, CA125 was shown to be a galectin receptor counter.7,37 CA125, a heavily glycosylated protein, is upregulated in heart failure and aortic stenosis.1,2,32,37 Given the previous statements, we postulate that the Gal-3-CA125 interaction may represent a crucial factor in disease progression in TAVI patients. Elevation of both biomarkers probably identified a subset of patients with more adverse left ventricular remodeling in which TAVI procedures might have less impact on the natural course of the disease.

In this study, we confirm the prognostic value of CA125 in TAVI patients and additionally confirm prior data from heart failure patients suggesting a differential prognostic effect of Gal-3 across CA125 values.37 Interestingly, elevated Gal-3 levels were associated with an increased rate and adjusted risk of the primary endpoint only when CA125 values were elevated but lacked prognostic value when CA125 values were not elevated. Even more importantly, the incorporation of elevated Gal-3 to a risk model containing unelevated CA125 values did not improve the predictive capacity of this model.

Thus, Gal-3 alone does not seem to be an ideal biomarker for risk prediction in patients undergoing TAVI, as it mainly depends on CA125 levels. In TAVI patients, it appears reasonable to only determine Gal-3, if CA125 is elevated, since in this situation it offers optimal prognostic information.

Currently, risk prediction and the optimal timing of TAVI procedures is insufficient. Biomarkers could potentially optimize risk prediction beyond conventional risk scores in TAVI patients. Recent evidence suggests improved outcomes in patients with severe aortic stenosis when there is early diagnosis and invasive treatment, even in lower-risk patients.38 The interaction of 2 biomarkers reflecting different pathophysiological pathways indicates the complexity of aortic stenosis pathophysiology and suggests an integration of a multimodal approach. CA125 seems to be a key player in this scenario and should be incorporated into clinical practice. In addition, further studies are warranted to explore the biological relevance of Gal-3-CA125 interaction as a potential therapeutic target.

This is an observational study conducted in a single center. The outcomes of patients undergoing interventional procedures is complex and multifactorial. Thus, potential additional clinical determinants and procedural factors may not have been recognized in this analysis.