Study design

In this prospective, multicenter cohort study, we evaluated the clinical outcomes after TAVI according to Lp(a) concentration. The patients were stratified into high- and low‑Lp(a) groups using a prespecified cutoff of 30 mg/dl consistent with recent TAVI data.¹⁵ Additional thresholds were explored in sensitivity analyses, as described below.

The study protocol was approved by the Ethics Committee at the Medical University of Warsaw (KB/128/2018, KB/4/A2021). Written informed consent was obtained from each study participant.

Patients

The patients with severe AS (aortic valve area <⁠1 cm² or indexed aortic valve area <⁠0.6 cm²/m²) who were qualified for TAVI by a local heart team were prospectively enrolled across 3 centers. Dobutamine stress echocardiography was performed to distinguish true‑severe from pseudo‑severe AS in patients with low‑flow, low‑gradient AS and reduced left ventricular ejection fraction (LVEF). Aortic calcium score was measured on computed tomography in the patients with low‑flow, low‑gradient AS and preserved LVEF.

The patients were excluded if they were scheduled for a valve‑in‑valve procedure, had chronic kidney disease (estimated glomerular filtration rate <⁠30 ml/min/1.73 m²), autoimmune disease, active neoplastic disease, or were pregnant or breast‑feeding. For the present analysis, Lp(a) level was measured only in the patients with an available stored plasma sample suitable for testing; therefore, the final analytic cohort comprised 82 individuals.

Procedure, hospitalization, and follow‑up

All patients underwent TAVI performed by an interventional cardiologist assisted by a cardiothoracic surgeon. Antithrombotic therapy was prescribed according to the 2017 ESC/EACTS guidelines for the management of valvular heart disease, which were binding at the time of enrollment.¹⁷ Other drugs were prescribed at the discretion of the treating physician.

Patient baseline data were collected at the time of the index hospitalization. Blood samples from each patient were collected and double‑centrifuged (2500 g, 15 min, 20 C) to obtain platelet‑depleted plasma that was stored at –20 C until further analyses. Lipoprotein(a) level was measured with enzyme‑linked immunosorbent assay (ELISA; ab212165, Abcam, Cambridge, United Kingdom). Absorbance was read at 450 nm, with 570 nm as the reference wavelength for correction. Sample concentrations were calculated from an 8‑point serial dilution standard curve (including blank) fitted using linear regression and corrected for the dilution factor.

MACCEs were recorded during the follow‑up visit scheduled approximately 1 year after the procedure. Long‑term survival follow‑up was completed on November 12, 2023. Data were obtained from medical records and telephone interviews.

Clinical end points

Two primary end points were evaluated in this study. The first was MACCE, defined as a nonhierarchical composite of all‑cause death, myocardial infarction, stroke, transient ischemic attack, heart failure (HF) worsening, valve reintervention, or valve thrombosis, truncated at 12 months. The components of MACCE were defined with reference to standardized definitions.^18,19 HF worsening was defined as a sudden or rapid exacerbation of HF signs and symptoms, requiring medical attention (outpatient intensification of diuretic therapy or hospitalization).

The second primary end point was overall survival truncated at the common longest follow‑up time in both groups. Secondary end points in this study were the individual components of the first primary composite end point.

Subgroup analysis

Subgroup analyses for the primary end points were performed according to sex, body mass index (<⁠27 kg/m² vs ≥27 kg/m²), atrial fibrillation, chronic kidney disease, diabetes mellitus, age (<⁠85 y vs ≥85 y), New York Heart Association class (I–II vs III–IV), and history of stroke.

Statistical analysis

Continuous variables were presented as mean (SD) or median (interquartile range [IQR]), as appropriate. Categorical variables were presented as count (percentage). Differences between the groups were calculated using the t test or Wilcoxon rank‑sum test for continuous variables and the χ² test or Fisher exact test for categorical variables, as appropriate.

The primary end points were analyzed using the Kaplan–Meier time‑to‑event curves and compared between the Lp(a) groups with the log‑rank test. Sensitivity analyses included exclusion of periprocedural events (≤30 days) and redefinition of Lp(a) cutoff to 50 and 75 mg/dl. Additionally, Cox proportional hazards models adjusted for age and sex were used to analyze the primary end points. Unadjusted Cox proportional hazards models were fitted for the secondary end points and for the subgroup analyses. P values for interaction between the subgroups and P values for all Cox proportional hazards models were derived from the likelihood ratio test. Given the limited sample size and statistical power, the subgroup analyses, sensitivity analyses, and multivariable Cox regression analyses should be considered exploratory and hypothesis‑generating. For all analyses, a P value below 0.05 was considered significant.

All statistical analyses were performed with R software, version 4.5.2 (R Foundation for Statistical Computing, Vienna, Austria).

Patients and procedures

From November 2018 to September 2021, TAVI was performed across 3 clinical sites. However, Lp(a) levels could only be assessed if an adequate stored plasma sample was available, which yielded an analytic cohort of 82 patients. Of these, 60 had high Lp(a) concentration (≥30 mg/dl) and 22 had low Lp(a) concentration (<⁠30 mg/dl). The baseline characteristics of the patients are presented in Table 1. Median (IQR) age of the patients was 80 (75–83) years, 51.2% were men, and median EuroSCORE II score was 3.6 (2.5–4.7). Median follow‑up time was 2.8 (2.1–3.8) years and the maximum follow‑up time was 4.7 years. The patients in the high‑Lp(a) group had higher concentrations of total cholesterol (TC; P = 0.03) and lower concentrations of triglycerides (P = 0.049) than the patients in the low‑Lp(a) group.

Information on the type of implanted valves and access routes is shown in Supplementary material, Table S1, with no differences between the study groups. Paravalvular leak was the most common periprocedural complication (Supplementary material, Table S2). Medications at discharge were also similar in both groups (Supplementary material, Table S3).

Primary end points

In the period from baseline to 12 months, MACCEs occurred in 4 of 22 patients (18.2%) in the low‑Lp(a) group and in 10 of 60 patients (16.7%) in the high‑Lp(a) group (log‑rank P = 0.95; Figure 1A). No marked difference regarding MACCEs was found between the groups in a Cox regression model adjusted for age and sex (hazard ratio [HR], 0.95; 95% CI, 0.3–3.05; P = 0.93; Table 2).

During long‑term follow‑up (median, 2.8 [2.1–3.8] y; maximum 4.7 y), death from any cause occurred in 3 of 22 patients (13.6%) in the low‑Lp(a) group and in 23 of 60 patients (38.3%) in the high‑Lp(a) group (log‑rank P = 0.045; Figure 1B). The difference lost significance after adjustments for age and sex in a Cox regression model (HR, 2.85; 95% CI, 0.85–9.55; P = 0.054; Table 2).

Secondary end points

In the period from baseline to 12 months, death from any cause was the most common MACCE component and it occurred in 6 patients (10%) in the high‑Lp(a) group and in 1 patient (4.5%) in the low‑Lp(a) group (Table 2). The second most common secondary end point was HF worsening that occurred in 3 patients (5%) in the high‑Lp(a) group and 3 patients (13.6%) in the low‑Lp(a) group (Table 2). None of the secondary end points differed significantly between the groups (Table 2).

Subgroup analyses

Effect sizes for MACCEs were heterogeneous in the analyzed subgroups, but no significant interaction was detected (Figure 2A). Effect sizes for all‑cause death were consistent across most subgroups, except for those defined by age and prior stroke (Figure 2B). High Lp(a) level was associated with an increased risk of death in the patients younger than 85 years (HR, 8.51; 95% CI, 1.14–63.45; Figure 2B), whereas no such association was observed in those aged 85 years or older (HR, 0.54; 95% CI, 0.09–3.26; P for interaction = 0.03; Figure 2B). High Lp(a) level was also associated with a higher risk of death in the patients without prior stroke (HR, 4.73; 95% CI, 1.11–20.14; Figure 2B), while this pattern was not evident in the patients with a history of stroke (HR, 0.45; 95% CI, 0.03–7.18; Figure 2B); however, the interaction did not reach significance (P for interaction = 0.11; Figure 2B).

Sensitivity analysis

No differences in MACCE‑free survival between the high- and low‑Lp(a) groups were observed in the following sensitivity analyses: exclusion of periprocedural (≤30 days) events (log‑rank P = 0.64; Figure 3A), Lp(a) cutoff of 50 mg/dl (log‑rank P = 0.79; Figure 3B), and Lp(a) cutoff of 75 mg/dl (log‑rank P = 0.81; Figure 3C).

No differences in long‑term overall survival between the high- and low‑Lp(a) groups were observed in the following sensitivity analyses: exclusion of periprocedural (≤30 days) events (log‑rank P = 0.07; Figure 4A), Lp(a) cutoff of 50 mg/dl (log‑rank P = 0.1; Figure 4B), and Lp(a) cutoff of 75 mg/dl (log‑rank P = 0.57; Figure 4C).

Limitations

There are several limitations to our study that should be acknowledged. First, the Lp(a) level measurements were available only in the patients with an adequate stored plasma sample, resulting in a reduced size of the analyzed cohort. This may have introduced selection bias and limit applicability of our findings to the general TAVI population. Second, the relatively small number of patients may have limited the statistical power to detect some associations. Nevertheless, we observed a significant difference in overall survival and identified other potentially important trends, which should be interpreted as exploratory. Further, we did not adjust our analyses for cholesterol or triglycerides levels, which differed between the groups, because full lipid panel was available only for about 70% of the patients. Nevertheless, the observed difference in TC is expected, as Lp(a)-cholesterol concentration contributes to TC level. Adjusting for TC could even lead to overadjustment and attenuation of the effect of Lp(a) level mediated via the cholesterol‑related pathway. Last, Lp(a) levels were measured several years after blood sample collection, using an ELISA in platelet‑depleted plasma. As a consequence, the absolute Lp(a) values may not be directly comparable with commonly used thresholds, and the distribution of Lp(a) concentrations in our cohort seems right‑shifted in comparison with previous studies. However, this systematic shift in Lp(a) levels likely impacts all patients in the same way, and thus the relative differences between the groups categorized by Lp(a) levels remain significant and clinically meaningful.

Conclusions

Patients with severe AS and elevated Lp(a) level have a comparable risk of 12‑month MACCE after TAVI to those with low Lp(a) level, while unadjusted analyses showed a trend toward worse long‑term survival in the high‑Lp(a) group. Based on the exploratory analyses, the difference in long‑term mortality appears more pronounced in younger patients with a lower burden of comorbidities. These findings require confirmation in larger cohorts.