Lipoprotein(a) and Incident Aortic Stenosis Over 30 Years: The ARIC Study

Elevated lipoprotein(a) [Lp(a)] is a genetically determined, causal risk factor for calcific aortic valve disease and is believed to promote calcification through an oxidized phospholipid-mediated pathway.¹ Mendelian randomization and prospective studies, predominantly in White European populations, have definitively established that Lp(a) concentrations above 50mg/dL are associated with aortic stenosis (AS).² Long-term data from racially diverse cohorts, however, are lacking, and the competing risk of death has not been fully addressed. Using 30-years of follow-up in the Atherosclerosis Risk in Communities (ARIC) cohort, we examined the association of midlife Lp(a) levels with incident clinical AS, accounting for the competing risk of death. We also assessed whether Lp(a) associations differ by race and sex.

ARIC enrolled 15,792 self-reported Black and White adults at visit 1 (1987–1989; age:45–64 years) from 4-US communities. De-identified data from 14,423 participants with a baseline Lp(a) measurement and complete covariates were downloaded from BioLINCC. Lp(a) levels at visit 1 were measured with an immunoassay (kringle-IV type-2 repeat-sensitive) that quantified apolipoprotein(a) protein mass and expressed as total particle mass. Lp(a) was modeled continuously (per-SD of the log-transformed and with restricted cubic splines using 4-knots at Harrell percentiles) or at prespecified thresholds. Incident AS was defined as an AS or aortic valve replacement ascertained through active surveillance for hospitalizations with abstraction of all hospitalization ICD discharge codes through 2019. Fine-Gray regression was used to assess associations of Lp(a) with incident AS accounting for the competing risk of death. Cox-regression was used for mortality. Covariates included age, sex, race, systolic blood pressure, diabetes, antihypertensive medication, body mass index, HDL-cholesterol, log-triglycerides, LDL-cholesterol, statin treatment, and cigarette smoking. Population attributable fraction(PAF) estimated by model-based standardization (g-formula) on the absolute-risk scale at 25-years. The analysis was performed using R software (version 4.4.3) using cmprsk (functions crr, cuminc), survival (function coxph) and rms packages (function rcs).

Among 14,423 participants (mean age:54 years; 54% female; 25% Black adults; 8% diabetes), 3007(21%) had Lp(a) ≥50mg/dL. During follow-up(median, 27-years), 1113 participants(8%) developed incident AS and 7310(51%) died.

Higher Lp(a) was associated with incident AS both per SD (adjusted subhazard ratio[sHR], 1.13 [95%CI, 1.05–1.20], cause-specific HR[csHR] 1.11(1.04–1.18)) and at ≥50 versus <50mg/dL (sHR, 1.30[95%CI, 1.13–1.51], csHR 1.26[1.10–1.45]), with a graded, monotonic dose-response (Figure A–C). The 30-year cumulative incidence of AS was higher with Lp(a) ≥50 than <50mg/dL (8.6% vs 7.4%; Gray P=.02; Figure [B–D]), despite higher competing mortality in the ≥50mg/dL group (Gray P=.01), which attenuated the subdistribution estimate. The all-cause mortality HR was 1.03 (95% CI, 1.00–1.05) per 1-SD higher log-Lp(a) and 1.05 (95% CI, 0.99–1.11) for ≥50 vs <50 mg/dL.

Figure.

Lipoprotein(a) [Lp(a)] and 30-year risk of incident aortic stenosis (AS) in the ARIC study. A and C, Multivariable adjusted restricted cubic spline (using 4 knots, at Harrell’s recommended percentiles, placed at the 5th, 35th, 65th, and 95th percentiles) dose response for incident AS across Lp(a), by sex and by race, with the corresponding sex- and race-specific Lp(a) density distributions shaded in the background (reference, 10 mg/dL; vertical dotted line, 50 mg/dL). Wald test of the nonlinear spline terms (P<.001 overall; by stratum, P=.002 for men, .084 for women, <.001 for White, and .136 for Black participants). B and D, Thirty-year cumulative incidence of AS by Lp(a) <50 versus ≥50 mg/dL (solid, ≥50; dashed, <50), accounting for the competing risk of death, stratified by sex and by race. E, Multivariable adjusted Fine-Gray subdistribution hazard ratios (sHR) for incident AS by Lp(a) category (reference, <30 mg/dL) within strata; filled markers denote significant elevation. All models adjusted for age, sex, race, systolic blood pressure, diabetes, antihypertensive use, body mass index, HDL cholesterol, log-triglycerides, LDL cholesterol, statin treatment and smoking.

Associations were directionally consistent across sex and race, without significant interaction (Lp(a)-x-sex P=.42; Lp(a)-x-race P=.08), without interaction between any other covariate. Excess risk was confined to the ≥50mg/dL category in every stratum except Black adults, whom estimate was null (sHR 0.98[0.71–1.36],Figure-E). When the endpoint was restricted to stenosis-specific codes (ICD-10 I35.0/.2+AVR), associations strengthened (sHR per-SD, 1.18[95%CI, 1.07–1.31]; ≥50mg/dL, 1.52[95%CI, 1.21–1.91]). Findings were consistent across diagnosis-based and replacement-based endpoints; the null estimate in Black adults was present in the diagnosis endpoint, which does not depend on procedure referral. In the whole cohort, Lp(a) above >50mg/dL was associated with a PAF 6.5%(3.2–9.8%).

Evidence linking Lp(a) to aortic valve disease has rested largely on genetic studies and on imaging of subclinical calcification, predominantly in White European populations. Importantly our analysis extends this association to include hard clinical endpoints accrued over three decades in a community-based cohort of Black and White US adults. Our analysis also shows that this association withstands the competing risk of death during follow-up. Risk was concentrated above 50mg/dL. In a community-based US cohort, Lp(a) accounted for roughly 1 in 15 cases of AS, offering a single, inexpensive marker of an at-risk subgroup decades before clinical disease.

High Lp(a) was roughly twice as prevalent among Black adults (≥50mg/dL in 38% vs 16%), yet we did not observe an excess of AS. This null association is likely explained by limited power (77events), healthcare access, referral patterns, and outcome ascertainment, and possibly differential misclassification due to the use of an isoform-sensitive assay. Misclassification at the 50mg/dL threshold may be differentially affected by race, potentially biasing the estimate toward the null.³ Whether the lower AS prevalence in Black individuals, despite higher Lp(a) at same isoform sizes and oxidized-phospholipids⁴, reflects possible underlying racial differences in Lp(a) biology should be examined with an isoform-insensitive assay in diverse cohorts.⁵

Limitations of this study include possible residual confounding attributable to its observational nature, ascertainment by administrative codes without echocardiographic adjudication, which underestimates true AS incidence and cannot exclude subclinical valve disease at baseline, the protein-mass assay does not capture apo(a) isoform size, Lp(a) lipid content, or oxidized-phospholipid burden, which may contribute to heterogeneity in risk, and its restriction to Black and White adults.

In summary, in a large US community-based cohort, higher Lp(a) was associated with a 30-year risk of incident aortic stenosis, independent of the competing risk of death, with risk concentrated above 50mg/dL. These findings support Lp(a) as a lifelong driver of calcific aortic valve disease and highlight the need for Lp(a)-lowering trials powered for aortic stenosis endpoints in diverse populations. If efficacious, these therapies will likely have high costs, and it is critical to develop strategies to ensure equitable access.

Data Access and Data Sharing:

Data from the ARIC can be accessed by application through the BioLINCC Management System.