[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Figure 1.
Mendelian Randomization of LDL-C and Risk of Aortic Stenosis in the Malmö Diet and Cancer Study
Mendelian Randomization of LDL-C and Risk of Aortic Stenosis in the Malmö Diet and Cancer Study

The aim of Mendelian randomization is to provide a robust test of the association between low-density lipoprotein cholesterol (LDL-C) and aortic stenosis (association 3). Association 3 can be tested simply using standard epidemiologic methods, but these methods may be biased (eg, confounding, reverse causality, etc). To overcome this bias, Mendelian randomization indirectly tests association 3 by first establishing via linear regression that LDL-C–related single-nucleotide polymorphisms (SNPs) increase LDL-C (association 1). These LDL-C SNPs are then tested for an association with aortic stenosis (association 2). Under the assumption that the entire effect of the LDL-C SNPs on aortic stenosis (association 2) is mediated by their effect on increasing LDL-C (association 1), an unconfounded assessment of association 3 can be obtained (ie, instrumental variable estimate).

Figure 2.
Magnitude of Genetic Increase in LDL-C and Odds of Aortic Valve Calcium Across All LDL-C SNPs in CHARGE Participants
Magnitude of Genetic Increase in LDL-C and Odds of Aortic Valve Calcium Across All LDL-C SNPs in CHARGE Participants

Each dot represents a single low-density lipoprotein cholesterol (LDL-C) single-nucleotide polymorphism (SNP). Across all 57 LDL-C–associated SNPs, a given genetic increase in LDL-C is correlated with a concomitant increase in the odds of aortic valve calcium (AVC). The solid line represents the best line of fit, and the dashed lines represent the 95% CI for this relationship. P value reported is for the linear association.

Table 1.  
Characteristics of CHARGE Participants
Characteristics of CHARGE Participants
Table 2.  
Baseline Characteristics for Participants of the Malmö Diet and Cancer Study
Baseline Characteristics for Participants of the Malmö Diet and Cancer Study
Table 3.  
Baseline Lipid Levels by Incident Aortic Stenosis Status and Associated Relative Risks Among MDCS Participants With Available Lipid Values
Baseline Lipid Levels by Incident Aortic Stenosis Status and Associated Relative Risks Among MDCS Participants With Available Lipid Values
Table 4.  
Association of Genetic Risk Scores With Aortic Valve Calcium, Aortic Stenosis, and Aortic Valve Replacementa
Association of Genetic Risk Scores With Aortic Valve Calcium, Aortic Stenosis, and Aortic Valve Replacementa
1.
Go  AS, Mozaffarian  D, Roger  VL,  et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee.  Heart disease and stroke statistics: 2013 update: a report from the American Heart Association. Circulation. 2013;127(1):e6-e245.
PubMedArticle
2.
Iung  B, Baron  G, Butchart  EG,  et al.  A prospective survey of patients with valvular heart disease in Europe: the Euro Heart Survey on Valvular Heart Disease. Eur Heart J. 2003;24(13):1231-1243.
PubMedArticle
3.
Stewart  BF, Siscovick  D, Lind  BK,  et al.  Clinical factors associated with calcific aortic valve disease: Cardiovascular Health Study. J Am Coll Cardiol. 1997;29(3):630-634.
PubMedArticle
4.
Thanassoulis  G, Campbell  CY, Owens  DS,  et al; CHARGE Extracoronary Calcium Working Group.  Genetic associations with valvular calcification and aortic stenosis. N Engl J Med. 2013;368(6):503-512.
PubMedArticle
5.
Cowell  SJ, Newby  DE, Prescott  RJ,  et al; Scottish Aortic Stenosis and Lipid Lowering Trial, Impact on Regression (SALTIRE) Investigators.  A randomized trial of intensive lipid-lowering therapy in calcific aortic stenosis. N Engl J Med. 2005;352(23):2389-2397.
PubMedArticle
6.
Rossebø  AB, Pedersen  TR, Boman  K,  et al; SEAS Investigators.  Intensive lipid lowering with simvastatin and ezetimibe in aortic stenosis. N Engl J Med. 2008;359(13):1343-1356.
PubMedArticle
7.
Chan  KL, Teo  K, Dumesnil  JG, Ni  A, Tam  J; ASTRONOMER Investigators.  Effect of lipid lowering with rosuvastatin on progression of aortic stenosis: results of the aortic stenosis progression observation: measuring effects of rosuvastatin (ASTRONOMER) trial. Circulation. 2010;121(2):306-314.
PubMedArticle
8.
Kamstrup  PR, Tybjaerg-Hansen  A, Steffensen  R, Nordestgaard  BG.  Genetically elevated lipoprotein(a) and increased risk of myocardial infarction. JAMA. 2009;301(22):2331-2339.
PubMedArticle
9.
Tikkanen  E, Tuovinen  T, Widén  E,  et al.  Association of known loci with lipid levels among children and prediction of dyslipidemia in adults. Circ Cardiovasc Genet. 2011;4(6):673-680.
PubMedArticle
10.
Do  R, Willer  CJ, Schmidt  EM,  et al.  Common variants associated with plasma triglycerides and risk for coronary artery disease. Nat Genet. 2013;45(11):1345-1352.
PubMedArticle
11.
Psaty  BM, O’Donnell  CJ, Gudnason  V,  et al; CHARGE Consortium.  Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium: design of prospective meta-analyses of genome-wide association studies from 5 cohorts. Circ Cardiovasc Genet. 2009;2(1):73-80.
PubMedArticle
12.
Smith  JG, Platonov  PG, Hedblad  B, Engström  G, Melander  O.  Atrial fibrillation in the Malmö Diet and Cancer study: a study of occurrence, risk factors and diagnostic validity. Eur J Epidemiol. 2010;25(2):95-102.
PubMedArticle
13.
Teslovich  TM, Musunuru  K, Smith  AV,  et al.  Biological, clinical and population relevance of 95 loci for blood lipids. Nature. 2010;466(7307):707-713.
PubMedArticle
14.
Willer  CJ, Schmidt  EM, Sengupta  S,  et al; Global Lipids Genetics Consortium.  Discovery and refinement of loci associated with lipid levels. Nat Genet. 2013;45(11):1274-1283.
PubMedArticle
15.
PLINK: whole genome association analysis toolset. http://pngu.mgh.harvard.edu/~purcell/plink/. Accessed October 6, 2014.
16.
rproject.org. http://cran.rproject.org/web/packages/gtx/index.html. Accessed October 6, 2014.
17.
Ehret  GB, Munroe  PB, Rice  KM,  et al; International Consortium for Blood Pressure Genome-Wide Association Studies; CARDIoGRAM Consortium; CKDGen Consortium; KidneyGen Consortium; EchoGen Consortium; CHARGE-HF Consortium.  Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature. 2011;478(7367):103-109.
PubMedArticle
18.
Burgess  S, Butterworth  A, Thompson  SG.  Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658-665.
PubMedArticle
19.
Pierce  BL, Burgess  S.  Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators. Am J Epidemiol. 2013;178(7):1177-1184.
PubMedArticle
20.
Cosmi  JE, Kort  S, Tunick  PA,  et al.  The risk of the development of aortic stenosis in patients with “benign” aortic valve thickening. Arch Intern Med. 2002;162(20):2345-2347.
PubMedArticle
21.
Coffey  S, Cox  B, Williams  MJA.  The prevalence, incidence, progression, and risks of aortic valve sclerosis: a systematic review and meta-analysis. J Am Coll Cardiol. 2014;63(25 pt A):2852-2861.
PubMedArticle
22.
Rajamannan  NM, Edwards  WD, Spelsberg  TC.  Hypercholesterolemic aortic-valve disease. N Engl J Med. 2003;349(7):717-718.
PubMedArticle
23.
Thanassoulis  G, Massaro  JM, Cury  R,  et al.  Associations of long-term and early adult atherosclerosis risk factors with aortic and mitral valve calcium. J Am Coll Cardiol. 2010;55(22):2491-2498.
PubMedArticle
24.
Owens  DS, Katz  R, Johnson  E,  et al.  Interaction of age with lipoproteins as predictors of aortic valve calcification in the Multi-Ethnic Study of Atherosclerosis. Arch Intern Med. 2008;168(11):1200-1207.
PubMedArticle
25.
Rosenhek  R, Rader  F, Loho  N,  et al.  Statins but not angiotensin-converting enzyme inhibitors delay progression of aortic stenosis. Circulation. 2004;110(10):1291-1295.
PubMedArticle
26.
Mundy  G, Garrett  R, Harris  S,  et al.  Stimulation of bone formation in vitro and in rodents by statins. Science. 1999;286(5446):1946-1949.
PubMedArticle
27.
O’Brien  KD, Reichenbach  DD, Marcovina  SM, Kuusisto  J, Alpers  CE, Otto  CM.  Apolipoproteins B, (a), and E accumulate in the morphologically early lesion of ‘degenerative’ valvular aortic stenosis. Arterioscler Thromb Vasc Biol. 1996;16(4):523-532.
PubMedArticle
28.
Otto  CM, Kuusisto  J, Reichenbach  DD, Gown  AM, O’Brien  KD.  Characterization of the early lesion of ‘degenerative’ valvular aortic stenosis: histological and immunohistochemical studies. Circulation. 1994;90(2):844-853.
PubMedArticle
29.
Hirsch  D, Azoury  R, Sarig  S, Kruth  HS.  Colocalization of cholesterol and hydroxyapatite in human atherosclerotic lesions. Calcif Tissue Int. 1993;52(2):94-98.
PubMedArticle
30.
Parhami  F, Morrow  AD, Balucan  J,  et al.  Lipid oxidation products have opposite effects on calcifying vascular cell and bone cell differentiation: a possible explanation for the paradox of arterial calcification in osteoporotic patients. Arterioscler Thromb Vasc Biol. 1997;17(4):680-687.
PubMedArticle
31.
Miller  JD, Weiss  RM, Serrano  KM,  et al.  Lowering plasma cholesterol levels halts progression of aortic valve disease in mice. Circulation. 2009;119(20):2693-2701.
PubMedArticle
32.
VanderWeele  TJ, Tchetgen Tchetgen  EJ, Cornelis  M, Kraft  P.  Methodological challenges in Mendelian randomization. Epidemiology. 2014;25(3):427-435.
PubMedArticle
Original Investigation
November 5, 2014

Association of Low-Density Lipoprotein Cholesterol–Related Genetic Variants With Aortic Valve Calcium and Incident Aortic Stenosis

Author Affiliations
  • 1Department of Cardiology, Clinical Sciences, Lund University, Lund, Sweden
  • 2Department of Heart Failure and Valvular Disease, Skåne University Hospital, Lund, Sweden
  • 3Department of Clinical Sciences, Lund University, Malmö, Sweden
  • 4Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, Massachusetts
  • 7Center for Human Genetic Research, Massachusetts General Hospital and Harvard Medical School, Boston
  • 5McGill University Health Center, Preventive and Genomic Cardiology, Montreal, Quebec, Canada
  • 6McGill University Health Center and Research Institute, Department of Medicine, Montreal, Quebec, Canada
  • 8Department of Medicine, University of Washington, Seattle
  • 9National Institute on Aging, Bethesda, Maryland
  • 10Department of Biostatistics, University of Washington, Seattle
  • 11Icelandic Heart Association Research Institute, Kopavogur, Iceland
  • 12Faculty of Medicine, University of Iceland, Reykjavik, Iceland
  • 13Los Angeles Biomedical Research Institute at Harbor–UCLA, Los Angeles, California
  • 14Department of Biostatistics, Boston University School of Public Health, Boston, Massachusetts
  • 15Framingham Heart Study, Framingham, Massachusetts
  • 16University of Virginia, Charlottesville
  • 17Cardiology Division, Massachusetts General Hospital, Boston
  • 18NHLBI Cardiovascular Epidemiology and Human Genomics Branch, National Heart, Lung, and Blood Institute, Bethesda, Maryland
  • 19Division of Cardiology, Department of Medicine, Johns Hopkins University, Baltimore, Maryland
JAMA. 2014;312(17):1764-1771. doi:10.1001/jama.2014.13959
Abstract

Importance  Plasma low-density lipoprotein cholesterol (LDL-C) has been associated with aortic stenosis in observational studies; however, randomized trials with cholesterol-lowering therapies in individuals with established valve disease have failed to demonstrate reduced disease progression.

Objective  To evaluate whether genetic data are consistent with an association between LDL-C, high-density lipoprotein cholesterol (HDL-C), or triglycerides (TG) and aortic valve disease.

Design, Setting, and Participants  Using a Mendelian randomization study design, we evaluated whether weighted genetic risk scores (GRSs), a measure of the genetic predisposition to elevations in plasma lipids, constructed using single-nucleotide polymorphisms identified in genome-wide association studies for plasma lipids, were associated with aortic valve disease. We included community-based cohorts participating in the CHARGE consortium (n = 6942), including the Framingham Heart Study (cohort inception to last follow-up: 1971-2013; n = 1295), Multi-Ethnic Study of Atherosclerosis (2000-2012; n = 2527), Age Gene/Environment Study–Reykjavik (2000-2012; n = 3120), and the Malmö Diet and Cancer Study (MDCS, 1991-2010; n = 28 461).

Main Outcomes and Measures  Aortic valve calcium quantified by computed tomography in CHARGE and incident aortic stenosis in the MDCS.

Results  The prevalence of aortic valve calcium across the 3 CHARGE cohorts was 32% (n = 2245). In the MDCS, over a median follow-up time of 16.1 years, aortic stenosis developed in 17 per 1000 participants (n = 473) and aortic valve replacement for aortic stenosis occurred in 7 per 1000 (n = 205). Plasma LDL-C, but not HDL-C or TG, was significantly associated with incident aortic stenosis (hazard ratio [HR] per mmol/L, 1.28; 95% CI, 1.04-1.57; P = .02; aortic stenosis incidence: 1.3% and 2.4% in lowest and highest LDL-C quartiles, respectively). The LDL-C GRS, but not HDL-C or TG GRS, was significantly associated with presence of aortic valve calcium in CHARGE (odds ratio [OR] per GRS increment, 1.38; 95% CI, 1.09-1.74; P = .007) and with incident aortic stenosis in MDCS (HR per GRS increment, 2.78; 95% CI, 1.22-6.37; P = .02; aortic stenosis incidence: 1.9% and 2.6% in lowest and highest GRS quartiles, respectively). In sensitivity analyses excluding variants weakly associated with HDL-C or TG, the LDL-C GRS remained associated with aortic valve calcium (P = .03) and aortic stenosis (P = .009). In instrumental variable analysis, LDL-C was associated with an increase in the risk of incident aortic stenosis (HR per mmol/L, 1.51; 95% CI, 1.07-2.14; P = .02).

Conclusions and Relevance  Genetic predisposition to elevated LDL-C was associated with presence of aortic valve calcium and incidence of aortic stenosis, providing evidence supportive of a causal association between LDL-C and aortic valve disease. Whether earlier intervention to reduce LDL-C could prevent aortic valve disease merits further investigation.

Introduction

Aortic valve disease remains the most common form of heart valve disease in Europe and North America and is the most common cause of valve replacement.1,2 Despite the heavy disease burden, no medical treatments are known to stop or retard disease progression. Although aortic valve disease shares several risk factors with vascular disease,3 it remains largely unknown which factors are causal and should be targeted to reduce valve disease. Our group recently described evidence for a causal association between a common variant in the LPA gene, via elevated plasma lipoprotein(a) (Lp[a]), and aortic valve disease.4 Whether other plasma lipids are causally associated with the development of aortic valve disease remains unclear.

Low-density lipoprotein cholesterol (LDL-C) is an important risk factor for aortic valve disease in epidemiologic studies3; however, large randomized trials of LDL-C–lowering therapy in patients with advanced aortic stenosis have failed to demonstrate effectiveness in reducing disease progression.57 Nonetheless, if LDL-C plays a causal role in the earlier stages of aortic valve disease, this could have important implications for prevention.

Because of the random allocation of genetic information that occurs at conception, genetic variation can be used as an effective tool to distinguish potentially causal from noncausal biomarkers. Termed “Mendelian randomization,” this approach has been successfully applied to assess for causality of several biomarkers with various clinical end points.4,8 Genetic risk scores (GRSs) for lipids, incorporating multiple genetic variants, have been shown to be strongly associated with their corresponding lipid levels in both children9 and adults,10 providing strong support for the contention that a higher GRS confers life-long exposure to higher lipid levels. Here, we used a Mendelian randomization approach to determine whether genetic contributions to elevations in LDL-C and other lipids were associated with early subclinical aortic valve disease and incident clinical aortic stenosis.

Methods

Associations of GRSs with aortic valve calcium were evaluated in the 3 CHARGE cohorts where data from computed tomographic (CT) imaging were available; associations with incident aortic stenosis were estimated in the population-based Malmö Diet and Cancer Study (MDCS).

CHARGE Cohorts

The Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium is an ongoing collaboration among several large, well-phenotyped, prospective, longitudinal cohort studies from the United States and Europe.11 This analysis included white European participants with aortic valve calcium data and genome-wide association study (GWAS) data from the Multi-Ethnic Study of Atherosclerosis (MESA, n = 2527; cohort inception to last follow-up: 2000-2012), the Age, Gene/Environment Susceptibility–Reykjavik Study (AGES-RS, n = 3120; from 2000-2012), and the Framingham Heart Study (FHS, n = 1295, from 1971-2013). Details of the CHARGE consortium and each of these cohorts have been previously described.4,11 The studies were approved by relevant regulatory bodies and participants provided informed consent.

Malmö Diet and Cancer Study

The MDCS is a prospective, population-based cohort study from the city of Malmö in southern Sweden. Data collection, sample characteristics, and clinical definitions for MDCS have been described previously.12 Briefly, 30 447 randomly selected men born between 1923 and 1945 and women born between 1923 and 1950 attended a baseline examination between 1991 and 1996. Follow-up for events was completed on December 31, 2010. For the present study, DNA was available for 28 461 participants. Participants underwent sampling of peripheral venous blood and measurement of blood pressure and anthropometric measures and filled out a questionnaire. Measurements of cholesterol (high-density lipoprotein cholesterol [HDL-C], triglycerides [TG], total cholesterol) were performed in a random subset of 5269 participants, on fresh blood samples according to standard procedures at the Department of Clinical Chemistry, Skåne University Hospital, Malmö.12 Levels of LDL-C were calculated according to the Friedewald formula, with exclusion of individuals where TG exceeded 4.52 mmol/L (to convert TG to mg/dL, divide by 0.0113). Informed consent was obtained from all participants and the study was approved by the ethics committee of Lund University, Sweden.

Genotyping, Imputation, and Construction of Lipid Genetic Scores
CHARGE Cohorts

Details of the genotyping and HapMap imputation methods used in the FHS, AGES-RS, and MESA have previously been described.4,11 To construct each lipid GRS, we included previously reported, linkage disequilibrium-pruned (ie, uncorrelated) single-nucleotide polymorphisms (SNPs) from large-scale lipid GWAS13,14 that were associated with the primary lipid traits (LDL-C, HDL-C, and TG) at P value <5.0 × 10−8 and were available in the aortic valve calcium GWAS data set. We constructed separate weighted genetic scores for LDL-C (57 SNPs), HDL-C (70 SNPs), and TG (39 SNPs) weighted by the β coefficients of each individual SNP from large-scale lipid GWASs.13,14 The possible range of the GRS was from 0 to 5.17, 0 to 6.94, and 0 to 3.66 for LDL-C, HDL-C, and TG, respectively. Because of the known association of aortic valve calcium and aortic stenosis with Lp(a), we specifically excluded the rs10455872 SNP, which is strongly associated with Lp(a).4 We also calculated secondary SNP scores that were specific for each single lipid trait by excluding SNPs that were associated with a secondary lipid trait (P value <5.0 × 10−8). The SNPs included in each score are summarized in eTables 1 through 3 in the Supplement.

Malmö Diet and Cancer Study

Genotyping and genotype quality control for the 116 SNPs associated with plasma concentrations of LDL-C, HDL-C, or TG in the Global Lipid Genetics consortium (GLGC) study13 in the MDCS are described in detail in the eMethods in the Supplement. The final SNPs included in the genetic scores for LDL-C (31 SNPs), HDL-C (41 SNPs), and TG (26 SNPs) are summarized in eTables 4-7 in the Supplement. Weighted GRSs were calculated using PLINK version 1.0715 as the sum of the number of alleles associated with increased LDL-C or TG or decreased HDL-C weighted by the β coefficients (expressed in mmol/L) of each individual SNP from the GLGC study. The possible range of the GRS was from 0 to 3.61, 0 to 1.70, and 0 to 2.84 for LDL-C, HDL-C, and TG, respectively. The secondary SNP scores specific for each lipid trait are described in eTables 4-6 in the Supplement.

Outcomes
Aortic Valve Calcium in CHARGE Cohorts

To assess aortic valve calcium, standard CT scanning was performed on all participants and images were analyzed for the presence of aortic valve calcium using offline digital software. Using standard Agatston methods, the presence of calcium was defined as 3 or more contiguous pixels with a brightness of at least 130 Hounsfield units.4 Aortic valve calcium was defined as calcium residing within the aortic-valve leaflets or commissures, excluding aortic annulus, proximal aorta, and coronary arteries.

Aortic Stenosis in MDCS

Prevalent and incident diagnoses of aortic stenosis and aortic valve replacement were identified by record linkage with nationwide registers on hospitalizations and causes of death as described previously.4 Incident aortic stenosis includes all cases of aortic stenosis identified in the MDCS whereas the aortic valve replacement outcome includes all aortic stenosis cases that underwent aortic valve replacement in the MDCS. High diagnostic validity of aortic stenosis in national Swedish registers has previously been confirmed, and most patients with an aortic stenosis diagnosis in national registers had moderate to severe aortic stenosis.4 A summary of case ascertainment procedures is provided in the eMethods in the Supplement.

Statistical Analysis
CHARGE Cohorts

To test the hypothesis of a possible causal association between each lipid trait and aortic valve calcium, we used GWAS summary-level data from the 3 CHARGE participating cohorts with aortic valve calcium data.4 The CHARGE aortic valve calcium GWAS has been previously described4 and was performed using an age- and sex-adjusted additive model. To estimate the association for each lipid GRS (LDL-C, HDL-C, and TG) with aortic valve calcium presence, we extracted summary-level SNP data from the CHARGE aortic valve calcium GWAS for all lipid SNPs in each GRS and used the Genetics ToolboX (gtx) R package version 0.0816 to generate GRS effect sizes (βGRS) expressed as odds ratios (ORGRS = eβGRS) with 95% confidence intervals for presence of aortic valve calcium. This method has been shown to provide GRS estimates that are equivalent to using individual participant data for GRS association studies.17,18 Details of the statistical methods can be found in the eMethods in the Supplement.

We also performed 3 sensitivity analyses to address the issue of genetic pleiotropy, in which a given lipid SNP may have several biological effects other than their known effects on plasma lipid levels. First, using the same approach outlined above, we reanalyzed the association between each lipid GRS and aortic valve calcium using a secondary “specific” GRS in which SNPs associated with more than 1 lipid were excluded. Second, we also addressed pleiotropy using a recently developed regression method called multivariable Mendelian randomization,10 which uses a linear regression model to adjust for any possible known pleiotropic effects of the included SNPs by other lipids traits. The details of this method are available in the eMethods in the Supplement. Third, as a final sensitivity analysis, we simulated 1000 LDL-C GRS randomly excluding 30% of the available LDL-C SNPs in each set. Each LDL-C GRS was then tested against aortic valve calcium and the β coefficient for the GRS association was plotted against the LDL-C effect for each of these LDL-C GRS. We also report the mean GRS association and the upper and lower 95% confidence intervals for the effect estimate across all 1000 simulated GRS.

Malmö Diet and Cancer Study

In contrast to the summary-level data used in the CHARGE cohorts, direct genotype-phenotype data were used in the MDCS. The association of each linear GRS with incident clinical aortic stenosis was examined using Cox proportional hazards regression with censoring at death from other causes, emigration, or aortic valve replacement from other causes, adjusting for age, sex, height, weight, diabetes, hypertension, and current smoking at baseline. Age was used as the time scale, and both prevalent and incident diagnoses at the first study examination were included. Wald tests were used for significance testing of the hazard ratios (HRs). A sensitivity analysis was also performed, restricted to participants who underwent aortic valve replacement due to aortic stenosis. Cumulative incidence by GRS quartiles was calculated using cumulative incidence functions (CIF macro in SAS).

To estimate the possible causal association of LDL-C on aortic stenosis risk, an instrumental variable analysis was performed in the MDCS that considered both the GRS association with aortic stenosis and the GRS association with LDL-C and provided an estimate for the increased risk of aortic stenosis per unit change in LDL-C. We used a 2-stage regression approach in which (1) the association between the LDL GRS and LDL-C was estimated in the subset of MDCS with cholesterol measurements (n = 5269) and (2) the association between the predicted LDL-C levels (from the first level regression) and incident aortic stenosis was estimated in the entire MDCS cohort (n = 28 461). Analyses were adjusted for age and sex. To account for error terms across both stages, we used the delta method to estimate standard errors of the instrumental variable ratio estimates.19 Analyses in the MDCS were performed using IBM SPSS version 21.0 (IBM Corporation), Stata version 12.0 (StataCorp), and SAS version 9.4 (SAS Institute). For all analyses, we considered a 2-tailed P value of <.05 to be statistically significant.

Results

Baseline characteristics of CHARGE and MDCS participants are detailed in Table 1 and Table 2. The numbers of individuals who qualified for analysis were as follows: FHS, n = 1295; AGES-RS, n = 3120; MESA, n = 2527; MDCS, n = 28 461. The prevalence of aortic valve calcium across the 3 CHARGE cohorts was 32% (n = 2245). In the MDCS, over a median follow-up time of 16.1 years (interquartile range, 14.8-17.7 years), aortic stenosis developed in 17 per 1000 participants (n = 473) and aortic valve replacement for aortic stenosis occurred in 7 per 1000 (n = 205). In the MDCS, we confirmed that each lipid GRS was associated with the corresponding lipid level as shown in eTable 8 in the Supplement. Each GRS explained 7.1%, 5.7%, and 4.3% of the variance of LDL-C, HDL-C, and TG, respectively. Pleiotropic associations with other lipid traits were also noted for all scores. However, the specific LDL-C GRS that excluded variants with associations with lipid fractions in the GLGC study (at P < 5.0 × 10−8) other than LDL-C was only significantly associated with LDL-C and total cholesterol (eTable 8 in the Supplement). The LDL-C GRS was also not associated with Lp(a) levels in the CHARGE cohorts with available Lp(a) data. As expected, Lp(a) levels were appropriately randomized across LDL-C GRS tertiles (eTables 9-10 in the Supplement).

Association Between Plasma Cholesterol Levels and Aortic Stenosis

In the subcohort of the MDCS where lipid fractions were measured (n = 5269), baseline LDL-C was significantly associated with incident aortic stenosis (HR per mmol/L, 1.28; 95% CI, 1.04-1.57; P = .02; incidence of aortic stenosis: 1.3% and 2.4% in lowest and highest LDL-C quartiles, respectively). Baseline HDL-C or TG levels were not significantly associated with incident aortic stenosis (Table 3).

Lipid GRSs and Aortic Valve Calcium in CHARGE Participants

In CHARGE participants, the LDL-C GRS, but not the HDL-C or the TG GRS, was significantly associated with presence of aortic valve calcium (OR per GRS increment, 1.38; 95% CI, 1.09-1.74; P = .007) (Table 4). After exclusion of pleiotropic SNPs that were associated with more than 1 lipid trait, a “specific” LDL-C GRS remained associated with presence of aortic valve calcium (OR per GRS increment, 1.39; 95% CI, 1.04-1.86; P = .03).

Lipid GRSs, LDL-C, and Aortic Stenosis in MDCS Participants

In the MDCS, the LDL-C GRS was strongly associated with LDL-C (+2.4 mmol/L per GRS increment; P = 4 × 10−93). Only the LDL-C GRS, but not the HDL-C or TG GRS, was significantly associated with increased aortic stenosis incidence (HR per GRS increment in LDL-C, 2.78; 95% CI, 1.22-6.37; P = .02; aortic stenosis incidence: 1.9% and 2.6% in lowest and highest GRS quartiles, respectively) (Table 4). For aortic stenosis requiring aortic valve replacement, the increased risk was not statistically significant (HR per GRS increment, 3.15; 95% CI, 0.90-11.03; P = .06; aortic valve replacement incidence: 0.8% and 1.4% in lowest and highest GRS quartiles, respectively). After exclusion of pleiotropic SNPs that were associated with more than 1 lipid trait, a “specific” LDL-C GRS remained associated with incident aortic stenosis (HR per GRS increment, 3.85; 95% CI, 1.37-10.79; P = .009; aortic stenosis incidence: 1.6% and 2.7% in lowest and highest GRS quartiles, respectively).

In formal instrumental variable analysis, which may provide a more robust causal association estimate between change in LDL-C and risk of aortic stenosis, genetic increases in LDL-C were significantly associated with an increased risk of aortic stenosis (HR per mmol/L, 1.51; 95% CI, 1.07-2.14; P = .02) (Figure 1).

Sensitivity Analyses

In CHARGE participants, we also used multivariable Mendelian randomization, a regression-based method, to assess the association between βLDL and βaortic valve calcium and observed a strong linear association (P for linear trend 7 × 10−13) indicating a significant relation between the magnitude of change in LDL-C and the increase in aortic valve calcium presence (Figure 2). Adjustments for βHDL and βTG did not materially change the observed relation between βLDL and βaortic valve calcium (P for linear trend 3 × 10−13) (eTable 11 in the Supplement). Finally, the results after simulating 1000 LDL-C GRS by random SNP exclusion were largely in keeping with the main analysis, demonstrating no major dependencies of the LDL-C GRS association with aortic valve calcium on any set of specific SNPs (eFigure 1 in the Supplement).

Discussion

In this study of 6942 participants with subclinical data on aortic valve calcium and more than 28 000 participants with more than 15 years of follow-up for aortic stenosis, we demonstrate that genetic elevation in LDL-C, as determined by a GRS, is associated with both presence of aortic valve calcium and incident aortic stenosis. We also found no evidence of association between genetically elevated HDL-C or TG with presence of valve calcium or aortic stenosis, indicating that a mechanism specifically attributable to LDL-C is likely responsible for the association with aortic valve disease. Our findings link a genetically mediated increase in plasma LDL-C with early subclinical valve disease, as measured by aortic valve calcium,20,21 and incident clinical aortic stenosis, providing supportive evidence for a causal role of LDL-C in the development of aortic stenosis. Our results are in keeping with prior observations of a higher risk of aortic stenosis in patients with familial hypercholesterolemia,22 due to marked elevations in LDL-C, caused by rare mutations in the LDL receptor gene (and other related genes), and extend these findings to individuals with common variants that modestly increase LDL-C in the general population. These data suggest that, in addition to the established risks for myocardial infarction and other vascular diseases, increases in LDL-C are also associated with increased risk for aortic stenosis.

Epidemiological studies have previously demonstrated that higher LDL-C is a risk factor for aortic valve calcium and aortic stenosis.3,23,24 Although observational studies have also reported a reduction in the incidence and progression of aortic stenosis in statin users,25 this was not observed in subsequent large randomized trials.57 The disappointing results of these trials have cast doubt on the LDL hypothesis in aortic valve disease and dampened enthusiasm for LDL-C lowering in the prevention of valve disease. So how do we reconcile our Mendelian randomization data supporting an association between LDL-C and the development of aortic valve disease with the failure of LDL-C lowering in randomized trials of aortic stenosis?

All trials of LDL-C lowering for aortic stenosis enrolled older participants with established valve disease (in most cases, moderate or severe aortic stenosis),57 in which LDL-C may no longer be an important mediator of progressive disease. Once valve calcification and remodeling are well established, factors other than LDL-C, including hemodynamic forces and other procalcifying mediators, may become more important for progression. We have previously shown that higher total cholesterol in early adulthood, prior to any valve disease, was strongly associated with presence of valve calcification 25 years later.23 Similarly, in MESA, the association between LDL-C and aortic valve calcium was strongest among younger, rather than older, participants.24 It also remains to be seen whether part of the failure of prior randomized clinical trials may be related to the use of statins for lipid lowering, which may promote calcification in advanced valve disease.26

Lipoprotein deposition is a seminal process in both valve disease and vascular atherosclerosis. In early valve lesions, apolipoprotein B (the carrier molecule of LDL-C) has been demonstrated to co-localize to regions of early leaflet remodeling and calcification.27 Although LDL-C may play a common role in both vascular and valvular disease, LDL-C may play a more important role in the early calcification and mineralization phase that is strongly characteristic of early valve lesions, as opposed to vascular lesions.28 Low-density lipoprotein cholesterol may promote calcification in these early lesions, via the formation of cholesterol microcrystals that act as nidi for initial calcification,29 and also via oxidized LDL, a potent proinflammatory and pro-oxidant mediator that is known to strongly induce an osteogenic phenotype in valvular cells.30 The importance of LDL-C in the early stages of aortic valve disease is also well supported by transgenic animal data demonstrating that the genetic induction of hypercholesterolemia leads to an early procalcific phenotype at the aortic valve that can be arrested, but not reversed, with reversion to normal cholesterol levels.31 Based on the known biology of valve calcification and the failure of prior randomized trials targeting advanced disease, our results suggest that early lipid lowering, prior to the development of even mild forms of aortic stenosis, may be required to prevent aortic valve disease; this hypothesis remains to be tested in a randomized trial.

Our study has limitations that deserve mention. First, the use of summary-level data from the CHARGE consortium precluded adjustment for additional covariates. However, large Mendelian randomization studies are unlikely to be confounded because of the inherent randomization in these analyses. Indeed, adjusting for several potential confounders in the MDCS did not materially change our findings. Second, although we performed several sensitivity analyses to exclude potential pleiotropic effects of the included SNPs, a non–LDL-C–related mechanism for the observed associations cannot be entirely excluded: an important caveat of all Mendelian randomization studies. In addition, our estimates from instrumental variable analysis may be biased due to time-varying LDL-C and other considerations; therefore, the magnitude of these estimates must be interpreted cautiously.32 Third, we used diagnostic codes to identify incident aortic stenosis in the MDCS. However, these codes have been validated and shown to have a predictive value greater than 90% when compared with echocardiography.4 Fourth, the LDL-C GRS in the CHARGE and MDCS analyses differed because of the lack of GWAS data and fewer SNPs available in the MDCS; however, results in CHARGE were not materially different when we used the same LDL-C GRS used in the MDCS. Finally, our study was performed only in individuals of European descent; whether our results apply to other ethnicities will require further study.

Conclusions

Genetic predisposition to elevated LDL-C was associated with presence of aortic valve calcium and incidence of aortic stenosis, providing evidence supportive of a causal association between LDL-C and aortic valve disease. Whether earlier intervention to reduce LDL-C could prevent aortic valve disease merits further investigation.

Back to top
Article Information

Corresponding Author: George Thanassoulis, MD, Preventive and Genomic Cardiology, McGill University Health Center and Research Institute, 687 Pine Ave W, H4.55, Montreal QC H3A 1A1, Canada (george.thanassoulis@mcgill.ca).

Published Online: October 26, 2014. doi:10.1001/jama.2014.13959.

Author Contributions: Drs J. G. Smith and Thanassoulis had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: J. G. Smith, Thanassoulis.

Acquisition, analysis, or interpretation of data: J. G. Smith, Luk, Schultz, Engert, Do, Hindy, Rukh, Dufresne, Almgren, Owens, Harris, Peloso, Kerr, Wong, A. V. Smith, Budoff, Rotter, Cupples, Rich, Kathiresan, Orho-Melander, Gudnason, Post, Thanassoulis.

Drafting of the manuscript: J. G. Smith, Luk, Thanassoulis.

Critical revision of the manuscript for important intellectual content: J. G. Smith, Schultz, Engert, Do, Hindy, Rukh, Dufresne, Almgren, Owens, Harris, Peloso, Kerr, Wong, A. V. Smith, Budoff, Rotter, Cupples, Rich, Kathiresan, Orho-Melander, Gudnason, O’Donnell, Post, Thanassoulis.

Statistical analysis: J. G. Smith, Luk, Schultz, Engert, Do, Dufresne, Almgren, Peloso, Kerr, A. V. Smith, Cupples, Kathiresan, Thanassoulis.

Obtained funding: J. G. Smith, Harris, Budoff, Rotter, Orho-Melander, Gudnason, O’Donnell, Thanassoulis.

Administrative, technical, or material support: J. G. Smith, Luk, Owens, Harris, Rotter, Rich, Kathiresan, Orho-Melander, Gudnason, Thanassoulis.

Study supervision: J. G. Smith, Thanassoulis.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Cupples reported having received funding from Boston Heart Diagnostics. Dr Thanassoulis reported having received speaker fees from Servier Canada. No other disclosures were reported.

Funding/Support: The Malmö Diet and Cancer study was supported by the Swedish Cancer Society, the Swedish Medical Research Council, the Swedish Dairy Association, the Albert Påhlsson and Gunnar Nilsson Foundations, and the Malmö city council. Dr J. G. Smith was supported by the Swedish Heart-Lung Foundation, governmental funding of clinical research within the Swedish National Health Service, the Crafoord Foundation, Skåne University Hospital, and the Thorsten Westerström Foundation. The Age, Gene/Environment Susceptibility–Reykjavik Study was supported by a grant from the National Institute on Aging (N01-AG-12100) and by the National Eye Institute, the National Institute on Deafness and Other Communication Disorders, the National Heart, Lung, and Blood Institute (NHLBI), the National Institute on Aging Intramural Research Program, Hjartavernd (the Icelandic Heart Association), and the Althingi (Icelandic parliament). The MESA, MESA CARe, and the MESA SHARe project were supported by contracts N01-HC-95159 through 95169, N01-HC-65226, and N02-HL-64278 and grants from the NHLBI (UL1 RR024156, R01-HL-071051, R01-HL-071205, R01-HL-071250, R01-HL-071251, R01-HL-071252, R01-HL-071258, R01-HL-071259). Computed tomographic measures in MESA were supported by a grant to Dr Budoff (R01 HL071739). The Framingham Heart Study was supported by contracts from the NHLBI (N01-HC-25195, N02-HL-64278) with Affymetrix for genotyping services. Dr Kathiresan was supported by the American College of Cardiology Foundation/Merck Adult Cardiology Research Fellowship Award and the GlaxoSmithKline Research and Education Foundation for Cardiovascular Disease Young Investigator Award. Dr Thanassoulis was supported by a Clinician-Scientist Salary Award from the Fonds de la Recherche du Québec–Santé and operating grants from the RI-MUHC/Merck Research Competition and the Canadian Institutes of Health Research (MOP-119380 and MOP-126033).

Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Go  AS, Mozaffarian  D, Roger  VL,  et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee.  Heart disease and stroke statistics: 2013 update: a report from the American Heart Association. Circulation. 2013;127(1):e6-e245.
PubMedArticle
2.
Iung  B, Baron  G, Butchart  EG,  et al.  A prospective survey of patients with valvular heart disease in Europe: the Euro Heart Survey on Valvular Heart Disease. Eur Heart J. 2003;24(13):1231-1243.
PubMedArticle
3.
Stewart  BF, Siscovick  D, Lind  BK,  et al.  Clinical factors associated with calcific aortic valve disease: Cardiovascular Health Study. J Am Coll Cardiol. 1997;29(3):630-634.
PubMedArticle
4.
Thanassoulis  G, Campbell  CY, Owens  DS,  et al; CHARGE Extracoronary Calcium Working Group.  Genetic associations with valvular calcification and aortic stenosis. N Engl J Med. 2013;368(6):503-512.
PubMedArticle
5.
Cowell  SJ, Newby  DE, Prescott  RJ,  et al; Scottish Aortic Stenosis and Lipid Lowering Trial, Impact on Regression (SALTIRE) Investigators.  A randomized trial of intensive lipid-lowering therapy in calcific aortic stenosis. N Engl J Med. 2005;352(23):2389-2397.
PubMedArticle
6.
Rossebø  AB, Pedersen  TR, Boman  K,  et al; SEAS Investigators.  Intensive lipid lowering with simvastatin and ezetimibe in aortic stenosis. N Engl J Med. 2008;359(13):1343-1356.
PubMedArticle
7.
Chan  KL, Teo  K, Dumesnil  JG, Ni  A, Tam  J; ASTRONOMER Investigators.  Effect of lipid lowering with rosuvastatin on progression of aortic stenosis: results of the aortic stenosis progression observation: measuring effects of rosuvastatin (ASTRONOMER) trial. Circulation. 2010;121(2):306-314.
PubMedArticle
8.
Kamstrup  PR, Tybjaerg-Hansen  A, Steffensen  R, Nordestgaard  BG.  Genetically elevated lipoprotein(a) and increased risk of myocardial infarction. JAMA. 2009;301(22):2331-2339.
PubMedArticle
9.
Tikkanen  E, Tuovinen  T, Widén  E,  et al.  Association of known loci with lipid levels among children and prediction of dyslipidemia in adults. Circ Cardiovasc Genet. 2011;4(6):673-680.
PubMedArticle
10.
Do  R, Willer  CJ, Schmidt  EM,  et al.  Common variants associated with plasma triglycerides and risk for coronary artery disease. Nat Genet. 2013;45(11):1345-1352.
PubMedArticle
11.
Psaty  BM, O’Donnell  CJ, Gudnason  V,  et al; CHARGE Consortium.  Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium: design of prospective meta-analyses of genome-wide association studies from 5 cohorts. Circ Cardiovasc Genet. 2009;2(1):73-80.
PubMedArticle
12.
Smith  JG, Platonov  PG, Hedblad  B, Engström  G, Melander  O.  Atrial fibrillation in the Malmö Diet and Cancer study: a study of occurrence, risk factors and diagnostic validity. Eur J Epidemiol. 2010;25(2):95-102.
PubMedArticle
13.
Teslovich  TM, Musunuru  K, Smith  AV,  et al.  Biological, clinical and population relevance of 95 loci for blood lipids. Nature. 2010;466(7307):707-713.
PubMedArticle
14.
Willer  CJ, Schmidt  EM, Sengupta  S,  et al; Global Lipids Genetics Consortium.  Discovery and refinement of loci associated with lipid levels. Nat Genet. 2013;45(11):1274-1283.
PubMedArticle
15.
PLINK: whole genome association analysis toolset. http://pngu.mgh.harvard.edu/~purcell/plink/. Accessed October 6, 2014.
16.
rproject.org. http://cran.rproject.org/web/packages/gtx/index.html. Accessed October 6, 2014.
17.
Ehret  GB, Munroe  PB, Rice  KM,  et al; International Consortium for Blood Pressure Genome-Wide Association Studies; CARDIoGRAM Consortium; CKDGen Consortium; KidneyGen Consortium; EchoGen Consortium; CHARGE-HF Consortium.  Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature. 2011;478(7367):103-109.
PubMedArticle
18.
Burgess  S, Butterworth  A, Thompson  SG.  Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658-665.
PubMedArticle
19.
Pierce  BL, Burgess  S.  Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators. Am J Epidemiol. 2013;178(7):1177-1184.
PubMedArticle
20.
Cosmi  JE, Kort  S, Tunick  PA,  et al.  The risk of the development of aortic stenosis in patients with “benign” aortic valve thickening. Arch Intern Med. 2002;162(20):2345-2347.
PubMedArticle
21.
Coffey  S, Cox  B, Williams  MJA.  The prevalence, incidence, progression, and risks of aortic valve sclerosis: a systematic review and meta-analysis. J Am Coll Cardiol. 2014;63(25 pt A):2852-2861.
PubMedArticle
22.
Rajamannan  NM, Edwards  WD, Spelsberg  TC.  Hypercholesterolemic aortic-valve disease. N Engl J Med. 2003;349(7):717-718.
PubMedArticle
23.
Thanassoulis  G, Massaro  JM, Cury  R,  et al.  Associations of long-term and early adult atherosclerosis risk factors with aortic and mitral valve calcium. J Am Coll Cardiol. 2010;55(22):2491-2498.
PubMedArticle
24.
Owens  DS, Katz  R, Johnson  E,  et al.  Interaction of age with lipoproteins as predictors of aortic valve calcification in the Multi-Ethnic Study of Atherosclerosis. Arch Intern Med. 2008;168(11):1200-1207.
PubMedArticle
25.
Rosenhek  R, Rader  F, Loho  N,  et al.  Statins but not angiotensin-converting enzyme inhibitors delay progression of aortic stenosis. Circulation. 2004;110(10):1291-1295.
PubMedArticle
26.
Mundy  G, Garrett  R, Harris  S,  et al.  Stimulation of bone formation in vitro and in rodents by statins. Science. 1999;286(5446):1946-1949.
PubMedArticle
27.
O’Brien  KD, Reichenbach  DD, Marcovina  SM, Kuusisto  J, Alpers  CE, Otto  CM.  Apolipoproteins B, (a), and E accumulate in the morphologically early lesion of ‘degenerative’ valvular aortic stenosis. Arterioscler Thromb Vasc Biol. 1996;16(4):523-532.
PubMedArticle
28.
Otto  CM, Kuusisto  J, Reichenbach  DD, Gown  AM, O’Brien  KD.  Characterization of the early lesion of ‘degenerative’ valvular aortic stenosis: histological and immunohistochemical studies. Circulation. 1994;90(2):844-853.
PubMedArticle
29.
Hirsch  D, Azoury  R, Sarig  S, Kruth  HS.  Colocalization of cholesterol and hydroxyapatite in human atherosclerotic lesions. Calcif Tissue Int. 1993;52(2):94-98.
PubMedArticle
30.
Parhami  F, Morrow  AD, Balucan  J,  et al.  Lipid oxidation products have opposite effects on calcifying vascular cell and bone cell differentiation: a possible explanation for the paradox of arterial calcification in osteoporotic patients. Arterioscler Thromb Vasc Biol. 1997;17(4):680-687.
PubMedArticle
31.
Miller  JD, Weiss  RM, Serrano  KM,  et al.  Lowering plasma cholesterol levels halts progression of aortic valve disease in mice. Circulation. 2009;119(20):2693-2701.
PubMedArticle
32.
VanderWeele  TJ, Tchetgen Tchetgen  EJ, Cornelis  M, Kraft  P.  Methodological challenges in Mendelian randomization. Epidemiology. 2014;25(3):427-435.
PubMedArticle
×