[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.
Pleiotropy and the Validity of Estimates Derived From Mendelian Randomization
Pleiotropy and the Validity of Estimates Derived From Mendelian Randomization

Single-nucleotide polymorphisms (SNPs) are used in a genetic instrument for an exposure to assess the association with risk of disease. For each exposure there is a true association, which we try to approximate from mendelian randomization (MR). For the purposes of simplicity, conventional MR is compared with MR-Egger. Vertical pleiotropy explains where the genetic instrument associates with biomarkers (other than the exposure) that are on the causal pathway from exposure through disease. Horizontal pleiotropy is where the genetic instrument associates with additional traits not on the causal pathway of the exposure of interest. When horizontal pleiotropy is balanced, there should be no bias in the effect derived from MR. In this scenario, the estimate obtained from conventional MR is similar to that from MR-Egger. When horizontal pleiotropy is unbalanced (also termed directional pleiotropy), the pleiotropy systematically biases the estimate (which can be exaggerated or diminished) in a naive analysis using conventional MR. In the example given in this figure, the unbalanced pleiotropy exaggerates the magnitude of the association. Conventional MR will derive a biased estimate, whereas MR-Egger, correcting for unbalanced pleiotropy, should yield a valid estimate. An example of unbalanced horizontal pleiotropy is the association between high-density lipoprotein cholesterol (HDL-C) and risk of coronary artery disease (CAD); the association derived from conventional MR is different from that of MR-Egger, with the latter indicating that, once unbalanced pleiotropy is accounted for, there is no effect of HDL-C on risk of CAD.

Figure 2.
Pipeline for Derivation of the Data Set Used for Mendelian Randomization Analyses of Lipid Subtypes With Risk of Coronary Artery Disease (CAD) and Diabetes
Pipeline for Derivation of the Data Set Used for Mendelian Randomization Analyses of Lipid Subtypes With Risk of Coronary Artery Disease (CAD) and Diabetes

CARDIoGRAM indicates Coronary Artery Disease Genome-wide Replication and Meta-analysis; C4D, Coronary Artery Disease Genetics; DIAGRAM, Diabetes Genetics Replication and Meta-analysis; GLGC, Global Lipids Genetics Consortium; HDL-C, high-density lipoprotein cholesterol; LD, linkage disequilibrium; LDL-C, low-density lipoprotein cholesterol; SNP, single-nucleotide polymorphism; and TGs, triglycerides.

Figure 3.
Associations of Routinely Measured Lipids With Risk of Coronary Artery Disease (CAD) and Diabetes From Mendelian Randomization (MR) Analyses
Associations of Routinely Measured Lipids With Risk of Coronary Artery Disease (CAD) and Diabetes From Mendelian Randomization (MR) Analyses

A description of the 3 mendelian randomization models is provided in the Methods section. Estimates for conventional MR were derived from 2-sample MR that forces the slope through the origin, thereby not accounting for pleiotropy. Multivariate MR statistically adjusts for other lipid traits, and MR-Egger adjusts for unbalanced pleiotropy. R2 refers to the proportion of variance of lipid traits explained by the genetic instrument; 95% CIs are Bonferroni-adjusted. To convert high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) to millimoles per liter, multiply by 0.0259; to convert triglycerides (TGs) to millimoles per liter, multiply by 0.0113. OR indicates odds ratio; SNPs, single-nucleotide polymorphisms.

Figure 4.
Crosshair Plot of a 1-SD Increase in Lipids and Risk of Coronary Artery Disease (CAD) and Diabetes
Crosshair Plot of a 1-SD Increase in Lipids and Risk of Coronary Artery Disease (CAD) and Diabetes

All estimates derived from mendelian randomization-Egger. Error bars represent 95% CIs that are Bonferroni-adjusted. HDL-C indicates high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; OR, odds ratio; and TGs, triglycerides.

Table.  
Details of the Consortia
Details of the Consortia
1.
Stein  EA, Mellis  S, Yancopoulos  GD,  et al.  Effect of a monoclonal antibody to PCSK9 on LDL cholesterol.  N Engl J Med. 2012;366(12):1108-1118.PubMedGoogle ScholarCrossref
2.
Akdim  F, Stroes  ES, Sijbrands  EJ,  et al.  Efficacy and safety of mipomersen, an antisense inhibitor of apolipoprotein B, in hypercholesterolemic subjects receiving stable statin therapy.  J Am Coll Cardiol. 2010;55(15):1611-1618.PubMedGoogle ScholarCrossref
3.
Cannon  CP, Blazing  MA, Giugliano  RP,  et al; IMPROVE-IT Investigators.  Ezetimibe added to statin therapy after acute coronary syndromes.  N Engl J Med. 2015;372(25):2387-2397.PubMedGoogle ScholarCrossref
4.
Baigent  C, Blackwell  L, Emberson  J,  et al; Cholesterol Treatment Trialists’ (CTT) Collaboration.  Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials.  Lancet. 2010;376(9753):1670-1681.PubMedGoogle ScholarCrossref
5.
Mihaylova  B, Emberson  J, Blackwell  L,  et al; Cholesterol Treatment Trialists’ (CTT) Collaborators.  The effects of lowering LDL cholesterol with statin therapy in people at low risk of vascular disease: meta-analysis of individual data from 27 randomised trials.  Lancet. 2012;380(9841):581-590.PubMedGoogle ScholarCrossref
6.
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.PubMedGoogle ScholarCrossref
7.
Holmes  MV, Asselbergs  FW, Palmer  TM,  et al.  Mendelian randomization of blood lipids for coronary heart disease.  Eur Heart J. 2015;36(9):539-550.PubMedGoogle ScholarCrossref
8.
Voight  BF, Peloso  GM, Orho-Melander  M,  et al.  Plasma HDL cholesterol and risk of myocardial infarction: a mendelian randomisation study.  Lancet. 2012;380(9841):572-580.PubMedGoogle ScholarCrossref
9.
Schwartz  GG, Olsson  AG, Abt  M,  et al; dal-OUTCOMES Investigators.  Effects of dalcetrapib in patients with a recent acute coronary syndrome.  N Engl J Med. 2012;367(22):2089-2099.PubMedGoogle ScholarCrossref
10.
Preiss  D, Seshasai  SR, Welsh  P,  et al.  Risk of incident diabetes with intensive-dose compared with moderate-dose statin therapy: a meta-analysis.  JAMA. 2011;305(24):2556-2564.PubMedGoogle ScholarCrossref
11.
Swerdlow  DI, Preiss  D, Kuchenbaecker  KB,  et al.  HMG-coenzyme A reductase inhibition, type 2 diabetes, and bodyweight: evidence from genetic analysis and randomised trials.  Lancet. 2015;385(9965):351-361.PubMedGoogle ScholarCrossref
12.
Ridker  PM, Pradhan  A, MacFadyen  JG, Libby  P, Glynn  RJ.  Cardiovascular benefits and diabetes risks of statin therapy in primary prevention: an analysis from the JUPITER trial.  Lancet. 2012;380(9841):565-571.PubMedGoogle ScholarCrossref
13.
Fall  T, Xie  W, Poon  W,  et al; GENESIS Consortium.  Using genetic variants to assess the relationship between circulating lipids and type 2 diabetes.  Diabetes. 2015;64(7):2676-2684.PubMedGoogle ScholarCrossref
14.
Haase  CL, Tybjaerg-Hansen  A, Nordestgaard  BG, Frikke-Schmidt  R.  HDL cholesterol and risk of type 2 diabetes: a mendelian randomization study.  Diabetes. 2015;64(9):3328-3333.PubMedGoogle ScholarCrossref
15.
Lawlor  DA, Harbord  RM, Sterne  JA, Timpson  N, Davey Smith  G.  Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.  Stat Med. 2008;27(8):1133-1163.PubMedGoogle ScholarCrossref
16.
Bowden  J, Davey Smith  G, Burgess  S.  Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.  Int J Epidemiol. 2015;44(2):512-525.PubMedGoogle ScholarCrossref
17.
Burgess  S, Dudbridge  F, Thompson  SG.  Re: Multivariable mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects.  Am J Epidemiol. 2015;181(4):290-291.PubMedGoogle ScholarCrossref
18.
Day  FR, Loh  PR, Scott  RA, Ong  KK, Perry  JR.  A robust example of collider bias in a genetic association study.  Am J Hum Genet. 2016;98(2):392-393.PubMedGoogle ScholarCrossref
19.
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.PubMedGoogle ScholarCrossref
20.
Morris  AP, Voight  BF, Teslovich  TM,  et al; Wellcome Trust Case Control Consortium; Meta-analyses of Glucose and Insulin-related traits Consortium (MAGIC) Investigators; Genetic Investigation of ANthropometric Traits (GIANT) Consortium; Asian Genetic Epidemiology Network–Type 2 Diabetes (AGEN-T2D) Consortium; South Asian Type 2 Diabetes (SAT2D) Consortium; DIAbetes Genetics Replication and Meta-analysis (DIAGRAM) Consortium.  Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes.  Nat Genet. 2012;44(9):981-990.PubMedGoogle ScholarCrossref
21.
Deloukas  P, Kanoni  S, Willenborg  C,  et al; CARDIoGRAMplusC4D Consortium; DIAGRAM Consortium; CARDIOGENICS Consortium; MuTHER Consortium; Wellcome Trust Case Control Consortium.  Large-scale association analysis identifies new risk loci for coronary artery disease.  Nat Genet. 2013;45(1):25-33.PubMedGoogle ScholarCrossref
22.
Johnson  T. Efficient calculation for multi-SNP genetic risk scores: presented at the American Society of Human Genetics Annual Meeting, San Francisco, November 6-10, 2012. http://cran.r-project.org/web/packages/gtx/vignettes/ashg2012.pdf. Accessed November 1, 2015.
23.
Burgess  S, Thompson  SG.  Multivariable mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects.  Am J Epidemiol. 2015;181(4):251-260.PubMedGoogle ScholarCrossref
24.
Brion  MJ, Shakhbazov  K, Visscher  PM.  Calculating statistical power in mendelian randomization studies.  Int J Epidemiol. 2013;42(5):1497-1501.PubMedGoogle ScholarCrossref
25.
Bonnefond  A, Yengo  L, Le May  C,  et al; DESIR study group.  The loss-of-function PCSK9 p.R46L genetic variant does not alter glucose homeostasis.  Diabetologia. 2015;58(9):2051-2055.PubMedGoogle ScholarCrossref
26.
Tragante  V, Asselbergs  FW, Swerdlow  DI,  et al.  Harnessing publicly available genetic data to prioritize lipid modifying therapeutic targets for prevention of coronary heart disease based on dysglycemic risk.  Hum Genet. 2016;135(5):453-467.PubMedGoogle ScholarCrossref
27.
Tirosh  A, Shai  I, Bitzur  R,  et al.  Changes in triglyceride levels over time and risk of type 2 diabetes in young men.  Diabetes Care. 2008;31(10):2032-2037.PubMedGoogle ScholarCrossref
28.
Ginsberg  HN, Zhang  YL, Hernandez-Ono  A.  Regulation of plasma triglycerides in insulin resistance and diabetes.  Arch Med Res. 2005;36(3):232-240.PubMedGoogle ScholarCrossref
29.
Klimentidis  YC, Wineinger  NE, Vazquez  AI, de Los Campos  G.  Multiple metabolic genetic risk scores and type 2 diabetes risk in three racial/ethnic groups.  J Clin Endocrinol Metab. 2014;99(9):E1814-E1818.PubMedGoogle ScholarCrossref
30.
Klimentidis  YC, Chougule  A, Arora  A, Frazier-Wood  AC, Hsu  CH.  Triglyceride-increasing alleles associated with protection against type-2 diabetes.  PLoS Genet. 2015;11(5):e1005204.PubMedGoogle ScholarCrossref
31.
Eli Lilly and Co. Lilly to discontinue development of evacetrapib for high-risk atherosclerotic cardiovascular disease. https://investor.lilly.com/releasedetail.cfm?ReleaseID=936130. Published October 12, 2015. Accessed November 1, 2015.
32.
Preiss  D, Sattar  N.  Lipids, lipid modifying agents and cardiovascular risk: a review of the evidence.  Clin Endocrinol (Oxf). 2009;70(6):815-828.PubMedGoogle ScholarCrossref
33.
Lee  M, Saver  JL, Towfighi  A, Chow  J, Ovbiagele  B.  Efficacy of fibrates for cardiovascular risk reduction in persons with atherogenic dyslipidemia: a meta-analysis.  Atherosclerosis. 2011;217(2):492-498.PubMedGoogle ScholarCrossref
34.
Swerdlow  DI, Kuchenbaecker  KB, Shah  S,  et al.  Selecting instruments for mendelian randomization in the wake of genome-wide association studies [published online June 24, 2016].  Int J Epidemiol.PubMedGoogle Scholar
35.
Soininen  P, Kangas  AJ, Würtz  P, Suna  T, Ala-Korpela  M.  Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics.  Circ Cardiovasc Genet. 2015;8(1):192-206.PubMedGoogle ScholarCrossref
Original Investigation
September 2016

Association of Lipid Fractions With Risks for Coronary Artery Disease and Diabetes

Author Affiliations
  • 1University College London Genetics Institute, University College London, London, England
  • 2Farr Institute at University College London, London, England
  • 3Department of Medicine, Imperial College London, London, England
  • 4Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Oxford, England
  • 5The Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, England
  • 6Center for Applied Genomics, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
  • 7Department of Surgery, University of Pennsylvania, Philadelphia, Pennsylvania
  • 8Division of Heart and Lungs, Department of Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands
  • 9Durrer Center for Cardiovascular Research, Netherlands Heart Institute, Utrecht, Netherlands
  • 10Institute of Cardiovascular Science, University College London, London, England
  • 11Institute of Cardiovascular and Medical Science, University of Glasgow, Glasgow, Scotland
  • 12Cardiovascular Genetics, British Heart Foundation Laboratories, Institute Cardiovascular Science, University College London, London, United Kingdom
 

Copyright 2016 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.

JAMA Cardiol. 2016;1(6):692-699. doi:10.1001/jamacardio.2016.1884
Abstract

Importance  Low-density lipoprotein cholesterol (LDL-C) is causally related to coronary artery disease (CAD), but the relevance of high-density lipoprotein cholesterol (HDL-C) and triglycerides (TGs) is uncertain. Lowering of LDL-C levels by statin therapy modestly increases the risk of type 2 diabetes, but it is unknown whether this effect is specific to statins.

Objective  To investigate the associations of 3 routinely measured lipid fractions with CAD and diabetes through mendelian randomization (MR) using conventional MR and making use of newer approaches, such as multivariate MR and MR-Egger, that address the pleiotropy of genetic instruments where relevant.

Design, Setting, and Participants  Published data from genome-wide association studies were used to construct genetic instruments and then applied to investigate associations between lipid fractions and the risk of CAD and diabetes using MR approaches that took into account pleiotropy of genetic instruments. The study was conducted from March 12 to December 31, 2015.

Main Outcomes and Measures  Coronary artery disease and diabetes.

Results  Genetic instruments composed of 130 single-nucleotide polymorphisms (SNPs) were used for LDL-C (explaining 7.9% of its variance), 140 SNPs for HDL-C (6.6% of variance), and 140 SNPs for TGs (5.9% of variance). A 1-SD genetically instrumented elevation in LDL-C levels (equivalent to 38 mg/dL) and TG levels (equivalent to 89 mg/dL) was associated with higher CAD risk; odds ratios (ORs) were 1.68 (95% CI, 1.51-1.87) for LDL-C and 1.28 (95% CI, 1.13-1.45) for TGs. The corresponding OR for HDL-C (equivalent to a 16-mg/dL increase) was 0.95 (95% CI, 0.85-1.06). All 3 lipid traits were associated with a lower risk of type 2 diabetes. The ORs were 0.79 (95% CI, 0.71-0.88) for LDL-C and 0.83 (95% CI, 0.76-0.90) for HDL-C per 1-SD elevation. For TG, the MR estimates for diabetes were inconsistent, with MR-Egger giving an OR of 0.83 (95%CI, 0.72-0.95) per 1-SD elevation.

Conclusions and Relevance  Routinely measured lipid fractions exhibit contrasting associations with the risk of CAD and diabetes. Increased LDL-C, HDL-C, and possibly TG levels are associated with a lower risk of diabetes. This information will be relevant to the design of clinical trials of lipid-modifying agents, which should carefully monitor participants for dysglycemia and the incidence of diabetes.

Introduction

Understanding the interplay between circulating lipids and the risk of type 2 diabetes and coronary artery disease (CAD) is of emerging public health importance and has implications for drug development for cardiovascular disease prevention.1,2 For example, a causal influence of low-density lipoprotein cholesterol (LDL-C) on CAD is widely accepted3-5 and the proposed causal role of triglycerides (TGs) in CAD is gaining acceptance.6,7 In contrast, the role of high-density lipoprotein cholesterol (HDL-C) in CAD remains in doubt.7-9

However, evidence has emerged that LDL-C reduction with statin therapy results in a modest increase in risk of type 2 diabetes10,11 (outweighed by the benefit of statins in protection from CAD).12 Whether this diabetogenic effect is a general consequence of LDL-C lowering or whether it is specific to inhibition of hydroxymethylglutaryl coenzyme A reductase remains unclear.13 Moreover, the role of TGs and HDL-C in the development of diabetes remains uncertain.14

Residual confounding and reverse causality can limit causal inference from observational studies. When a genetic instrument can be used as an instrument for an exposure, mendelian randomization (MR) generates unbiased, unconfounded effect estimates that are sometimes taken as evidence of a causal role. This interpretation is because genotype is not modifiable by disease and the random allocation of alleles at gametogenesis helps to avoid bias in studies arising due to reverse causality and confounding, respectively.

A critical assumption of the MR paradigm is that the genetic instrument influences disease risk exclusively through the exposure of interest. However, genetic variants used to proxy the exposure of interest can also associate with other traits, a phenomenon termed pleiotropy. When pleiotropy arises as a downstream consequence of genetic perturbation of the biomarker of interest, it is referred to as vertical pleiotropy and the MR assumption is preserved.15 However, when pleiotropy arises because of the association of genetic variants with additional phenotypes in alternative disease pathways (termed horizontal pleiotropy), the assumption is compromised. When MR analysis is based on multiple single-nucleotide polymorphisms (SNPs) drawn from different regions of the genome selected systematically for their association with the biomarker of interest, additional nonsystematic effects on any other biomarkers might be balanced and the MR effect estimate could still be valid. However, if horizontal pleiotropy is unbalanced, as might occur when the set of biomarkers concerned comes from closely connected pathways, MR estimates may become systematically biased (termed unbalanced or directional pleiotropy16), resulting in invalid effect estimates (Figure 1).

Recent methodologic advances in MR analysis, including multivariate MR17 and MR-Egger,16 provide new approaches for dealing with pleiotropic genetic instruments. In multivariate MR, adjustment is made for genetic associations with measured traits but may not fully account for unbalanced pleiotropy.18 In contrast, MR-Egger can detect and correct for unbalanced pleiotropy of the genetic instrument, even when unbalanced pleiotropy is mediated through unmeasured or unknown traits; a disadvantage being a reduction in power.

We used summary data from multiple major cardiometabolic genome-wide association studies to investigate the underlying associations between lipid levels, diabetes, and CAD using 3 MR approaches: (1) conventional MR, which does not account for pleiotropy; (2) multivariate MR, which adjusts for traits that may mediate unbalanced pleiotropy; and (3) MR-Egger, which more fully accounts for unbalanced pleiotropy.

Box Section Ref ID

Key Points

  • Question Are levels of routinely measured lipids associated with the risk of cardiometabolic disease?

  • Findings In this mendelian randomization analysis, a lifelong higher low-density lipoprotein cholesterol (LDL-C) and triglyceride (TG) levels was found to be associated with higher risk of coronary artery disease. In contrast, higher levels of all 3 lipid traits (LDL-C, high-density lipoprotein cholesterol, and TGs) was associated with a reduced risk of diabetes.

  • Meaning Lower LDL-C and TG levels may increase risk of diabetes; clinical trials of lipid-modifying agents should carefully monitor for the incidence of diabetes.

Methods
Data Sources

We used summary-level data for lipids from the Global Lipids Genetics Consortium,19 diabetes data from the Diabetes Genetics Replication and Meta-analysis,20 and CAD data from the Coronary Artery Disease Genome-wide Replication and Meta-analysis plus Coronary Artery Disease Genetics.21 Details of the consortia are provided in the Table. All data sets were limited to individuals of European ancestry; β coefficients and SEs were obtained for the per-allele association of each SNP with all exposures and outcomes from these data sources. If SNPs were not present in a data set, we used proxies (R2 > 0.9) as indicated in Figure 2. The study was conducted from March 12 to December 31, 2015.

Because this report used published genome-wide association studies data available in the public domain, specific ethical review and/or consent from study participants was not sought (and had been obtained in the original studies).

Selection of SNPs

We used 185 lipid-associated SNPs identified by Willer et al19 to generate a series of genetic instruments for each of the exposures: LDL-C, HDL-C, and TGs. This process was conducted by first restricting to a set of SNPs in low linkage disequilibrium (pairwise R2<0.2). We then organized these SNPs by descending order of proportional variance (R2, estimated from the summary statistics using the gtx package in R [https://www.r-project.org/]) between SNPs with the corresponding lipid exposure to generate a range of instruments from 5 to 150 SNPs. The process used to determine the final tally of SNPs for inclusion in a genetic instrument for each lipid trait is described below.

Handling of SNPs

We matched SNPs across the data sources by aligning them to the same effect allele. Effect allele frequencies were checked for concordance.

MR Analyses

We used 3 MR approaches. First, we conducted conventional 2-sample instrumental variable analyses, which do not make any allowance for pleiotropy. Basing our approach on the method first proposed by Johnson,22 we incorporated the bootstrap suggested by Bowden et al16 as a way to incorporate the error in the published estimates of SNP effect on both exposure and outcome.

Second, we conducted multivariate MR analyses, which statistically adjust for pleiotropy with additional phenotypes measured in the data set.23 Multivariate MR is an extension of the conventional weighted regression in which the β values for additional phenotypes are included as covariates. In this case we used all 3 lipid traits in the model (eg, for the HDL-C instrument, we included LDL-C and TGs, thereby adjusting for them).

Third, we used MR-Egger,16 which accounts for unbalanced pleiotropy of a genetic instrument. MR-Egger is a linear regression of estimated SNP effects (for the exposure-raising allele) on exposure against the corresponding estimates of SNP on outcome weighted by the inverse variance of the SNP on outcome effect estimates. This approach differs from conventional 2-sample MR in that the regression line is not forced through the origin. Bowden et al16 showed that the MR-Egger estimate is unaffected by net pleiotropic effects of the instrument and that the presence of unbalanced pleiotropy can be inferred if the intercept term is not zero.

For all 3 approaches (conventional MR, multivariate MR, and MR-Egger), we conducted 10 000 bootstraps, and our effect estimate is the mean of the bootstraps with the CI determined empirically and set to Bonferroni-adjusted (for 6 tests) 95% (ie, 99.2%).

Quantifying the Proportion of Variance Explained by the Genetic Instruments
R Trend

The proportion of variance (R2) of the trait explained by the genetic instrument will rise with the addition of more SNPs. However, the improvement beyond the optimum number of SNPs in the instrument will occur increasingly as a result of model overspecification. We examined the ratio of R2 for the current instrument to R2 for an instrument comprising 30 more SNPs (we term this the R trend). The trend in the ratio gives an indication of the transition from useful additional information to overspecification since it becomes asymptotic when each new SNP adds the same amount of information than the last. We judged the beginning of the asymptotic phase of the line to mark the largest useful instrument obtainable from the available data. The 30-SNP window was chosen empirically because it emphasizes trend; a smaller window gave a more erratic line, obscuring the trend. Calculation and use of the R trend effectively limited the analysis to instruments comprising 155 or fewer SNPs, further restricted to 150 for presentational purposes.

Gain From Adding a SNP to the Instrument

We estimated the benefit to R2 from adding the current SNP to the previous instrument by bootstrapping the summary statistics and calculating R2 values for the instruments with n and (n − 1) SNPs. While performing 10 000 bootstraps, we noted the number of occasions when the current instrument gave higher R2 than the previous instrument. This value was summarized as a percentage; the point at which the current instrument was no better than the previous instrument was when 50% of the runs showed a benefit.

Selection of Optimal Number of SNPs

The optimum number of SNPs was chosen by consideration of the R trend and the gain from adding the current SNP when presented graphically (eFigures 1-6 in the Supplement). The optimum instrument was identified when both estimates of R2 gain were asymptotic. As discussed below, the exact point (±20 SNPs) makes little difference to the conclusions. After independent consideration, 2 investigators (J.W. and M.V.H.) reached a consensus as to the number of SNPs to include for each genetic instrument for each lipid trait.

Selection of MR Model to Derive Estimates of the Underlying Association

Once we determined the optimal number of SNPs to incorporate in each instrument, we used the following decision tree to select the MR approach to derive the estimate:

  1. if there was no evidence of unbalanced pleiotropy using the intercept derived from MR-Egger, we selected the conventional MR instrumental variable estimate as the most reliable indicator to the underlying association since it retains maximal power and makes the fewest assumptions;

  2. if there was evidence of unbalanced pleiotropy, we used the estimate from MR-Egger; and

  3. in cases of discordance between conventional MR and MR-Egger, we used multivariate MR to inform whether differences could arise from pathways shared between the 3 lipid traits.

The inSIDE Assumption

Because the underlying models assume a linear dose response, instrumental variable effect estimates must be independent of the exposure effect in MR analysis (ie, the Instrument Strength Independent of Direct Effect [inSIDE] rule).16 We tested the null hypothesis that the instrumental variable effect (derived from the ratio of outcome to exposure) estimates for the SNPs in an instrument were independent of the exposure (lipid) effect estimates for the same SNPs for both CAD and diabetes. In all scenarios, the inSIDE assumption was satisfied (eTable 1 in the Supplement).

Power

We followed the method of Brion et al24 (http://cnsgenomics.com/shiny/mRnd/). Using the mean number of individuals and estimated R2 for the instrument together with the reported proportion of cases, we adjusted the estimate of the true effect of exposure on outcome to obtain the value for which we had 80% power at a Bonferroni-adjusted α value of .05/6.

Results

The pooled data set included 188 577 individuals with measures of blood lipids, 63 158 CAD cases, and 34 840 diabetes cases. The optimal number of SNPs for each lipid trait was 130 for LDL-C (explaining 7.9% of its variance), 140 for HDL-C (6.6% of the variance), and 140 for TG (5.9% of the variance) (eFigures 1-6 in the Supplement).

Low-Density Lipoprotein Cholesterol

The genetic instrument for LDL-C showed unbalanced pleiotropy for CAD and diabetes. For CAD, the estimate derived from MR-Egger was an OR of 1.68 (95% CI, 1.51-1.87) per 1-SD (equivalent to 38 mg/dL [to convert to millimoles per liter, multiply by 0.0259]) genetically instrumented higher LDL-C. This was of greater magnitude but directionally consistent with conventional and multivariate MR estimates (Figure 3 and eFigure 1 in the Supplement). For diabetes, the OR was 0.79 (95% CI, 0.71-0.88) per 1-SD higher LDL-C from MR-Egger, which was again of greater magnitude yet directionally consistent with conventional and multivariate MR estimates (Figure 3 and eFigure 2 in the Supplement).

High-Density Lipoprotein Cholesterol

A 1-SD genetically instrumented elevation in HDL-C (equivalent to 16 mg/dL [to convert to millimoles per liter, multiply by 0.0259]) did not provide conclusive evidence of an association between HDL-C and risk of CAD. There was evidence of unbalanced pleiotropy of the HDL-C genetic instrument, and the estimate for CAD from MR-Egger was an OR of 0.95 (95% CI, 0.85-1.06). There was a stepwise weakening of the effect toward the null from conventional MR (OR, 0.80; 95% CI, 0.75-0.86) through adjusting for LDL-C and TGs in multivariate MR (OR, 0.86; 95% CI, 0.78-0.96) to the MR-Egger estimate (Figure 3 and eFigure 3 in the Supplement).

For diabetes, there was no evidence of unbalanced pleiotropy of the genetic instrument comprising 140 SNPs. The conventional MR provided an estimate consistent with estimates from both multivariate MR and MR-Egger (OR, 0.83; 95% CI, 0.76-0.90) (Figure 3 and eFigure 4 in the Supplement).

Triglycerides

The TG genetic instrument showed unbalanced pleiotropy for both CAD and diabetes. A 1-SD genetically instrumented increase in TG (equivalent to 89 mg/dL [to convert to millimoles per liter, multiply by 0.0113]) yielded an OR for CAD from MR-Egger of 1.28 (95% CI, 1.13-1.45), which was weaker than the multivariate MR estimate (OR, 1.38; 95% CI, 1.19-1.59) and roughly half the magnitude of the conventional MR estimate (OR, 1.49; 95% CI, 1.37-1.62) (Figure 3 and eFigure 5 in the Supplement).

Triglycerides were associated with reduced risk of diabetes (OR, 0.83; 95% CI, 0.72-0.95 from MR-Egger). This was dissimilar to both conventional and multivariate MR estimates (Figure 3). The scatterplot identified that the intercept of the MR-Egger slope was positive (eFigure 6 in the Supplement).

Power

There was adequate power to detect the reported estimates (eTable 2 in the Supplement). The level of power made it unlikely that associations arose from chance.

Putting the Pieces Together: Framework of Associations

Our findings demonstrate that elevations in LDL-C, TG, and HDL-C levels are associated with a reduced risk of diabetes, with the magnitude (per 1-SD increase) being greatest for LDL-C, then TG followed by HDL-C (although the 95% CI for the effect on diabetes for the 3 lipids overlap) (Figure 4). In contrast, only LDL-C and TG levels were associated with an increased risk of CAD (with the magnitude again stronger for LDL-C vs TG).

Discussion

We exploited data from multiple genome-wide association studies to conduct MR analyses exploring the associations between lipids and risk of diabetes and CAD. Our findings reveal a series of associations that will help inform potential downstream consequences of pharmacologic modification of lipid levels.

Although all 3 lipids were associated with reduced risk of diabetes, it does not necessarily follow that lowering of LDL-C or TG levels through use of drugs that target specific proteins (eg, PCSK9) will alter the risk of diabetes. Large-scale genetic and clinical investigations are needed to clarify the effects of pharmacologic lowering of LDL-C and TG levels to gauge dysglycemic associations.25,26

Our findings are complementary to those from a study by Fall et al13 that, to address pleiotropy, excluded SNPs showing strong associations with diabetes, glycemia-related traits, or potential confounders, such as adiposity. This manual pruning weakened the associations, yielding inconsistent conclusions. In our study, we applied novel approaches for SNP selection (to optimize the SNPs in each genetic instrument) and MR (using MR-Egger, obviating the need to manually prune SNPs); these approaches collectively allow us to make more robust conclusions about the role of lipids in diabetes.

To our knowledge, the protective effect of TGs in the risk of diabetes that we describe is novel, yet potentially counterintuitive. Observational studies27 report that increases in TG levels are associated with an increase in the risk of diabetes; however, insulin resistance results in perturbations in TG metabolism,28 meaning that the direction of the causal association is not clear. Although our data (suggesting that TGs may provide protection from diabetes) should be interpreted with caution, especially given that effect estimates from the 3 MR approaches were not consistent with one another, the findings are consistent with those of recent genetic studies29,30 in both Europeans and African Americans. Further investigations are needed to identify which TG pathways, if any, may lead to a reduction in the risk of diabetes.

The effects of LDL-C and TGs on the risk of CAD were robust; however, the evidence on HDL-C was far less convincing, with the estimate from MR-Egger failing to identify an effect. This finding is in keeping with prior MR studies,7,8 including that of Voight et al,8 that manually pruned pleiotropic SNPs. However, choosing only nonpleiotropic SNPs could introduce selection bias in the genetic instrument by focusing on a subset of SNPs that is not representative of any meaningful proxy of HDL-C, with the removal of potentially informative HDL-C–related pathways. Our data show that adjusting for TGs and LDL-C in multivariate MR does not fully account for the unbalanced pleiotropy of the HDL-C genetic instrument. Use of MR-Egger identifies the likely underlying association is a genetically determined higher HDL-C that does not result in a reduced risk of CAD. Although these findings are consistent with those of recent trials of therapeutics targeting HDL-C,9,31 the results do not preclude the possibility that a drug modifying HDL-C (or HDL particles) could reduce the risk of CAD or other outcomes, such as stroke.

The association of TGs with CAD recapitulates findings from prior MR and genetic studies.6,7 The MR-Egger estimate for TG levels was less than half the magnitude for an equivalent increase in LDL-C levels (OR [95% CI], 1.28 [1.13-1.45]; and 1.68 [1.51-1.87] for TGs and LDL-C, respectively, per 1-SD increment). Specific TG-lowering approaches have had, at best, modest efficacy, whereas statin trials have had consistently positive results.32,33 Our data suggest that pharmacologic lowering of TG levels should translate into CAD benefit.

The present study has several advantages. First, we used the most up-to-date data available for lipids to generate the most comprehensive genetic instruments available. Second, MR-Egger enabled inclusion of genome-wide association study–identified lipid-related SNPs in the genetic instruments regardless of the presence of unbalanced pleiotropy. Third, use of summary-level data from different sources represents an efficient study design to facilitate original investigations such as these without the cost or need for de novo phenotyping or genotyping.

The study also has some limitations. First, estimates could be sensitive to SNPs that were included in the genetic instruments. However, MR estimates were stable at the point at which we selected the genetic instrument. Second, our MR analyses pertain to biomarkers rather than specific drug targets. Mendelian randomization for validating individual targets is best achieved using SNPs in genes encoding the drug target of interest.34 Third, patients targeted for lipid modification may be at risk for other diseases, such as heart failure or atrial fibrillation; the relevance and direction of effects on these and other end points could be important but were not evaluated in this study. Fourth, we are not able to account for statin treatment in the analyses; given that we detected the protective effect of LDL-C on the risk of diabetes (the scenario that statins are most likely to confound), major bias is unlikely to arise in this setting. Finally, although our data cast further doubt on the relevance of HDL-C in development of CAD, it remains possible that individual HDL lipoproteins and/or lipid compositions could play an important role. New methods, such as hydrogen 1 nuclear magnetic resonance metabolomics,35 that quantify lipoprotein subclasses and lipid compositions are likely to facilitate future MR studies of HDL subclasses.

Conclusions

Our comprehensive MR investigations identify distinct associations between major lipid subfractions and the risk of CAD and diabetes. Increased LDL-C and TG levels increase the risk of CAD. In contrast, an increase in LDL-C and HDL-C levels is likely to provide protection from diabetes, with new evidence suggesting that TGs may also play a protective role. Although further studies are needed to examine whether specific pathways or lipid subtypes are implicated, our findings inform potential expected downstream consequences of interventions affecting lipid traits and provide cautionary evidence that therapeutics that lower LDL-C and TG levels may have dysglycemic effects.

Back to top
Article Information

Corresponding Author: Michael V. Holmes, MD, PhD, Clinical Trial Service Unit & Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Roosevelt Dr, Richard Doll Bldg, Old Rd Campus, Oxford OX3 7LF, England (michael.holmes@ndph.ox.ac.uk).

Accepted for Publication: May 13, 2016.

Published Online: August 3, 2016. doi:10.1001/jamacardio.2016.1884

Author Contributions: Dr White had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: White, Hingorani, Holmes.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: White, Holmes.

Critical revision of the manuscript for important intellectual content: Swerdlow, Preiss, Fairhurst-Hunter, Keating, Asselbergs, Sattar, Humphries, Hingorani, Holmes.

Statistical analysis: White, Holmes.

Administrative, technical, or material support: Keating.

Study supervision: Hingorani, Holmes.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Swerdlow has been a paid consultant to Pfizer on work unrelated to the present analysis. Dr Sattar has received honoraria for advisory boards or lectures for Amgen, Sanofi, Boehringer Ingelheim, Novo Nordisk, Merck, Janssen, and Astrazeneca. No other disclosures were reported.

Funding/Support: Clinical Trial Service Unit (University of Oxford) is the central coordinating center for the REVEAL (Randomized Evaluation of the Effects of Anacetrapib Through Lipid-Modification) trial of anacetrapib; REVEAL is funded through a grant to the University of Oxford by Merck Sharp & Dohme Corp but was designed and is being conducted independently of the funder. Support has been provided by a Genomic Medicine and Statistics Wellcome Trust DPhil Studentship (Ms Fairhurst-Hunter), a Dekker scholarship–Junior Staff Member 2014T001–Netherlands Heart Foundation and University College London (UCL) Hospitals National Institute for Health Research (NIHR) Biomedical Research Centre (Dr Asselbergs), British Heart Foundation (BHF) professorship and funding by grant PG08/008 and the NIHR University College London Hospitals Biomedical Research Centre (Professor Humphries), an NIHR Academic Clinical Fellowship (Dr Swerdlow), and the UCL Hospitals NIHR Biomedical Research Centre (Dr Hingorani) with work in the laboratory supported by BHF Programme and Special Project Grants.

Role of the Funder/Sponsor: The funding sources 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.

Additional Information: Dr Hingorani is an NIHR senior investigator.

References
1.
Stein  EA, Mellis  S, Yancopoulos  GD,  et al.  Effect of a monoclonal antibody to PCSK9 on LDL cholesterol.  N Engl J Med. 2012;366(12):1108-1118.PubMedGoogle ScholarCrossref
2.
Akdim  F, Stroes  ES, Sijbrands  EJ,  et al.  Efficacy and safety of mipomersen, an antisense inhibitor of apolipoprotein B, in hypercholesterolemic subjects receiving stable statin therapy.  J Am Coll Cardiol. 2010;55(15):1611-1618.PubMedGoogle ScholarCrossref
3.
Cannon  CP, Blazing  MA, Giugliano  RP,  et al; IMPROVE-IT Investigators.  Ezetimibe added to statin therapy after acute coronary syndromes.  N Engl J Med. 2015;372(25):2387-2397.PubMedGoogle ScholarCrossref
4.
Baigent  C, Blackwell  L, Emberson  J,  et al; Cholesterol Treatment Trialists’ (CTT) Collaboration.  Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials.  Lancet. 2010;376(9753):1670-1681.PubMedGoogle ScholarCrossref
5.
Mihaylova  B, Emberson  J, Blackwell  L,  et al; Cholesterol Treatment Trialists’ (CTT) Collaborators.  The effects of lowering LDL cholesterol with statin therapy in people at low risk of vascular disease: meta-analysis of individual data from 27 randomised trials.  Lancet. 2012;380(9841):581-590.PubMedGoogle ScholarCrossref
6.
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.PubMedGoogle ScholarCrossref
7.
Holmes  MV, Asselbergs  FW, Palmer  TM,  et al.  Mendelian randomization of blood lipids for coronary heart disease.  Eur Heart J. 2015;36(9):539-550.PubMedGoogle ScholarCrossref
8.
Voight  BF, Peloso  GM, Orho-Melander  M,  et al.  Plasma HDL cholesterol and risk of myocardial infarction: a mendelian randomisation study.  Lancet. 2012;380(9841):572-580.PubMedGoogle ScholarCrossref
9.
Schwartz  GG, Olsson  AG, Abt  M,  et al; dal-OUTCOMES Investigators.  Effects of dalcetrapib in patients with a recent acute coronary syndrome.  N Engl J Med. 2012;367(22):2089-2099.PubMedGoogle ScholarCrossref
10.
Preiss  D, Seshasai  SR, Welsh  P,  et al.  Risk of incident diabetes with intensive-dose compared with moderate-dose statin therapy: a meta-analysis.  JAMA. 2011;305(24):2556-2564.PubMedGoogle ScholarCrossref
11.
Swerdlow  DI, Preiss  D, Kuchenbaecker  KB,  et al.  HMG-coenzyme A reductase inhibition, type 2 diabetes, and bodyweight: evidence from genetic analysis and randomised trials.  Lancet. 2015;385(9965):351-361.PubMedGoogle ScholarCrossref
12.
Ridker  PM, Pradhan  A, MacFadyen  JG, Libby  P, Glynn  RJ.  Cardiovascular benefits and diabetes risks of statin therapy in primary prevention: an analysis from the JUPITER trial.  Lancet. 2012;380(9841):565-571.PubMedGoogle ScholarCrossref
13.
Fall  T, Xie  W, Poon  W,  et al; GENESIS Consortium.  Using genetic variants to assess the relationship between circulating lipids and type 2 diabetes.  Diabetes. 2015;64(7):2676-2684.PubMedGoogle ScholarCrossref
14.
Haase  CL, Tybjaerg-Hansen  A, Nordestgaard  BG, Frikke-Schmidt  R.  HDL cholesterol and risk of type 2 diabetes: a mendelian randomization study.  Diabetes. 2015;64(9):3328-3333.PubMedGoogle ScholarCrossref
15.
Lawlor  DA, Harbord  RM, Sterne  JA, Timpson  N, Davey Smith  G.  Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.  Stat Med. 2008;27(8):1133-1163.PubMedGoogle ScholarCrossref
16.
Bowden  J, Davey Smith  G, Burgess  S.  Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.  Int J Epidemiol. 2015;44(2):512-525.PubMedGoogle ScholarCrossref
17.
Burgess  S, Dudbridge  F, Thompson  SG.  Re: Multivariable mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects.  Am J Epidemiol. 2015;181(4):290-291.PubMedGoogle ScholarCrossref
18.
Day  FR, Loh  PR, Scott  RA, Ong  KK, Perry  JR.  A robust example of collider bias in a genetic association study.  Am J Hum Genet. 2016;98(2):392-393.PubMedGoogle ScholarCrossref
19.
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.PubMedGoogle ScholarCrossref
20.
Morris  AP, Voight  BF, Teslovich  TM,  et al; Wellcome Trust Case Control Consortium; Meta-analyses of Glucose and Insulin-related traits Consortium (MAGIC) Investigators; Genetic Investigation of ANthropometric Traits (GIANT) Consortium; Asian Genetic Epidemiology Network–Type 2 Diabetes (AGEN-T2D) Consortium; South Asian Type 2 Diabetes (SAT2D) Consortium; DIAbetes Genetics Replication and Meta-analysis (DIAGRAM) Consortium.  Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes.  Nat Genet. 2012;44(9):981-990.PubMedGoogle ScholarCrossref
21.
Deloukas  P, Kanoni  S, Willenborg  C,  et al; CARDIoGRAMplusC4D Consortium; DIAGRAM Consortium; CARDIOGENICS Consortium; MuTHER Consortium; Wellcome Trust Case Control Consortium.  Large-scale association analysis identifies new risk loci for coronary artery disease.  Nat Genet. 2013;45(1):25-33.PubMedGoogle ScholarCrossref
22.
Johnson  T. Efficient calculation for multi-SNP genetic risk scores: presented at the American Society of Human Genetics Annual Meeting, San Francisco, November 6-10, 2012. http://cran.r-project.org/web/packages/gtx/vignettes/ashg2012.pdf. Accessed November 1, 2015.
23.
Burgess  S, Thompson  SG.  Multivariable mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects.  Am J Epidemiol. 2015;181(4):251-260.PubMedGoogle ScholarCrossref
24.
Brion  MJ, Shakhbazov  K, Visscher  PM.  Calculating statistical power in mendelian randomization studies.  Int J Epidemiol. 2013;42(5):1497-1501.PubMedGoogle ScholarCrossref
25.
Bonnefond  A, Yengo  L, Le May  C,  et al; DESIR study group.  The loss-of-function PCSK9 p.R46L genetic variant does not alter glucose homeostasis.  Diabetologia. 2015;58(9):2051-2055.PubMedGoogle ScholarCrossref
26.
Tragante  V, Asselbergs  FW, Swerdlow  DI,  et al.  Harnessing publicly available genetic data to prioritize lipid modifying therapeutic targets for prevention of coronary heart disease based on dysglycemic risk.  Hum Genet. 2016;135(5):453-467.PubMedGoogle ScholarCrossref
27.
Tirosh  A, Shai  I, Bitzur  R,  et al.  Changes in triglyceride levels over time and risk of type 2 diabetes in young men.  Diabetes Care. 2008;31(10):2032-2037.PubMedGoogle ScholarCrossref
28.
Ginsberg  HN, Zhang  YL, Hernandez-Ono  A.  Regulation of plasma triglycerides in insulin resistance and diabetes.  Arch Med Res. 2005;36(3):232-240.PubMedGoogle ScholarCrossref
29.
Klimentidis  YC, Wineinger  NE, Vazquez  AI, de Los Campos  G.  Multiple metabolic genetic risk scores and type 2 diabetes risk in three racial/ethnic groups.  J Clin Endocrinol Metab. 2014;99(9):E1814-E1818.PubMedGoogle ScholarCrossref
30.
Klimentidis  YC, Chougule  A, Arora  A, Frazier-Wood  AC, Hsu  CH.  Triglyceride-increasing alleles associated with protection against type-2 diabetes.  PLoS Genet. 2015;11(5):e1005204.PubMedGoogle ScholarCrossref
31.
Eli Lilly and Co. Lilly to discontinue development of evacetrapib for high-risk atherosclerotic cardiovascular disease. https://investor.lilly.com/releasedetail.cfm?ReleaseID=936130. Published October 12, 2015. Accessed November 1, 2015.
32.
Preiss  D, Sattar  N.  Lipids, lipid modifying agents and cardiovascular risk: a review of the evidence.  Clin Endocrinol (Oxf). 2009;70(6):815-828.PubMedGoogle ScholarCrossref
33.
Lee  M, Saver  JL, Towfighi  A, Chow  J, Ovbiagele  B.  Efficacy of fibrates for cardiovascular risk reduction in persons with atherogenic dyslipidemia: a meta-analysis.  Atherosclerosis. 2011;217(2):492-498.PubMedGoogle ScholarCrossref
34.
Swerdlow  DI, Kuchenbaecker  KB, Shah  S,  et al.  Selecting instruments for mendelian randomization in the wake of genome-wide association studies [published online June 24, 2016].  Int J Epidemiol.PubMedGoogle Scholar
35.
Soininen  P, Kangas  AJ, Würtz  P, Suna  T, Ala-Korpela  M.  Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics.  Circ Cardiovasc Genet. 2015;8(1):192-206.PubMedGoogle ScholarCrossref
×