Advertisement
American Journal of Kidney Diseases

Challenges in Interpreting Multivariable Mendelian Randomization: Might “Good Cholesterol” Be Good After All?

  • Michael V. Holmes
    Correspondence
    Address for Correspondence: Michael V. Holmes, MD, PhD, Medical Research Council Population Health Research Unit at the University of Oxford, Nuffield Department of Population Health, Big Data Institute, Old Road Campus, Oxford, OX3 7LF, United Kingdom.
    Affiliations
    Medical Research Council Population Health Research Unit, University of Oxford, Oxford, United Kingdom

    Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom

    National Institute for Health Research, Oxford Biomedical Research Centre, Oxford University Hospital, Oxford, United Kingdom

    Medical Research Council Integrative Epidemiology Unit, University of Bristol, Bristol, United Kingdom
    Search for articles by this author
  • George Davey Smith
    Correspondence
    George Davey Smith, MD, PhD, DSc, Medical Research Council Integrative Epidemiology Unit, University of Bristol, Oakfield House, Oakfield Grove, Clifton BS8 2BN, Bristol, United Kingdom KZ.
    Affiliations
    Medical Research Council Integrative Epidemiology Unit, University of Bristol, Bristol, United Kingdom

    Population Health Sciences, University of Bristol, Bristol, United Kingdom
    Search for articles by this author
      Related Article, p. 166
      Understanding the causal basis of disease is important so that we approach disease prevention and treatment using a valid etiologic framework. Blood lipids play an important role in the shuttling of nutrients (in the form of triglycerides and fatty acids) and cholesterol from the diet to the peripheral tissues. Certain types of blood lipids (eg, low-density lipoprotein cholesterol [LDL-C] and probably triglycerides [TG]) are atherogenic and lead to higher risks for coronary heart disease (CHD).
      • Silverman M.G.
      • Ference B.A.
      • Im K.
      • et al.
      Association between lowering LDL-C and cardiovascular risk reduction among different therapeutic interventions: a systematic review and meta-analysis.
      • Emerging Risk Factors Collaboration.
      • Di Angelantonio E.
      • Sarwar N.
      • Perry P.
      • et al.
      Major lipids, apolipoproteins, and risk of vascular disease.
      • Baigent C.
      • Blackwell L.
      • et al.
      Cholesterol Treatment Trialists' Collaborators
      Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials.
      • Mihaylova B.
      • Emberson J.
      • et al.
      Cholesterol Treatment Trialists' 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.
      • Holmes M.V.
      • Asselbergs F.W.
      • Palmer T.M.
      • et al.
      Mendelian randomization of blood lipids for coronary heart disease.
      • Ference B.A.
      • Yoo W.
      • Alesh I.
      • et al.
      Effect of long-term exposure to lower low-density lipoprotein cholesterol beginning early in life on the risk of coronary heart disease: a Mendelian randomization analysis.
      The role of high-density lipoprotein cholesterol (HDL-C) to date has been more elusive.
      • Emerging Risk Factors Collaboration.
      • Di Angelantonio E.
      • Sarwar N.
      • Perry P.
      • et al.
      Major lipids, apolipoproteins, and risk of vascular disease.
      Its central purported role is the reverse transport of cholesterol, which theoretically should lead to a net reduction in atheroma in the tunica intima of the arterial vasculature (with supposedly commensurate reductions in risk for vascular disease). This idea has come under scrutiny in recent years in response to accumulating scientific evidence
      • Holmes M.V.
      • Asselbergs F.W.
      • Palmer T.M.
      • et al.
      Mendelian randomization of blood lipids for coronary heart disease.
      • Voight B.F.
      • Peloso G.M.
      • Orho-Melander M.
      • et al.
      Plasma HDL cholesterol and risk of myocardial infarction: a Mendelian randomisation study.
      • White J.
      • Swerdlow D.I.
      • Preiss D.
      • et al.
      Association of lipid fractions with risks for coronary artery disease and diabetes.
      • Schwartz G.G.
      • Olsson A.G.
      • Abt M.
      • et al.
      Effects of dalcetrapib in patients with a recent acute coronary syndrome.
      suggesting that increases in conventional measures of HDL-C may not lead to tangible benefits to CHD. However, this does not rule out a potentially important role of HDL-C in other diseases (including other vascular diseases, such as abdominal aortic aneurysm).
      • Lanktree M.B.
      • Thériault S.
      • Walsh M.
      • Paré G.
      HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
      • Burgess S.
      • Davey Smith G.
      Mendelian randomization implicates high-density lipoprotein cholesterol-associated mechanisms in etiology of age-related macular degeneration.
      • Harrison S.C.
      • Holmes M.V.
      • Burgess S.
      • et al.
      Genetic association of lipids and lipid drug targets with abdominal aortic aneurysm: a meta-analysis.
      In the accompanying study by Lanktree et al,
      • Lanktree M.B.
      • Thériault S.
      • Walsh M.
      • Paré G.
      HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
      the authors aimed to dissect the nature of the relationship between blood lipid concentrations and chronic kidney disease (CKD). Although traditional observational data provide evidence that HDL-C concentration is inversely associated with risk of kidney disease,
      • Baragetti A.
      • Norata G.D.
      • Sarcina C.
      • et al.
      High density lipoprotein cholesterol levels are an independent predictor of the progression of chronic kidney disease.
      such findings need to be interpreted with caution because the inherent limitations of observational research (namely, confounding and reverse causality) can distort findings. For example, the inverse association of HDL-C concentration with CHD seen in conventional observational studies
      • Emerging Risk Factors Collaboration.
      • Di Angelantonio E.
      • Sarwar N.
      • Perry P.
      • et al.
      Major lipids, apolipoproteins, and risk of vascular disease.
      has not been validated in clinical trials
      • Bowman L.
      • Hopewell J.C.
      • et al.
      HPS3/TIMI55-REVEAL Collaborative Group
      Effects of anacetrapib in patients with atherosclerotic vascular disease.
      • Holmes M.V.
      • Davey Smith G.
      Dyslipidaemia: Revealing the effect of CETP inhibition in cardiovascular disease.
      ; similar confounding could be at play in the reported associations of HDL-C with other diseases, including CKD. Mendelian randomization (MR) is an alternative analytical approach that uses genetic variants that are inherited at random and are nonmodifiable to make causal inference that should be relatively free of confounding and reverse causality.
      • Davey Smith G.
      • Ebrahim S.
      'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease?.
      • Davey Smith G.
      • Hemani G.
      Mendelian randomization: genetic anchors for causal inference in epidemiological studies.
      However, just as traditional observational studies have their inherent limitations, MR studies also make assumptions and have potential limitations that can cloud their interpretation.
      • Bowden J.
      • Del Greco M.F.
      • Minelli C.
      • Davey Smith G.
      • Sheehan N.
      • Thompson J.
      A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization.
      • Holmes M.V.
      • Ala-Korpela M.
      • Davey Smith G.
      Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
      • Hartwig F.P.
      • Davies N.M.
      • Hemani G.
      • Davey Smith G.
      Two-sample Mendelian randomisation: avoiding the downsides of a powerful, widely applicable but potentially fallible technique.
      • Haycock P.C.
      • Burgess S.
      • Wade K.H.
      • Bowden J.
      • Relton C.
      • Davey Smith G.
      Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies.
      Recent methodological developments in MR have allowed a relaxation of some of these assumptions and provide sensitivity analyses with which to scrutinize estimates from MR in further detail.
      • Bowden J.
      • Del Greco M.F.
      • Minelli C.
      • Davey Smith G.
      • Sheehan N.
      • Thompson J.
      A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization.
      • Bowden J.
      • Davey Smith G.
      • Burgess S.
      Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.
      • Bowden J.
      • Davey Smith G.
      • Haycock P.C.
      • Burgess S.
      Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator.
      • Hartwig F.P.
      • Davey Smith G.
      • Bowden J.
      Robust inference in two-sample Mendelian randomisation via the zero modal pleiotropy assumption.
      • Burgess S.
      • Dudbridge F.
      • Thompson S.G.
      Re: “Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects”.
      Improved methodologies, together with the availability of data platforms such as MR-Base
      • Hemani G.
      • Zheng J.
      • Wade K.H.
      • et al.
      MR-Base: a platform for systematic causal inference across the phenome using billions of genetic associations.
      and its extensions implemented in “MR of everything vs everything”
      • Hemani G.
      • Bowden J.
      • Haycock P.C.
      • et al.
      Automating Mendelian randomization through machine learning to construct a putative causal map of the human phenome.
      facilitate the conduct of MR analyses for multiple exposures and multiple outcomes.
      Lanktree et al
      • Lanktree M.B.
      • Thériault S.
      • Walsh M.
      • Paré G.
      HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
      used genetic variants identified from genome-wide association analyses that are associated with different concentrations of the 3 main blood lipid fractions (namely LDL-C, HDL-C, and TG) reported by the Global Lipids Genetics Consortium (GLGC).
      • Willer C.J.
      • Schmidt E.M.
      • et al.
      Global Lipids Genetics Consortium
      Discovery and refinement of loci associated with lipid levels.
      These genetic variants were used to gauge insight into the causal relationships of blood lipids with 3 markers of kidney function reported from genome-wide association studies by the CKD Genetics (CKDGen) consortium
      • Pattaro C.
      • Teumer A.
      • Gorski M.
      • et al.
      Genetic associations at 53 loci highlight cell types and biological pathways relevant for kidney function.
      : (1) estimated glomerular filtration rate (eGFR) as a continuous trait (percent difference), (2) dichotomized eGFR (odds ratio of eGFR < 60 mL/min/1.73 m2), and (3) albumin-creatinine ratio (ACR; percent difference). Using a 2-sample MR framework (in which the single-nucleotide polymorphism [SNP]-to-exposure and SNP-to-outcome estimates were obtained from predominantly nonoverlapping data sets, with the authors reporting that <10% of data overlapped between GLGC and CKDGen), Lanktree et al
      • Lanktree M.B.
      • Thériault S.
      • Walsh M.
      • Paré G.
      HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
      provide evidence in support of blood lipid concentrations being linked to kidney function.
      The authors identified that genetically elevated HDL-C concentrations were associated with better eGFRs (a higher percent difference in eGFR and lower risk for eGFR < 60 mL/min/1.73 m2) and lower ACR using genetic instruments for HDL-C. These associations remained robust to adjustment for the association of the genetic variants with LDL-C, TG, and hemoglobin A1c concentrations and blood pressure in so-called multivariable MR.
      • Burgess S.
      • Dudbridge F.
      • Thompson S.G.
      Re: “Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects”.
      Such findings are commensurate with HDL-C having a potentially protective role in kidney function. However, the authors note that treatment trials of drugs that increase HDL-C concentrations (such as the Atherothrombosis Intervention in Metabolic Syndrome With Low HDL/High Triglycerides: Impact on Global Health Outcomes [AIM-HIGH]
      • Kalil R.S.
      • Wang J.H.
      • de Boer I.H.
      • et al.
      Effect of extended-release niacin on cardiovascular events and kidney function in chronic kidney disease: a post hoc analysis of the AIM-HIGH trial.
      trial of niacin) had no discernible effect on kidney function. This discrepancy between a genetic instrument for HDL-C versus a specific therapy may arise for various reasons. First, a trial of an individual drug (such as niacin) may have a separate effect on kidney function compared with that of the overall causal effect of HDL-C concentration, the latter being a broad biomarker that has a genetic architecture comprising multiple independent loci.
      • Willer C.J.
      • Schmidt E.M.
      • et al.
      Global Lipids Genetics Consortium
      Discovery and refinement of loci associated with lipid levels.
      Second, blood lipids may play a role in the cause of kidney disease only at certain periods of life and not at others, a so-called “critical period” effect.
      • Holmes M.V.
      • Ala-Korpela M.
      • Davey Smith G.
      Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
      As an extension to the critical period effect, lipids could be a causal risk factor for kidney disease progression rather than disease onset; disease initiation and progression could have distinct causes, meaning that exposures causal for disease onset may not be necessarily causal for progression (and vice versa).
      • Paternoster L.
      • Tilling K.M.
      • Davey Smith G.
      Genetic epidemiology and Mendelian randomization for informing disease therapeutics: conceptual and methodological challenges.
      For LDL-C–related SNPs, a relationship of genetically elevated LDL-C concentration with a higher percentage difference in eGFR (ie, better kidney function) was identified when LDL-C SNPs were examined on their own. However, incorporating the relationship of the LDL-C SNPs with other lipids, hemoglobin A1c, and blood pressure, the association between genetically elevated LDL-C concentration with percentage difference in eGFR became less pronounced. For TG, the relationship of genetically elevated TG concentration with percentage difference in eGFR, while weak on its individual analysis, became more pronounced on adjustment for these other traits.
      These relationships are nontrivial to tease apart. For example, adjusting the LDL-C SNPs for hemoglobin A1c concentration (a marker of dysglycemia) could adjust for a potential mediating effect of diabetes on CKD, thereby resulting in the attenuation of the relationship between LDL-C concentration and percentage difference in eGFR that the authors report. Prior MR studies have shown that higher LDL-C concentration is related to lower risk for type 2 diabetes mellitus,
      • White J.
      • Swerdlow D.I.
      • Preiss D.
      • et al.
      Association of lipid fractions with risks for coronary artery disease and diabetes.
      • Fall T.
      • Xie W.
      • Poon W.
      • et al.
      Using genetic variants to assess the relationship between circulating lipids and type 2 diabetes.
      meaning that a causal pathway could exist from higher LDL-C concentration to lower risk for type 2 diabetes mellitus and lower risk for CKD. Alternatively, the wider 95% confidence intervals and resultant attenuated effect on percentage difference in eGFR in multivariable adjustment could simply arise from the imprecision introduced by the multivariate model, meaning that a true relationship might exist.
      The pattern of consistency of the association of the lipid traits with the 3 kidney traits (percent difference in eGFR, risk for eGFR < 60 mL/min/1.73 m2, and percent difference in ACR) is where the complexity becomes further apparent. In the case of HDL-C, there is, as one might expect, a directionally consistent relationship of HDL-C concentration with percent difference in eGFR, risk for low eGFR, and percent difference in ACR. The consistency across these traits (although 2 are essentially marking the same entity; ie, eGFR) for HDL-C lends weight to a potential protective role of HDL-C in CKD. The same is not the case for LDL-C or TG, for which both traits appear to associate with higher percent differences in both eGFR and ACR, potentially indicating a physiologic phenomenon for which there is higher filtration yet deteriorating function.
      This study raises several questions about how to reliably interpret these various strands of evidence. Undoubtedly the main challenge in the wake of abundant genome-wide data and large-scale resources such as the UK Biobank is how to address the potential for genetic pleiotropy to confound the estimates derived from MR. At a June 2017 MR conference hosted by the Medical Research Council Integrative Epidemiology Unit in Bristol, United Kingdom, more than 30 MR methodologies were presented, the majority of which are new and have yet to be subjected to the same scrutiny as those that are becoming more established in the MR field.
      • Hartwig F.P.
      • Davey Smith G.
      • Bowden J.
      Robust inference in two-sample Mendelian randomisation via the zero modal pleiotropy assumption.
      Although exciting for those of us active in this field, it also poses major challenges; for example, which approaches should we use in our battery of tests when conducting MR, and what is the added value of the newer methodologies? This will no doubt be the subject of many narrative reviews to follow, but allow us to synthesize a few points below, based on those MR approaches that are now commonly used.
      In the absence of genetic pleiotropy (the scenario in which ≥1 genetic variant used in a genetic instrument associates with >1 phenotype, described in detail in a recent review
      • Holmes M.V.
      • Ala-Korpela M.
      • Davey Smith G.
      Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
      ), the conventional MR estimate ought to provide a reliable guide to the causal relationship of an exposure on an outcome. Established sensitivity analyses include MR-Egger–,
      • Bowden J.
      • Davey Smith G.
      • Burgess S.
      Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.
      median-,
      • Bowden J.
      • Davey Smith G.
      • Haycock P.C.
      • Burgess S.
      Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator.
      and mode-based
      • Hartwig F.P.
      • Davey Smith G.
      • Bowden J.
      Robust inference in two-sample Mendelian randomisation via the zero modal pleiotropy assumption.
      approaches. Each has its own assumptions on the type and amount of genetic confounding (see Table 5 in Hartwig et al
      • Hartwig F.P.
      • Davey Smith G.
      • Bowden J.
      Robust inference in two-sample Mendelian randomisation via the zero modal pleiotropy assumption.
      ), meaning that if generally consistent findings are identified across the various approaches (eg, as was the case for the relationship of education and CHD
      • Tillmann T.
      • Vaucher J.
      • Okbay A.
      • et al.
      Education and coronary heart disease: Mendelian randomisation study.
      ), gross pleiotropy is unlikely to account for the findings. These updated methods, although invaluable for testing the assumptions implicit in instrumental variable analyses, are not a panacea, and critical issues arise in selecting appropriate variants for MR and interpreting their findings as described in a recent review.
      • Holmes M.V.
      • Ala-Korpela M.
      • Davey Smith G.
      Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
      Multivariable MR is increasingly being used to genetically “adjust” for other traits.
      • Burgess S.
      • Dudbridge F.
      • Thompson S.G.
      Re: “Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects”.
      This has been used, for example, in clarifying the role of lipids in heart disease, in which the association of HDL-C concentration with risk for CHD diminished on adjustment for TG concentration.
      • White J.
      • Swerdlow D.I.
      • Preiss D.
      • et al.
      Association of lipid fractions with risks for coronary artery disease and diabetes.
      Challenges in interpreting findings from multivariable MR analyses can arise, as exemplified by the following 2 scenarios. First, when a genetic instrument shows associations with another trait (Fig 1, scenario 1), which may be a causal intermediate, a well-conducted multivariable MR should in theory provide a “direct” (rather than “total”) effect of the exposure on risk for disease; the direct effect is that which is due only to the exposure of interest and not mediated by another trait, even when the latter trait may be influenced by the exposure of interest (eg, in the case of adjusting the relationship of age at menarche with cancer for adult body mass index, for which it is known that age at menarche influences adult body mass index
      • Burgess S.
      • Thompson D.J.
      • Rees J.M.B.
      • Day F.R.
      • Perry J.R.
      • Ong K.K.
      Dissecting causal pathways using Mendelian randomization with summarized genetic data: application to age at menarche and risk of breast cancer.
      • Day F.R.
      • Thompson D.J.
      • Helgason H.
      • et al.
      Genomic analyses identify hundreds of variants associated with age at menarche and support a role for puberty timing in cancer risk.
      ). In clinical and public health terms, the total (rather than direct) effect is, of course, what modifying the exposure would produce. In the multivariable MR framework, measurement error in the intermediate phenotypes can influence the findings and may not produce entirely reliable adjusted estimates. This leads to multivariable MR not being an optimal approach for mediation analysis. Two-step MR
      • Relton C.L.
      • Davey Smith G.
      Two-step epigenetic Mendelian randomization: a strategy for establishing the causal role of epigenetic processes in pathways to disease.
      can provide quantifiable estimates of the proportion of mediation and, in contrast to multivariable MR, 2-step MR does not have error in the exposure to mediating phenotype relationship.
      • Richmond R.C.
      • Hemani G.
      • Tilling K.
      • Davey Smith G.
      • Relton C.L.
      Challenges and novel approaches for investigating molecular mediation.
      Second, in the case of overlapping traits (eg, non–HDL-C and TG
      • Helgadottir A.
      • Gretarsdottir S.
      • Thorleifsson G.
      • et al.
      Variants with large effects on blood lipids and the role of cholesterol and triglycerides in coronary disease.
      ), such multivariate MR analyses do not provide meaningful estimates because they in effect adjust for the same trait that is measured, abrogating the underlying relationships
      • Holmes M.V.
      • Ala-Korpela M.
      • Davey Smith G.
      Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
      (Fig 1, scenario 2). Given these (nonexhaustive) scenarios, multivariable MR is an approach that can generate more questions than it answers. A recent modification to multivariable MR in the form of multivariable MR-Egger
      • Rees J.M.B.
      • Wood A.M.
      • Burgess S.
      Extending the MR-Egger method for multivariable Mendelian randomization to correct for both measured and unmeasured pleiotropy.
      enhances the conventional multivariable MR in additionally adjusting for unbalanced genetic pleiotropy of the genetic instruments and can quantify the extent to which adjustment for other traits addresses unbalanced pleiotropy.
      Figure thumbnail gr1
      Figure 1Challenges in interpreting multivariable Mendelian randomization (MR) analyses. Scenario 1: use of multivariable MR when the second biomarker (in addition to the exposure) lies on the causal pathway to disease. Adjusting for a potential mediator as a form of mediation analysis is suboptimal because error terms in the exposure to intermediate relationship are not properly accounted for in the analysis. Scenario 2: use of multivariable MR when the second biomarker measures the same entities as the primary exposure. Adjusting an exposure for an overlapping trait has the net effect of autoadjusting, meaning that the findings from multivariable MR are unreliable. In scenario 1, biomarker 2 is a mediating phenotype of the exposure (eg, investigating the degree to which triglycerides mediate the relationship of body mass index with risk for coronary heart disease
      • Xu L.
      • Borges M.C.
      • Hemani G.
      • Lawlor D.A.
      The role of glycaemic and lipid risk factors in mediating the effect of BMI on coronary heart disease: a two-step, two-sample Mendelian randomisation study.
      ); in scenario 2, biomarker 2 is an overlapping trait with the exposure (eg, assessing the role of triglycerides and non–high-density lipoprotein cholesterol in risk for coronary heart disease
      • Helgadottir A.
      • Gretarsdottir S.
      • Thorleifsson G.
      • et al.
      Variants with large effects on blood lipids and the role of cholesterol and triglycerides in coronary disease.
      ). Abbreviation: SNP, single-nucleotide polymorphism.
      Returning to the article by Lanktree et al,
      • Lanktree M.B.
      • Thériault S.
      • Walsh M.
      • Paré G.
      HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
      the findings provide genetic support for the hypothesis that HDL-C may be causally protective for kidney disease. How might these findings be potentially translated to impact on patient care? As described, MR of a complex phenotype (such as HDL-C) is distinct to MR of a drug target (eg, cholesteryl ester transfer protein [CETP]). Natural areas for further investigation should now include individual drug targets that alter HDL-C concentrations (such as CETP) for which genetic studies
      • Thompson A.
      • Di Angelantonio E.
      • Sarwar N.
      • et al.
      Association of cholesteryl ester transfer protein genotypes with CETP mass and activity, lipid levels, and coronary risk.
      • Ference B.A.
      • Kastelein J.J.P.
      • Ginsberg H.N.
      • et al.
      Association of genetic variants related to CETP inhibitors and statins with lipoprotein levels and cardiovascular risk.
      have successfully predicted (albeit after a tortuous path
      • Schwartz G.G.
      • Olsson A.G.
      • Abt M.
      • et al.
      Effects of dalcetrapib in patients with a recent acute coronary syndrome.
      • Lincoff A.M.
      • Nicholls S.J.
      • Riesmeyer J.S.
      • et al.
      Evacetrapib and cardiovascular outcomes in high-risk vascular disease.
      ) the findings of the recently announced phase 3 trial of the CETP inhibitor anacetrapib.
      • Bowman L.
      • Hopewell J.C.
      • et al.
      HPS3/TIMI55-REVEAL Collaborative Group
      Effects of anacetrapib in patients with atherosclerotic vascular disease.

      References

        • Silverman M.G.
        • Ference B.A.
        • Im K.
        • et al.
        Association between lowering LDL-C and cardiovascular risk reduction among different therapeutic interventions: a systematic review and meta-analysis.
        JAMA. 2016; 316: 1289-1297
        • Emerging Risk Factors Collaboration.
        • Di Angelantonio E.
        • Sarwar N.
        • Perry P.
        • et al.
        Major lipids, apolipoproteins, and risk of vascular disease.
        JAMA. 2009; 302: 1993-2000
        • Baigent C.
        • Blackwell L.
        • et al.
        • Cholesterol Treatment Trialists' Collaborators
        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: 1670-1681
        • Mihaylova B.
        • Emberson J.
        • et al.
        • Cholesterol Treatment Trialists' 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: 581-590
        • Holmes M.V.
        • Asselbergs F.W.
        • Palmer T.M.
        • et al.
        Mendelian randomization of blood lipids for coronary heart disease.
        Eur Heart J. 2015; 36: 539-550
        • Ference B.A.
        • Yoo W.
        • Alesh I.
        • et al.
        Effect of long-term exposure to lower low-density lipoprotein cholesterol beginning early in life on the risk of coronary heart disease: a Mendelian randomization analysis.
        J Am Coll Cardiol. 2012; 60: 2631-2639
        • Voight B.F.
        • Peloso G.M.
        • Orho-Melander M.
        • et al.
        Plasma HDL cholesterol and risk of myocardial infarction: a Mendelian randomisation study.
        Lancet. 2012; 380: 572-580
        • White J.
        • Swerdlow D.I.
        • Preiss D.
        • et al.
        Association of lipid fractions with risks for coronary artery disease and diabetes.
        JAMA Cardiol. 2016; 1: 692-699
        • Schwartz G.G.
        • Olsson A.G.
        • Abt M.
        • et al.
        Effects of dalcetrapib in patients with a recent acute coronary syndrome.
        N Engl J Med. 2012; 367: 2089-2099
        • Lanktree M.B.
        • Thériault S.
        • Walsh M.
        • Paré G.
        HDL cholesterol, LDL cholesterol, and triglycerides as risk factors for CKD: a Mendelian randomization study.
        Am J Kidney Dis. 2018; 71: 166-172
        • Burgess S.
        • Davey Smith G.
        Mendelian randomization implicates high-density lipoprotein cholesterol-associated mechanisms in etiology of age-related macular degeneration.
        Ophthalmology. 2017; 124: 1165-1174
        • Harrison S.C.
        • Holmes M.V.
        • Burgess S.
        • et al.
        Genetic association of lipids and lipid drug targets with abdominal aortic aneurysm: a meta-analysis.
        JAMA Cardiol. 2017; (https://doi.org/10.1001/jamacardio.2017.4293)
        • Baragetti A.
        • Norata G.D.
        • Sarcina C.
        • et al.
        High density lipoprotein cholesterol levels are an independent predictor of the progression of chronic kidney disease.
        J Intern Med. 2013; 274: 252-262
        • Bowman L.
        • Hopewell J.C.
        • et al.
        • HPS3/TIMI55-REVEAL Collaborative Group
        Effects of anacetrapib in patients with atherosclerotic vascular disease.
        N Engl J Med. 2017; 377: 1217-1227
        • Holmes M.V.
        • Davey Smith G.
        Dyslipidaemia: Revealing the effect of CETP inhibition in cardiovascular disease.
        Nat Rev Cardiol. 2017; 14: 635-636
        • Davey Smith G.
        • Ebrahim S.
        'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease?.
        Int J Epidemiol. 2003; 32: 1-22
        • Davey Smith G.
        • Hemani G.
        Mendelian randomization: genetic anchors for causal inference in epidemiological studies.
        Hum Mol Genet. 2014; 23: R89-R98
        • Bowden J.
        • Del Greco M.F.
        • Minelli C.
        • Davey Smith G.
        • Sheehan N.
        • Thompson J.
        A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization.
        Stat Med. 2017; 36: 1783-1802
        • Holmes M.V.
        • Ala-Korpela M.
        • Davey Smith G.
        Mendelian randomization in cardiometabolic disease: challenges in evaluating causality.
        Nat Rev Cardiol. 2017; 14: 577-590
        • Hartwig F.P.
        • Davies N.M.
        • Hemani G.
        • Davey Smith G.
        Two-sample Mendelian randomisation: avoiding the downsides of a powerful, widely applicable but potentially fallible technique.
        Int J Epidemiol. 2016; 6: 1717-1726
        • Haycock P.C.
        • Burgess S.
        • Wade K.H.
        • Bowden J.
        • Relton C.
        • Davey Smith G.
        Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies.
        Am J Clin Nutr. 2016; 103: 965-978
        • 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: 512-525
        • Bowden J.
        • Davey Smith G.
        • Haycock P.C.
        • Burgess S.
        Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator.
        Genet Epidemiol. 2016; 40: 304-314
        • Hartwig F.P.
        • Davey Smith G.
        • Bowden J.
        Robust inference in two-sample Mendelian randomisation via the zero modal pleiotropy assumption.
        Int J Epidemiol. 2017; (https://doi.org/10.1093/ije/dyx102)
        • Burgess S.
        • Dudbridge F.
        • Thompson S.G.
        Re: “Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects”.
        Am J Epidemiol. 2015; 181 ([letter]): 290-291
        • Hemani G.
        • Zheng J.
        • Wade K.H.
        • et al.
        MR-Base: a platform for systematic causal inference across the phenome using billions of genetic associations.
        bioRxiv. 2016; (https://doi.org/10.1101/078972)
        • Hemani G.
        • Bowden J.
        • Haycock P.C.
        • et al.
        Automating Mendelian randomization through machine learning to construct a putative causal map of the human phenome.
        bioRxiv. 2017; (https://doi.org/10.1101/173682)
        • Willer C.J.
        • Schmidt E.M.
        • et al.
        • Global Lipids Genetics Consortium
        Discovery and refinement of loci associated with lipid levels.
        Nat Genet. 2013; 45: 1274-1283
        • Pattaro C.
        • Teumer A.
        • Gorski M.
        • et al.
        Genetic associations at 53 loci highlight cell types and biological pathways relevant for kidney function.
        Nat Commun. 2016; 7: 10023
        • Kalil R.S.
        • Wang J.H.
        • de Boer I.H.
        • et al.
        Effect of extended-release niacin on cardiovascular events and kidney function in chronic kidney disease: a post hoc analysis of the AIM-HIGH trial.
        Kidney Int. 2015; 87: 1250-1257
        • Paternoster L.
        • Tilling K.M.
        • Davey Smith G.
        Genetic epidemiology and Mendelian randomization for informing disease therapeutics: conceptual and methodological challenges.
        PLoS Genet. 2017; 13: e1006944
        • Fall T.
        • Xie W.
        • Poon W.
        • et al.
        Using genetic variants to assess the relationship between circulating lipids and type 2 diabetes.
        Diabetes. 2015; 64: 2676-2684
        • Tillmann T.
        • Vaucher J.
        • Okbay A.
        • et al.
        Education and coronary heart disease: Mendelian randomisation study.
        BMJ. 2017; 358: j3542
        • Burgess S.
        • Thompson D.J.
        • Rees J.M.B.
        • Day F.R.
        • Perry J.R.
        • Ong K.K.
        Dissecting causal pathways using Mendelian randomization with summarized genetic data: application to age at menarche and risk of breast cancer.
        Genetics. 2017; 207: 481-487
        • Day F.R.
        • Thompson D.J.
        • Helgason H.
        • et al.
        Genomic analyses identify hundreds of variants associated with age at menarche and support a role for puberty timing in cancer risk.
        Nat Genet. 2017; 49: 834-841
        • Relton C.L.
        • Davey Smith G.
        Two-step epigenetic Mendelian randomization: a strategy for establishing the causal role of epigenetic processes in pathways to disease.
        Int J Epidemiol. 2012; 41: 161-176
        • Richmond R.C.
        • Hemani G.
        • Tilling K.
        • Davey Smith G.
        • Relton C.L.
        Challenges and novel approaches for investigating molecular mediation.
        Hum Mol Genet. 2016; 25: R149-R156
        • Helgadottir A.
        • Gretarsdottir S.
        • Thorleifsson G.
        • et al.
        Variants with large effects on blood lipids and the role of cholesterol and triglycerides in coronary disease.
        Nat Genet. 2016; 48: 634-639
        • Rees J.M.B.
        • Wood A.M.
        • Burgess S.
        Extending the MR-Egger method for multivariable Mendelian randomization to correct for both measured and unmeasured pleiotropy.
        Stat Med. 2017; 36: 4705-4718
        • Thompson A.
        • Di Angelantonio E.
        • Sarwar N.
        • et al.
        Association of cholesteryl ester transfer protein genotypes with CETP mass and activity, lipid levels, and coronary risk.
        JAMA. 2008; 299: 2777-2788
        • Ference B.A.
        • Kastelein J.J.P.
        • Ginsberg H.N.
        • et al.
        Association of genetic variants related to CETP inhibitors and statins with lipoprotein levels and cardiovascular risk.
        JAMA. 2017; 318: 947-956
        • Lincoff A.M.
        • Nicholls S.J.
        • Riesmeyer J.S.
        • et al.
        Evacetrapib and cardiovascular outcomes in high-risk vascular disease.
        N Engl J Med. 2017; 376: 1933-1942
        • Xu L.
        • Borges M.C.
        • Hemani G.
        • Lawlor D.A.
        The role of glycaemic and lipid risk factors in mediating the effect of BMI on coronary heart disease: a two-step, two-sample Mendelian randomisation study.
        Diabetologia. 2017; 60: 2210-2220

      Linked Article