Research Techniques Made Simple| Volume 139, ISSUE 7, P1416-1421.e1, July 01, 2019

# Research Techniques Made Simple: Using Genetic Variants for Randomization

Open Archive
Observational epidemiological studies have identified associations between a number of modifiable exposures and outcomes, including in dermatology, such as between smoking and psoriasis. However, it is challenging to determine if such relationships are causal because of the potential of confounding and reverse causation. Mendelian randomization (MR) is a statistical method that can be used to investigate the causal relationships between an exposure and outcome by using a genetic instrument that proxies the exposure. The resulting estimate (under certain assumptions) can be interpreted as the causal estimate, free of confounding and reverse causation. In this review, we provide an overview of how to undertake an MR analysis, with examples from the dermatology literature. We also discuss the challenges and future directions of this method.

#### Abbreviations:

BMI (body mass index), GWAS (genome-wide association study), MR (Mendelian randomization)
CME Activity Dates: 20 June 2019
Expiration Date: 19 June 2020
Estimated Time to Complete: 1 hour
Planning Committee/Speaker Disclosure: Lavinia Paternoster is a consultant/advisor for Merck. All other authors, planning committee members, CME committee members and staff involved with this activity as content validation reviewers have no financial relationships with commercial interests to disclose relative to the content of this CME activity.
Commercial Support Acknowledgment: This CME activity is supported by an educational grant from Lilly USA, LLC.
Description: This article, designed for dermatologists, residents, fellows, and related healthcare providers, seeks to reduce the growing divide between dermatology clinical practice and the basic science/current research methodologies on which many diagnostic and therapeutic advances are built.
Objectives: At the conclusion of this activity, learners should be better able to:
• Recognize the newest techniques in biomedical research.
• Describe how these techniques can be utilized and their limitations.
• Describe the potential impact of these techniques.
CME Accreditation and Credit Designation: This activity has been planned and implemented in accordance with the accreditation requirements and policies of the Accreditation Council for Continuing Medical Education through the joint providership of Beaumont Health and the Society for Investigative Dermatology. Beaumont Health is accredited by the ACCME to provide continuing medical education for physicians. Beaumont Health designates this enduring material for a maximum of 1.0 AMA PRA Category 1 Credit(s)™. Physicians should claim only the credit commensurate with the extent of their participation in the activity.
Method of Physician Participation in Learning Process: The content can be read from the Journal of Investigative Dermatology website: http://www.jidonline.org/current. Tests for CME credits may only be submitted online at https://beaumont.cloud-cme.com/RTMS-Jul19 – click ‘CME on Demand’ and locate the article to complete the test. Fax or other copies will not be accepted. To receive credits, learners must review the CME accreditation information; view the entire article, complete the post-test with a minimum performance level of 60%; and complete the online evaluation form in order to claim CME credit. The CME credit code for this activity is: 21310. For questions about CME credit email [email protected] .

## Introduction

### Summary Points

• Mendelian randomization (MR) is a statistical method for investigating causality between exposure and outcome variables in observational epidemiology.
• Unlike traditional observational studies, MR uses genetic variants as instruments (or proxies) for the exposure, hence avoiding confounding and reverse causation.
• Application of such methods in the field of dermatology is a promising area of research.
• Future directions and developments will allow MR to be a valuable tool for investigating causal pathways for disease, as well as providing insight into therapeutic interventions.
Observational epidemiological studies have uncovered relationships between disease and various explanatory factors known as exposures (Table 1). Notable examples in dermatology include the association of psoriasis with smoking (
• Armstrong A.W.
• Harskamp C.T.
• Dhillon J.S.
• Armstrong E.J.
• Armstrong A.W.
Psoriasis and smoking: a systematic review and meta-analysis funding sources.
) and, more recently, the association of atopic dermatitis with cardiovascular traits (
• Standl M.
• Tesch F.
• Baurecht H.
• Rodríguez E.
• Müller-Nurasyid M.
• Gieger C.
• et al.
Association of atopic dermatitis with cardiovascular risk factors and diseases.
). However, traditional observational studies are prone to biases such as confounding, where the observed association may be due to the exposure being related to other lifestyle or socioeconomic factors that have a casual influence on disease. Furthermore, the observed associations may be due to reverse causation, where disease is actually influencing the assumed exposure (
• Lawlor D.A.
• Harbord R.M.
• Sterne J.A.C.
• Timpson N.
• Davey Smith G.
Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.
); for example, having psoriasis could influence an individual’s propensity to smoke. Mendelian randomization (MR) presents a method for evaluating causality in an observational study setting. We aim to provide an overview of the principle of MR and the statistical methods used.
Table 1Glossary
TermDescription
ConfounderA variable that is a common cause of both the exposure and the outcome.
ExposureAn explanatory variable used to explain or predict an outcome variable, such as a trait or disease.
F-statisticObtained from the regression of a response variable on a predictor variable, for example, the regression of the exposure of interest on an instrumental variable (IV). This can be used as a measure of the strength of association between an IV and the exposure, thereby giving an indication of the strength of the instrument. The further away the F-statistic is from 1, the stronger the instrument. The F-statistic also depends on the size of the sample.
GWASGenome-wide association study. Involves analyzing genetic variants across the genome, such as single-nucleotide polymorphisms for association with a disease or trait of interest.
Instrumental variable (IV)A variable that is associated with an exposure of interest but not the outcome. In MR studies, genetic variants are used as IVs. A valid IV must also be independent of confounders of the exposure-outcome association and must affect only the outcome via the exposure.
Mendelian randomizationA method for assessing the causal effect of an exposure on an outcome by using genetic variants as instruments or proxies for the exposure variable.
MR-baseA centralized database of summary GWAS data and an analytical platform to perform Mendelian randomization and sensitivity analyses.
PheWASPhenome-wide association study. Involves analyzing the association between genetic variants and multiple phenotypic variables (on a phenome-wide scale) rather than a single phenotype.
PleiotropyOccurs when a genetic instrument is independently associated with multiple risk factors for the outcome, in addition to the exposure of interest. This results in the third IV assumption being violated, which assumes that the genetic instrument affects only the outcome via the exposure.
Reverse causalityWhere an association is due to the assumed outcome variable influencing the exposure variable rather than the exposure influencing the outcome.
Sensitivity analysisPerformed to assess the robustness of the main analysis or the validity of the main results.

## The Principle of MR

MR is a form of instrumental variable analysis whereby genetic variants are used as instruments (or proxies) for an exposure of interest (Table 1). Because genetic variants are randomly segregated at conception and cannot be influenced by confounding factors or the outcome itself, they can be used to estimate the causal effect of the exposure upon an outcome (
• Lawlor D.A.
• Harbord R.M.
• Sterne J.A.C.
• Timpson N.
• Davey Smith G.
Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.
) (Figure 1).
Performing MR requires two pieces of information: (i) the effect of the genetic instrument on the exposure ($βXZ$) and (ii) the effect of the genetic instrument on the outcome ($βYZ$). These can then be used to estimate the causal effect of the exposure on the outcome ($causalβYX$) with the following ratio (
• Wald A.
The fitting of straight lines if both variables are subject to error.
): $causalβYX=βYZβXZ$.
For a genetic variant to qualify as an instrumental variable, three core assumptions must be satisfied: the variants (i) must be truly associated with the exposure of interest, (ii) must not be associated with confounders of the exposure-outcome relationship, and (iii) must affect only the outcome via the exposure and not through an alternative pathway (
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
). The use of genetic variants in an MR framework can be compared with a randomized controlled trial, where genotypes are used to randomize individuals to different subgroups (
• Lawlor D.A.
• Harbord R.M.
• Sterne J.A.C.
• Timpson N.
• Davey Smith G.
Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.
). The effect of the genetic instrument on the outcome ($βYZ$) is analogous to an intention-to-treat effect from an association between randomization and an outcome in a randomized controlled trial (
• Burgess S.
• Small D.S.
• Thompson S.G.
A review of instrumental variable estimators for Mendelian randomization.
).
Because MR requires estimates of the associations between genetic variants and the exposure and genetic variants and the outcome, the rise of genome-wide association studies (GWASs) (
• Tsoi L.C.
• Patrick M.T.
• Elder J.T.
Research techniques made simple: using genome-wide association studies to understand complex cutaneous disorders.
) provides a wealthy resource of genetic instruments for MR. Published summary GWAS data can be obtained from various sources such as the GWAS catalogue (www.ebi.ac.uk/gwas/) and MR-base (www.mrbase.org) or directly from the authors of the GWAS (Figure 2). Commonly, independent single-nucleotide polymorphisms (SNPs) that have been reported to be associated with an exposure on a genome-wide significance level (P-value < 5 × 10–8) are used as genetic instruments for the exposure (
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
), but MR analyses can be conducted by using just a single genetic variant or even using all variants in the genome (appropriately weighted by their effect on the exposure). Published MR studies in dermatology include those investigating causal relationships between fatty acids and melanoma (
• Liyanage U.E.
• Law M.H.
• Ong J.S.
• Cust A.E.
• Mann G.J.
• Ward S.V.
• et al.
Polyunsaturated fatty acids and risk of melanoma: a Mendelian randomisation analysis.
), vitamin D levels and AD risk (
• Manousaki D.
• Paternoster L.
• Standl M.
• Moffatt M.F.
• Farrall M.
• Bouzigon E.
• et al.
Vitamin D levels and susceptibility to asthma, elevated immunoglobulin E levels, and atopic dermatitis: a Mendelian randomization study.
) as well as skin aging (
• Noordam R.
• Hamer M.A.
• Pardo L.M.
• van der Nat T.
• Kiefte-de Jong J.C.
• Kayser M.
• et al.
No causal association between 25-hydroxyvitamin D and features of skin aging: evidence from a bidirectional Mendelian randomization study.
), and, most recently, body mass index (BMI) and psoriasis risk (
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
), which will be referred to throughout this review.

## MR Approaches and Statistical Methods

### MR study designs

A basic MR study design involves obtaining all information required from the same set of individuals, meaning that the genetic, exposure, and outcome data are all available from the same study. This is known as one-sample MR (Table 2). Large population-based studies such as the UK Biobank provide ideal data sets for such analyses to be carried out. However, it may not always be possible to gather exposure and outcome measures from the same data set. Two-sample MR is therefore more commonly adopted, whereby the effect of genetic variants on the exposure is obtained from one sample, and the effect of genetic variants on the outcome is obtained from another (Table 2). This approach has been greatly facilitated by the increasing availability of summary GWAS data, as well as analytical platforms to perform two-sample MR, such as MR-base. The steps for a two-sample MR are shown in Figure 2 (
• Hemani G.
• Zheng J.
• Elsworth B.
• Haberland V.
• Baird D.
• et al.
The MR-base platform supports systematic causal inference across the human phenome.
).
Table 2Methods and approaches for MR analysis
CategoryDescription
MR study design
One-sample MRPerformed with genetic instruments, exposure and outcome data that have been measured in the same sample population.
Two-sample MRThe effect of the genetic instruments on the exposure and the effect of the genetic instruments on the outcome are obtained from a non-overlapping sample populations.
Bidirectional MRThe causal relationship between two traits is investigated in both directions. This approach can be applied to one-sample or two-sample MR.
Statistical methods
Wald ratio methodPerformed with a single genetic instrument (or genetic risk score) by dividing the coefficient of the outcome-instrument association by the coefficient of the exposure-instrument association.
Two-stage least squares (2SLS) regressionInvolves two regression stages where the exposure is regressed on the genetic instruments. The outcome is then regressed on the genetically predicted exposure values from the first-stage regression.
Combining multiple variants
Inverse-variance weighted (IVW) estimatorCombination of ratio estimates from individual variants in a fixed-effect meta-analysis. The contribution of each instrument is the inverse of the variance of its effect on the outcome.
Genetic risk score (GRS)Multiple genetic instruments for an exposure are combined into a genetic risk score. This can then be used as a single instrument to perform MR.
Sensitivity analysis
MR-Egger regressionSensitivity analysis to perform MR with multiple instruments. This can be used to detect pleiotropy and provide a causal estimate that is robust to pleiotropy.
Weighted-median estimatorSensitivity analysis to perform MR with multiple instruments. Will provide consistent causal estimates when at least 50% of the information in the analysis comes from valid genetic instruments.
Mode-based estimatorAn MR sensitivity analysis that will provide a robust causal estimate in the presence of pleiotropy, if the most common pleiotropy value is zero across the genetic instruments.
Latent causal variable analysisDistinguishes between genetic correlation and causation by mediating the genetic correlation between two traits with a latent causal variable that itself has a causal effect on each trait.
Abbreviation: Mendelian randomization.
We recently investigated causality between BMI and psoriasis using both one-sample MR with individual-level data from the UK Biobank and Nord-Trøndelag Health Study (i.e., HUNT) and two-sample MR with published summary GWAS data. Consistent results were obtained from both analyses. The combined causal estimate suggested a 9% increase in the risk of psoriasis per 1 unit increase in BMI (
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
) (Figure 3). This finding supports previous reports of weight loss improving the prognosis of psoriasis (
• Maglio C.
• Peltonen M.
• Rudin A.
• Carlsson L.M.S.
Bariatric surgery and the incidence of psoriasis and psoriatic arthritis in the Swedish obese subjects study.
) and could suggest weight control as an intervention to prevent or treat psoriasis.
A bidirectional MR approach can also be adopted that investigates causal effects in both directions (Table 2). This requires suitable genetic instruments to be available for both traits. Such analysis can help uncover the direction of causality that explains the observational association. For example, when considering the relationship between BMI and psoriasis, we performed bidirectional MR and found evidence that the observational relationship is largely due to the causal effect of higher BMI on psoriasis risk rather than a causal effect of psoriasis influencing BMI (
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
).

### MR statistical methods

The simplest method to perform MR involves dividing the effect of the genetic instrument on the outcome by the effect of the genetic instrument on the exposure. This is commonly termed the ratio of coefficients method or the Wald ratio method (as described earlier) and can be performed with either summarized or individual-level data (
• Burgess S.
• Small D.S.
• Thompson S.G.
A review of instrumental variable estimators for Mendelian randomization.
). Two-stage methods can also be applied, such as two-stage least squares, as used in the BMI and psoriasis article by
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
(Table 2). This method involves regressing the exposure on the genetic instruments and then regressing the outcome on the genetically predicted values from the first regression, which allows for the true standard error to be estimated. Additional MR methods have been previously discussed elsewhere (
• Burgess S.
• Small D.S.
• Thompson S.G.
A review of instrumental variable estimators for Mendelian randomization.
).

### Combining multiple variants

Where multiple genetic instruments are available for an exposure, these can be combined into a genetic risk score and used as a single instrument to perform MR (
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
). Alternatively, an inverse-variance–weighted approach can be applied, whereby the ratio estimate from each independent genetic variant is combined by using a fixed-effect meta-analysis model, where each variant is assumed to provide independent information, and the contribution of each variant is the inverse of the variance of its effect on the outcome () (Table 2).

### Sensitivity methods

One major potential problem with MR occurs when the genetic instrument affects the outcome through an alternative pathway that is distinct from the exposure of interest (termed pleiotropy) (Table 1), which violates the third assumption (as outlined earlier). Various sensitivity methods have been developed to detect and address pleiotropy, including MR-Egger regression, weighted-median analysis, the mode-based estimate, and the latent causal variable method (Table 2). These methods have different assumptions, but they aim to estimate the true causal effect in the presence of modest levels of pleiotropy (
• O’Connor L.J.
• Price A.L.
Distinguishing genetic correlation from causation across 52 diseases and complex traits.
,
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
).

### Challenges and limitations of MR studies

Although MR has proven to be a useful tool for estimating causality, there are instances where MR may be limited or the instrumental variable assumptions may be violated. In some cases, there may be only weak genetic instruments available for the exposure of interest. Genetic instruments that explain very little of the variance in exposure can result in weak instrument bias, where the causal estimates can be biased toward the null in a two-sample MR setting and toward the observational estimate in a one-sample MR setting (
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
). This highlights the need for GWASs to uncover associated variants and strong, reliable instruments to perform MR. The F-statistic from the regression of the exposure on the genetic instrument indicates the strength of the instrument (Table 1). It is recommended that genetic variants with an F-statistic greater than 10 be used (
• Burgess S.
• Butterworth A.
• Thompson S.G.
Mendelian randomization analysis with multiple genetic variants using summarized data.
,
• Lawlor D.A.
• Harbord R.M.
• Sterne J.A.C.
• Timpson N.
• Davey Smith G.
Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.
). Because the F-statistic is dependent on sample size, weak instrument bias can also be addressed by using larger sample sizes (
• Burgess S.
• Small D.S.
• Thompson S.G.
A review of instrumental variable estimators for Mendelian randomization.
). Additionally, combining individual variants into a genetic risk score increases the instrument strength. The instrument for BMI in our psoriasis analysis had an F-statistic of 7,091, indicating a strong instrument for BMI (
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
).
Although it is assumed that a genetic instrument is independent of confounders, this cannot be tested for all potential confounders. However, it is sensible to test for association between the genetic instrument and any available measured potential confounders.

### Applications and future directions for MR

MR is commonly performed to investigate the causality of established observational associations. However, a “hypothesis-free” approach can also be adopted to uncover novel causal relationships. This involves performing MR on a phenome-wide scale, known as MR-pheWAS, where the effect of a single exposure on multiple outcomes is evaluated. This has been shown by
• Haycock P.C.
• Burgess S.
• Nounu A.
• Zheng J.
• Okoli G.N.
• Bowden J.
• et al.
Association between telomere length and risk of cancer and non-neoplastic diseases.
, who found that telomere length increased the risk of several cancers and reduced the risk of nonneoplastic diseases.
MR can also be applied to investigate the causal role of molecular traits, such as gene expression, methylation, and protein biomarkers, on disease. In doing so, genetic variants associated with expression (expression quantitative trait loci), methylation (methylation quantitative trait loci), or plasma protein levels (protein quantitative trait loci) are used as genetic instruments for the exposure and can provide insight into the causal pathways that underlie disease. This has been shown for AD, where MR analysis with protein quantitative trait loci gave evidence that IL1RL2 and IL18R1 are causal proteins for AD risk (
• Sun B.B.
• Maranville J.C.
• Peters J.E.
• Stacey D.
• Staley J.R.
• Blackshaw J.
• et al.
Genomic atlas of the human plasma proteome.
).
Many MR studies are performed in cohorts with limited ethnic variation. As shown by
• Ogawa K.
• Stuart P.E.
• Tsoi L.C.
• Suzuki K.
• Nair R.P.
• Mochizuki H.
• et al.
A trans-ethnic Mendelian randomization study identifies causality of obesity on risk of psoriasis [e-pub ahead print].
, transethnic MR studies can make the causal estimate more robust to confounding by population stratification and more generalizable to broader ethnic backgrounds (
• Ogawa K.
• Stuart P.E.
• Tsoi L.C.
• Suzuki K.
• Nair R.P.
• Mochizuki H.
• et al.
A trans-ethnic Mendelian randomization study identifies causality of obesity on risk of psoriasis [e-pub ahead print].
).
We also expect that MR methods will begin to be applied to outcomes of disease progression (as opposed to onset), to enable them to be more informative for the treatment of patients (
• Paternoster L.
• Tilling K.
• Davey Smith G.
Genetic epidemiology and Mendelian randomization for informing disease therapeutics: conceptual and methodological challenges.
). Such studies have begun to emerge in other disease areas, such as Parkinson disease (
• Simon K.C.
• Eberly S.
• Gao X.
• Oakes D.
• Tanner C.M.
• Shoulson I.
• et al.
Mendelian randomization of serum urate and Parkinson disease progression.
), and could potentially uncover novel therapeutic targets or drug repurposing opportunities in dermatology.

## Conclusion

MR has proven to be a robust statistical method to infer causal relationships in observational studies. In this review, we have presented strategies for performing MR, as well as the limitations and promising extensions of this method. As large GWAS summary statistics and open-access data sets become increasingly available and additional methods continue to be developed, the potential for MR analysis to produce further evidence of causality for dermatological traits will increase. This, in turn, will aid in the understanding of underlying mechanisms of disease and inform disease prevention and treatment.

## Conflict of Interest

LP has received personal fees from Merck for Scientific Input Engagement related to MR methodology.

### Multiple Choice Questions

• 1.
Which of the following is a limitation of observational studies that can be addressed with MR?
• A.
Publication bias
• B.
Selection bias
• C.
Confounding
• D.
• 2.
Which of the following is NOT an assumption for a valid MR instrument?
• A.
The instrument must be truly associated with the exposure and the outcome.
• B.
The instrument must be truly associated with the exposure.
• C.
The instrument must not be associated with confounders of the exposure-outcome relationship.
• D.
The instrument must affect only the outcome via the exposure.
• 3.
Which of the following can be used to uncover the direction of a causal relationship?
• A.
Two-sample MR
• B.
Observational analysis
• C.
One-sample MR
• D.
Bidirectional MR
• 4.
Which of the following can be used to address pleiotropy in MR?
• A.
Wald ratio method
• B.
MR-Egger regression
• C.
Inverse-variance weighted estimator
• D.
Two-stage least squares
• 5.
Which of the following statements is FALSE?
• A.
MR can be performed in a hypothesis-free manner.
• B.
MR estimates represent the effect of long-term exposures.
• C.
Pleiotropic genetic instruments cannot be included in MR analyses.
• D.
MR can be used to investigate the causal role of molecular phenotypes.

## Acknowledgments

AB-A is funded by a grant awarded by the British Skin Foundation (8010 Innovative Project), awarded to LP. AB-A and LP work in a research unit funded by the UK Medical Research Council (MC_UU_00011/1).

• 1.
Which of the following are limitations of observational studies that can be addressed with MR?
• Traditional observational studies are limited by confounding, reverse causation, and measurement error. MR can be used to evaluate causality in observational studies while avoiding these limitations.
• 2.
Which of the following is NOT an assumption for a valid MR instrument?
• Correct answer: A. The instrument must be truly associated with the exposure and the outcome.
• A valid MR instrument must satisfy three core assumptions. The instrument must be truly associated with the exposure, must not be associated with confounders of the exposure-outcome relationship, and must affect only the outcome via the exposure and not through an alternative pathway.
• 3.
Which of the following can be used to uncover the direction of a causal relationship?
• Correct answer: D. Bidirectional MR
• Bidirectional MR involves investigating the causal effect of an exposure on an outcome, as well as evaluating the effect in the reverse direction of the outcome on the exposure. In doing so, the direction of the causal relationship can be determined.
• 4.
Which of the following can be used to address pleiotropy in MR?
• Correct answer: B. MR-Egger regression
• MR-Egger regression can be performed to detect the presence of pleiotropy and also to obtain a causal estimate that is robust to pleiotropy.
• 5.
Which of the following statements is FALSE?
• Correct answer: C. Pleiotropic genetic instruments cannot be included in MR analyses.
• MR can be performed on a phenome-wide scale to investigate the causal effect of a single exposure on multiple outcomes with MR-pheWAS. MR estimates also represent the effect of long-term exposures rather than short-term interventions. In addition, MR can be extended to investigate the causal effect of molecular traits on disease, where expression quantitative trait loci, methylation quantitative trait loci, or protein quantitative trait loci are used as genetic instruments. Genetic instruments that are pleiotropic are not valid for MR analysis; however, MR methods have been developed to address pleiotropy which allows for both unpleiotropic and pleiotropic variants to be included. These include MR-Egger regression, weighted-median analysis, and the mode-based estimate.

## Supplementary Material

• Teaching Slides

## References

• Armstrong A.W.
• Harskamp C.T.
• Dhillon J.S.
• Armstrong E.J.
• Armstrong A.W.
Psoriasis and smoking: a systematic review and meta-analysis funding sources.
Br J Dermatol. 2014; 170: 304-314
• Budu-Aggrey A.
• Brumpton B.
• Tyrrell J.
• Watkins S.
• Modalsli E.H.
• Celis-Morales C.
• et al.
Evidence of a causal relationship between body mass index and psoriasis: a Mendelian randomization study.
PLoS Med. 2019; 16: e1002739
• Burgess S.
• Butterworth A.
• Thompson S.G.
Mendelian randomization analysis with multiple genetic variants using summarized data.
Genet Epidemiol. 2013; 37: 658-665
• Burgess S.
• Small D.S.
• Thompson S.G.
A review of instrumental variable estimators for Mendelian randomization.
Stat Methods Med Res. 2017; 26: 2333-2355
1. Burgess S, Thompson SG. Mendelian randomization: methods for using genetic variants in causal estimation, https://books.google.co.uk/books?id=WYSbBgAAQBAJ&printsec=frontcover&dq=Mendelian+Randomization+Methods+for+using+Genetic+variants+in+Casual+Estimation+CRC+Press&hl=en&sa=X&ved=0ahUKEwjbwuT9k5reAhUqBMAKHYvmBlMQ6AEIMjAB#v=onepage&q&f=false; 2015 (accessed 22 Oct 2018).

• Haycock P.C.
• Burgess S.
• Nounu A.
• Zheng J.
• Okoli G.N.
• Bowden J.
• et al.
Association between telomere length and risk of cancer and non-neoplastic diseases.
JAMA Oncol. 2017; 3: 636
• Hemani G.
• Zheng J.
• Elsworth B.
• Haberland V.
• Baird D.
• et al.
The MR-base platform supports systematic causal inference across the human phenome.
eLife. 2018; 7: e34408
• Lawlor D.A.
• Harbord R.M.
• Sterne J.A.C.
• Timpson N.
• Davey Smith G.
Mendelian randomization: using genes as instruments for making causal inferences in epidemiology.
Stat Med. 2008; 27: 1133-1163
• Liyanage U.E.
• Law M.H.
• Ong J.S.
• Cust A.E.
• Mann G.J.
• Ward S.V.
• et al.
Polyunsaturated fatty acids and risk of melanoma: a Mendelian randomisation analysis.
Int J Cancer. 2018; 143: 508-514
• Maglio C.
• Peltonen M.
• Rudin A.
• Carlsson L.M.S.
Bariatric surgery and the incidence of psoriasis and psoriatic arthritis in the Swedish obese subjects study.
Obesity (Silver Spring). 2017; 25: 2068-2073
• Manousaki D.
• Paternoster L.
• Standl M.
• Moffatt M.F.
• Farrall M.
• Bouzigon E.
• et al.
Vitamin D levels and susceptibility to asthma, elevated immunoglobulin E levels, and atopic dermatitis: a Mendelian randomization study.
PLoS Med. 2017; 14: e1002294
• Noordam R.
• Hamer M.A.
• Pardo L.M.
• van der Nat T.
• Kiefte-de Jong J.C.
• Kayser M.
• et al.
No causal association between 25-hydroxyvitamin D and features of skin aging: evidence from a bidirectional Mendelian randomization study.
J Invest Dermatol. 2017; 137: 2291-2297
• O’Connor L.J.
• Price A.L.
Distinguishing genetic correlation from causation across 52 diseases and complex traits.
Nat Genet. 2018; 50: 1728-1734
• Ogawa K.
• Stuart P.E.
• Tsoi L.C.
• Suzuki K.
• Nair R.P.
• Mochizuki H.
• et al.
A trans-ethnic Mendelian randomization study identifies causality of obesity on risk of psoriasis [e-pub ahead print].
J Invest Dermatol. 2019; 139: 1397-1400
• Paternoster L.
• Tilling K.
• Davey Smith G.
Genetic epidemiology and Mendelian randomization for informing disease therapeutics: conceptual and methodological challenges.
PLoS Genet. 2017; 13: e1006944
• Simon K.C.
• Eberly S.
• Gao X.
• Oakes D.
• Tanner C.M.
• Shoulson I.
• et al.
Mendelian randomization of serum urate and Parkinson disease progression.
Ann Neurol. 2014; 76: 862-868
• Standl M.
• Tesch F.
• Baurecht H.
• Rodríguez E.
• Müller-Nurasyid M.
• Gieger C.
• et al.
Association of atopic dermatitis with cardiovascular risk factors and diseases.
J Invest Dermatol. 2017; 137: 1074-1081
• Sun B.B.
• Maranville J.C.
• Peters J.E.
• Stacey D.
• Staley J.R.
• Blackshaw J.
• et al.
Genomic atlas of the human plasma proteome.
Nature. 2018; 558: 73-79
• Tsoi L.C.
• Patrick M.T.
• Elder J.T.
Research techniques made simple: using genome-wide association studies to understand complex cutaneous disorders.
J Invest Dermatol. 2018; 138: e23-e29
• Wald A.
The fitting of straight lines if both variables are subject to error.
Ann Math Stat. 1940; 11: 284-300
• Zheng J.
• Baird D.
• Borges M.-C.
• Bowden J.
• Hemani G.
• Haycock P.
• et al.
Recent developments in mendelian randomization studies.
Curr Epidemiol Rep. 2017; 4: 330-345