Meta-analysis of Mendelian randomization studies incorporating all three genotypes

Multivariate Methods for Meta-Analysis of Genetic Association Studies

Multivariate meta-analysis of genetic association studies and genome-wide association studies has received a remarkable attention as it improves the precision of the analysis. Here, we review, summarize and present in a unified framework methods for multivariate meta-analysis of genetic association studies and genome-wide association studies. Starting with the statistical methods used for robust analysis and genetic model selection, we present in brief univariate methods for meta-analysis and we then scrutinize multivariate methodologies. Multivariate models of meta-analysis for a single gene-disease association studies, including models for haplotype association studies, multiple linked polymorphisms and multiple outcomes are discussed. The popular Mendelian randomization approach and special cases of meta-analysis addressing issues such as the assumption of the mode of inheritance, deviation from Hardy-Weinberg Equilibrium and gene-environment interactions are also presented. All available methods are enriched with practical applications and methodologies that could be developed in the future are discussed. Links for all available software implementing multivariate meta-analysis methods are also provided.

Insight into rheumatological cause and effect through the use of Mendelian randomization

Establishing causality of risk factors is important to determine the pathogenetic mechanisms underlying rheumatic diseases, and can facilitate the design of interventions to improve care for affected patients. The presence of unmeasured confounders, as well as reverse causation, is a challenge to the assignment of causality in observational studies. Alleles for genetic variants are randomly inherited at meiosis. Mendelian randomization analysis uses these genetic variants to test whether a particular risk factor is causal for a disease outcome. In this Review of the Mendelian randomization technique, we discuss published results and potential applications in rheumatology, as well as the general clinical utility and limitations of the approach.

The association between plasminogen activator inhibitor type 1 (PAI-1) levels, PAI-1 4G/5G polymorphism, and myocardial infarction: a Mendelian randomization meta-analysis

BACKGROUND

The circulating levels of plasminogen activator inhibitor type 1 (PAI-1) are increased in individuals carrying the 4G allele at position -675 of the PAI-1 gene. In turn, overexpression of PAI-1 has been found to affect both atheroma and thrombosis. However, the association between PAI-1 levels and the incidence of myocardial infarction (MI) is complicated by the potentially confounding effects of well-known cardiovascular risk factors. The current study tried to investigate in parallel the association of PAI-1 activity with the PAI-1 4G/5G polymorphism, with MI, and some components of metabolic syndrome (MetS).

METHODS

Using meta-analytical Mendelian randomization approaches, genotype-disease and genotype-phenotype associations were modeled simultaneously.

RESULTS

According to an additive model of inheritance and the Mendelian randomization approach, the MI-related odd ratio for individuals carrying the 4G allele was 1.088 with 95% confidence interval (CI) 1.007, 1.175. Moreover, the 4G carriers had, on average, higher PAI-1 activity than 5G carriers by 1.136 units (95% CI 0.738, 1.533). The meta-regression analyses showed that the levels of triglycerides (p=0.005), cholesterol (p=0.037) and PAI-1 (p=0.021) in controls were associated with the MI risk conferred by the 4G carriers.

CONCLUSIONS

The Mendelian randomization meta-analysis confirmed previous knowledge that the PAI-1 4G allele slightly increases the risk for MI. In addition, it supports the notion that PAI-1 activity and established cardiovascular determinants, such as cholesterol and triglyceride levels, could lie in the etiological pathway from PAI-1 4G allele to the occurrence of MI. Further research is warranted to elucidate these interactions.

A selective review of the first 20 years of instrumental variables models in health-services research and medicine

BACKGROUND

The method of instrumental variables (IV) is useful for estimating causal effects. Intuitively, it exploits exogenous variation in the treatment, sometimes called natural experiments or instruments. This study reviews the literature in health-services research and medical research that applies the method of instrumental variables, documents trends in its use, and offers examples of various types of instruments.

METHODS

A literature search of the PubMed and EconLit research databases for English-language journal articles published after 1990 yielded a total of 522 original research articles. Citations counts for each article were derived from the Web of Science. A selective review was conducted, with articles prioritized based on number of citations, validity and power of the instrument, and type of instrument.

RESULTS

The average annual number of papers in health services research and medical research that apply the method of instrumental variables rose from 1.2 in 1991-1995 to 41.8 in 2006-2010. Commonly-used instruments (natural experiments) in health and medicine are relative distance to a medical care provider offering the treatment and the medical care provider's historic tendency to administer the treatment. Less common but still noteworthy instruments include randomization of treatment for reasons other than research, randomized encouragement to undertake the treatment, day of week of admission as an instrument for waiting time for surgery, and genes as an instrument for whether the respondent has a heritable condition.

CONCLUSION

The use of the method of IV has increased dramatically in the past 20 years, and a wide range of instruments have been used. Applications of the method of IV have in several cases upended conventional wisdom that was based on correlations and led to important insights about health and healthcare. Future research should pursue new applications of existing instruments and search for new instruments that are powerful and valid.

Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators

Mendelian randomization (MR) is a method for estimating the causal relationship between an exposure and an outcome using a genetic factor as an instrumental variable (IV) for the exposure. In the traditional MR setting, data on the IV, exposure, and outcome are available for all participants. However, obtaining complete exposure data may be difficult in some settings, due to high measurement costs or lack of appropriate biospecimens. We used simulated data sets to assess statistical power and bias for MR when exposure data are available for a subset (or an independent set) of participants. We show that obtaining exposure data for a subset of participants is a cost-efficient strategy, often having negligible effects on power in comparison with a traditional complete-data analysis. The size of the subset needed to achieve maximum power depends on IV strength, and maximum power is approximately equal to the power of traditional IV estimators. Weak IVs are shown to lead to bias towards the null when the subsample is small and towards the confounded association when the subset is relatively large. Various approaches for confidence interval calculation are considered. These results have important implications for reducing the costs and increasing the feasibility of MR studies.

Severity of bias of a simple estimator of the causal odds ratio in Mendelian randomization studies

Mendelian randomization studies estimate causal effects using genetic variants as instruments. Instrumental variable methods are straightforward for linear models, but epidemiologists often use odds ratios to quantify effects. Also, odds ratios are often the quantities reported in meta-analyses. Many applications of Mendelian randomization dichotomize genotype and estimate the population causal log odds ratio for unit increase in exposure by dividing the genotype-disease log odds ratio by the difference in mean exposure between genotypes. This 'Wald-type' estimator is biased even in large samples, but whether the magnitude of bias is of practical importance is unclear. We study the large-sample bias of this estimator in a simple model with a continuous normally distributed exposure, a single unobserved confounder that is not an effect modifier, and interpretable parameters. We focus on parameter values that reflect scenarios in which we apply Mendelian randomization, including realistic values for the degree of confounding and strength of the causal effect. We evaluate this estimator and the causal odds ratio using numerical integration and obtain approximate analytic expressions to check results and gain insight. A small simulation study examines finite sample bias and mild violations of the normality assumption. For our simple data-generating model, we find that the Wald estimator is asymptotically biased with a bias of around 10% in fairly typical Mendelian randomization scenarios but which can be larger in more extreme situations. Recently developed methods such as structural mean models require fewer untestable assumptions and we recommend their use when the individual-level data they require are available. The Wald-type estimator may retain a role as an approximate method for meta-analysis based on summary data.

Bayesian methods for meta-analysis of causal relationships estimated using genetic instrumental variables

Genetic markers can be used as instrumental variables, in an analogous way to randomization in a clinical trial, to estimate the causal relationship between a phenotype and an outcome variable. Our purpose is to extend the existing methods for such Mendelian randomization studies to the context of multiple genetic markers measured in multiple studies, based on the analysis of individual participant data. First, for a single genetic marker in one study, we show that the usual ratio of coefficients approach can be reformulated as a regression with heterogeneous error in the explanatory variable. This can be implemented using a Bayesian approach, which is next extended to include multiple genetic markers. We then propose a hierarchical model for undertaking a meta-analysis of multiple studies, in which it is not necessary that the same genetic markers are measured in each study. This provides an overall estimate of the causal relationship between the phenotype and the outcome, and an assessment of its heterogeneity across studies. As an example, we estimate the causal relationship of blood concentrations of C-reactive protein on fibrinogen levels using data from 11 studies. These methods provide a flexible framework for efficient estimation of causal relationships derived from multiple studies. Issues discussed include weak instrument bias, analysis of binary outcome data such as disease risk, missing genetic data, and the use of haplotypes.

TP53 codon 72 polymorphism and cervical cancer: a pooled analysis of individual data from 49 studies

BACKGROUND

Cervical cancer is caused primarily by human papillomaviruses (HPV). The polymorphism rs1042522 at codon 72 of the TP53 tumour-suppressor gene has been investigated as a genetic cofactor. More than 80 studies were done between 1998 and 2006, after it was initially reported that women who are homozygous for the arginine allele had a risk for cervical cancer seven times higher than women who were heterozygous for the allele. However, results have been inconsistent. Here we analyse pooled data from 49 studies to determine whether there is an association between TP53 codon 72 polymorphism and cervical cancer.

METHODS

Individual data on 7946 cases and 7888 controls from 49 different studies worldwide were reanalysed. Odds ratios (OR) were estimated using logistic regression, stratifying by study and ethnic origin. Subgroup analyses were done for infection with HPV, ethnic origin, Hardy-Weinberg equilibrium, study quality, and the material used to determine TP53 genotype.

FINDINGS

The pooled estimates (OR) for invasive cervical cancer were 1.22 (95% CI 1.08-1.39) for arginine homozygotes compared with heterozygotes, and 1.13 (0.94-1.35) for arginine homozygotes versus proline homozygotes. Subgroup analyses showed significant excess risks only in studies where controls were not in Hardy-Weinberg equilibrium (1.71 [1.21-2.42] for arginine homozygotes compared with heterozygotes), in non-epidemiological studies (1.35 [1.15-1.58] for arginine homozygotes compared with heterozygotes), and in studies where TP53 genotype was determined from tumour tissue (1.39 [1.13-1.73] for arginine homozygotes compared with heterozygotes). Null results were noted in studies with sound epidemiological design and conduct (1.06 [0.87-1.29] for arginine homozygotes compared with heterozygotes), and studies in which TP53 genotype was determined from white blood cells (1.06 [0.87-1.29] for arginine homozygotes compared with heterozygotes).

INTERPRETATION

Subgroup analyses indicated that excess risks were most likely not due to clinical or biological factors, but to errors in study methods. No association was found between cervical cancer and TP53 codon 72 polymorphism when the analysis was restricted to methodologically sound studies.

FUNDING

German Research Foundation (DFG).