Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors.

Burgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG, EPIC- InterAct Consortium. · 2015

Read Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors. by Burgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG, EPIC- InterAct Consortium. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4516908.txt.

Introduction

Stephen Burgess and four co-authors set out a recipe for causal studies that use only published numbers, in the European Journal of Epidemiology in 2015. Mendelian randomization uses gene variants as natural tests of whether a risk factor truly causes a disease. Big gene studies now publish summary results freely, but hunting down each person's raw data can be hard.

Their plan takes gene links to the risk factor and to the outcome from two separate sources. New math in the paper merges the links of many gene variants, related or not, into one causal estimate. A worked example asks whether blood calcium changes fasting glucose. The authors note the trade-off: the method is fast and precise, yet checking the assumptions behind causal claims gets harder with summary data alone.

Catalog record

CREATORBurgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG, EPIC- InterAct Consortium.
YEAR2015
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4516908

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors By Stephen Burgess, Robert A. Scott, Nicholas J. Timpson, George Davey Smith, Simon G. Thompson Published in: European Journal of Epidemiology DOI: 10.1007/s10654-015-0011-z ABSTRACT Finding individual-level data for adequately-powered Mendelian randomization analyses may be problematic. As publicly-available summarized data on genetic associations with disease outcomes from large consortia are becoming more abundant, use of published data is an attractive analysis strategy for obtaining precise estimates of the causal effects of risk factors on outcomes. We detail the necessary steps for conducting Mendelian randomization investigations using published data, and present novel statistical methods for combining data on the associations of multiple (correlated or uncorrelated) genetic variants with the risk factor and outcome into a single causal effect estimate. A two-sample analysis strategy may be employed, in which evidence on the gene-risk factor and gene-outcome associations are taken from different data sources. These approaches allow the efficient identification of risk factors that are suitable targets for clinical intervention from published data, although the ability to assess the assumptions necessary for causal inference is diminished. Methods and guidance are illustrated using the example of the causal effect of serum calcium levels on fasting glucose concentrations. The estimated causal effect of a 1 standard deviation (0.13 mmol/L) increase in calcium levels on fasting glucose (mM) using a single lead variant from the CASR gene region is 0.044 (95 % credible interval −0.002, 0.100). In contrast, using our method to account for the correlation between variants, the corresponding estimate using 17 genetic variants is 0.022 (95 % credible interval 0.009, 0.035), a more clearly positive causal effect. Electronic supplementary material The online version of this article (doi:10.1007/s10654-015-0011-z) contains supplementary material, which is available to authorized users. INTRODUCTION Mendelian randomization is a technique which uses genetic variants to assess whether a risk factor, such as a biomarker, has a causal effect on a disease outcome in a non-experimental (observational) setting [1, 2]. We assume that the chosen genetic variants are associated with the risk factor, but not associated with any confounder of the risk factor–outcome relationship, nor associated with the outcome via any pathway other than that through the risk factor of interest [3]. These three assumptions form the definition of an instrumental variable [4]. A variant satisfying these assumptions divides a study population into subgroups which are analogous to treatment arms in a randomized controlled trial, in that they differ systematically with respect to the risk factor of interest, but not with respect to confounding factors [5]. An association between the genetic variant and the outcome therefore implies that the risk factor has a causal effect on the outcome. Mendelian randomization is a valuable approach for identifying risk factors as potential targets for clinical or behavioural intervention [6]. Evidence from Mendelian randomization has been used to prioritize investigation of certain biomarkers as causal risk factors for cardiovascular disease: for example lipoprotein(a) [7], and interleukin-6 receptor [8]; and to de-prioritize others: fibrinogen [9], C-reactive protein (CRP) [10], and uric acid [11]. However, it may be hard to find a suitable study population with sufficient data on the genetic variants, and both the risk factor and outcome of interest. As many genetic variants only explain a small proportion of the variation in the risk factor, large sample sizes (in some cases comprising tens of thousands of individuals [12]) may be required for adequately-powered Mendelian randomization investigations. Several consortia with large numbers of participants, such as CARDIoGRAMplusC4D for coronary artery disease [13] and DIAGRAM for type 2 diabetes [14], have published data on the association of catalogues of genetic variants with either risk factors or disease status (a list of consortia is given in Web Table A1). These provide precise estimates of genetic associations which can be used to obtain causal estimates based on Mendelian randomization in a fast and cost-effective way. In this paper, we provide a blueprint for this approach. METHODS The steps involved in a Mendelian randomization investigation are: (1) specification of the dataset(s) for analysis, (2) search for candidate instrumental variables, (3) validation of the instrumental variable assumptions, (4) estimation of the causal effect (if appropriate), (5) supplementary and sensitivity analyses. A schematic diagram of the relevant components in a Mendelian randomization analysis is given in Fig. 1. We proceed to outline each of these steps.Fig. 1Schematic diagram outlining the Mendelian randomization approach Specification of the dataset(s) for analysis Traditionally, Mendelian randomization analyses have been performed on a single study or studies containing data on genetic variants, and both the risk factor and outcome of interest. The main advantages of using published data rather than individual-level data are their size and scope. The associations of these variants with the risk factor and outcome in large consortia are likely to be more precisely estimated than in a single study. However, it is unlikely that published data on the genetic associations with the risk factor, with the outcome, and with potential confounders are available on the same set of studies. Two-sample Mendelian randomization is a strategy in which evidence on the associations of genetic variants with the risk factor and with the outcome comes from non-overlapping data sources [15]. The limiting factor for the power of a Mendelian randomization analysis using a given set of genetic variants is the precision in the estimate of the genetic association with the outcome, as this association is typically much weaker than the genetic association with the risk factor. Published data on genetic associations with the outcome can therefore be combined with individual-level data from a cross-sectional study on genetic variants and the risk factor to obtain precise Mendelian randomization estimates. If the study used to estimate genetic associations with the risk factor is included in the estimate of the genetic association with the outcome, then this is a subsample rather than a two-sample analysis strategy. Alternatively, published data can be used in all aspects of the analysis. In this case, the two published data sources may overlap (for example, they both constitute meta-analyses and some studies are included in both sources). In any case, it is likely that the sets of individuals used in the gene-risk factor and gene-outcome arms of the analysis will not be identical. An important assumption to ensure the validity of the analysis is that the two sets represent samples taken from the same underlying population. If this is not the case, then inferences may be misleading, as the association of the genetic variants with the risk factor may not be replicated in the set of individuals in which the association with the outcome is estimated, or a variant may not be a valid instrumental variable in both sets. Search for candidate instrumental variables Genetic variants are sought which are associated with the risk factor of interest. These can be obtained from available individual-level data or from the catalogues of genetic variants identified by genome-wide association studies (GWAS) that have been compiled [16]. It is important that estimates of both the gene-risk factor and the gene-outcome associations are available for each of these variants, or for proxies of the variants (a proxy is a variant in complete or near complete

Read the complete text