Bias due to participant overlap in two‐sample Mendelian randomization

Abstract

Mendelian randomization analyses are often performed using summarized data. The causal estimate from a one‐sample analysis (in which data are taken from a single data source) with weak instrumental variables is biased in the direction of the observational association between the risk factor and outcome, whereas the estimate from a two‐sample analysis (in which data on the risk factor and outcome are taken from non‐overlapping datasets) is less biased and any bias is in the direction of the null. When using genetic consortia that have partially overlapping sets of participants, the direction and extent of bias are uncertain. In this paper, we perform simulation studies to investigate the magnitude of bias and Type 1 error rate inflation arising from sample overlap. We consider both a continuous outcome and a case‐control setting with a binary outcome. For a continuous outcome, bias due to sample overlap is a linear function of the proportion of overlap between the samples. So, in the case of a null causal effect, if the relative bias of the one‐sample instrumental variable estimate is 10% (corresponding to an F parameter of 10), then the relative bias with 50% sample overlap is 5%, and with 30% sample overlap is 3%. In a case‐control setting, if risk factor measurements are only included for the control participants, unbiased estimates are obtained even in a one‐sample setting. However, if risk factor data on both control and case participants are used, then bias is similar with a binary outcome as with a continuous outcome. Consortia releasing publicly available data on the associations of genetic variants with continuous risk factors should provide estimates that exclude case participants from case‐control samples.

Differential associations of depression‐related phenotypes with cardiometabolic risks: Polygenic analyses and exploring shared genetic variants and pathways

“…It is worthwhile to note that a recent study by Burgess, Davies, and Thompson (2016) As we shall see in the "Results" section, sample overlap for all except one pair of traits was non-significant as evaluated by LDSR, hence our conclusions are unlikely to be substantially affected by overlap issues.…”

Response to Hartwig and Davies

“…For example, around 71% of participants are shared between the Genetic Investigation of Anthropometric Traits (GIANT) consortium 10 and the Global Lipids Genetics Consortium (GLGC) 11 . For the IVW method, the direction of weak instrument bias varies linearly as the proportion of sample overlap increases from the two-sample setting (where bias is in the direction of the null) to the single-sample setting (where bias is in the direction of the observational association) 12 . Further work is needed to see if a similar pattern holds for the MR-Egger method.…”

Cannabis use and risk of schizophrenia: a Mendelian randomization study

“…Conventionally, when using data sets that overlap for the SNP-exposure and SNP-outcome, this can generate biased estimates and yield a spurious causal estimate (arising from correlation of the error terms of SNP-exposure and SNP-outcome). 24 However, in our case, there was only minimal overlap (o5%) between the data sets used to derive the effect estimates for SNPs with ever use of cannabis and risk of schizophrenia (Supplementary Figure S2), minimizing the possibility of weak instrument bias yielding a false positive association. In contrast to overlapping datasets where weak instrument bias can lead to a false positive result, use of non-overlapping data sets in MR can lead to a false negative association.…”