Powerful three-sample genome-wide design and robust statistical inference in summary-data Mendelian randomization

Abstract

Background. Mendelian randomization (MR) uses genetic variants as instrumental variables to estimate the causal effect of risk exposures in epidemiology. Two-sample summary-data MR that uses publicly available genome-wide association studies (GWAS) summary data have become a popular design in practice. With the sample size of GWAS continuing to increase, it is now possible to utilize genetic instruments that are only weakly associated with the exposure.Methods. To maximize the statistical power of MR, we propose a genome-wide design where more than a thousand genetic instruments are used. For the statistical analysis, we use an empirical partially Bayes approach where instruments are weighted according to their true strength, thus weak instruments bring less variation to the estimator. The final estimator is highly efficient in the presence of many weak genetic instruments and is robust to balanced and/or sparse pleiotropy.Results. We apply our method to estimate the causal effect of body mass index (BMI) and major blood lipids on cardiovascular disease outcomes. Compared to previous MR studies, we obtain much more precise causal effect estimates and substantially shorter confidence intervals. Some new and statistically significant findings are: the estimated causal odds ratio of BMI on ischemic stroke is 1.19 (95% CI:1.07-1.32, p-value ≤ 0.001); the estimated causal odds ratio of high-density lipoprotein cholesterol (HDL-C) on coronary artery disease (CAD) is 0.78 (95% CI 0.73-0.84, p-value ≤ 0.001). However, the estimated effect of HDL-C becomes substantially smaller and statistically non-significant when we only use the strong instruments.Conclusions. By employing a genome-wide design and robust statistical methods, the statistical power of MR studies can be greatly improved. Our empirical results suggest that, even though the relationship between HDL-C and CAD appears to be highly heterogeneous, it may be too soon to completely dismiss the HDL hypothesis. Further investigations are needed to demystify the observational and genetic associations between HDL-C and CAD.

Age at menarche and osteoporosis: A Mendelian randomization study

“…Hence, we carried out a recently proposed method called Robust Adjusted Profile Score (MR.RAPS) [34] which considers the measurement error in SNP-exposure effects and is unbiased when there are many (e.g. hundreds of) weak instruments, and is robust to systematic and idiosyncratic pleiotropy.…”

“…hundreds of) weak instruments, and is robust to systematic and idiosyncratic pleiotropy. Detailed information about this method please refer to the original paper [34]. …”

A Mendelian randomization study of the role of lipoprotein subfractions in coronary artery disease

“…Because IVW and weighted median estimates are biased towards 0 when there are weak instruments [1,6], we only use these methods with the set of SNPs that are genome-wide significant in the selection dataset. For RAPS we used the implementastion in the mr.raps package (https://github.com/qingyuanzhao/mr.raps), using the empirical partially Bayes estimator with Huber's loss function as described in [5]. RAPS does not suffer from weak instrument bias as long as the average instrument strength is not too weak [6], so we applied RAPS to both sets of SNPs.…”

“…For univariate Mendelian randomization we follow the three-sample summary-data design as described in [5]. This design requires three non-overlapping GWAS summary datasets which will be referred to as the selection, exposure, and outcome datasets.…”

“…Lipoprotein subfractions have been increasingly used in epidemiological studies and even in clinical practice to predict the risk of cardiovascular diseases (CVD) (1)(2)(3). Some studies have identified potentially novel lipoprotein subfraction predictors for CVD (2,(4)(5)(6)(7)(8) and demonstrated that the addition of lipoprotein subfraction measures can significantly improve the prediction of CVD (1,(9)(10)(11). However, observational studies of lipoprotein subfractions have provided conflicting evidence.…”