Research Synthesis Methodology for Normative Data and Genetic Data

Abstract

A reference interval represents the normative range for measurements from a healthy population. It can be interpreted as a prediction intervalfor a new individual from the overall population. The reference interval based on one study might not be applicable to a broader population. Meta-analysis can provide a more generalizable reference interval based on the combined population by synthesizing results from multiple studies. However, existing random effects methods may give imprecise estimates of the between-study variation with only a few studies. In addition, the normal distribution of underlying study-specific means, and equal within-study variance assumption in these methods may be inappropriate in some settings. In the first paper, we develop a mixture distribution method using the fixed effects model. It combines studies by assuming the overall population is a mixture of sub-populations comprised of individual studies. This mixture distribution method does not explicitly estimate the between-study heterogeneity, which is difficult for a random effects model with few studies. In the second paper, We propose a Bayesian nonparametric (NP) model with more flexible assumptions to extend random effects meta-analysis for estimating reference intervals. The simulation studies show the performance of the mixture distribution and NP approaches when the assumptions of normally distributed study mean and equal within-study variances do not hold. Both methods are applied to real datasets and provide more reasonable estimates for reference intervals compared with existing methods. The third paper focuses on developing a new Mendelian randomization (MR) approach, which leverages genetic data to estimate the causal effect of an exposure factor on an outcome from observational studies. We utilize genetic correlations to summarize information on a large set of genetic variants associated with the exposure factor. Our proposed two-stage random effects approach (TS-RE) can accommodate many weak and pleiotropic effects. Our approach quantifies the variation explained by all included instrumental variables instead of estimating the individual effects and thus could accommodate weak IVs. This is useful for performing MR estimation in small studies where the selection of valid IVs is unreliable and thus has a large influence on the MR estimation. Through simulation and real data analysis, we demonstrate that our approach provides a robust alternative to the existing methods.