Two sample Mendelian Randomisation using an outcome from a multilevel model of disease progression.

Area of Science:

• Epidemiology
• Genetics
• Biostatistics

Background:

• Identifying causal factors for disease progression, particularly in neurodegenerative diseases, is crucial.
• Disease progression is often modeled as a linear trajectory with an intercept (initial severity) and slope (rate of change).
• Two-sample Mendelian randomization (2SMR) is a method to infer causal relationships from observational data, mitigating confounding bias.

Purpose of the Study:

• To develop and evaluate a multivariate two-sample Mendelian randomization (2SMR) approach for analyzing disease progression.
• To estimate the causal effect of an exposure on both the intercept and slope of a linear disease progression trajectory.
• To compare the performance of the multivariate 2SMR approach against a univariate 2SMR method.

Main Methods:

• A multivariate 2SMR method was developed using a multilevel model for disease progression.
• A simulation study was conducted comparing the multivariate 2SMR approach with a univariate 2SMR approach.
• The simulation involved scenarios where an exposure affected both the intercept and slope of a linearly progressing outcome.
• The methods were applied to two Parkinson's disease cohorts to assess the effect of body mass index (BMI) on disease progression.

Main Results:

• Simulation results indicated that both univariate and multivariate 2SMR approaches showed no significant evidence against non-zero bias.
• Confidence interval coverage for intercept (93.4-96.2%) and slope (94.5-96.0%) was appropriate in simulations.
• The multivariate approach provided better joint coverage for both intercept and slope effects.
• Analysis of Parkinson's cohorts found no strong evidence that BMI causally affects disease progression, though confidence intervals were wide.

Conclusions:

• The developed multivariate 2SMR approach is a viable method for estimating causal effects on disease progression trajectories.
• While simulation results were comparable to univariate methods, the multivariate approach offers improved joint estimation of progression parameters.
• Further research with larger cohorts is needed to confirm the effect of BMI on Parkinson's disease progression due to wide confidence intervals.