Exploiting collider bias to apply two-sample summary data Mendelian randomization methods to one-sample individual...

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Mendelian randomization (MR) using summary statistics is widely adopted.
• Applying these methods to individual-level data presents challenges.
• There's a need for sophisticated causal inference methods accounting for non-linearity and effect modification.

Purpose of the Study:

• To develop a general procedure for optimally applying two-sample summary data MR methods to one-sample individual-level data.
• To investigate the causal role of sleep disturbance on HbA1c levels using this novel approach.

Main Methods:

• A meta-analysis of summary data estimates intentionally contaminated by collider bias is performed.
• These estimates are used to correct standard observational associations.
• Simulations validate the method's performance against naive approaches.

Main Results:

• The developed procedure optimally applies two-sample summary data MR methods to one-sample data.
• The method successfully corrects for collider bias, improving causal inference.
• Application to UK Biobank data investigates the sleep disturbance-HbA1c relationship.

Conclusions:

• The proposed method offers a robust framework for one-sample MR analysis.
• It generalizes existing techniques for bias adjustment in genetic studies.
• This approach advances causal inference by artificially inducing and correcting for collider bias.