Constrained instruments and their application to Mendelian randomization with pleiotropy.

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Mendelian randomization (MR) infers causal relationships using genetic variants as instrumental variables.
• A key MR assumption is that genetic variants influence the outcome solely through the phenotype of interest.
• Pleiotropy, where genetic variants affect multiple phenotypes, can violate this assumption and bias causal inference.

Purpose of the Study:

• To develop novel methods for constructing valid instrumental variables in the presence of pleiotropy.
• To enable adjusted causal effect estimation when pleiotropic phenotypes are known.
• To compare the performance of new constrained instrumental variable (CIV) methods against existing approaches.

Main Methods:

• Introduction of constrained instrumental variable (CIV) methods for instrumental variable construction.
• Development of a smoothed CIV version for approximate selection of valid genetic instruments.
• Comparative performance analysis of CIV methods and existing approaches via extensive simulations.
• Application of CIV methods to Alzheimer's Disease Neuroimaging Initiative (ADNI) data.

Main Results:

• Smoothed CIV methods provide unbiased causal effect estimates by effectively selecting valid instruments.
• CIV methods demonstrate robust performance across various types of pleiotropic violations.
• The study successfully disentangled causal relationships between biomarkers and Alzheimer's disease progression using ADNI data.

Conclusions:

• Constrained instrumental variable (CIV) methods offer a robust solution for Mendelian randomization studies affected by pleiotropy.
• These methods improve the accuracy of causal inference in genetic epidemiology.
• CIV methods are applicable to real-world datasets, such as those investigating Alzheimer's disease biomarkers.