Causal inference for semi-competing risks data

Area of Science:

• Biostatistics
• Epidemiology
• Genetics

Background:

• The Apolipoprotein E epsilon4 allele (APOE) is a significant genetic risk factor for late-onset Alzheimer's disease (AD).
• Defining the causal impact of APOE on both AD onset and mortality is challenging due to the interplay between these two events.
• AD occurrence can influence the age of death, complicating direct causal inference.

Purpose of the Study:

• To propose novel statistical methods for estimating the causal effects of APOE on both Alzheimer's disease diagnosis and death.
• To address the complexities arising from the semi-competing risks nature of these dual outcomes.
• To develop a framework that accounts for the order of AD diagnosis and death.

Main Methods:

• Utilizing a semi-competing risks framework for time-to-event data analysis.
• Proposing new estimands stratified by the order of AD diagnosis and death.
• Introducing a novel assumption leveraging time-to-event data properties, offering more flexibility than traditional monotonicity assumptions.
• Developing and implementing nonparametric and semiparametric estimation methods.

Main Results:

• Derivation of results on partial identifiability of causal effects.
• Development of a sensitivity analysis approach to assess the impact of assumption violations.
• Identification of conditions under which full identification of causal effects is achievable.
• Presentation of estimation methods for right-censored semi-competing risks data.

Conclusions:

• The proposed semi-competing risks framework provides a robust method for disentangling the complex causal relationships between APOE, Alzheimer's disease, and mortality.
• The novel estimands and assumptions offer a more nuanced understanding of genetic risk factors in the presence of competing events.
• The developed statistical methods facilitate more accurate causal inference in complex epidemiological and genetic studies.