Random-effects Cox proportional hazards model: general variance components methods for time-to-event data.

Area of Science:

• Biostatistics
• Genetic Epidemiology
• Statistical Genetics

Background:

• Proportional hazards regression models are standard for time-to-event data.
• Genetic diseases often show variation in age at onset, requiring analysis of correlated observations.
• Existing variance components methods are not readily available for survival data.

Purpose of the Study:

• To outline a method for variance component analyses under general random effects proportional hazards models.
• To enable survival analyses for correlated time-to-event data in genetic studies.
• To provide a computational method for analyzing age-at-onset data analogous to quantitative trait analyses.

Main Methods:

• Developed a variance component analysis method for random effects proportional hazards models.
• Utilized a Laplace approximation for computational feasibility with correlated time-to-event data.
• Applied correlated frailty models for genetic and structured random effects analyses.

Main Results:

• Demonstrated significant, moderate heritability of breast cancer age-at-onset in a familial cohort.
• Successfully performed variance component linkage analyses on simulated data (GAW12).
• Identified the locations of simulated disease genes impacting age-at-onset.

Conclusions:

• The proposed method makes variance component analyses computationally feasible for time-to-event endpoints, even in large datasets.
• This approach allows for genetic analyses of age-at-onset data analogous to quantitative trait analyses.
• The method is effective for both heritability estimation and linkage analysis in complex genetic diseases.