Case-sibling gene-association studies for diseases with variable age at onset

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Family-based genetic studies are crucial for identifying associations between genetic loci and phenotypic traits.
• Traditional binary disease models used in family studies can be inefficient or biased for complex diseases with variable age at onset.
• Population stratification is a key confounder that family-based studies help to mitigate.

Purpose of the Study:

• To review and assess analytical methods for censored age-at-onset data in family studies.
• To compare the performance of stratified Cox regression, genotype-decomposition regression, and a novel retrospective likelihood method.
• To provide guidance on appropriate methods based on disease prevalence and analytic goals.

Main Methods:

• Review of existing methods: stratified Cox regression and genotype-decomposition regression.
• Introduction of a new retrospective likelihood method for censored age-at-onset data.
• Performance assessment via simulation studies, evaluating type I error rates and confidence interval coverage.

Main Results:

• Stratified Cox regression and the retrospective likelihood method demonstrated appropriate type I error rates.
• Genotype-decomposition regression was often anti-conservative.
• Stratified Cox regression confidence intervals were anti-conservative for rare diseases; the retrospective likelihood showed better efficiency and coverage, particularly for rare diseases.

Conclusions:

• Stratified Cox regression is suitable for common diseases when estimating genotype relative risks.
• The retrospective likelihood method is more appropriate for rare diseases due to its efficiency and accurate confidence interval coverage.
• The choice of method depends on disease rarity and the primary analytical objective in family-based genetic studies.