Working within a survival analysis framework, a model was developed to investigate genetic and environmental influences on the time to onset of developmental milestones and multifactorial threshold traits. Briefly, the accelerated failure time model, a regression model commonly used for the analysis of failure time data with measured covariates, was modified to include latent sources of genetic and environmental variation. The gamma distribution was chosen as the baseline distribution for this regression model since it provides a flexible density function over the positive range of values and is defined by parameters which are heuristically appealing.; For developmental milestones, both multifactorial and major locus gamma accelerated failure time models were considered. In each, a single latent aging effect, determined by genetic and, in the multifactorial case, environmental effects, was specified. For threshold traits, there were two sources of latent variation: an aging effect and an individual's inherited liability to the trait. It was shown that estimating the contribution of each of these sources to variation in the time to onset of threshold traits is a critical step in determining risks to relatives conditional upon a proband's age of onset.; Using simulated data sets of monozygotic and dizygotic twin pairs and nuclear families, maximum likelihood parameter estimation was carried out for the multifactorial and major locus models. For all models, several different orders of the gamma distribution were considered. Results indicate that the models are tractable, although the computing time required for the multifactorial threshold case may prohibit extensive hypotheses testing. In addition, the results suggest that the power to reject wrong gamma models and wrong genetic models depends upon the order of the true gamma process.; As an application of the multifactorial gamma accelerated failure time model, the age at menarche in 379 pairs of Australian twins was considered. Results indicate that there is substantial population heterogeneity in the latent aging variable, and 95% of this population variation is due to additive gene action.; Suggestions are made for future work in integrating survival analysis and quantitative genetic methods.
展开▼
