Because age at onset is a truncated quantitative trait – censored for those individuals for whom age at onset was not observed, we used semiparametric proportional hazards model as implemented in SAS [6] software to estimate the survival distribution function after adjusting for significant covariates – age, sex, maximum number of drinks in 24-hour period, and smoking status. Our assumption for using this truncated quantitative trait model was that all individuals would eventually get the disease if they lived or if the follow-up was long enough.
