UBC Theses and Dissertations
Empiric risk estimation in Alzheimer disease Irwin, Mark Edward
Alzheimer disease is believed to be the most common cause of dementia. The main cause is presently unknown, with genetic and environmental factors suggested. It appears that 10-15% of Alzheimer disease is due to an autosomal dominant gene and it has been hypothesized that this is the cause for all Alzheimer's. Alzheimer's variable age of onset makes it more difficult to determine the validity of this and other genetic models. Empiric risk estimates for Alzheimer disease in relatives can used to test the plausibility of various genetic models. Three types of procedures for estimating the risk of Alzheimer disease are discussed. Three nonparametric, product-limit type estimators (Kaplan-Meier, Life-table, Weinberg) for age-specific risks are discussed first. Then three estimators for lifetime risk of disease using a predetermined weight function believed to approximate the true age of onset distribution (Stromgren, Modified Stromgren, maximum likelihood) are compared. Finally a maximum likelihood procedure to estimate lifetime risk and the age of onset distribution is presented. The properties of these estimators are discussed using a data set from the Alzheimer Clinic, University Hospital - U.B.C. Site. In addition, the results of a Monte-Carlo study of the maximum likelihood procedure for estimating the lifetime risk and age of onset distribution are discussed. The most useful of these estimators appear to be the Kaplan-Meier and the life-table estimators for age-specific risks and the maximum likelihood procedure for estimating lifetime risk and the age of onset distribution. The Weinberg estimator appears to be biased and the fixed age of onset estimators for lifetime risk appear to be too dependent on the choice of the age of onset distribution to be useful in general.
Item Citations and Data