Jeanine J Houwing-Duistermaat,
Dept. of Medical Statistics and Bioinformatics, LUMC, Leiden.

Title: "A score statistic for linkage analysis of censored data"

Abstract:
AIM: To derive a score statistic for linkage analysis of censored outcomes such
as life span. Several linkage methods exist that model censored outcomes by
extending the Cox model with a frailty. Normal and gamma distributions have
been proposed for the frailty distribution. Because likelihood ratio statistics
are used for testing, the obtained results may be affected by the assumed
distribution. A robust score statistic is needed.

LINKAGE ANALYSIS: The goal of linkage is to identify chromosomal regions which 
may harbour a gene involved in the aetiology of the studied trait. One commonly 
used approach is to type genetic markers along the genome in affected relative 
pairs. Then it is tested whether these relative pairs share more alleles 
identical by descent (IBD) than expected under Mendelian transmission. Regions 
which show excess IBD sharing may harbour genes of interest. The standard score 
statistic does not take into account additional information such as current age.

DATA. A candidate region spanning 20 cM is typed in 160 nonagenarian sibling 
pairs from the Leiden Longevity Study. There is no evidence for linkage when the
whole group is analyzed, but when subsets of older pairs are analyzed, the 
evidence for linkage increases. Selection of subsets is quite arbitrary and a 
more sound method is needed. 

In this presentation, linkage analysis will be briefly introduced. Based on a 
gamma frailty model, a score statistic will be derived. Finally the statistic 
will be illustrated using data from the Leiden Longevity Study.