Model-free age-of-onset methods applied to the linkage of bipolar disorder.

We performed Haseman-Elston regression on a set of bipolar pedigrees using each of three dependent variables: a binary trait indicating disease concordance or discordance, a binary trait adjusted for age-of-onset, and the residuals from a survival analysis. The latter two methods, which both adjust for age-of-onset, gave smaller p-values when previous analyses suggested linkage between disease and marker, but not when previous analyses were not suggestive of linkage.