Efforts to map susceptibility loci for complex human diseases frequently result in weak evidence for linkage, followed by failures to convincingly replicate. If a disease locus influences both affection status (AF) and a quantitative trait, a variant of the extreme discordant sib pair (EDSP) strategy may be used to judiciously sample families for a replication study. This approach was evaluated by conducting joint segregation and linkage analysis of four bivariate phenotypes, each comprising AF and one quantitative trait (Q2, Q3, Q4, Q5), and undertaking a genomic scan of the GAW10 Problem 2B data set. Suggestive evidence was found for linkage of AF/Q2 to marker D8G26 (lod = 1.63, P = 7.52, p = 3.05 ร 2 10 ). Sets of 23 EDSP families were selected based on Q2 values that demarcated -3 the bottom 1%, 2%, 5%, 10%, or 20% of the distribution. Highly significant evidence for linkage was found when the 1% or 2% cutoffs were used (lod > 4, P 2 > 20, p < 4 ร 10 ), but more than 1,000 families had to be screened. Significant -6 evidence for linkage (lod = 3.24, P = 14.91, p = 5.63 ร 10 ) was found in the 2 -5
EDSP sample obtained using a 5% cutoff; about 300 families had to be screened. Only suggestive evidence for linkage (lod = 1.65, P = 7.59, p = 2.93 ร 10 ) was 2 -3
found in non-EDSP families. Use of EDSP sampling variants can permit convincing replications not otherwise obtainable in the genetic analysis of complex diseases.