Recent reports of a linkage trend in 6p24-22 for schizophrenia (SZ), in different samples, were tempered by the concurrent evidence of negative reports in other samples. In the studies showing positive results, different definitions of affection and a wide spectrum of diagnoses were used. Our objectives were not only to test for linkage at 6p24-22 in the Eastern Quebec population, but also to test whether this putative vulnerability locus was either selectively linked to schizophrenia (SZ), or to bipolar disorder (BP), or to both major psychoses. Parametric and nonparametric linkage analyses with 12 microsatellite markers in 6p24-p22 were performed on a sample of 18 large multigenerational pedigrees (N = 354) either affected by SZ, or by BP, or equally affected by both major psychoses (i.e., mixed pedigrees). Three affection definitions were usually tested in our program: one on schizophrenia (SZ), one on bipolar disorder (BP), and one that comprised SZ and BP under the hypothesis of a susceptibility locus common to both in major psychoses (common locus, CL). The results of parametric analyses did not support a major gene hypothesis. However, in one large mixed pedigree (#151), we observed with the common locus phenotype (CL) lod scores of 2.49 and 2.15, respectively, at the D6S296 and D6S277 loci under a dominant model. Our data suggest the presence of a potential vulnerability locus at 6p24-22 that could be related to both schizophrenia and bipolar affective disorder. These results may be seen as congruent with former studies that used schizoaffective as well as schizophrenia diagnoses as entry criteria for the affected families, and used an affection definition that comprised affective psychoses as well as schizophrenia.

We have performed initial nonparametric sib‐pair genome scans in the early (N = 52) and late (N = 53) onset subgroups of the COGA pedigrees, stratified near the median value of pedigree mean age of onset for ALDX1 diagnosis of alcoholism. Because the early group contained a higher proportion of smokers, traits of alcoholism, smoking, and addiction (defined as either alcoholism or smoking) were examined. Subgroups and phenotypic definitions influenced initial linkage results, corrected for the number of analyzed traits. Evidence for linkage to the ALDX1 alcoholism phenotype at the ADH3 functional candidate gene was increased in the late onset subgroup (Bonferroni corrected significance level < 0.002), as compared with the unstratified sample that replicated COGA linkage obtained in the same analysis; there was no evidence for linkage at this locus in the early onset subgroup. The theoretical implication of this result is that the loss of power due to contracting sample size through stratification may in some cases be more than offset by extraction of a more homogenous subgroup from the etiologically complex trait.

We believed that subtyping alcoholism might be an efficient strategy for mapping susceptibility genes. Cluster analysis is one of the possible statistical techniques for such a purpose. We required that, ideally, the variables to be used in cluster analysis should be: 1) related to alcoholism, 2) related to the severity of alcoholism, and 3) familial, i.e., correlated within families. Only three variables met all three conditions. Those included age of onset of ALDX1, smoking, and TPQ-HA. A global score of symptoms of alcoholism was systematically introduced as one of the variables composing a subset for cluster analysis, although this score did not show any familial aggregation. Our strategy led to a strong evidence of linkage at D15S230 in only 20 families whose members are mainly characterized by heavy smoking. 1999 Wiley-Liss, Inc.