To the Editor: In his letter Dr. Harrison points out the discrepancies between our findings—which suggested a relationship between variation in the 5-HT_2A receptor gene (5HTR2A) and the level of [³H]ketanserin 5-HT_2A receptor binding in the prefrontal cortex—and two other studies by authors who failed to observe such a relationship. A possible explanation for this inconsistency may be related to the characteristics of the patients included in our study, who were of French Canadian origin.

The existence of a founder effect in the French Canadian population has been confirmed by many studies (1). The French Canadian population currently living in Quebec is descended from approximately 7,000 founding individuals who came to "Nouvelle France" before 1760 (2). Because this is a relatively young population (approximately 12 generations), the background linkage disequilibrium is considerable. For instance, studies of unrelated subjects who carry a mutation for oculopharyngeal muscular dystrophy suggest that for a 5-centimorgan interval, 75% of the subjects of French Canadian origin share all alleles (3). These findings have been replicated in analyses conducted with different diseases (4).

Dr. Harrison reminds the reader that the polymorphisms used to investigate genetic variation in our study (and others) are of unknown function. Indeed, the T102C marker is a silent T→C base substitution (5), and the variants at the A-1438G locus present similar basal promoter activity (6). Therefore, it is possible that our positive results with the 5HTR2A polymorphisms investigated may in fact be a consequence of underlying linkage disequilibrium between these markers and a functional genetic variant located in a different coding or regulatory region of the 5HTR2A gene. Furthermore, because of the extent of the linkage disequilibrium observed in French Canadians, this functional variant could be located relatively far from the tested polymorphisms (regulatory sequences and enhancers). If this is the case, the other studies may have failed to detect such an effect because, although testing larger samples, they investigated patients selected from nonisolated, panmictic populations, in which the power to detect associations due to linkage disequilibrium is considerably reduced. Certainly, other explanations may be also proposed for the reported inconsistency. Among these are type I error and the different analytical procedures employed. To better understand these issues, we are in the process of collecting a larger and independent sample of French Canadian suicide cases and will attempt to replicate our previous findings.