BACKGROUND:

Currently there is substantial interest in finding genetic loci linked to human obesity. However, the complexity of the genetic architecture for human body fatness and body weight make this a very challenging task. Recently, several authors have proposed the use of discordant sibling pairs to improve the power to detect linkage for obesity and other continuously distributed complex phenotypes. The use of discordant siblings pairs can markedly increase the power of statistical analyses to detect linkage. However, the use of discordant sibling pairs also presents several practical concerns. These include both the difficulty and cost of locating markedly discordant sibling pairs and the possibility that when such pairs are located, they may in fact be only half-siblings due to non-paternity. In this paper, I consider both of these issues.

RESULTS:

Quantitative estimates of the probability of finding pairs of a given degree of discordance are presented based on the assumption of a normal distribution as well as on actual data. Quantitative estimates of the odds of selecting pairs that are actually half siblings as opposed to full siblings are also presented for different degrees of discordance. Finally, percentile scores for BMI are tabulated as these are necessary for researchers to implement certain sibling selection strategies.