In contrast to linkage studies,  association studies aim to detect co-segregation of a marker with a specific obesity phenotype in closely matched samples of affected and non-affected individuals who need not be related.

The simplest application of association is its use in testing the hypothesis that a specific polymorphism in a candidate gene is present in obese probands at a frequency greater than would be expected by chance (using X² or similar tests) and is thus either causal or at least capable of conferring susceptibility to the development of obesity.

It is clearly essential that controls and probands are from closely matched populations so that the overall population polymorphism frequency is similar,  otherwise a positive result may occur due to population stratification.  Where there is no obvious candidate gene,  controls and obese probands may be genotyped for the presence of one or more marker alleles and differences in respective frequencies are compared to determine whether particular markers are significantly associated with the phenotype of interest.

Association studies have traditionally been used further to evaluate signals detected by linkage methods and the greater a priori likelihood of a true association confers additional statistical confidence in a positive result. As they are usually more convenient to perform and can localize genes with only moderate or small effects on phenotype,  association studies are increasingly being used as a first line of investigation, significant association suggesting that the allele either affects the phenotype of interest or at least is close enough to be in linkage disequilibrium with the allele that is actually responsible.

As with linkage studies,  there are considerable inherent problems in determining the significance of an apparently positive finding and the literature is littered with uncorroborated reports of significant association,  many of which are likely to prove spurious. Factors that have led to poor replication in association studies include latent population stratification, lack of detailed and accurate ascertainment of phenotype and lack of power to confirm or refute an association especially where there may be multiple single nucleotide polymorphisms (SNPs)  in intronic,  exonic,  regulatory and upstream regions of the locus all of which require investigation in multiple populations before disease relevance may confidently be excluded (Hirschhorn and Altshuler, 2002).

Warden CH and Fisler JS
Katsanis N, Beales PL, Woods MO