Selecting a contingency table in a population-based association study: allele frequency or positivity?

In population-based association studies, the significance of the association between a candidate gene and a disease is usually examined by a simple chi(2) test. Such studies require a 2 x 2 contingency table made up of either allele frequencies or positivities in affected and control groups. In order to investigate the influence of each 2 x 2 table on the power of the chi(2) test, P values were calculated for two penetrance models (multiplicative and additive). When the value of penetrance was small and not markedly different among genotypes, a large difference in the power of the chi(2) test was observed between the two tables. In a multiplicative model, the allele frequency table was superior to the positivity table for detecting significance. In contrast, in an additive model, the positivity table was most suitable. Selecting a contingency table was especially important for detecting true association, when the required significance level was corrected to avoid the problem of multiple hypothesis testing. The chi(2) test, therefore, should be performed for both tables to identify a susceptible gene with a low penetrance, even when no significant difference is observed in one table.