There are three genetic methods often used for detecting genes contributing to susceptibility or resistance to multifactorial diseases: nonparametric linkage analysis, case-control association analysis, and transmission disequilibrium test. In this review, we present the theoretical basis that the case-control association study has the highest power of detecting disease genes if there is no population stratification between patients and controls. Taking advantage of the high power, we have carried out extensive case-control association analyses of candidate genes for the search of susceptibility genes to rheumatic diseases in the Japanese as well as in some other populations. Several new associations have been disclosed, including those of TNFR2, FCGR2B, and CD19 gene polymorphisms with systemic lupus erythematosus, in addition to some unexpected findings such as the common occurrence of NKG2-C null allele in the healthy population. Genome-wide association studies using single nucleotide polymorphisms (SNPs) or microsatellite polymorphisms have become realistic, and development of new high-throughput and cost-effective SNP typing technologies is urgently needed. At the same time, our observations may indicate that the 'classical' candidate gene approach will remain a strong alternative, even in the age of 'post genome-sequence'.