Graves¡¦ disease (GD), an organ specific autoimmune disease, requires two signals to activate the T cells. In addition to the specific binding of T cell receptor to the antigenic peptide-MHC complex, an antigen-independent costimulatory pathway reportedly require generate subsequent cytokines and cell surface molecules. This regulation of T-cell response is a highly-organized multiple step program. T cell costimulatory signals is found to regulate the magnitude and duration of various type of autoimmune diseases. This study is to test whether genetic polymorphism of these costimulatory genes is related with the disease susceptibility or progression. We anticipated that the candidate genetic makers are beneficial for importing GD management. We recruited 262 GD patients from the Outpatient Department of Endocrine and 200 healthy controls from the Health Screening Center of Chang Gung Memorial Hospital in Kaohsiung.The GD patients were divided into three groups: recurred within 9 months (n=91), between 10-36 months (n=65), and more than 36 months (n=106). Clinical and laboratory attributes included: the genotypes of CTLA-4, CD28, ICOS, PD-1 and CD40; serum levels of T4, T3 and TSH; goiter size and TSH-receptor antibodies at the beginning and end of treatment. Genomic DNA was extracted from peripheral blood leucocytes by kit. The single nuclotide polymorphisms of the candidate genes were genotyped by polymerase chain reaction- restriction fragment length polymorphism and TaqMan® SNP Genotyping Assays with specific primers. Linkage disequilibuium between pairs of polymorphism was estimated by Haploview software. Haplotype analyses were performed using the Hap-Clustering program. Variance and correlation of data was statistically analyzed by Chi-square, general liner model, multiple logistic regression analysis and Kaplan-Meier plot. A p value <0.01 was considered significant. The results showed:(1) Genetic polymorphism within the costimulatory molecules affected the susceptibility and progression of GD; (2) GD patients carried more risk alleles than the controls; (3) Within the GD group, patients harboring more risk alleles wound relapse earlier after drug withdrawal; (4) Number of risk alleles, goiter size and TBII levels at end of treatment were independent predictors of disease relapse; (5) A risk score calculation based on odds ratio of risk alleles correlated with patients¡¦ relapse time after drug withdrawal. We concluded that patients¡¦ genetic makers of costimulatory molecules may be helpful in choosing appropriate treatment for GD.