The purpose of the study was to investigate the association of single-nucleotide polymorphisms (SNPs) within excision repair cross-complementation (ERCC) gene polymorphisms, additional gene-gene interaction, and haplotype combination with osteosarcoma risk. Generalized multifactor dimensionality reduction (GMDR) was used to screen the best interaction combination among SNPs. Logistic regression was performed to investigate the association between six SNPs within ERCC gene, additional gene-gene interaction on osteosarcoma risk. Haplotype analysis was performed using SNPstats ( http://bioinfo.iconcologia.net/SNPstats ). Osteosarcoma risk was significantly higher in carriers with the T allele of ERCC2-rs1799793 than those with GG genotype (GT+ TT vs. GG), adjusted OR (95% CI) = 1.56 (1.13-2.01), and higher in carriers with the A allele of ERCC3-rs4150441 than those with GG genotype (GA+ AA vs. GG), adjusted OR (95% CI) = 1.63 (1.25-2.09). GMDR model indicated a significant two-locus model (p = 0.0107) involving rs1799793 and rs4150441; cross-validation consistency of the two-locus model was 9/10; and the testing accuracy was 60.11%. Participants rs1799793-GT or -TT and rs4150441-GA or -AA genotype have the highest osteosarcoma risk, compared to subjects with rs1799793-GG and rs4150441-GG genotype, OR (95% CI) = 2.87 (1.21-4.63), after covariates adjustment. Haplotype containing the rs1799793-T and rs11615-T alleles was associated with a statistically increased osteosarcoma risk, OR (95% CI) = 1.47 (1.12-1.92). We found that the T allele of ERCC2-rs1799793 and the A allele of ERCC3-rs4150441, interaction between rs1799793 and rs4150441, and haplotype containing the rs1799793T and rs11615-T alleles were all associated with increased osteosarcoma risk.