Lung cancer is the leading cause of cancer-related deaths worldwide. By now, genome-wide association studies (GWAS) have identified numerous loci associated with the risk of developing lung cancer. However, these loci account for only a small fraction of the familial lung cancer risk. We hypothesized that epistasis may contribute to the missing heritability. To test this hypothesis, we systematically evaluated the association of epistasis of genetic variants with risk of lung cancer in Han Chinese cohorts. We conducted a pair-wise genetic interaction analysis of 591,370 variants, using BOolean Operation-based Screening and Testing (BOOST), in an ongoing GWAS of lung cancer that includes 2,331 cases and 3,077 controls. Pairs of epistatic loci with PBOOST ≤ 1.00 × 10(-6) were further evaluated by a logistic regression model (LRM) with covariate adjustment. Four promising epistatic pairs identified at the screening stage (PLRM ≤ 2.86 × 10(-13)) were validated in two replication cohorts: the first from Beijing (1,534 cases and 1,489 controls), and the second from Shenyang and Guangzhou (2,512 cases and 2,449 controls). Using this combined analysis, we identified an interaction between rs2562796 and rs16832404 at 2p32.2 that was significantly associated with the risk of developing lung cancer (PLRM = 1.03 × 10(-13) in total 13,392 subjects). This study is the first investigation of epistasis for lung cancer on a genome-wide scale in Han Chinese. It addresses part of the missing heritability in lung cancer risk and provides novel insight into the multifactorial etiology of lung cancer.