Genetic susceptibility and Epstein-Barr virus (EBV) infection are important etiological factors in nasopharyngeal carcinoma (NPC). In this study, in southern China, where NPC is endemic, a single nucleotide polymorphism (SNP) in the EBV-encoded RPMS1 gene (locus 155391: G > A [G155391A]) and seven host SNPs (rs1412829, rs28421666, rs2860580, rs2894207, rs31489, rs6774494, and rs9510787) were confirmed to be significantly associated with NPC risk in 50 NPC cases versus 54 hospital-based controls with throat washing specimens and 1925 NPC cases versus 1947 hospital-based controls with buffy coat samples, respectively. We established a strategy to detect the NPC-associated EBV and host SNPs using saliva samples in a single test that is convenient, noninvasive, and cost-effective and displays good compliance. The potential utility of this strategy was tested by applying a risk prediction model integrating these EBV and host genetic variants to a population-based case-control study comprising 1026 incident NPC cases and 1148 controls. Receiver operating characteristic (ROC) curve analysis revealed an area under the curve of the NPC risk prediction model of 0.74 (95% CI: 0.71-0.76). Net reclassification improvement (NRI) analysis showed that inclusion of the EBV SNP significantly improved the discrimination ability of the model (NRI = 0.30, P < 0.001), suggesting the promising value of EBV characteristics for identifying high-risk NPC individuals in endemic areas. Taken together, we developed a promising NPC risk prediction model via noninvasive saliva sampling. This approach might serve as a convenient and effective method for screening the population with high-risk of NPC.