This analysis evaluated the association between serum retinol levels and risk of head and neck cancer (HNC) and whether the association is modulated by the use of alcohol, betel quid, or cigarette. In addition, we also examined the association between HNC risk and 2 single nucleotide polymorphisms, TTR rs1667255 and RBP4 rs10882272, that have been associated with serum retinol levels. Unconditional logistic regression was performed to evaluate the association between serum retinol levels and HNC risk among 160 HNC cases and 198 controls. The associations between TTR rs1667255 and RBP4 rs10882272 and serum retinol levels or HNC risk were evaluated by linear regression and unconditional logistic regression, respectively, for 418 HNC cases and 497 controls. The results showed that HNC cases had a lower mean serum retinol level compared with controls (845.3 μg/L vs 914.8 μg/L, P = 0.03). An inverse association between serum retinol levels and HNC risk occurred among never/occasional alcohol drinkers but not among regular drinkers. TTR rs1667255 was associated with serum retinol levels; however, neither TTR rs1667255 nor RBP4 rs10882272 was associated with HNC risk. In summary, this study showed an inverse association between serum retinol levels and HNC risk, specifically among never/occasional alcohol drinkers. More studies are needed to establish the underlying biologic mechanisms for the inverse association between serum retinol levels and HNC risk and the modulation of this relationship by alcohol drinking.