Early diagnosis of head and neck cancer can improve therapeutic outcomes but remains a challenge. The blood proteome can comprise a key source of biomarkers that enable the early diagnosis and precision medicine in head and neck cancer, but blood protein biomarkers of head and neck cancer are not well delineated. Here we applied two-sample Mendelian randomization to a GWAS dataset of 1478 blood proteins and large dataset of head and neck cancer cases and controls to identify blood proteome traits associated with head and neck cancer. Multiple two-sample Mendelian randomization (MR) methods were used to assess causal effects of the exposures, including: Inverse-variance weighted (IVW), Mendelian randomization-Egger method, Weight Median method, simple mode, weight mode. Sensitivity analysis was performed by using heterogeneity test, pleiotropy test and one-by-one exclusion test. Multivariable MR analyses were performed to assess the effects of obesity, diabetes mellitus, and smoking. A significant causal association between A Disintegrin and metalloproteinase domain-containing protein 23 (ADAM23) and head and neck cancer was noted. The sensitivity analysis indicated no significant bias. Multivariate analysis showed that the effect for ADAM23 remained significant after adjusting for the indirect effects of obesity, diabetes mellitus and smoking. In sum, this study showed a significant causal role of genetically dysregulated ADAM23 protein with head and neck cancer risk. The specific mechanisms underlying the role of ADAM23 in mediating head and neck cancer risk, and its role as a potential therapeutic target and biomarker, need further investigation.