Serum alpha-fetoprotein (AFP) has long been used as a diagnostic marker for hepatocellular carcinoma (HCC), albeit controversially. Although it remains widely used in clinics, the value of AFP in HCC diagnosis has recently been challenged due to its significant rates of false positive and false negative findings. To improve the efficacy of AFP as HCC diagnostic marker, we developed a method of measuring total and glycosylated AFP by multiple reaction monitoring (MRM)-MS. In this study, we verified the total amount of AFP (nonglycopeptide levels) and the degree of glycosylated AFP (deglycopeptide levels) in 60 normal (41 men and 19 women; mean age 53 years; range 32-74 years), 35 LC (23 men and 12 women; mean age 56 years; range 43-78 years; HBV-related), and 60 HCC subjects (42 men and 18 women; mean age 58 years; range 38-76 years; HBV-related; 30 stage I, 15 stage II, and 10 stage III). By MRM-MS analysis, the nonglycopeptide had 56.7% sensitivity, 68.3% specificity, and an AUC of 0.687 [cutoff value: ≥0.02 (light/heavy ratio)], comparing the normal and HCC group, whereas the deglycopeptide had 93.3% sensitivity, 68.3% specificity, and an AUC of 0.859 [cutoff value: ≥0.02 (light/heavy ratio)]. In comparing the stage I HCC subgroup with the LC group, the nonglycopeptide had a sensitivity of 66.7%, specificity of 80.0%, and an AUC of 0.712 [cutoff value: ≥0.02 (light/heavy ratio)], whereas the deglycopeptide had a sensitivity of 96.7%, specificity of 80.0%, and an AUC of 0.918 [cutoff value: ≥0.02 (light/heavy ratio)]. These data demonstrate that the discriminatory power of the deglycopeptide is greater than that of the nonglycopeptide. We conclude that deglycopeptide can distinguish cancer status between normal subjects and HCC patients better than nonglycopeptide.