AIM:To evaluate reliability and diagnostic value of retinal nerve fiber layer(RNFL) thickness by GDx measurements in the diagnosis of glaucoma, and to identify the most important RNFL thickness parameters of GDx for early glaucoma diagnosis with stepwise discrimination analysis, sensitivity, specificity, accuracy, positive predictive value, negative predictive value,positive likelihood ratio,negative likelihood ratio of RNFL parameters on glaucoma diagnosis. ·METHODS: This study included 94 normal subjects (188 eyes) and 88 glaucoma patients (173 eyes). Sensitivity was calculated for all glaucoma patients(173 eyes) and again for two separate subgroups: early glaucoma (n=122 eyes) and moderation or advanced glaucoma(n=51 eyes) according to the mean defect of visual field. Receive operating characteristic curves(ROC) of discrimination function were plotted.·RESULTS: The mean age of the normal subjects and patients with glaucoma were 41.7±8.8(ranged form 25 to 75) and 52.8±14.6 (ranged form 19 to 73), respectively. There were 47 males and 47 females in normal individuals and 37 males and 51 females in patients with glaucoma. Mean defect of visual field of patients with glaucoma was from -1.6 to 23.2dB (Octopus 1-2-3 automatic perimeter). There was significant difference between normal subjects and glaucoma patients in all parameters of RNFL with t-test and ANOVA (P<0.01). Using single GDx VCC value as a cut-off point, the sensitivity:74.0%, specificity: 74.0%, accuracy: 86.8%, positive predictive value: 73.0%, negative predic-tive value: 76.0%,positive likelihood ratio: 2.96,negative likelihood ratio: 0.33 for TSNIT; sensitivity: 71.1%, speci-ficity: 84.6%, accuracy: 77.5%, positive predictive value: 80.9%,negative predictive value:76.1%,positive likeli-hood ratio:4.62,negative likelihood ratio:0.34 for SA; sensitivity:76.3%, specificity:82.4%,accuracy:78.8%,positive predictive value:80.0%,negative predictive value:79.1%,positive likelihood ratio:4.34,negative likelihood ratio:0.35 for NFI. When using two GDx VCC values as cut-off point, the sensitivity: 76.3%, specificity 74.0%, accuracy: 93.3%, positive predictive value: 87.7%, negative predictive value: 63.2%,positive likelihood ratio: 3.05,negative likelihood ratio: 0.32 for TSNIT and NFI, sensitivity:83.8%, specificity 79.3%, accuracy: 80.8%, positive predictive value: 78.8%, negative predictive value: 84.2%,positive likelihood ratio: 2.43,negative likelihood ratio: 0.29 for NFI and SA; sensitivity:88.4%, specificity 79.3%, accuracy: 83.0%, positive predictive value: 80.0%, negative predictive value: 88.2%,positive likelihood ratio: 4.27,negative likelihood ratio: 0.15 for NFI and IA; sensitivity:79.2%, specificity 84.6%, accuracy: 81.3%, positive predictive value: 82.5%, negative predictive value: 84.6%,positive likelihood ratio: 5.14,negative likelihood ratio: 0.25 for SA and IA. sensitivity:86.7%, specificity 85.6%, accuracy: 85.4%, positive predictive value: 84.7%, negative predictive value: 87.5%,positive likelihood ratio: 6.02,negative likelihood ratio: 0.16 for the integral of NFI, TSNIT, SA and IA. Under the same specificity, the sensitivity of early glaucoma stage with GDx was 66.4%, accuracy: 58.5%, positive predictive value: 92.0%, negative predictive value: 77.5%,positive likelihood ratio: 2.56,negative likelihood ratio: 0.45. The sensitivity of modera-tion and advanced glaucoma stage with GDx was 86.3%, accuracy: 77.4%, positive predictive value:48.4%, negative predictive value: 95.3%,positive likelihood ratio: 3.45,negative likelihood ratio: 0.18. When the "cut-off" point of NFI≥20, the sensitivity was 78.3%, specificity: 78.7%, accuracy: 78.7%, positive predictive value: 77.3%, negative predictive value: 80.0%,positive likelihood ratio: 3.68,negative likelihood ratio: 0.28. When the "cut-off" point of NFI≥23, sensitivity was 75.1%, specificity: 84.0%, accuracy: 79.8%, positive predictive value: 81.3%, negative predictive value: 74.9%,positive likelihood ratio: 4.69,negative likelihood ratio: 0.30. When the "cut-off" value of NFI≥27, sensitivity was 64.7%, specificity: 91.0%, accuracy: 35.7%, positive predictive value: 86.8%, negative predictive value: 73.7%,positive likelihood ratio: 7.19,negative likelihood ratio: 0.39. The area under ROC curve of GDx parameter were: NFI: 0.84, IA:0.79, TSNIT: 0.78, SA: 0.77, ISE: 0.76, respectively. Among them, the NFI discriminated best. In general, the area under the ROC curve increased from early to moderate and advanced glaucoma. Using stepwise discrimination analysis, NFI and IA discriminated the best for detecting early glaucoma diagnosis (F test: P<0.01), sensitivity and specificity with IA and NFI were 88.4% and 74.6%, respectively.·CONCLUSION: GDx can provide quantitative para-meters for clinic detecting RNFL thickness; NFI and IA are the best discrimination indexes to distinguish normal subjects and patients with glaucoma. GDx is helpful for early glaucoma diagnosis in clinic practice.