A total of 45 POAG patients (80 eyes) were enrolled in this study, including 25 males and 20 females, with an average age of 61.7 ± 14.6 years (25–84 years). Twenty normal controls (20 eyes), including 8 males and 12 females, with an average age of 43.7 ± 19.4 years old (23–79 years old) were enrolled in this study. Demographics and clinical characteristics are displayed in Table 1. There were no significant differences in gender or IOP between the POAG group and the control group (P > 0.05), but there were significant differences in age, visual acuity, MS, MD, LV, and mean RNFL thickness (P < 0.05).

Table 1 Demographics and clinical characteristicsTest–retest repeatability of the EFT

The reproducibility of EFT was good with the test contrast of 3%, 5%, and 12%. And the reproducibility was moderate for the EFT with the test contrast of 25% and 35%. For the control group, the ICC were 0.819, 0.765, 0.767, 0.660, and 0.643 for the EFT with a test contrast of 3% (P < 0.001), 5% (P = 0.003), 12% (P = 0.003), 25% (P = 0.017), and 35% (P = 0.008), respectively.

RT in the control group

In this study, the average age of the control group was 43.7 ± 19.4 years, ranging from 23 to 79 years. There was no significant correlation between RT3%, RT5%, RT12%, RT25%, RT35%, and age (rs = 0.175, P = 0.462; rs = 0.034, P = 0.887; rs = 0.392, P = 0.088; rs = 0.372, P = 0.107; rs = 0.413, P = 0.07). The distribution of the RT in different contrasts in the control group is shown in Fig. 1. As the contrast of the EFT test stimulus increased, the RT gradually decreased.

Fig. 1

Distribution of RT in different contrasts in the control group

Analysis of EFT results in POAG subgroups

POAG eyes were divided into 5 groups according to eGSS (Table. 2). The results of structural and functional tests for glaucoma eyes were compared among different groups. There were significant differences in the BCVA, MS, MD, and LV among the different POAG subgroups. The mean RNFL thickness and the RNFL thickness of each quadrant also showed significant differences in different POAG subgroups (P < 0.05).

Table 2 Clinical characteristics of POAG subgroups

Within 50 s, 1 of the eyes in the advanced glaucoma group could not perceive test stimuli at 12% and 25% contrast. Among the eyes with end-stage glaucoma, 8 eyes could not perceive the test stimulus at 12% contrast, 6 eyes could not perceive the test stimulus at 25% contrast, and 3 eyes could not perceive the test stimulus at 35% contrast. With increasing test contrast, the ability of the POAG eye to perceive the test stimulus in the late and end stages improved.

The Kruskal–Wallis test was used to compare the differences between the POAG subgroups and the control group. There were significant differences between the RT in the severe, end-stage POAG groups, and the control group at 12%, 25%, and 35% test contrast (P < 0.001). However, there were no significant differences between RT in the early, moderate, advanced groups, and the control group (P > 0.05) (Table. 3). The results were Bonferroni corrected.

Table 3 Analysis of EFT results in POAG subgroups

RT in different test contrasts in the control group and different stages of POAG groups were shown in Fig. 2. Kendall’s tau-b correlation analysis was used to analyze the correlation between the different stages of POAG and RT with different test contrasts. RT12%, RT25%, and RT35% were positively correlated with POAG visual field staging, respectively (Kendall’s tau-b = 0.546, P < 0.001; Kendall’s tau-b = 0.591, P < 0.001; Kendall’s tau-b = 0.561, P < 0.001).

Fig. 2

RT in different test contrasts in the control group and different stages of POAG groups

The Kruskal–Wallis test was used to compare the differences in the EFT results among different POAG subgroups (Fig. 3). The results were Bonferroni corrected. At contrasts of 25% and 35%, the RT of end-stage glaucoma was quite different from that of the other POAG subgroups.

Fig. 3

RT12%, RT25% and RT35% in different POAG subgroups. **P < 0.001; *P < 0.05

Correlations between RT and structural and functional measurements

Generalized estimating equation was used to analyze the correlations between the RT at different contrasts ​​and glaucomatous structural and functional monitoring indicators (BCVA, IOP, MD, and RNFLT). The results are shown in Table 4. The RT at different contrasts was significantly correlated with BCVA (P ≤ 0.001). There was no significant correlation between IOP and RT12%, RT25%, and RT35% (P = 0.721, P = 0.973, P = 0.189). Among the visual field indicators, MS was significantly correlated with RT12%, RT25%, and RT35% (P < 0.001); MD was significantly correlated with RT12%, RT25%, and RT35% (P < 0.001); and LV was significantly correlated with RT12%, RT25%, and RT35% (P < 0.05). The general thickness of the RNFL and the thickness of the RNFL in each quadrant were significantly correlated with the RT at different contrasts (P < 0.05).

Table 4 Correlations between RT at different contrasts and glaucomatous structural and functional measures

Correlation models between RT and structural and functional measurements were analyzed by regression analysis (Table. 5). The scatterplots of RT12% and MD were best fitted by a cubic model with a higher R2 (Fig. 4A):

Table 5 Correlation models between RT and structural and functional measurementsFig. 4

Scatterplot and correlation models between RT and MD. A Scatterplot and correlation models between RT12% and MD. B Scatterplot and correlation models between RT25% and MD. C Scatterplot and correlation models between RT35% and MD

The scatterplots of RT25% and MD were best fitted by an exponential model with a higher R2 (Fig. 4B):

The relationship between RT35% and MD was well explained by an exponential model with a higher R2 (Fig. 4C):

ROC analysis

ROC analysis was performed for RT12%, RT25%, RT35%, and MD. Among the RTs, RT12% showed the best performance for diagnosing POAG, with an AUC of 0.863, followed by RT25% with an AUC of 0.855. RT35% had the worst diagnostic value with an AUC of 0.772. And the AUC for MD was 0.940 (Fig. 5). DeLong’s test for ROC curves demonstrated significant differences between RT12% and RT35% (Z = 2.409, P = 0.016) and RT25% and RT35% (Z = 2.595, P = 0.0095). And there was no significant difference in the ROC curve between RT12% and RT25% (Z = 0.188, P = 0.85).

Fig. 5

ROC curves for RT12%, RT25%, RT35%, and MD

POAG staging based on EFT and OCT

POAG eyes were further divided into two groups according to the eGSS: the first group included early, moderate, advanced, and severe glaucoma eyes, and the second group included end-stage glaucoma eyes. The characteristics of end-stage glaucoma were analyzed. Compared to the first group, the RT25% of the end-stage glaucoma was significantly increased, and the general RNFL thickness was significantly reduced (P < 0.05) (Table. 6). Therefore, this study used RT25% and average RNFL thickness to develop a nomogram model. These two indicators correspond to the scores. A probability axis was drawn based on the relationship between total points and the probability of end-stage glaucoma (Fig. 6A). The nomogram demonstrated relatively good performance, with a C index of 0.77, and the calibration plot showed that the model curve fit well with the standard curve (Fig. 6B). According to the nomogram, the POAG eyes could be further divided into 8 stages (Fig. 6C): Patients with or fewer than 45 points were defined as stage 1, patients with 45–60 points or with 60 points were defined as stage 2, patients with 60–75 points or with 75 points were defined as stage 3, patients with 75–85 points or with 85 points were defined as stage 4, patients with 85–95 points or with 95 points were defined as stage 5, patients with 95–100 points or with 100 points were defined as stage 6, patients with 100–115 points or with 115 points were defined as stage 7, and patients with more than 115 points were defined as stage 8. According to the probability distribution (the red line), this model was less effective in evaluating the visual function of early glaucoma. As the disease progressed, it became more accurate.

Table 6 The multivariate logistic regression analysis of end-stage glaucomaFig. 6

A For predicting the probability of end-stage glaucoma, RT25% and average RNFL thickness were included and placed on the variable axes. Draw vertical lines of each risk factor towards the total point scale, which was assigned to be the point as each parameter. Then sum up the points of the parameters as total points. B The calibration plot showed that the model curve fit well with the standard curve. C Risk curve based on the total points refers to the probability of end-stage glaucoma