We tested 30 eyes of 30 healthy subjects and 45 eyes of 35 glaucoma patients. Mean ages were as follows: healthy subjects group, 47.9 ± 13.3 years; combined glaucoma patients group, 65.0 ± 11.0 years; mild glaucoma patients group, 60.9 ± 10.6 years; moderate glaucoma patients group, 66.4 ± 10.4 years; and severe glaucoma patients group, 67.5 ± 11.2 years.
Comparing the AUC results of the ROC analysis shown in
Table 1, the SD-tVEP's P100 Lc latency parameter was found to have the greatest performance as a classifier for discrimination. Using an SD-tVEP P100 Lc latency cutoff value of 122.06 ms,
Table 2 tabulates the resulting sensitivities, specificities, NPV, and PPV for discriminating healthy subjects' eyes from glaucomatous eyes. The 95% confidence interval for sensitivity and specificity is included.
14
A linear regression was performed to investigate the confounding effect of age on the selected SD-tVEP parameter; Lclatency = 101.5 + 0.207 * Age (P < 0.05). There is a 2.07 ms per decade increase in the SD-tVEP Lc latency due to age. For completeness in the evaluation of the possible confounding effect of age, the same linear regression was performed on the remaining three SD-tVEP parameters: P100 Hc amplitude, Hc latency, and Lc amplitude. The results were as follows: Hcamplitude = 6.011 + 0.086 * Age (P = 0.329); Hclatency = 105.1 − 0.012 * Age (P = 0.919); Lcamplitude = 4.536 + 0.036 * Age (P = 0.418). Considering only the SD-tVEP parameter selected as a classifier, a two-way ANOVA was performed with the MD and patient's age as independent variables and the SD-tVEP P100 Lc latency as the dependent variable. The following conclusions were obtained from the two-way ANOVA. First, there was a significant main effect for the SAP_MD diagnostic group (F (3, 56) = 18.49, P < 0.05; η 2 = 44.01%; F critical (3, 56) = 2.77). Secondly, no significant main effect was found for the age group (F (6, 56) = 0.522, P > 0.05; η 2 = 2.48%; F critical (6, 56) = 2.27). Finally, there was not a significant interaction effect between the MD and age groups (F (9, 56) = 1.27, P > 0.05; η 2 = 9.04%; F critical (6, 56) = 2.05). Since there was no significant main effect for either the SD-tVEP Lc latency parameter or the MD, no statistical adjustments was made to age match the results of the healthy subjects and glaucoma patients.
Figure 2 is a scatter plot with the linear regression of MD (explanatory variable) and Lc latency (dependent variable).
Figure 3 is a plot of the resulting standardized residual of the linear regression. Ninety-five percent of the results fell within ±2 of the standardized residual plot and thus were considered normally distributed. Any data point falling outside this range was considered an outlier. There was a single significant outlier. The Pearson coefficient showed strong and significant correlation between the MD and the Lc latency (
r = −0.60,
P < 0.0001). For each unit increase in SD-tVEP Lc latency, the MD worsened by 2.18 dB.
The mean SD-tVEP low-contrast (Lc) latency parameters for each group were normal group, 112.6 ± 8.0 ms; mild glaucoma, 128.5 ± 7.4 ms; moderate glaucoma, 138.1 ± 17.0 ms; and severe glaucoma, 170.9 ± 42.1 ms (P < 0.0001).