Scanning laser polarimetry: logistic regression analysis for perimetric glaucoma diagnosis

To calculate and validate a linear discriminant function (LDF) for scanning laser polarimetry (SLP) with variable corneal compensation (GDx-VCC) to increase the diagnostic accuracy when using isolated retinal nerve fibre layer (RNFL) parameters to discriminate between healthy and glaucomatous eyes with visual field loss.

We prospectively selected two independent samples (teaching and validating sets). The teaching set, comprising 71 consecutive healthy subjects and 73 patients with open-angle glaucoma, was used to calculate the LDF. The validating set, comprising 72 consecutive normal eyes and 76 glaucoma patients, was used to test the performance of the LDF in an independent population. Receiver operating characteristic (ROC) curves were plotted for the validating set to evaluate the diagnostic accuracy of the LDF and the SLP parameters.

The obtained function was LDF=-12.20+(0.15 x nasal average)-(23.85 x normalized inferior area)+(1.18 x maximum modulation). The areas under the ROC curve were 0.901 and 0.893 for our LDF and 0.893 and 0.877 for the nerve fibre indicator (NFI) in the teaching and validating populations, respectively. There were no significant differences between these values (P=0.743 in the teaching set, and P=0.458 in the validating set). NFI was the SLP-provided parameter with the best sensitivity-specificity balance. Sensitivities were 57.89% for the LDF and 48.68% for NFI at 95% fixed specificity.

The LDF and NFI were the most accurate SLP parameters for diagnosing glaucoma. The LDF yielded the highest sensitivity at 95% fixed specificity to discriminate between normal and glaucoma subjects.