advertisement
Editors Selection IGR 9-3
Visual Function and Structure: GDX-VCC linear discriminant function for case finding
Comment by Gábor Holló on:
23884 Scanning Laser Polarimetry with Variable Corneal Compensation to Detect Preperimetric Glaucoma Using Logistic Regression Analysis, Sánchez-Cano A; Baraibar B; Pablo LE et al., Ophthalmologica, 2009; 223: 256-262
Can the diagnostic accuracy of GDx-VCC be further improved? This important question has been addressed several times by different authors in the last years since scanning laser polarimetry with variable cornea compensator (GDx-VCC) was introduced into clinical practice to detect and monitor glaucoma. It was repeatedly found that the GDx-VCC software is not optimally sensitive for small localized retinal nerve fibre layer defects, even though those defects are visible on the polarimeric image. In this interesting study, in order to increase the accuracy of classification, Sánchez-Cano et al. (674) used 12-clock-hour retinal nerve fibre layer sectors analysed with linear discriminant function, instead of using the usual software provided parameters (TSNIT average, superior and inferior thickness and NFI) for separation of preperimetric glaucoma eyes from normal eyes. They grouped the sectors, identified the ones with best separation ability (superior and inferior clock-hour sectors), and created a new variable, for which the receiver-operating characteristics (ROC) curve was determined. When this ROC curve was compared to those calculated for the software provided parameters, significant differences were found in favor of the new parameter. The area under the curve was 0.710 for the new parameter and 0.691 or less for all others. At a specificity of 82.0% the new parameter had higher (47.4%) sensitivity than any of the software provided parameters (39.0% or less). In addition, the correlation with optic nerve head findings was better with the new parameter based on small sectors than any of the other parameters based on wider areas.
In fact, the success of this approach is not unexpected in preperimetric glaucoma with narrow and localized nerve fiber layer defects. However, even the improved sensitivity found by the authors is below optimal. Although age, IOP and central corneal thickness did not differ between the pre-perimetric glaucoma eyes and the control eyes, and all eyes with a spherical refractive error above ± 5.0 diopters were excluded from the study, no information on the presence or absence of atypical retardation pattern (ARP) in the study population is available in the article. ARP has been shown to influence both the measured retinal nerve fiber layer thickness and the diagnostic accuracy with GDx-VCC. The lack of information on ARP in the current study populations allows speculation on the potential role of ARP in the performance of the author's new parameter. It would be good to know if GDx with enhanced cornea compensation (GDx-ECC), the other commercially available software developed to neutralize ARP, performs even better in the same population when the same linear discriminate function is employed.
Comments
The comment section on the IGR website is restricted to WGA#One members only. Please log-in through your WGA#One account to continue.Log-in through WGA#One
