Structure-function relationship depends on glaucoma severity
- Marta Gonzalez-Hernandez (martagh{at}jet.es),
- Luis E Pablo (lpablo{at}unizar.es),
- Karintia Armas-Domingue,
- Ricardo Rodriguez de la Vega,
- Antonio Ferreras (aferreras{at}msn.com),
- Manuel Gonzalez de la Rosa (mgdelarosa{at}telefonica.net)
- University Hospital of the Canary Islands. University of La Laguna, Spain
- Department of Ophthalmology. Miguel Servet University Hospital. University of Zaragoza, Spain
- Department of Ophthalmology. University Hospital of the Canary Islands. University of La Laguna, Spain
- Department of Ophthalmology. University Hospital of the Canary Islands. University of La Laguna, Spain
- Department of Ophthalmology. Miguel Servet University Hospital. University of Zaragoza, Spain
- Department of Ophthalmology. University Hospital of the Canary Islands. University of La Laguna, Spain
- Published Online First 2 June 2009
Abstract
Background/Aims: To determine the structure-function relationship throughout the different stages of glaucoma.
Methods: Subjects: 228 controls and 1007 suspected, early, moderate, or advanced glaucomas. In controls, Pearson’s coefficient of variation (PCV) of standard automated perimetry mean sensitivity (SAP-MS) and global mean retinal nerve fiber layer thickness (RNFL, Heidelberg Retina Tomograph) were calculated. Piecewise linear regression was used to evaluate if the data are related better to two separate regression lines in all population.
Results: PCV for SAP-MS and RNFL was 6.19% and 29.27%, respectively. For SAP-MS >22.42 dB (piecewise linear regression-obtained breakpoint, p<0.05), the logarithmic and linear relationships between SAP-MS and mean RNFL thickness were not different (r=0.182, r=0.185 respectively; p=0.950). For SAP-MS <22.42 dB, the logarithmic and linear relationships between SAP-MS and mean RNFL thickness were not different (r=0.353, r=0.344 respectively; p=0.884).
Conclusion: The logarithmic and linear relationships between SAP-MS and mean RNFL thickness did not differ when individuals were stratified using the piecewise linear regression-obtained cut-off point. The curvilinear relationship between the morphologic and perimetric results may be due to the wide variability in normal morphology and limitations in the dynamic range of the morphologic tests in cases with moderate and severe defects.
