Purchase this article with an account.
A. S. Raza, D. C. Hood, R. H. Kardon; Are We Ready to Estimate Retinal Ganglion Cell Number from SAP Sensitivity and OCT Thickness?: A Comparison of Models Relating Structural and Functional Loss Due to Glaucoma. Invest. Ophthalmol. Vis. Sci. 2009;50(13):4401.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare models that relate RNFL thickness measured with optical coherence tomography to loss of visual sensitivity due to glaucoma.
Sixty-six eyes of 66 patients (mean MD -6.5±4.4 dB) with glaucoma (G) and 44 eyes of 44 controls (C) were tested with static automated perimetry (SAP) (24-2 SITA, Zeiss) and time domain OCT (Fast RNFL scan, Zeiss). For the OCT, the RNFL thickness profiles were exported and the average thickness was determined for the 2 arcuate regions of the Garway-Heath et al (GH) map  and for the 4 equivalent regions of the Harwerth et al map ; these are the critical regions for glaucoma and they fall within the 24-2 field. The models described in Harwerth et al (HWR)  and Hood and Kardon (HK)  were evaluated by comparing the percentage of the variance in the data accounted for by the model. An optimization of the HWR parameters was performed to improve the fit of a modified HWR model that contains a correction for stage of disease.
First, for both maps, the HK model accounted for a greater proportion of the variance in the data, even after modifying the HWR model to include a new disease stage correction . Second, for the arcuate regions of the GH map, optimizing the parameters of the modified HWR model brought the performance of the models close. Third, neither model provides a complete explanation for the extensive variability seen in the data.
While the HK model outperformed the HWR model when using published parameters, optimization of the HWR parameters brought the performance of the models close. However, the important difference between the models is not how well they fit the data. Rather, there are important differences in the scope and assumptions. The HWR model has more parameters and assumptions, but is more ambitious in scope as it converts both OCT and SAP data into retinal ganglion cell count. Thus, the HWR model represents an advancement over the HK model, but only if its assumptions re axon thickness, SAP to OCT mapping, as well as glial cell contribution to RNFL thickness with age and disease stage, can be confirmed with independent (e.g. histological) evidence. 1. Garway-Heath et al. Ophthal, 2000; 2. Harwerth et al. IOVS, 2007; 3. Harwerth et al. IOVS, 2008; 4. Hood & Kardon, PRER, 2007; 5. Harwerth (personal communication).
This PDF is available to Subscribers Only