Purchase this article with an account.
J. Caprioli, D. Mock, E. Bitrian, A. Afifi, F. Yu, K. Nouri-Madhavi, A. Coleman; Rates of Damage in Glaucoma: Techniques to Measure Rates of Regional Visual Field Change. Invest. Ophthalmol. Vis. Sci. 2010;51(13):2053.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To evaluate statistical models that yield spatial rates of progressionfor the assessment of visual field defects for glaucoma
We collected data from patients treated for primary-open angle glaucoma who received visual field testing (SAP 24-2) for a minimum follow-up time of 6 years. The absolute threshold levels was regressed with three statistical models (linear-1st order, quadratic-2nd order nonlinear, exponential- 1st order log-linear) for each visual field (VF) location. The Akaike information criterion (AIC) and multiple-R2 is used for post-regression diagnostics model selection that bests describes individual data trends. Regression-based cluster analysis is performed to assess the spatial field location correlation for the progression rates.
From the total number of eyes [n=389, mean age = 68.7 yrs + 9.5; average Mean Deviation (MD)= -10.8 dB+5.4]. The distribution for the best-fit models with AIC were: exponential 96.8%; linear 2.4%; and quadratic 0.8%. With the multiple R-squared criterion, the relative number of linear, quadratic and exponential best-fits for these models showed a better performance centrally  with linear and quadratic fits while peripheral test locations were better fitted with an exponential model [p<.025]. A dendrogram of the location rates of progression corresponded statistically with known arcuate nerve bundle patterns.
Regional modeling of rates of progression may enhance the determination of rates of glaucomatous visual field loss, facilitate the identification of "fast progressors", and provide clinically useful predictions of visual field outcomes.Examples: Progression for a single VF location with trends line exhibiting the exponential model and display of the calculated % rate per year as determined by the exponential fit model.
This PDF is available to Subscribers Only