Age as a predictor of future rate of visual field change in glaucoma: Traditional statistics and tree-based modeling

Abstract

Purpose: In perimetry, mean deviation (MD) is corrected for age. However, age is also a risk factor for glaucomatous progression. We therefore aim to determine whether age influences cross-sectional and longitudinal structure-function relations in eyes with glaucoma and glaucoma suspect eyes.

Methods: Structural (OCT, Spectralis), functional (HFA 24-2, Carl Zeiss Meditec), ocular (IOP, central corneal thickness; CCT) and systemic blood pressure data were collected every six months from participants enrolled in the longitudinal Portland Progression Project. Data from the most recent 7 visits where perimetry was reliable and OCT was of acceptable quality were used (n=173, both eyes if available). We investigated the cross-sectional structure-function relation by using structural (global retinal nerve fiber layer thickness; RNFLT), ocular and systemic data from the 1st visit to predict MD at the 1st visit (MD₁). We investigated the longitudinal structure-function relation by using the rate of RNFLT change over time, together with the average ocular and systemic measures during the sequence, to predict the rate of MD change over time (MDR). Age at the 1st visit was used in both analyses. Traditional multivariate statistics (generalized estimating equations) and tree-based models were used.

Results: Global RNFLT (P<0.001) and age (P=0.027) were the only significant predictors of MD₁. This suggests that age influences the cross-sectional structure-function relation. For each 1µm thinner RNFLT, MD₁ was 0.1dB worse and for each decade older, MD₁ was 0.3dB worse. RNFLT and age were the most important predictors of MD₁ in tree-based models. The rate of change of RNFLT (P=0.004) and baseline age (P<0.001) were the only significant predictors of MDR. This suggests that age also influences the longitudinal structure-function relation. For each 1 µm/yr thinning of RNFLT, MDR was 0.05 dB/yr worse and for each decade older, MDR was 0.02 dB/yr worse. Age and the rate of RNTLT change were the most important predictors of MDR in tree-based models.

Conclusions: Age significantly influences both the cross-sectional and longitudinal structure-function relation in glaucoma. Furthermore, the rate of visual field change is significantly influenced by age, whereas it was not influenced by IOP, CCT or blood pressure in this clinically managed cohort.