The Ocular Hypertension Treatment Study (OHTS): Longitudinal analyses of mean deviation (MD) loss and its association with visual acuity (VA) and contrast sensitivity (CS) in eyes newly diagnosed with primary-open angle glaucoma (POAG)

Abstract

 

Purpose

 

1.To use latent class analysis (LCA) to identify mutually-exclusive classes of eyes by their rate of MD loss 2.To test the association of rate of MD loss with VA and contrast sensitivity (CS).

 

Methods

 

Analysis of 13 years of data collected prospectively from 277 participants (359 eyes) that developed POAG (5 years post-POAG f/up) and 1243 participants (2480 eyes) that did not. Visual fields were performed semi-annually, ETDRS visual acuity (VA) annually and contrast sensitivity (CS) at study close-out. At least 4 VF’s were required for inclusion. Data were censored after cataract surgery. LCA was performed to divide POAG eyes into subgroups based on rate of MD loss. Bayesian Information Criterion (BIC) was used to select optimal LCA. To assess the association of MD loss with VA and CS, multivariable linear mixed-effect models were fitted to compare between-group differences in ETDRS and CS, assuming the LCA subgroups as known. For POAG eyes, length of f/up was centered at the date of POAG diagnosis and, for non-POAG eyes, at 90th month the median time of POAG diagnosis. All the mixed models allow each eye to have its own random intercept and (if included) slope.

 

Results

 

In multivariable, linear mixed-effects models, no differences between POAG and non-POAG eyes were detected in VA (52.4 vs. 52.6 letters, p=0.75) or CS (1.58 vs. 1.61, p=0.17), though POAG eyes showed slightly faster decrease in VA (-0.61 vs. -0.44 letters/year, p=0.002). LCA and BIC identified 5 classes best describing MD loss (Figure). Mixed models compared non-POAG eyes with LCA classes and found classes 4 and 5 (15% of POAG eyes) had significantly worse VA and CS compared to non-POAG eyes and classes 1&2 (Table). The estimated intercepts and slopes of MD from mixed model were also presented in the Table.

 

Conclusions

 

Overall comparisons of outcomes by POAG status may fail to detect differences affecting a small proportion of participants. LCA provided formal criteria for identifying heterogeneity among participants in whom rapid rate of MD loss was associated with loss in other visual functions.

 

 

Estimated intercepts and slopes from linear mixed models

 

 

MD (means, 95% CI) for 5 latent classes of MD loss

 

Keywords: 459 clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • 462 clinical (human) or epidemiologic studies: outcomes/complications