Purpose:
1. To evaluate factors for the development of POAG such as optic disc area, axial length, follow-up IOP, blood pressure (BP), systemic conditions and medications. 2. To construct a prediction model for the development of POAG with the highest predictive accuracy using the fewest variables.
Methods:
Data from 156 incident POAG cases from 1636 participants in OHTS with 6.5 years median follow-up through June 2002. Univariate and multivariate Cox proportional hazard models stratified by randomization group were fit. C-statistic and calibration chi-square were used to assess predictive accuracy and model fit. We identified the "best" single variable model based on the c-statistic, then added a 2nd variable that most improved the c-statistic until no statistically significant factors were left.
Results:
Factors that were not statistically significant in univariate models included axial length, blood pressure variables (perfusion pressure, systolic BP, diastolic BP, pulse pressure) diabetes, migraine, cardiovascular disease and calcium channel blocker usage. Factors that were statistically significant in univariate models included age, central corneal thickness (CCT), vertical cup to disc ratio (VCD), disc area, follow-up IOP (mean, standard deviation, range, maximum, percent change from baseline), pattern standard deviation and systemic beta blocker usage. The best 1, 2, 3, 4 and 5 variable prediction models are reported below.
Conclusions:
A 3-variable model with CCT, baseline VCD and mean f/up IOP provides similar predictive accuracy (c-statistic=0.76) to our previously reported 5 variable prediction model (c-statistic=0.75)1.1. Ocular Hypertension Treatment Study Group, European Glaucoma Prevention Study Group. Validated prediction model for the development of primary open-angle glaucoma in individuals with ocular hypertension. Ophthalmology 2007;114:10-19.
Clinical Trial:
http://www.clinicaltrials.gov NCT00000125
Keywords: optic nerve • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • visual fields