Classification trees can predict progressive glaucomatous optic neuropathy prognosis

The authors of this study used decision trees to predict progressive glaucomatous optic neuropathy (pGON) outcomes using data on baseline standardized automated perimetry (SAP) exams, central corneal thickness, baseline IOP and age. Results of the decision trees had an average 65 percent sensitivity and 87 percent specificity. When data from 100 normal subjects were used in the models, average specificity was 69 percent. The results suggested that using information regarding the percentile associated with threshold values may assist in assessing the functional status of glaucoma patients and their risk of optic nerve head progressive change.

The study was conducted using age-adjusted baseline SAP thresholds and other clinical variables from the worse eye of 168 individuals with high-risk ocular hypertension or early glaucoma. Longitudinally gathered stereo optic nerve head photographs were used to diagnose pGON.

Forty percent of the subjects were diagnosed with pGON during annual follow-up visits. The authors' classification and regression tree model was able to divide this group into high- and low-risk cohorts using only age-adjusted baseline SAP thresholds at six locations. The inferior visual field, particularly along the nasal horizontal meridian, appeared most important when visual field locations were ranked by pGON predictive strength.