advertisement

Editors Selection IGR 11-1

Perimetry: SAP and progression

Comment by Boel Bengtsson on:

22536 Predicting progressive glaucomatous optic neuropathy using baseline standard automated perimetry data, Demirel S; Fortune B; Fan J et al., Investigative Ophthalmology and Visual Science, 2009; 50: 674-680

Find related abstracts

In a longitudinal study, Demirel et al. (144) investigate the ability for Standard Automated Perimetry (SAP) fields to predict progressive glaucomatous optic neuropathy (pGON) in patients with early glaucoma or high-risk ocular hypertension followed for a median of 6.1 years. The authors apply a very interesting technique called Classification and Regression Tree (CART). CART is a tool for predictive modeling, a decision tree, learning from input data. Age-corrected SAP threshold values assessed at baseline, together with baseline IOP, baseline age and central corneal thickness, were used as input data. A fairly large number of patients (n = 168) and healthy subjects (n = 100) were included. Threshold values, also being within normal limits, at certain test point locations and pattern in the visual field were identified as important for prediction of pGON. The inferior part of the visual field seemed more important than the superior part. The sensitivity to predict pGON was 65%, which is not particular high, but certainly better than by chanceonly. The specificity was 87% on the average, surprisingly decreasing to 69% when healthy subjects were used as controls. A ten-fold cross validation was applied to avoid training and evaluation using the same dataset. Nevertheless, as pointed out by the authors, a validation in an independent dataset is needed before the general applicability of the method can be determined. The most intriguing feature in this paper is the potential possibility to identify a subset of test points in the visual field that is important for predicting glaucoma progression, here pGON. It would be even more interesting to see whether this technique also can be applied to predict visual field progression. It could then be possible to follow glaucoma patients with fields where fewer locations are tested and thereby shorten test time to enable more frequent testing in newly detected glaucoma patients, and patients needing intensified treatment to reduce the risk for developing serious field loss.

• ← Previous
• Next →