Purchase this article with an account.
Harsha L. Rao, Uday K. Addepalli, Ravi K. Yadav, Nikhil S. Choudhari, Sirisha Senthil, Anil K. Mandal, Chandra S. Garudadri; Accuracy Of Ordinary Least Squares And Empirical Bayes Estimates Of Short Term Visual Field Progression Rates To Predict Long Term Outcomes In Glaucoma. Invest. Ophthalmol. Vis. Sci. 2012;53(14):182.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare the accuracy of ordinary least squares (OLS) and empirical Bayes (EB) estimates of short term visual field (VF) progression rates to predict long term outcomes in glaucoma.
Short term rates of VF progression were calculated using the first 5 Visual Field Index (VFI) values of 170 eyes (136 patients) of primary glaucoma patients with >5 VF examinations during follow-up. Rates were estimated by calculating slopes using OLS estimates and, EB estimates of best linear unbiased predictions. VFI values predicted at all subsequent VF examinations using these two slopes were compared with actual VFI values.
Median follow-up of patients was 8.7 years (1st and 3rd quartiles: 6.5, 11.2). Median number of VF examinations was 8 (7, 10). Median short term progression rate was -0.39% per year (-1.32, 0.47) by OLS and -0.23% per year (-0.56, 0.04) by Bayes estimates. Median absolute prediction error was significantly lesser (p<0.001) with Bayes (3.6 VFI units) compared to OLS estimates (4.3 VFI units). Predicted values were within ±5 units of actual VFI values in 56% of VFs using OLS estimates and 63% using Bayes estimates. Predicted values were within ±10 units of actual VFI values in 79% of VFs using OLS estimates and 84% using Bayes estimates.
Prediction accuracy of future VFI values were statistically significantly better with EB estimates of short term rate of progression compared to the OLS estimates. The differences in prediction errors with the 2 methods however, were small and unlikely to be clinically significant.
This PDF is available to Subscribers Only