Purchase this article with an account.
Richard A. Russell, David P. Crabb, David F. Garway-Heath; The ‘Deviation’ of Mean Deviation: Variability Estimates Derived from One Million Simulated Visual Fields. Invest. Ophthalmol. Vis. Sci. 2012;53(14):173.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Evaluation of visual field (VF) damage is often based on summary indices from standard automated perimetry such as mean deviation (MD); however, VF measurements are highly variable (as shown by frequency-of-seeing and test-retest studies), which significantly impedes assessment of progression. The aim of this study was to characterize the variability of MD using real VF data and statistical modelling.
A sample of 1,000 Humphrey 24-2 SITA-Standard VFs from 1,000 patients from Moorfields Eye Hospital from 1998-2009 were studied. One million VF simulations were generated using each real VF as ‘ground-truth’ (1,000 simulations per real VF); pointwise sensitivity was simulated where the amount of noise in the model was derived from previous estimates of pointwise VF variability. 1 The MD of each simulated VF was then calculated and the standard deviation of simulated MDs was compared with the ground-truth MD.
The median (interquartile range) patient age and MD was 66 (56 to 75) years and -3.5 (-8.3 to -1.11) dB, respectively. The estimated variability in MD is shown in Figure 1. The considerable scatter in Figure 1 suggests that pointwise sensitivity has a significant impact on MD variability. For example, VFs with global diffuse damage tend to be more variable than VFs with focal blinding damage and other regions that are healthy (see Patients ‘A’ and ‘B’ in Figure 1).
This study highlights that, on average, MD variability increases as the level of damage increases; however, the variability in MD is also dependent on the type of VF damage. The results from this study are important for helping clinicians to distinguish real VF progression from noise and will be used in statistical and health economic models for glaucoma progression detection.1 Russell, RA et al., ARVO 2010: 5490
This PDF is available to Subscribers Only