Purchase this article with an account.
D. P. Crabb, V. M. F. Owen, D. F. Garway-Heath; Poor Agreement Between Current Tests of Structural and Functional Progression in Glaucoma Can Be Explained by Measurement Noise. Invest. Ophthalmol. Vis. Sci. 2007;48(13):1615. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To demonstrate that results from recent longitudinal studies 1,2 indicating that poor agreement exists between progression detection rates based on structural measures (Heidelberg retina tomography imaging) and functional measures (automated perimetry) may be due to measurement noise.
A computer simulated ‘patient’ was given an identical linear rate of loss (R%) for both function (F) and structure (S) ‘measurements’ sampled from a clinically realistic population of rates of loss described by a Weibull distribution. R% was degraded with measurement noise of known characteristics, once for the functional measure and once for the structural measures. Simulation experiments were conducted where the magnitude and independence of the noise for F and S were varied. Significant rates of loss, as detected by linear regression on time with strict control of type 1 errors, were compared with outcomes being the percentage of ‘patients’ that progressed by F alone, by S alone, or by both (F&S).
Venn diagrams with areas proportional to % of patients classified as progressing are shown in the figure with the area outside the circles representing patients not detected by either method. The modelled result when change in F and S is identical but noise was high and unrelated (panel f) is strikingly similar to those from published data on real patients (a-b).
An explanation for the apparent lack of concordance between current tests in detecting structural and functional progression lies in measurement noise, with the unique characteristic of the noise in the different modalities of testing contributing to the disassociation as much as the simple magnitude of noise. Moreover, the results indicate that measurement noise is certainly unrelated: a patient demonstrating highly variable structural measures will not necessarily give highly variable functional measurements and vice versa.1. Artes & Chauhan. (2005) Prog Retin Eye Res. 24:333-54.2. Strouthidis et al (2006) IOVS. 47:2904-10.
This PDF is available to Subscribers Only