Abstract
Purpose: :
A linear structure–function relationship has been reported when perimetric data are expressed in linear units in a sample composed of glaucoma patients and normals (Garway–Heath et al, 2002). However this effect was inconsistent when only glaucoma suspects and glaucoma patients were studied (Leung et al, 2005). We investigated this further in a large sample of normals and patients covering a wide range of glaucoma stages.
Methods: :
Three–hundred–and–eighty–five participants were included: 195 glaucoma patients (100 with both visual field (VF) and photo defects, 80 with photo defects alone and 15 with VF defects alone), 77 ocular hypertensives (OH) and 113 participants with a normal ocular examination. All had Heidelberg Retina Tomograph (HRTII) imaging and a reliable standard automated perimetry (SITA 24–2) test performed within 6 months. VF data were expressed in logarithmic decibel (dB) and linear 1/Lambert units. Linear and quadratic functions were fitted to the data in each of the six HRT sectors and in the central VF. The independent variable was rim area and the dependent variable was the mean of the VF thresholds in the areas corresponding to the HRT sectors. We determined that quadratic fits explained significantly more variance than linear fits when the coefficient associated with the x2 term had a p–value < 0.05. Analyses were performed on the entire sample and on subsets in which specific subgroups (normals and OH) were excluded.
Results: :
The strongest correlations between structural and functional measures were obtained in the supero–temporal HRT sector (R2 ranged between 0.08 to 0.32). When VF data were expressed in dB units, curvilinear relationships were obtained. Expressing the VF data in linear units also yielded curvilinear relationships in each sector. No difference was observed between central and eccentric areas and the results were similar for all subgroup analyses.
Conclusions: :
Relatively small percentages of the overall variance were explained by the linear and curvilinear fits, with only small differences between the two. This may be due to a weak association between structural and functional measures at any given point in time, in addition to the variability associated with using surrogate measures of structure and function, large patient variability or the statistical methods used. While improving the precision of the structure–function relationship is an important goal, transforming perimetric data into linear units does not produce a clinically meaningful improvement over the use of the dB scale.
Keywords: visual fields • imaging/image analysis: clinical