Abstract
Purpose: :
To investigate the relationship between the number of small, medium and large drusen and the total drusen area in patients with dry AMD. While total drusen area should necessarily equal the number of drusen of all sizes multiplied by the size of the individual drusen, historical methods mandate that individual sizes be categorized. This leads to some approximations. Furthermore, because drusen are typically not round, druse size is taken to be its longest linear dimension. We also investigated whether the presence of a given number of small, medium, large, and very large drusen was reflective of total drusen area.
Methods: :
Using a validated computer algorithm to semi-automatically detected drusen (Ophthalmic Surg Lasers Imaging. 2007 Mar-Apr;38(2):126-34), we measured the total drusen area in 949 eyes with dry AMD. The algorithm could detect a druse as small as 9 microns in diameter. However, as the numbers of very tiny drusen was substantial and could be considered noise in the detection method, we ignored this category. We explored the relationship between quantitative drusen counts (medium, 125-249 microns; large, 250-499 microns; and very large, 500-999 microns) and the total drusen area using linear mixed effects models to account for the clustering within subjects.
Results: :
The relative importance of the number of medium, large, and very large drusen in predicting area was determined by fitting models with independent variables standardized so the fixed effect regression coefficients could be compared. Surprisingly perhaps, for drusen located within the central 3000 micron region of the macula, when comparing single explanatory factor models, the total number of medium-sized drusen (those 125-249 microns across) was a better predictor of total drusen area than the number of large (250-499 microns) or very large (500-999 microns) drusen using the generally accepted Akaike Information Criterion (AIC). Analogously for the fovea (1000 micron diameter region), the total number of large drusen was superior to models with only medium or very large drusen counts. Including all three sizes in the model, however, produced a much better model than single factor models. In this model very large drusen had more weight than large drusen, and large drusen had more weight than medium drusen (1000: very large 0.037, large 0.031, medium 0.0159, p<0.0001; 3000: very large 0.245, large, 0.117, medium 0.077, p<0.0001).
Conclusions: :
Assuming that total drusen area is a risk factor for visual loss in dry AMD, from a clinical perspective, the number of medium drusen in the central 300 microns of the posterior pole is a better reflection of the total drusen area than the number of large or very large drusen.
Keywords: drusen • age-related macular degeneration • imaging/image analysis: clinical