Abstract
Purpose::
To determine if baseline morphologic features predict which eyes are more likely to develop CNVM events
Methods::
522 participants over the age of 50 from the AREDS trial were pooled with the PTAMD (Prophylactic treatment of AMD) participants at the Pittsburgh site. All subjects were followed photographically for at least three visits over a 2 year period (baseline, 12months, and 24months). Digitized fundus images were assessed for drusen by a trained technician using the Drusen Analyzer (Iridex, Mountain View, CA) program which quantitates drusen parameters. Choroidal neovascular events were confirmed on fluroscein angiography by readers for the respective trials.The following parameters were evaluated for each study: total drusen area in the central 1000 microns of the macula, drusen area in the central 3000 microns, the presence of atrophy within 1000 microns, the presence of atrophy within 3000 microns, the presence of pigment within 1000 microns and the presence of pigment within 3000. Those parameters were also pooled from both studies to create a larger population. We fitted a mixed effect longitudinal model to drusen area and a mixed effect model with a random intercept to the event data.
Results::
At baseline, statistically significant factors (P<0.05) for predicting the occurrence of an event within the studies were total drusen area within the central 1000micron diameter field and the presence of pigment within the central of 3000 microns. The study assignment was also a significant factor. The correlation coefficient between the occurrence of an event and the predicted probability of occurrence was 0.18 for the model which incorporated drusen area within 1000 microns and the study assignment. The presence of pigment at 3000 microns adds 0.04 to the correlation coefficient when it was included in the model. Using a full model that includes age, study assignment, drusen area within the central 1000 microns and the presence or absence of pigment, an increase in drusen area within the central 1000 microns from 0.1 to 0.1064 mm2 increased the odds of an event occurring by about 42%. Note that a single druse 125 microns in diameter has an area of 0.012 mm2 . Having pigment alone increased the odds for an event by 313% in this model.
Conclusions::
Drusen analysis of baseline characteristics may help predict the likelihood of a CNV event occurring in an eye. The presence or absence of pigment in the macula was a consistent risk factor.The odds of an event occurring in the PTAMD study subset was 35 times that of an event occurring for the AREDS study, but PTAMD participants all were selected because of their higher risk characteristics.
Keywords: drusen • age-related macular degeneration • imaging/image analysis: clinical