Background: Patients with age-related macular degeneration (ARMD) have several imaging techniques carried out regularly. In this study we introduce a new grading model of autofluorescence images (AF), compare it with fluorescein angiography (FFA) and digital colour fundus photos (COL) and test for inter- and intraobserver reliability.
Methods: A total of 71 eyes of 54 patients with bilateral or unilateral CNV had COL, FFA and AF, fulfilling the inclusion criterion of having all 3 types of imaging carried out on the same day or within 14 days. The grading of COL was performed by a trained grader based on the International ARM classification; FFA and AF images were independently graded by two trained retinal specialists in order to assess inter-observer reliability. Overall, 30% of all images were regraded after at least 14 days interval to assess intra-observer variability.
Results: The intergrader agreement was exact for classification of CNV (k = 1.00); almost perfect for FFA features (k = 0.83) and correspondence of decreased AF to COL (k = 0.94); substantial for patterns of decreased and increased AF (k = 0.80, k = 0.78), correspondence of patterns of increased AF to FFA and to COL (k = 0.78, k = 0.74) and background AF (k = 0.72); moderate for CNV diameter in FFA (k = 0.45), FFA pattern (k = 0.43), dimension of increased and decreased AF (k = 0.5, k = 0.56); fair for quality of FFA and AF images (k = 0.21, k = 0.26) respectively. The intragrader agreement varied from exact to substantial for all categories. Diffuse and reticular patterns of decreased AF and reticular pattern of increased AF correlated well, with visual acuity worse than 6/24.
Conclusion: The combined grading system was reliable for evaluating the three imaging techniques, and might be suitable for epidemiological studies and therapeutic trials where such grading is warranted. Certain AF patterns seem to predict VA outcome better than one might have predicted based on FFA. Further studies are needed to evaluate its usefulness in clinical settings for predicting outcomes for patients receiving therapy for end-stage disease.