AMD and diet

  Richard K Parrish II MD, presenting the Jackson Memorial Lecture Award to Emily Chew MD Eating a Mediterranean diet, particularly a lot of fish, may be beneficial for those with early or even intermediate age-related macular degeneration (AMD), said Emily Y Chew MD in her Jackson Memorial Lecture at AAO 2019 in San Francisco, USA. Research shows that a diet high in fish can reduce the chances of developing late AMD by 65% for patients who also have protective genes, Dr Chew said. In the general AMD population, a high fish diet reduced progression of intermediate AMD, with bilateral large drusen, to geographic atrophy by 31%. High adherence to a Mediterranean diet reduced progression from intermediate to late AMD, with geographic atrophy or neovascularisation, by 25-to-40%. A Mediterranean diet was defined as one high in fruits, vegetables, nuts and legumes, moderate in fish, white meat and whole grains, moderate-to-low in alcohol and low in red meat and refined sugar, with a high ratio of mono-unsaturated to saturated fat intake. As well as reducing progression from intermediate to late AMD, it also reduced progression from the small drusen of early AMD to large drusen of intermediate disease. “It’s never too late to start,” she said. Genes and AMD risk The findings Dr Chew presented are based on data from 7,765 patients without baseline late AMD from the landmark Age-Related Eye Diseases (AREDS) and AREDS2 studies, supplemented by diet surveys, and analysis of patient records, retinal images and genetic tests of patients followed for 10 years each. A study team including researchers from the University of Pittsburgh, the US National Eye Institute, the Oregon Health and Science University and the University of Michigan combined these data with findings from previous large-scale studies of the genetics of AMD to assess the effects of genetics on AMD progression, and to develop and test prediction models that include genetic data as well as other AMD risk factors, Dr Chew said. The AREDS/AREDS2 analysis confirmed earlier findings that the gene for complement factor H (CFH) on chromosome 1 was associated with greatly increased risk of drusen formation, while ARMS2 on chromosome 10 is associated with haemorrhage and poor visual acuity (Ophthalmology 2018;125(4):559-568). An earlier study by the AMD Gene Consortium found these two genes accounted for 60% of hereditary AMD risk, and these were included with another 50 previously identified AMD genetic risk markers in a genetic risk score tool (Nat Genet 2016;48:134-143). The AREDS/AREDS2 analysis identified additional AMD genetic risk alleles, Dr Chew said. These analyses are important for research and understanding the pathways of AMD progression, Dr Chew said. But are they clinically useful? To find out, the group tested several AMD progression prediction models with and without a genetic risk score. Other factors in the models were baseline age, education, smoking status and baseline AMD severity scores in both the study and fellow eyes. The most powerful predictors turned out to be baseline severity scores for both the study and fellow eyes, with genetic risk scores only minimally improving performance when added to such a model. (Genetics 2017;206:119-133) The results “point to the importance of an eye exam”, Dr Chew said. No interaction was observed between genetic factors and a Mediterranean diet for AMD progression, Dr Chew said. However, for patients with a protective gene for CFH, consumption of fish was 
strongly associated with reduced AMD progression risk. Even low fish intake reduced risk of progression from intermediate AMD to geographic atrophy by 50%, while high fish intake reduced it 65%. While these findings are important for research, Dr Chew sees little clinical use for genetic testing for AMD. Commercial gene testing services offer AMD gene evaluations, but no prospective data support testing for predicting progression or response to therapies, or for dietary recommendations. For these reasons the AAO does not recommend genetic testing for AMD, she said. AI and AMD The Simplified Severity Scale developed through AREDS classifies the severity and risk of progression to late AMD on a person level based on the presence of large drusen, pigmentary changes and neovascularisation or geographic atrophy in fundus images of both eyes. However, it requires expert analysis by retinal specialists. To broaden access to AMD screening, NEI researchers harnessed AI to automate analysis, Dr Chew said. Based on a training set of more than 58,000 AREDS images, the DeepSeeNet project developed deep learning neural networks to assess each of the three image subgroups; drusen, pigmentary changes and late AMD. In a test of 900 images, DeepSeeNet was more accurate in detecting drusen and pigmentary changes and only slightly less accurate in detecting late AMD than an AREDS retinal specialist panel assessing the same images, Dr Chew said. (Ophthalmology 2019;126:565-575) She believes DeepSeeNet’s accuracy in detecting late AMD will improve with more training images and is likely to surpass human readers. Dr Chew does not see deep learning systems replacing ophthalmologists, but they may soon improve diagnosis and clinical management. “We highlight the potential for this system in enhancing decision-making in AMD,” she concluded. Emily Chew: echew@nei.nih.gov