Influence of Risk Factors and Individualized Risk Scores for Age-Related Macular Degeneration through a Variable Influence Analysis Model

Abstract

INTRODUCTION: To explore the relative influence of different risk factors in age-related macular degeneration (AMD) development and progression in an epidemiologic-based study through the novel variable influence analysis (VIA) model, aiming to compute personalized AMD risk scores.

METHODS: Population-based 2-visit epidemiologic study (Coimbra Eye Study) on AMD prevalence and 6.5-year incidence. Participants were imaged with color fundus photography at both visits and additionally with NIR, FAF, and OCT at the follow-up visit. Data on medical history and risk factors were obtained, including a food frequency questionnaire to calculate adherence to the Mediterranean diet. Blood samples were collected, and 69 SNPs were genotyped with the EYE-RISK genotype assay. A novel VIA model was developed to calculate the influence score of each risk factor in the transition between AMD stages. A global patient ‘risk score’ for AMD was then computed.

RESULTS: We included 948 subjects, 243 with AMD and 705 controls. The transition from no AMD (Rotterdam stages 0 or 1) to AMD (Rotterdam stages 2, 3 or 4) was mainly predicted by baseline AMD stage, age, risk variants CFHrs35292876, CFHrs10922109 and ARMS2/HTRA1rs3750846. To predict the transition to more severe stages (stages 3 and 4), the influence score of smoking almost doubles, and new influential factors that were negligible become more relevant, including diabetes, high blood pressure at baseline, and variant C3rs2230199. For root cause analysis, the most influential variables, explaining what caused these transitions, were variant C2rs429608, adherence to Mediterranean diet, physical exercise, body mass index, and arterial hypertension. Based on the influence scores obtained, global AMD risk scores were computed for each participant.

CONCLUSION: The risk of AMD development was mainly predicted by baseline age, and risk variants in CFH and AMRS2/HTRA1, while clinical and lifestyle factors, including diet, were influential in causing such transition from no disease to disease. As the disease progressed to more severe stages, other clinical risk factors such as smoking almost doubled its influence score, and clinical factors like diabetes and high blood pressure became more relevant. This approach enables personalized AMD risk scores, allowing targeted interventions to risk reduction by addressing modifiable risk factors.