Danielle L Weiler1, Carla B Engelke, Anna L O Moore, Wendy W Harrison. 1. *OD, FAAO †OD ‡OD, PhD, FAAO Southern Arizona Veterans Affairs Health Care System, Tucson, Arizona (DLW, CBE, ALOM); Southern California College of Optometry at Marshall B. Ketchum University, Fullerton, California (DLW, CBE, ALOM); and Midwestern University, Arizona College of Optometry, Glendale, Arizona (WWH).
Abstract
PURPOSE: Earlier identification of diabetic eye disease is an important research effort. Retinopathy is widely acknowledged but retinal vessel changes are not evaluated as stringently. Here, we create a multivariate model for the association between retinal vessel tortuosity (RVT) and other health factors in patients with diabetes. METHODS: Three hundred eyes of 150 patients with diabetes were included. Three investigators independently reviewed telemedicine fundus photographs and scored the level of diabetic retinopathy (DR) and RVT. These scores were evaluated for agreement and averaged. Also collected were age, duration of diabetes, presence or absence of diabetic nephropathy or neuropathy, blood pressure, total cholesterol, and hemoglobin A1c. A regression model evaluating the association of tortuosity with other factors was created. RESULTS: There was very high agreement between the three graders for level of DR (κ = 0.84). Agreement between the three graders for RVT varied substantially: poor for venous tortuosity (κ = 0.23) and fair for arteriole tortuosity (κ = 0.44) and overall gut tortuosity (κ = 0.42). The overall gut tortuosity was the most reproducible for the graders with a correlation coefficient of 0.923. There were univariate associations between arteriole tortuosity and venous tortuosity, DR level, and cholesterol. The selected best multivariate model found arteriole tortuosity to be associated with DR and cholesterol levels. CONCLUSIONS: First, RVT, particularly for venules, is difficult to grade consistently; therefore, future studies examining tortuosity should focus on arterioles. Second, the model indicates that there is an association between vessel changes, DR, and systemic cholesterol levels. Although DR and RVT are readily available to assess concurrently on a photograph, the addition of cholesterol to this model indicates that patients with RVT may warrant further follow-up on health factors, such as cholesterol levels.
PURPOSE: Earlier identification of diabetic eye disease is an important research effort. Retinopathy is widely acknowledged but retinal vessel changes are not evaluated as stringently. Here, we create a multivariate model for the association between retinal vessel tortuosity (RVT) and other health factors in patients with diabetes. METHODS: Three hundred eyes of 150 patients with diabetes were included. Three investigators independently reviewed telemedicine fundus photographs and scored the level of diabetic retinopathy (DR) and RVT. These scores were evaluated for agreement and averaged. Also collected were age, duration of diabetes, presence or absence of diabetic nephropathy or neuropathy, blood pressure, total cholesterol, and hemoglobin A1c. A regression model evaluating the association of tortuosity with other factors was created. RESULTS: There was very high agreement between the three graders for level of DR (κ = 0.84). Agreement between the three graders for RVT varied substantially: poor for venous tortuosity (κ = 0.23) and fair for arteriole tortuosity (κ = 0.44) and overall gut tortuosity (κ = 0.42). The overall gut tortuosity was the most reproducible for the graders with a correlation coefficient of 0.923. There were univariate associations between arteriole tortuosity and venous tortuosity, DR level, and cholesterol. The selected best multivariate model found arteriole tortuosity to be associated with DR and cholesterol levels. CONCLUSIONS: First, RVT, particularly for venules, is difficult to grade consistently; therefore, future studies examining tortuosity should focus on arterioles. Second, the model indicates that there is an association between vessel changes, DR, and systemic cholesterol levels. Although DR and RVT are readily available to assess concurrently on a photograph, the addition of cholesterol to this model indicates that patients with RVT may warrant further follow-up on health factors, such as cholesterol levels.
Authors: Lucía Ramos; Jorge Novo; José Rouco; Stephanie Romeo; María D Álvarez; Marcos Ortega Journal: BMC Med Res Methodol Date: 2018-11-20 Impact factor: 4.615