Purchase this article with an account.
Eveline Kersten, Yara Lechanteur, Nicole T.M. Saksens, Tina Schick, Sascha Fauser, Anneke I Den Hollander, Carel C B Hoyng; Validation of a prediction model in families with age-related macular degeneration. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):794.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Prediction models for age-related macular degeneration (AMD) are based on common environmental and genetic risk factors. In AMD families these risk factors may be distributed differently and rare genetic variants with a high risk for disease have been identified. This raises the question if these models can be used in AMD families. We created a prediction model based on a case-control dataset. The aim of this study is to evaluate the validity of this model in an AMD family dataset.
A total of 1156 advanced AMD cases and 1393 controls enrolled in the European Genetic Database (EUGENDA) were included in this study. The dataset contained a family set of 119 advanced AMD cases and 103 controls from 96 AMD families. The remaining individuals were randomly subdivided into a training set (816 cases, 984 controls) and a validation set (221 cases, 306 controls). A prediction model for the development of advanced AMD was created in the training set using logistic regression analyses. Subsequently, validation of this model was performed using receiver operating characteristics curves in the validation and the family set.
The final model included the factors age, smoking, physical exercise, education, family history and seven single nucleotide polymorphisms in CFH, ARMS2, CETP, VEGFA and TIMP3. This model showed an area under the curve (AUC) of 0.873 (95% CI 0.851-0.895) in the training set. The AUCs in the validation and the family set were 0.854 (95% CI 0.808-0.900) and 0.842 (95% CI 0.787-0.897), respectively. Further analyses of individual families showed clustering of risk factors in most families, with predicted values above 0.8 for all affected individuals. We also identified several densely affected families with low predicted risk scores for the individual affected family members.
In general, these results indicate that our model is able to estimate risks of developing advanced AMD in families based on clustering of known risk factors. However, in a subset of densely affected families the risk of advanced AMD is not explained by these risk factors. This suggests that in these families other factors, such as rare genetic variants, play a role in the development of advanced AMD. Prediction models can therefore aid in the identification of families that are of interest for additional genetic testing.
This PDF is available to Subscribers Only