Purchase this article with an account.
Linda C. McCarthy, Paul Newcombe, John Whittaker, John Wurzelmann, Michael Fries, Nancy Burnham, Gengqian Cai, Sandra Stinnett, Trupti Trivedi, Chun-Fang Xu; Predictive Models Of Conversion To Choroidal Neovascularisation (CNV). Invest. Ophthalmol. Vis. Sci. 2011;52(14):6670.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Our aim was to develop models for predicting Age-Related Maculopathy (ARM) patients at high risk of progressing to CNV, and to assess their value in screening patients for participation in clinical trials of CNV prevention. In particular, this work addressed whether there is additional predictive value in the inclusion of genetic markers in the models, and if so, whether there is sufficient additional predictive value to make the generation of genetic data at screening cost effective.
Clinical, environmental and genetic factors were explored in the predictive modelling, which was conducted using data from 2039 subjects from the Age-Related Eye Disease Study (AREDS) study (dbGaP: phs000001.v2.p1). Genetic variants in the following genes were tested: CFH, ARMS2, C2, CFB, C3, and VEGF. Using simulations based on the AREDs data we explored and compared relative trial costs under clinical and genetic enrichment, across three trial lengths (2, 3 and 4 years).
There was very strong evidence for genetic effects on progression (p < 1E-7), with ARMS2 (A69S) (p=0.00002) and CFH (Y402H) (p=0.0002) demonstrating the strongest associations. Overall, the inclusion of genetics gave only marginal improvements in predictive performance compared to non-genetic models, e.g. as measured by the cross-validated AUC for 3 year progression to CNV (88% for the clinical + genetics model, vs. 87% for the clinical only model, vs. 75% for a genetics only model). Across all trial lengths it was estimated that including genetics in an enrichment design would result in no more than a 5% reduction in trial costs.
Despite strong evidence for genetic effects, inclusion of genetics adds little discriminatory power to predictive models for CNV conversion. Consequently, our simulations suggest genetic enrichment would lead only to a marginal increase in trial efficiency.
This PDF is available to Subscribers Only