Purchase this article with an account.
L. Feiner, J. Ehrlich, J. Ward, L. Tuomi; Classification and Regression Tree Analysis to Identify Predictive Factors for Visual Acuity Outcomes in Ranibizumab Trials. Invest. Ophthalmol. Vis. Sci. 2009;50(13):2358.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Guiding expectations for treatment outcome is a key aspect of patient counseling. Classification and Regression Tree (CART) analysis is one technique to identify relative similarities among individuals within large groups. In ophthalmology, CART analysis has previously been used to develop decision rules for glaucoma diagnosis. Here, we have utilized CART modeling to identify factors predictive of visual acuity outcomes in studies of ranibizumab for neovascular age-related macular degeneration (AMD).
Patient baseline characteristics (such as age, gender, study treatment arm, and CNV lesion type) were extracted from clinical trials databases. CART modeling algorithms were then applied using SAS and S+ software, with change in visual acuity (VA) at month 12 as the specified outcome of interest.
We have first applied CART modeling to identify features predictive of a >= 15 letter (3 line) gain in VA in the ANCHOR study of ranibizumab compared against Verteporfin/PDT. Overall, 27% (114/422) of patients experienced >= 15 letter gain in VA in ANCHOR. 106 of these patients were in ranibizumab treatment arms (0.3mg or 0.5mg). In those patients, baseline CNV size was the first branch point in one classification tree model, followed by baseline VA and age. We will also present CART models for features predictive of poor visual outcomes in ranibizumab clinical trials.
CART can be used to identify baseline factors associated with treatment outcomes. This can be informative for physicians and patients, although not all neovascular AMD patients may be similar to those in the clinical trials.
Clinical Trial: :
This PDF is available to Subscribers Only