November 2011 Seminar 'Prognostic Factors for Epilepsy'
- School of Mathematics Statistics Group, Seminar room 310, University of Birmingham, Watson Building
- Medical and Dental Sciences
'Prognostic Factors for Epilepsy'
Prognostic factors identify patients at varying degrees of risk for specificoutcomes which facilitates treatment choice and aids patient counselling. Usingdata from clinical trials of epilepsy patients the influence of prognosticfactors was considered to build prognostic models for time to eventoutcomes.
A list of potential prognostic factors was established and for each,univariate and multivariate analyses used log-rank tests and Cox models wereundertaken. Variables were centred to diminish multicollinearity and continuousones were investigated by log or fractional polynomial transformations. Parsimonious multivariate models were produced with variable reduction byAkaike's Information Criterion. Multivariable models identified a number ofsignificant factors. Evidence of treatment-covariate interactions was alsoidentified. Subject to external validation, these prognostic models should aidindividual patient risk stratification and the design and analysis of futureepilepsy trials.
Department of Biostatistics,
University of Liverpool