Related Topics
 
Dr. Arzu Onar

Seizure recurrence and risk factors after anti-epilepsy drug withdrawal in brain tumor patients

In this article we studied seizure outcome after antiepilepsy drug (AED) withdrawal in pediatric brain tumor patients and explored risk factors for seizure recurrence. The data set included brain tumor patients treated during January 1985 and December 2004 at St Jude who had at least one unequivocal witnessed seizure. A large number of variables were prospectively identified to be explored as risk factors for seizure recurrence with the intent to be used in multivariable models. Based on standard statistical approaches (e.g. Log-rank tests and single-variable Cox proportional-hazards models) two multivariable Cox models were identified: the first model was developed to generate hypotheses for future studies and attempted to maximize the number of variables that could reasonably be supported by the data. The second model was developed to identify variables with the most robust effects by emphasizing parsimony and applying stricter variable-selection criteria, which included covariate-specific sample-size considerations, monitoring of changes to diagnostics with each candidate variable, and Bonferroni-adjusted p-values in the final model.

Guided by the single-variable analyses, both multivariable models were developed by using exhaustive step-wise methods. In addition to individual factor effects, all two-way interactions were also explored. Ridge-regression methods were used to reduce the effects of correlations among the risk factors, and robust variance estimates were used to reduce the effects of outliers. The Cox proportional-hazards assumption was verified with log-minus-log plots.

Our analyses indicated that more than one tumor resection and whole-brain radiation treatment might be associated with seizure recurrence. Further posterior fossa tumor location was correlated with reduced seizure recurrence risk in our models. Interestingly in the subgroup of patients who had an EEG before AED withdrawal, we did not find any association between spikes or slow waves and seizure recurrence.

This article appears in Epilepsia 2006. The co-first author is Raja Khan (Baptist Hospital-Memphis).