Only five randomized controlled trials (RCTs) have tested the effects of anticonvulsants for the primary prophylaxis of seizures in patients with newly diagnosed brain tumours. No studies have been disease-specific, and all included a mixture of both primary and secondary brain tumours. All were heterogeneous with respect to inclusion criteria and the anticonvulsants used in the study.
Anticonvulsants are problematic in brain tumour patients. The studies of Glantz et al and Forsyth et al. demonstrate that the rate and intensity of anticonvulsant-related side effects may be higher in patients with brain tumours when compared to patients with seizure disorders from other causes. In addition, the anticonvulsants tested in those RCTs were in the class of enzyme-inducing anticonvulsants and are expected to have pharmacodynamic interactions with other medications commonly used in the treatment of those patients. In particular, interactions between enzyme-inducing anticonvulsants and chemotherapy are of significant concern. No studies have tested the newer generation of anticonvulsants that, in general, are characterized by fewer adverse effects and minimal drug interactions.
Because those clinical trials were all small, it is not possible to determine if there are special subgroups of patients at increased risk of seizures. Intuitively, patients with tumours near the motor strip, with cortically based tumours, or hemorrhagic tumours may be at increased risk of seizures; however there are no data to support this assumption. Clinicians must therefore individualize treatment for those patients. Tumour-related factors such as location must be integrated with patient preferences and quality of life (QOL), in addition to concomitant medications, in order to reach treatment decisions.
From the larger RCTs and the meta-analysis, it is clear that conventional anticonvulsants (phenytoin, valproic acid, and phenobarbital) confer at best a 25% reduction in the risk of seizures, with the incremental risk of adverse effects and drug interactions. The Neurology Disease Site Group (DSG) felt that the statistical power of those trials reliably excluded a clinically important reduction in seizure risk for seizure-naïve patients with newly diagnosed primary and secondary brain tumours. Thus, the routine use of postoperative anticonvulsants is not recommended in those patients, especially in light of a significant risk of serious adverse effects and problematic drug interactions. This recommendation is in agreement with the American Academy of Neurology (AAN) practice parameter.
The newer anti-epileptic medications may overcome some of these issues but have not been tested in clinical trials. No recommendations can be made regarding the use of those medications; however, a randomized controlled clinical trial, informed by the analysis and the suggestions by Forsyth et al., should be considered.
In the survey of Ontario practitioners, 74% of respondents indicated that they did not recommend the routine use of anticonvulsants in seizure-naïve patients with newly diagnosed brain tumours. Thus, the current practice is consistent with the recommendation of this guideline. The pre-guideline survey is presented in Appendix 1 in the original guideline document, and the results of the two case scenarios are shown in Table 3 in the original guideline document. Respondents were asked to indicate which response most closely matched their usual practice.
For patients who are already on anticonvulsants but have never had a seizure, there is very little evidence to guide treatment. The AAN practice parameter recommends considering a taper and discontinuation of anticonvulsants; however, only one small retrospective clinical trial has attempted to address that issue. There is insufficient evidence to recommend for or against the tapering of anticonvulsants in this situation, and, therefore, treatment must be individualized. In the Ontario practice survey, essentially equal numbers of practitioners would maintain anticonvulsants, withdraw them, or have a discussion and allow a patient-based decision. Thus, current practice appears to reflect the lack of data that addresses this very specific question.