Abstracts

Rationale:
Drug-resistant epilepsy (DRE), defined as incomplete seizure control after adequate trials of two or more antiseizure medications, affects approximately one third of individuals with epilepsy and is associated with reduction in quality of life, increased hospital admissions and resource utilization. Although a recent meta-analysis found epilepsy surgery to be more effective in controlling seizures in individuals with focal DRE than medical management, epilepsy surgery remains an underutilized treatment option. Limiting factors may be access to specialized centers and misconceptions about its risks and benefits. Our aim was to facilitate medical decision making in patients with drug-resistant focal epilepsy by developing an algorithm based on a mathematical model for sequential decision processes.
Method:
To simulate the sequential decision process when deciding about medical treatment versus epilepsy surgery we used a robust Markov decision process (MDP) model which analyzes average “cost”, i.e. likelihood of adverse outcome, depending on action taken. Input variables were possible states (examples are “seizure free”, “incomplete response”, “no response”, “adverse effects”), and possible actions (examples are “add medication”, “stop medication”, “presurgical evaluation”, “surgery”) were considered for each state. Sensitivity analysis was performed to determine thresholds for changes in recommended actions based on varying outcome probabilities. Outcome measures were based on quality adjusted life years (QALYs), and likelihood of adverse outcome was based on medical literature.
Results:
As expected, the algorithm recommended adding a second anti-seizure drug if response to the first drug was incomplete and risk of adverse effects was low (< 20%). Presurgical evaluation plus adding a third anti-seizure drug was the optimal action if a second drug failed to control seizures completely, and epilepsy surgery was deemed superior to medical treatment in surgical candidates who did not respond to three or more drugs if probability of adverse outcome was < 23%.
Conclusion:
The MDP model we developed provides a mathematical algorithm that recommends continued medical treatment versus epilepsy surgery based on outcome. This may help facilitate decision making when considering referrals for epilepsy surgery evaluation and deciding about epilepsy surgery.
Funding:
:n/a