Real-Time Management of Seizures
Efforts have been made for decades to predict when seizures will occur and provide an immediate intervention to either prevent or terminate the seizure.69 These efforts rely on a range of recording devices, computational algorithms to identify at-risk periods, and active response in the form of electrical/optical stimulation or administration of a drug. Although there has been steady progress with these strategies over the decades, many approaches are maturing to the point where they seem poised to provide a workable and effective therapy for a larger number of patients.70 Indeed, the introduction in 2013 of an FDA-approved closed loop device that detects seizures and aborts them by deep brain stimulation has spawned many efforts to refine stimulation parameters for better seizure control.71 New seizure prediction algorithms8 as well as new devices may allow intravenous injection or even direct infusion of antiseizure agents into the brain at the onset of or immediately before a seizure is predicted.72 This approach has the capacity to harness the utility of proven pharmacological treatments without the side effects of chronic exposure to drug in blood and brain. Taken a step further, introduction of active drug locally into the epileptic focus could provide more selective treatment of certain epileptic conditions, including refractory epilepsies, localization-related epilepsies, and status epilepticus. Increasing power of computational algorithms7 should allow enhanced ability to predict seizures from multiple streams of data, including electrical recordings and peripheral readouts. In addition, miniaturization of devices can improve the ability to deliver electrical, light, or pharmacological stimuli to specific regions both inside and outside of the central nervous system. The emergence of new animal models of genetic epilepsies provides another opportunity to test detection and seizure interruption strategies in homogeneous models that share some basis with human epilepsy and thus might provide robust data that can be translated to patients harboring these variants. A range of models might stimulate improvements in the low signal to noise ratio in seizure prediction and in the abortion of seizures, such as evaluation of new biomarkers that change prior to seizure initiation73 and consideration of circadian rhythms.74 Ultimately, these systems need to be suitable for self-management in the home and other nonmedical settings in order to improve adherence and efficacy.
Taken to its logical albeit futuristic conclusion, one might envision a paradigm shift from ASMs in the form of multiple doses of a drug per day and steady-state blood levels (with attendant side effects) to delivery systems that provide anticonvulsants to the brain at the site they are needed and only when they are needed, improving the quality of life of the patient. Further, each patient’s treatment would be customized based on genetic and molecular profiles. This form of precision medicine would eliminate the need for chronic and systemic nonspecific and side effect-laden pharmacotherapy, improving efficacy and possibly reducing the development of pharmacoresistance.