Using existing data to optimise adaptive intervention in epilepsy

Approximately half of patients with newly diagnosed epilepsy who start antiepileptic drug (AED) monotherapy fail on their first treatment, after which the clinician and patient will discuss the likely overall prognosis and decide on the next AED to start. Failure on the second AED may lead to further substitution of AED as monotherapy, or the addition of further AEDs as polytherapy, with the potential for a large number of alternative treatment sequences and possible adaptations. The probability of achieving seizure control reduces with every failed AED and approximately 30% of patients will have a drug-resistant form of epilepsy. Traditionally trial designs for epilepsy focus on randomising newly diagnosed patients or drug-resistant patients as separate populations, and there is a dearth of evidence to support adaptive treatment decisions between these two points in an individual patient's pathway.
The Sequential Multiple Assignment Randomised Trial (SMART) is an innovative clinical trial design, to provide high-quality data that can be used to inform the development of adaptive interventions across the pathway. A SMART involves multiple intervention stages where each stage corresponds to one of the critical decisions involved in the adaptive intervention and randomises participants at each state. Although such a SMART trial would be the gold standard for causal inference, the approach would be expensive and lengthy. Utilising existing real-world data by applying analytic techniques to emulate a SMART target trial could be an efficient use of currently available data and evidence. Personalising optimal pathways may also be possible by incorporating data about patient characteristics where this is available.
The aim is to establish how real-world evidence can be used to optimise treatment decisions for patients with epilepsy and inform the design of a future SMART trial. Specific aims include:
1. Summarise the evidence for treatment decisions following treatment failure.
2. Identify risk factors for outcome following treatment failure in patients with epilepsy
3. Evaluate how previous SMART trials have used existing evidence in their design and analysis
4. Implement an analytic approach that emulates a target SMART trial of AED treatments using existing real-world data
5. Design a SMART trial to explore the most promising treatment sequences in epilepsy incorporating existing evidence