Abstract
Friedrichs-Maeder C, Proix T, Tcheng TK, Skarpaas T, Rao VR, Baud MO. Ann Neurol. 2024;95(4):743–753
Objective: This study was undertaken to determine the effects of antiseizure medications (ASMs) on multidien (multiday) cycles of interictal epileptiform activity (IEA) and seizures and evaluate their potential clinical significance. Methods: We retrospectively analyzed up to 10 years of data from 88 of the 256 total adults with pharmacoresistant focal epilepsy who participated in the clinical trials of the RNS System, an intracranial device that keeps records of IEA counts. Following adjunctive ASM trials, we evaluated changes over months in (1) rates of self-reported disabling seizures and (2) multidien IEA cycle strength (spectral power for periodicity between 4 and 40 days). We used a survival analysis and the receiver operating characteristics to assess changes in IEA as a predictor of seizure control. Results: Among 56 (33.3%) of the 168 adjunctive ASM trials suitable for analysis, ASM introduction was followed by an average 50% to 70% decrease in multidien IEA cycle strength and a concomitant 50% to 70% decrease in relative seizure rate for up to 12 months. Individuals with a ≥50% decrease in IEA cycle strength in the first 3 months of an ASM trial had a higher probability of remaining seizure responders (≥50% seizure rate reduction, p < 10−7) or super-responders (≥90%, p < 10−8) over the next 12 months. Interpretation: In this large cohort, a decrease in multidien IEA cycle strength following initiation of an adjunctive ASM correlated with seizure control for up to 12 months, suggesting that fluctuations in IEA mirror “disease activity” in pharmacoresistant focal epilepsy and may have clinical utility as a biomarker to predict treatment response.
Commentary
Despite having more than 30 medications to choose from, each with differing mechanisms of action, selecting an efficacious anti-seizure medication still involves significant trial and error. This approach can be very frustrating for both clinicians and patients alike because it requires substantial time and may result in several failures. Failing to achieve seizure control promptly places patients at risk of morbidity and mortality.
Determining medication success relies on patient self-reporting of seizures and monitoring for a duration sufficient to exclude normal fluctuations in seizure patterns. 1 Unfortunately, we have not yet identified clinically useful markers of treatment response that enable us to personalize treatment decisions and determine when to switch to a different medication.
Recently, there have been efforts to use surface EEG (Pharmaco-EEG) to predict treatment response with variable success. 2 It remains to be seen whether such an approach can be successful since the surface EEG data is cross-sectional and limited. On the other hand, longitudinal data from patients with responsive nerve stimulation (RNS) have provided us insights into epilepsy, revealing that interictal epileptiform activity (IEA) in persons with epilepsy oscillates with person-specific multidien (multi-day) periods. 3 Additionally, modeling this cyclical IEA activity enables forecasting of seizures up to 3 days in advance. 4 The current study 5 leverages this knowledge to try to predict response to anti-seizure medications and create an RNS crystal ball.
The authors analyzed continuous data from patients implanted with the RNS system to determine the impact of ASMs on IEA cycles and whether this predicts seizure control. The cohort consisted of 88 out of 256 adult persons with epilepsy enrolled in the neuropace trial, with data spanning from 2004 to 2018. Patients with low seizure counts, absence of clear IEA multidien rhythms, and very slow (>1.5 months) titrations of medications such as lamotrigine were excluded.
A total of 168 ASM trials in these 88 individuals were analyzed, with multiple trials allowed per individual if there was at least a 3-month follow-up. The mean age of the cohort was 35.2 years, 46.6% had bilateral ictal onsets, and 75% had temporal lobe epilepsy, 67.1% had mesial temporal lobe epilepsy, with an average seizure frequency of 7.3 per month. The most commonly analyzed medication trials were Lacosamide (36), Clobazam (28), and Levetiracetam (20). Patients were categorized as responders or super-responders if there was seizure reduction ≥50% or ≥90%, respectively. The primary focus was on focal seizures with impaired awareness, and focal to bilateral tonic-clonic seizures based on seizure diaries. The RNS data was modeled to extract the dominant multidien power, that is, IEA cycle strength, between 4 and 40 days, and then the phase-locking value of seizures to the IEA cycle. The primary predictor of interest was an IEA response defined by a ≥50% power decrease in the IEA cycle strength, that is, how successful was the medication in suppressing the IEA cycle power between months 1 and 2 after achieving a therapeutic dose or last medication adjustment. The phase locking value of seizures reflects the timing of seizures in relation to the IEA cycle.
A ≥50% seizure reduction at 3 months was seen in 64.3% in the presence of an IEA response versus 27.7% in the absence of a response, while 26.8% of super-responders had an IEA response versus 3.6% of those who did not. Using this IEA response measure at 2 months was enough to predict responders up to 12 months out, and this predictor was better than using absolute IEA counts alone. The sensitivity of the IEA response as a predictor was 57%, with a specificity of 81%. Interestingly the ASM did not affect the phase coupling of seizures to the IEA cycle, so the days in the cycle when the individual was most likely to have a seizure remained the same, it was only the likelihood of seizure occurrence that was affected. Of note, the findings persisted even after controlling for changes to the RNS settings such as detection parameters and stimulation parameters. There was also no effect of ASM mechanism of action or dosage.
There is much to learn from this paper. First, this paper re-emphasizes the utility of conceptualizing and modeling epilepsy as a cyclic disease, with most individuals having individual patterns of IEA cycles and seizure cycles. Understanding these cycles will allow us to forecast seizure risk and in this case medication responsiveness. Second, characterizing IEA burden and cycles from the EEG can guide treatment. In the surface EEG world, monitoring the frequency and duration of IEAs in generalized epilepsy is a good indicator of ASM response. 6 We also know that in the epilepsy monitoring unit, withdrawal of ASMs increases IEA burden and increases seizure risk. 7 However, efforts in focal epilepsy have largely failed because of undersampling and the poor resolution of EEG for mesial and deep structures. The RNS system bypasses this issue and provides continuous data. The study findings add fuel to the ongoing debate of whether we should be chasing IEAs with medications and “treating the EEG.” Will these persistent IEAs that did not respond to the medication affect sleep, cognition, and mental health? These would have also been outcomes of interest. Third, the findings are significant because even in a refractory population with several failed medication trials and an implanted neurostimulator, treatment response could still be predicted. Similarly, observations of prolonged remission periods in up to a third of patients with drug-resistant epilepsy have been documented with medication changes alone. This highlights the dynamic nature of drug-resistant epilepsy and challenges the attribution of these “honeymoon periods” to a specific intervention. 8
There are several caveats that must be kept in mind; the prediction model was only applied to patients who did not have clear multidien rhythms and who did not have low seizure counts. The data did not include more recent medications such as cenobamate which has been found to have good efficacy in refractory epilepsy, 9 and most patients had mesial temporal epilepsy. Is there also a synergistic effect of RNS and a specific medication, which might not apply to the population without RNS?
The sensitivity of the IEA response as measured was low, so there were several ASM responders who exhibited a clinical response but not an IEA response. The authors posit that ASMs may prevent seizures through different mechanisms, one of them being increasing the seizure threshold without necessarily suppressing IEAs. It would have been interesting to see whether this predictor also predicted the number of long episodes on RNS rather than patient self-reported seizures since these are known not to always be reliable. Finally, the most obvious limitation of this study is that it could only be applied in patients with an implanted device, and the time frame needed to predict medication responsiveness at best was 2 months. This time frame is still not ideal for the clinician tailoring medications, but at least it is a step in the right direction in terms of forecasting ASM responsiveness in focal epilepsy. Prior studies using a machine learning approach incorporating clinical, noninvasive EEG, and neuroimaging data achieved modest discrimination values of 0.65 when predicting seizure freedom. 10 Findings from this study are comparable to seizure response in a refractory patient population.
Ultimately, we might have to incorporate several other measures to predict ASM responsiveness quicker and more accurately. Factors such as genetics, underlying pathology, circadian rhythm, and other biological factors in addition to the EEG will likely be needed. If only we could obtain continuous and reliable EEG data without requiring brain surgery.
