Time to prerandomization seizure count design sufficiently assessed the safety and tolerability of perampanel for the treatment of focal seizures

Abstract

Objective

In traditionally designed randomized clinical trials of antiseizure medications, participants take a blinded treatment for a prespecified number of weeks, irrespective of continued seizures. The alternative design time to prerandomization monthly seizure count (T‐PSC) allows participants to end the blinded treatment after an individually prespecified number of seizures, which shortens exposure to placebo and ineffective treatment. Previous reanalyses have shown that T‐PSC replicated the efficacy conclusions of trials; therefore, we evaluated whether T‐PSC also could replicate tolerability and safety conclusions.

Methods

We retrospectively applied the T‐PSC design to analyze treatment‐emergent adverse events (TEAEs) from three blinded, placebo‐controlled trials of perampanel for focal onset seizures (NCT00699972, NCT00699582, NCT00700310). We evaluated the incidences of TEAEs, treatment‐related TEAEs, serious TEAEs, and TEAEs that prompted medication adjustment compared to those observed during the full‐length trial.

Results

Of the 1480 participants in the three trials, 1093 experienced any TEAE, of whom 1006 (92%) had onset prior to T‐PSC. When evaluating the differences in each type of TEAE for each dose of perampanel from placebo within each trial, there was no consistent pattern of under‐ or overestimation. Across the three studies, 23 of 79 (29%) serious TEAEs, most requiring hospitalization, occurred after T‐PSC.

Significance

Almost all TEAEs occurred before T‐PSC. Similar conclusions regarding the tolerability and safety of perampanel would have been reached if the T‐PSC design had been used. This suggests that the T‐PSC design may potentially benefit participants by allowing earlier change from an ineffective treatment to an alternate treatment, which could reduce the risk of serious consequences of ineffective treatment, such as hospitalization.

Key points.

• Use of the T‐PSC design would have resulted in almost identical tolerability and safety conclusions as the full‐length trials.

• Almost all (92%) of any TEAEs occurred before T‐PSC.

• In the subset of serious TEAEs, 29% were avoided, as they occurred after T‐PSC.

• The difference in rate of each TEAE between perampanel and placebo was not consistently over‐ or underestimated.

• A time to TEAE analysis in each dose of perampanel versus placebo also did not identify systematic over‐ or underestimation.

1. INTRODUCTION

Placebo‐controlled randomized clinical trials (RCTs) of adjunct antiseizure medications (ASMs) are designed to rigorously evaluate efficacy, tolerability, and safety. ¹ Traditionally designed trials include periods of screening, prerandomization baseline, titration, and maintenance, each of which has a prespecified duration irrespective of participants' seizure frequency. ¹ However, participants randomized to placebo experience increased risks due to ineffective seizure control including, but not limited to, a 5.8 times increased risk of sudden unexpected death in epilepsy (SUDEP). ¹ , ² These risks may contribute to the increasing challenges in recruiting participants for RCTs, where sites recently recruited an average of fewer than four participants. ³ , ⁴ This observation and other factors have prompted development of modified trial designs that evaluate efficacy, tolerability, and safety similar to the traditional designs, while also reducing placebo exposure. ¹

An alternative design of time to prerandomization monthly seizure count (T‐PSC) appears to accomplish each of those goals. ¹ , ⁵ In a T‐PSC design trial, the prerandomization baseline period defines a participant's prerandomization monthly seizure count (PSC; e.g., 6 seizures/month). ¹ , ⁵ After randomization and completion of the titration period, participants receive blinded treatment until (1) their postrandomization seizure count exceeds their PSC, (2) they reach the traditional duration of maintenance, or (3) they withdraw from the trial for another reason. After the participant reaches T‐PSC, they have completed the blinded phase and could enter the open label phase so that they may try the study drug on a known dose or choose to pursue other noninvestigational treatments. When more than 20 previous trials of ASMs were reanalyzed as if they had used the T‐PSC design, the efficacy conclusions of the trials were unchanged, without substantial evidence of reduced statistical power, despite the trials being 40% shorter and including 60% fewer seizures. ¹ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ Importantly, the T‐PSC design differs from designs with shorter maintenance periods (which have reduced statistical power ¹² ) because, in the T‐PSC design, responders to blinded treatment may complete the full traditional trial. The full duration of the maintenance period evaluates sustainability of efficacy and differentiation of long‐term ASMs from rescue medications like short‐acting benzodiazepines. The T‐PSC design also differs from early stopping or adaptive designs that limit the number of participants recruited and alter randomization ratio based on interim statistical analyses. Because of inclusion of fewer participants, those early stopping designs acquire less efficacy, tolerability, and safety data. In contrast, the T‐PSC design aims to maintain the quality of the efficacy, tolerability, and safety evaluation.

To date, only one trial for regulatory approval (lacosamide for primary generalized tonic–clonic [PGTC] seizures) used a related time to second seizure design. ¹³ A main concern about the T‐PSC design is that it may reduce the placebo‐controlled evaluation of tolerability and safety. ¹ , ¹⁴ The full assessment of tolerability and safety requires long‐term open‐label and postmarketing studies, but the blinded period of placebo‐controlled trials is the only time that rigorous prospective placebo‐controlled data are acquired regarding treatment‐emergent adverse events (TEAE). As suggested by the variability of placebo efficacy across trials, ³ , ¹⁵ , ¹⁶ concurrently collected placebo‐controlled information remains beneficial to address unanticipated and unmeasurable sources of variability.

Despite concern about shortening of tolerability and safety data, empiric evidence suggests that the T‐PSC design likely is sufficient to evaluate tolerability and safety. By design, the T‐PSC design encourages but does not require at least 1 month of maintenance observation. In many previous trials that reported the timing of TEAEs, the vast majority of TEAEs occurred during titration or the first month of maintenance. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² In our first analysis of tolerability and safety with the T‐PSC design, we reanalyzed the trial of perampanel for PGTC seizures to demonstrate that T‐PSC was sufficient for evaluation of tolerability and safety. ²³ However, there was concern than the T‐PSC design may overestimate the difference between active treatment and placebo by accurately measuring TEAE rate for active treatment, whereas the shorter T‐PSC for placebo may underestimate the TEAE rate for placebo.

In this study, we aimed to replicate and build upon that previous work through reanalysis of three placebo‐controlled randomized clinical trials of perampanel for focal onset seizures to evaluate the hypothesis that T‐PSC was sufficient to observe tolerability and safety. Building upon that previous analysis also is important because there are fundamental differences between PGTC seizures and focal onset seizures. A previous analysis of these three trials of perampanel for focal onset seizures demonstrated that almost all TEAEs occurred within the first 6 weeks of treatment (titration), ²⁴ but this timing was not compared to PSC. In addition to evaluating the T‐PSC design, this study provided further clinically relevant information on when adverse events occur during treatment with active treatment or placebo.

2. MATERIALS AND METHODS

2.1. Trial protocol information

This is an exploratory analysis of three previously published multicenter, double‐blind, randomized placebo‐controlled trials (Eisai study protocols: E2007‐G000‐304, E2007‐G000‐305, E2007‐G000‐306, abbreviated as Study 304, Study 305, and Study 306, respectively; ClinicalTrials.gov identifiers: NCT00699972, NCT00699582, NCT00700310). ²⁵ These trials were performed in accordance with the Declaration of Helsinki, Good Clinical Practice International Council for Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use 6th Edition (ICH‐E6) Guideline CPMP/ICH/135/95, European Good Clinical Practice Directive 2005/28/EC, Clinical Trial Directive 2001/20/EC, and US Code of Federal Regulations Part 21. Before participating, all participants gave written informed consent. We describe the details of the trial in brief here. For additional details regarding the trials, please refer to the original publications of the primary efficacy, safety, and tolerability endpoints. ²⁶ , ²⁷ , ²⁸ , ²⁹

Many design aspects of the three trials were shared. These trials included participants aged 12 years and older diagnosed with partial onset seizures according to the 1981 International League Against Epilepsy (ILAE) classification of epileptic seizures and the 1989 ILAE classification of epilepsies and epileptic syndromes. Partial onset seizures correspond to focal onset seizures based on the 2017 ILAE classification of seizures. ³⁰ Participants were required to have at least five focal unaware seizures with or without bilateral tonic–clonic seizures per month during prerandomization baseline and to be taking stable doses of 1–3 approved ASMs. The prerandomization phase comprised screening (≤4 weeks) and baseline (4–8 weeks) periods. Studies 304 and 305 included placebo, perampanel 8 mg daily, and perampanel 12 mg daily randomized 1:1:1. Study 306 included placebo, perampanel 2 mg daily, perampanel 4 mg daily, and perampanel 8 mg daily randomized 1:1:1:1. The subsequent randomization phase comprised titration (weeks 1–6) and maintenance (weeks 7–19) periods. During randomization, participants, investigators, site personnel, and sponsor staff were blinded to treatment.

During titration, participants in the perampanel group received an initial daily dose of 2 mg, before uptitration in weekly 2‐mg increments to the targeted daily dose. Participants entered the maintenance period at the last dose achieved during titration. Dose adjustment during the maintenance period was not recommended; however, according to the investigator's clinical judgment, participants who experienced intolerable TEAEs could have their dose decreased by one 2‐mg increment.

2.2. Exploratory endpoint assessment

This work evaluates the exploratory endpoint of the time to exceed the prerandomization monthly seizure count (T‐PSC). ⁶ The prerandomization monthly seizure count, n, was calculated as the number of seizures during screening and baseline divided by the duration of screening plus baseline multiplied by 28 days. Postrandomization, T‐PSC was the days from the first dose of maintenance to the day of the (n + 1)th seizure plus 1 day.

To evaluate the ability of T‐PSC to measure the tolerability and safety of perampanel during the blinded phase of trials, we simulated the T‐PSC design by truncating data after T‐PSC. Due to the complexity of combining results across three trials, we focus our analysis on any TEAEs, treatment‐related TEAEs, TEAEs requiring dose adjustment, serious TEAEs, and TEAEs that occurred often enough to appear on the package label (5% or more in any perampanel dose and, if <5% in perampanel then >1% more common in perampanel than placebo). There have been multiple statistical analyses of TEAEs using full‐length trial data. ²⁵ , ³¹

To allow for rigorous comparison between trial conclusions with the simulated T‐PSC design compared to the full‐length trial, we evaluated the relative incidence of each category of TEAE between each dose of perampanel and placebo with Fisher exact tests. We also evaluated whether each category of TEAE was more likely to occur prior to T‐PSC in placebo versus perampanel. In addition, we evaluated the time to any TEAE with Cox proportional hazards analyses. To evaluate the time to any TEAE before T‐PSC, we also used cumulative incidence function (CIF) based on the Fine–Gray method. CIF accounts for the competing risk of withdrawal due to T‐PSC in the evaluation of TEAE rate. ³² Due to reduced efficacy of placebo, T‐PSC was shorter for participants randomized to placebo than perampanel. ⁵ These survival analyses were best suited for events that occurred in at least 10% of participants, so we did not perform similar analyses for subcategories of TEAEs.

3. RESULTS

The participants included in this exploratory analysis matched the participants who were included in the primary safety analysis (see original publications for demographics and clinical information). ²⁶ , ²⁷ , ²⁸ , ²⁹ Of 1491 randomized participants, 1480 were included (placebo = 442; perampanel 2 mg = 180, 4 mg = 172, 8 mg = 431, 12 mg = 255). Of these 1480 participants, 1363 (93%) completed the 6‐week titration phase.

The seizure efficacy outcomes and details of T‐PSC were described previously. ⁵ Briefly, the median T‐PSC on placebo was 37 days in Study 304, 30 days in Study 305, and 43 days in Study 306. The median T‐PSC for perampanel 2 mg daily in Study 306 was 43 days. The median T‐PSC for perampanel 4 mg daily in Study 306 was 56 days. The median T‐PSC for perampanel 8 mg daily was >100 days in Study 304, 43 days in Study 305, and 91 days in Study 306. The median T‐PSC for perampanel 12 mg daily was >100 days in Study 304 and 48 days in Study 305.

The rates of the categories of TEAEs are illustrated in Figures 1, 2, 3 (see Table S1 for exact numbers). Of the 1093 participants who experienced any TEAEs in the full‐length trial, 1006 (92%) had onset before T‐PSC (by Study 304: 318/340 [94%], Study 305: 290/309 [94%], Study 306: 398/444 [90%]). Although T‐PSC was shorter in placebo, the proportion of any TEAEs that occurred prior to T‐PSC was the same for placebo and any dose of perampanel for Studies 305 and 306. In Study 304, 86% of any TEAEs reported on placebo occurred before T‐PSC, compared to 97% on perampanel (p = .001). In each trial, treatment‐related TEAEs were more common with perampanel 4 mg, 8 mg, or 12 mg than placebo, both based on full‐length trial data and based on data prior to T‐PSC (see Table S1). Treatment‐related TEAEs were not different on perampanel 2 mg compared to placebo using data prior to T‐PSC or full‐length data. The differences in rate of TEAEs that prompted dose adjustment, and the rate of each action taken (withdrawal, reduction, or interruption), were the same when compared with full‐length data and data prior to T‐PSC, with two exceptions (Table S1). In Study 304, the rate of interruption of 12 mg versus placebo had a p‐value of .12 with full‐length trial data and .062 with data prior to T‐PSC. In Study 305, the rate of interruption of 8 mg versus placebo had a p‐value of .055 with full‐length trial data and .236 with data prior to T‐PSC.

FIGURE 1.

The rate of any and of any treatment‐related treatment‐emergent adverse events (TEAEs) observed before time to exceed prerandomization monthly seizure count (T‐PSC; darker overlay) was comparable to the full‐length trial (colors). Symbols (*, +) denote statistical significance between perampanel and placebo groups based on Fisher exact tests (p < .05); + indicates significance using data prior to T‐PSC only, and * indicates significance using full‐length trial data.

FIGURE 2.

The rate of serious or severe treatment‐emergent adverse events (TEAEs) observed before time to exceed prerandomization monthly seizure count (T‐PSC; darker overlay) was comparable to the full‐length trial (colors). Symbols (*, +) denote statistical significance between perampanel and placebo groups based on Fisher exact tests (p < .05); + indicates significance using data prior to T‐PSC only, and * indicates significance using full‐length trial data.

FIGURE 3.

The rate of each specific treatment‐emergent adverse events (TEAEs) observed before time to exceed prerandomization monthly seizure count (T‐PSC; darker overlay) was comparable to the full‐length trial (colors). Symbols (*, +) denote statistical significance between perampanel and placebo groups based on Fisher exact tests (p < .05); + indicates significance using data prior to T‐PSC only, and * indicates significance using full‐length trial data.

The comparison of incidence of all TEAEs in perampanel versus placebo was more complex. With a time to all TEAEs approach (Figure 4), there was a significantly increased rate of all TEAEs on perampanel 8 mg and 12 mg and no significant difference on perampanel 2 mg, each versus placebo (also listed in Figure 1, Table S2). In Figure 4, the colored censoring lines reflect when a participant reached T‐PSC before they had a TEAE. The green censoring lines indicate that they had no TEAE that started after T‐PSC. The red censoring lines indicate that there was at least one TEAE that started after T‐PSC, which therefore would have been excluded from the safety and tolerability analysis if the trial had used a T‐PSC design. We evaluated whether that truncation of data at T‐PSC would impact the comparison between perampanel and placebo. In the intermediate perampanel 4 mg dose of Study 306, the hazards ratio of time to all TEAEs in perampanel versus placebo was 1.3 (95% confidence interval [CI] = .97–1.66, p = .08) based on full‐length data, but it was 1.2 (95% CI = .91–1.6, p = .20) before T‐PSC. There was not a substantial correction of this difference with the competing risk approach (hazard ratio = 1.2, 95% CI = .91–1.61, p = .18; Figure S1, Table S2).

FIGURE 4.

The time to any treatment‐emergent adverse events (TEAEs) after the first treatment dose showing the almost identical results from the full‐length trial data as compared to TEAE data censored, or truncated, at the time to exceed prerandomization monthly seizure count (T‐PSC). The large vertical dashed line reflects the transition between titration (first 6 weeks) and maintenance treatment periods. Smaller vertical lines reflect when data were censored from this time to TEAE analysis. Smaller vertical lines reflect when data were censored from this time to TEAE analysis. Each small vertical line reflects one participant, and some overlap. Black censoring lines indicate study withdrawal unrelated to T‐PSC or TEAE. Colors reflect censoring of TEAE information due to simulated truncation at T‐PSC, where green indicates no TEAE occurred after T‐PSC and red indicates a TEAE started after T‐PSC.

When analyzing the incidence of all TEAEs, irrespective of timing, there were no significant differences between any perampanel dose for any study versus placebo based on data prior to T‐PSC, with one exception. A higher proportion of all TEAEs were observed before T‐PSC during Study 304 for perampanel (8 mg: 98%, p = .0005; 12 mg: 95%, p = .031) as compared to placebo (86%). The impact of that difference in observation was that the rate of all TEAEs between placebo and perampanel 8 mg was not significantly different with full‐length data (8 mg: 88%, placebo: 83%, p = .29), but it was prior to T‐PSC (8 mg: 86%, placebo: 71%, p = .0032). For perampanel 12 mg, the rate of all TEAEs was significantly higher using both full‐length data and data prior to T‐PSC (full‐length: 92%, p = .037; T‐PSC: 87%, p = .0017). The proportion of all TEAEs that were observed before T‐PSC during Studies 305 and 306 were not different when comparing perampanel to placebo (Study 305: 94% placebo, 93% 8 mg, 95% 12 mg; Study 306: 90% placebo, 86% 2 mg, 87% 4 mg, 94% 8 mg).

One of the primary goals of the T‐PSC design is to reduce the rate of serious TEAEs that occur during the blinded phase of trials by limiting exposure to ineffective treatment. Of the 79 participants who experienced any serious TEAEs in the full‐length trial, 23 (29%) had onset after T‐PSC, which may have been avoided if treatment had been changed by transitioning to open label extension or pursuing treatments outside the context of the trial (results by Study 304: 7/23 [30%], Study 305: 8/29 [28%], Study 306: 8/27 [30%]). All serious TEAEs that occurred after T‐PSC required hospitalization. There were significantly more serious TEAEs after T‐PSC in Study 306 for any perampanel dose versus placebo (perampanel: 44% [8/18] after T‐PSC, placebo: 0% [0/9] after T‐PSC, p = .03). These serious TEAEs that occurred after T‐PSC were evenly split between each perampanel dose (before T‐PSC/full‐length: 2 mg 33% [2/6], 4 mg 50% [3/6], 8 mg 50% [3/6]). In Studies 304 and 305, there was no difference observed (p > .50) in the proportion of serious TEAEs that occurred after T‐PSC in perampanel versus placebo. No deaths occurred in any of these trials.

When evaluating the incidence of specific types of TEAE, the overall conclusions matched at T‐PSC compared to the full‐length trial for almost all TEAEs listed on the US Food and Drug Administration package insert, with two exceptions. In Study 305, fatigue occurred more often in perampanel 12 mg than placebo before T‐PSC (p = .041), and that difference did not cross the p < .05 threshold when calculated with full‐length data (p = .054). Fatigue was not significantly different between perampanel and placebo in any of the other trials. In Study 304, there was greater weight increase with perampanel 8 mg versus placebo based on full‐length data (p = .038), but that difference was not seen based on data before T‐PSC, nor was it seen for other doses or other trials.

4. DISCUSSION

The evaluation of tolerability and safety of perampanel for focal onset seizures utilizing the T‐PSC design was comparable to the full‐length blinded trials. Specifically, 92% of any TEAEs occurred prior to T‐PSC, and the differences observed between treatment‐related TEAEs were measured before T‐PSC as well as with full‐length trial data. In addition, prospective use of the T‐PSC design may have allowed for treatment changes prior to 29% of serious TEAEs, all of which required hospitalization. This analysis thus suggests that the T‐PSC design sufficiently evaluated tolerability and safety. Moreover, in combination with the strong retrospective evidence that the T‐PSC design replicated the efficacy conclusions of trials, ¹ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ the totality of T‐PSC analyses suggests that the T‐PSC design may be appropriate for regulatory approval trials.

There was concern that the only time that placebo‐controlled tolerability and safety data are obtained was during phase 3 trials. ¹ , ¹⁴ The value of this prospective data collection of TEAEs for each trial was highlighted by the differences in overall TEAE rate, where the trial with lower perampanel doses (Study 306) also had a lower TEAE rate. These prospective data allow for comparison of TEAE rate on active treatment as compared to placebo, but placebo is not a benign treatment, because it has been associated with increased rates of SUDEP, among other serious consequences of uncontrolled seizures. ¹ , ² Therefore, a benefit of the T‐PSC design is reducing exposure to both placebo and ineffective treatments. These data suggest that both competing interests can be satisfied, because the T‐PSC design effectively reduces exposure to placebo and ineffective treatment, and this reduction does not substantially impact the rigorous comparisons of TEAE rate between perampanel compared to placebo.

While replicating the evaluation of tolerability, the T‐PSC design also accomplished its goal of improving the safety of trials by reducing the rate of serious TEAEs by 29%. These serious TEAEs, all of which required hospitalization, can have substantial and long‐lasting impact upon participants' quality of life and are major events that everyone involved in the care of people with seizures would like to avoid. If these trials had used the T‐PSC design, then the participants' treatment could have changed by entering open‐label extension (and potentially increasing perampanel dose) or pursuing other treatment options outside of the trial that may have had a better chance of improving seizures, thus limiting exposure to ineffective treatment and mitigating safety risk. ¹ However, we recognize that some of these serious TEAEs may not have been preventable.

One of the concerns about the T‐PSC design is that the data available to evaluate whether serious TEAEs are associated with active treatment compared to placebo could be reduced (in this trial the rate of serious TEAEs was reduced by 29%). In Study 306 but not Study 304 or Study 305, more serious TEAEs occurred after T‐PSC in perampanel compared to placebo. That relative underestimation of serious TEAE rate based on using data before T‐PSC could overestimate the safety of perampanel, but the overall rate of serious TEAEs was low (Study 306, full length: 5.5% in perampanel, 5.0% in placebo; before T‐PSC: 3.7% in perampanel, 4.1% in placebo). It is common for serious TEAE rates to be insufficient for statistical comparison within single‐blinded trials. For example, a meta‐analysis of 112 trials was needed to demonstrate a higher mortality rate in participants randomized to placebo compared to effective ASM doses. ² Thus, irrespective of trial design, reliable assessment of rare safety events requires further evaluation with open‐label extension and postmarketing surveillance studies. ³³ However, when considered in totality, the T‐PSC design may improve safety of trial participants within blinded trials by successfully reducing the rate of serious TEAEs while also providing sufficient data compared to the full‐length blinded phrase.

The results of this tolerability and safety reanalysis with T‐PSC in focal onset seizures differs from our analogous analysis for PGTC seizures. ²³ The T‐PSC in these focal onset trials was substantially shorter than in the PGTC trial, which may reflect increased postrandomization seizure freedom rates and reduced prerandomization seizure frequency of PGTC seizures. ⁵ , ⁶ In the reanalysis of the PGTC seizure trial, there were more TEAEs after T‐PSC in participants randomized to placebo than perampanel. ²³ That systematic difference led to an overestimation of the difference between TEAE rate in perampanel and in placebo. In contrast, for Study 306 only, the T‐PSC design observed all serious TEAEs that occurred in patients randomized to placebo, but it underestimated the serious TEAE rate on perampanel (Fisher exact p = .017). No such over‐ or underestimation was observed in other types of TEAEs for Study 306, or for any category of TEAEs for Studies 304 and 305. These inconsistencies of results suggested that the T‐PSC design did not systematically over‐ or underestimate the difference in TEAEs between perampanel and placebo.

In addition to evaluating the T‐PSC design, this analysis demonstrated that the timing of any TEAEs was comparable for placebo and perampanel in participants with focal onset seizures. In Study 304 and the PGTC seizure analysis, we observed that the rate of any TEAEs was initially lower for participants randomized to placebo, but over the course of the trial, the placebo rate “caught up” so that the final incidence of any TEAEs was comparable between placebo and perampanel. ²³ That nonsignificantly higher “catch‐up” rate of any TEAEs on placebo occurring later in the trial compared to earlier TEAEs on perampanel suggested that TEAEs before the first month of maintenance may be more associated with treatment than placebo, whereas TEAEs after the first month of maintenance may be more associated with placebo than treatment. However, this “catch‐up” increase in any TEAE rate was not observed in Studies 305 and 306.

Despite the variability in these observations between these focal onset studies and the PGTC study, this analysis of the timing of TEAEs replicates similar results to suggest that most TEAEs occur in the first month of titration. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² In all three studies, a large portion of TEAEs occurred in the first week of titration, equally in placebo and perampanel. This observation suggests that most TEAEs occur within the first week of treatment and may be associated with a nocebo effect, where participants anticipate TEAEs. ³⁴ , ³⁵ , ³⁶

The time‐to‐any‐TEAEs curves of placebo and perampanel began to diverge between weeks 1 and 3 of titration. These time‐to‐TEAE curves then stabilized as participants entered the maintenance phase. That timeline further builds on data suggesting that most treatment‐related TEAEs occur in the first month of treatment, and therefore, participants who tolerate the first month of titration are likely to tolerate the ASM for at least 19 weeks.

This analysis and the T‐PSC design have important limitations. The evaluation of the long‐term tolerability and safety of ASMs requires combination of data from the blinded phase of trials with open‐label extension and postmarketing studies. For example, the weight gain seen with full‐length trial data for participants taking perampanel 8 mg only within Study 304 only may reflect a delayed onset TEAE that required longer term and larger studies to measure reliably. Additionally, we highlighted that TEAEs occurred within the titration phase, but the time before steady state serum level was not uniform across all doses. The maintenance dose of participants randomized to 2 mg daily or 4 mg daily was started on week 1 or 2, respectively, and the 105‐h serum half‐life of perampanel suggests that steady state was attained 3 weeks after reaching the assigned dose.

One suggested modification of the T‐PSC design is to require at least 4 weeks of maintenance observation, which may have led to the observation of some of the TEAEs that occurred after T‐PSC. ¹ , ¹³ However, visual analysis of the time‐to‐TEAE curves (Figure 4) and the previous PGTC seizure trial analysis suggested that modification would not lead to a substantial difference in conclusions.

Although this study builds upon the evidence that the T‐PSC was sufficient to evaluate the tolerability and safety of perampanel in participants 12 years or older with focal or generalized onset seizures, these analyses also should be performed on trials of ASMs that have different inclusion and exclusion criteria (e.g., pediatrics and developmental epileptic encephalopathies), as well as different titration durations and serum half‐lives. Our discussions with patients and their advocates suggest that the T‐PSC design would improve participant recruitment by reducing the risk of participation, but formal patient‐centered studies to confirm this have not been conducted.

5. CONCLUSIONS

The tolerability and safety of perampanel for focal onset seizures was sufficiently evaluated when participants were able to stop the blinded period at T‐PSC, as compared to the full‐length trial. This builds upon the evidence supporting the use of this design for trials for regulatory approval of ASMs. If implemented prospectively, this trial design may substantially reduce the risk of participation by reducing exposure to placebo and individually ineffective treatments, which in turn may improve participation recruitment.

AUTHOR CONTRIBUTIONS

Wesley T. Kerr designed the statistical comparison, wrote the manuscript, and created the figures. Leock Y. Ngo coordinated data analysis and contributed to all phases of manuscript preparation. Liang Zhu and Anna Patten performed the statistical analysis. Jocelyn Y. Cheng provided disease‐ and industry‐relevant comments and contributed to the design and framing of the manuscript. Jacqueline A. French supervised and contributed to all aspects of the manuscript. Lavanya Biju assisted with statistical analysis and figure creation. All authors reviewed manuscript drafts and approved the final version of the work.

CONFLICT OF INTEREST STATEMENT

W.T.K. has received compensation as Associate Editor of Epilepsia; writes review articles for Medlink Neurology; is a paid consultant for SK Life Sciences, the Epilepsy Study Consortium, UCB Pharmaceuticals, Cerebral Therapeutics, Jazz Pharmaceuticals, Neurelis, Ventus, Harmony, Epygenix, EpiTel, Neurona Therapeutics, and Biohaven Pharmaceuticals; and has collaborative or data use agreements with Eisai, Janssen, Johnson & Johnson, Radius Health, Glaxo‐Smith‐Kline, and Praxis. L.Y.N., L.Z., A.P., and J.Y.C. were employees of Eisai when this analysis was performed. J.A.F. receives salary support from the Epilepsy Foundation and for consulting work and/or attending scientific advisory boards on behalf of the Epilepsy Study Consortium for Aeonian/Aeovian, Alterity Therapeutics Limited, Anavex, Arkin Holdings, Angelini Pharma, Arvelle Therapeutics, Athenen Therapeutics/Carnot Pharma, Autifony Therapeutics Limited, Baergic Bio, Biogen, Biohaven Pharmaceuticals, BioMarin Pharmaceutical, BioXcel Therapeutics, Bloom Science, BridgeBio Pharma, Camp4 Therapeutics Corporation, Cerebral Therapeutics, Cerevel, Clinical Education Alliance, Coda Biotherapeutics, Corlieve Therapeutics, Eisai, Eliem Therapeutics, Encoded Therapeutics, Encoded Therapeutics, Engage Therapeutics, Engrail, Epalex, Epihunter, Epiminder, Epitel, Equilibre BioPharmaceuticals, Greenwich Biosciences, Grin Therapeutics, GW Pharma, Janssen Phamaceutica, Jazz Pharmaceuticals, Knopp Biosciences, Lipocine, LivaNova, Longboard Pharmaceuticals, Lundbeck, Marinus, Mend Neuroscience, Marck, NeuCyte, Neumirna Therapeutics, Neurocrine, Neuroelectives USA Corporation, Neuronetics, Neuropace, NxGen Medicine, Ono Pharmaceutical Co., Otsuka Pharmaceutical Development, Ovid Therapeutics, Paladin Labs, Passage Bio, Pfizer, Praxis, PureTech LTY, Rafa Laboratories, SK Life Sciences, Sofinnova, Stoke, Supernus, Synergia Medical, Takeda, UCB, Ventus Therapeutics, Xenon, Xeris, Zogenix, and Zynerba. J.A.F. also has received research support from the Epilepsy Study Consortium (funded by Andrews Foundation, Eisai, Engage, Lundbeck, Pfizer, SK Life Science, Sunovion, UCB, and Vogelstein Foundation), the Epilepsy Study Consortium/Epilepsy Foundation (funded by UCB), GW/FACES, and NINDS. She is on the editorial board of Lancet Neurology and Neurology Today. She is Chief Medical/Innovation Officer of the Epilepsy Foundation. She has received travel reimbursement related to research, advisory meetings, or presentation of results at scientific meetings from the Epilepsy Study Consortium, the Epilepsy Foundation, Angelini Pharma, Clinical Education Alliance, NeuCyte, Neurocrine, Praxis, and Xenon. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Figure S1.

Tables S1–S2.

Kerr WT, Ngo LY, Zhu L, Patten A, Cheng JY, Biju L, et al. Time to prerandomization seizure count design sufficiently assessed the safety and tolerability of perampanel for the treatment of focal seizures. Epilepsia. 2025;66:3207–3218. 10.1111/epi.18460

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.