Abstracts

Rationale: The impact of missing data due to ET on the statistical methods used to evaluate the efficacy of anti-epileptic drugs (AEDs) in clinical trials is not well understood. ESL is a once-daily (QD) oral AED, approved in the EU, US, and Canada as adjunctive treatment for partial-onset seizures (POS). Here we report results of various pre-specified sensitivity analyses from a Phase III study that evaluated the safety and efficacy of ESL as adjunctive therapy for refractory POS (BIA-2093-304; NCT00988429). The sensitivity analyses were designed to examine the robustness of the primary efficacy analysis result, under different assumptions about missing data due to ET.Methods: Subjects ≥16 years entered an 8-week baseline (BL) phase, at the end of which, eligible subjects were randomized 1:1:1 to receive placebo (PBO), ESL 800 mg, or ESL 1200 mg (all QD). Double-blind (DB) treatment included a 2-week titration phase (TP) followed by a 12-week, fixed-dose maintenance phase (MP). Seizure data were captured by event entry diaries in 29% of subjects, and by daily entry diaries in 71% of subjects. The primary efficacy variable was standardized seizure frequency (SSF) per 28 days during the MP. In the primary analysis this variable was calculated using all available seizure data collected during the study. Three imputation methods were used to address ET: 1. TP data carried forward (TDCF); 2. BL data carried forward (BDCF); 3. last 2-week data carried forward (LDCF). TDCF imputes missing seizure frequency data during MP (due to ET during TP) with SSF during TP. BDCF imputes missing seizure frequency data between the date of the last available seizure assessment and the scheduled end of MP with SSF during BL. LDCF uses SSF over the last 2 weeks prior to ET to impute missing seizure frequency data. Finally, a completer analysis was performed, using seizure data from only those subjects who completed the MP. In both the primary and sensitivity analyses, log-transformed MP SSF data were analyzed using an ANCOVA model, with log-transformed BL SSF and diary version as covariates, and treatment as the fixed effect. Pair-wise comparisons were between each ESL dose versus PBO. LS means were back-transformed via the exponential function.Results: The intention-to-treat (ITT) population (n=640) comprised randomized subjects who received ≥1 dose of study treatment and had ≥1 post-BL seizure frequency assessment. 43 (7%) and 93 (15%) subjects discontinued during TP and MP, respectively. The percentages of subjects completing DB treatment were 86% (PBO), 80% (ESL 800 mg), and 69% (ESL 1200 mg). The completer analysis, TDCF, BDCF, and LDCF produced results similar to those of the primary analysis (Table 1).Conclusions: While little is known about seizure frequency after subjects withdrew from the study, sensitivity analyses under different scenarios showed that the primary analysis of the efficacy of adjunctive ESL treatment in subjects with POS was robust. Study sponsored by Sunovion Pharmaceuticals Inc.