Abstracts

Rationale: The ideal POP study provides a strong signal of efficacy over the shortest duration, while exposing the fewest patients possible. Availability of an appropriate POP study design for refractory partial seizures could increase the numbers of new chemical entities. We used data from a large Pfizer database of trials which studied add-on pregabalin treatment (N=1825),AED approved as adjunctive therapy for adult patients with partial onset seizures. We model how duration of baseline, post-randomization treatment period and number of subjects impact on likelihood of an interpretable efficacy signal. Methods: We combined data from five double-blind, randomized, placebo controlled add-on treatment studies. The baseline period was divided into weekly intervals (-8 to -1 weeks prior to randomization). The post-randomization period was also divided into weekly periods (week 1 to 12). For a fixed sample size, we varied the length of the baseline and post-randomization periods and calculated effect sizes, p-values, and adjusted R2 for the data available for each combination of baseline and post-randomization period length.The probability of a positive study defined as Ppos = Pr (p-value < 0.05)was assesed. Our models used Rratio as dependent variable with treatment (drug vs placebo), study, and baseline seizure rate as predictors. We did full sample size and sample sizes of N=20 to 120 per group in steps of 10. We withdrew samples without replacement from the full population 1000 times. We repeated these processes for all subjects, for subjects with moderate seizures at baseline (> 9.5 seizures /month) and for subjects with high seizure frequency activity at baseline (> 21 seizures /month). Optimal baseline and post-randomization periods were based on clinical and statistical inputs.Only PGB 600 and PGB 300 mg/day with a 1-week or shorter titration were used. Results: Four weeks of baseline with two weeks post-randomization periods were selected as clinically and statistically important periods as compared to full length of study period (recall titration was <1week). For all subjects, with a sample size of N=60 per group average treatment difference between PGB 600 and Placebo was a reduction of 28.2% (95% CI [-45.6, -10.1] and average p-value of 0.0447 (95% CI [<0.00001,0.21]) and Ppos = 80%. For subjects with moderate seizure frequency at baseline, with a sample size of N=50 per group average treatment difference between PGB 600 and Placebo was a reduction of 27.4% (95% CI [-43.5, -10.9] and average p-value of 0.0365 (95% CI [<0.00001,0.27]) and Ppos = 83.5%. For subjects with high seizure frequency at baseline, with a sample size of N=50 per group average treatment difference between PGB 600 and Placebo was a reduction of 26.0% (95% CI [-41.9, -9.9] and average p-value of 0.0452 (95% CI [<0.00001,0.26]) and Ppos = 79.5%. Conclusions: As compared to full period of observations, four weeks at baseline followed by two weeks of post-treatment seemed to be sufficient to provide an interpretable efficacy signal. Sample size was also considerably reduced.