GROUP SEQUENTIAL SAMPLE SIZE: DOES STOPPING EARLY REQUIRE COMPROMISING SAFETY? Zmira Silman1 and Diklah Geva2 1 Industry Consultnt, Netanya, 2 IntegriStat, TelAviv, Israel EMR 2007 1 Introduction Stopping the trial early, either due to efficacy or futility, is the purpose of Group Sequential designs, whilst still fully controlling the prespecified type I error. These designs are usually employed to test a single efficacy endpoint1. A concern is raised as to the safety and tolerability of a product when the clinical trial was concluded early2. 2 Aim We propose an approach for integrating the two endpoints Efficacy and Safety for the group sequential study design. 3 Application-Field Example A study aiming to compare Novel and Standard Acne treatment in a population of youngsters. The standard treatment is efficient but has low tolerability. The novel treatment is expected to have similar therapeutic effect but will be more tolerable. 4 Application-Field Example The study endpoints are: Efficacy (E): Percent of reduction in Acne lesions compared to baseline (Average µ) Safety (S): Incidence of side effects (Proportion π ) A sequential design was considered in order to obtain an early stop if striking results are achieved. Definitions: TN-Novel Treatment Ts-Standard Treatment 5 Hypotheses The two hypotheses will be tested with an overall significance level of α=0.05 and 80% power. Each hypothesis; HE and HS will be tested with α=0.025. 1. HE : TN is not inferior TS wrt Efficacy. 2. HS: TN is superior to TS wrt Safety. Formally these translates to: 1. HE: µ T - µ T ≤-δ (δ - non-inferiority margin) N S 2. HS: π T –πT >0 N S 6 Methods Group sequential design was applied to obtain test boundaries using O’Brien & Fleming’s method with k=4 stops and α=0.025. Efficacy and Safety boundaries were plotted in two panels, along with the corresponding sample sizes. The boundaries and sample size were calculated for the pre-specified clinically meaningful effect size. The upper boundaries represent the efficacy/safety α-spending criteria The lower boundaries represent the futility criteria 7 Panel 1: Efficacy Boundaries Efficacy Boundaries and Sample Size Test Statistic 5.0 Efficacy Futility 54 4.0 109 3.0 2.0 163 218 1.0 0.0 40 70 100 130 160 190 220 Overall Number of patients 250 280 310 O’Brien & Fleming’s design, with k=4 stops, α=0.025, ∆=0.3, σ=1,δ= -0.1 8 Panel 2: Safety Boundaries Saftey Bounaries and Sample Size 5.0 Safety Futility 77 4.0 3.0 154 2.0 231 308 1.0 0.0 40 70 100 130 160 190 220 250 Overall Number of patients 280 310 O’Brien & Fleming’s design, with k=4 stops, α=0.025, π1=0.15 and π 2=0.05 9 How to decide on an early stop? Rule: Early stop will occur only when the boundaries are crossed for the two panels simultaneously. Problem: How to apply the rule as the two panels call for different sample size on the pre-specified time point. Solution: We propose to fix the boundaries with the larger-n, i.e. safety panel and to extrapolate boundaries of the lower-n, i.e. Efficacy panel. Such extrapolation can follow the given boundary also called the α spending curve. 10 Calculation procedure Obtain the number of subjects at each k. (k=4) Find the α-spending boundary function (power function) Re-calculate the lower-n boundaries by extrapolating the larger-n on the α-spending boundary Draw an integrated boundaries graphs. Decide to stop early if E and S boundaries are crossed. 11 Re-Calculation of efficacy boundaries Test Statistic Efficacy Boundaries Extrapulated at Safety S-Sizes 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 54 Safety Futility y = 29.32x-0.5 (77) 109 163 218 (308) 40 70 100 130 160 190 220 250 280 310 Overall Number of patients Adjusted boundaries based on the power function (O’brien & Fleming) 12 Integrated boundaries for Safety & efficacy Safety Efficacy Futility OBS-S OBS-E Intigrated Saftey& Efficacy Boundariesl Test Statistic 5.0 4.0 77 154 3.0 231 2.0 308 1.0 0.0 40 70 100 130 160 190 220 250 280 310 Overall Number of patients Stooping Time K1 K2 K3 K4 N Boundaries Observed Statistic Efficacy Safety Safety EfficacyFutility OBS-S OBS-E 54 77 4.0 3.3 0 2.78 2.5 109 154 2.8 2.4 0 2.87 2.55 163 231 2.3 1.9 0 2.95 2.65 218 308 2.0 1.7 1.99 13 Discussion We present here a method to simultaneously monitor efficacy and safety, in a group sequential design of clinical trial. This is a simple method may lead to an early stop without compromising safety information on one hand or inflating the efficacy information on the other hand. This method provides a combined interim test for both efficacy & safety even though the initial evaluation indicated different group sizes for each outcome. We used the O’Brien & Fleming’s method here, however other methods to calculate boundaries can also benefit from this procedure. 14 References 1. Chow S. C., Shao J., Wang H.: Sample Size Calculations in Clinical Research. CRC Press of Taylor & Francis Group 2003, Chapter 8. 2. EMEA: Reflection paper on methodological issues in confirmatory clinical trials with flexible design and analysis plan, CHMP/EWP2459/02, London March 2003. 3. The ADDPLAN software, www.addplan.com Zmira E-Mail: [email protected] Diklah E-Mail: [email protected] 15