A three‐stage design for allergen immunotherapy trials

Background Clinical trials of allergen immunotherapy (AIT) may require up to 5 years to complete. These lengthy trials may be complicated by high and potentially differential dropouts, especially among participants who perceive that they are receiving placebo. We propose a three‐stage design in which the placebo group in Stage 1 crosses over to receive active treatment in Stage 2. In Stage 3, AIT is discontinued to determine whether benefit is maintained post‐treatment. We apply inferential statistics to support the three‐stage design for clinical trials to determine clinical efficacy, treatment response over time, and sustained response to AIT. Methods The proposed framework constitutes a series of hypothesis tests for comparing treatment responses at the end of each stage. A simulation study was performed to illustrate the statistical properties under varying statistical missing mechanisms and effect sizes. Results The statistical properties in terms of bias and statistical power were consistent with what are expected from conventional analyses. Specifically, the extent of bias depended on the missing mechanism and magnitude. The statistical powers were largely driven by effect and sample sizes as well as prespecified success margins. As an illustration, assuming relative treatment differences of 25% and stagewise dropout rate of 15%, a sample size of 200 per group may achieve 93% power to demonstrate a treatment effect and 60% power to demonstrate a maintained response post‐treatment. Conclusions Inferential statistics support our proposed study design for evaluating benefits of AIT over time and inform clinical understanding and decisions. We introduce a three‐stage design to characterize the effects of an allergen immunotherapy product through stage‐wise statistical evaluations. These effects include initial treatment effect (Stage 1), effect associated with time of introduction and length of treatment (Stage 2), and sustained effect after treatment is withdrawn (Stage 3). Simulations illustrated the statistical properties (bias and power) under various scenarios.Abbreviations: A, stage 2 scenario; AIT, allergen immunotherapy; B, stage 2 scenario; DBPCR, double‐blind placebo‐controlled randomized; MAR, missing at random; MNAR, missing not at random; NULL, stage 2/3 scenario; RD, relative difference