PT - JOURNAL ARTICLE AU - Polina Kurbatova AU - Agathe Bajard AU - Harm Tiddens AU - Vitaly Volpert AU - Catherine Cornu AU - Nicolai Bessonov AU - Behrouz Kassai AU - Sylvie Chabaud AU - Patrice Nony AU - Daan Caudri TI - Modelling and simulation of experimental designs to find the best design of randomized clinical trials in a rare disease: Cystic fibrosis DP - 2014 Sep 01 TA - European Respiratory Journal PG - P1220 VI - 44 IP - Suppl 58 4099 - http://erj.ersjournals.com/content/44/Suppl_58/P1220.short 4100 - http://erj.ersjournals.com/content/44/Suppl_58/P1220.full SO - Eur Respir J2014 Sep 01; 44 AB - Introduction: The parallel group randomized controlled trial is the gold standard in clinical trials, but in rare diseases alternative trial designs may be more suitable. An in silico approach with modelling and simulation has been proposed in the CRESim project to find the most appropriate design in cystic fibrosis (CF).Methods: A mathematical model was built to describe the effect of Dornase alfa on mucocillary clearance in CF. Secondly, a virtual patient population was created and the randomization of these patients in clinical trials was simulated. Several experimental designs were simulated: Parallel, Cross-over, Randomized withdrawal, Early escape, N of 1, and Adaptative randomizations such as “Play the winner” (PW) and “Drop the loser” (DL). The results of simulations of these different experimental designs were compared with the traditional parallel group design, in terms of precision of the estimation of treatment effect, statistical power, and trial duration.Results: One thousand trials were simulated for each design with a sample size of 50 patients. The Cross-over design showed the best estimation of treatment effect with 87% statistical power, a low coefficient of variation (36%) but with a high trial duration (2 years). Parallel, PW and DL designs all had a similar power (about 60%) and coefficient of variation (about 45%) but PW and DL had a higher trial duration (2 years vs. 1.5 years). The N of 1 design had the lowest power (22%), which could be improved by a meta-analysis of N of 1 trials.Conclusion: This in silico approach could be used to find the most effective clinical trial design, especially in rare diseases such as CF.