Presentation (CTI)

OBJECTIVES: Non-interventional studies (NIS) are real-world studies which aim to gain deeper insights into how treatments work (1). Regulatory Guidelines for NIS have been published (2-3); however further effort is needed at the study design stage to complement the usual precision calculation approach (1). Herein, we address the assumptions in the design of a two-cohort NIS in the context of a marketed drug in one cohort and conventional therapy in another, examining a real-world time-to-event primary endpoint, rwTTNT1.
METHODS: Through simulation studies we explore the potential impact of altering assumptions such as the operationally fixed sample size. The consequence of utilising different median rwTTNT1 and annual study withdrawal rates are also assessed for study duration, along with bias and the precision associated to median rwTTNT1. Furthermore, the probabilities of having shorter median rwTTNT1 and the overlap probabilities of 95% CI of the rwTTNT1 between the two cohorts of patients are inspected, alongside the number of events observable at specific data cuts, together with the operational impact of lower recruitment.
RESULTS: A few key findings are presented here but full results will be shown. For each value of median rwTTNT1 and withdrawal rate, we observed the expected patterns under different scenarios and, importantly, were able to quantify these differences based on 1000 simulations. For instance, where the true median was 8 months, the study duration was approximately 26 months with no withdrawals, and approximately 28 months with a 20% annual withdrawal rate. Additionally, where the true median was 6 months, only 15% of simulations observed a median less than 5.5 months. The bias was acceptable across all scenarios; however, the precision of estimates was greatly affected across the differing assumptions.
CONCLUSIONS: We believe that this framework can provide valuable information at a NIS design stage, complementing the usual precision calculation approach.