A Target Trial Emulation Design to Seek EMA Approval for a Reformulated Drug With a Non-Inferiority, Pivotal, Retrospective Study: A Success Story

OBJECTIVES: Reformulation of drugs is common to enhance convenience and simplification of regimens, which increases efficacy by improving compliance and tolerability. Regulatory agencies still request phase-III randomized clinical trials (RCT) to support the reformulation approval when it cannot be excluded that differences in the new formulation pharmacokinetic profile will influence efficacy and safety. Target trial emulation (TTE) mimics the principles of RCTs into observational studies allowing for causal inference. We designed a TTE for a non-inferiority, pivotal RCT of a new extended-release (ER) drug compared to its immediate-release (IR) equivalent for a central nervous system (CNS) disease, to seek EMA approval using available real-world data (RWD) from the United States.

METHODS: We extracted RWD from claims and electronic medical records and used natural language processing on unstructured notes. Newly initiated ER patients were matched to IR patients using propensity scores (PS) based on relevant covariates. Effectiveness and safety endpoints, adapted to the real-world setting, were assessed using multivariable models, adjusted by intercurrent events (non-adherence). TTE framework was deployed through the statement of a causal research question, PS-matching to emulate randomization, aligning eligibility, treatment assignment, and follow-up start at same point-in-time (IR/ER initiation), and defining the causal estimand and statistical analyses on as-per-protocol strategy. Negative controls (outcomes with no causal relationship with studied drugs or outcomes) were defined by CNS experts.

RESULTS: Of 20,000 eligible subjects, the 2,200 subjects included in the PS-matched cohorts confirmed the non-inferiority of ER to IR in terms of effectiveness, and indicated superiority of ER for most endpoints, including safety and healthcare-resource utilization.

CONCLUSIONS: Our TTE design minimized confounding, time-related biases (immortal time or selection of prevalent users) by aligning epidemiological aspects at time-zero, and residual systematic error. EMA scientific advice corroborated the TTE design to support the application of the new formulation.