Presentation (CTI)

OBJECTIVES: Efficacy to effectiveness bridging can provide early insights on long-term outcomes, supporting evidence generationfor newly approved therapies. We propose a simulation framework to project long-term treatment effects in real-world (RW) settings under different scenarios. It was applied to estimate the effectiveness of adding zilucoplan to standard of care (SoC) on severe exacerbations in generalized myasthenia gravis.
METHODS: A two-step simulation framework was developed to estimate effectiveness under specified RW scenarios. First, a Cox proportional hazards disease-progression model is trained and validated to estimate baseline event-free survival under SoC. Second, individual survival curves are simulated in scenario-specific populations under SoC alone and the investigational treatment by combining the disease-progression model with a treatment-effect model derived from trials. Uncertainty is quantified by sampling the treatment-effect coefficient across repeated simulations. Results are summarized as scenario-specific mean event-free survival curves with 95% confidence intervals and associated risk reductions at a prespecified time point. In our application, zilucoplan added to SoC was the investigational treatment, with two-year effectiveness assessed across scenarios reflecting different SoC therapy profiles. Its treatment effect was estimated from clinical trial data; while scenario populations were informed by MarketScan claims data.
RESULTS: In our application, the disease-progression model showed acceptable predictive performance (C-index: 0.71; mean time-dependent AUC: 0.73; integrated Brier score: 0.08). Across SoC-based scenarios, the framework simulated lower risks of severe exacerbation over two years when adding zilucoplan to SoC vs SoC alone, although confidence intervals overlapped. Two-year risk reductions ranged from 8.6% to 15.1% across SoC-based scenarios with heterogeneous baseline predicted risks.
CONCLUSIONS: This study highlights the value of RW simulation-based approaches for anticipating long-term outcomes when evidence is limited. By combining predictive modelling with trial efficacy and uncertainty estimation this framework can support early evidence generation and characterize long-term outcome variation across clinically relevant scenarios.