Menu Back to Poster-Presentations-Details

P339: Estimand Framework in Phase 1b Autologous CAR T-cell Therapy for SLE

Poster Presenter

Meng-Ru Cheng

Statistical Science Associate Director
AstraZeneca
United States

Objectives

Evaluate and implement an ICH E9(R1)–aligned estimand framework in a Phase 1b study of investigational autologous CAR T-cell therapy for systemic lupus erythematosus (SLE), addressing cell therapy–specific intercurrent events (ICEs) to strengthen early phase evidence and align with pivotal trials.

Method

We reviewed SLE trials and mapped cell-therapy ICEs, covering vein-to-vein manufacturing failures, bridging therapy, lymphodepletion, and CAR T-cell pathway. We addressed arm-specific ICEs and event timing, reviewed oncology conventions, contrasting them with SLE needs to inform Phase 1b choices.

Results

We identified ICEs seen in SLE trials—such as treatment discontinuation and rescue medication use—alongside those distinctive to autologous chimeric antigen receptor (CAR) T-cell therapies, including manufacturing failures preventing infusion (no CAR T-cell dose administered), products not meeting release specifications across the vein-to-vein process, responses to bridging therapy, receipt of alternative therapies, CAR T-cell expansion failures, and deaths before outcome assessment. By linking these ICEs to estimand attributes, we refined the clinical questions and ensured analysis strategies harmonized with ICH E9(R1) principles. Because autologous cell therapies were first developed in oncology, we reviewed oncology conventions for ICE handling and contrasted them with SLE needs. This highlighted differences in ICE characteristics. Specifically, in oncology, the focus is often on terminal or irreversible events handled via while-on-treatment strategies. Whereas in chronic autoimmune diseases, flares and ongoing disease activity are typical ICEs, and treatment-policy or composite approaches are preferred. In early-phase settings, estimand implementation differs from later stages: Phase 1 treats them as optional and less formalized, particularly since primary endpoints focus on safety and tolerability, while the estimand framework centers on efficacy. Nonetheless, for CAR T-cell assets, there is high interest in efficacy endpoints already in Phase 1. Also, clinical development plans are accelerated to go directly to pivotal phases after Phase 1. This prompted early consideration of estimand strategies. On the other hand, small sample sizes limit the applicability of rigorous estimand implementation. In the end, standard Phase 1b “as-observed” data reporting was chosen as the main descriptive analysis, complemented by supplementary analyses to cover loss-to-follow-up events.

Conclusion

Early consideration of ICH E9(R1) estimands in Phase 1b autologous CAR T-cell for SLE improves consistency, supports regulatory readiness, and strengthens cross-functional alignment, while helping reduce protocol ambiguities and refine missing data and post-ICE strategies. Explicitly defining estimand attributes (population, variable, ICE strategy, summary measure) for multi-step pathways as well as manufacturing and timing complexity enhances interpretability and helps translate early signals into pivotal trial planning. Adding supplementary estimands—often composite ones that include clinically meaningful outcomes after an ICE—further focuses results on clinical priorities without reducing the flexibility needed in early development. Overall, this approach Improves data interpretability, bolsters regulatory discussions, and strengthens planning for future registrational trials.