(B-231) Comparison of prevalent new-user and sequential trial designs for studying the real-world effects of treatment discontinuation in cystic fibrosis

Background: The optimal strategy for applying target trial emulation to estimate causal effects of treatment discontinuation is unclear, especially given potential key challenges, e.g. defining time zero.

Objectives: To establish and compare the estimands, implementation, and interpretation of two innovative study designs in this setting; the prevalent new-user design (PNUD) and sequential trials design.

Methods: To illustrate and compare the two study designs, we present a cohort study estimating causal effects of treatment discontinuation in cystic fibrosis (CF), based on the CF-WISE randomised controlled trial (RCT), using data from the UK CF Registry (a database with near-whole population coverage of people with CF in the UK). We estimate the effect of discontinuation of inhaled corticosteroids (ICS) on time to first pulmonary exacerbation using the trial emulation framework combined with the PNUD and sequential trials designs. Following the study design phase, the analyses use Cox regression with control for baseline confounders through weighting or adjustment.
A key part of this work is to clarify key differences in the underlying estimands and analytical approaches implied by the two designs.

Results: We identified 536 people discontinuing ICS treatment in the study period (Jan 2016 – Jun 2018) after ≥90 days sustained use. Review of the ICS start and stop dates, and outcome data indicated they were reliably recorded.
The PNUD targets an average treatment effect in the treated (ATT), i.e. “how would the outcomes in those discontinuing ICS have differed, on average, had they continued on ICS?”. The sequential trials design targets the average treatment effect in the whole population (ATE), i.e. “how would the outcomes in the population differ, on average, had everyone discontinued ICS versus continued on ICS?”.
Further results from the applied study will be presented at the conference.

Conclusions: Studying causal effects of treatment discontinuation involves further complexities than studies of treatment initiators, with the optimal approaches being unclear. This emphasises the importance of clear specification of target trial protocols in this context to ensure emulations reflect a relevant research question.