(C-170) Quantifying the Effect of Treatment Switch: An Illustration with the Prevalent New User Design

Background: In real-world evidence (RWE) studies, patients often switch from one treatment to another, complicating the use of traditional new user design that restricts the analysis to treatment-naïve patients to estimate treatment effects. There is a lack of comprehensive tutorials on how to implement the prevalent new user design, which accounts for treatment history and confounding factors among all prevalent treatment users.

Objectives: To provide a step-by-step guide for implementing the prevalent new user design, with an illustration and programming example using real-world evidence data.

Methods: The prevalent new user design pairs cases (patients who have switched from the initial treatment to the new one treatment switchers) and controls (patients who remain on the initial treatment) at the time of the treatment switch, creating exposure time sets. The design steps include, 1) the identification of cases (treatment switchers) and eligible control candidates for each exposure set, 2) calculation of time-conditional propensity scores to account for covariates and treatment history at the time of the treatment switch, 3) matching cases to controls within each exposure set based on the propensity scores and 4) analysis of outcomes on matched set. The illustration demonstrates these steps using a real-world evidence dataset of patients treated for respiratory diseases. The effect of treatment switches on mortality was taken as the main outcome of interest.

Results: A total of 1,827 cases were are available and pooled to24,628 eligible controls, resulting in 6,262,488 potential controls across all exposure sets. The matching was performed exactly on year of inclusion and duration between initiation of first treatment and treatment switch and using propensity score on time-varying confounding factors measured at each exposure set (number of previous events of interest, age, comorbidities). After matching at a 1:1 ratio, 1,652 cases (90.4%) were successfully paired. The occurrence of death was lower in the case group.

Conclusions: The prevalent new user design offers a powerful approach for evaluating treatment switches in RWE studies. This step-by-step guide demonstrates its potential and feasibility using standard programming software.