Impact of Event-Related Treatment Modifications in Exposure-Response Analyses of Repeated Time-to-Event Data

Area of Science:

• Pharmacometrics
• Biostatistics
• Clinical Trial Analysis

Background:

• Exposure-response (ER) analyses in repeated time-to-event (RTTE) data face confounding when treatment modifications are linked to the event of interest.
• Event-related treatment discontinuations can introduce reversed causality, biasing ER estimates.

Purpose of the Study:

• To evaluate methods for unbiased ER estimation in RTTE data with event-related treatment modifications.
• To assess the performance of average exposure, time-varying exposure, and a novel ranked exposure approach.

Main Methods:

• Simulated placebo-controlled studies were used to assess three ER analysis approaches.
• Approach 1: Average exposure.
• Approach 2: Time-varying exposure.
• Approach 3: Average exposure combined with a percentile-rank of blinded study treatment received.

Main Results:

• Average exposure (Approach 1) was biased in all scenarios with event-related treatment discontinuations.
• Time-varying exposure (Approach 2) was generally unbiased, except when events followed treatment discontinuation.
• The novel ranked exposure approach (Approach 3) provided unbiased estimates, particularly for linear ER relationships.

Conclusions:

• Treatment modifications linked to events are a significant confounder in RTTE ER analyses.
• A ranked measure of average blinded study treatment received can mitigate confounding from event-related treatment modifications.