Instrumental Variable Methods to Target Hypothetical Estimands With Longitudinal Repeated Measures Data: Application...

Area of Science:

• Clinical Trials
• Pharmacology
• Biostatistics

Background:

• The STEP 1 trial assessed semaglutide's efficacy for weight reduction over 68 weeks.
• Nonadherence to treatment protocols is common in long-term clinical studies.
• Existing analyses used the Estimand Framework, with primary analysis adopting a treatment policy strategy.

Purpose of the Study:

• To propose and evaluate an alternative instrumental variable (IV) method for adjusting semaglutide treatment effects in the presence of nonadherence.
• To define and estimate two distinct hypothetical treatment effects under adherence scenarios.
• To apply this novel IV method to the STEP 1 trial data.

Main Methods:

• Utilized an instrumental variable (IV) approach to adjust for nonadherence, relaxing the "unconfoundedness" assumption.
• Employed a method that leverages repeatedly measured outcome data and allows for time-varying treatment adherence effects.
• Defined two hypothetical estimands: universal adherence to semaglutide and universal adherence to both semaglutide and placebo.

Main Results:

• The proposed IV method provides valid results even under guaranteed nonadherence conditions.
• It effectively utilizes longitudinal data for a more robust analysis of treatment effects.
• Application to STEP 1 data suggests a sustained, gradually diminishing weight loss effect of semaglutide.

Conclusions:

• The novel instrumental variable method offers a robust alternative for analyzing treatment effects in the presence of nonadherence.
• This approach provides valuable insights into hypothetical treatment outcomes under full adherence.
• Semaglutide demonstrates a sustained weight loss effect, even after accounting for nonadherence in the STEP 1 trial.