Improving categorical endpoint longitudinal exposure-response modeling through the joint modeling with a related...

Area of Science:

• Pharmacometrics
• Clinical Pharmacology
• Biostatistics

Background:

• Exposure-response modeling is crucial for optimizing drug dosage in clinical trials.
• Primary endpoints in trials can be limited; additional informative endpoints may exist.
• Joint modeling can leverage all available data for more robust analysis.

Purpose of the Study:

• To investigate the benefits of joint modeling for an ordered categorical primary endpoint and a related continuous endpoint.
• To apply latent variable indirect response (IDR) modeling for parameter sharing between endpoints.
• To analyze data from a ustekinumab psoriasis trial using this joint modeling approach.

Main Methods:

• Developed and compared separate and joint latent variable Type I IDR models.
• Utilized data from a Phase 3b trial including Physician's Global Assessment (PGA) and Psoriasis Area and Severity Index (PASI) scores.
• Analyzed longitudinal data up to 116 weeks.

Main Results:

• A separate model for the Physician's Global Assessment (PGA) showed substantial structural bias.
• Joint modeling of PGA and Psoriasis Area and Severity Index (PASI) scores corrected the bias observed in the separate PGA model.
• The joint model provided a more accurate representation of the exposure-response relationship.

Conclusions:

• Joint modeling of related endpoints enhances the accuracy of exposure-response models in clinical trials.
• This approach can mitigate structural bias present in models using only primary endpoints.
• Integrating multiple data streams, like PGA and PASI, leads to improved pharmacokinetic/pharmacodynamic (PK/PD) analyses.