Nonlinear mixed-effects modeling as a method for causal inference to predict exposures under desired within-subject...

Area of Science:

• Pharmacometrics
• Causal Inference
• Clinical Trial Analysis

Background:

• The estimand framework separates clinical questions from estimation methods, necessitating valid estimation techniques.
• Causal inference is crucial for validating estimation methods beyond intention-to-treat analyses.
• Mixed-effects models are common in pharmacometrics but seldom discussed for causal inference.

Purpose of the Study:

• To evaluate nonlinear mixed-effects modeling and simulation (NLME M&S) as a causal inference method.
• To demonstrate NLME M&S as a standardization technique for longitudinal data with confounders.

Main Methods:

• Nonlinear mixed-effects modeling and simulation (NLME M&S) was employed.
• Standardization principles from causal inference were applied.
• A simulated clinical trial with dose titration was used to illustrate the approach.

Main Results:

• Nonlinear mixed-effects modeling was shown to be a specific implementation of standardization.
• The method conditions on individual parameters (random effects) to correct for confounding.
• Unbiased estimates were obtained by conditioning on individual parameters or prior outcomes.

Conclusions:

• NLME M&S provides a valid causal inference approach for longitudinal data.
• This modeling strategy effectively implements standardization for handling confounders.
• The framework allows for unbiased estimation of treatment effects under adherence assumptions.