Confidence and Prediction Intervals for Pharmacometric Models.

Area of Science:

• Pharmacometrics and Drug Development
• Quantitative Pharmacology
• Clinical Decision Support

Background:

• Pharmacometric models are vital for drug development, predicting drug exposure (pharmacokinetics) and effects (pharmacodynamics).
• Current practices often focus on parameter estimation uncertainty, which is less relevant for clinical application.
• Accurate uncertainty quantification is essential for informed clinical decision-making.

Purpose of the Study:

• To review confidence and prediction intervals in pharmacometric modeling.
• To provide calculation methods for prediction intervals.
• To encourage routine reporting of prediction uncertainty by pharmacometricians.

Main Methods:

• Review of statistical concepts: confidence intervals vs. prediction intervals.
• Discussion of calculation methodologies for prediction intervals.
• Illustrative examples (not explicitly detailed in abstract but implied by 'tutorial' and 'methods').

Main Results:

• Parameter estimation uncertainty is commonly reported but less critical than prediction uncertainty.
• Prediction uncertainty is the key metric for supporting clinical decisions.
• Confidence and prediction intervals offer distinct information crucial for different decision-making contexts.

Conclusions:

• Pharmacometricians should prioritize reporting prediction uncertainty for robust clinical decision support.
• Understanding and calculating prediction intervals are essential skills for pharmacometricians.
• Routine reporting of prediction intervals will enhance the utility of pharmacometric models in drug development.