The visual predictive check and real-world data

Real-world data is a rich source of information for developing pharmacometric models, but it comes with unique challenges. This publication in Clinical Pharmacology & Therapeutics (CPT)  by researchers from InsightRX and Pharmetheus highlights a major issue: traditional VPCs and prediction-corrected VPCs (pcVPCs) can produce misleading results when applied to RWD shaped by adaptive clinical decision-making, such as dosing, sampling, and follow-up.

At the PAGE 2026 conference, two posters present novel methodological solutions to accurately diagnose models applied to RWD:

• Reference-Corrected VPCs: Addressing Model Evaluation Challenges with Real-World Data and Adaptive Designs

This poster addresses adaptive dosing and sampling by normalizing both observed and simulated data to a fixed reference dataset, such as a fixed BID dosing regimen.

• Visual Predictive Checks for Real-World Data using Propensity-Score-Matching

This poster describe how corrects simulation biases by matching drawn parameter sets to the observed subject’s empirical Bayes estimates (EBEs), restoring the correlation between parameters and study design.