Data Visualization for Quality Control in NONMEM Data set

Non Linear M ixed Ef fects Model (NONMEM) data set is w idely used for pharmacokinetics (PK) / pharmacodynamics (PD) modeling and simulation, w hich studies the drug concentration in the body over time (measured in terms of absorption, distribution, metabolism, and excretion [ADME]) and the body’s pharmacological response to a drug (measured in terms of adverse events [AE] and eff icacies). In a very specif ic pre-defined format, the NONMEM data set includes a chronological mixture of dosing records, PK/PD observations and covariates of the dosing and observation records. To create NONMEM data sets, it takes tremendous programming efforts for programmers to derive dosing history, order PK/PD observations and merge various types of covariates. The variables required for NONMEM data are often complicated and come from different source data sets. It is a tough challenge to perform data validation and cleaning. Good quality NONMEM data is critical in PK/PD analysis and errors from a small portion of the data can redirect the conclusion of a study. To guarantee the accurate and meaningful PK/PD analysis, data cleaning is essential and crucial for quality control in NONMEM data set production. Graphs are visual summaries of data and very effective to describe essential features than tables of numbers. This paper illustrates some commonly used graphs to virtualize the data errors and questionable records in both raw clinical data and NONMEM data set. Scientif ic programmers and pharmacometricians w ith minimal programming skills can apply these graphs to check data issues and examine data thoroughly.