Estimation problem due to multiple solutions in pharmacokinetic curve fitting to two-compartment model and its...

Area of Science:

• Pharmacokinetics
• Computational Chemistry
• Biostatistics

Background:

• Accurate curve fitting of plasma drug concentration data is crucial for pharmacokinetic analysis.
• Non-linear least-squares regression is commonly used, but susceptible to data errors and model complexity.
• Two-compartment models are frequently employed but can present estimation challenges.

Purpose of the Study:

• To investigate data errors in curve fitting for pharmacokinetic models.
• To identify and address estimation problems in two-compartment models.
• To evaluate the performance of non-linear least-squares regression programs (SALS, NONLIN) and explore solutions for improved accuracy.

Main Methods:

• Analysis of 151 datasets from 11 drugs to quantify errors in intravenous and oral data.
• Simulation of drug concentration data with added random errors (25% CV) for two-compartment models.
• Application of non-linear least-squares regression programs (SALS, NONLIN).
• Utilized the SIMPLEX algorithm for initial parameter estimation in curve fitting.

Main Results:

• Intravenous bolus data showed 5-10% errors, while oral data exhibited 10-25% errors.
• Two-compartment models with tri-exponential equations produced unreasonable estimates (e.g., 'flip-flop' phenomenon) when data errors exceeded 10% or parameters were >= 5.
• The SIMPLEX method successfully provided reliable initial estimates, avoiding erroneous results.

Conclusions:

• Data errors significantly impact the accuracy of pharmacokinetic model fitting.
• Complex models like the two-compartment model are sensitive to data quality and parameter count.
• The SIMPLEX algorithm offers a robust solution for obtaining accurate initial estimates in pharmacokinetic curve fitting, enhancing reliability.