A nonlinear multiple regression program, MULTI2 (BAYES), based on Bayesian algorithm for microcomputers

Area of Science:

• Pharmacokinetics and Pharmacodynamics
• Computational Biology
• Biostatistics

Background:

• Individual patient data is often insufficient for accurate pharmacokinetic modeling.
• Existing pharmacokinetic software may lack flexibility in model definition and parameter estimation.
• Integrating population pharmacokinetic parameters with individual data can improve predictive accuracy.

Purpose of the Study:

• To develop a user-friendly microcomputer program (MULTI2(BAYES)) for nonlinear multiple regression analysis in pharmacokinetics.
• To implement a Bayesian algorithm that combines individual and population pharmacokinetic data.
• To provide a flexible platform for pharmacokinetic model development and analysis.

Main Methods:

• Development of MULTI2(BAYES) using Microsoft BASIC for broad microcomputer compatibility.
• Incorporation of a Bayesian algorithm for data fusion.
• Implementation of four nonlinear least squares algorithms: Gauss-Newton, damping Gauss-Newton, Fletcher's modified Marquardt, and simplex methods.
• User-defined pharmacokinetic models with unrestricted parameter, independent, and dependent variables.

Main Results:

• MULTI2(BAYES) enables prediction of patient plasma time courses by integrating limited individual data with published population pharmacokinetic parameters.
• The program supports flexible pharmacokinetic model definition by the user.
• Multiple nonlinear regression algorithms are available for parameter estimation.
• Calculation of 95% confidence limits for predicted time courses is provided.

Conclusions:

• MULTI2(BAYES) offers a powerful and flexible tool for pharmacokinetic analysis on microcomputers.
• The Bayesian approach enhances predictive accuracy by leveraging both individual and population data.
• The software's adaptability to various pharmacokinetic models and algorithms makes it valuable for research and clinical applications.