Individualized Drug Dosage

Martin Fransson is a PhD student at the Programming Environments Laboratory at Linköping University. He explains that his research field is individualized drug dosage, meaning basically that he is trying to build time-dynamical models of the disposition and metabolism of different drugs. “Building such models are in many cases a difficult task due to the complexity of the system and the often relatively sparse data sets, but the combination of sophisticated software solutions and new measurement technology now starts to make it possible”, Fransson explains.

The goal with Fransson’s research is to investigate the possibilities to individualize drug dosage so that a patient undergoing treatment receives an optimal dose based on her personal physiological profile. Recently Fransson has simulated and analyzed different models describing the distribution and metabolism of a cytotoxic drug used to treat women with ovarian cancer. Although data from 33 patients had previously been collected during a clinical study, this data was not complete for all the necessary modeling purposes, something that was first a problem. “However, the combination of MathModelica System Designer’s user-friendly modeling environment and the possibilities to analyze the results with Mathematica’s rich functionality proved to be a powerful tool”, he says, “the solution provided a fast way to create simulated data that could complement the clinical study.”

However, Martin noticed that adding the simulated data was not enough to provide a stable model. A sensitivity analysis on the existing models was also needed to reveal which parameter that should be fixed to a prior value. The specialized built-in Mathematica function NIntegrate could be applied on the simulation results from MathModelica System Designer, and gave a measure of the model sensitivity for different scenarios. “Thanks to MathModelica System Designer Professional the parameter that had the least influence on the model could be identified and fixed to a prior value, thus stabilizing the model.“, Fransson concludes.