BACKGROUND AND PURPOSE Enhancing the precision of drug-drug
interaction (DDI) prediction is essential for mitigating potential drug
interactions and enhancing drug safety and efficacy. This study aims to
investigate the impact of in vitro and in silico approaches for
calculating the fraction metabolized by CY3A4 (fm) on DDI prediction
accuracy and identified the most effective method for improving DDI
prediction using physiologically based pharmacokinetic (PBPK) models.
EXPERIMENTAL APPROACH Both in vitro and in silico methods were utilized
to determine fm values for 33 approved drugs, or fm values were assumed
to be 100%. These fm values were then integrated into PBPK models.
Subsequently, the PBPK models were combined with a PBPK model of
ketoconazole to predict potential DDIs. Finally, the accuracy of these
predictions was assessed. KEY RESULTS The integration of in vitro fm had
remarkable precision in predicting CmaxR of 31 drugs and accurately
predicting AUCRs of 28 drugs out of 33 drugs, both within 2 times of the
measured values. However, using 100% fm and in silico fm resulted in
lower prediction accuracy that was comparable to each other. CONCLUSIONS
AND IMPLICATIONS Our study highlights the importance of incorporating in
vitro fm data into PBPK models to improve the accuracy of predicting
DDIs. While in silico fm may have some potential, its influence on
predictions appears to be limited. Additionally, our findings suggest
that drugs with high Clliver levels (>15 L·h-1) and high fm
(>75%) are particularly susceptible to the impact of
CYP3A4 inhibitor ketoconazole.