Risk assessment for drug degradation products using physiologically-based pharmacokinetic models

[摘要] Degradation product toxicity is a critical quality issue for a small group of useful drug products--e.g. lidocaine, isoniazid, chlorhexidine, gabapentin. In the traditional risk assessment approaches, a no-observed-adverse-effect level (NOAEL) derived from animal data is determined with the use of generic (and arbitrary) uncertainty factors to obtain an acceptable daily intake. The effects of compound-specific biological complexities and pharmacokinetics are typically not part of the risk calculations. The selection of uncertainty factors that account for interspecies or intraspecies difference concerning biokinetics and biodynamics has also generally failed to consider chemical-specific mechanism information or pharmacokinetics data. The use of combining in-vitro biopharmaceutical characterization methods and physiologically-based pharmacokinetic modeling has undergone extensive study and validation for predicting clinical drug blood level time profiles. The rationale for the proposed research is that a PBPK modeling utilizing rat to human scaling for target tissue toxicity in combination with the Monte Carlo method for estimating human target exposure distributions provides a rational basis for assessing drug stability safety issues for drug substances that potentially degrade to toxic compounds. PBPK models for rats and humans were developed to simulate drug exposure time profiles after oral administration of model compounds including aniline, p-chloroaniline, 2,6-dimethylaniline, o-toluidine and p-aminophenol. The PBPK models were parameterized using a combination of literature values, computational models and standard in vitro experiments. Microsomal and hepatocyte metabolism studies were used to estimate the metabolic constants, and ultrafiltration was used to measure protein binding. Intestinal permeability was predicted using a set of related compound data to correlate measured Caco-2 permeability with molecular descriptors by multivariate regression. Sensitivity analyses were conducted to evaluate the impact of PBPK model parameters on plasma level predictions. To evaluate patient population effects on exposure profiles, the PBPK model parameters were varied in meaningful ways using Monte Carlo methods. Based on population PBPK models, distributions of target tissue exposure in rats and humans were simulated and compared to derive human safe dose. As results, rat PBPK model-predicted aniline concentration time profiles were in reasonable agreement with published profiles. Distributions of target tissue exposure in rats and humans were generated and compared based on a criterion. A human reference dose was then selected at a value of 1% criteria. This approach was compared to traditional risk assessment calculations. In conclusion, the PBPK modeling approach resulted in drug degradation product risk specifications that were less stringent than those estimated by conventional risk assessment approach.

[发布机构] University of Iowa