摘要：A cocktail study is an in vivo evaluation method to assess multiple CYP activities via a single trial and single administration of a cocktail drug that is a combination of multiple CYP substrates. However, multiple blood samples are required to evaluate the pharmacokinetics of a CYP probe drug. A limited-point sampling method is generally beneficial in clinical studies because of the simplified protocol and reduced participant burden. The aim of this study was to evaluate whether a limited-point plasma concentration analysis of CYP substrates in a cocktail drug could predict their area under the curve ( AUC ). We created prediction models of five CYP substrates (caffeine, losartan, omeprazole, dextromethorphan, and midazolam) using multiple linear regressions from the data of two cocktail studies, and then performed predictability analysis of these models using data derived from data in the co-administration with inducer (rifampicin) and inhibitors (fluvoxamine and cimetidine). For the administration of inhibitors, the AUC prediction accuracy (mean absolute error (MAE)) were <39.5% in Model 1 and <26.2% in Model 2 which were created using 1- and 4-point sampling data. MAE shows larger values in the administration of inducer in compared with the administration of inhibitors. The accuracy of the prediction in Model 2 could be acceptable for screening of inhibitions. MAE for caffeine, dextromethorphan, and midazolam were acceptable in the model that used 4 sampling points from all data. The use of this method could reduce the burden on the subject and make it possible to evaluate each AUC in a minimally invasive manner.
