Background and objective: Pharmacokinetic drug-drug interactions (DDIs) are one of the major causes of adverse events in pharmacotherapy, and systematic prediction of the clinical relevance of DDIs is an issue of significant clinical importance. In a previous study, total exposure changes of many substrate drugs of cytochrome P450 (CYP) 3A4 caused by coadministration of inhibitor drugs were successfully predicted by using in vivo information. In order to exploit these predictions in daily pharmacotherapy, the clinical significance of the pharmacokinetic changes needs to be carefully evaluated. The aim of the present study was to construct a pharmacokinetic interaction significance classification system (PISCS) in which the clinical significance of DDIs was considered with pharmacokinetic changes in a systematic manner. Furthermore, the classifications proposed by PISCS were compared in a detailed manner with current alert classifications in the product labelling or the summary of product characteristics used in Japan, the US and the UK.
Methods: A matrix table was composed by stratifying two basic parameters of the prediction: the contribution ratio of CYP3A4 to the oral clearance of substrates (CR), and the inhibition ratio of inhibitors (IR). The total exposure increase was estimated for each cell in the table by associating CR and IR values, and the cells were categorized into nine zones according to the magnitude of the exposure increase. Then, correspondences between the DDI significance and the zones were determined for each drug group considering the observed exposure changes and the current classification in the product labelling. Substrate drugs of CYP3A4 selected from three therapeutic groups, i.e. HMG-CoA reductase inhibitors (statins), calcium-channel antagonists/blockers (CCBs) and benzodiazepines (BZPs), were analysed as representative examples. The product labelling descriptions of drugs in Japan, US and UK were obtained from the websites of each regulatory body.
Results: Among 220 combinations of drugs investigated, estimated exposure changes were more than 5-fold for 41 combinations in which ten combinations were not alerted in the product labelling at least in one country; these involved buspirone, nisoldipine and felodipine as substrates, and ketoconazole, voriconazole, telithromycin, clarithromycin and nefazodone as inhibitors. For those drug combinations, the alert classifications were anticipated as potentially inappropriate. In the current product labelling, many inter-country differences were also noted. Considering the relationships between previously observed exposure changes and the current alert classifications, the boundaries between 'contraindication' and 'warning/caution' were determined as a 7-fold exposure increase for statins and CCBs, and as a 4-fold increase for BZPs. PISCS clearly discriminated these drug combinations in accordance with the determined boundaries. Classifications by PISCS were expected to be valid even for future drugs because the classifications were made by zones, not by designating individual drugs.
Conclusion: The present analysis suggested that many current alert classifications were potentially inappropriate especially for drug combinations where pharmacokinetics had not been evaluated. It is expected that PISCS would contribute to constructing a leak-less alerting system for a broad range of pharmacokinetic DDIs. Further validation of PISCS is required in clinical studies with key drug combinations, and its extension to other CYP and metabolizing enzymes remains to be achieved.