DISCUSSION
Accurate prediction of potential DDIs is crucial in ensuring patient safety and efficacy of drugs during the drug discovery and development process. PBPK-DDI models rely on the accurate representation of various physiological parameters and processes to predict the DDIs. We used two methods to determine the metabolism and fm of test articles for improvement of DDI prediction are as follows: i) In vitroapproach using HLM. This method directly measures the fraction metabolized by simulating the metabolic process in the human liver. ii) in silico approach using mathematical models. This method predicts the metabolism and fm based on the molecular structure and properties of the substance, without the need for laboratory experimentation. By incorporating differentfm values by in vitro method or in silico method, instead of assuming 100% values, the accuracy of PBPK models can be improved, leading to more precise predictions. The use of actualin vitro determined values for predicting DDIs leads to improved accuracy compared to using in silico methods. This can help to better understand the pharmacokinetics of drugs in different populations and inform drug development and regulatory decisions.
In addition to liver microsomes, other in vitro systems such as hepatocytes and recombinant enzymes can be used to determine the metabolic fate of drugs, fm , and evaluate their potential for DDIs (Lindmark, Lundahl, Kanebratt, Andersson & Isin, 2018; Youdim et al., 2008). Each system has its own advantages and limitations, and the choice of system depends on the specific research question being addressed and the properties of the drug under investigation. Hepatocytes contain a much broader range of metabolic enzymes such as alcohol dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH), which are involved in the metabolism of alcohol, and glucuronosyltransferases, sulfotransferases, and glutathione S-transferases (Klammers et al., 2022; Lindmark, Lundahl, Kanebratt, Andersson & Isin, 2018). Herein, the hepatocytes have been considered better for in vitro evaluation on fm as they provide a more accurate representation of the metabolic activities of the liver (Lindmark, Lundahl, Kanebratt, Andersson & Isin, 2018). However, microsomes are still widely used due to their ease of preparation, cost-effectiveness, and stability, and they still provide valuable information about drug metabolism and evaluation onfm . In this paper, the prediction of DDIs using the PBPK-DDI model with incorporation of fmvalues from HLM is relatively accurate and does not significantly overestimate the results. Since most of these drugs are primarily metabolized by the P450 enzyme system in the liver, the in vitrostudies using HLM can provide a good prediction of the potential for DDIs in vivo . If other enzymes, such as ADH, ALDH, glucuronosyltransferases, sulfotransferases, and glutathione S-transferases are involved into drug metabolism, hepatocytes are a better choice for determining the fm as they contain these various enzyme systems and can provide a comprehensive view of the drug metabolism process.
The study found that Clliver and fraction metabolized by CYP3A4 are significant factors in understanding DDI as showed in Table 1 and Figure 1. Drugs with high Clliver are more likely to be affected by CYP3A inhibitors, which can result in changes in the drug’s exposure and potential interactions. There is total 6 victim drugs which exposure increased higher than 3 times among 33 compounds when co-administered with ketoconazole and all of these six victim drugs possess higher Clliver (>15 L h-1). On the other hand, drugs with low Clliver may not see significant changes in clearance when exposed to CYP3A inhibitors as the body’s elimination processes for the drug are already at their limit. There are 18 drugs with lower Clliver (<15 L h-1) among the 33 drugs and their exposure increase less than 3 times after co-administered ketoconazole. When the fm is low, it indicates that the CYP3A isoform is not a major contributor to the metabolism of a drug, and therefore, inhibitors of CYP3A such as ketoconazole may not have a significant effect on the metabolism of the drug. Herein, the above 6 victim drugs with exposure increased higher than 3 times not only possess higher clearance (>15 L h-1) but also higher fm(>75%).
The tested fm values from the two methods are not consistent with the actual, real-life conditions of the biological system being studied, as showed in Figure S5. It seems to indicate a problem with the methods used or a limitation in the ability of the methods to accurately reflect the true situation in vivo . However, the predictions of DDIs were improved much by integration ofin vitro fm , which are indeed more accurate than predictions based on in silico fm or 100% offm . The underlying cause of this phenomenon is that variation between in vitro and in vivo measurements of fm is lower than that between in silicofm and in vivo fm for the drugs with high clearance as showed in Figure 2. The difference betweenin vitro fm and in vivo fmis unlikely to have a significant impact on the assessment of potential DDIs since the evaluation of low-clearance compounds for DDIs is not sensitive to variations in fm . The discrepancy between in vitro and in vivo measurements offm is always observed for compounds with low clearance. The lower the clearance value, the slower the rate of drug metabolism, leading to smaller changes in drug concentration in HLM system. The limited accuracy of detection (±15%) can make it challenging to determine the effect of inhibitors on the drug’s metabolism.
We used the remaining amount of substrate and in vitro clearance as an index to calculate the fraction metabolized by CYP3A4 with or without an inhibitor respectively. The first method is simple and straightforward and provides a rough estimate of the fraction metabolized by CYP3A4 by using the remaining amount of substrate as an index. On the other hand, the second method provides a more controlled and direct way to evaluate the metabolic fate of the substrate by measuring the rate of metabolism in a laboratory setting and provide more accurate and precise results compared to the first method.
It is important to note that abiraterone and naloxegol were not enrolled in this paper. Abiraterone is a pro-drug, which metabolism process may involve multiple enzyme systems. Therefore, a more comprehensive evaluation of multiple enzyme systems is necessary to accurately assess the fm of abiraterone. The present PBPK-DDI model considers the metabolic pathways and enzymes involved in drug metabolism, but transporters also play a significant role in DDIs and cannot be overlooked. Naloxegol is a substrate for both CYP3A and a transporter, the developed PBPK-DDI model may not accurately predict the potential for DDI, as the interplay of metabolic and transport mechanisms is complex. The DDI risk of naloxegol is obviously under-estimated using the PBPK-DDI model. Therefore, it may further improve the prediction to consider both metabolic and transport mechanisms when evaluating the potential for DDIs, to provide a more comprehensive understanding of the interactions between drugs.