Drug–drug interaction (DDI) is a phenomenon of modification of pharmacological activity of one drug by another co-administered drug (Min and Bae, 2017; Rekić et al., 2017). Pharmacokinetic (PK) DDIs are the most common type and happens when the absorption, distribution, metabolism, or excretion (ADME) of the substrates are affected by co-administered drugs, resulting in an increased or decreased exposure of the substrates (Min and Bae, 2017; Chung and Kesisoglou, 2018; Wu et al., 2020). Understanding how certain co-administered drugs affect the exposure changes of substrate is essential for drug development and application (Zhang et al., 2012). The Japanese Medicines and Medical Devices Administration (PMDA), European Medicines Agency (EMA), and U.S. Food and Drug Administration (FDA) have issued separate DDI guidelines, emphasizing the predictive use of comprehensive mechanism methods, such as physiologically based pharmacokinetic (PBPK) models, for quantitative evaluation of potential DDI risk in different periods of drug discovery and development (Committee for Human Medicinal Products, 2012; Drug Interaction Studies, 2012; Drug Interaction Studies, 2014). Due to the fact that most drugs are primarily eliminated via liver and kidneys, the impairment of organ function of liver and kidney is one of the important intrinsic factors for regulating drug exposure, that is, drug–disease interaction (DDZI). Consequently, the FDA and EMA also published guidelines to assess the impact of hepatic and/or renal impairment (HIs and RIs, respectively) on the pharmacokinetics of drugs, providing the dosing and labeling recommendations (Guidance for Industry, 2003; Guidance for Industry, 2010; Committee for Human Medicinal Products, 2014; Committee for Human Medicinal Products, 2015).

Apatinib is an oral small-molecule tyrosine kinase inhibitor (TKI), which can competitively bind to the tyrosine ATP binding site in vascular endothelial growth factor receptor (VEGFR), selectively inhibit the activity of VEGFR tyrosine kinases, and block the downstream signaling, therefore suppressing the tumor angiogenesis (Roviello et al., 2016; Geng et al., 2018; Meng et al., 2020). Apatinib shows good oral bioavailability and tolerance in humans. In 2014, apatinib was approved by the National Medical Products Administration for the treatment of advanced gastric adenocarcinoma or gastric esophageal junction adenocarcinoma (Roviello et al., 2016; Geng et al., 2018). Clinical trials have shown that apatinib has certain objective remission rate and survival benefit in the treatment of a variety of malignant tumors, i.e., breast cancer, non–small cell lung cancer, hepatocellular carcinoma, and colorectal cancer (Zhang, 2015; Xue et al., 2018; Fathi Maroufi et al., 2020). In vitro metabolism studies showed that apatinib was the substrate of P450 enzymes, which was mainly metabolized via CYP3A4/5 and, to a lesser extent, by CYP2D6 (Ding et al., 2013). The plasma exposure of apatinib was reported to be significantly affected by co-administration with itraconazole (AUC_0–t increased by 75%) or rifampin (AUC_0–t decreased by 83%) in humans, indicating the high potential of DDIs by affecting drug metabolism enzymes for apatinib (Liu et al., 2018). However, the effects of moderate and weak enzyme inhibitors or inducers on the exposures of apatinib have not been evaluated since cancer patients normally receive other drugs, i.e., hypoglycemics, antihypertensive, anticoagulants, and antifungal, many of which are enzyme inhibitors or inducers. Further research is needed to evaluate its potential involvement in DDIs for the rational clinical administration. In addition, the potential risk assessments of apatinib’s pharmacokinetic changes in RIs or HIs will also be of great significance to recommend appropriate dosing regimens.

PBPK modeling is a mathematical predictive approach, which can quantitatively predict ADME process of chemical substances in human or animal by extrapolating in vitro, in situ, and in silico drug-dependent parameters (Min and Bae, 2017; Wu et al., 2020). Some excellent researches and reviews have illustrated the ability of PBPK modeling in drug development (Li et al., 2021; Nauwelaerts et al., 2021; Peng et al., 2021). Hence, a growing emphasis on the application of PBPK modeling has been placed to quantitatively evaluate the potential risks of DDI and DDZI of drugs in vivo (Ono et al., 2017; Rasool et al., 2017; Rasool et al., 2021; Rasool and Läer, 2021). Accordingly, developing a PBPK model of apatinib to quantitatively assess the influence of external and internal factors on apatinib pharmacokinetics would be of great benefit. In the present research, we aimed to establish and verify PBPK models of apatinib on the basis of the reported clinical trials, then the verified PBPK model was applied to evaluate the potential DDI risks with CYP3A4 and CYP2D6 inhibitors or inducers, as well as the DDZI outcomes in HIs and RIs. We focused on gaining insights into the rational clinical administration of apatinib by using the dynamic modeling approach.

In the present study, we first developed a PBPK model for apatinib by incorporating both in vitro and in vivo data. The pharmacokinetics of apatinib at an oral dose level of 750 and 250 mg in healthy volunteers, as well as 750 mg dose level in cancer patients, were predicted. The results showed that the pharmacokinetic profiles of apatinib matched well with the clinically observed one. Baker et al. have reported that age, sex, and body size have negligible influence on the CYP3A activity in 134 cancer patients (Baker et al., 2004). Consistently, midazolam, a CYP3A probe substrate, was predicted to have comparable exposure between healthy volunteers and cancer patients using PBPK modeling (Cheeti et al., 2013). The PBPK model developed by Chiho et al. captured comparable pharmacokinetics of bosutinib in healthy volunteers with cancer patients as well (Ono et al., 2017). Therefore, the Simcyp default “Chinese Healthy Volunteers” population file was used both in simulation with healthy volunteers and cancer patients. Our study also demonstrated the similar simulation in these two populations. Afterwards, the validation of the apatinib DDI model was conducted using the reported clinical study, which showed the successful simulation of the exposure changes of apatinib in the presence or absence of itraconazole and rifampicin (Liu et al., 2018).

Then, the potential risks of DDI and DDZI of apatinib were estimated using the verified PBPK model. Owing to that cancer patients are frequently treated with multiple medications, the potential DDI risks of apatinib co-administered with other drugs must be determined. Apatinib was reported to be metabolized primarily via CYP3A4/5 and CYP2D6 (Ding et al., 2013). Therefore, the verified PBPK model was applied to evaluate the potential DDI risks involving CYP3A4 and CYP2D6-mediated metabolizing pathways. Erythromycin and verapamil were moderate CYP3A4 inhibitors (Zhou, 2008; Yamazaki et al., 2015). In our simulations, apatinib exposure was obviously increased in subjects concomitantly treated with erythromycin and verapamil. The predicted AUC of DDI ratio was 3.60 and 2.17 with co-administration of erythromycin and verapamil, respectively. The presence of paroxetine, a strong CYP2D6 inhibitor, could also lead to a moderate increase of the exposure of apatinib by 1.29-fold. The simulated result was also consistent with the fact that the metabolism of apatinib via CYP3A4 was much higher than that of CYP2D6. Thus, attention should be paid to the co-administration of apatinib with moderate CYP3A4 inhibitors and the corresponding dose adjustment in clinic. In the DDI simulation with CYP3A4 inducers, the predicted apatinib exposure decreased by 71.8, 61.6, and 63.9% in the presence of carbamazepine, efavirenz, and phenytoin, respectively, indicating that we should be careful about the pharmacokinetic interactions of apatinib with CYP3A4 inducers, which is likely to impair the efficacy.

It has been reported that cirrhosis could decrease gastrointestinal absorption and reduce the expression of drug-metabolizing enzymes as well as transporters (Verbeeck, 2008). The renal impairment could cause decreases in renal clearance along with other changes such as activity of drug-metabolizing enzyme, especially in CYP3A4 and reduced plasma protein binding (Yeung et al., 2014). Thus, the DDZI risks should be assessed in HIs and RIs for appropriate dosing recommendations. In the DDZI simulation with HIs, the AUC of apatinib was significantly increased by 2.25-fold and 3.04-fold for Child–Pugh B and Child–Pugh C, respectively, with slight but consistent decrease of C_max by 1.54- and 1.67-fold, respectively. The results were in accordance with the alteration of absorption and the decrease in CYP enzyme-mediated clearance of apatinib due to decreased enzyme abundances in patients with HIs. The model-predicted AUC ratios in moderate and severe RIs were 1.06 and 1.01, respectively, suggesting the negligible impact of renal impairment on the exposure of apatinib. In view of the challenges to conduct various DDI and DDZI studies in cancer patients, the PBPK model would be of great benefit to quantitatively predict pharmacokinetic changes under different circumstances which might be difficult to evaluate clinically, so as to avoid some DDI and DDZI risks in advance.

However, some limitations in this study warrant further discussion. First, the present study focused on the pharmacokinetic changes of apatinib as a victim drug; the impact of apatinib as perpetrators on other drugs being mainly metabolized via CYP3A4 or CYP2D6 has not been evaluated. It has been demonstrated that apatinib inhibited the metabolism of gefitinib in vitro and in vivo via CYP2D6 and CYP3A4 (Wang et al., 2021). Zhu et al. have investigated the potential of apatinib as a perpetrator on the pharmacokinetics of nifedipine and warfarin in advanced solid tumor patients. The results indicated that concomitant administration of apatinib led to an obvious increase to nifedipine and warfarin exposure (Zhu et al., 2020). Due to the risk of pharmacokinetic DDI based on enzyme inhibition by apatinib, a dynamic PBPK model is needed for further dosing recommendations when apatinib is co-administered with CYP3A4, CYP2D6, or CYP2C9 substrates. Second, we only evaluated the DDZI outcomes at a single dose level; the multiple-dose DDZIs were not incorporated in the simulation, although the comparable changes of exposures in HIs between single- and multiple-dose administrations were found in some drugs such as bosutinib (Ono et al., 2017). Third, the model-predicted impacts of HIs and RIs on the pharmacokinetics of apatinib required further clinical verification, considering potential physiological changes resulted from impaired hepatic or renal function. There exist considerable challenges to accurately incorporate the complex physiological changes into PBPK models due to the changes of hepatic architecture in HIs. As a result, bosutinib exposures could not be well predicted in HIs compared with the clinical observed data by Chiho et al. (Ono et al., 2017). The apatinib PBPK model is expected to be further refined using the post-marketing observed data in HIs and RIs, and therefore more adequately evaluating the potential DDZI risks.

All in all, the present study showed that the verified PBPK model of apatinib could provide mechanistic insights for further understanding the potential risk of DDIs and DDZIs, providing the reasonable dosing recommendations.

In the present study, we have developed and verified a dynamic PBPK model of apatinib based on the currently available data. The models can be used to predict apatinib exposures in the single-dose DDIs with other CYP inhibitors and DDZIs in HIs or RIs. The DDI prediction indicated significant changes (2–4-fold) in apatinib exposures with moderate CYP3A4 inhibitors or CYP3A4 inducers. A moderate increase of apatinib exposure (1.25–2-fold) was found with strong CYP2D6 inhibitor. Clinicians should be cautious about the necessary dose adjustment of apatinib being used with these perpetrators. In the DDZI simulation with HIs, the AUC of apatinib was significantly increased by 2.25-fold and 3.04-fold for Child–Pugh B and Child–Pugh C, respectively, with slight but consistent decrease of C_max by 1.54- and 1.67-fold, respectively. The DDZI prediction suggested a negligible increase in apatinib exposures in RIs. The PBPK models could be of great benefit to give some recommendations in DDI and DDZI risk assessments, and replace some clinical studies as well. The verified PBPK model could provide a rough understanding of the extrinsic and intrinsic factors on the exposure of apatinib.