Discussion

To the best of our knowledge, this is the first study that measured CPI kinetics in a FIH dose escalation study to evaluate the transporter-mediated DDI potential of an NME.

Most PK parameters for GDC-0810 administered after a 600 mg dose to BC patients in this study were similar to those in healthy subjects reported in the DDI study with pravastatin (Liu et al., 2018). Food intake numerically increased the C_max and area under the concentration-time curve at 24 hours, but statistical significance was not assessed given the limited sample size and exhibited large variability. The PK of 800 mg once daily was similar to that of 600 mg once daily, yet 800 mg was deemed intolerable due to gastrointestinal side effects (safety data on file). Collectively, the recommended phase II dose was determined to be 600 mg once daily administered under fed conditions.

The in vitro inhibition study suggested that the two major glucuronide metabolites of GDC-0810 were potent inhibitors of OATP1B with IC₅₀ comparable to that of the parent drug (data on file). However, considering the much lower plasma concentrations (3% and 10% of parent drug at 600 mg for N-glucuronide and acyl-glucuronide, respectively), we concluded that these two metabolites would not contribute significantly to the OATP1B-mediated DDI. Therefore, only the impact of GDC-0810 on OATP1B-mediated DDI was considered in subsequent analyses.

A positive trend between the plasma concentrations of CPI and GDC-0810 exposure was observed (Fig. 3). As expected, the extent of the change in CPI plasma concentrations was comparable to that observed in the dedicated DDI study, illustrating that CPI is a sensitive biomarker for the OATP1B-mediated DDI pathway (Liu et al., 2018). There are a few reported studies evaluating the interaction between endogenous biomarkers and multiple dose levels of transporter inhibitors (Tsuruya et al., 2016; Takehara et al., 2018). This is the first time that the extent of DDI potential of a perpetrator was studied for a wide range of drug exposures in patients using an endogenous compound. For NMEs that cause potent in vivo inhibition of transporters, strategies to mitigate the risk of DDI at various doses of the perpetrator drug can be formulated based on the change in CPI plasma concentrations from baseline and predicted magnitude of DDI at different exposures observed in the FIH study.

The kinetic profiles and the estimated parameters for CPI, such as the rate of degradation, C_max, and T_max in this study, were compared with those from the previously reported dedicated DDI study in healthy volunteers (Liu et al., 2018; Yoshida et al., 2018), and were found to be comparable in the presence of GDC-0810. When the BC patients were given 600 mg GDC-0810, they had similar CPI baseline values as healthy subjects in the clinical DDI study (Fig. 4); however, in general, when all patients were considered in the population kinetic analysis as reported in Table 5, the CPI baseline levels were slightly lower in patients than that in healthy female volunteers (Yoshida et al., 2018). This difference may be due to interindividual variability instead of the disease state. Among healthy volunteers, previous studies reported that the CPI baseline levels could range from approximately 0.6 nM (Kunze et al., 2018; Shen et al., 2018) to 0.97 nM (Lai et al., 2016). The estimated K_i,u value in the present study was more than 2-fold higher than the previous study, accompanied by much larger interindividual variability, indicating that the point estimate is less reliable. However, the estimated K_i,u value in this study was still approximately 100-fold smaller than the in vitro IC₅₀ values [<0.3 µM (OATP1B1) and 0.9 µM (OATP1B3)], which is consistent with the previous observation (Liu et al., 2018; Yoshida et al., 2018). Although part of the discrepancy might be explained by the reduced effective concentration in the in vitro study for this highly bound molecule (the plasma unbound fraction is 0.008%) (Chen et al., 2018), quantitative extrapolation of in vitro IC₅₀ to in vivo K_i remained challenging. This once again highlights an advantage of using biomarker kinetic profiles to estimate the K_i,u value, which results in a more accurate prediction of clinical DDI. Notably, the T_max value of CPI in the single dose analysis (Fig. 4) was slightly delayed to 6 hours from the median of 3 hours. This difference can be explained by the PK of GDC-0810, which changed the plasma concentration of CPI. A delayed peak after one dose of 600 mg GDC-0810 was seen in two patients in this cohort, which consequently led to the delayed CPI peak. It is uncertain if other disease states would affect the CPI levels and its kinetic parameters; for instance, reduced OATP1B1/3 expression has been reported in patients with hepatocellular carcinoma (Obaidat et al., 2012). Therefore, measuring CPI levels in patients in a FIH study instead of a DDI study, which typically enrolls healthy volunteers, may provide a better reflection of the extent of an OATP1B-mediated DDI in prospective patients.

The simulated fold change in CPI plasma concentrations from baseline at 3 hours postdose using a developed model (Fig. 5) showed a positive trend despite the large empirical bootstrap confidence intervals. Furthermore, our PBPK DDI simulation based on the estimated K_i,u value resulted in underestimation of the point estimate, and again with a large extent of uncertainty in the predicted AUCR of the victim probe drug, which ranged from no DDI to mild DDI (AUCR = 1.02–1.35) (Fig. 6). This uncertainty in the simulation was likely attributed to the relatively small sample size per cohort in this substudy. Therefore, and more importantly, since the magnitude of interindividual variability is also wide, it is challenging to quantitatively characterize the interaction between GDC-0810 and CPI from the current data. Interestingly, Barnett et al. (2018) evaluated sample size requirement for CPI to detect different levels of inhibition and found that 15 subjects are needed to achieve 80% power at α = 0.05 or 0.01 for weak (AUCR > 1.25) inhibition. Indeed, we had <5 patients in the 600 or 800 mg groups with CPI measurement. As a comparison, in the clinical DDI study where 15 healthy subjects were enrolled and administered 600 mg GDC-0810, the uncertainty in the predicted AUCR of pravastatin was much narrower (Fig. 6). These observations support the need for adequate sample size to accurately quantify the magnitude of DDI, especially when the effect of the NME on OATP1B is small. In contrast, for strong DDIs, which are more likely to be clinically important, a standard FIH cohort can be sufficient to capture the interaction. Collectively, our study illustrates the utility of model-based analysis of an endogenous biomarker in a FIH study to provide a quantitative estimation of the extent of DDIs. The proposed approach, accompanied by sufficient sample size, may pave the way to justifying that GDC-0810 has low clinically relevant potential of driving OATP1B-mediated DDI.

Recently, a PBPK model for CPI has been established that predicts the DDI caused by rifampicin via the OATP1B and multidrug resistance-associated protein pathway (Mori et al., 2019). This model enables the mechanistic understanding of the complex DDI mediated by the studied perpetrator based on in vitro information; however, it also requires in-depth understanding of the substrate-dependent, and potentially patient-dependent, biomarker kinetic processes for model development. Our current approach complements this PBPK model; it allows the evaluation of interindividual variability in the estimated parameters by using a population PK approach, and hence the magnitude of the DDI can be better characterized and would be more reflective of the expected DDI potential of the perpetrator in a real patient population. Finally, it is noteworthy that in addition to CPI, there are other endogenous biomarkers for OATP1B that are being studied, which may be measured concurrently with CPI in the future to predict transporter-mediated DDIs with greater confidence (Takehara et al., 2018; Barnett et al., 2019).

Here, we mention two limitations of this study. First, the effect of OATP1B polymorphism on the CPI baseline and kinetics in the presence of GDC-0810 was not investigated. OATP1B polymorphism has been associated with many clinically relevant impacts on the disposition of drug substrates (Niemi et al., 2005; Ho et al., 2007; Deng et al., 2008), and recently with CPI and other biomarkers (Mori et al., 2019). Therefore, it is of great interest to study how the interactions between CPI and inhibitors for OATP1B would be affected by the transporter pharmacogenomics. The second limitation is the inevitably confounding effect of concomitant medications; all patients enrolled in this study reported taking at least one medication before or after study initiation. While known strong OATP1B inhibitors, such as cyclosporine, were restricted in the study, some medications might weakly inhibit the transporters.

In conclusion, the PK results of GDC-0810 and its two glucuronide metabolites in ER+ human epidermal growth factor receptor 2-negative (HER2-) BC patients were characterized in this FIH dose escalation study. Furthermore, the dose-dependent increase in CPI concentration suggested the presence of weak in vivo inhibition of OATP1B. The prediction of DDI potential based on population PK and PBPK modeling approaches resulted in some underestimation with relatively large uncertainty, likely due to limited sample size. This study provides important insights into the novel avenue to quantitatively predict transporter-mediated DDI as early as in FIH studies of the NME using endogenous biomarkers.