There is an increasing trend of earlier globalization of clinical trials to allow for faster patient recruitment and to gather clinical data across ethnically diverse patient populations (eg, including Asia) that can eventually be used for global registration and early access to new medicines.1 This is particularly the case for diseases with unmet medical need where speed is vital for access to life-saving medicines throughout the world. Ethnic differences in study populations and their impact on the drug's pharmacokinetics (PK) and pharmacodynamics are important considerations for selection of an appropriate safe and efficacious dose across populations to optimize the benefit:risk profile and to decrease the lag times for global access to novel therapies.2 A key consideration is the influence of genetic polymorphisms in drug-metabolizing enzymes (DMEs) and/or transporters that may contribute to clinically relevant differences in drug clearance. Regional/racial differences in the frequencies of major genetic polymorphisms in DMEs and associated extensive metabolizer (EM)/poor metabolizer (PM) phenotypic incidence are well recognized for multiple cytochrome P-450 (CYP) enzymes (eg, CYP2C19, CYP2D6, CYP3A5) .3-7 Additionally, racial differences in the abundance and/or intrinsic activity of DMEs or transporters can translate to ethnic differences in drug exposure, as established for the OATP1B1 substrate rosuvastatin, for which the higher exposures in Asian patient populations are linked not to a specific genetic polymorphism but rather to a lower intrinsic activity/capacity of hepatic uptake transport.8 Of note, for a polymorphic enzyme with genetically determined PM and EM phenotypes, when racial differences in enzyme abundance or functional activity/capacity are observed, we posit that this can lead to complex race-by-genotype interactions that may require special consideration in population PK covariate analyses. Herein, we illustrate the discovery of such a race-by-genotype interaction hypothesis for the DME CYP2C19 using population physiologically based PK (PBPK) modeling and simulation and discuss broader implications of these observations for global drug development strategies and covariate analyses in population PK modeling. CYP2C19 is polymorphic with ∼18% frequency of poor metabolizers (PMs) in Japanese and ∼2% in whites.9 In addition, it has been reported that the liver abundance of CYP2C19 is 3.5-fold lower in Japanese compared to whites.10 The lower abundance of enzyme could mean that the relative contribution of CYP2C19, and therefore the impact of genetic polymorphism on the PK of CYP2C19 substrates, may differ between these populations, although there is little published literature directly addressing this potential for a race-by-genotype interaction. This hypothesis finds some support in the observation that the PM/EM ratio in systemic exposures (AUC) of omeprazole, a CYP2C19 substrate, is greater in whites (∼12-fold) than in Chinese subjects (∼5-fold).11 Furthermore in a drug-drug interaction study of the partial CYP2C19 substrate diazepam with the selective CYP2C19 inhibitor omeprazole, Caraco et al 12 reported differential inhibition of diazepam clearance in whites versus Chinese. The effect of a selective and potent CYP2C19 inhibitor on diazepam PK may be viewed in similar light as comparing diazepam exposure in PMs vs EMs.13 The interaction magnitude with omeprazole appears to be larger in whites (39% reduction in clearance) than in Chinese subjects (22% reduction in clearance), suggesting again that the relative contribution of CYP2C19 to overall diazepam clearance is greater in whites compared to the Chinese population, consistent with lower CYP2C19 abundance in Asian than in white populations.10 These examples, taken together with biochemical reports of lower CYP2C19 hepatic abundance in Asian populations, lend support to the hypothesis of a race-by-genotype interaction that we have quantitatively explored further using a PBPK approach. A PBPK model-based approach was used, and a series of 8 virtual compounds, assumed to be cleared exclusively by hepatic metabolism (via CYP2C19 and CYP3A4) were created. The compounds had a fixed CYP2C19 intrinsic clearance (CLint) of 5 μL/(min·pmol microsomal protein) but varied in their CYP3A4 CLint (0.01 to 1.0 μL/[min·pmol microsomal protein]) to reflect a range of relative contributions of CYP2C19. Virtual trials (10 trials of 16 subjects each) were simulated using the North European white and Japanese populations in SimCYP® V14 (Certara, Princeton, New Jersey) that are based on meta-analyses of population-specific physiologic and biochemical parameters.9 As expected, these simulations (Figure 1) show that CLpo values of all compounds are greater in white vs Japanese EMs, explained at least in part by higher absolute abundances of both CYP2C19 and CYP3A4 in whites (liver and gastrointestinal abundance of 14 and 1.5 pmol/mL, respectively, in whites vs 4.77 and 0.511 pmol/mL in Japanese9, 10). Importantly, Japanese are predicted to have smaller PM:EM AUC ratios because of the lower relative contribution of CYP2C19 in Japanese vs whites when differences in enzyme abundance across the populations are considered (as illustrated in the representative pie charts for compounds 1, 4, and 8). The results of these in silico PBPK simulations support the hypothesis that isoform-selective differences in enzyme abundance between Japanese and white populations for polymorphic enzymes can result in quantitatively different CYP2C19 PM:EM exposure ratios by race. Further, as expected the PM vs EM differences in systemic exposure are greater when the percentage relative contribution for CYP2C19 is greater. Based on this analysis, the impact of genetic polymorphism on the PK of CYP2C19 substrates is predicted to be smaller in Japanese compared to whites, suggesting that a dose modification for CYP2C19 PMs recommended based on PM:EM exposure ratios in 1 population may not necessarily be quantitatively translatable to the other population. This presents an interesting dilemma in the context of global drug development. These considerations of possible interactions between race and CYP2C19 genotype effects on PK can be especially important for narrow-therapeutic-index drugs where precise dose determination in subpopulations is crucial for safe and effective pharmacotherapy. It should be noted that because limited data exist to support the differences in DME abundance between the races, these findings should be considered as hypothesis generating and serve as a call for further research at both the biochemical and clinical pharmacologic levels to evaluate this hypothesis further and refine inputs for PBPK modeling and simulation of drugs in heterogeneous global populations. Looking ahead, PBPK modeling and simulation may represent an efficient approach to bridging these gaps without necessarily studying the PK of investigational agents in all 4 race × genotype groups (eg, through prediction of exposures in white PMs based on a PBPK model that is adequately qualified for predictive performance of clinical PK in white and Japanese EMs and Japanese PMs). A proposed flowchart for such model-based efficient bridging across races and genotypes for a CYP2C19 substrate is offered in Figure 2. The approach begins with construction of a PBPK model based on physicochemical and in vitro preclinically scaled inputs of absorption, distribution, metabolism, and excretion (ADME) information (ie, bottom-up approach, step 1). This is followed by calibration of the model using observed clinical PK/ADME data in the white EM population in early clinical development (ie, top-down approach), assuming the common scenario of clinical development initiating in the West (step 2). This PBPK model is then used to predict PK in Asian populations to guide dose selection for Asian EMs and PMs (step 3). PK studies in the Asian population (eg, Japanese) are then conducted in genotyped cohorts of EMs and PMs, with dose selection guided by the PBPK model-based predictions. The PBPK model should subsequently be assessed for its predictive performance of PK in the Japanese EM and PM populations, with further recalibration, as needed, using the observed Japanese EM and PM data to qualify predictive capability (step 4). These iterative model refinement steps should consider and apply strategies discussed for PBPK model development and verification based on emerging clinical PK data.14 For example, an objective comparison via overlay of 90% prediction intervals from the population PBPK and population PK models can help qualify the ability of the PBPK model to recover not only the typical concentration-time profile but also the associated intersubject variability. Finally, when Japanese PM PK data become available, the final model should be qualified/recalibrated as needed using all available mechanistic information on ADME of the molecule and clinical PK data to adequately predict all 3 studied population groups (white EM, Japanese EM, Japanese PM), thereby establishing readiness to forecast PK in the white PM population (step 5). Importantly, because the CYP2C19 PM frequency in whites is very low, it would be impractical to conduct a PM vs EM PK study in whites with a sample size adequate to precisely estimate the PM/EM exposure ratio. Accordingly, the above-described model-based “bridging” approach presents an efficient approach to inform expectations in this low-incidence population. The model-based estimates of PK, viewed in context of the compound's therapeutic index and overall extent and sources of PK variability, can be used to guide the decision of whether the dose determined in Asian PMs can be safely evaluated in PMs at large, including whites, with population PK and safety characterization during phase 3 (step 5) for underwriting the safe use of the drug at appropriate doses across all race × genotype groups. Although the focus of this commentary is on CYP2C19, the general principles that follow may apply to other polymorphic enzymes with racial differences in enzyme abundance/activity. For example, the reported differences in hepatic microsomal abundance of CYP2B6 between Japanese (3 pmol/mg protein) and whites (17 pmol/mg protein)3-7 may suggest a hypothesis for differential effects of pharmacogenetic variation in CYP2B6 on the PK of substrates of this enzyme in Japanese vs whites. In conclusion, the quantification of race × genotype covariate interactions based on an adequately qualified PBPK model can facilitate efficient and safe global drug development across race and genotype groups for substrates of polymorphic molecular determinants of drug disposition. The outputs of PBPK modeling and simulation can provide valuable prior information to the exploration of covariate effects in population PK analyses, thereby enhancing knowledge management in population pharmacology analyses of heterogeneous global populations in clinical drug development.