Model-based meta-analysis of ethnic differences and their variabilities in clearance of oral drugs classified by clearance mechanism

Abstract

In this study, the ethnic ratios (ERs) of oral clearance between Japanese and Western populations were subjected to model-based meta-analysis (MBMA) for 81 drugs evaluated in 673 clinical studies. The drugs were classified into eight groups according to the clearance mechanism, and the ER for each group was inferred together with interindividual variability (IIV), interstudy variability (ISV), and inter-drug variability within a group (IDV) using the Markov chain Monte Carlo (MCMC) method. The ER, IIV, ISV, and IDV were dependent on the clearance mechanism, and, except for particular groups such as drugs metabolized by polymorphic enzymes or their clearance mechanism is not confirmative, the ethnic difference was found to be generally small. The IIV was well-matched across ethnicities, and the ISV was approximately half of the IIV as the coefficient of variation. To adequately assess ethnic differences in oral clearance without false detections, phase I studies should be designed with full consideration of the mechanism of clearance. This study suggests that the methodology of classifying drugs based on the mechanism that causes ethnic differences and performing MBMA with statistical techniques such as MCMC analysis is helpful for a rational understanding of ethnic differences and for strategic drug development.

FIGURE 1

Multi‐layered error model in this study. The parameters ${\mathrm{\sigma }}_{\mathrm{IIV},\mathrm{g},\mathrm{d},\mathrm{s}}^2$, ${\mathrm{\sigma }}_{\mathrm{VM},\mathrm{g},\mathrm{d},\mathrm{s}}^2$, ${\mathrm{n}}_{\mathrm{g},\mathrm{d},\mathrm{s}}$, ${\mathrm{\omega }}_{\mathrm{ISV},\mathrm{g}}^2$, ${\mathrm{\omega }}_{\mathrm{IDV},\mathrm{g}}^2$, ${\mathrm{ER}}_{\mathrm{g}}$, and ${\mathrm{CL}}_{\mathrm{W},\mathrm{g},\mathrm{d}}$ represent the unbiased estimate of inter‐individual variance, variance of mean oral clearance (CL_oral), number of subjects, variance for interstudy variability, variance for inter‐drug variability, ethnic ratio of CL_oral, and estimated CL_oral of Western populations, respectively. The subscripts g, d, and s represent groups g, d, and s, respectively. The parameters attached to the hat represent the parameters inferred by the Markov chain Monte Carlo method.

FIGURE 2

Interindividual variability (IIV), interstudy variability (ISV), and inter‐drug variability (IDV) of oral clearance in Japanese and Western populations for 81 drugs. (A) Renal clearance (CL) drugs; (B) mixed hepatic and renal CL drugs; (C) CYP3A substrates; (D) CYP2D6 substrates; (E) CYP2C19 substrates; (F) other hepatic metabolism drugs; (G) OATP1B substrates; and (H) extremely low absorption drugs. We subtracted 1.0 from exponential of the standard deviation of logarithmic CL_oral and shown as a percentage. (a) IDV and (b) ISV were inferred by the Markov chain Monte Carlo (MCMC) method. The boxes and bars represent the interquartile and 2.5–97.5 percentile ranges of MCMC samples, respectively. IDV for group (H) was fixed to zero because only two drugs were included in this group. NE, not estimated. (c) IIV was calculated from the standard deviation or analogous information described in the data source. The 95% confidence interval for the variance of IIV was calculated as {(n − 1)*s ²/χ ² (0.025), (n − 1)*s ²/χ ² (0.975)}, where n refers to the total number of subjects included in studies, s ² to the study‐level weighted average of the unbiased variance of lnCL, and χ ² (0.025) and χ² (0.975) to 2.5 and 97.5 percentiles of a chi‐square distribution with one degree of freedom, respectively. CV, coefficient of variation.

FIGURE 3

Inferred ethnic ratio of oral clearance (CL) of Japanese to Western populations for 81 drugs with the Markov chain Monte Carlo (MCMC) method. The drugs are arranged in descending order of ethnic ratio (ER) in each group. The boxes and bars represent the interquartile and 0.5–99.5 percentile (i.e., 99% credible interval) ranges of MCMC samples, respectively. The interval of 99% was adopted as a significant criterion based on a simulation study. The bar at the top of each group indicates the ER of the group representative. Boxes marked with an asterisk indicate drugs for which both Japanese and Western subjects are included in the same study. *Gefitinib is classified in the CYP3A substrate group in this study but also metabolized by CYP2D6 with almost equally contribution.

FIGURE 4

Simulation analysis for validation of the Markov chain Monte Carlo (MCMC) method. (a) Comparison of true and estimated ethnic ratios for every 25 virtual drugs in the 100 datasets. (b) Comparison of true and estimated ISV in Western (red diamond) and Japanese (blue circle) studies. (c) Dependency of the estimation of ISV on IIV in the Western study and (d) that in the Japanese study. Note that the numbers of virtual studies in a dataset were four and two for Western and Japanese studies, respectively; thus, the variability was larger in the Japanese study. CV, coefficient of variation; ER, ethnic ratio; IIV, interindividual variability; ISV, interstudy variability.