- Имитационная оценка соответствия режима дозировки многочисленным миеломным лекарствам, отражающим демографические и этнические различия.

Introduction

Several countries, including South Korea, have introduced bridge trial exemption policies to accelerate access to treatments for rare diseases [ 1 ]. Rare diseases such as multiple myeloma encounter inherent challenges in conducting large‐scale clinical trials due to the limited and heterogeneous patient populations involved in recruitment. Consequently, regulatory approvals are often based on the clinical trial data generated primarily from populations outside the local region. However, applying dosing regimens derived from multinational trials to populations that were underrepresented in those studies remains a persistent challenge. Differences in demographic characteristics and regional factors can influence pharmacokinetics and pharmacologic response, raising concerns about whether dosing strategies established in one population are fully appropriate for another [ 2 , 3 ].

Population‐specific factors such as body weight, body surface area, renal function, and body composition have been shown to influence drug exposure and response. These differences may be particularly relevant when extrapolating clinical trial results from Western populations to Asian populations, where demographic characteristics can differ substantially. This issue is particularly relevant in Korea. Clinical development programs conducted by multinational pharmaceutical companies often include only a small number of Korean participants, raising uncertainty about whether dosing strategies established in global trials adequately reflect the characteristics of Korean patients. Despite increasing clinical experience with multiple myeloma therapies in Korea, there remains a limited quantitative evaluation of whether currently approved dosing regimens provide comparable exposure in Korean patients to that observed in the populations studied in registration trials. Addressing this knowledge gap requires approaches that can quantitatively evaluate potential exposure differences using available data sources. Model‐informed approaches, including population PopPK modeling and simulation, provide a framework for exploring how demographic differences may influence drug exposure across populations [ 4 ].

In this study, PopPK‐based simulations were combined with real‐world demographic data derived from the Korean National Health Insurance Service (KNHIS) database and disease‐specific patient registries to evaluate the potential impact of population characteristics on drug exposure in Korean patients with multiple myeloma. We hypothesized that demographic differences between Korean patients and populations represented in global clinical trials could lead to measurable differences in predicted drug exposure.

The objective of this study was therefore to perform a quantitative post‐marketing evaluation of currently approved dosing regimens for multiple myeloma therapies in Korean patients using model‐informed simulations and real‐world population data. The results are intended to provide evidence that may help identify drugs or indications for which further pharmacokinetic evaluation or focused clinical studies in Korean or East Asian populations may be warranted.

Methods

Study Design and Ethical Considerations

The present study used a simulation‐based analysis using published PopPK models to evaluate the appropriateness of standard dose regimens approved in Korea. The analytical workflow consisted of four sequential steps (file S1 ). First, a literature search was conducted to identify major multiple myeloma therapies with published PopPK models incorporating demographic covariates. Second, demographic data for KP were obtained from KNHIS database and the Korean MM Registry, whereas demographic characteristics of the TP were extracted from previously published clinical trial. Third, pharmacokinetic simulations were performed using the published PopPK models under identical dosing regimens to compare predicted exposure between KP and TP. Fourth, differences in simulated exposure were evaluated and interpreted in the context of previously reported exposure–response relationships. For drugs in which simulated exposure in KP was ≥ 10% higher than in TP, potential safety implications were explored, whereas potential efficacy implications were considered when exposure in KP was ≥ 10% lower than in TP. A threshold of 10% difference in exposure was used as a conservative exploratory criterion to identify potential population‐related differences, given that regulatory bioequivalence assessments typically allow exposure differences of approximately ±20% without a clinically meaningful impact. This study was approved by the Institutional Review Board of The Catholic University of Korea Seoul St. Mary's Hospital (IRB No. KC23ZISI0463). All data were anonymized and derived from publicly available databases and published studies; therefore, informed consent was waived.

Literature Search and Selection of Drugs With PK Models

Drugs included in the analysis were selected based on their relevance to current clinical practice and the availability of suitable pharmacokinetic models. Candidate therapies were first identified according to the drug classes recommended in the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Multiple Myeloma (Version 1.2025) [ 5 ].

PopPK models were then selected according to the following criteria:

(1) Availability of previously published per‐reviewed PopPK models developed using clinical trial data, and (2) inclusion of demographic covariates that could be identified within Korean real‐world clinical datasets. When multiple PopPK models were available for a given drug, the model incorporating covariates that could be matched to variables available in Korean clinical data sources was selected for simulation. This approach allowed direct incorporation of observed demographic and clinical characteristics of the Korean population into the simulations without introducing additional assumptions for unavailable covariates. The selected models were previously evaluated and qualified in the original publications using standard population pharmacokinetic model evaluation procedures, including goodness‐of‐fit diagnostics and visual predictive checks, supporting their suitability for simulation‐based analyses. Detailed literature search strategies, including databases, search terms, and screening procedures, are provided in the Figure S1 .

Demographic Data Collection and Confirmation

The demographic and clinical data for KP were systematically collected from two primary data sources. First, comprehensive demographic information was obtained from the KNHIS, a database encompassing approximately 97% of the Korean population [ 6 ]. The KNHIS database offers a comprehensive array of demographic data, health examination records, and prescription information. Secondly, the present study collected disease‐specific clinical data from the Korean MM Registry, which includes clinical data of 1801 myeloma patients enrolled at Seoul St. Mary's Hospital between 2011 and 2017. The collected clinical variables included body weight, height, gender, MM immunoglobulin type (IgG or non‐IgG), serum creatinine levels, and drug usage history.

Demographic and laboratory characteristics for the TP were extracted from previously published clinical trial reports. To construct a representative TP dataset for simulation, demographic variables were randomly sampled according to the reported distributions of covariates in each clinical trial. A series of statistical analyses were performed to compare the demographic and clinical variables between the two groups. The distribution of covariates between the two groups was assessed using Student's t ‐test, with a significance level of 0.05 set a priori. In addition, differences in key covariates were interpreted in relation to their impact on pharmacokinetic parameters (e.g., clearance and volume) based on the covariate–parameter relationships defined in the selected PopPK models. For the melphalan‐specific analysis, demographic data from the melphalan‐treated subgroup within the Korean MM Registry were used because the demographic characteristics of this subgroup differed from those of the overall registry population.

To account for correlations among demographic and clinical covariates, simulation datasets were generated using covariance matrices implemented with the MASS package in R (version 3.4.1; The R Foundation for Statistical Computing, Vienna, Austria). Sampling was performed within the observed clinical ranges using truncated distributions.

Simulation and Calculation of Drug Exposure Indicators

Each published PopPK model was implemented for simulation in NONMEM (version 7.5.1) and R (mrgsolve package), based on the structural model, parameter estimates, and covariate relationships reported in the original NONMEM model publications. NONMEM was used to reproduce each original model exactly as reported in the source publication, whereas mrgsolve was used to perform the large‐scale virtual‐population simulations for KP and TP; simulation outputs from the two platforms were cross‐checked to confirm consistent implementation before application to the study datasets. The implementation preserved the original model structure and parameterization as described in the source publications. Inter‐individual variability was incorporated using the reported between‐subject variability terms from the original models. Where necessary for simulation, model components were translated into the mrgsolve framework while maintaining the same structural equations and covariate relationships. Prior to application to the Korean population dataset, the model was implemented and verified by reproducing key exposure metrics reported in the original studies or, when unavailable, in relevant Phase 1 clinical studies. Agreement between simulated and reported PK metrics was used to confirm the reproducibility and adequacy of the implemented models as described in file S2 . All model parameter values and the NONMEM control stream codes used for the simulations are provided in Figure S1 to ensure transparency and reproducibility of the model implementation.

Virtual populations were generated separately for KP and TP using multivariate normal sampling with the mvrnorm function from the MASS package in R (version 3.4.1). For KP, the means, standard deviations, and covariance structure of the demographic and clinical covariates were computed by pooling the individual‐level data obtained from the KNHIS database and the Korean MM Registry. Because the individual‐level pool available for several covariates was limited in size, multivariate normal sampling was used in preference to direct resampling of observed records, to generate a virtual population that more broadly represented the plausible covariate space of Korean MM patients. For TP, covariate means, standard deviations, and correlation structures were reconstructed from the published source clinical trials comprising 729 participants for carfilzomib, 125 participants for lenalidomide, 100 participants for melphalan, and 239 participants for daratumumab. For both KP and TP, 1000 individual covariate vectors were generated per drug, and sampled covariates were constrained to the observed minimum and maximum values of each source dataset using truncated distributions to avoid physiologically implausible values. Correlations among covariates (e.g., body weight, body surface area, fat‐free mass, and creatinine clearance) were preserved through the covariance matrix applied during multivariate sampling.

The simulations were executed using dosing regimens that were identical to those reported in the respective clinical studies. These simulations were performed using the “mrgsolve” package in R (version 3.4.1). PK exposure indicators—including the area under the concentration‐time curve ( AUC ), peak concentration ( C max ), and time to peak concentration ( T max )—were calculated from the simulated time‐concentration profiles using the “NonCompart” package in R. The exposure indicators from KP and TP were summarized using descriptive statistics (mean, median, interquartile range) and subsequently compared to quantitatively evaluate PK differences attributable to demographic variations.

Assessment of Safety and Efficacy

To assess the clinical impact of differences in drug exposure between KP and TP, additional analyses were conducted for drugs exhibiting a mean exposure difference exceeding 10%. A threshold of ≥ 10% difference in predicted exposure was used as a pragmatic screening criterion to identify drugs that may warrant further pharmacokinetic or clinical evaluation. When an exposure–response (E–R) relationship was available in the same publication as the selected PopPK model, graphical data were extracted from the relevant figures using the WebPlotDigitizer program. The comparison between KP and TP was based on the mean exposure values. If the selected PopPK model publication did not include E–R data, graphical information was similarly extracted from a separate published study on the same drug using the WebPlotDigitizer program. For carfilzomib, where the simulated exposure in KP was lower by 10% or more compared to TP, the E‐R relationship result was adopted from the same publication that the selected carfilzomib PopPK model [ 7 ]. The efficacy was evaluated by examining the relationship between drug exposure and overall response rate (ORR), as defined by the International Myeloma Working Group Uniform Response Criteria (IMWG‐URC) [ 8 ]. For lenalidomide, where the simulated exposure in KP was higher by 10% or more than in TP, a safety evaluation was performed to assess potential risks associated with increased drug levels. Because the selected PopPK model publication for lenalidomide did not provide E–R analysis, the E–R relationship was instead obtained from a separate clinical study conducted in Japanese patients with newly diagnosed multiple myeloma [ 9 ]. Although this exposure–response data was derived from a study conducted in Japanese patients, this was used for contextual interpretation under the assumption that broadly comparable exposure–response relationships exist across East Asian populations. The correlation between drug exposure and hematologic adverse events (AEs) was evaluated, with AEs graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI‐CTCAE, version 6.0) [ 10 ].

Results

Selected Drugs, PK Models, and Covariates

Five drugs were included in the present analysis: carfilzomib, daratumumab, lenalidomide, melphalan, and panobinostat. Previously published PopPK models were identified for each drug and implemented for the simulations [ 7 , 11 , 12 , 13 , 14 ]. The models consisted of one‐, two‐, or three‐compartment structures and incorporated demographic covariates known to influence pharmacokinetic parameters such as clearance and volume of distribution. Key covariates included body size metrics (e.g., body surface area, body weight, or fat‐free mass), renal function indicators (e.g., creatinine clearance), and selected clinical variables (e.g., albumin level, immunoglobulin type, hematocrit, sex, and age). Detailed information on the PopPK model structures, covariates, and their relationships with pharmacokinetic parameters is summarized in Table 1 .

Demographic and Clinical Differences Between Populations

A comprehensive analysis revealed substantial disparities in both demographics and clinical characteristics between KP and TP across all examined drugs (Table 2 ). For carfilzomib, in trial PX‐171‐009, which predominantly involved Caucasian participants (71%), KP exhibited a markedly lower mean body surface area (BSA) (17.1% difference). This difference corresponds to approximately a 14% lower predicted clearance in KP than in TP. In the daratumumab trial ( NCT03180736 ), with a predominantly Caucasian participant population (89%), KP demonstrated a 20.6% reduction in body weight and an 8.5% decrease in albumin levels. These covariate differences correspond to an approximately 21% lower predicted clearance and central volume ( V1 ) in KP. Additionally, the proportion of patients with IgG‐type multiple myeloma was 26.8% higher in KP. For lenalidomide, KP exhibited a 23.2% lower FFM and a 22.2% lower CrCl than TP. These differences correspond to an approximately 38% lower predicted clearance and an approximately 23% lower volume of distribution in KP. Melphalan (ACTRN0126000231549), similarly lacking ethnicity details, exhibited substantial covariate disparities. Specifically, FFM was 25.3% lower and CrCl 45.4% lower in KP than in TP. These differences correspond to an approximately 36% lower predicted clearance and an approximately 25% lower central volume ( V1 ) in KP. Panobinostat ( NCT02654990 ), which primarily enrolled Caucasian participants (85%), showed that KP had a lower BSA (by 12.6%) and a higher age than TP. These differences correspond to an approximately 17% lower predicted clearance, with a similar reduction observed in V2 .

The demographic and clinical characteristics of TP and KP are summarized in Table 2 and illustrated in Figure 1 . Additional baseline characteristics available from the source studies and Korean datasets are summarized in the Table S1 for each drug. These comparisons highlight population‐level variability that may influence pharmacokinetic exposure and clinical outcomes.

Simulated Pharmacokinetic Exposure Differences

The simulated PK exposure metrics ( AUC, C max , T max ) were compared between KP and TP, and the key findings are summarized in Table 3 . Variability in exposure was evaluated using prediction intervals and distributional analyses (Figure 2 and Supplementary file 1 ).

Drugs with Decreased Exposure in KP: Carfilzomib exhibited lower simulated exposure in KP across two dosing regimens. For the 20/27 mg/m 2 dosing regimen, KP demonstrated a mean AUC cycle1 of 817 ng·h/mL compared with 926 ng·h/mL in TP, corresponding to a 12% lower exposure. Under the 20/56 mg/m 2 regimen, the mean AUC cycle1 in KP was 1983 ng·h/mL compared with 2301 ng·h/mL in TP, corresponding to a 15% difference (Figure 2A.1 ). These differences were consistent with a leftward shift of the exposure distribution in KP relative to TP. However, substantial overlap between the two populations was observed (Figure 2A.2 ).

Drugs with Increased Exposure in KP: Lenalidomide demonstrated higher simulated exposure in KP, with a mean AUC 0–24 of 3644 ng·h/mL compared with 2745 ng·h/mL in TP, representing a 24.7% difference. The peak concentration (C max,0–24 ) was also higher in KP by 21.8% (Figure 2B.1 ). Distributional analyses showed a rightward shift in KP, while still demonstrating partial overlap with TP (Figure 2B.2 ).

Drugs with Minimal Exposure Difference: Daratumumab, melphalan, and panobinostat demonstrated relatively small differences in simulated exposure between KP and TP. Daratumumab showed a mean AUC C3D1 difference of 0.8% (KP: 11,686 μg·h/mL vs. TP: 11,782 μg·h/mL, Figure 2E ). Melphalan showed a mean AUC 0–4 difference of 7.0% (KP: 7.18 mg·h/L vs. TP: 6.68 mg·h/L, Figure 2C ), whereas panobinostat showed a 7.7% lower exposure in KP (96 ng·h/mL) compared with TP (104 ng·h/mL, Figure 2D ).

Exposure‐Response Outcomes for Efficacy and Safety

Exposure‐response analyses were conducted specifically for carfilzomib and lenalidomide, as these drugs demonstrated clinically significant PK exposure differences (≥ 10%) between KP and TP.

Differences in exposure to carfilzomib resulted in discrepancies in the predicted overall ORR. For the 20/27 mg/m 2 dosing regimen, the predicted ORR was 0.44 for TP and 0.42 for KP, indicating approximately a 4% lower probability of response in KP. In the 20/56 mg/m 2 regimen, the ORR was 0.69 for TP and 0.63 for KP, reflecting a 9% reduction in predicted efficacy in the KP group. These findings suggest that reduced carfilzomib exposure in KP may be associated with a 4%–9% lower probability of achieving overall response (Figure 3 ).

A safety assessment of lenalidomide is presented. Lenalidomide demonstrated increased exposure in KP, correlating with a higher risk of hematologic AEs. At KP's mean AUC 0‐24 (3644 ng·h/mL), the estimated probability of experiencing grade ≥ 3 hematologic AEs was 63%, compared to 50% for TP (2745 ng·h/mL), representing a 1.27‐fold increased risk of significant AEs (Figure 4 ). These findings underscore the importance of vigilant monitoring and evaluation of lenalidomide safety in the KP population due to elevated exposure‐related toxicity risks.

Discussion

To our knowledge, this is the first study to evaluate potential differences in drug exposure indicators between multinational clinical trial populations and the Korean patients using simulation based on the published PopPK models and Korean RWD. Our analysis indicated that exposure to carfilzomib was modestly lower in the Korean patients, whereas exposure to lenalidomide was relatively higher compared with the multinational clinical trial populations. Overall, these findings suggest that population‐specific demographic and clinical characteristics may influence pharmacokinetic variability and could potentially affect treatment response and safety outcomes. This highlights the value of evaluating population differences using simulation approaches that incorporate real‐world population characteristics.

Key determinants such as body size metrics (e.g., BSA or FFM), renal function, and laboratory parameters differed between the Korean population and clinical trial participants, which may contribute to variability in exposure metrics. Previous studies have emphasized that dosing strategies should consider demographic and clinical characteristics in addition to ethnicity when translating clinical trial results to real‐world populations. The FDA has also noted that ethnicity encompasses cultural and lifestyle‐related factors that may influence physiological characteristics such as body composition [ 18 ]. Together, these findings emphasize the importance of considering population‐specific characteristics when evaluating pharmacokinetic variability and treatment outcomes.

For carfilzomib, exposure differences between the KP and TP were observed. These differences may be partly explained by variations in BSA distributions between populations, as it was included as a covariate influencing drug clearance in the PopPK model. Although multiple factors influence treatment response, this result suggests that population‐related PK variability may represent one contributing factor to observed differences in clinical outcomes.

For lenalidomide, the simulations suggested relatively higher systemic exposure in the KP. Renal function and body composition were key determinants of lenalidomide clearance in the PopPK model. Previous studies have shown that increased lenalidomide exposure may be associated with a higher risk of hematologic adverse events rather than improved efficacy [ 9 ]. Our findings are consistent with these observations and highlight the importance of evaluating population‐specific safety considerations when interpreting lenalidomide exposure in Korean patients. In contrast, no meaningful exposure differences were observed for melphalan, daratumumab, or panobinostat between the Korean and trial populations. Previous analyses have similarly reported minimal effects of population difference on exposure or response for daratumumab and panobinostat [ 13 , 14 ].

Lastly, our findings highlight the importance of evaluating population variability beyond broad racial or ethnic classifications. Although ethnicity has been evaluated as a covariate in several pharmacokinetic analyses, it is often not retained as a significant predictor of PK variability. Instead, differences in demographic characteristics, clinical factors, and healthcare environments across regions may contribute to variability in drug exposure and treatment outcomes. Simulation approaches that integrate real‐world population data may therefore provide a useful framework for evaluating these differences and supporting evidence‐based therapeutic decisions.

This study has several limitations. First, a single published PopPK model was selected for each drug when multiple models were available. Although these models were developed from clinical trial datasets and reported in per‐reviewed publications, alternative model structures or covariate relationships described in other studies were not evaluated. They could potentially lead to different exposure predictions. In addition, simulations were performed using the reported typical parameter estimates together with between‐subject variability from the published models. Parameter uncertainty associated with fixed‐effect estimates was not propagated because the parameter variance–covariance matrices required for uncertainty propagation were not available for most of the selected published models.

Second, the exposure–response relationship applied for lenalidomide was derived from a previously published study conducted in Japanese patients with newly diagnosed multiple myeloma. Although East Asian populations share certain demographic similarities, applying exposure–response relationships across populations and disease stages may introduce uncertainty, and the extrapolation should be interpreted cautiously.

Third, this study primarily evaluated pharmacokinetic differences between populations. Although comparable drug exposure was observed, this does not necessarily ensure comparable clinical outcomes, as therapeutic response is determined by multiple factors beyond systemic exposure. These include pharmacodynamic variability as well as differences in treatment patterns and supportive care in real‐world clinical settings. Therefore, while the present findings support pharmacokinetic similarity, further studies incorporating clinical outcome data would be valuable to determine whether these similarities translate into comparable therapeutic effects.

Finally, genetic factors that may influence drug metabolism or drug response were not explicitly incorporated into the simulation models. Future studies integrating pharmacogenomic data and prospective clinical validation may further improve the interpretation of population‐specific drug exposure differences.

These findings are consistent with current regulatory perspectives emphasizing the evaluation of population differences in drug response. International regulatory guidance, such as the ICH E5 guideline, highlights the importance of assessing ethnic factors when extrapolating clinical data across regions [ 19 ]. In addition, recent FDA guidance encourages systematic collection and analysis of race and ethnicity data in clinical studies [ 20 ]. Within this context, model‐informed approaches integrating clinical trial data and real‐world evidence may provide a practical framework for evaluating population‐specific pharmacokinetic differences and supporting regulatory decision‐making [ 21 ].

Our findings highlight the importance of evaluating pharmacokinetic variability when applying clinical trial evidence to real‐world populations. Simulation approaches integrating real‐world data from target populations may provide a useful framework for identifying drugs or indications that warrant further evaluation in bridging studies. In addition, such analyses may contribute to the assessment of potential ethnic sensitivity and support the design of focused pharmacokinetic or exposure–response studies in Korean or broader East Asian populations. Furthermore, the proposed framework may enable exploratory evaluation of alternative dosing strategies in Korean patients by simulating dose–exposure relationships, providing a basis for future studies.

Author Contributions

All authors wrote the manuscript. S.C. and S.H. designed the research. M.K. performed the research. M.K. analyzed the data.

Funding

No funding was received for this work.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Flow Diagram of Simulation and Analysis.

Supplementary Figure 2. Simulated AUC distributions in the multinational clinical trial population (TP) and Korean patients (KP).

Table S1: Comparison of simulated pharmacokinetic exposure metrics with reported drugs' values.

File S3: NONMEM code file for population pharmacokinetic model.

File S4: Supplementary Methods.

File S5: Supporting Information.

Acknowledgments

ChatGPT (OpenAI) was used solely for language editing and improving readability. No AI tools were used for data analysis or interpretation. The authors are fully responsible for all scientific content.

References

1 M.o.F.a.D. Safety , Regulation on Designation of Orphan Drugs ( Ministry of Food and Drug Safety (MFDS) , 2018 ).

2 A. Ramamoorthy , H. H. Kim , E. Shah‐Williams , and L. Zhang , “ Racial and Ethnic Differences in Drug Disposition and Response: Review of New Molecular Entities Approved Between 2014 and 2019 ,” Journal of Clinical Pharmacology 62 , no. 4 ( 2022 ): 486 – 493 . 34608640 10.1002/jcph.1978

3 A. S. Gross , A. C. Harry , C. S. Clifton , and O. Della Pasqua , “ Clinical Trial Diversity: An Opportunity for Improved Insight Into the Determinants of Variability in Drug Response ,” British Journal of Clinical Pharmacology 88 , no. 6 ( 2022 ): 2700 – 2717 . 35088432 10.1111/bcp.15242 PMC9306578

4 R. Madabushi , P. Seo , L. Zhao , M. Tegenge , and H. Zhu , “ Review: Role of Model‐Informed Drug Development Approaches in the Lifecycle of Drug Development and Regulatory Decision‐Making ,” Pharmaceutical Research 39 , no. 8 ( 2022 ): 1669 – 1680 . 35552984 10.1007/s11095-022-03288-w PMC9097888

5 S. K. Kumar , N. S. Callander , K. Adekola , et al., “ NCCN Guidelines(R) Insights: Multiple Myeloma, Version 1.2025 ,” Journal of the National Comprehensive Cancer Network 23 , no. 5 ( 2025 ): 132 – 140 . 40340857 10.6004/jnccn.2025.0023

6 S. O. Song , C. H. Jung , Y. D. Song , et al., “ Background and Data Configuration Process of a Nationwide Population‐Based Study Using the Korean National Health Insurance System ,” Diabetes and Metabolism Journal 38 , no. 5 ( 2014 ): 395 – 403 . 25349827 10.4093/dmj.2014.38.5.395 PMC4209354

7 Y. Ou , S. Doshi , A. Nguyen , et al., “ Population Pharmacokinetics and Exposure‐Response Relationship of Carfilzomib in Patients With Multiple Myeloma ,” Journal of Clinical Pharmacology 57 , no. 5 ( 2017 ): 663 – 677 . 27925676 10.1002/jcph.850

8 B. G. Durie , B. G. M. Durie , J. L. Harousseau , et al., “ International uniform response criteria for multiple myeloma ,” Leukemia 20 , no. 9 ( 2006 ): 1467 – 1473 . 16855634 10.1038/sj.leu.2404284

9 T. Kobayashi , M. Miura , T. Niioka , et al., “ Phase II Clinical Trial of Lenalidomide and Dexamethasone Therapy in Japanese Elderly Patients With Newly Diagnosed Multiple Myeloma to Determine Optimal Plasma Concentration of Lenalidomide ,” Therapeutic Drug Monitoring 40 , no. 3 ( 2018 ): 301 – 309 . 29494421 10.1097/FTD.0000000000000499

10 Common Terminology Criteria for Adverse Events (CTCAE) ( U.S.D.o.H.a.H. Service , 2025 ).

11 C. E. Nath , P. J. Shaw , J. Trotman , et al., “ Population Pharmacokinetics of Melphalan in Patients With Multiple Myeloma Undergoing High Dose Therapy ,” British Journal of Clinical Pharmacology 69 , no. 5 ( 2010 ): 484 – 497 . 20573084 10.1111/j.1365-2125.2010.03638.x PMC2856049

12 J. H. Hughes , M. A. Phelps , R. N. Upton , et al., “ Population Pharmacokinetics of Lenalidomide in Patients With B‐Cell Malignancies ,” British Journal of Clinical Pharmacology 85 , no. 5 ( 2019 ): 924 – 934 . 30672004 10.1111/bcp.13873 PMC6475687

13 A. G. Dosne , X. Li , M. M. Luo , et al., “ Population Pharmacokinetics and Exposure‐Response Analyses of Daratumumab Plus Pomalidomide/Dexamethasone in Relapsed or Refractory Multiple Myeloma ,” British Journal of Clinical Pharmacology 89 , no. 5 ( 2023 ): 1640 – 1655 . 36484341 10.1111/bcp.15628

14 M. Savelieva , M. M. Woo , H. Schran , S. Mu , J. Nedelman , and R. Capdeville , “ Population Pharmacokinetics of Intravenous and Oral Panobinostat in Patients With Hematologic and Solid Tumors ,” European Journal of Clinical Pharmacology 71 , no. 6 ( 2015 ): 663 – 672 . 25939707 10.1007/s00228-015-1846-7 PMC4430599

15 R. D. Mosteller , “ Simplified Calculation of Body‐Surface Area ,” New England Journal of Medicine 317 , no. 17 ( 1987 ): 1098 . 3657876 10.1056/NEJM198710223171717

16 D. W. Cockcroft and M. H. Gault , “ Prediction of Creatinine Clearance From Serum Creatinine ,” Nephron 16 , no. 1 ( 1976 ): 31 – 41 . 1244564 10.1159/000180580

17 S. Janmahasatian , S. B. Duffull , S. Ash , L. C. Ward , N. M. Byrne , and B. Green , “ Quantification of Lean Bodyweight ,” Clinical Pharmacokinetics 44 , no. 10 ( 2005 ): 1051 – 1065 . 16176118 10.2165/00003088-200544100-00004

18 Collection of Race and Ethnicity Data in Clinical Trials and Clinical Studies for FDA ( F.a.D. Administration , 2024 ).

19 “ICH E5 Ethnic Factors in the Acceptability of Foreign Clinical Data” . 10180134

20 “Collection of Race and Ethnicity Data in Clinical Trials and Clinical Studies for FDA‐Regulated Medical Products,” 2024 Fishers Lane, Rm 1061 Rockville, MD 20852.

21 Y. Wang , H. Zhu , R. Madabushi , Q. Liu , S. M. Huang , and I. Zineh , “ Model‐Informed Drug Development: Current US Regulatory Practice and Future Considerations ,” Clinical Pharmacology and Therapeutics 105 , no. 4 ( 2019 ): 899 – 911 . 30653670 10.1002/cpt.1363