Human Pharmacokinetic Prediction of Antibody‐Drug Conjugates Using Human FcRn Transgenic Mice Authors: Keitaro Nakagawa, Kenta Haraya Published: 2026 ABSTRACT ABSTRACTAntibody‐drug conjugates (ADCs) have progressed significantly in recent years, particularly in the oncology field. However, predicting human pharmacokinetics (PK) of ADCs in the early stages of drug development remains challenging. Human neonatal Fc receptor (FcRn) transgenic mice have been established as a reliable pre‐clinical PK model for predicting human PK profiles of antibody‐based therapeutics. In this study, we investigated the applicability of human FcRn transgenic mice for predicting the human PK of ADCs in a pre‐clinical setting. Eight FDA‐approved ADCs were administered to human FcRn transgenic mice, and PK parameters were calculated using two‐compartment model analysis. The clearance (CL) values of total ADCs in human FcRn transgenic mice showed a strong correlation with human CL (R 2 = 0.888). Single‐species allometric scaling using data from human FcRn transgenic mice indicated an optimal exponent of 1.0 for predicting human CL, with 87.5% of ADCs falling within twofold of the human CL. These findings indicate that allometric scaling from human FcRn transgenic mice serves as a promising translational tool for predicting human CL of ADCs from the early development stage. Introduction In recent years, the development of ADCs in oncology has progressed rapidly, and as of 2025, 15 ADCs have been approved by the U.S. Food and Drug Administration (FDA) [1, 2]. These approved ADCs cover a wide range of indications, including hematologic malignancies and solid tumors such as lung, gastric, head and neck, and breast cancer. Antibody‐drug conjugate represents a next‐generation therapeutic strategy designed to selectively deliver cytotoxic agents to tumor cells. ADCs are designed to selectively target tumor‐associated surface antigens and deliver cytotoxic payloads to malignant cells. Upon antigen engagement, ADCs undergo internalization and are trafficked to lysosomes, where proteolytic degradation facilitates payload release and ultimately, induces cell death [2]. This targeted delivery mechanism enhances anti‐tumor efficacy while minimizing off‐target toxicity, thereby expanding the therapeutic window [2]. Structurally, ADCs comprise three key components: a monoclonal antibody, a chemical linker, and a cytotoxic payload [3]. Drug conjugation generally changes physicochemical properties such as hydrophobicity and isoelectric point, which enhances non‐specific binding and induces higher clearance in vivo [4, 5]. Moreover, ADCs exhibit clearance pathways that differ from those of monoclonal antibodies, including deconjugation of the linker and payload [4]. Cleavable linkers are engineered to respond to tumor‐specific conditions including acidic pH, enzymatic activity, or reducing environments, enabling selective payload release [6]. Except for trastuzumab emtansine and belantamab mafodotin, all approved ADCs incorporate cleavable linkers as a structural component [2]. Trastuzumab emtansine and belantamab mafodotin have non‐cleavable linkers and require intracellular degradation of the antibody to release the payload [4, 7]. In both types of linkers, undesirable cleavage in systemic circulation accelerates ADC clearance via deconjugation and compromises tumor exposure, implying that the linker stability in blood directly affects drug efficacy [8]. The Fc region, composed of CH2 and CH3 domains, mediates binding to Fc receptors, such as Fcγ receptor and FcRn. FcRn is a heterodimeric protein composed of a class I major histocompatibility complex (MHC‐I)‐like α‐chain (α‐FcRn) and β2‐microglobulin (β2m). The expression profile of FcRn is conserved across species and tissues, with predominant expression reported in endothelial cells, stromal macrophages, Kupffer cells, alveolar macrophages, intestinal epithelial cells, and choroid plexus epithelium [9]. Additionally, FcRn is a membrane protein localized to specific intracellular components, particularly endosomes, and exhibits pH‐dependent binding to the Fc region of IgG [9]. FcRn binds IgG with high affinity under acidic conditions, whereas its affinity markedly decreases at neutral pH. Consequently, antibody‐based therapeutics internalized by vascular endothelial cells bind FcRn within endosomes and are subsequently released into the circulation via exocytosis as the pH returns to neutral. This recycling mechanism enables prolonged systemic exposure to antibody‐based therapeutics. In FcRn knockout mice, the half‐life of antibodies is significantly reduced [10]. Conversely, engineered mutations that enhance FcRn binding at acidic pH have been shown to substantially extend antibody half‐life in both pre‐clinical models and humans, indicating that FcRn binding affinity significantly regulates antibody pharmacokinetics [11]. The binding affinity of human IgG for FcRn varies across species. In particular, KD for rodent FcRn is approximately twofold stronger than that for human FcRn [12]. Therefore, predicting human pharmacokinetics of antibody‐based therapeutics from early non‐clinical species such as mice and rat is limited by species differences in FcRn affinity. By contrast, cynomolgus monkey FcRn shares high amino acid sequence homology and comparable binding affinity to human FcRn [12]. Thus, cynomolgus monkeys have been often used for translational research [12]. Several studies have suggested that the clearance of monoclonal antibodies in humans can be predicted from non‐human primate data [13, 14, 15, 16]. Consistent with these findings, it has been reported that allometric scaling approaches are applicable to Fc‐fusion proteins as well as antibodies [17]. Human clearance (CL) has been predicted using cynomolgus monkey data not only for monoclonal antibodies, but also for ADCs. Mahmood showed the utility of an interspecies scaling method that predicts the human CL of ADCs using three animal species including cynomolgus monkey [18]. Li et al. reported that human CL of ADCs could be accurately predicted from cynomolgus monkey data using an allometric exponent of 1.0 [19]. However, ethical considerations, cost, and throughput limit the use of cynomolgus monkeys for human prediction. To overcome these problems, human FcRn transgenic mouse models, Tg276 and Tg32, which murine α‐FcRn is replaced with human α‐FcRn, have been developed [20]. These models have been shown to be valuable tools for predicting human PK profiles of biotherapeutics. Haraya et al. reported that the plasma half‐lives of marketed monoclonal antibodies and Fc‐fusion proteins in Tg276 mice correlated more closely with human data than those in wild‐type (WT) mice or cynomolgus monkeys [21]. Avery et al. demonstrated that human CL of monoclonal antibodies could be accurately predicted using Tg32 mice with an allometric exponent of 0.93 [22]. Subsequently, Betts et al. proposed scaling factors to predict human two compartment model parameters (i.e., CL, Vc; central compartment volume, Q; intercompartmental clearance, Vp; peripheral compartment volume) for monoclonal antibodies based on data from Tg32 mice [23]. Haraya et al. showed that Tg32 mice are also suitable for predicting PK parameters of monoclonal antibodies engineered for extended half‐life via Fc modifications [24]. These findings highlight the utility of human FcRn transgenic mice as a translational research tool. However, studies utilizing human FcRn transgenic mice for allometric scaling of ADCs remain limited. Considering the above‐described limitations of using cynomolgus monkeys, human FcRn transgenic mice are valuable for human prediction. In this study, we aimed to evaluate the utility of human FcRn transgenic mice for predicting the pharmacokinetics of ADCs in humans, and to assess their applicability in the pre‐clinical stage. Methods ADCs To investigate the prediction of human pharmacokinetic parameters using Tg32 mouse data, we selected eight marketed drug products with diverse linkers and payloads because clinical pharmacokinetic parameters after intravenous administration are available for these agents. Tisotumab vedotin (Tivdak) was purchased from Seagen. Loncastuximab tesirine (Zynlonta) was acquired from Swedish Orphan. Sacituzumab govitecan (Trodelvy) was sourced from Gilead. Inotuzumab ozogamicin (Besponsa) was purchased from Pfizer. Trastuzumab deruxtecan (Enhertu) and datopotamab deruxtecan (Datroway) were acquired from Daiichi Sankyo Co. Ltd. Polatuzumab vedotin (Polivy) and trastuzumab emtansine (Kadcyla) were prepared at Chugai Pharmaceutical Co. Ltd. Pharmacokinetic Study All animal experiments in this study were conducted under the AAALAC approved guidelines. Male 5‐week‐old human FcRn transgenic mice, strain number 32 (B6.Cg‐FcgrtTg(FCGRT)32Dcr/DcrJ, The Jackson Laboratory Japan, Kanagawa, Japan), and male 5 or 6‐week‐old C57BL/6J WT mice (The Jackson Laboratory Japan) were purchased and housed in specific pathogen‐free cages. Following an acclimation period, pharmacokinetic studies were initiated. Intravenous injection into the tail vein was conducted at a dose of 1 mg/10 mL/kg (n = 3 per group) except for inotuzumab ozogamicin. Inotuzumab ozogamicin was injected at 0.1 mg/10 mL/kg (n = 3 per group) in accordance with AAALAC approved guidelines. Blood samples were collected from the jugular vein at 5 min, 4 h, 7 h, 1 day, 3 days, 7 days, 14 days, 21 days, and 28 days after administration. Blood samples were treated with sodium heparin (Mochida Pharmaceutical Co. Ltd., Tokyo, Japan). Extracted plasma was stored at −80°C until the ligand‐binding assay. Electrochemiluminescence Immunoassay for Quantification of Total Antibody and ADC Ligand‐binding assays were performed using 384‐well uncoated SECTOR plates (MESO SCALE DIAGNOSTICS, Rockville, MD, USA). For the total antibody assay, goat anti‐human IgG antibodies (1 μg/mL; MERCK, Tokyo, Japan) were immobilized on the plate surface. For the total ADC assay, anti‐payload antibodies (1 μg/mL as a final concentration) specific to each ADC were immobilized. Anti‐DM1 (Thermo Fisher Scientific, Waltham, MA, USA) antibodies were used for trastuzumab emtansine; anti‐MMAE antibodies (ACROBiosystems, Newark, DE, USA) for tisotumab vedotin and polatuzumab vedotin; anti‐PBD antibodies (Creative Diagnostics, Shirley, NY, USA) for loncastuximab tesirine; anti‐SN‐38 antibodies (ACROBiosystems, Newark, DE, USA) for sacituzumab govitecan; anti‐Dxd antibodies (antibodies, Davis, CA, USA) for trastuzumab deruxtecan and datopotamab deruxtecan; and anti‐calicheamicin antibodies (Creative Diagnostics, Shirley, NY, USA) for inotuzumab ozogamicin. After blocking with bovine serum albumin (MERCK) and Block Ace (Sumitomo Pharma Co. Ltd., Osaka, Japan), collected plasma was added and incubated for 1 h at room temperature. Biotin‐labeled anti‐human IgG antibodies were added and incubated for 1 h at room temperature, followed by 1 h incubation with SULFO‐TAG labeled streptavidin (MESO SCALE DIAGNOSTICS) at room temperature. MSD Read Buffer T (MESO SCALE DIAGNOSTICS) was added, and electrochemiluminescence signals were measured using the MSD MESO SECTOR S 600 instrument (MESO SCALE DIAGNOSTICS). Can Get Signal Solution (TOYOBO, Osaka, Japan) was used for the detection of inotuzumab ozogamicin as a dilution buffer. Concentrations were calculated using SoftMax Pro. A four‐parameter logistic model was used for calibration curves, with an accuracy threshold of ±30% and a coefficient of variation threshold of ±20% for duplicate measurements. Pharmacokinetic Analysis Non‐compartmental analysis and two compartment analysis were performed using WinNonlin version 8.5.2.4 (Certara). The area under the curve of plasma concentration (AUC) was calculated as AUC infinity (AUCinf). Clinical Data Total ADC CL values in humans were collected from reports which performed population pharmacokinetic analyses (Table 1). For models that constructed linear CL and deconjugation CL separately, total ADC CL was calculated as the sum of linear CL and deconjugation CL values. TABLE 1PK parameters extracted from reports which perform population pharmacokinetics analysis.Drug nameCLVcQVpReferencesValueUnitValueUnitValueUnitValueUnitTisotumab vedotin1.42L/day3.1L4.01L/day4.47L[25]Loncastuximab tesirine0.2621L/day3.86L1.16L/day3.53L[26]Sacituzumab govitecan0.133L/h2.77L0.006L/h0.908L[27]Trastuzumab deruxtecan0.421L/day2.77L0.199L/day5.16L[28]Polatuzumab vedotin0.0344L/h3.15L0.0145L/h3.98L[29]Inotuzumab ozogamicin0.113L/h6.7L0.0405L/h5.1L[30]Trastuzumab emtansine0.676L/day3.127L1.534L/day0.66L[31]Datopotamab deruxtecan0.416L/day3.02L0.341L/day2.84L[32] PK parameters extracted from reports which perform population pharmacokinetics analysis. Allometric Scaling The coefficients of determination and scaling exponents were calculated using GraphPad Prism version 10.1.2 and Microsoft Excel, respectively. Human CL values were predicted using the following allometric scaling equation: CLhuman=CLTg32×BWhumanBWTg32eCL Vchuman=VcTg32×BWhumanBWTg32eVc Qhuman=QTg32×BWhumanBWTg32eQ Vphuman=VpTg32×BWhumanBWTg32eVp Human body weight (BW) data were obtained from the same sources used to extract the CL values (Table 1). Also, 20 g was set as a fixed body weight for mice. To determine the optimal exponents, e CL was examined by titrating in 0.05 increments. Prediction errors for CL, Vc, Q, and Vp were calculated using the following equations, respectively: PE=CLpredicted−CLobservedCLobserved×100%for overprediction PE=CLpredicted−CLobservedCLpredicted×100%for underprediction PE=Vcpredicted−VcobservedVcobserved×100%for overprediction PE=Vcpredicted−VcobservedVcpredicted×100%for underprediction PE=Qpredicted−QobservedQobserved×100%for overprediction PE=Qpredicted−QobservedQpredicted×100%for underprediction PE=Vppredicted−VpobservedVpobserved×100%for overprediction PE=Vppredicted−VpobservedVppredicted×100%for underprediction Twofold and threefold prediction accuracy ranges are −100% to 100% and −200% to 200%, respectively. Twofold difference criterion has been utilized in the study conducted by Li et al., which showed the single‐species allometric scaling method using cynomolgus monkey [19]. To compare with the twofold difference calculated in this prior study, the same criterion was selected. Data Analysis The relative bias and root mean squared error (RMSE) were calculated by following equations, respectively: Relative bias=1N∑i=1Nθ^i−θiθi×100% RMSE=1N∑i=1Nθ^i−θi2 where θ^i is the predicted parameters, and θi is the observed parameters. N refers to the number of evaluated ADCs. Results Plasma Concentration‐Time Profiles of Total ADC and Total Antibody in Mice Among the 15 FDA‐approved ADCs, eight commercially available ADCs with diverse linker and payloads were used to evaluate the applicability of Tg32 mice for predicting human PK parameters (Table 2). The absence of cross‐reactivity with mice antigens for all ADCs used in this study enabled the evaluation of target‐independent pharmacokinetic profiles. After intravenous administration of ADCs to Tg32 mice and C57BL/6J WT mice, plasma concentrations of total human antibody and total ADC were quantified using two types of ligand‐binding assays: anti‐human IgG antibody sandwich format and anti‐payload antibody capture/anti‐human IgG detection format, respectively. The anti‐payload antibody capture assay is theoretically capable of detecting analytes that have more than one payload. As shown in Figures 1 and 2, all ADCs exhibited biphasic elimination profiles for both total antibody and total ADC. PK parameters calculated by non‐compartmental analysis are summarized in Tables S1 and S2. Except for sacituzumab govitecan, all ADCs showed higher total ADC CL than total antibody CL. Inotuzumab ozogamicin was rapidly eliminated, resulting in a clear discrepancy between total ADC CL and total antibody CL. Sacituzumab govitecan did not exhibit the discrepancy between total ADC CL and total antibody CL, and showed the highest total antibody CL among the evaluated ADCs. TABLE 2Panel of evaluated FDA‐approved ADCs, including target antigens, payloads, DAR, and linker types.Drug nameTrade nameTargetPayloadDARLinkerLinker typeTisotumab vedotinTivdakTFMMAE4Valine–citrullineCleavableLoncastuximab tesirineZynlontaCD19PBD2.3Valine–alanineCleavableSacituzumab govitecanTrodelvyTROP2SN‐387.6CL2ACleavableTrastuzumab deruxtecanEnhertuHER2Dxd7.8Glycine–glycine–phenylalanine–glycineCleavablePolatuzumab vedotinPolivyCD79bMMAE3.5Valine–citrullineCleavableInotuzumab ozogamicinBesponsaCD22Calicheamicin6AcBut acyl hydrazone‐disulfideCleavableTrastuzumab emtansineKadcylaHER2DM13.5SMCCNon‐cleavableDatopotamab deruxtecanDatrowayTROP2Dxd4Glycine–glycine–phenylalanine–glycineCleavable Panel of evaluated FDA‐approved ADCs, including target antigens, payloads, DAR, and linker types. Mean ± SD plasma concentration–time profiles of ADCs in Tg32 mice. Open squares indicate total ADC concentration‐time profiles and closed squares indicate total antibody concentration‐time profiles. Mean ± SD plasma concentration–time profiles of ADCs in WT mice. Open squares indicate total ADC concentration‐time profiles and closed squares indicate total antibody concentration‐time profiles. Comparative Analysis Between Mice and Human Because human CL values were extracted from population pharmacokinetic reports which performed two compartment model analyses (Table 1), two compartment analysis was performed in the mouse study. The results of two compartment model analysis are shown in Tables 3 and S3–S5. To compare mice and human, the correlation of total ADC CL was evaluated (Figure 3). Mice CL used in these correlation plots were from two compartment analysis without scaling. WT mice did not show a clear correlation for total ADC CL (R 2 = 0.308), whereas Tg32 mice showed a much stronger coefficient of determination (R 2 = 0.888). Regarding other two compartment model parameters (Vc, Q, Vp), significant correlations with human parameters were not observed (Figure S1). TABLE 3Estimated CL parameters based on two compartment model analysis. Values are presented as mean ± SD.Drug nameTg32WTTotal ADC CL (mL/day/kg)Total antibody CL (mL/day/kg)Total ADC CL (mL/day/kg)Total antibody CL (mL/day/kg)Tisotumab vedotin11.9 ± 0.36.2 ± 0.219.5 ± 3.611.3 ± 1.0Loncastuximab tesirine9.2 ± 1.14.9 ± 1.48.3 ± 2.07.5 ± 1.3Sacituzumab govitecan28.6 ± 2.637.7 ± 3.214.7 ± 1.415.4 ± 0.5Trastuzumab deruxtecan8.5 ± 0.55.8 ± 0.77.9 ± 0.68.2 ± 0.4Polatuzumab vedotin14.9 ± 2.08.7 ± 2.224.8 ± 14.613.0 ± 9.1Inotuzumab ozogamicin27.2 ± 1.012.6 ± 3.036.4 ± 1.411.1 ± 0.8Trastuzumab emtansine8.4 ± 0.46.5 ± 0.612.3 ± 1.56.8 ± 0.8Datopotamab deruxtecan12.8 ± 1.19.8 ± 0.810.3 ± 1.05.5 ± 1.1 Estimated CL parameters based on two compartment model analysis. Values are presented as mean ± SD. Correlation analysis for CL between human and Tg32 mice or WT mice, based on reported parameters in clinical trials. Prediction of Human CL Based on Mouse CL Regarding the CL for Tg32 mice, the optimal allometric scaling exponent was calculated by titration, in which the increment was set as 0.05 (Table 4). As a result, the optimal exponent for total ADC CL was calculated as 1.0. Prediction accuracy within a twofold difference of observed human total ADC CL was 87.5%. Under this condition, the root mean square error (RMSE) which represents the difference between observed values and predicted values was 8.12. Only loncastuximab tesirine fell outside of this range; its human CL was overestimated with a 2.75‐fold prediction error. The optimal allometric scaling exponent for other parameters (Vc, Q, Vp) was also calculated (Tables S6–S8). The optimal exponents that maximize the twofold prediction accuracy of Vc, Q, and Vp are 1.05, 0.65–0.8, and 0.95, respectively. TABLE 4Estimated human CL of total ADC based on allometric scaling from Tg32 mice.Human CL (mL/day/kg)Exponent: 0.85Exponent: 0.9Exponent: 0.95Exponent: 1Exponent: 1.05Exponent: 1.1CL (mL/day/kg)Prediction error (%)CL (mL/day/kg)Prediction error (%)CL (mL/day/kg)Prediction error (%)CL (mL/day/kg)Prediction error (%)CL (mL/day/kg)Prediction error (%)CL (mL/day/kg)Prediction error (%)Tisotumab vedotin18.93.45−448%5.21−263%7.86−141%11.9−59.6%17.9−5.76%27.042.7%Loncastuximab tesirine3.362.67−25.9%4.0420.1%6.1081.6%9.23175%13.9315%21.1528%Sacituzumab govitecan45.68.42−442%12.7−260%19.0−140%28.6−59.3%43.0−5.93%64.742.0%Trastuzumab deruxtecan7.282.58−182%3.85−89.2%5.74−27.0%8.5417.3%12.774.7%19.0160%Polatuzumab vedotin11.04.34−154%6.55−68.2%9.88−11.4%14.935.4%22.5104%33.9208%Inotuzumab ozogamicin37.07.95−365%12.0−209%18.1−105%27.2−35.9%41.010.9%61.967.2%Trastuzumab emtansine9.662.47−292%3.71−160%5.58−73.2%8.39−15.1%12.630.6%19.096.4%Datopotamab deruxtecan6.483.83−69.3%5.73−13.1%8.5832.4%12.898.3%19.2197%28.8345%Within 2‐fold prediction accuracy25.0%50.0%62.5%87.5%62.5%50.0%Within 3‐fold prediction accuracy50.0%62.5%100.0%100.0%87.5%62.5%Relative bias−62.9%−44.1%−16.0%26.4%90.2%186.1%RMSE (Root mean square error)18.0015.6812.348.127.6618.07 Estimated human CL of total ADC based on allometric scaling from Tg32 mice. The optimal exponents for CL, Vc, Q, and Vp in WT mice were also calculated as 0.95, 0.95–1.0, 0.65–0.75, and 0.9, respectively (Tables S9–S12). The within twofold prediction accuracy for CL was 87.5%, which is comparable to that for Tg32 mice. The CL of sacituzumab govitecan in humans was not predicted within twofold accuracy and was underestimated by 4.67‐fold. Also, the RMSE value was 13.8, which is higher than that for Tg32 mice. Discussion Although Tg32 mice have been utilized for pharmacokinetic prediction of monoclonal antibodies, the application of this human FcRn transgenic mice for predicting the CL of ADC has not been reported [22, 23, 24]. The present study was conducted to evaluate the applicability of Tg32 mice as a pre‐clinical model for predicting the human PK of ADCs. Analysis of FDA‐approved ADCs revealed that total ADC CL in Tg32 mice correlated strongly with human CL, and that the prediction accuracy within twofold of CL was 87.5% when the exponent was set as 1.0 (Figure 3). These findings indicated that human FcRn transgenic mice serve as a translational tool for predicting the human ADC PK as well as monoclonal antibodies. The optimal allometric scaling exponent for predicting human ADC CL from Tg32 mice was determined to be 1.0, which is higher than the values previously reported for monoclonal antibodies (0.93 by Avery et al. and 0.9 by Betts et al.) [22, 23]. Although the underlying mechanisms remain unclear, this difference is likely attributed to ADC‐specific clearance pathways. The pharmacokinetics of monoclonal antibodies are regulated by FcRn‐mediated recycling, target‐mediated drug disposition, and non‐specific pinocytosis and binding, whereas the pharmacokinetics of ADCs are additionally influenced by the properties of conjugated linker and payload. It is well established that deconjugation contributes to faster systemic elimination of total ADC [33]. Consistent with this, shorter half‐lives for total ADC relative to total antibody were observed for most ADCs in the present study (Table S1). In addition, the conjugated linker and payload also affect ADC PK. It has been reported that ADCs conjugating hydrophobic linkers are eliminated by nonparenchymal liver cells [34]. Studies using severe combined immunodeficiency (SCID) mice have shown that CL increases with higher DAR for ADCs generated by MMAE conjugation to cysteine residues of anti‐CD30 monoclonal antibody [35]. In non‐obese diabetic scid gamma (NSG) mice, plasma samples collected 48 h after administration of EGFRvIII mAb‐MMAF with DAR0, DAR4, DAR6, and DAR8 revealed that ADCs with a higher number of payload conjugations exhibited lower plasma concentrations [36]. Given this mechanism, the difference in the allometric scaling exponent between ADCs and monoclonal antibodies in the present study would be caused by the conjugated linker and payload. In summary, these human FcRn binding‐independent clearance mechanisms may explain the observed difference in scaling exponents between ADCs and monoclonal antibodies in human FcRn transgenic mice. In the present study, not only Tg32 mice but also WT mice are used for the evaluation. As for the predictability of human CL, both Tg32 mice and WT showed 87.5% accuracy within twofold range. Given that the optimal exponents were derived from datasets comprising a limited number of ADCs, it will be necessary to assess a greater number of ADCs to fully determine which mice are more valuable as a prediction tool. Chunze Li et al. reported an allometric scaling exponent of 1.0 for predicting human ADC CL from cynomolgus monkeys, which is consistent with the exponent identified in the present study [19]. This scaling factor was calculated from 11 ADCs, including VC‐MMAE ADCs and mcc‐DM1, and the prediction accuracy within a twofold range was 100% [19]. On the other hand, the present study showed that only loncastuximab tesirine was overestimated by 2.75‐fold when calculating human CL using the optimal exponent condition. This anti‐CD19 ADC is the only ADC that utilizes VC‐PBD as its cytotoxic payload. Differences in prediction accuracy between Tg32 mice and cynomolgus monkeys would be caused by species differences in FcRn expression, deconjugation‐related proteins, pH environment, and blood flow rate. However, it is important to note that the diversity in physicochemical properties of ADCs used for exponent calculation differs between Tg32 mice in the present study and the cynomolgus monkey dataset reported by Chunze Li et al. In the cynomolgus monkey dataset, ADCs had an average DAR of 3.5, except for brentuximab vedotin, and were largely conjugated with VC‐MMAE via disulfide bond. Conversely, the ADCs dataset in the present study showed greater diversity in DAR, cytotoxic payload, linker type, and conjugation method. Therefore, the observed difference in prediction performance is insufficient to conclude that cynomolgus monkeys are superior for extrapolating human CL for ADCs. Moreover, considering advantages in availability, Tg32 mice represent a practical model for pharmacokinetic prediction, especially in the early drug discovery stage where high throughput and cost efficiency are required. Predicting the time‐concentration profile of ADC in humans from pre‐clinical animals is required when first‐in‐human dose is selected. However, not only the predicted CL but also other predicted pharmacokinetic parameters are essential to predict PK profile. In this study, optimal exponents of all two‐compartment model parameters were calculated. The two‐compartment model parameters estimated from these exponents have the possibility that enables the PK prediction in clinical. Further investigation is required to confirm the applicability. There are limitations to the present allometric scaling method using Tg32 mice. First, as described above, the optimal exponent was derived from a dataset comprising a limited number of ADCs. To confirm the robustness of exponents calculated in this study and to establish a more generalizable scaling factor, it will be necessary to assess a greater number of ADCs with diverse molecule formats. There is also the possibility that the scaling exponent varies depending on ADC attributes such as DAR, payload, linker type, and conjugation site. To obtain a more precise scaling exponent, additional studies stratified by such ADC attributes are needed. Second, compared with in vivo approaches such as allometric scaling, ex vivo and in vitro‐based prediction methods offer advantages in throughput. Several reports have described pharmacokinetic prediction methods for ADCs that do not require animal studies. One study reported that a whole mouse blood incubation assay correlated with in vivo stability for payload‐conjugated antibodies in mice (R 2 = 0.87) [37]. In that report, 13 THIOMAB antibody‐drug conjugates with 3 types of payload were used for incubation assay [37]. In another report, the CL of PEGylated ADCs with varying PEG lengths in a rat PK study correlated well with uptake assays using Kupffer cells (R 2 = 0.94), implying conjugation‐dependent clearance enhancement in in vivo can be predicted from a cell‐based assay [38]. To further accelerate the ADC drug discovery, the development of in vitro and ex vivo methods which accurately predict human ADC PK is required, in parallel with in vivo approaches. In conclusion, this study proposes Tg32 mice as a model for predicting human pharmacokinetics of ADCs. These human FcRn transgenic mice enable high‐throughput, low‐cost evaluation of ADC potential for clinical translation. The ability to predict human PK from Tg32 mice at early stages of drug discovery can accelerate the development of ADCs. Author Contributions K.N. and K.H. wrote the manuscript; K.N. and K.H. designed the research; K.N. performed the research; K.N. analyzed the data. Funding This work was supported by the Chugai Pharmaceutical Co. Ltd. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Correlation analysis for Vc, Q, and Vp between human and Tg32 mice or WT mice, based on reported parameters in clinical trials. Table S1: Summary of Tg32 mice PK parameters calculated by non‐compartmental analysis. Values are presented as mean ± SD. Table S2: Summary of WT mice PK parameters calculated by non‐compartmental analysis. Values are presented as mean ± SD. Table S3: Estimated Vc parameters based on two compartment model analysis. Values are presented as mean ± SD. Table S4: Estimated Q parameters based on two compartment model analysis. Values are presented as mean ± SD. Table S5: Estimated Vp parameters based on two compartment model analysis. Values are presented as mean ± SD. Table S6: Estimated human Vc of total ADC based on allometric scaling from Tg32 mice. Table S7: Estimated human Q of total ADC based on allometric scaling from Tg32 mice. Table S8: Estimated human Vp of total ADC based on allometric scaling from Tg32 mice. Table S9: Estimated human CL of total ADC based on allometric scaling from WT mice. Table S10: Estimated human Vc of total ADC based on allometric scaling from WT mice. Table S11: Estimated human Q of total ADC based on allometric scaling from WT mice. Table S12: Estimated human Vp of total ADC based on allometric scaling from WT mice.