Prediction of Human Pharmacokinetic Profiles of the Antituberculosis Drug Delamanid from Nonclinical Data: Potential Therapeutic Value against Extrapulmonary Tuberculosis

ABSTRACT

Delamanid has been studied extensively and approved for the treatment of pulmonary multidrug-resistant tuberculosis; however, its potential in the treatment of extrapulmonary tuberculosis remains unknown. We previously reported that, in rats, delamanid was broadly distributed to various tissues in addition to the lungs. In this study, we simulated human plasma concentration-time courses (pharmacokinetic profile) of delamanid, which has a unique property of metabolism by albumin, using two different approaches (steady-state concentration of plasma-mean residence time [C_ss-MRT] and physiologically based pharmacokinetic [PBPK] modeling). In C_ss-MRT, allometric scaling predicted the distribution volume at steady state based on data from mice, rats, and dogs. Total clearance was predicted by in vitro-in vivo extrapolation using a scaled albumin amount. A simulated human pharmacokinetic profile using a combination of human-predicted C_ss and MRT was almost identical to the observed profile after single oral administration, which suggests that the pharmacokinetic profile of delamanid could be predicted by allometric scaling from these animals and metabolic capacity in vitro. The PBPK model was constructed on the assumption that delamanid was metabolized by albumin in circulating plasma and tissues, to which the simulated pharmacokinetic profile was consistent. Moreover, the PBPK modeling approach demonstrated that the simulated concentrations of delamanid at steady state in the lung, brain, liver, and heart were higher than the in vivo effective concentration for Mycobacterium tuberculosis. These results indicate that delamanid may achieve similar concentrations in various organs to that of the lung and may have the potential to treat extrapulmonary tuberculosis.

INTRODUCTION

Tuberculosis (TB) is a serious health problem and a leading cause of death worldwide. In 2019, approximately 10 million people developed TB, and 1.5 million people died from the disease (1). Although the pulmonary form occurs in about 85% of TB patients, almost all organs are represented in the remaining 15% of cases (1). TB treatment always consists of a multidrug regimen to prevent resistance development, but current standard treatment regimens require a long duration and the use of drugs with poor safety profiles. Therefore, to improve treatment outcomes, more effective and safer drugs are needed.

Delamanid (Deltyba, OPC-67683, nitro-dihydroimidazooxazole derivative; developed by Otsuka Pharmaceutical Co. Ltd., Tokyo, Japan) (2, 3) has been approved in Europe, Japan, South Korea, Hong Kong, China, and Russia for the treatment of pulmonary multidrug-resistant TB (MDR-TB). The World Health Organization has also provided guidance on how to use delamanid in the treatment of pulmonary MDR-TB, namely, when an effective regimen cannot be constructed due to resistance or tolerability (1). However, TB drugs, including delamanid, have been studied exclusively in pulmonary TB patients, and there is no clinical evidence regarding the potential use of delamanid for the treatment of extrapulmonary TB diseases. Thus, treatment decisions for patients with extrapulmonary TB disease are often empirical, without supportive clinical evidence. This is, of course, not ideal and likely affects treatment success. According to a retrospective cohort study conducted in the United States, extrapulmonary TB is a significant burden for some health departments due to prolonged treatment and increased hospitalizations (4). In 2013, the onset ratio of extrapulmonary TB in Korea reached 20.4% of all cases of TB (5). The risk factors for extrapulmonary TB are independent of patient age, sex, foreign birth, and human immunodeficiency virus infection. In extrapulmonary TB, the most commonly infected organs are the central nervous system, lymphatic system, pleura, gastrointestinal organs, and genitourinary organs (4, 5). In particular, meningitis, brain TB, is one of the most severe forms of extrapulmonary TB, resulting in death or neurological disability in 50% of patients (6, 7).

Due to the diversity of extrapulmonary TB disease, it is challenging to obtain clinical evidence on the efficacy of extrapulmonary TB treatment in clinical trials; however, one possible bridge from the clinical evidence obtained in pulmonary TB to other forms of TB is to examine the pharmacokinetic (PK) properties based on a drug’s pharmacodynamic driver in the target organ and compare those with what has been proven to be effective in the lung. Therefore, tissue distribution is an important determinant factor in the discussion of the effectiveness against extrapulmonary TB. Since tissue collection is limited in humans, modeling and predicting tissue distribution of drugs in human using animal data can provide necessary information to guide the use of TB drugs in extrapulmonary TB.

We previously reported that delamanid is broadly distributed to the lungs, brains, and various tissues in rats (8). In addition, delamanid has been shown to have potent inhibition of growth against MDR-TB strains in the liver and spleen in addition to the lung in a mouse model (9). These studies have suggested that delamanid is expected to distribute to many tissues and effectively eliminate TB bacteria infecting the tissues.

For the quantitative examination of tissue drug concentrations, it is essential to understand the metabolism and distribution of drugs in humans to predict the total clearance (CL) and the distribution volume at steady state (V_ss). Various prediction methods for human CL (CL_human) and V_ss (V_ss,human) have been reported and verified by comparing the observed value with the predicted value. Examples of the prediction methods for CL_human are in vivo allometric scaling methods using the animal CL (CL_animal) of single or multiple species and in vitro-in vivo extrapolation (IVIVE) methods with a scaling factor (SF) of the main metabolic enzyme (10, 11). In this study, we focused on the unique metabolic property of delamanid, which is mainly metabolized to the main metabolite by albumin (12, 13). With regard to V_ss,human, there are in vivo allometric scaling methods using the animal V_ss (V_ss,animal) and the semimechanistic Øie-Tozer methods with the physiological description (14, 15).

A simple approach for simulating the human plasma concentration-time courses (PK profile) using CL_human and V_ss,human obtained by the various methods is the steady-state plasma drug concentration-mean residence time (C_ss-MRT) approach (16). This approach is based on the assumption that PK profiles are similar and linear across species and that the transformed curves from a variety of animal species, including humans, can be superimposed. For example, using the C_ss-MRT approach, human PK profiles of four in-house compounds metabolized by the liver and four cephem-type antibiotics excreted through the urine could be accurately simulated (16, 17).

An alternative approach to simulating human profiles is the physiologically based PK (PBPK) modeling approach, which incorporates the PK data of the drug, including the tissue-to-plasma partition coefficient (K_p,tissue) for each organ and the intrinsic CL and the physiological parameters such as the organ blood flow and tissue volume (18). Although the PBPK model requires a number of data for model construction, it is a useful approach for simulating the time course of drug concentration in the tissues of interest and plasma. The PBPK model of rifapentine was used to assess the potential efficacy of regimens by simulating lung concentrations of drugs at steady state and by comparing those with the MIC for Mycobacterium tuberculosis (19). In the case of efavirenz, associations between central nervous system toxicity and the simulated brain profile have been discussed (20).

This study aimed to simulate the plasma and tissue concentrations of delamanid in humans and to discuss its potential effects against extrapulmonary TB. Hence, on the assumption of no distribution difference of delamanid among animal species and the metabolism by human albumin, we simulated the human PK profile of delamanid using the C_ss-MRT approach based on the V_ss and CL values for the whole-body and the PBPK models, which were constructed using the data on K_p,tissue, intrinsic CL, and physiological parameters for each organ.

RESULTS

Animal PK studies and allometric scaling for prediction of human V_ss.

Delamanid was intravenously administered at a dose of 3 mg/kg to mice, rats, and dogs and subsequently exhibited a low CL and high V_ss. There was no remarkable difference observed among the species (Fig. 1; Table 1). Using the animals’ intravenous (i.v.) data, the V_ss,human was predicted by allometric scaling. The logarithmic V_ss was approximately proportional to the logarithmic body weight of the species by applying the allometric scaling equation, and the regression line exhibited linearity with good correlation (R² = 0.99; Fig. 2). The allometric coefficient (a) and exponent (b) were 3.2036 and 1.1434, respectively. The V_ss,human was extrapolated to be 5.89 liter/kg for a typical subject with 70 kg body weight, and the fold error difference between the predicted value and the observed value was 1.35 (Fig. 2; Table 2).

FIG 1.

Plasma concentrations after intravenous administration of a single dose of delamanid (3 mg/kg) in mice, rats, and dogs. Values represent the mean + standard deviation of three mice and rats and four dogs.

TABLE 1.

Pharmacokinetic parameters of delamanid after a single intravenous administration in mouse, rat, and dog^a

Parameterb	Value for:
	Mouse	Rat	Dog
Dose (mg/kg)	3	3	3
C0 (ng/ml)	1,198.9	13,923.4	1,320.6
AUCt (ng·h/ml)	13,388	21,551	13,822
AUC∞ (ng·h/ml)	13,491	21,566	14,140
t1/2 (h)	6.3	9.2	17.6
CL (ml/min/kg)	3.7	2.3	3.6
Vss (liter/kg)	2.3	1.7	5.3
MRT (h)	10.5	12.5	24.5

^aValues are the mean of n = 3 (mouse and rat) and n = 4 (dog).

^bC₀, predose concentration; AUC_t, area under the curve to the last data point.

FIG 2.

Linear regression analysis of log distribution volumes at steady state versus log body weights for mice, rats, and dogs after intravenous administration of 3 mg/kg delamanid. The symbols represent mice (○), rats (△), dogs (□), and human (×, predicted). The body weights were 0.02, 0.25, 10, and 70 kg in mice, rats, dogs, and humans, respectively.

TABLE 2.

Predicted human pharmacokinetic parameters of delamanid and their accuracy

Prediction method	Predicted value	Observed valuec	Fold error differenced
Vssa (liter/kg)
Allometric scaling	5.89	4.37	1.35
CLb (ml/min/kg)
IVIVEe using SF of plasma albumin	0.774	2.71	0.29
IVIVE using SF of whole-body albumin	1.86		0.69

^aPredicted V_ss values were calculated using V_ss of mouse, rat, and dog after intravenous administration and each body weight (21).

^bPredicted CL were calculated using in vitro half-life in human plasma (0.64 h [13]) and SFs in Table 3.

^cCL and V_ss values were calculated using F = 0.36 of the average oral absorption fraction in humans (22).

^dPredicted value versus observed value ratio was calculated.

^eIVIVE, in vitro-in vivo extrapolation; SF, scaling factor.

Prediction of human CL.

The CL_human of delamanid was predicted using the IVIVE method. In detail, the intrinsic clearance of delamanid by human (CL_int,albumin) of 0.432 μl/min/mg was calculated from the in vitro half-life (t_1/2) in human plasma (13) and physiological albumin content values (21) (Table 3). To predict CL_human, the CL_int,albumin and two SFs (Table 3) were used based on the concept of IVIVE (Table 2). When the SF of the plasma albumin amount alone was used, the predicted CL_human was 0.774 ml/min/kg, and the fold error difference between predicted value versus observed was 0.29. On the other hand, when the SF of whole-body albumin amount was used, the predicted CL_human was 1.86 ml/min/kg, and the fold error difference was 0.69. The predicted CL_human with the SF of whole-body albumin was closer to the observed values than the SF of plasma albumin amount alone.

TABLE 3.

Calculation of scaling factor for human albumin

Parameter	Value	Reference no.
Body wt (kg)	70	21
Plasma albumin (g/100 ml)	4.18	21
Plasma vol (ml)	3,000	21
Extravascular-to-intravascular ratio of albumin	1.4	38
Scaling factors (mg albumin/kg body wt)
Plasma albumin	1,791
Whole-body albumin	4,299

Simulation of human PK profile using the C_ss-MRT approach.

The transformed PK profiles of delamanid in mice, rats, and dogs were almost superimposable (Fig. 3). Based on the assumption that the transformed PK profiles in humans were also superimposed, the human PK profiles after intravenous administration were simulated using the predicted V_ss,human (5.89 liter/kg) and CL_human (1.86 ml/min/kg) as described above. Then, the elimination rate constant (k_el) and volume of distribution (V) were estimated to be 0.0177 h⁻¹ and 7.73 liter/kg, respectively, by fitting to the one-compartment model. After a single oral dose of 100 mg of delamanid, the human PK profile was simulated using the one-compartment model with first-order absorption kinetics. The simulated and observed oral PK profiles were reasonably similar (Fig. 4). The time to maximum concentration of drug in serum (t_max), area under the curve to infinity (AUC_∞), and t_1/2 from the simulated PK profile were within 2-fold error to the observed values with acceptable prediction accuracy (Table 4). The t_1/2 value from the simulated PK profile was slightly higher than the observed t_1/2, and the C_max was lower than the observed C_max.

FIG 3.

Transformed pharmacokinetic profile of delamanid in animals after intravenous administration.

FIG 4.

Simulated and observed human plasma PK profile of oral administration of 100 mg delamanid using the C_ss-MRT and PBPK modeling approaches. The open circles with error bar represent the observed clinical data (n = 6; mean ± SD). The above figure shows the concentrations up to 24 h after administration.

TABLE 4.

Simulated and observed human PK parameters of delamanid after 100 mg oral dose and the prediction accuracy^c

Parameter	Predicted PK parameter of:		Observed value	Fold error differencea of:
	Css-MRTb	PBPK modeling		Css-MRTb	PBPK modeling
Cmax (ng/ml)	46.6	203.4	198.4	0.23	1.02
tmax (h)	5	2	4	1.25	0.50
AUC168h (ng·h/ml)	2,826	4,381	3,150	0.90	1.39
AUC∞ (ng·h/ml)	3,004	4,385	3,168	0.95	1.38
t1/2 (h)	41.0	16.7	23.7	1.73	0.71

^aPredicted value versus observed value ratio was calculated.

^bIn the C_ss-MRT approach, the human PK profile after a single oral administration of delamanid was simulated using the one-compartment model with first-order absorption kinetics.

^cThe k_a, lag time, k_el, V, and F used values of 0.85 h⁻¹, 0.69 h, 0.0177 h⁻¹, 7.73 liter/kg, and 0.36, respectively.

Simulation of human PK profile using PBPK modeling.

Based on the assumption that delamanid would be eliminated by the albumin in plasma and tissues, we simulated the human PK profile of delamanid after oral administration using the constructed PBPK model (Fig. 5). The simulated human PK profile of delamanid was able to capture the observed profile (Fig. 4). The C_max, t_max, AUC_∞, and t_1/2 from the simulated PK profile were within 2-fold error to the observed values with acceptable prediction accuracy (Table 4).

FIG 5.

Physiologically based pharmacokinetic model of delamanid, including metabolism by albumin. The line and dotted arrows represent the blood flow rate (Q) of tissues and clearance by albumin (CL_albumin).

Simulation of human tissue profile using PBPK modeling.

The lung, brain, heart, and liver concentration-time profiles of delamanid in humans after the clinical dosage for pulmonary MDR-TB (i.e., repeated oral dosing of 100 mg twice a day for 5 days) were simulated using the PBPK model. Figure 6 shows the simulated tissue profiles of delamanid and the effective concentration in the lung (268 ng/g) (8). In these tissues, the delamanid concentrations reached steady state at 96 h after dosing, and the concentrations were higher than the effective concentration.

FIG 6.

Simulated lung (A), brain (B), heart (C), and liver (D) concentration-time profiles of delamanid in humans after 5 days of repeated 100 mg oral dosing twice a day using PBPK modeling. The dotted line shows the speculated effective concentration (268 ng/g) in a previous report (8).

DISCUSSION

The assessment and prediction of the expected PKs in humans have become increasingly important in drug discovery and development. Various prediction methods for human PKs have been reported and verified by comparing observed values with predicted values.

To predict human exposure to delamanid, which has a unique property of metabolism, we used two different approaches (C_ss-MRT and PBPK modeling) in this study to simulate the PK profile of delamanid in humans and compared this with actual observations. In addition, to emphasize previous findings of the effectiveness of delamanid against extrapulmonary TB (8, 9), various tissue concentration profiles were simulated, and we discussed the effect of delamanid on extrapulmonary TB.

The key parameters for understanding the PK properties of the drug are V_ss and CL, which describe the degree of tissue distribution in the whole body and the drug’s elimination rate from the body, respectively. For the simulation used in the C_ss-MRT approach, the predicted V_ss,human and CL_human were calculated as C_ss,human and MRT_human (see the C_ss-MRT approach in Materials and Methods). Thereby, the relationship between body weight and V_ss was established with good correlation by allometric scaling based on the data from mice, rats, and dogs (R² = 0.99; Fig. 2). Furthermore, the predicted V_ss,human from the regression line was within 2-fold error to the observed value with good prediction accuracy (Table 2). These results indicate that there are no species differences in V_ss and that the differences of V_ss are mainly driven by the body size. With regard to CL, two types of SFs were applied to predict CL_human using the IVIVE concept (Tables 2 and 3). The prediction accuracy using the SF of whole-body albumin was superior to that using the SF of plasma albumin (Table 2). Taking into consideration that delamanid is mainly metabolized by albumin and the contribution of other metabolic pathways or urinary/biliary excretion can be negligible (22), these results suggest that not only human albumin in plasma but also various tissues contribute to the metabolism of delamanid. Moreover, the predicted CL_human based on albumin activity in the in vitro experiments and the albumin amount in the body was almost identical to the observed CL_human, indicating that the metabolic activity of albumin could well determine in vivo CL.

In the C_ss-MRT approach, the transformed intravenous PK profiles obtained from mice, rats, and dogs were well superimposed (Fig. 3), which suggests that the PK profiles among animals are similar. In this study, we used the predicted V_ss,human value from allometric scaling and the predicted CL_human value from the IVIVE concept to calculate the C_ss,human (dose/V_ss,human) and MRT_human (V_ss,human/CL_human). Sequentially, the human oral PK profile of delamanid was simulated in combination with the transformed intravenous PK profiles from animals, the calculated C_ss,human and MRT_human, and the absorption parameters. The human PK profile obtained was almost identical to the observed profile (Fig. 4; Table 4), suggesting that there are no remarkable differences in PK profiles of delamanid. However, the goodness-of-fit plot in the C_ss-MRT approach showed overprediction at low concentrations and underprediction at high concentrations around the C_max (Fig. 7A). The discrepancy between the observed and predicted values may be caused by the elimination profile. The observed plasma PK profile of delamanid after oral administration to humans is biexponential elimination. The volume of distribution in the central compartment and that in the peripheral compartment were 1.14 liter/kg and 3.63 liter/kg, respectively, as estimated by fitting the observed human PK profile after oral administration to the two-compartment model. There might be species differences between humans and animals in the distribution rate from the central compartment to the peripheral compartment after administration. However, the predicted human V_ss by allometric scaling was almost the same value as the V_ss, which is the sum of the volume of distribution in the central compartment and that in the peripheral compartment, suggesting that human V_ss of delamanid was almost predictable.

FIG 7.

Goodness-of-fit plots of observed and predicted concentrations in the C_ss-MRT approach (A) and PBPK modeling approach (B).

In the PBPK modeling approach, based on the assumption that delamanid is metabolized by albumin in blood and each organ (12, 13), the 2 blood (arterial and venous), 10 organs (lung, brain, heart, liver, spleen, small intestine, kidney, muscle, adipose, and skin), and rest compartments were adopted in the model (Fig. 5). In all compartments, delamanid was considered to be metabolized depending on the amount of albumin. To incorporate tissue distribution, each K_p,tissue of the mass balance equations (see the “PBPK modeling and simulation” in Materials and Methods) used the K_p values of the rat that received delamanid (8). Delamanid is thought to distribute through passive diffusion, as it is not a substrate for the influx and efflux transporters related to tissue distribution (23). It is generally thought that there is no species difference in passive distribution. This is supported by the fact that the C_ss-MRT approach adequately predicted the observation in which a species difference was not taken into account. Therefore, we used the K_p,tissue values of the rat in the PBPK model equations.

For the K_p,rest value in the compartment, we performed the sensitivity analysis in the range of one-third or 3-fold of the observed mean rat K_p values in all tissues. The results did not demonstrate a remarkable impact on the simulated PK profiles or their PK parameters (data not shown), which suggests that the rest compartment in this PBPK model has a negligible contribution to distribution and metabolism. In other words, this PBPK model, which includes 10 tissues with the K_p,tissue of the rat, was considered to be sufficient for explaining the human PK of delamanid. Using this PBPK model, the simulated human PK profile of delamanid after oral administration was well consistent with the observed profile (Fig. 4; Table 4). The goodness-of-fit plot in the PBPK modeling approach showed good prediction at all concentrations (Fig. 7B). This shows that the model adequately describes the physiological process, including the metabolism and disposition of delamanid. Similar to delamanid (12, 13), the metabolism by albumin has been reported for several other compounds, namely, aspirin (24, 25), olmesartan medoxomil (26), p-nitrophenyl esters (27–31), cinnamoyl imidazole (32), carbaryl (33), organophosphate insecticides (34), and long- and short-chain fatty acid esters (35). Currently, there are no reports on human PK simulations for these drugs involving the metabolism by albumin. Considering that these compounds share the same metabolic pathway, the clearance and PK profile in humans may be adequately predicted by either the IVIVE method or the PBPK modeling approach, similar to delamanid.

Using this PBPK model, the concentration-time profiles were simulated in the lung, brain, heart, and liver after repeated oral administration for 5 days using the treatment regimen against MDR-TB in the clinical setting (i.e., 100 mg orally twice a day; Fig. 6). The existence ratio of the unchanged form of these tissues was clarified in a rat study (8). We proposed in previous studies that, based on the relationship between PKs and a pharmacodynamic pulmonary TB model in mice and rats, the effective concentration in the lung is 268 ng/g (8, 9, 36). In an MDR-TB study of a mouse model, the researchers reported that delamanid reduced CFU by more than 99% against MDR-TB in the liver, spleen, and lungs after an approximate 1.625-mg/kg oral dose, demonstrating the strong pharmacologic effect of delamanid on MDR-TB in extrapulmonary tissues (9). The liver and spleen concentrations in mice at this time were expected to be sufficiently higher than 268 ng/g based on tissue distribution data in rats (8). The nonclinical data indicated that delamanid can eliminate TB bacterial infection in tissues where the exposure of delamanid is comparable to or greater than that in the lungs. In this study, the steady-state concentration of delamanid in the lung was simulated to be sufficiently greater than 268 ng/g. Thereby, using this PBPK model, the clinical efficacy of delamanid for pulmonary TB was also supported via the consideration of the effective and simulated tissue concentration. In the brain profile simulation, the exposure of delamanid in the brain was even higher than that in the lung. In fact, one clinical study demonstrated that delamanid exists in human cerebrospinal fluid (37). These results suggest that delamanid might be effective for the treatment of brain meningitis TB. Similarly, simulated exposures in the liver and heart, which are organs that represent extrapulmonary TB, were higher than that in the lung (Fig. 6), suggesting that the pharmacological activity of delamanid is also expected in these tissues.

Due to the diversity of disease backgrounds and the limited number of patients, the treatment of patients with extrapulmonary TB is challenging. The PK of delamanid differed minimally between normal and TB conditions in both mice (3, 12) and human studies (22). Therefore, these simulated results may support the clinical benefit of delamanid for both the pulmonary and extrapulmonary form of TB. We hope that, in the near future, the efficacy of delamanid against extrapulmonary TB will be confirmed through clinical trials.

The PBPK model constructed in this study could not be adapted to rats because the metabolic activity of albumin in rats was less than that in human (13). However, in the PBPK modeling approach, the human PK profile using the model incorporating the distribution using rat K_p,tissue values, and the organ albumin-dependent metabolic clearance was successfully simulated in terms of the observed human PK profile. This result may be considered one of the model validations. In this study, the pharmacokinetic parameters (i.e., CL and V_ss) of delamanid in humans are assumed to be linear. As per nonclinical data, the linear kinetics were considered to be maintained at exposure levels of 3 mg/kg intravenously or 3 to 10 mg/kg orally in animals because the t_1/2 values of delamanid were similar after intravenous or oral administration (12). Clinically, the t_1/2 of single doses of delamanid in humans at 100, 200, and 400 mg (approximately 1.4, 2.6, and 5.7 mg/kg) were similar (22), suggesting that linearity of CL and V_ss is maintained in this dose range. This nonclinical and clinical information indicated that the linearity of PK is established in the dose level of this study (i.e., 3 mg/kg in nonclinical experiment and approximately 1.4 mg/kg in clinical). In nonlinear conditions (e.g., saturation of V_ss and CL), it may be necessary to incorporate the appropriate mathematical model. The exposure after single oral administration of delamanid in humans (22) increased with dose escalation, but the increase was less than dose proportional, suggesting the saturation of absorption. It may be difficult to estimate the absorption rate for delamanid due to physicochemical properties (e.g., high lipophilicity) and differences in the intestinal environment.

In conclusion, the human PK profiles of delamanid were simulated by C_ss-MRT and PBPK modeling approaches using animal PK data and in vitro metabolism by albumin. The simulated PK profiles were almost identical to the observed PK profile. Both the distribution and metabolism of delamanid in humans were well predicted; that is, there was no remarkable difference in the distribution among species, and the metabolism by albumin was involved in the whole body. These factors could be used to reliably simulate the human PK profile of delamanid. Therefore, this approach may be applicable to other compounds which are also metabolized by albumin. In addition, after repeated oral administration of the clinical treatment regimen against MDR-pulmonary TB, the simulated steady-state concentration of delamanid in the lung, brain, liver, and heart in humans was higher than the in vivo effective concentration for M. tuberculosis. Therefore, delamanid was considered to have potential therapeutic value for the treatment of both extrapulmonary and pulmonary TB.

MATERIALS AND METHODS

Chemicals and reagents.

Delamanid was obtained from Otsuka Pharmaceutical Co., Ltd. (Tokyo, Japan). All other chemicals used in the study were commercially available.

Animals.

Male ICR mice aged 5 weeks were purchased from Japan SLC, Inc. (Shizuoka, Japan). Male Sprague-Dawley rats aged 5 weeks were purchased from Charles River Laboratories (Yokohama, Japan). Male and female beagle dogs aged 6 months were purchased from Covance Research Products, Inc. (Cumberland, VA). All animal experimental protocols and procedures were reviewed in accordance with the Guidelines for Animal Care and Use at Otsuka Pharmaceutical Co., Ltd. and were approved by the in-house animal ethics committee.

Animal PK studies.

The intravenous dosing solution of delamanid was prepared by dissolving delamanid in dimethyl sulfoxide on the day of the experiment. The mice, rats, and dogs were administered intravenous doses of 3 mg/kg in the tail vein for the rodents and in the cephalic vein for dogs. Blood was collected at 0.08, 0.25, 0.5, 1, 2, 4, 6, 8, 12, 24, 32, and 48 h in mice and additionally at 72 and 96 h in rats and at 3, 72, and 96 h in dogs after dosing (n = 3 for the rodents at each time point and n = 4 for the dogs). To separate the plasma, heparinized blood was centrifuged (centrifuge model 05PR-22; Hitachi, Japan) at 1,870 × g for 10 min at 4°C to avoid metabolism by albumin. All plasma samples were stored at −20°C until analysis by liquid chromatography-tandem mass spectrometry to avoid metabolism by albumin.

Determination of delamanid in animal plasma samples.

Liquid chromatography-tandem mass spectrometry was used to determine the delamanid concentrations in plasma samples, as described in a previous report (22). The high-performance liquid chromatography eluate was directly introduced into a TSQ 7000 triple quadrupole mass spectrometer (Thermo Fisher Scientific, Inc., San Jose, CA), and the mass spectrometer was operated in the positive electrospray ionization selected reaction monitoring mode. The selected reaction monitoring mode of delamanid was used with the following transitions: m/z 535.2 (precursor ion) and m/z 352 (product ion). Data acquisition and processing were performed using Xcalibur software (version 1.2; Thermo Fisher Scientific Inc.). The lower to upper limits of quantification of the delamanid concentration in plasma samples were 3 to 500 ng/ml in rats and dogs and 6 to 1,000 ng/ml in mice.

Allometric scaling for human V_ss.

The V_ss values of mice, rats, and dogs were scaled allometrically to humans using the body weight of the animals. Data from mice, rats, and dogs were fitted to the following equation, and the coefficient (a) and exponent (b) were obtained. Then, human V_ss was calculated using the typical human body weight (BW) (i.e., 70 kg), as follows:

$$V_{\text{ss}}=a\times {\left(\text{BW}\right)}^b$$

IVIVE calculation for human CL.

Using the IVIVE concept, the human CL value was calculated by the following equations. First, the intrinsic clearance of delamanid in human albumin (CL_{int, albumin}) was calculated from the t_1/2 in human plasma (0.64 h) (13) and physiological albumin content values (4.18 g/100 ml plasma) (21), and then, the human CL was calculated with CL_{int, albumin} using two SFs of albumin amount (21, 38), namely, the amount of plasma albumin (intravascular albumin) and the whole-body albumin (intravascular and extravascular albumin) in humans (Table 3).

$${\text{CL}}_{\text{int,albumin}}=\frac{\text{ln}2}{t_{1/2}}\times \left(\text{physiological albumin content}\right)$$

$$\text{CL}={\text{CL}}_{\text{int,albumin}}\times \left(\text{SF of plasma albumin}\right)\text{or}\left(\text{SF of whole}–\text{body albumin}\right)$$

C_ss-MRT approach.

Transformed PK profiles of delamanid were derived by dividing the y axis of the animals’ plasma concentration by each C_ss,animal (equal to dose/V_ss,animal) and by dividing the x axis of the times by each MRT_animal (CL_animal/V_ss,animal; Fig. 3). In addition, we confirmed that in three animals, the transformed PK profiles of delamanid were almost superimposable. Based on the assumption that the transformed PK profile in humans is identical to the transformed PK profiles of animals with monoexponential elimination, the transformed PK profile was multiplied by the predicted C_ss,human (dose/[predicted V_ss,human]) and MRT_human ([predicted V_ss,human]/[predicted CL_human]) to predict the human PK profile after intravenous administration. The predicted V_ss,human and CL_human values were predicted using an allometric scaling method and the IVIVE concept, respectively.

From the obtained intravenous PK profile, the human PK profile after a single 100-mg oral administration of delamanid was simulated with the one-compartment model based on first-order absorption kinetics, as shown in the following equation:

$$C_t=\frac{F\times \text{dose}\times k_a}{V_d\times \left(k_a-k_{\text{el}}\right)}\times \left(e^{-k_{\text{el}}\times t}-e^{-k_a\times t}\right)$$

We used the oral absorption rate constant (k_a) of 0.85 h⁻¹, which was obtained by fitting the observed human PK profiles (22; interview form of Deltyba and internal data) to the above equation with the lag time (0.69 h). We used a bioavailability (F) of 0.36, which is the average oral absorption fraction in human study with labeled delamanid (22).

PBPK modeling and simulation.

We constructed a human PBPK model of delamanid that consisted of several compartments corresponding to the different physiological organs of the body and linked by the circulating blood. Each compartment is described by the organ volume and blood flow rate. These physiological parameters are referred to in the literature (21, 39). This model consisted of 10 tissue types that contain a high amount of albumin or large volumes. Figure 5 shows a schematic of the PBPK model. All tissues, except for the 10 tissues above, were compiled into one compartment as the “rest” tissue. The volume or blood flow of the rest compartment was obtained by subtracting the sum of the 10 tissue types from the total volume or blood flow rate. In this PBPK model, concentrations of delamanid in plasma (C_plasma) and tissues (C_tissue) were calculated as follows:

$$C_{\text{plasma}}=\frac{C_{\text{blood}}}{R_b}$$

$$C_{\text{tissue}}=\frac{A_{\text{tissue}}}{V_{\text{tissue}}}$$

where C_blood, R_b, A_tissue, and V_tissue are the delamanid concentration in the blood, blood-to-plasma concentration coefficient, delamanid amount in tissue, and tissue volume, respectively. R_b (0.73) used the values in the in vivo rat study with radiolabeled delamanid (8).

The elimination of delamanid was assumed to be metabolized by only albumin in the whole body. Metabolic clearance by albumin is dependent on the amount of albumin contained in the blood and tissue compartment. In the mass balance equations, the clearances by albumin in plasma (CL_{albumin,plasma}) and tissues (CL_{albumin,tissue}) were represented as follows:

$${\text{CL}}_{\text{albumin,plasma}}=k_{\text{el,plasma}}\times V_{\text{venous}}\times \left(1-\text{HT}\right)$$

$${\text{CL}}_{\text{albumin,tissue}}=k_{\text{el,plasma}}\times K_{p\text{,albumin,tissue}}\times V_{\text{tissue}}$$

The k_el,plasma, HT, and K_{p,albumin,tissue} are the elimination rate constant of delamanid in human plasma, physiological hematocrit value, and tissue-to-plasma ratio of the albumin concentration, respectively. We obtained the k_el,plasma value of 1.083 h⁻¹ from in vitro t_1/2 in human plasma (13). The K_{p,albumin,tissue} was used in the literature (40). The k_p,albumin value of 0.112 of the rest compartment was used, which is the average value in tissues.

We obtained the time course of the amount in each compartment by solving the mass balance equations. Equations in typical compartments, such as venous, lung, gut, liver, and tissue, were as follows:

Venous

$$\begin{array}{l}\frac{{\text{dA}}_{\text{venous}}}{\text{dt}}=\frac{C_{\text{brain}}}{\left(\raisebox{1ex}{$K_{p\text{,brain}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{brain}}+\frac{C_{\text{heart}}}{\left(\raisebox{1ex}{$K_{p\text{,heart}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{heart}}+\frac{C_{\text{liver}}}{\left(\raisebox{1ex}{$K_{p\text{,liver}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{liver}}\\ +\frac{C_{\text{kidney}}}{\left(\raisebox{1ex}{$K_{p\text{,kidney}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{kidney}}+\frac{C_{\text{muscle}}}{\left(\raisebox{1ex}{$K_{p\text{,muscle}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{muscle}}+\frac{C_{\text{adipose}}}{\left(\raisebox{1ex}{$K_{p\text{,adipose}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{adipose}}+\\ \frac{C_{\text{skin}}}{\left(\raisebox{1ex}{$K_{p\text{,skin}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{skin}}+\frac{C_{\text{rest}}}{\left(\raisebox{1ex}{$K_{p,\text{rest}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{rest}}-C_{\text{venous}}\times Q_{\text{venous}}-\frac{C_{\text{venous}}}{R_b}\times {\text{CL}}_{\text{albumin, plasma}}\end{array}$$

Lung

$$\frac{{\text{dA}}_{\text{lung}}}{\text{dt}}=C_{\text{venous}}\times Q_{\text{venous}}-\frac{C_{\text{lung}}}{\left(\raisebox{1ex}{$K_{p\text{,lung}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{lung}}-\frac{C_{\text{lung}}}{K_{p\text{,lung}}}\times {\text{CL}}_{\text{albumin, lung}}$$

Gut

$$\frac{{\text{dA}}_{\text{gut}}}{\text{dt}}=C_{\text{arterial}}\times Q_{\text{gut}}+k_a\times F\times A_{\text{dose}}\times \frac{C_{\text{gut}}}{\left(\raisebox{1ex}{$K_{p,\text{gut}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{gut}}-\frac{C_{\text{gut}}}{K_{p,\text{gut}}}\times {\text{CL}}_{\text{albumin, gut}}$$

Liver

$$\begin{array}{l}\frac{{\text{dA}}_{\text{liver}}}{\text{dt}}=C_{\text{arterial}}\times Q_{\text{hepa}}+\frac{C_{\text{spleen}}}{\left(\raisebox{1ex}{$K_{p\text{,spleen}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{s\text{pleen}}+\frac{C_{\text{gut}}}{\left(\raisebox{1ex}{$K_{p,\text{gut}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{gut}}-\frac{C_{\text{liver}}}{\left(\raisebox{1ex}{$K_{p\text{,liver}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\\ \times Q_{\text{liver}}-\frac{C_{\text{liver}}}{K_{p\text{,liver}}}\times {\text{CL}}_{\text{albumin, liver}}\end{array}$$

Tissues

$$\frac{{\text{dA}}_{\text{tissue}}}{\text{dt}}=C_{\text{arterial}}\times Q_{\text{tissue}}-\frac{C_{\text{tissue}}}{\left(\raisebox{1ex}{$K_{p,\text{tissue}}$}\!\left/ \!\raisebox{-1ex}{$R_b$}\right.\right)}\times Q_{\text{tissue}}-\frac{C_{\text{tissue}}}{K_{p,\text{tissue}}}\times {\text{CL}}_{\text{albumin, tissue}}$$

K_p,tissue was based on the values in the in vivo rat study with the radiolabeled delamanid (8). The K_p,tissue values of the lung (1.86), brain (2.00), heart (2.92), and liver (10.33) were the AUC ratio of delamanid concentration in tissue to plasma. The K_p of the tissues, namely, the spleen (2.78), gut (2.55), kidney (6.28), muscle (1.86), adipose (3.94), and skin (1.99), were the AUC ratio of radioactivity in tissue to plasma. The K_p,rest was 3.62 because the average value of the ratio of the AUC in the 29 tissues in the in vivo rat study. Q_tissue, k_a, F, and A_dose represent the tissue blood flow rate, absorption rate constant, absorption fraction, and dose amount of delamanid, respectively. Delamanid was administered in the gut compartment and absorbed with an appropriated lag time. Table S1 in the supplemental material shows the details of the physiological parameters of this PBPK.

Using the constructed PBPK model, we simulated the plasma concentration in humans after a single oral administration of 100 mg of delamanid. Sequentially, we simulated the concentration-time profiles in the lung, brain, heart, and liver after 5 days of repeated oral administration of delamanid by clinical regimen (100 mg twice a day).

PK data and analysis.

The information on the clinical PK data for delamanid was obtained from an oral dosing study in six normal-health volunteers in Japan (22; interview form of Deltyba and internal data). Table S2 shows the PK parameters in humans. The V_ss and CL were calculated using the oral absorption fraction in a human study with labeled delamanid (22). The noncompartmental analysis, compartmental analysis, and PBPK model building and simulation in this study were performed using Phoenix WinNonlin version 8.0 (Certara LP). The goodness-of-fit plot was used to assess the predictive performance of each approach.