Population Pharmacokinetic Modeling and Simulation of Pudexacianinium (ASP5354) for Dose Setting of a Phase 2 First‐in‐Patient Study: A Novel Imaging Agent for Intraoperative Ureter Visualization during Abdominopelvic Surgery

Pudexacianinium (ASP5354) chloride is an indocyanine green derivative designed to enable enhanced ureter visualization during surgery. The objective of the present analysis was to determine appropriate doses of pudexacianinium for a phase 2, dose‐ranging study (NCT04238481). Real‐time urine pudexacianinium concentration is considered a good pharmacodynamic surrogate marker, since ureter visualization likely depends on its concentration in the ureter. Using plasma and urine concentrations of pudexacianinium from a phase 1 single‐ascending‐dose (0.1‐24.0 mg) study in healthy participants, a 3‐compartment population pharmacokinetic model with a urine output compartment was developed and effectively described the concentration‐time profiles. The individual estimated glomerular filtration rates had a significant impact on drug clearance. Simulations suggested that a 1.0 mg intravenous injection would achieve target urine concentrations over 1 μg/mL (determined from previous nonclinical studies) for 3 hours postdose, assuming a urine production rate of 1.0 mL/min. Based on this simulation, doses of 0.3, 1.0, and 3.0 mg were proposed for the phase 2 study. The observed plasma concentrations were generally consistent with model predictions. For urine, although only limited data could be obtained due to the difficulties of spot urine collection from surgical patients, intraoperative ureter visualization was successful at 1.0 and 3.0 mg.

2][3] If left undiagnosed and untreated, IUI can result in severe complications, including urinary fistula, renal dysfunction, or sepsis, 2 contributing to hospital readmission, increased healthcare costs, and increased mortality. 3dentifying the ureters during surgery can be challenging, especially in patients with advanced endometriosis, inflammatory disease, cancer, or a history of previous surgery or radiotherapy.Preoperative placement of ureteral stents can be utilized for highrisk procedures, 4 but their use necessitates longer operating time and the risk of IUI from the stenting itself cannot be completely avoided. 3,5Moreover, lack of tactile feedback available in minimally invasive and laparoscopic surgeries have been reported to increase the risk of IUI. 6 Imaging techniques such as intravenous or retrograde pyelography and urologic computed tomography can be employed presurgically to reduce the risk of IUI, 7 but they do not provide intraoperative ureter identification and include some radiation exposure.
Image-guided surgery using near-infrared fluorescence (NIR-F) imaging technology with contrast agents is an emerging technique for enhancing real-time visualization of biologic tissues and structures. 8Advantages of NIR-F imaging include its high tissue penetration, 9 low tissue autofluorescence, weak absorption, and facilitation of a strong visual signal without alteration of the appearance of the surgical field. 10Currently, 2 NIR-F dyes, methylene blue (MB) and indocyanine green (ICG), are clinically available, although neither has been approved by the US Food and Drug Administration for ureter visualization. 8Intravenous injection of MB for ureter visualization has been reported. 9However, MB is suboptimal for ureter identification. 10Since MB has an excitation peak of about 700 nm, which translates to a tissue penetration depth of 3-5 mm, improvement in penetration is expected by utilizing a contrast agent with a longer fluorescence wavelength. 11As for ICG, its clearance exclusively by the liver allows only for studies of retrograde ureteral ICG injection, 12,13 which requires expertise, instrumentation of the bladder, and additional surgical time.
Thus, neither MB nor ICG possesses optimal properties for intraoperative ureter imaging, and the need to develop improved agents for ureter visualization remains.][16] Pudexacianinium chloride (ASP5354 chloride, formerly TK-1) is a novel ICG derivative conjugated with β-cyclodextrin.][22] Pudexacianinium is currently being developed for real-time intraoperative ureter visualization during surgery with multiple clinical studies completed to date. 23,24hoto-optically, pudexacianinium shows NIR-F properties similar to those of ICG, with an absorption of 780 nm and a fluorescence emission of 820 nm but, due to its hydrophilicity, it is excreted rapidly and entirely into the urine after intravenous adminis-tration.Furthermore, the real-time urine pudexacianinium concentration is considered a theoretically good pharmacodynamic surrogate marker for ureter visualization.Ex vivo ureter imaging studies in minipigs suggested that a 1.0 μg/mL urine pudexacianinium concentration in the urinary tract should be sufficient for ureteral visualization. 19Intravenous administration of 0.01 mg/kg pudexacianinium to minipigs allowed sufficient intraoperative fluorescent visualization of the ureter up to 3 hours after administration, with a urine pudexacianinium concentration over 1 μg/mL sustained throughout the period.
In a phase 1, double-blinded, randomized, placebo-controlled study conducted in healthy participants (NCT03698305), the safety, tolerability, and pharmacokinetics (PKs) of pudexacianinium were investigated. 23Briefly, safety and tolerability were confirmed up to a single dose of 24.0 mg; no treatment-emergent adverse events were related to pudexacianinium.Observed PK parameters suggested a dose-proportional plasma concentration increase in the investigated dose range.Almost all pudexacianinium (76.8%-100%) was excreted unchanged in urine.The mean renal clearance rate ranged from 3.57 to 4.92 L/h, which was nearly equivalent to the mean total clearance rate (4.38-6.01L/h).In vivo metabolite profiling of pudexacianinium using plasma and urine samples showed no metabolites.
To our knowledge, none of the other uretervisualization agents under development have sought to apply clinical pharmacology principles, including modeling and simulation, when determining the dose for phase 2 dose-ranging first-in-patient studies.Rather, it seems that the efficacious doses in preclinical studies, adjusted with body weight calculations, are used directly for clinical dose setting.
The objectives of this analysis were to develop a population PK (PPK) model using data from the phase 1 study to assess the effects of covariates on pudexacianinium PK parameters and to determine a dose range for the phase 2 study by using the target urine concentration based on nonclinical studies and simulating individual plasma and urine pudexacianinium concentration-time profiles for patients under anesthesia during surgery.Finally, model qualifications were checked visually by comparing model results (predictions) with the PK findings from the phase 2 study (NCT04238481) at the proposed doses.

Study and Data Description
A phase 1, first-in-human, double-blinded, randomized, controlled, single-ascending-dose study in healthy adult participants (NCT03698305) 23  Data from the phase 1 study were used for PPK model development.
A total of 260 observed pudexacianinium concentrations for both plasma and interval urine collections and 160 spot urine pudexacianinium concentrations were obtained in the study.For model development, plasma and interval urine pudexacianinium concentrations were used.Data at predose sampling, missing data, and any concentration below the lower limit of quantification (BLQ) were excluded from the modeling data set.One missing plasma sample (sample not collected) in the 0.1 mg group, 1 implausible BLQ plasma concentration at 0.25 hours in the 0.5 mg group, and 1 outlier urine concentration in the 8 mg group were excluded from the modeling.
The phase 2 study (NCT04238481) was a randomized, open-label, dose-ranging clinical trial in adult participants undergoing laparoscopic, minimally invasive colorectal surgery during which the need for anatomic visualization of the ureter was anticipated. 24A total of 12 participants were randomly assigned and received an intravenous pudexacianinium dose of 0.3 mg (n = 3), 1.0 mg (n = 6), or 3.0 mg (n = 3).For the first 3 participants, plasma PK samples were collected predose, 10, 30, and 60 minutes postdose, and then every 30 minutes thereafter until a final collection at the end of surgery (or at 180 minutes if the surgery was completed in <105 minutes).For the 9 participants enrolled after the pro-tocol was amended, blood was collected predose, at 10 minutes and either 30 or 60 minutes after administration, and at the end of surgery (or at 180 minutes if the surgery was completed in <105 minutes).For urine PKs, total urine samples for the amount of pudexacianinium excreted in urine during surgery were collected.Spot urine samples collected at the same time as blood had been planned, but due to feasibility issues, point urine samples were collected only predose and at the end of surgery, after a protocol amendment.Data from the phase 2 study were used for model qualification and were not included in model development.

Analytical Methods
Pudexacianinium plasma and urine concentrations were measured using a validated high-performance liquid chromatography-tandem mass spectrometry assay.The lower limit of quantification was 0.001 μg/mL for plasma concentrations and 0.02 μg/mL for urine concentrations.Details of the analytical methods were reported by Murase et al. 23 Computer Software Data sets were prepared using SAS for Windows v.9.4 (SAS Institute, Inc., Cary, NC, USA).The modeling, simulations, and postprocessing were conducted in Metworx v.20.12.2 (Metrum Research Group, Tariffville, CT, USA).PPK parameters were estimated by nonlinear mixed-effects modeling with NONMEM v.7.5.0 (ICON Development Solutions, Ellicott City, MD, USA) compiled with GNU Fortran v.7.5.0.NON-MEM execution, model evaluation, and stepwise covariate modeling were performed with Perl Speaks NONMEM (PsN) v.4.9.0.R v.4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) was used for data management and postprocessing, including plot creation.Pirana v.2.9.9 (Certara, Princeton, NJ, USA) was used for run record management.

Model Structure
Plasma and interval urine pudexacianinium concentrations were analyzed by NONMEM using the first-order conditional estimation method with eta-epsilon interaction.The base model was identified by comparing different structural PK models: a 2-compartment model versus a 3-compartment model.The selection of initial structural models was guided by exploratory analyses.Once the basic structural model for plasma pudexacianinium concentration was identified, urine pudexacianinium concentrations were analyzed simultaneously using the output compartment.

Statistical Model
The statistical model included the random and residual variance models.Population parameters assumed a lognormal distribution: where θ is the typical population value of parameter P, subscript i denotes the ith patient, P i is the value of parameter P for the ith patient, and η i denotes the deviation for the ith patient's parameter value from the typical value θ.The random-effects η values were assumed to have a normal distribution, that is, η ∼ N(0, ω 2 P ).ω 2 P is the variance estimate for interindividual variability (IIV).

Covariate Exploration
Body weight (WT), sex (SEXF, 0:male, 1:female where P i is the individual parameter, θ 1 is the typical value parameter (intercept), COV is the covariate value, and θ 2 is the estimated covariate parameter.The effect of the categorical covariate SEXF was assessed using a proportional change model as follows: where θ 2 represents the proportional change in the covariate effect for females (X 1 = 1) compared with the male reference value of X 1 = 0.

Calculation of Half-Life
Individual participants' half-lives of the distribution phase (t 1/2,α ), elimination phase (t 1/2,β ), and terminal elimination phase (t 1/2,γ ) were estimated using the R package PKconverter 25 with post hoc PK parameters from the best PPK model.t 1/2 is defined as log(2) divided by the corresponding exponent α, β, or γ that represents each slope of the decay in the plot of drug concentration versus time.

Model Evaluations
Models were evaluated by assessing goodness of fit (plots and metrics), parameter uncertainty as the relative standard error, shrinkage, condition number, and/or predictive performance.Prediction-corrected visual predictive checks 26 were performed using 1000 simulations to determine if the developed models can simulate data that are consistent with the observed data.

Simulation
Ureter visualization theoretically depends on the urine concentration of pudexacianinium in the ureter.It is likely that the (real-time) urine concentration of pudexacianinium can be a good pharmacodynamic (visualization) surrogate marker.In nonclinical studies using Göttingen minipigs, it was observed that ureter visualization can be achieved at a urine pudexacianinium concentration of 1 μg/mL or higher, 19 which therefore was set as the target urine concentration for the present analysis.
For dose setting in the phase 2 study, the individual plasma and urine pudexacianinium concentrationtime profiles following a single intravenous injection of 0.1, 0.3, 1.0, 2.0, or 3.0 mg in patients under anesthesia during surgery were simulated for 1000 patients per each treatment assuming a normal distribution of body weight and fixed normal renal function (individual eGFR of 90 mL/min).The urine concentration in the ureter was simulated using the following equation: where C u and C p represent urine and plasma pudexacianinium concentrations, respectively, CL R represents renal clearance (same with total CL), and R urine represents urine production rate.The proportion of patients with a urine concentration above the target (1 μg/mL) over a certain duration was calculated from the simulated concentration-time curves.The simulations were performed assuming that the urine production rate during surgery would be controlled at 1.0 mL/min, 27,28 that there was no IIV in the urine production rate, and that there would be no delay observed between plasma concentration and excretion into urine.The impact of urine production rate change was also tested using a sensitivity analysis by varying R urine from 0.5 to 2.0 mL/min.

Model Qualification
After the phase 2 study was conducted with the doses proposed based on the simulations, the clinical data were compared with those in the model.The ability of the developed PPK model to simulate plasma and spot urine concentration-time profiles in the phase 2 study 24 was confirmed using overlay plots of observations and 90% prediction intervals using generated data for 1000 simulated patients whose urine production rate ranged from 0.2 to 2.0 mL/min (low urine production rate under anesthesia as observed in the phase 2 study) or 1.0 to 4.5 mL/min (normal or high urine production rate as observed in the phase 1 study).The generated urine production rate assuming a uniform distribution in the population was used as a constant value per patient in the simulation.

Demographics
The demographics of participants in the phase 1 and 2 studies are summarized in Table 1.There were 30 participants enrolled in the phase 1 study.Of the 20 participants who were administered a single dose of pudexacianinium and included in the PPK analysis, 50% were male and 50% were female.Of the 20 participants, 13 participants were White, 5 were Black or African American, 1 was Asian, and 1 was Native Hawaiian or other Pacific Islander.Of the 12 participants who received pudexacianinium in the phase 2 study, 2 were male and 10 were female; 10 participants were White, 1 was Black or African American, and 1 was Asian.There were no significant differences in the demographics of participants between the 2 studies, except for urine production rate.

Observations
A total of 218 plasma observations and 216 interval urine observations from 20 participants were included in the modeling.The observed pudexacianinium concentration-time profiles for plasma and urine were reported by Murase et al. 23 Briefly, plasma PKs were linear in the dose range from 0.1 to 24.0 mg.Almost all pudexacianinium administered was excreted unchanged into urine in all tested doses.

Base Model
A 3-compartment model with log-normally distributed IIV on clearance (CL), central volume of distribution (V 1 ), shallow peripheral volume of distribution (V 2 ), and deep peripheral volume of distribution (V 3 ) was selected as the base model.IIV could not be estimated for the intercompartment clearances (Q 2 and Q 3 ).For the plasma concentrations, a proportional error model was used.For the urine concentrations, a proportional plus additive error model was used.The structure of the 3-compartment model incorporating the urine output compartment is shown in Figure 1.

Covariate Exploration
All tested covariates except for ALB were statistically significant (P < .05)after the forward addition step of SCM.The final model after the forward step included EGFRI, SEXF, and UVOLT on CL, SEXF and WT on V 1 , and WT on V 2 .After the backward elimination step (P > 0.01), UVOLT on CL and SEXF on V 1 were removed.Considering the PK characteristics of renal elimination, all selected variables and the magnitude of coefficients were reasonable.Finally, EGFRI and SEXF on CL and WT on V 1 and V 2 were included in the best model.

Best Model
The parameter estimates of the best model are shown in Table 2. Body weight had significant impact on V 1   The magnitude of eGFR change on CL was clinically meaningful.
In the individual post hoc parameter estimates from the best model, the overall mean (n = 20 in the phase 1 study) of estimated t 1/2,α was very short (7.56 minutes) and the t 1/2,β and t 1/2,γ were 1.19 and 3.55 hours, respectively.

Model Evaluations
The best model adequately described the observed data.The goodness-of-fit plots for the best model displayed no bias and no trends in the residuals (Figures S1  and S2).Although the number of observations was limited, prediction-corrected visual predictive checks showed that the median time course of observations was generally within the range of the simulated interval of median time course (Figure 2), suggesting that the best model had good predictive performance for both plasma and interval urine concentrations.

Simulation
A total of 1000 patients were simulated with a median (minimum-maximum) body weight of 78.4 kg (34.0-117.0kg), male-to-female ratio of 1:1, and fixed eGFR (90 mL/min) to a simulated pudexacianinium concentration-time profile for each dose level (0.1, 0.3, 1.0, 2.0, and 3.0 mg).The simulated plasma and urine pudexacianinium concentration-time courses for patients under anesthesia are presented in Figure 3.The proportions of patients above the target urine concentration (1 μg/mL) are presented in Figure 4.The simulation results suggested that a single 1.0 mg intravenous dose of pudexacianinium for the patients with urine production at 1.0 mL/min would result in a urine concentration over 1 μg/mL for 3 hours in more than 99% of patients.

Model Qualification
The observed plasma and spot urine concentrations in the phase 2 study 24 were generally within the range of the model predictions, although high IIV was observed in patients during surgery compared with those in healthy participants (Figure 5).There were difficulties in spot urine PK sample collection in patients under anesthesia because their urine production rates were very low and intermittent, while urine production rates in the phase 1 study were very high because participants were allowed to consume water ad libitum throughout the study.A time course of spot urine concentrations was obtained from only 2 participants in the 1.0 mg group in the phase 2 study.A time lag of urine concentration (drug excretion into urine) was observed in both patients.

Discussion
In the present analysis, we took a PPK approach, utilizing available nonclinical and clinical data to determine a dose range to be used in the phase 2 study.The significance of accurately predicting the phase 2 dose range is that the study can be conducted efficiently with a compact sample size, which benefits pharmaceutical companies and minimizes the number of patients exposed to the experimental drug.
In the model development, it is notable that a urine output compartment was incorporated to simulate the urine pudexacianinium concentration-time profile, which is considered a good pharmacodynamic surrogate marker of ureter visualization.
From a 3-compartment model established as the best model, the estimated half-life in the terminal elimination phase (median t 1/2,γ = 3.55 hours) matched the estimates in the phase 1 study by noncompartmental analysis (3.4-3.6 hours at high doses of 8 and 24 mg, in which plasma concentrations were quantifiable up to 24 hours postdose). 23ince the results of covariate exploration suggest that patients with renal impairment will have slower elimination of pudexacianinium, fluorescence intensity and duration of ureter visualization may be affected.However, the contribution of glomerular filtration to the elimination of pudexacianinium is yet to be fully clarified because the estimated population mean of CL (5.31 L/h) was higher than the GFR of unbound drug, and the observed eGFR distribution used to construct the model had a narrow range within normal renal function.Further investigation on fraction unbound and renal excretion is currently planned in a dedicated PK study in patients with renal impairment.
In the simulations, while urine production rate during surgery was assumed to be controlled at 1.0 mL/min, the impact of changing the rate of urine production over a range of 0.5 to 2.0 mL/min was also evaluated.This evaluation was performed to conservatively predict the urine concentration level, given that surgical patients under anesthesia are usually maintained in a hypovolemic state, resulting in a low rate of urine production. 27,29imulation results (Figures 3 and 4) suggested that urine production rate greatly impacted urine pudexacianinium concentrations, and that assuming the rate of 1.0 mL/min, administration of 1.0 mg would result in a urine concentration >1 μg/mL for 3 hours postdose in >99% of patients.Therefore, by selecting 1.0 mg as the central dose, 3 dose levels of 0.3, 1.0, and 3.0 mg were proposed for the phase 2 study using a common ratio of 3 to differentiate between doses.To describe the results of this phase 2 study briefly, successful intraoperative ureter visualization occurred in 2 of 3, 5 of 6, and 3 of 3 participants who received the 0.3, 1.0, or 3.0 mg dose, respectively. 24Pudexacianinium was safe and well tolerated, with only 1 adverse event considered related to treatment.
After the phase 2 study was completed, the simulated results were compared against the observed clinical data.The observed plasma concentrations were generally consistent with model predictions, although high variability was observed in the phase 2 study compared with the phase 1 study.The observed urine concentrations were higher than those from the phase 1 study, which is an expected finding since we  had assumed that surgical patients under anesthesia would have a lower urine production rate.Figure 5 shows that the concentrations basically fell within the range of model predictions in the pink zones, where a low urine production rate (0.2-2.0 mL/min) is assumed.However, the higher variability in urine production in the phase 2 study was not anticipated, which also may have led to the higher variability in the urine concentrations.
Regarding the time lag observed for urine concentration to reach its maximum for 2 patients at the 1.0 mg dose, similar observations of a time lag in maximum urine PK or visualization efficacy have been found in other nonclinical and clinical pudexacianinium studies. 19,23,24These occurrences may be explained by the volume of urine contained in the renal pelvis, as the human renal pelvis is able to accommodate a certain volume of urine. 29,30For some patients with a low urine production rate, it is expected to take several tens of minutes to push out the drug-free urine already existing in the renal pelvis and replace it with urine newly produced after the pudexacianinium injection.
Overall, although the model qualification with urine PKs from the phase 2 study could not be sufficiently achieved due to the reasons discussed above, the observed ureter-visualization results from the phase 2 study matched well with what was suggested by the PK simulations, that is, that the 1.0 mg dose would result in ureter visualization for sufficient duration.
This analysis has some limitations.First, the target urine concentration was set based on ureter imaging in minipigs. 19Although the ureters of minipigs are considered most similar to human ureters in terms of their size and anatomy, 31 the target concentration in minipigs may not directly translate to that in humans during abdominopelvic surgery.Also, the NIR-F devices used for visualization in the phase 2 clinical study were different from those used in nonclinical studies.The actual target concentration may vary by several parameters, including the NIR-F source used, performance or settings of the camera, and distance between the ureter to the source/camera.While many NIR-F devices are used to detect ICG with the same dose regimen, and pudexacianinium possesses photo-optic properties similar to those of ICG, we have yet to confirm in the clinic that pudexacianinium can be visualized using all NIR-F devices with ICG.

Conclusions
A 3-compartment PPK model of pudexacianinium was successfully developed to describe the plasma and urine concentration-time profiles observed in healthy participants.Simulations were conducted to guide the dose setting for the phase 2 study, which suggested that an intravenous injection of 1.0 mg would result in a urine concentration over 1 μg/mL for 3 hours after dosing in more than 99% of patients under anesthesia during surgery, assuming a urine production rate of 1.0 mL/min and that patients have normal renal function.The developed PPK model had a sufficient prediction performance against observed plasma concentrations in the phase 2 study.For urine, although only limited data could be obtained due to the difficulties of spot urine collection from surgical patients, the observed ureter-visualization results matched well with what was suggested by the PK simulations, that is, a 1.0 mg dose would result in ureter visualization for sufficient duration.

Figure 1 .
Figure 1.Schematic population pharmacokinetics model.CL, clearance for the linear elimination; CL R , renal clearance; Q 2 and Q 3 , intercompartmental clearances; V 1 , central volume of distribution; V 2 , shallow peripheral volume of distribution; V 3 , deep peripheral volume of distribution.

Figure 2 .
Figure 2. Prediction-corrected visual prediction checks of the best model.Blue circles, observations; red lines, median (solid) and 5th and 95th percentiles (dashed) of the observations; shaded areas, 95% CIs around the median (pink) and 5th and 95th percentiles (blue) based on model simulations.

Figure 3 .
Figure 3. Simulated plasma and urine pudexacianinium concentration-time profiles.Red solid lines, median of prediction; pink zones, 90% prediction intervals; green lines, target urine concentration of 1 μg/mL; horizontal and vertical dashed lines, lower limit of quantification and target duration of 3 hours, respectively.

Figure 5 .
Figure 5. Simulated and observed plasma and urine pudexacianinium concentration-time profiles for patients in the phase 2 study.Circles, observations; pink zones, 90% prediction intervals of plasma concentration and urine concentration under low urine production rate (0.2-2.0 mL/min); grey zones, 90% prediction intervals of urine concentration under high urine production rate (1.0-4.5 mL/min).

Table 2 .
Population Pharmacokinetics Parameter Estimates for the Best Model